Depressive symptomatology in Obstructive Sleep Apnoea ... fileDepressive symptomatology in Obstructive Sleep Apnoea

Depressive symptomatology in Obstructive Sleep Apnoea

Shenooka Nanthakumar

Bachelor of Arts (Honours)

This thesis is submitted in partial fulfilment of the requirements for the degree

of Doctor of Philosophy (Ph.D.) at the University of Western Australia

School of Psychological Science

2017

Depressive symptomatology in OSA i

I, Shenooka Nanthakumar certify that:

This thesis has been substantially accomplished during enrolment in the degree.

This thesis does not contain material which has been accepted for the award of any

other degree or diploma in my name, in any university or other tertiary institution.

No part of this work will, in the future, be used in a submission in my name, for any

other degree or diploma in any university or other tertiary institution without the prior

approval of The University of Western Australia and where applicable, any partner

institution responsible for the joint-award of this degree.

This thesis does not contain any material previously published or written by another

person, except where due reference has been made in the text.

The work(s) are not in any way a violation or infringement of any copyright, trademark,

patent, or other rights whatsoever of any person.

The research involving human data reported in this thesis was assessed:

Study 2 (Chapter 3) and Study 4 (Chapter 5): Both studies used data from the

Obstructive Sleep Apnoea and Related Symptoms (OSARS) study. The study

was approved by the Sir Charles Gairdner Hospital Research Ethics

Committee and reciprocal approval was provided by UWA HREC (NHMRC

1031575). The study was registered with the ANZ Clinical Trials Register:

ACTRN12612001136897. This study was performed in accordance with the

protocol; ethical principles that have their origin in the Declaration of

Helsinki (revised 1996), and adhered to the National Statement of Ethical

Conduct in Human Research (2007) and in accordance with the ICH

Harmonised Tripartate Guidelines for Good Clinical Practice (GCP).

Study 3 (Chapter 4): Ethics Committee approval was obtained from

University of Western Australia Human Research Ethics Committee;

approval number: RA/4/1/2203.

Written patient consent has been received and archived for the research involving

patient data reported in this thesis.

The work described in this thesis was supported by funding from two sources. Study

2 (Chapter 3) and Study 4 (Chapter 5) were supported through a grant to R.S. Bucks

and co-investigators from the NH&MRC (APP1031575), ‘Obstructive Sleep Apnoea

and Depression’ (2012). Study 3 (Chapter 4) was supported through a grant to R.S.

Bucks and co-investigators from the West Australian Government, to support the,

‘The Busselton Healthy Ageing Study’.

This thesis contains published work and/or work prepared for publication, some of

which has been co-authored.

Signature: [ ] Date: [12/4/2017]

Depressive symptomatology in OSA ii

Abstract

The overall aim addressed of this thesis was to investigate the symptomatology

of depression in Obstructive Sleep Apnoea (OSA), specifically, to determine the impact

of symptom overlap on the expression and treatment of depressive symptoms in OSA.

It begins with a general introduction (Chapter 1), which defines depression and OSA,

and then considers factors underlying the relationship between these two conditions,

symptom overlap, and the treatment of depressive symptoms in OSA by Continuous

Positive Airway Pressure (CPAP). Four research studies (Chapters 2, 3, 4, and 5) are

then reported, concluding with a general discussion (Chapter 6).

Chapter 2 (Study 1) examined the impact of symptom overlap on the prevalence

of depression in individuals with OSA. OSA is a frequent and often under-diagnosed

disorder of breathing during sleep that involves partial or full obstruction to the upper

airway. It is often characterised by symptoms that are also characteristic of depression

(e.g. fatigue, psychomotor retardation, loss of libido). The assessment of depression in

OSA is potentially confounded by symptom overlap, which may lead to the

overestimation of depression in OSA. It is unclear whether the variable prevalence

estimates found in the literature are a function of symptom overlap. A systematic

literature search was conducted, identifying 13 studies that met eligibility criteria.

Meta-analysis was used to synthesize the results from studies examining the prevalence

of depression in OSA. Additionally, symptoms of depression within depression

questionnaires were categorised (e.g. anhedonia, cognitive, cognition). The aggregated

prevalence of depression in OSA was 27%, with estimates ranging from 8% to 68%,

reflecting marked heterogeneity. OSA severity, BMI, and age did not moderate

prevalence rates, whereas, depression questionnaire type, sample type (opportunity

samples versus population-based samples), and gender, significantly moderated

Depressive symptomatology in OSA iii

prevalence estimates. Importantly, prevalence estimates based on questionnaires with

greater symptom overlap between OSA and depression were higher, whereas

questionnaires with a higher proportion of anhedonia symptoms were associated with

lower prevalence estimates. These data suggest that symptom overlap poses a difficulty

in understanding the expression of depression in OSA.

As the assessment of depression in OSA is confounded by symptom overlap, it

is unclear whether CPAP ameliorates both overlapping (OSA-related symptoms) and

non-overlapping depressive symptoms (OSA-independent). Chapter 3 (Study 2)

examined the effect of CPAP in ameliorating overlapping and non-overlapping

depressive symptoms from the Hamilton Rating Scale for Depression (HAM-D), in a

12-week, single arm study (Obstructive Sleep Apnoea and Related Symptoms

(OSARS)) study in individuals with OSA (N = 178 intention to treat data, and N = 135

per protocol data), using latent growth curve modeling (LGCM). At baseline,

individuals endorsed more severe overlapping compared to non-overlapping depressive

symptoms. Both overlapping and non-overlapping symptom severity significantly

decreased over time, but with a greater decrease in the severity of overlapping

depressive symptoms. Critically, greater CPAP use and higher BMI were associated

with faster decline in overlapping depressive symptoms, in a dose-response manner, but

no such relationship was found with the decline of non-overlapping symptom scores.

These findings suggest that, although overlapping and non-overlapping depressive

symptoms significantly decrease over time, overlapping symptoms are more responsive

to CPAP treatment.

Chapter 4 (Study 3) examined the effect of symptom overlap in the Depression,

Anxiety, and Stress Scale (DASS-21). The DASS-21 was examined for inclusion as an

outcome measure in Study 4 (Chapter 5). There are two underlying factors predicting a

Depressive symptomatology in OSA iv

person’s score on a depression measure in OSA, 1) the degree to which there are true

differences in the degree to which the person is genuinely depressed, and 2) the degree

to which they are also endorsing items (particularly overlapping depressive symptoms)

as a function of their OSA. Therefore, questionnaires that consist of overlapping

symptoms may not be assessing genuine depression in OSA: that is, they may not be

measurement invariant, due to differential item functioning (DIF). The study examined,

a) if Lovibond & Lovibond’s (1995) correlated three-factor structure of the DASS-21

holds in an OSA sample (N = 185), compared to an age and gender matched non-OSA

sample (N = 185), and b) measurement invariance across these samples. The OSA and

non-OSA samples were recruited from the Busselton Healthy Ageing Study (BHAS).

The correlated 3-factor structure of the DASS-21 was a better fit in the non-OSA

sample. There was a degree of DIF for each of the subscales, especially for the Anxiety

subscale, in which 2 symptoms (that are also physiological symptoms of OSA)

produced lower severity scores in the OSA sample compared to the non-OSA sample.

However, the effect of DIF on scale totals was very small. Such findings suggest that

the DASS-21 is substantively measurement invariant in OSA and therefore suitable for

use in an OSA sample.

Chapter 5 (Study 4) then examined the effect of CPAP on the DASS-21. As the

DASS-21 is measurement invariant in OSA, any change in depressive symptom scores

over time represents a “true change” in depressive symptoms, rather than the effect of

DIF, due to symptom overlap. The data used in this study were the per protocol data (N

= 134) from the OSARS study (Chapter 3; Study 2). This study examined, 1) the effect

of CPAP in ameliorating DASS-21 symptoms (anxiety, depression and stress), and 2)

whether there was a dose-response relationship in the amelioration of depression,

anxiety, and stress symptoms, and CPAP use, using latent change modelling.

Depressive symptomatology in OSA v

Additionally, this study also examined whether individuals with OSA made a clinically

significant improvement in depressive symptoms at post-treatment, using clinical

significance analyses. Based on latent change models, during the course of CPAP,

depression and anxiety symptoms significantly decreased over time, albeit the decrease

in scores was very small. However, this decline in scores was not a function of CPAP

use. Stress scores did not significantly decline over time. No covariates predicted

individual change in any subscales. In regards to clinical significance, the majority of

individuals did not make a clinical significant improvement over time, albeit most of

the individuals were in the normal range for anxiety, depression, and stress at baseline.

Overall, results suggest that anxiety and depression significantly decrease over time,

however this is not a function of CPAP use.

Chapter 6 discusses the findings with reference to the wider literature, future

directions, originality of the thesis, and recommendations for clinicians.

Depressive symptomatology in OSA vi

Contents

Declaration………………………………………………………………….i

Abstract .............................................................................................................. ii

List of tables ...................................................................................................... ix

List of figures .................................................................................................... xi

List of appendices.............................................................................................. xii

Acknowledgements ........................................................................................... xiv

Statement of original contribution ................................................................... xv

Published work .................................................................................................. xviii

Publications ....................................................................................................... xx

List of abbreviations.......................................................................................... xxii

Chapter 1: General Introduction ....................................................................... 1

Human sleep ............................................................................................... 1

Adult OSA .................................................................................................. 3

Depression in OSA..................................................................................... 4

The relationship between depression and OSA........................................ 6

Symptomatology of depression in OSA ................................................... 9

Differential item functioning (DIF) and measurement invariance.......... 10

OSA, depression, and CPAP ..................................................................... 11

Clinical significance methodology ........................................................... 14

Inter-individual factors moderating the relationship between

depression in OSA ..................................................................................... 15

Summary and content of chapters ............................................................. 18

Chapter 2: Prevalence of depression in OSA .................................................. 22

Preface ........................................................................................................ 22

Depressive symptomatology in OSA vii

Abstract ....................................................................................................... ..23

Introduction ................................................................................................... 24

Method .......................................................................................................... 26

Results ........................................................................................................... 41

Discussion ..................................................................................................... 44

Chapter 3: Effect of CPAP on the HAM-D symptoms in OSA using Latent

Growth Curve Modelling (LGCM) .................................................................... 49

Preface ...................................................................................................... 49

Abstract .................................................................................................... 50

Introduction .............................................................................................. 51

Method ..................................................................................................... 54

Results ...................................................................................................... 62

Discussion ................................................................................................ 75

Chapter 4: Examining the use of the DASS-21 in OSA .................................... 81

Preface ...................................................................................................... 81

Abstract .................................................................................................... 83

Introduction .............................................................................................. 84

Method ..................................................................................................... 87

Results ...................................................................................................... 91

Discussion ................................................................................................ 98

Chapter 5: Effect of CPAP on the DASS-21 symptoms in OSA using latent

change modelling and clinical significance analysis ......................................... 102

Preface ...................................................................................................... 102

Abstract .................................................................................................... 103

Introduction .............................................................................................. 104

Depressive symptomatology in OSA viii

Method………………………………………………………………107

Results ................................................................................................... 114

Discussion ............................................................................................. 125

Chapter 6: General Discussion ......................................................................... 130

Prevalence of depression in OSA and depression

questionnaires ....................................................................................... 130

The effect of CPAP on depressive symptoms in OSA ....................... 132

Symptom-based approach of depression in OSA................................ 133

Clinical Depression in OSA………………………………………...138

Inter-individual covariates moderating the relationship

between OSA and depression .............................................................. 138

Implications for clinicians .................................................................... 141

Conclusion…………………………………………………………..147

References.......................................................................................................... 148

Appendices ....................................................................................................... 180

Appendix A ........................................................................................... 180

Appendix B ............................................................................................ 183

Depressive symptomatology in OSA ix

List of Tables

Table 2.1: Categorisation of depression items into symptom categories ................... 33

Table 2.2: Symptom proportions within depression questionnaires ........................... 38

Table 2.3: Participant and study characteristics for each study of

the meta-analysis .......................................................................................... 40

Table 3.1: Study measures assessed at each visit/week in the CPAP trial ................. 57

Table 3.2: Patient characteristics for intention to treat and per protocol

individuals .................................................................................................... 63

Table 3.3: Descriptive statistics for the HAM-D ......................................................... 64

Table 3.4: HAM-D severity classification for visit 1 and visit 4 for

per protocol individuals ............................................................................... 65

Table 3.5: Best fitting single quadratic growth curve models .................................... 70

Table 3.6: Single quadratic growth curve models with covariates

(…average CPAP use) by per protocol analysis ........................................ 72

Table 3.7: Single quadratic growth curve models with covariates

(…CPAP use 4 hours) by per protocol analysis ..................................... 74

Table 4.1: DASS-21 characteristics for the Non-OSA and OSA samples ................. 92

Table 4.2: DASS-21 CFA in the Non-OSA and OSA samples................................... 93

Table 4.3: Item parameters for the Depression, Anxiety, and Stress subscales ......... 96

Table 4.4: Measurement equivalence of the DASS-21 in the Non-OSA

and OSA samples .......................................................................................... 98

Table 5.1: Characteristics of the OSA sample.............................................................. 114

Table 5.2: Pre and post treatment depression, anxiety and stress subscale scores .... 115

Table 5.3: Descriptive statistics for the DASS-21 from the latent

change modelling analysis .......................................................................... 116

Depressive symptomatology in OSA x

Table 5.4: Latent change models for the DASS-21 mean subscale scores ................ 119

Table 5.5: Latent change models for the DASS-21 mean subscale scores

and covariates (…average CPAP use) ........................................................ 120

Table 5.6: Latent change models for the DASS-21 mean subscale scores

and covariates (…CPAP use 4 hours)...................................................... 121

Table 5.7: Information relevant for calculating clinical significance......................... 122

Table 5.8: Clinical significance classifications ............................................................ 123

Depressive symptomatology in OSA xi

List of Figures

Figure 2.1: Flow chart of search, retrieval and inclusion process

of the meta-analysis ...................................................................................... 28

Figure 2.2: Forest plot for each study in the meta-analysis ........................................... 42

Figure 2.3: Moderated regression plot of OSA items against the

depression prevalence estimates .................................................................. 44

Figure 3.1: Change model for HAM-D total scores for per protocol data ................... 66

Figure 3.2: Change models for non-overlapping and overlapping

depressive symptoms for per protocol data................................................. 69

Figure 4.1: DIF in three DASS-21 anxiety items ........................................................... 97

Figure 5.1: Latent change model design ......................................................................... 111

Figure 6.1: Subgroups of individuals with OSA, based on

Latent Profile Analysis ................................................................................. 135

Depressive symptomatology in OSA xii

List of Appendices

Appendix A ............................................................................................................ 180

Figure S2.1: Forest plot for the effect of sample type on depression

prevalence .................................................................................... 180

Figure S2.2: Moderated regression plot for OSA items and depression

prevalence .................................................................................... 180

Figure S2.3: Moderated regression plot for loss of energy/fatigue items

and depression prevalence ........................................................... 181

Figure S2.4: Moderated regression plot for sleep items and depression

prevalence .................................................................................... 181

Figure S2.5: Moderated regression plot for anhedonia items and

depression prevalence ................................................................... 182

Figure S2.6: Moderated regression plot for proportion of males

and depression prevalence............................................................ 182

Appendix B ............................................................................................................ 183

Table S3.1: Single growth curve model building for total HAM-D

scores, by per protocol analysis..................................................... 183

Table S3.2: Single growth curve model building for overlapping

symptom scores, by per protocol analysis ................................... 184

Table S3.3: Single growth curve model building for non-overlapping

symptom scores by per protocol analysis ..................................... 185

Table S3.4: Single growth curve model building for total HAM-D scores,

by intention to treat analysis .......................................................... 186

Table S3.5: Single growth curve model building for non-overlapping

symptom scores, by intention to treat analysis ............................. 187

Depressive symptomatology in OSA xiii

Table S3.6: Single growth curve model building for overlapping symptom

scores, by intention to treat analysis .................................................. 188

Table S3.7: Quadratic growth curve models with covariates

(…average CPAP use) by intention to treat analysis ....................... 189

Table S3.8: Quadratic growth curve models with covariates

(…CPAP use 4 hours) by intention to treat analysis ................... 190

Acknowledgements

This research was supported by an Australian Government Research Training

Program (RTP) Scholarship, awarded by the Australian Government, University of

Western Australia, and funds from the School of Psychology, University of Western

Australia.

I would like to thank my amazing supervisors, Professors Romola Bucks,

Timothy Skinner, and Sergio Starkstein: you have taught me so much about research,

writing, and life! I could not have completed this thesis without your continued support,

inspiration, kind words, and guidance. I will be forever grateful. Romola and Tim,

thank you for always replying to my emails quickly and making supervision enjoyable!

My amazing mother, you mean everything to me and I could not have reached

this milestone without your continued care, wise words, love, patience, and support. To

my beloved Dad and best friend Lexie dog, who are both watching over me from

heaven, I wish you both were here to celebrate with me, but knowing that you are

always watching over me allowed me to reach this milestone.

My fantastic friends who have always been there to listen to me when I am

having a “PhD crisis”. Thank you for your encouragement and support during the past 4

years.

Finally, a big thank you to Professor David Hillman, Clinical Professor Alan

James, Dr Michael Hunter, Katherine Hatch, Dr Nigel McArdle, and Professor Matthew

Knuiman, and the other wonderful staff at Sir Charles Gairdner Hospital, West

Australian Sleep Disorders Institute, and the Busselton Health Ageing Study Team,

who provided me with guidance, knowledge, and support. It was a great experience

working with you all.

Depressive symptomatology in OSA xv

Statement of Original Contribution

The work contained in this thesis has not previously been submitted for a degree

or diploma at any other higher education institution. This thesis is entirely my own

work and has been accomplished during enrolment in the degree of Doctor of

Philosophy. All external sources have been acknowledged. The preparation of this

thesis has been completed by the candidate alone, with guidance from her supervisors.

The articles that were published as a result of the work undertaken for this thesis are

included in Chapters 2 (Study 1) and 4 (Study 3). The study discussed in Chapter 3

(Study 2) is currently being prepared for publication using the full data set.

In Chapter 2 (Study 1), the student undertook the data collection, set up the

database of articles to be screened, screened and selected all titles, abstracts and

articles, extracted all data, completed all analyses, and formulated and wrote the paper.

RSB and TCS provided intellectual input, conducted screening of the titles, abstracts,

and full-papers, conducted paper quality assessment, advised on the analyses and

interpretation, assisted with the formulation and editing of the paper, and assisted with

the review process for publishing the final manuscript in Health Psychology (IF: 1.95,

top 20% journal).

In Chapters 3 (Study 2) and 5 (Study 4), the research was conducted as part of a

larger study, the Obstructive Sleep Apnoea and Related Symptoms (OSARS) study, in

collaboration with Professor David Hillman, West Australian Sleep Disorders Research

Institute, and Department of Pulmonary Physiology and Sleep Medicine, Sir Charles

Gairdner Hospital; Clinical Professor Alan James, School of Medicine and

Pharmacology, University of Western Australia, and West Australian Sleep Disorders

Research Institute, Sir Charles Gairdner Hospital (SCGH); Dr Nigel McArdle, School

of Medicine and Pharmacology, University of Western Australia; Professor Matthew

Depressive symptomatology in OSA xvi

Knuiman, University of Western Australia; and Katherine Hatch, University of Western

Australia, as outlined in the Acknowledgements section. The student conducted weekly

visits to Sir Charles Gairdner Hospital (SCGH) to: create 20 online questionnaires using

Qualtrics; help set up and configure the patient database; collect questionnaire data;

enter questionnaire data into SPSS (~200 participant data files); define variables and

variable labels and their coding criteria; and, check the database entries, for a total of 21

months (~400 hours). The student also gave a presentation to the hospital staff at SCGH

about the results/updates and to obtain feedback from the staff. The student also

completed the data analyses and wrote the papers. RSB, TCS, and SS, provided

intellectual input, advised on the analyses and interpretation, and edited the final

manuscript. The other authors on the paper provided intellectual input and edited the

final manuscript.

In Chapter 4 (Study 3), the research was conducted as part of a larger study, the

Busselton Healthy Ageing Study (BHAS), in collaboration with Professor David

Hillman, West Australian Sleep Disorders Research Institute, and Department of

Pulmonary Physiology and Sleep Medicine, Sir Charles Gairdner Hospital; Clinical

Professor Alan James, School of Medicine and Pharmacology, University of Western

Australia, and West Australian Sleep Disorders Research Institute, Sir Charles Gairdner

Hospital; Dr Michael Hunter, School of Population Health, University of Western

Australia, and Busselton Population Medical Research Institute, Sir Charles Gairdner

Hospital, as outlined in the Acknowledgements section. The student visited the BHAS

Centre in Busselton (2+ hours from Perth) for a week to enter details from participant

questionnaires into the participant database (+250 questionnaires; 23 hours). The

student entered a large proportion of the data into the participant database, organised

the database for analyses, completed the analyses, and formulated and wrote the paper.

Depressive symptomatology in OSA xvii

RSB, TCS and SS provided intellectual input, advised on the analyses and

interpretation, assisted with the formulation and editing of the paper, and assisted with

the review process for publishing the final manuscript in Psychological Assessment (IF:

2.90). The other authors on the paper provided intellectual input and assisted with

editing the final paper.

For the remaining chapters, Chapters 1 (general introduction) and 6 (general

discussion), the student wrote the chapters, and RSB and TCS assisted with

formulation, provided intellectual input, and assisted with editing these chapters.

Outside the scope of my thesis, but during my PhD, I was invited to be co-

author twice: on a manuscript that examined cognitive and mood dysfunction in

individuals with OSA, which has been published, and on a manuscript looking at the

prevalence of depression in type 2 diabetes, which is currently being prepared for

publication.

Depressive symptomatology in OSA xviii

This thesis contains work that has been published and/or prepared for publication.

Details of the work:

Are we overestimating the prevalence of depression in chronic illness using

questionnaires? Meta-analytic evidence in obstructive sleep apnoea.

Nanthakumar, S., Bucks, R. S., & Skinner, T. C. (2016). Are we overestimating

the prevalence of depression in chronic illness using questionnaires? Meta-analytic

evidence in obstructive sleep apnoea. Health Psychology, 35(5), 423-432.

Location in thesis:

Chapter 2 (Study 1)

Student contribution to work:

The student undertook the data collection, set up the databases of articles to be

screened, completed all analyses, formulated and wrote the paper.

Co-author signatures and dates:

8/4/2017

Details of the work:

Assessment of the Depression, Anxiety, and Stress Scale (DASS-21) in Untreated

Obstructive Sleep Apnoea (OSA).

Nanthakumar, S., Bucks, R. S., Skinner, T. C., Starkstein, S., Hillman, D., James, A.,

& Hunter, M. (2017). Assessment of the Depression, Anxiety, and Stress Scale

(DASS-21) in untreated obstructive sleep apnea (OSA). Psychological Assessment,

29(10), 1201-1209. http://dx.doi.org/10.1037/pas0000401.

Location in thesis:

Chapter 4 (Study 3)

Student contribution to work:

The student attended the BHAS centre in Busselton (2+ hours from Perth) for a

week to enter details from participant questionnaires to the participant database

(+250 questionnaires; 23 hours). The student entered a large proportion of the data

into the participant database, organized the database for analyses, completed the

analyses, and formulated and wrote the paper. RSB and TCS provided intellectual

input, advised on the analyses and interpretation, assisted with the formulation and

editing of the paper, and assisted with the review process for publishing the final

manuscript in Psychological Assessment (IF: 2.90, top 20% journal). The other

Depressive symptomatology in OSA xix

authors on the paper provided intellectual input and assisted with editing the final

paper.

Co-author signatures and dates:

8/4/2017

Student signature:

Date: 10/4/2017

I, Romola S. Bucks certify that the student statements regarding their contribution

to each of the works listed above are correct

Coordinating supervisor signature:

Date: 8/4/2017

Depressive symptomatology in OSA xx

Publications

Components of this thesis presented at conferences

Nanthakumar, S., Bucks, R.S., Skinner, T.C. (2015). Are we overestimating the

prevalence of depression in obstructive sleep apnoea? Meta-analytic evidence

from questionnaire studies. 7th World Congress of the World Sleep Federation,

Istanbul, Turkey, 31st October – 4th November, 2015. Platform presentation.

Nanthakumar, S., Bucks, R.S., Skinner, T.C. (2015). Are we overestimating the

prevalence of depression in obstructive sleep apnoea? Meta-analytic evidence

from questionnaire studies. Sleep DownUnder, Australasian Sleep Association

Annual Conference, Melbourne, Australia, 22 - 24th October, 2015. Sleep and

Biological Rhythms, 13(Suppl. 1), 28. Poster presentation.

Other abstracts published during the course of this PhD

Bucks, R. S, Nanthakumar, S., Starkstein, S.E., Hillman, D.R., James, A.L., Knuiman,

M, & Hatch, K., Skinner, T. C (2017). The Effect of Continuous Positive

Airway Pressure (CPAP) on HAM-D Depressive Symptoms in Obstructive

Sleep Apnoea (OSA), using Latent Growth Curve Modelling. World Sleep

2017, Prague, Czech Republic, 7th October – 11th October, 2017. Poster

presentation.

Clarke, P., Notebaert, L., Nanthakumar, S., Blackwell, S., Holmes, E., & MacLeod, C.

Is emotional ambiguity necessary in imagery-based interpretive bias

modification? 7th World Congress of Behavioural and Cognitive Therapies

(WCBCT) Conference; 22nd – 24th July, 2013, Lima, Peru. Platform

presentation.

Peer reviewed publications (Publications from this thesis are marked with an

asterix

Depressive symptomatology in OSA xxi

(Chapter 2; Study 1) *Nanthakumar, S., Bucks, R. S., & Skinner, T. C. (2016). Are We

overestimating the prevalence of depression in chronic illness using

questionnaires? Meta-analytic evidence in obstructive sleep apnoea. Health

Psychology, 35(5), 423–432. https://doi.org/10.1037/hea0000280.

(Chapter 4; Study 3) * Nanthakumar, S., Bucks, R. S., Skinner, T. C., Starkstein, S.,

Hillman, D., James, A., & Hunter, M. (2017). Assessment of the Depression,

Anxiety, and Stress Scale (DASS-21) in untreated obstructive sleep apnea

(OSA). Psychological Assessment, 29(10), 1201-1209.

http://dx.doi.org/10.1037/pas0000401.

Olaithe, M., Nanthakumar, S., Eastwood, P. R., & Bucks, R. S. (2015). Cognitive and

mood dysfunction in adult obstructive sleep apnoea (OSA): Implications for

psychological research and practice. Invited review for Translational Issues in

Psychological Science – Special edition ‘The Science of Sleep’. 1(1), 67-78.

http://dx.doi.org/10.1037/tps0000021.

Clarke, P. J. F., Nanthakumar, S., Notebaert, L., Holmes, E., Blackwell, S.E., &

McLeod, C. (2013). Simply imagining sunshine, lollipops and rainbows will not

budge the bias: The role of ambiguity in Interpretive Bias

Modification. Cognitive Therapy and Research, 37(3).

Depressive symptomatology in OSA xxii

List of Abbreviations

AHI Apnoea-Hypopnoea Index

BMI Body Mass Index

BHAS Busselton Healthy Ageing Study

DASS-21 Depression, Anxiety, Stress, 21 item version

DIF Differential Item Functioning

LGCM Latent growth curve modelling

HAM-D Hamilton Rating Scale for Depression

OSA Obstructive Sleep Apnoea

OSARS Obstructive Sleep Apnoea and Related Symptoms

RDI Respiratory Distress Index

PSG Polysomnography

Depressive symptomatology in OSA 1

CHAPTER 1: General Introduction

Depressive symptomatology in Obstructive Sleep Apnoea (OSA)

Human sleep

Sleep is a state that involves brain activity, and the relaxation and inactivity of

muscles (Kryger, Roth, & Dement, 2016). Human sleep features two broad stages:

Rapid Eye Movement sleep (REM), which is characterised by the presence of eye

movements, and Non-Rapid Eye Movement (NREM) sleep (Kryger et al., 2016).

NREM sleep is divided into 3 sub-stages: NREM stage 1 (N1), NREM stage 2 (N2),

and NREM stage 3 (N3; AASM, 2012). These sub-stages differ in terms of brain wave

patterns, eye movements, and muscle tone activity (Carskadon & Dement, 2011).

Stages N1 and N2 are regarded as shallow, transitory sleep stages, whereas stage N3 is

considered deep sleep, which is characterised by slow, rhythmic brain activity (AASM,

2012; Kryger et al., 2016). Over the course of a period of sleep, NREM and REM sleep

alternates in 90-120 minute cycles, approximately totaling 4-6 cycles per period of

sleep, in a healthy adult (Carskadon & Dement, 2011). Typically, a sleep episode

begins with stage N1, then proceeds to stages N2, N3, and then to REM (Altevogt &

Colten, 2006; Kryger et al., 2016). NREM sleep constitutes about 75 to 80 percent of

total time spent in sleep, and REM sleep constitutes the remaining 20 to 25 percent

(Kryger et al., 2016). Sometimes, this pattern is not a neat progression, as an individual

can experience periods of wake or shallow sleep during, or before deeper sleep states,

or REM sleep (Kryger et al., 2016).

The physiological and neurological features of the different stages of sleep are

measured and recorded using polysomnography (PSG; Carskadon & Dement, 2011).

An overnight, laboratory PSG is the gold standard (Carskadon & Dement, 2011). PSG

uses a number of monitoring techniques (e.g. electroencephalography (EEG),


electroculogram (EOG)) to provide an assessment of the physiological and neurological

processes that occur during sleep (Carskadon & Dement, 2011; Kryger et al., 2016).

During an overnight PSG sleep study, measurements of oxygen saturation, brain

activity, muscle tone, heart activity and rhythm, breathing rhythm, airflow, eye

movements, and sound and gross body movements are obtained (Pagel & Pandi-

Perumal, 2014; Kryger et al., 2016). These physiological measurements are used to

‘stage’ sleep, and disturbances to sleep (Pagel & Pandi-Perumal, 2014; Kryger et al.,

2016). Sleep stages are defined during PSG by the type of waveform present in an

epoch (30 seconds of recorded sleep) in the EEG signal (Kryger et al., 2016). The

pattern of waveforms over the course of a sleep period is known as sleep architecture.

Disturbances to sleep impact sleep architecture because they change the waveforms and

length of sleep stages seen in healthy sleep (Kryger et al., 2016).

The definitions of sleep patterns are outlined in the ‘The American Academy of

Sleep Medicine Manual for the Scoring of Sleep and Associated Events: Rules,

Terminology, and Technical Specifications’ (AASM, 2012). This guide provides a

comprehensive set of rules for equipment application and signal evaluation in a PSG.

Additionally, the AASM manual also defines the sleep stages and sleep-disorder events

(AASM, 2012).

A sleep-disordered event is the occurrence of a disturbance to regular sleep

architecture (Kryger et al., 2016). An event may be closing of the upper airway

(apnoea), irregular and periodic movement of an individual’s legs (restless legs),

frequent wakening (sleep fragmentation), or other disturbances (e.g. nocturea) (Kryger

et al., 2016). These events occur in healthy sleepers: however, when these events occur

many times per night and affect sleep architecture, these disturbances may be indicative

of a sleep disorder (Kryger et al., 2016). Sleep-disordered breathing (SDB) is a group of


sleep disorders characterised by abnormalities in respiratory patterns during sleep,

which encompass a range of disorders: Central Sleep Apnoea (CSA), sleep-related

hypoventilation, and Obstructive Sleep Apnoea (OSA), which is the most common (Al

Lawati, Patel, & Ayas, 2009). These SDB disorders are distinct clinical syndromes with

specific diagnostic criteria, and differ in terms of the biological/physiological processes

involved, presentation, symptoms, treatment and overall management (Yap &

Fleetham, 2001; Eckert, Jordan, Merchia, & Malhotra, 2007). As such, the correct

diagnosis is essential for overall optimal care and management.

Adult OSA

Adult OSA is a common and often under-diagnosed condition (Motamedi,

McClary, & Amedee, 2009). It is characterised by recurrent episodes of partial

(hypopnoea) or near-complete (apnoea) obstruction of the pharyngeal airway during

sleep (Spicuzza, Caruso, & Di Maria, 2015). Obstructive events are associated with

drops in oxygen saturation and are usually terminated by arousals, resulting in

fragmentation of sleep (Schröder & O’Hara, 2005; Eckert & Malhotra, 2008).

Diagnosis of OSA is by PSG, where an individual’s blood oxygen levels, nasal

airflow, and Apnoea/Hypopnoea Index (AHI) are measured during a sleep period

(Kryger et al., 2016). OSA severity is commonly represented by the AHI value, which

is calculated using the total number of apnoeas and hypopnoeas, divided by the total

number of hours of sleep (Epstein et al., 2009; Kryger et al., 2016). Based on AHI

values, OSA is classified as: 1) mild, AHI = 5 – 14.9/hr; 2), moderate, AHI= 15-

29.9/hr; and, 3) severe, AHI ≥ 30/hour (Schröder & O’Hara, 2005; Kryger et al., 2016).

As discussed, the scoring of respiratory events (apnoeas and hypopnoeas) is defined in

the AASM scoring manual (AASM, 2012). An apnoea is defined as a reduction in

airflow by >90% of baseline, that lasts for at least 10 seconds, and at least 90% of the


event’s duration meets the reduction in airflow by > 90%. Importantly, there is

variability in the definition of a hypopnoea event. The AASM scoring manual

recommended definition outlines that the changes in flow must be associated with a 3%

oxygen desaturation or a cortical arousal, however the manual allows an alternative

definition, which requires a 4% oxygen desaturation without consideration of cortical

arousals (Kapur et al., 2017). Within this thesis, the prior definition was used for

hypopnoeas (3% oxygen desaturation or a cortical arousal). Common, observable

symptoms of OSA are snoring, nocturnal gasping and choking, excessive daytime

sleepiness, reduced quality of life, and fatigue (Gupta, Chandra, Verm, & Kumar,

2010).

Peppard et al., (2013) conducted an epidemiological study and found that the

prevalence of individuals who had OSA of at least moderate severity (AHI ≥ 15) was

10% among 30-49 year old men, 17% among 50-70 year old men, 3% among 30 – 49

year old women, and 9% among 50-70 year-old women. Untreated OSA is associated

with medical/health issues, increased healthcare use, reduced work performance,

occupational injuries, and motor-vehicle accidents (Gupta et al., 2010). The economic

burden related to untreated OSA is substantial (Gupta et al., 2010). Hillman, Murphy,

Antic and Pezzulo (2006) estimated that the total economic burden of sleep disorders in

Australia was $US 7.5. billion per year, representing 1.4% of the total burden of disease

for Australia. Additionally, Sassani et al., (2004) estimated that in the year 2000, in the

United Sates, 810,000 motor-vehicle collisions and 1400 fatalities were attributable to

sleep apnoea, which resulted in a total cost of 15.9 billion US dollars. They believe that

appropriately treating sleep apnoea would result in an overall saving of 7.9 billion US

dollars.

Depression in OSA


According to the Diagnostic and Statistical Manual of Mental Disorders version

5 (DSM-5, American Psychiatric Association, 2013), cardinal symptoms of depression

include low mood, anhedonia, hypersomnia/insomnia, fatigue/loss of energy, feelings

of worthlessness/guilt, concentration difficulty/ indecisiveness, psychomotor

agitation/retardation, suicidal ideation, appetite changes, and weight gain/loss.

Depression frequently presents with symptoms of anxiety and stress, such as excessive

worry, phobia, and irritability (DSM-5, American Psychiatric Association, 2013).

Individuals can present with the above-mentioned depressive symptoms, and not

be diagnosed with a psychological disorder. For individuals to be diagnosed with Major

Depression (a psychological disorder) they must meet the symptom criteria according to

the DSM-5, their depressive symptoms impact upon their daily functioning, and their

symptoms cannot be attributed to any physical condition. The assessment of clinical

depression is via a structured clinical interview, such as the Structured Clinical

Interview (SCID; First, Spitzer, Gibbon, & Williams, 2012) for DSM-5. A recent

review reported that 29.70% of individuals with OSA met criteria for comorbid Major

Depression, based on the Diagnostic and Statistical Manual of Mental Disorders (DSM-

IV; using the SCID; El-Sherbini, Bediwy, & El-Mitwalli, 2011), in comparison to 6% -

7% of the general adult population (Kessler et al., 2003).

Questionnaires (e.g. Beck Depression Inventory (BDI), Depression and Anxiety

Stress Scale (DASS-21), Zung Self Rating Depression Scale (SDS)) are commonly

used to assess depressive symptoms, and calculate prevalence estimates in clinical and

epidemiological settings, when structured clinical interviews are not practicable or

feasible (Coyne, Thompson, Klinkman, & Nease, 2002). Some of the depression

symptoms assessed by these questionnaires are the same symptoms outlined in the

DSM-5 (e.g. suicidal ideation, low mood), whereas some are not (e.g. loss of libido,


cognitive difficulties). The specific symptoms assessed by these questionnaires are

understood to be a measure of the underlying construct: depression (Fried et al., 2016).

Depression questionnaire cut-offs are used to indicate whether an individual has

clinically significant depressive symptoms. Prevalence of mild or more severe

depressive symptoms in an OSA population, when using depression questionnaires (e.g.

BDI, SDS), ranges from 7% to 63% (Saunamäki & Jehkonen, 2007). This thesis

examined the presence and severity of depressive symptoms using depression

questionnaires, rather than depression, as a clinical psychological disorder.

The relationship between OSA and depression

OSA is a possible risk factor for developing depression. Peppard, Szklo-Coxe,

Hla, & Young (2006) found, in a large-scale longitudinal study of individuals with

OSA, that an increase of one OSA severity level (from minimal [AHI = 0 to 4.99] to

mild [AHI = 5 to 14.99], and mild to moderate or worse [AHI 15] SDB) was

associated with a 1.8-fold increase in the adjusted odds for the development of

depression. More recently, Chen, Keller, Kang, Hsieh, & Lin. (2013) conducted a

longitudinal study to look at the impact of untreated OSA and physician diagnosed

depression (based on ICD-9-CM codes). At 1-year follow-up, the incidence of

Depression was approximately twice as high in individuals with OSA, compared to

those without OSA.

In regards to the sleep architecture of individuals with depression and OSA,

results have indicated that individuals with depression alone experience increased sleep

latency, frequent awakenings, decrease in slow wave sleep, and shortened REM latency

(Benca, Obermeyer, & Thisted, 1992), whereas for individuals with OSA alone, sleep

architecture and sleepy latency are shorter and REM% is decreased (Bardwell, Moore,

Ancoli-Israel, & Dimsdale, 2000). For individuals with comorbid OSA and depressive


symptoms, Bardwell et al., (2000) found that individuals who had OSA and higher

depressive symptoms (based on the CES-D), had higher REM% than individuals with

lower depressive symptom scores. Therefore, there may be a causal relationship

between REM% and the onset of depressive symptoms in individuals with OSA. More

studies examining sleep architecture in individuals with OSA who present with

depressive symptoms are warranted.

There are multiple underlying processes relating to the development of mood

dysfunction in OSA, including the critical roles of sleep fragmentation, blood gas

abnormalities (intermittent hypoxemia), chemical changes (changes to the serotonergic

system and increased levels of cytokines). Disruption of these processes lead to a

difficulty for the body to return to a balanced state, impacting upon the nervous system,

which subsequently results in mood disturbances, daytime dysfunction, and cognitive

deficits (e.g. attention, executive functioning deficits; Olaithe, Nanthakumar, Bucks &

Eastwood, 2015).

Sleep fragmentation is a direct consequence of the repetitive arousals following

apnoeas and hypopnoeas (Schröder & O’Hara, 2005; Harris, Glozier, Ratnavadivel, &

Grunstein, 2009). Sleep fragmentation has been found to be the primary cause of

excessive daytime sleepiness (EDS) in individuals with OSA and is suggested to result

in depressive symptoms (Schröder & O’Hara, 2005).

Intermittent hypoxemia is the subsequent, temporary drop in circulating oxygen

that is caused by diminished saturation of oxyhemoglobin, due to the reduction or

cessation of breathing (Cohen-Zion et al., 2001). Preliminary neuroimaging data

suggest that hypoxemia may play a role in mood disorders. Hypoxemia has been linked

to white matter hypertensities (WMH) in individuals with OSA (Patel, Hanly, Smith,

Chan, & Coutts, 2015). Aloia, Arnedt, Davis, Riggs, and Byrd (2004) found that more


subcortical white matter hypertensities were found in MRIs of patients with severe

OSA, compared to individuals who had mild OSA. They also found a trend for a

positive correlation between subcortical hypertensities and depressive symptoms, as

assessed by the HAM-D. Additionally, Sforza, de Saint Hilaire, Pelissolo, Rochat, and

Ibanez (2002) found a positive correlation between mean nadir nocturnal SaO2 (an

index of hypoxemia) and elevations in the Hospital Depression and Anxiety Scale -

Depression subscale (HADS-D). Although there is limited research, the findings from

these studies suggest a possible causal pathway between intermittent hypoxemia and

depressive symptoms in OSA.

The serotoninergic system is implicated in upper airway muscle tone control

during sleep, the sleep-wakefulness cycle, and the regulation of mood (Schröder &

O’Hara, 2005). Serotonin delivery to the upper airway dilatator motor neurons is

reduced in sleep (Veasey, 2003). This may lead to dilator muscle activity during sleep,

which may contribute to the development of sleep apnoea (Veasey, 2003). Additionally,

changes to serotonin levels have led to alterations in sleep architecture in individuals

with depression (Schröder & O’Hara, 2005).

OSA is associated with elevated levels of inflammatory markers, such as in

cytokines (e.g. interleukin 6 (IL-6; Nadeem et al, 2013)). Major depression is also

associated with elevated levels of IL-6 (Frommberger et al.,1997). Furthermore,

individuals who are both obese and have OSA appear to have elevated levels of IL-6

compared to healthy volunteers (no obesity and without OSA) (Roytblat et al., 2000).

Although no study has explored their association, these findings suggests a possible

shared pathway.

Beebe and Gozal (Beebe, 2005; Beebe & Gozal, 2002) proposed a theoretical

framework about the critical roles of sleep fragmentation and blood gas abnormalities


in the development of mood, as well as cognitive dysfunction, in OSA. In their model,

sleep is considered a vital restorative activity, which regulates important processes such

as modulating neuroendocrine demands, learning and memory. Disruption of these

important processes due to sleep fragmentation, blood gas abnormalities, and

neurochemical changes lead to the body being unable to return to a balanced state, thus

damaging the nervous system. This damage results in a dysfunctional cognitive profile

(e.g. impairment to executive function, memory, attention), mood disturbances and

other daytime problems (see Olaithe, Nanthakumar, Eastwood, & Bucks, 2015; Bucks,

Olaithe & Eastwood, 2013; Lal, Strange, Bachman, 2012). More recently, Kerner and

Roose (2017) developed a model describing the pathophysiologic mechanisms that

underlie the associations between intermittent hypoxemia and sleep fragmentation in

response to oxygen desaturation. Similarly, they postulate that OSA, depression, and

cognitive impairment are related via pathologic processes.

Due to these potential shared pathways, both depression and OSA have common

risk factors including metabolic syndrome, cardiovascular disease, type-2 diabetes and

obesity (Harris et al., 2009).

Symptomatology of depression in OSA

Despite the strong association between OSA and depression noted above, the

assessment and diagnosis of depression in OSA is challenging due to the degree of

overlap in symptoms between the two conditions, such as insomnia, fatigue, decreased

libido, weight changes, and cognitive difficulties (Harris et al., 2009). Due to symptom

overlap, it cannot be assumed that overlapping depressive symptoms in individuals with

OSA indicate the presence of depression, independent of OSA. While this may be the

case in some individuals, it is also possible that these symptoms are attributable to

OSA, rather than to depression. Therefore, the identification of symptoms that reliably


determine the presence of depression, independent of OSA, is difficult. When using

depression questionnaires, the proportion of shared symptoms within depression

questionnaires may impact on prevalence estimates. It is plausible that the high

proportion of shared symptoms leads to the overestimation of depression in OSA.

Chapter 2 (Study 1) of this thesis uses meta-analytic techniques to examine whether the

proportion of overlapping symptoms within depression questionnaires is associated

with higher prevalence estimates of depression in OSA.

Attempts to disentangle the extent to which individuals experience depression,

in the presence of OSA have included researchers modifying questionnaires by

selecting only items that are considered highly characteristic of depression (e.g. feelings

of guilt, worthlessness, suicidal ideation) or removing items that are known to overlap

between the two conditions (e.g. sleepiness, fatigue items; Hashmi, Giray, &

Hirshkowitz, 2006). Millman, Fogel, McNamara & Carlisle (1989) proposed a profile

of the ‘depressed sleep apnoea patient’ using the Zung Depression Inventory (SDS).

They argued that symptoms such as fatigue, sleep disruption, and psychomotor

retardation (i.e. overlapping depressive symptoms), are indicative of underlying sleep

apnoea, whereas symptoms such as confusion, hopelessness, personal devaluation,

indecisiveness, crying, and suicidal ideation are better indicators of depression,

independent of OSA. However, no study to date has first examined, statistically,

whether these overlapping symptoms impact upon the assessment of depression in

individuals with OSA.

Differential item functioning (DIF) and measurement invariance

Scores on items within depression measures are typically added together to

create a sum-score, where an individual’s score reflects the severity of their depressive

symptoms (Fried, 2015). Two factors predict a person’s score when using a depression


measure in OSA: 1) the degree to which they are endorsing genuine depressive

symptoms, independent of OSA; and, 2) the degree to which they are endorsing items

as a function of their OSA. This may lead to differential item functioning (DIF) where

individuals with OSA respond differently to overlapping symptoms, compared to those

without OSA.

Measurement invariance assesses the equivalence of the factor structure

parameters (i.e. factor structure, factor loadings, factor means, and item loadings)

across different samples (Bryne & Watkins, 2003; Nye, Roberts, Saucier & Zhou, 2008;

Steenkamp & Baumgartner, 1998). Without measurement invariance, one cannot

determine if an observed score represents the “true”, underlying construct (i.e.

depression), or systematic biases in the way people from different samples perceive and

respond to items (Horn & McArdle, 1992; Cheung & Rensvold, 2002; Byrne &

Watkins, 2003). Therefore, when examining depressive symptomatology in OSA, it is

important to examine if questionnaires are impacted by DIF/measurement invariance,

due to symptom overlap. No study, to date, has examined if depression questionnaires

are measurement invariant in OSA. Chapter 4 (Study 3) of this thesis examines

measurement invariance in OSA, compared to matched non-OSA individuals, using the

short-form of the Depression, Anxiety, and Stress Scale (DASS-21; Lovibond &

Lovibond, 1995).

OSA, depression, and CPAP

The gold standard treatment for OSA is Continuous Positive Airway Pressure

(CPAP) (Harris et al., 2009). CPAP involves fitting a face or nose mask that is

connected to a device that maintains positive pressure in the respiratory system (relative

to the atmosphere) during sleep, therefore preventing collapse of the upper airway

during inhalation (Bachour & Maasilta, 2004).


Studies that have examined the impact of CPAP on depressive symptomatology

have found mixed results. A number of treatment studies have demonstrated that CPAP

improves depressive symptomatology short-term (4 weeks - 3 months; e.g. Means et al.,

2003; Schwartz, Kohler, & Karatinos, 2007; Edwards et al., 2015; El-Sherbini, Bediwy,

& El-Mitwalli; 2011), and long-term (e.g. 1 year; Yamamoto, Akashiba, Kosaka, Ito, &

Horie, 2000), by comparing scores on depression questionnaires (such as the BDI, Beck

Depression Inventory-Fast Screen for Medical Patients (BDI-II), HAD-S, and SDS)

before and after CPAP treatment. By contrast, other CPAP treatment studies (e.g.

Bardwell, Ancoli-Israel, & Dimsdale, 2007; Munoz, Mayoralas, Barbe, Pericas, &

Agusti., 2000; Borak, Cieślicki, Koziej, Matuszewski, & Zieliński, 1996) have found no

improvement in depressive symptoms (using questionnaires such as the BDI, Brief

Symptom Inventory (BSI): Depression) post-treatment. Critically, many of the

questionnaires (e.g. BDI, SDS) are confounded by overlapping depressive symptoms,

making it difficult to determine what is being improved (or not) with CPAP treatment:

depression, OSA, or both.

Few studies have examined which, specific types of depressive symptoms,

within depression questionnaires, are ameliorated with CPAP treatment. Means et al.,

(2003) found a small, but significant decline in somatic, affective, and cognitive

symptoms (assessed by the BDI), of equal magnitude, at post-treatment. Similarly, El-

Sherbini et al., (2015) found a significant decrease in HAM-D scores over time. When

they then excluded four items of the HAM-D, which could be caused by OSA (items

related to sleep and fatigue; albeit, not all OSA-related symptoms were excluded), they

continued to find a significant decrease in symptoms over time, but to a lesser degree.

However, no study to date has conceptually divided depressive symptoms into


overlapping and non-overlapping depressive symptoms, based on theory, and examined

the effect of CPAP on these symptoms, over time.

A further, important, question is whether decline in depressive symptoms over

time is actually related to CPAP use. Some studies (e.g. Douglas & Engleman, 2000;

Edwards et al., 2015) report improvement in depressive symptoms at post-treatment

when participants were CPAP adherent (4 hours of CPAP per night; Weaver &

Grunstein, 2008). In contrast, other studies (e.g. Wells, Freedland, Carney, Duntley, &

Stepanski, 2007, Kingshott, Vennelle, Hoy, Engleman, Deary, & Douglas, 2000; Lee,

Bardwell, Ancoli-Israel, Loredo, & Dimsdale, 2012) have found that the improvement

in depressive symptoms (e.g. on BDI, Center for Epidemiological Studies Depression

Scale (CES-D)) at post-treatment is not associated with CPAP use. Indices such as

Mean CPAP daily use and 4 hours of CPAP per night, are commonly used to assess

CPAP use (Weaver & Grunstein, 2008; Ballard, Gay, & Strollo, 2007). Overall, it is

still unknown if CPAP use has a different effect on overlapping and non-overlapping

depressive symptoms in OSA. Therefore, it is unclear which symptoms are more

responsive to CPAP treatment.

If CPAP reduces overlapping depressive symptoms, but has no effect on non-

overlapping depressive symptoms, then the decline in overlapping symptoms may

suggest that 1) overlapping depressive symptoms are more responsive to CPAP

treatment, and 2) CPAP only treats OSA, rather than depression. Whereas, if there is a

decline in both non-overlapping and overlapping depressive symptoms over time, and

this is a function of CPAP use, then CPAP may treat both OSA and depression.

The relationship between CPAP use and depression in OSA has not been

consistent in the literature. Two main features may explain these variable results: 1)

comparison between studies is difficult due to differences in study sample, sample size,


participant age, baseline depression scores, depression measure used, exclusion criteria,

OSA severity, and CPAP duration, and 2) symptom overlap in OSA. Due to symptom

overlap, it is unclear if, 1) CPAP ameliorates both overlapping and non-overlapping

depressive symptoms in OSA, and 2) whether decline in symptoms with treatment

represents a “true change” in depressive symptoms, or is due to DIF/measurement

invariance of the measure used. Therefore, the next steps are: 1) conceptually to divide

depression questionnaires into overlapping and non-overlapping depressive symptoms,

and compare the effect of CPAP on these symptoms, and 2) examine the effect of

CPAP on depressive symptoms assessed by a depression measure that is measurement

invariant.

Chapter 3 (Study 2) of this thesis examines the effect of CPAP using the

Hamilton Depression Rating Scale (HAM-D), by dividing the items into overlapping

and non-overlapping depressive symptoms, based on theory, and examines whether

CPAP use is associated with any change in symptom severity scores over time. Chapter

5 (Study 4), then examines, a) the effect of CPAP using the DASS-21 (a measure that

was found to be measurement invariant in Chapter 4 (Study 3)), and b) whether CPAP

use is associated with any change in depression, anxiety and stress symptoms, over

time.

Clinical significance methodology

Although research has examined whether CPAP reduces depressive symptoms

in individuals with OSA, at a group level, no research to date has examined the effect of

CPAP on depression in OSA at an individual level. Examining change in group mean

scores across treatment is potentially problematic. By focusing on the change in mean

scores, alone, this approach does not capture how many people show the desired effect.

A small, average change, may arise only because a few individuals had a significant


improvement with treatment, whilst most showed little or no change (Guyatt Osoba,

Wu, Wyrwich, & Norman, 2002).

Clinical significance methodology (Jacobson, Follette, & Revenstorf, 1984)

provides a valuable way of understanding and evaluating the symptomatic changes an

individual makes between two time periods (e.g. pre-treatment and post-treatment).

This methodology takes into consideration: a) whether an individual makes a change

over time, that is considered statistically reliable, and b) whether an individual’s post-

treatment score resembles the score of an individual in the functional, healthy

population, or the dysfunctional, treatment-seeking population (Jacobson & Truax,

1991). Clinical significance methodology has been used in clinical trials to assess the

effectiveness of treatments (e.g. Asarnow et al., 2005), and is a useful way to examine

whether improvement in scores is clinically meaningful (Jacobson & Truax, 1991).

Although research suggests that CPAP ameliorates depressive symptoms in

OSA, no study, to date, has examined whether CPAP leads to a clinically significant

improvement in depressive symptoms, or what proportion of individuals with OSA

make such a change. Clinical significance methodology was applied in Chapter 5

(Study 4) of this thesis as an additional analysis.

Inter-individual factors moderating the relationship between depression in OSA

Previous research has highlighted that the following inter-individual factors are

associated with the relationship between depression and OSA.

Age. Participant age is an important variable when exploring the relationship

between OSA and depression. The risk of having both OSA and depression,

independently, increases with age (Alchanatis et al., 2008). Interestingly, research has

found that age is not correlated to depressive symptom severity (e.g. Macey, Woo,


Kumar, Cross, & Harper, 2010; Akashiba et al., 2002), AHI severity (Macey et al.,

2010) or excessive daytime sleepiness (Macey et al., 2010).

Gender. There are gender differences in the way OSA is manifested in females

and males. Clear gender differences in upper airway anatomy and function, fat

distribution, control of ventilation, and hormonal differences have been proposed to

account for the higher risk of OSA in males (Ye, Pien, & Weaver, 2009; Kapsimalis, &

Kryger, 2002). Other risk factors such as alcohol consumption and smoking have been

associated with the higher prevalence of OSA in men (Young, Peppard, & Gottlieb,

2002). Additionally, Ye, Pien, Ratcliffe, & Weaver (2009) found that women (N = 24)

with OSA reported more severe depressive symptoms (assessed by the Profile of Mood

State (POMS) than males (N = 152) with OSA in a sample of middle-aged (46.7 ± 8.8

years) and obese individuals (BMI = 38.0 ± 8.2 kg/m2), whereas Aloia et al., (2005)

found that men (N = 61) and women (N = 32) did not differ significantly on the severity

of their depression scores in their sample of OSA individuals (BMI = 34.8 ± 8.4, Age =

52.2 ± 11.1). Limited research has examined gender differences and CPAP use,

however some studies (e.g. Pelletier-Fleury, Rakotonanahary, & Fleury, 2001) have

found that female gender predicted CPAP non-compliance, whereas, other studies (e.g.

Sin, Mayers, Man, & Pawluk, 2002) found that female gender was associated with

higher CPAP use.

Body Mass Index (BMI). Obesity (measured by BMI) is considered a risk

factor for OSA (Harris et al., 2009). The exact mechanism by which body weight

predisposes or worsens OSA has not been defined, but a narrowing of the upper airway

and altered pharyngeal mechanics has been identified in obese OSA individuals

(Schwab, Gefter, Hoffman, Gupta, & Pack, 1993). BMI is a risk factor for sleep apnoea,

and is positively associated with more severe OSA (based on AHI; Sharafkhaneh,


Giray, Richardson, Young, & Hirshkowitz, 2005; Macey et al., 2010). Higher BMI is

also associated with greater severity of depressive symptoms (e.g. Edwards et al., 2015;

Reyes-Zúñiga et al., 2012). Aloia et al., (2005) found that higher BMI was positively

correlated with cognitive dimension scores (e.g. sadness, pessimism items), but not to

somatic dimension scores (e.g. loss of energy, tiredness or fatigue items), of the BDI-II.

Therefore, BMI may have different effects, depending on the type of depressive

symptom.

AHI. AHI (or Respiratory Disturbance Index (RDI)) is the gold standard index

for OSA severity (Harris et al., 2009). A relationship between the severity of OSA and

depression severity would be expected if the two were related. Studies have found that

more severe depressive symptomatology (based on total scores on a depression

questionnaire) is associated with greater OSA severity (Edwards et al., 2015; Macey et

al., 2010). However, other researchers have argued that AHI may be differentially

associated with types of depressive symptoms, for example Aloia et al., (2005) found

that there was a positive relationship between higher RDI and a higher somatic factor

scores (e.g. loss of energy, tiredness or fatigue items), but not the cognitive dimension

scores (e.g. sadness, pessimism items) of the BDI-II, after controlling for the variance

in RDI. Other researchers (e.g. Macey et al., 2010) argue that AHI and RDI are

inappropriate measures for assessing the impact of OSA, and that the frequency of

arousal or degree of oxygen desaturation may be a better predictor of some OSA

symptoms, such as daytime sleepiness, poor sleep quality, and depression, than the

number and frequency of hypoxic events. This argument is consistent with studies

finding no relationship between AHI and depression symptom severity (Lee, Lee,

Chung, & Kim, 2015; Macey et al., 2010; Asghari, Mohammadi, Kamrava, Tavakoli, &

Farhadi, 2012; El Sherbini et al., 2011). Interestingly, Pillar and Lavie (1998) found no


association between OSA severity and depression severity in males, whereas women

with more severe OSA had higher depression scores.

As demonstrated, there is contrasting evidence regarding the relationships

between depression, OSA and these covariates (age, gender, BMI, and AHI/RDI).

Therefore, it is important that these covariates are taken into consideration when

examining the relationship between OSA and depression as they can impact upon the

findings in contrasting ways. For all studies within this thesis, the relationships between

these covariates (e.g. gender, age, BMI, AHI) are explored. Previous research has not

compared the effects of these covariates on overlapping and non-overlapping

depressive symptoms, separately, which will be a unique strength of this thesis.

Summary and content of chapters

The studies reviewed above highlight the complex relationship between

depression and OSA in terms of clinical presentation, underlying pathophysiology, and

treatment. While it appears that depressive symptomatology does play a role in the

overall expression of OSA, the true nature of its involvement in OSA remains

uncertain. This is largely due to the fact that we have yet to understand the

phenomenology of depression in OSA, or the impact of symptom overlap on the

expression and treatment of depression in OSA.

This thesis examines the symptomatology of depression and OSA, specifically

the impact of symptom overlap. In order to do this, Chapter 2 (Study 1) aimed to clarify

whether shared symptomatology within commonly-used depression questionnaires has

a significant effect on the variable prevalence estimates that have been reported in the

literature, using meta-analytic techniques. This study was published in Health

Psychology (2016), 35(5), 423-432.

Second, as the assessment of depression in Obstructive Sleep Apnoea (OSA) is


potentially confounded by symptom overlap, it is unclear if CPAP has a different effect

in ameliorating overlapping (OSA-related) and non-overlapping (OSA-independent)

depressive symptoms. Accordingly, Chapter 3 (Study 2) aimed to compare, a) the effect

of CPAP ameliorating overlapping and non-overlapping depressive symptoms, assessed

by the Hamilton Rating Scale for Depression (HAM-D), and b) examine, if change in

depressive symptoms (whether overlapping or non-overlapping) is related to CPAP use,

in a 12-week, single arm study, using growth curve modelling. The HAM-D is a

commonly used outcome measure in depression treatment trials (e.g. Detke, Lu,

Goldstein, Hayes, & Demitrack, 2002; Goldstein, Mallinckrodt, & Demitrack, 2002). It

was hypothesized that: (1) the severity of overlapping depressive symptom scores

would be higher than non-overlapping depressive symptom scores, at baseline, (2)

HAM-D total scores, and both overlapping and non-overlapping symptom scores would

significantly decrease over time, (3) there would be a faster rate of decline of

overlapping symptoms than non-overlapping symptom scores, over time, and (4)

greater CPAP use would be associated with a greater decrease in HAM-D total scores,

and in both overlapping and non-overlapping symptoms, in a “dose-dependent”

manner, however, the strength of this relationship would be weaker for the non-

overlapping symptoms.

Third, given the potential problem of symptom overlap, it is important in future

research in OSA to have a depression measure for which patients’ responses do not

differ as a function of whether or not they have OSA. Thus, Chapter 4 (Study 3) aimed

to examine whether the DASS-21 is an appropriate measure to assess depressive

symptomatology in OSA: that is, whether the DASS-21 is measurement invariant. The

DASS-21 was chosen because it was designed to have very few somatic symptoms

(Lovibond & Lovibond, 1995), and contains no sleep items. Nonetheless, we


hypothesized that there may be a lack of measurement invariance across samples,

primarily due to differential fit in the Anxiety and/or Stress subscales, due to symptom

overlap. As outlined in Chapter 4 (Study 3), the DASS-21 was found to be

substantially measurement invariant. This study was published in Psychological

Assessment, 29(10), 1201-1209.

Whilst dividing the HAM-D into overlapping and non-overlapping symptoms

and exploring the effect of CPAP on symptom change was one solution to the problem

of symptom overlap, it is not ideal. This is because such a division is determined by the

investigator, based on theory, rather than by exploring the ways in which individuals

with and without OSA respond to individual items statistically. Given that Chapter 4

(Study 3) does the latter, and reveals that the DASS-21 can be used with individuals

with OSA with some confidence, the final, investigative chapter of this thesis (Chapter

5, Study 4) aimed to examine the effect of CPAP on the DASS-21 by comparing pre

and post-treatment scores on the three subscales, and exploring the association between

CPAP usage rates and change in depressive symptoms, in the same sample of

individuals used in Chapter 3 (Study 2). Additionally, this chapter used clinical

significance methodology to examine the proportion of individuals who made, and did

not make, a clinically significant improvement, in depressive symptoms, over time. It

was hypothesized that, 1) there would be a significant reduction in depression, anxiety,

and stress scores, 2) and that greater reduction in subscale scores would be associated

with greater CPAP use. In regards to clinical significance, it was hypothesized that

CPAP treatment would lead to a greater proportion of OSA individuals making

clinically-significant improvement than not, across all subscales.


As inter-individual differences in age, BMI, gender, and AHI appear to

influence the relationship between OSA and depressive symptoms, these are taken into

consideration in all of the studies within this thesis.

The final chapter (Chapter 6) of the thesis presents a general discussion of the

findings, the strengths of this thesis, and future directions.


CHAPTER 2: PREVALENCE OF DEPRESSION IN OSA

Preface

The assessment of depression in OSA is potentially confounded by symptom

overlap. The prevalence of depression in OSA is highly variable; this variability may be

a function of symptom overlap between depression and OSA. The present study

investigated whether symptom overlap moderates the prevalence of depression in OSA.

The paper was written for publication in Health Psychology, therefore the focus of the

paper was on depression and chronic illness, with OSA as an exemplar condition. The

rationale, analysis, and results of this study can be applied to any chronic illness

population in which depression is common.

This chapter was published in the journal Health Psychology:

Nanthakumar, S., Bucks, R. S., & Skinner, T. C. (2016). Are we

overestimating the prevalence of depression in chronic illness using

questionnaires? Meta-analytic evidence in obstructive sleep apnoea.

Health Psychology, 35, 423–432.

http://dx.doi.org/10.1037/hea0000280.

It is presented below, as published, but formatted, for consistency, as the

rest of the thesis.

It is presented below, as published, but formatted consistency with the rest

of the thesis.


Abstract

Introduction: Depression is common in chronic illness, albeit prevalence can

be highly variable. This variability may be a function of symptom overlap between

depression and chronic illness. Using Obstructive Sleep Apnoea (OSA) as an exemplar,

this meta-analysis explored whether the proportion of overlapping symptoms between

OSA and depression, within different depression questionnaires, moderates prevalence

estimates.

Method: A systematic search identified 13 studies meeting eligibility

criteria.

Results: Based on depression questionnaires, the prevalence of depression in

OSA ranged from 8% to 68%, reflecting marked heterogeneity. Prevalence estimates

based on questionnaires with greater symptom overlap between OSA and depression

were higher, whereas questionnaires with a higher proportion of anhedonia symptoms

were associated with lower prevalence estimates.

Discussion: Overall, these data suggest that when using depression

questionnaires to assess the prevalence of depression in OSA, questionnaires that have a

lower proportion of symptom overlap between OSA and depression, as well as a higher

proportion of anhedonia symptoms, reduce the likelihood of overestimating the

prevalence of depression in OSA. This study has implications for other chronic illnesses

with symptom overlap with depression, for example diabetes, chronic kidney disease,

or heart disease, as well as suggesting that depression questionnaires are not equally

appropriate for assessing depression symptomatology in chronic illness populations.


Are we overestimating the prevalence of depression in chronic illness using

questionnaires? Meta-analytic evidence in obstructive sleep apnoea

Depression is a common, debilitating disorder characterized by loss of energy,

anhedonia, poor concentration, decreased libido, and feelings of sadness and

hopelessness (American Psychiatric Association, 2013). Evidence suggests greater

incidence of comorbid depression or more severe depression symptoms in individuals

with chronic medical illness, such as diabetes mellitus (Anderson et al., 2001), sleep

disorders (Vandeputte & de Weerd, 2003), and heart disease (Rutledge, Reis, Linke,

Greenberg, & Mills, 2006). The relationship is bidirectional, such that depression

increases risk of developing chronic illness, and chronic illness increases risk of

depression (Katon, 2003).

Combined depression with chronic illness is associated with increased symptom

burden and functional impairment, noncompliance with treatment, and poorer health

and quality of life (Katon, 2003). A growing number of meta-analyses has examined the

prevalence of depression in different chronic illnesses, including heart failure (Rutledge

et al., 2006), chronic kidney disease (Palmer et al., 2013), and diabetes mellitus

(Anderson et al., 2001). All find that depression prevalence estimates vary widely,

attributing this to the methodologies used and individual characteristics, such as the

type of depression assessment, type of sample, gender, age, and chronic illness severity.

Although these are all plausible explanations for differing prevalence estimates, no

research to date has examined whether depression prevalence estimates in individuals

with chronic illnesses are also moderated by the proportion of symptom overlap

between the chronic illness and depression, within the assessment tools used to assess

depression symptomatology.


Stringent identification of depression relies on structured diagnostic interviews

applied against criteria, such as those from the DSM-5 (American Psychiatric

Association, 2013), for example, the Structured Interview for Clinical Disorders (SCID;

First, Spitzer, Gibbon, & Williams, 2012). However, structured interviews are rarely

used in epidemiological and case– control studies (Coyne, Thompson, Klinkman, &

Nease Jr, 2002). Instead, validated psychometric questionnaires are used to assess self-

reported depressive symptomatology (Coyne et al., 2002), from which prevalence

estimates are estimated (Harris, Glozier, Ratnavadivel, & Grunstein, 2009). A number

of depression questionnaires are used to assess depression symptomatology, for

example, the Geriatric Depression Scale (15 items scale; Yesavage & Sheikh, 1986),

BDI (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961), and the SDS (Zung, 1965).

Each questionnaire consists of a different number of items assessing symptoms of

depression. However, many are also symptoms commonly reported by individuals with

a particular chronic medical illness, such as fatigue, sleep disturbance, and

weight/appetite changes (Benton, Staab, & Evans, 2007). Given this symptom overlap,

assessing depression, per se, becomes difficult, as there is an increased risk of

overestimating the prevalence of depression in chronic illness. To overcome this,

researchers/clinicians (e.g. Peppard, Szklo-Coxe, Hla, & Young, 2006) have eliminated

crossover symptoms when assessing depression, as endorsement of these symptoms

may not necessarily indicate depression but, rather, the chronic illness.

OSA is a chronic medical condition for which there is considerable controversy

regarding the prevalence of comorbid depression and high levels of symptom overlap

(Harris et al., 2009). Epidemiological studies show that between 2% and 4% of middle-

aged men, 1% to 2% of women (Young et al., 1993), and 11% to 62% of older adults

(>60 years old; Janssens, Pautex, Hilleret, & Michel, 2000) are diagnosed with OSA.


Diagnosis of OSA is by PSG and severity is usually based on the average number of

times per hour an individual’s upper airway partially (hypopnea) or completely

(apnoea) blocks as a result of upper airway collapse (Peppard et al., 2006).

Prevalence of mild or more severe depression in an OSA population ranges

from 7% to 63% (Saunamäki & Jehkonen, 2007) in comparison with 6% to 7% in the

general adult population (Kessler et al., 2003). However, assessment of depression in

OSA is confounded by overlap in symptoms, such as insomnia, irritability, decreased

libido, fatigue, and poor concentration, all of which are included in many of the

depression questionnaires used to assess prevalence estimates in the OSA literature

(Harris et al., 2009).

Therefore, this study sought to determine whether the prevalence of depression

in OSA is moderated by the degree of overlap between OSA and depressive symptoms

within the depression questionnaires used. We addressed three questions: (a) What is

the range of prevalence estimates of depression in OSA, when using depression

questionnaires? (b) Is the prevalence of depression in OSA moderated by the proportion

of symptoms shared between OSA and depression within the depression questionnaire

used? (c) Are these effects moderated by age, OSA severity, BMI, sample type, or

gender?

Method

Search Strategy

A scoping search failed to find sources before 1980 that met quality criteria.

Thus, data for this meta-analysis consisted of articles published in peer-reviewed

journals, non-published articles, and conference abstracts between January 1980 and

July 2013. Procedural details of the methodology employed in this meta-analysis are

outlined in Figure 2.1.


The terms ‘apnoea OR apnea OR sleep-disordered breathing’ were combined

with ‘depression OR major depressive disorder OR depressive disorder OR dysthymic

disorder’. The terms chosen covered a wide range of terms to capture studies that have

explored the relationship between depression and OSA. An extensive, computer-

assisted, systematic literature search was conducted using electronic databases

(Keyword and MeSH explode) for published articles and conference abstracts (Medline

R, PsychInfo, PubMed, EMBASE, CINAHL, CCTR), gray literature (SIGLE),

dissertations and theses (Proquest Dissertations and Theses), and via handsearching

(references of the included articles). Authors of conference abstracts who reported the

prevalence of depression in OSA were contacted to ask for further information

regarding their study, and for any unpublished studies relevant to the field of OSA and

depression. Additionally, relevant articles were retrieved from the reference lists of

studies that were included in the final analysis. This search methodology produced

11,226 papers/conference proceedings.

Study Selection Criteria and Quality Assessment

Abstracts retrieved from the databases and subsequent papers were examined

for suitability (outlined in Figure 2.1).


Figure 2.1. Flow chart of search, retrieval and inclusion process.


This meta-analysis included group-based empirical studies that assessed the

prevalence of depression in adults with OSA. In all instances, studies were excluded if

OSA participants were not diagnosed with OSA using an overnight sleep study (PSG),

split night PSG, in-home PSG, or portable ambulatory diagnosis. However, an

exception to this was for Reyes-Zúñiga et al. (2012) where some participants were

administered a simplified, respiratory polygraph instead of a PSG. Simplified

respiratory polygraphs have been found to be a sensitive and accurate method for

diagnosing OSA (Ballester et al., 2000). As the present focus was on OSA, studies

where it was clear that individuals had central sleep apnoea (CSA) were excluded.

Research demonstrates that the epidemiology, pathophysiology, and clinical

characteristics of CSA and OSA are distinct (Young et al., 2002). Studies that assessed

individuals under the category of sleep apnoea or sleep-disordered breathing (SDB)

were included, as CSA events are only prevalent in 9% to 10% of individuals with SDB

(Young et al., 2002).

This article considered only studies with adult participants (18 years). For

articles where the age range or inclusion criteria for age were not stated, 95%

confidence intervals using the provided mean and SD of the age were calculated.

Studies were excluded if the lower confidence interval fell below 18, reducing the

chance of including participants aged below 18. Only studies that assessed depression

using validated questionnaires were included. Studies were also excluded if the authors

failed to report the cut-off indicative of a depression. Although diagnostic and statistical

interviews of clinical disorders (DSM–III and DSM–IV) are the gold standard for

assessing depression, the focus of this study was exploring whether symptom overlap in

depression questionnaires influenced prevalence estimates, therefore studies that used

clinical interviews only were excluded. Studies were excluded if they did not include all


individuals with depression at baseline. Studies that excluded individuals who may

have had depression (e.g. individuals with current psychiatric illness, axis I diagnoses,

currently taking antidepressants, past diagnosis of depression etc.) were excluded.

Samples were also excluded if individuals were from special populations (e.g.

individuals with traumatic brain injury, major depression, terminal illness, chronic

kidney disease, Alzheimer’s disease etc.). Studies were excluded if patients had

engaged in any OSA treatment: CPAP, oxygen therapy, weight loss, mandibular

advancement splint and positional therapy, since the severity of depression

symptomatology may decrease after OSA treatment (Wells, Freedland, Carney,

Duntley, & Stepanski, 2007). As depressive symptomatology has been found to

influence CPAP use, studies were excluded if they reported baseline depression

prevalence only for individuals who engaged successfully with an OSA treatment

(Wells et al., 2007). Studies were excluded if authors provided insufficient information

regarding their sampling methodology. For studies where there was an overlap between

the populations assessed, studies that had the earlier publishing date or largest N were

selected; the remaining studies were excluded. If insufficient data were provided to

calculate prevalence rates (percentages, frequencies, or odds ratios) and the authors

could not be contacted, these studies were excluded.

Twenty-nine authors of conference abstracts that reported the prevalence of

depression in individuals with OSA were contacted for further information, and for

unpublished and published papers that were in the field of depression and OSA; seven

replied. Five provided more information about their study, however these studies were

later excluded for failure to meet the inclusion criteria. Four provided full journal

articles, but only one paper (Castro et al., 2013) was included, as the other studies were

either articles that were already found in the systematic search or did not meet inclusion


criteria. One author provided extra, but insufficient, information about their study,

leading to exclusion. Additionally, one unpublished thesis could not be sourced. This

left 13 studies to be included in the final analysis. The authors (S.N., R.S.B., T.C.S.)

independently reviewed 10% of the abstracts for abstract screening, 10% of the articles

for the paper screening, and 10% of the articles for the paper quality assessment. The

approach of assessing 10% of the articles within screening phases and data extraction

has been commonly used (see Hartling et al., 2013; Chang, Connelly, & Geeza, 2012).

Where there were disagreements the authors discussed and came to an agreement. The

data for all included studies were extracted and coded by the first author. The second

author extracted and coded the data for 5 randomly selected articles. There was 100%

agreement on data extraction between the first and second author.

Categorization of the items from the depression questionnaires

Table 2.1 reports the allocation of the items from each depression questionnaire

(from studies in the final analysis) into the symptom categories and subcategories. The

symptom categories were as follows:

OSA: Symptoms that are considered symptoms of both OSA and

depression;

Somatic/behavioral consequences of depression: Symptoms that are

physical or behavioral manifestations of depression;

Sleep disturbance: Symptoms that directly relate to sleep quality or

quantity;

Loss of energy or fatigue: Symptoms that reflect loss of energy or

fatigue;

Cognitive: Symptoms that relate to conscious mental activities;

Negative affect: Symptoms that refer to the experience of an internal


feeling or emotion;

Cognition: Symptoms that reflect mental processes about oneself, for

example, ‘I am a failure’;

Anhedonia: Symptoms that relate to the inability to experience pleasure

from different experiences or activities;

Anxiety: Somatic and cognition symptoms that are characteristic of

anxiety.


Table 2.1

The categorization of the items (numbers under each depression questionnaire corresponds to the item number within the questionnaire)

into symptom categories and subcategories

Symptom categories (in

bold) and subcategories Depression questionnaire

BDI HADS-D CES-D CES-D

(10 items)

SDS GDS QUIDS

OSA items

Increase in appetite 1 18 7

Increase in irritability 2 11 1 1 15

Weight gain 1 9

Diurnal variation 2 2

Psychomotor changes 3 8 15

Insomnia 4 16 11 7 4 1, 2, 3, 4

Loss of energy 4 10 13 14

Tiredness or fatigue 4 17

Cognitive difficulties

(concentration and/or

memory) 2

5 2 11 10 10

Loss of libido 5 21 6

Somatic/behavioural consequence of depression items

Loss of appetite 186 2 5 6

Appetite gain 186 7

Irritability 11 1 1 15




BDI HADS-D CES-D CES-D (10 items)

SDS GDS QUIDS

Weight loss 19 7 8

Weight gain 9

Agitation 13 16

Diurnal variation 2

Psychomotor retardation 8 15

Crying 10 17 3

Talking 13

Bowel changes 8

Sleep disturbance items

Insomnia 16 11 7 4 1, 2, 3, 4

Loss of energy or fatigue items

Loss of energy 10 13 14

Tiredness or fatigue 17

Apathy 15 7, 20 4, 10 12

Cognitive items

Indeciveness 13 16 107

Concentration difficulty 5 2 11 107

Memory issues 10

Negative affect items

Sad 1 6 18, 3, 6, 12 3, 8 1 5, 7 5

Lonely 14 9

Pessimism and hope 2 8 5 14 14





SDS GDS QUIDS

Guilt 5

Cognition items

Sense of humor 4

Useless 17

Bored 4

Helpless 8

Suicidal thoughts and

ideation

9 12

Failure 3 9

Worthless 12

Punishment 6

Self judgment 7, 8 11

Body image 14

Interpersonal

relationships

15, 19, 4 19 15

Anhedonia items

Full of life and

satisfaction

18 3, 11, 1

Loss of pleasure in

activities once enjoyed 4 2 20

Loss of interest in general 12 12 16 2, 9 13

Loss of interest in

appearance

10





SDS GDS QUIDS

Loss of interest in

entertainment

14

Loss of libido 21 6

Anxiety items

Somatic 9

Cognition 20 10 6 6

Note. BDI = Beck Depression Inventory (Beck et al., 1961); HADS-D = Hospital Anxiety and Depression Scale – Depression Scale (Snaith &

Zigmond, 1983); CES-D = Center for Epidemiological Studies Depression Scale (Radloff, 1977); CES-D (10 items) = Center for Epidemiological

Studies Depression Scale (10-item short form; Andresen, Malmgren, Carter, & Patrick, 1994); SDS = Zung Rating Scale for Depression (Zung, 1965); GDS = Geriatric Depression Scale (15-item short form; Yesavage & Sheikh, 1986); QUIDS = Quick Inventory for Depressive Symptoms (Rush et al.,

2003). QUIDS score categories: Sleep (items 1-4), sadness (item 5), appetite/weight changes (items 6-9), concentration/decision-making (item 10),

view of myself (item 11), suicide (item 12), general interest (item 13), energy level (item 14), and psychomotor retardation (items 15 and 16). For the

OSA items, the superscript numbers represent the articles that indicate that the symptom that the item is assessing is also a symptom of OSA. 1 Phillips, Kato, Narkiewicz, Choe, & Somers (2000); 2 Schröder & O’Hara (2005); 3 Mathieu et al., (2008); 4 Chervin (2000); 5 Karacan & Karatas (1995). 6 The

appetite item on the BDI asks about appetite change, which could be appetite gain or loss. Therefore, this symptom was only counted once for that

symptom category. 7 The concentration/decision-making item on the QUIDS relates to two symptom subcategories within the cognitive category,

concentration difficulty or indeciveness. Therefore, this symptom was only counted once for that symptom category.


The number of items in each symptom category was divided by the total number

of items in the questionnaire to produce a proportion for the symptom category. Items

that belonged to two or more symptom categories were counted separately for each

symptom category. For example, changes in fatigue and appetite are symptoms of OSA

as well as somatic/behavioral consequences of depression, thus they were counted in

both symptom categories. The QUIDS contains 16 items, collapsed into 9 category

scores (sleep, sadness, weight/appetite change, concentration/decision making, view of

myself, suicidal ideation, general interest, energy changes, and psychomotor changes).

Thus, the proportion for the QUIDS was based on 9 rather than 16 points. The

categorization was conducted by the primary author, and checked independently by the

other authors (R.S.B., T.C.S). Any disagreements were discussed and resolved. Table

2.2 reports the proportions of symptoms by symptom category for each questionnaire.


Note. BDI = Beck Depression Inventory; HADS = Hospital Anxiety and Depression Scale – Depression Scale; CES-D = Center for Epidemiological Studies Depression Scale; CES-D (10 items) = Center for Epidemiological Studies Depression Scale (10-item short form); SDS = Zung Rating Scale for Depression; GDS =

Geriatric Depression Scale (15-item short form); QUIDS = Quick Inventory for Depressive Symptoms (16 items divided into 9 categories: proportions of symptom

categories reported). 0.01 was added to each of the proportions in the categories (except the OSA item category) to prevent zero values on Comprehensive Meta

analysis.

Table 2.2

Number (%) of items belonging to each symptom category for each depression questionnaire

Symptom Categories Depression Questionnaire

BDI

HADS-D

CES-D

CES-D

(10 items)

SDS

GDS

QUIDS

OSA items 5 (23.81) 1 (14.29) 3 (15.00) 3 (30.00) 6 (30.00) 2 (13.33) 5 (55.56)

Somatic/behavioural

consequence of depression

items

4 (19.05) 1 (14.29) 4 (20.00) 1 (10.00) 7 (35.00) 0 2 (22.22)

Sleep disturbance items 1 (4.76) 0 1 (5.00) 1 (10.00) 1 (5.00) 0 1 (11.00)

Loss of energy or fatigue

items 2 (9.52) 0 2 (10.00) 2 (20.00) 2 (10.00) 1 (6.67) 1 (11.00)

Cognitive items 1 (4.76) 0 1 (5.00) 1 (10.00) 2 (10.00) 1 (6.67) 1 (11.00)

Negative affect items 3 (14.29) 1 (14.29) 6 (30.00) 4 (40.00) 2 (10.00) 3 (20.00) 1 (11.00)

Cognition items 6 (28.57) 1 (14.29) 4 (20.00) 0 2 (10.00) 4 (26.67) 2 (22.22)

Anhedonia items 3 (14.29) 4 (57.14) 1 (5.00) 0 3 (15.00) 5 (33.33) 1 (11.00)

Anxiety items 1 (4.76) 0 1 (5.00) 1 (10.00) 1 (5.00) 1 (6.67) 0


Data Extraction and Analysis

Data extracted and coded from the final articles included author/s, publication

status, year of publication, sample size, sample type (opportunity or population-based),

participant details when available (gender, body mass index [BMI], age, AHI/RDI

(Respiratory Desaturation Index), depression questionnaire used, and prevalence

estimate of depression. Opportunity samples were made up of participants with OSA,

usually recruited from a hospital clinic or through medical records. Population-based

studies were sampled using stringent sampling methodologies from a fixed population.

Pooled mean and SDs for AHI/RDI, BMI, and age that were not reported were

calculated if sufficient data were provided, using the following formulae for pooled

means [(x1 n1 x2 n2)/(n1 n2)] and pooled standard deviations (x1- x2 = [12/n1

22/n2]). Individual study and participant characteristics are provided in Table 2.3.


Note. Articles are displayed alphabetically by type of depression measure used; ‘Publication status’ corresponds to the publication status of the study (P= peer reviewed published

article; D= dissertation; ‘Study type’ indicates whether the study was an opportunity sample (O) or population-based study (P), xx indicates that these details were not provided in the study. A indicates that the median value was used as an AHI was not reported (see. Hozo, Djulbegovic, & Hozo, 2005). B indicates that only the individuals who were met the

cut-off for the HADS-D were used, not the individuals who reported anxiety plus depression. BDI = Beck Depression Inventory; HADS-D = Hospital Anxiety and Depression

Scale – Depression Scale; CES-D = Center for Epidemiological Studies Depression Scale; CES-D (10 items) = Center for Epidemiological Studies Depression Scale (10-item

short form); SDS = Zung Rating Scale for Depression); GDS = Geriatric Depression Scale (15-item short form); QUIDS = Quick Inventory for Depressive Symptoms. .

Table 2.3

Participant and study characteristics for each study

First author

of paper Year

Publication

status

Sample

Type N

Age

M (SD) % Male

AHI

(M/SD)

BMI

(M/SD) Depression measure

Prevalence of depression

(%)

Castro 2013 P P 354 xx xx xx xx BDI 9.28

Liu 2011 P O 142 47.52 (10.62) 80.28 10A 28.41 (3.82) BDI 31.69

Hashmi 2006 P O 98 53.20 (10.10) 100 43.50 (29.80) 35.40 (6.80) BDI 35.70

McCall 2006 P O 121 54.70 (14.10) 76.03 56.20 (27.20) xx BDI 44.60

Svaldi 2003 P O 54 48.70 (9.20) 61.11 49.70 (31.00) 37.40 (7.00) BDI 55.56

Stepnowsky 2011 P O 240 57.00 (12.10) 96.67 37.50 (20.10) 33.40 (6.30) CES-D (10 – items) 68.33

Kuo 2000 D P 52 60.1

(9.00) 59.60 33.50 (7.40) xx CES-D 11.50

Bardwell 2001 P O 64 49.30 (7.90) 84.38 51.50 (28.50) 29.80 (4.50) CES-D 33.00

Yaffe 2011 P P 105 82.60 (3.10) 0 xx 28.70 (5.30) GDS 8.57

Kristiansen 2012 P P 296 45.14 64.53 49.99 xx HADS-D 9.80

Reyes-

Zuniga 2012 P O 382 50.80 (13.60) 61.78 54.18 (37.55) 32.00 (6.00) HADS-D 7.59B

Mysliwiec 2013 P O 69 35.68 (7.64) 100 20.18 (17.56) 31.48 (3.86) QUIDS 47.82

Ye 2008 P O 108 47.60 (10.80) 80.56 38.70 (25.10) 27.30 (3.30) SDS 42.60


Data Processing and Quantitative Methodology

Comprehensive Meta-Analysis version 2.2.064 (Borenstein, Hedges, Higgins, &

Rothstein, 2011) was used to synthesize data, calculate effect sizes, conduct moderator

and meta-regression analyses, and create forest plots. All probability values that were

less than .05 were deemed significant.

Results

Data are reported from 2085 individuals with OSA (72% male; mean age, 52.69

9.83; mean BMI, 31.77 5.32; mean AHI, 40.45 24.91).

Calculation of Pooled Prevalence of Depression in OSA

We computed prevalence point estimates and 95% confidence intervals using

the formulas: Logit Event Rate = Log [Event rate/ (1 Event rate)], Logit Event Rate

SE = SQRT[1/(Event rate * Total)] [1/((1 Event rate) * Total)], and 95%

confidence intervals as Lower Limit = Logit Event Rate (1.96 Logit Event Rate SE)

and Upper Limit Logit Event Rate = (1.96 Logit Event Rate SE) (see Rutledge et al.,

2006). A random effects model was used. The random effects model assumes that each

study has a different underlying ‘true’ effect size because of differing sample

demographic variables (Rosenthal, 1995). A random effects model accounts for these

between-studies differences, as well as within study participant differences (Rosenthal,

1995). In the present meta-analysis, samples differed in age, OSA severity, BMI,

sample type, gender, and depression questionnaire used. Table 2.3 illustrates the

prevalence estimates of depression among individuals with OSA reported for each of

the 13 studies. The prevalence estimates reported across these studies varied widely,

from 7.59% to 68.33% (see Figure 2.2). The highest depression prevalence estimate

was found using the CES-D (10 items), and the lowest were from the GDS and HADS-

D. Rosenthal’s Fail-safe N, which represents the number of studies needed to create an


overall, non-significant effect, was 765.

Figure 2.2. Forest plot of the event rates, 95% confidence intervals, Z- values

(the standardized effect size), and P-values (the significance of the standardized

effect, indicating if the standardized effect is significantly greater than zero) for

each study, calculated using a random effects model.

Moderator Analysis

Categorical moderator analyses were conducted to examine whether age (Group

1 < 50 years old, Group 2 50 years old), OSA severity (Group 1 = Severe [AHI/RDI

30], Group 2 = Mild-Moderate [AHI/RDI < 30]), BMI (Group 1 = Overweight [BMI

25 – 29.99], Group 2 = Obese [BMI 30]), and sample type (Group 1 = opportunity

sample, Group 2 = population-based study) contributed to the heterogeneity of the

prevalence of depression in OSA. We had planned to explore depression questionnaire

and publication status as moderators, but there were insufficient studies to do so

statistically. Age, OSA severity and BMI were not significant moderators (p > .05) of

the prevalence of depression in OSA. Study type was a significant moderator of

depression prevalence estimates (Q (1) = 10.13, p < .05), with prevalence estimates


from population-based studies being lower (N = 4; 9.70%) than estimates from

opportunity samples (N = 9; 38.60%). See Figure S2.1 (Appendix A) for the forest plot

for the effect of study type on depression prevalence estimates.

Meta Regression Analysis

Meta regression analyses were conducted on the proportion of items for each of

the different categories of items for each depression assessment. A mixed effects

regression with unrestricted maximum-likelihood analysis (UML) was used. The

proportion of items within the questionnaires for the cognition, cognitive, negative

affect, anxiety, and somatic items were not significant predictors (p < .05) of depression

prevalence estimates in OSA. By contrast, the proportion of OSA symptoms was a

significant predictor, with increasing proportion of OSA symptoms associated with

increased prevalence rates, Q (1) = 6.96, p < .05; T2 = .71 (see Figure 2.3). Likewise,

the same positive association held for the proportion of sleep items, Q (1) = 17.32, p <

.05; T2 = .44, and for the proportion of loss of energy/fatigue items, Q (1) = 18.18, p <

.05; T2 = .42. The greatest proportion of symptom overlap with OSA was found for the

QUIDS and the CESD-10, the lowest was for the HADS and GDS (see Table 2.2). The

proportion of anhedonia items was also a significant predictor, Q (1) = 11.11, p < .05;

T2 = .56, except that the relationship with prevalence of depression was negative. That

is, as the proportion of anhedonia items increased (greatest in the GDS), prevalence of

depression decreased.


Figure 2.3. Moderated Regression plot of the proportion of OSA items against the

prevalence of depression (as a logit event rate). Note: Larger logit event rate = higher

prevalence estimate.

Finally, increasing proportion of males in a sample (N = 12 studies) was a

significant predictor of increasing depression estimates, Q (1) = 9.01, p < .05; T2 = .58.

See Appendix A (Figures S2.2 – S2.6) for the moderated regression plots of the

significant predictors.

Discussion

The current article builds on evidence surrounding the relationship between

depression and chronic illness, with a focus on OSA. Prevalence estimates ranged from

7.59% to 68.33%. Exploration of this heterogeneity considered depression

questionnaire characteristics (proportion of items that belong to the different symptom

categories within each depression questionnaire), and study and participant

characteristics (age, OSA severity, BMI, sample type, and gender).

Nature of the Depression Measure

Studies that utilized the HADS-D and GDS reported the lowest prevalence


estimates (7.59% – 9.80%), while the study that utilized the CES-D (10 items) reported

the highest prevalence estimate (68.33%). Proportion of items belonging to the

somatic/behavioral, cognition, cognitive, anxiety, and negative affect categories did not

moderate prevalence estimates. Proportion of depression symptoms that were also

symptoms of OSA moderated prevalence estimates of depression. The same was found

for sleep and loss of energy/fatigue items (both symptoms characteristic of OSA).

Interestingly, the higher the proportion of anhedonia items, the lower the prevalence of

depression. Questionnaires with very few overlapping symptoms with OSA, and a

higher proportion of anhedonia items, such as the HADS-D and GDS produced lower

prevalence estimates (between 7.59% and 9.80%).

Intriguingly, this prevalence estimate appears to be similar to that found in the

general population (6% to 7%) when a stringent diagnostic interview (Composite

International Diagnostic Interview [CIDI]; World Health Organization, 1990) was used

to assess the prevalence of depression (Kessler et al., 2003). Thus, all meta regressions

tell a consistent story: depression questionnaires that have a high proportion of shared

symptomatology and a low proportion of anhedonia items are associated with higher

prevalence estimates, and therefore may lead to an overestimation in the prevalence of

depression in OSA.

Study Characteristics

Studies created from opportunity samples (38.60%) were associated with higher

depression prevalence estimates than population-based studies (9.70%). This may be

attributable to differences in depression severity, biases in subject selection, quality of

life, or comorbid disease in individuals with OSA in these two types of samples (Harris

et al., 2009). Age and OSA severity were not significant moderators. These findings

were consistent with previous research that investigated the relationship between age


(e.g. Millman, Fogel, McNamara, & Carlisle, 1989), and OSA severity (e.g. Kripke et

al., 1997), but inconsistent with research exploring the relationship between BMI and

depression symptomatology (e.g. Aloia et al., 2005). Nonetheless, by assessing BMI,

age, and AHI/RDI, using group averages, we may have obscured the effects. Therefore,

further research addressing the effect of BMI, OSA severity, and age on the relationship

between depression and OSA is warranted.

This study suggests that males have a higher rate of depression than females,

which contrasts findings from previous studies. However, as males generally have more

severe OSA than females, they may endorse more symptoms on depression

questionnaires attributable to OSA than females, which may account for this result

(Wahner-Roedler et al., 2007). Studies reporting depressive symptoms by gender and

OSA severity (AHI/RDI) are required to assess this intriguing idea.

Other factors, such as antidepressant use, presence of hypertension, and

presence of other comorbid diseases (e.g. diabetes and heart failure), were not explored

in this study, and therefore could not be considered for their effect on prevalence

estimates (Benton et al., 2007). Finally, the meta-analysis was not pre-registered with

Prospero, which is an international database of prospectively registered systematic

reviews. Finally, the meta-analysis was not pre-registered with Prospero, which is an

international database of prospectively registered systematic reviews. Prospero provides

a comprehensive inventory of systematic reviews registered at inception, which aims to

reduce study duplication and the opportunity for reporting bias. Whilst we did not make

any changes to the protocol implemented from inception, the failure to register does not

allow us to provide evidence of this. Nevertheless, this is the first, comprehensive

study that has assessed whether the prevalence of depression in OSA is affected by the

degree of symptom overlap between the two disorders, within different depression


questionnaires.

Implications and future research

These results paint a compelling story regarding the influence of different

depression questionnaires used to assess the prevalence and symptomatology of

depression in OSA. These findings suggest that what is needed is the development of a

questionnaire to measure depression in OSA. At the most basic level, such a

questionnaire could assess symptoms that do not overlap with OSA, albeit this might

ignore important features of depression as a result. Any new questionnaire will need

validating against gold standard diagnostic interview measures such as the SCID (First

et al., 2012). Until such a questionnaire has been developed, however, and based on our

findings, we would recommend the use of the GDS (15-item short form; Yesavage &

Sheikh, 1986) or the HADS-D (Snaith & Zigmond, 1986) to assess depression

symptomatology in individuals with OSA, when clinical interviews are unavailable in

clinical or research settings.

The impact of depression questionnaires may have implications for the

assessment of CPAP outcomes. There has been contrasting evidence in the literature

regarding whether CPAP diminishes depression symptomatology in individuals with

OSA (Schröder & O’Hara, 2005). However, no study to date has examined whether the

proportion of overlapping symptoms in the different depression questionnaires used to

examine CPAP outcomes moderates the aggregated results regarding the efficacy and

effectiveness of CPAP on depressive symptomatology in OSA. Additionally,

comparing which symptoms within depression questionnaires change after OSA

treatment will allow us to explore which groups of symptoms are more tractable with

OSA treatment, and which (perhaps negative self beliefs, hopelessness, desire to self

harm) are less amenable to OSA treatment. Together, these studies will help to shed


more light on the phenomenon of depression in OSA. These findings also have

significant implications for other chronic illness populations in which depression is

comorbid (e.g. diabetes, Anderson et al., 2001; chronic kidney disease, Palmer et al.,

2013; and heart failure, Rutledge et al., 2006). Our findings suggest that not all

depression questionnaires may be appropriate for assessing depressive symptomatology

in chronic illness populations. Studies replicating this methodology in other chronic

illnesses, such as diabetes, chronic kidney disease, and heart failure, are now warranted.


CHAPTER 3: EFFECT OF CPAP ON THE HAM-D SYMPTOMS IN OSA

Preface

As the meta-analysis in Chapter 2 (Study 1) suggested that the assessment of

depression in OSA is confounded by symptom overlap, it is unclear if CPAP

ameliorates both overlapping (OSA-related) and non-overlapping depressive (OSA-

independent) symptoms to the same degree. Therefore, this chapter (Study 2) examined

the effect of CPAP in ameliorating overlapping and non-overlapping depressive

symptoms assessed in the Hamilton Rating Scale for Depression (HAM-D), in a 12-

week, single arm study, in individuals with OSA, using latent growth curve modeling

(LGCM).

This was a preliminary analysis using a proportion (46.30%) of the data from

the Obstructive Sleep Apnoea and Related Symptoms (OSARS) study. The paper using

the full dataset is being prepared for publication.


Abstract

The assessment of depression in Obstructive Sleep Apnoea (OSA) is

confounded by symptom overlap, making it difficult to determine whether Continuous

Positive Airway Pressure (CPAP) ameliorates both overlapping and non-overlapping

depressive symptoms in OSA. This study examined the effect of CPAP ameliorating

overlapping and non-overlapping depressive symptoms, within the Hamilton Rating

Scale for Depression (HAM-D), in a 12-week, single arm study using Latent Growth

Curve Modelling (LGCM). At baseline, individuals endorsed proportionally more

severe overlapping compared to non-overlapping, depressive symptoms. Both

overlapping and non-overlapping symptoms significantly decreased over time, but with

a greater decrease in the severity of overlapping than non-overlapping depressive

symptoms. Critically, greater CPAP use was associated with a faster rate of decline, but

in overlapping depressive symptoms alone. Therefore, overlapping depressive

symptoms are more responsive to CPAP treatment. Implications are discussed.


The effect of Continuous Positive Airway Pressure (CPAP) on HAM-D depressive

symptoms in Obstructive Sleep Apnoea (OSA), using Latent Growth Curve

Modelling (LGCM)

The assessment of depression in OSA is potentially confounded by the overlap

in symptoms between the two conditions, such as insomnia, fatigue, decreased libido,

irritability, weight gain, and cognitive difficulties (Nanthakumar, Bucks, & Skinner,

2016). Based on our meta-analysis (Nanthakumar et al., 2016, Chapter 2), depressive

symptoms that do not appear to overlap markedly with OSA include negative affect,

anhedonia, and depressive cognitions (e.g. suicidal ideation, hopelessness), whereas

physical consequences of anxiety (e.g. rapid heartbeat, dry mouth), loss of energy, sleep

difficulties, cognitive difficulties (e.g. concentration difficulty), and

somatic/behavioural consequences of depression (e.g. sexual dysfunction, psychomotor

retardation) are common to both depression and OSA. Critically, increasing symptom

overlap is associated with increasing prevalence of depression in OSA (Nanthakumar et

al., 2016).

A number of studies have found improvement in depressive symptoms with

CPAP treatment (e.g. Means et al., 2003; Edwards et al., 2015; El-Sherbini, Bediwy, &

El-Mitwalli, 2009), whereas other studies have not (e.g. Bardwell, Ancoli-Israel, &

Dimsdale, 2007; Munoz, Mayoralas, Barbe, Pericas, & Agusti., 2000; Borak, Cieślicki,

Koziej, Matuszewski, & Zieliński, 1996). A Cochrane meta-analysis by Giles et al.,

(2006) found a small, but significant improvement in depressive symptoms, at post-

treatment (with CPAP) when using the Hospital Anxiety and Depression Scale –

Depression subscale (HADS-D) to assess depressive symptomatology. Povitz et al.,

(2014) conducted a meta-analysis of randomized controlled trials of CPAP, and found

that depressive symptoms had improved post-treatment (using the Beck Depression


Inventory (BDI), Geriatric Depression Scale (GDS), Montgomery-Asberg Depression

Rating Scale (MADRAS), the Profile of Mood States (POMS) – depression subscale,

the Center for Epidemiologic Studies Depression Scale, and HADS-D), but found

considerable heterogeneity (variability) between trials, which may be explained by the

proportion of symptom overlap within measures, as suggested by our work in Chapter 2

(Study 1). Due to symptom overlap, it is unclear whether CPAP ameliorates both

overlapping and non-overlapping depressive symptoms to the same extent.

Few studies have examined which types of depressive symptoms are

ameliorated with CPAP treatment. Edwards et al., (2015) found that all but one

symptom (concentration item) in the Patient Health Questionnaire (9-item PHQ),

significantly decreased over time. They argued that the concentration symptom, which

is an OSA-related symptom, did not significantly reduce over time due to floor effects.

Similarly, El-Sherbini et al., (2015) found a significant decrease in HAM-D scores over

time. When they then excluded four items, which could be caused by OSA (items

related to insomnia and fatigue), they still found a significant decrease in depressive

symptoms over time, but to a lesser degree. Although this study excluded four items

that overlapped with OSA (insomnia and fatigue), they did not exclude other

overlapping symptoms, such as psychomotor retardation and genital symptoms (e.g.

loss of libido), which may have confounded the results.

Additionally, it is unclear whether decline in depressive symptoms over time is

related to CPAP use. Some studies (e.g. Douglas & Engleman, 2000; Edwards et al.,

2015) have found improvement in depressive symptoms at post-treatment only for

individuals who were CPAP adherent (4 hours of CPAP per night; Weaver &

Grunstein, 2008). In contrast, other studies (e.g. Wells, Freedland, Carney, Duntley, &

Stepanski, 2007, Kingshott, Vennelle, Hoy, Engleman, Deary, & Douglas, 2000; Lee,


Bardwell, Ancoli-Israel, Loredo, & Dimsdale, 2012) found the improvement in

depressive symptoms (e.g. using the BDI, Center for Epidemiological Studies

Depression Scale (CES-D)) at post-treatment was not associated with CPAP use. If

depressive symptoms improve over time, regardless of CPAP use, then the amelioration

of depressive symptoms over time could, potentially, represent a placebo effect

(Engleman, Martin, Douglas, & Deary, 1994).

Although studies have examined the effect of CPAP on depressive symptoms,

critically, no published study has conceptually divided depressive symptoms within a

depression measure into non-overlapping and overlapping symptoms (based on theory),

to compare the effect of CPAP in ameliorating these symptoms. If CPAP only reduces

overlapping depressive symptoms, then this would suggest that overlapping symptoms

are more responsive to OSA treatment, and perhaps, CPAP only treats OSA, not

depression. However, if there is also a decline in non-overlapping depressive symptoms

with treatment, then CPAP may improve both depression and OSA.

Latent Growth curve modeling (LGCM) is a statistical method that allows for

the estimation of inter-individual variability in intra-individual patterns of change over

time (i.e. trajectory; Curran, Obeidat, & Losardo, 2011). Instead of comparing mean

scores at different time points (as in repeated measures analysis of variance), LGCM

allows the prediction of an individual’s growth trajectory and examination of what

predictors (i.e. covariates) influence the rate of change during treatment (Curran et al.,

2011). Additionally, LGCM allows for the comparison between overlapping and non-

overlapping patterns of change. Characteristics such as BMI, age, antidepressant use,

gender, and baseline AHI have been found to influence the prevalence of depression in

OSA (Nanthakumar et al., 2016), therefore, they are important covariates to take into


consideration to examine if they predict the trajectory of change in overlapping and

non-overlapping symptoms over time.

Therefore, LGCM allows us to: 1) compare overlapping and non-overlapping

symptom scores at baseline, 2) compare the direction and rate of change of overlapping

and non-overlapping depressive symptoms, over time and 3) examine which predictors

(covariates: CPAP use and patient characteristics) influence the rate of change of

overlapping and non-overlapping depressive symptoms over time.

The present study examined the effect of CPAP (assessed by (a) mean

usage/hour, and (b) CPAP adherence ( 4 hours CPAP use per night)) in ameliorating

overlapping and non-overlapping symptoms depressive symptoms, using the HAM-D,

in a 12-week, single arm study. Although the HAM-D was not examined in Chapter 2

(Study 1), the HAM-D was used because it was the outcome measure for the larger

project (Obstructive Sleep Apnoea and Related Symptoms Study (OSARSS)), where

the data were from. We applied the same criteria outlined in Chapter 2 (Study 1) to

divide the HAM-D symptoms into overlapping and non-overlapping depressive

symptoms. It was hypothesized that: (1) the severity of overlapping depressive

symptom scores would be higher than non-overlapping depressive symptom scores, at

baseline, (2) HAM-D total scores, and both overlapping and non-overlapping symptom

scores would significantly decrease over time, (3) there would be a faster rate of decline

of overlapping symptoms than non-overlapping symptom scores, over time, and (4)

greater CPAP use would be associated with a greater decrease in HAM-D total scores,

and in both overlapping and non-overlapping symptoms, in a “dose-dependent”

manner, however, the strength of this relationship would be weaker for the non-


Method


Protocol and participants

The Obstructive Sleep Apnoea and Related Symptoms (OSARS) study is a

longitudinal, 12-week, single arm trial of CPAP. The study was approved by the Sir

Charles Gairdner Hospital Research Ethics Committee and reciprocal approval was

provided by UWA HREC (NHMRC 1031575). The study was registered with the ANZ

Clinical Trials Register: ACTRN12612001136897.

All participants provided written, informed consent. Participants, male and

female, aged 18 years old and over, were referred between January 2014 and March

2016 to the Sleep Disorders Clinic at the Queen Elizabeth II Medical Centre for

diagnosis of suspected OSA. As part of usual care, a sleep-health questionnaire was

mailed to all newly-referred patients to complete prior to review by a sleep physician.

After their first visit with a sleep physician, an overnight sleep study

(polysomnography, PSG) was arranged. Following PSG, patients were reviewed by a

sleep physician and treatment options discussed. Patients with OSA (AHI >5; derived

from overnight PSG), and who met the inclusion criteria (see below), were invited to

participate in the study.

Exclusion criteria: Patients were excluded if, they had (1) AHI < 5; (2) severe

nocturnal desaturation (more than 50% of respiratory events occurring during sleep are

non-obstructive); (3) central (non-obstructive) sleep apnoea (>50% of apnoeic/hypnoeic

events are non-obstructive); (4) predominantly Cheynes-Stokes respiration; (5) were

currently in treatment for OSA (e.g. oral appliance); (6) likely, non-OSA primary

diagnosis e.g. symptomatic restless legs, insomnia, narcolepsy, delayed sleep phase

disorder, sleep hypoventilation, neurological disorders; (7) they had any other deficit

that would prevent self-administration of the CPAP mask; (8) presence of severe

medical morbidity that may compromise 3-month survival; (9) severe respiratory


disease; (10) cerebrospinal fluid leak; (11) recent cranial surgery or trauma; (12) either

English was a second language (ESL) or there was poor command of the English

language; (13) they met diagnostic criteria for Psychotic Mood Disorders or Substance

use Disorders, as assessed by Structured Clinical Interview (SCID-I) at visit 1; (14)

presence of significant cognitive deficits as assessed with the Mini Mental State

Examination (MMSE) at visit 1; (15) severe suicidality as measured on the Columbia

Suicide Severity Rating Scale (CSSRS) at visit 1, (16) they were not able to attend

study visits, or (17) they reported that their dose of antidepressants or anxiolytics had

been adjusted within the 12 weeks prior to the screening phase, or were to be adjusted

during the term of treatment.

The study spanned 12 weeks, consisting of 5 clinic visits (visit 1, visit 1a, visit

2, visit 3, and visit 4), and any unscheduled sleep visits. Table 3.1 outlines the measures

and information, relevant to this paper, gathered at each visit.


Table 3.1

Study measures assessed at each visit/week in the OSARS trial

Visit

Week Prior to

recruitment

(Screening

phase)

Visit 1

Week 0

Visit 1a

Week 1

Visit 2

Week 3

Visit 3

Week 8

Visit 4

Week 12

(3 months)

Unscheduled sleep

visits

Between weeks 0 – 12

Measures Spirometry

Overnight

PSG

HAM-D

SCID-I

MMSE

CSSRS

Demographic

questionnaires

Height and

weight

measurements

Participants

given a CPAP

machine.

CPAP

usage

HAM-D

CPAP

usage

HAM-D

CPAP

usage

HAM-D

CPAP usage

CPAP usage

Note. Information regarding age, gender, BMI, education years were collected at visit 1. CPAP data were collected at Visit 1 to Visit 4 and Unscheduled sleep visits.


In total, 378 individuals were recruited at baseline, of whom 187 (51.94%)

individuals had completed the trial (including those who withdrew) and their data

outcomes were available for analysis as of the 3rd June 2016. Of these 187 individuals,

11 were excluded (due to insufficient English fluency, N = 1; central sleep apnoea, N =

1; severe nocturnal desaturation, N = 1; severe suicidality, N = 1; respiratory condition,

N = 1; or met criteria for Psychotic Mood Disorders, N = 3; or Alcohol Dependence, N

= 3), at the screening assessment phase, leaving 176 participants. Of these, 135

completed the trial per protocol, and 41 withdrew or were withdrawn during the course

of treatment due to, change in dose of antidepressants, N = 15; unable to attend visits, N

= 5; CPAP intolerance, N = 15; personal medical issues, N = 2; declined to continue, N

= 4.

For the purposes of this study, two sets of analyses were conducted 1) intention

to treat analysis (N = 176); and, 2) per protocol analysis (N = 135). In the intention to

treat analysis, missing data on the outcome variables for participants who withdrew or

were withdrawn were imputed using the conservative method of “last visit carried

forward”, whereby each participant’s outcome values at their last visit prior to

withdrawal were carried forward (Porta, Bonet, & Cobo., 2007). Patient characteristics

by intention to treat and per protocol are outlined in Table 3.2.

Measures

Demographic details. Participants completed questionnaires related to

demographic information (e.g. gender, birth date, marital status), and their general

health (e.g. smoking history, medical history, alcohol consumption). Height and weight

measurements were completed at Visit 1 and were used to calculate BMI.

Overnight sleep study. All patients underwent a standard, 12-channel PSG.

The PSG characterized sleep and includes recordings of brain activity/arousal


(electroencephalography), cardiac rhythm (electrocardiography), respiratory patterns,

movement (electromyography: limbs and eye), respiratory effort (electromyography:

thorax and abdomen), sound, airflow (nasal), oxygen saturation (oximetry), and body

position. Data were analysed by Sleep Technologists using standard criteria to identify

apnoeas, hypopnoeas, and desaturations to obtain an apnoea/hypopnoea index (AHI),

which was derived using the American Academy of Sleep Medicine scoring criteria

(AASM; 2012). Apnoeas required a reduction in airflow by >90% of baseline, that

lasted for at least 10 seconds, and at least 90% of the event’s duration met the reduction

in airflow by > 90%. Hypnopneas required a reduction in airflow by > 50% of baseline,

that lasted for at least 10 seconds, a > 3% desaturation from pre-event baseline, or there

was an arousal, and at least 90% of the event’s duration met the reduction in airflow by

> 50%. AHI was the average number of apnoeic and hypopnoeic events observed per

hour of monitoring.

Hamilton Rating Scale for Depression (HAM-D). The HAM-D (Hamilton,

1960) is a 17-item measure that is used to examine the burden (type and magnitude) of

depressive symptoms, in the form of a semi-structured interview. The HAM-D is one of

the most widely used outcome measures for depressive symptoms in clinical trials and

has been shown to be a valid and reliable measure (Cusin, Yang, Yeung., & Fava,

2009). Nine items are scored on a 5-point scale ranging from 0 = not present to 4 =

severe, and eight items are scored on a 3-point scale ranging from 0 = not present to 2 =

severe. Scores range from 0 to 54, with higher scores representing more severe

depressive symptoms. HAM-D severity classification cut-offs are: No depression (0-7),

mild depression (8-16), moderate depression (17-23), and severe depression (≥24)

(Zimmermann, Martinez, Young, Chelminski & Dairymple, 2013). The HAM-D was


assessed at Visit 1, Visit 2, Visit 3 and Visit 4, and was administered by trained

clinicians (who were trained by a Consultant Psychiatrist).

In a recent meta-analysis by Nanthakumar et al., (2016; Chapter 2; Study 1),

symptoms of depression in commonly-used depression measures were categorised into

symptoms that overlapped (OSA-related depression) with OSA, or symptoms that did

not overlap with OSA (non-overlapping depressive symptoms) symptoms. These

categories were used to divide the HAM-D items. Overlapping symptoms (HAM-D

items: 4, 5, 6, 7, 8, 13, 14, 10, and 11) assess behavioural consequences (e.g. genital

symptoms, psychomotor retardation, subjective tension/irritability, physical anxiety

symptoms, for example, heart palpitations), sleep difficulties, loss of energy, and

cognitive difficulties, whereas non-overlapping symptoms (HAM-D items: 1, 2, 3, 9,

12, 15, 16, and 17), assess agitation, loss of weight and appetite, negative affect (e.g.

depressed mood), cognition (e.g. helplessness, guilt, suicidal ideation, insight),

anhedonia, and hypochondriasis.

Two, separate subscale scores were created for overlapping depressive

symptoms (9 items; scores can range from 0-26) and non-overlapping depressive

symptoms (8 items; scores can range from 0-26). Overlapping symptoms and non-

overlapping symptom subscale scores were then converted to percentages to allow

comparison.

CPAP device: usage and compliance. Participants were provided with a CPAP

device (Resmed S9 Autoset – Australian Register or Therapeutic Goods (ARTG)),

humidifier and tube with embedded heating circuit at Visit 1. A trained Sleep Scientist

fit a nasal mask (for comfort) and instructed the participant how to use their CPAP

device with heated humidification, in accordance with a standardised protocol. Data

were downloaded from the CPAP device/smart card at each visit (including 1A and


unscheduled visits). For the purposes of this study, CPAP use was defined as 1) the

mean, daily usage of CPAP in hours/per night, and 2) the proportion of days that CPAP

use was 4 hours per night (CPAP adherent according to clinical guidelines; Weaver &

Grunstein, 2008). Where there were missing CPAP use data due to technical faults (N =

7 visits; 4 (2.27%) participants), CPAP usage was based on the CPAP data that were

available for the other visits. CPAP use data were not available for 10 (5.70%)

participants who withdrew prior to their 1A visit. The CPAP equipment was returned to

the study centre at the participant’s final visit.

Data analysis plan. Data screening and descriptive analyses were performed

using SPSS 19.0 for Windows, IBM (2015B). Latent Growth Curve Modeling was

performed using MPlus, version 7.3 (Muthén & Muthén, 1998-2011).

Latent Growth Curve Modeling. Analyses were conducted using Maximum

Likelihood Estimation with robust standard errors (MLR), as this approach is robust to

non-normality and is able to handle missing values (Muthén & Muthén, 1998-2011).

Different functional forms for trajectories were considered, including linear, quadratic,

and logarithmic growth trends. These different trajectories were modeled on the

following outcome variables: i) HAM-D total scores, ii) HAM-D overlapping symptom

percentage scores, and iii) HAM-D non-overlapping symptom percentage scores, across

visits 1 to 4. Separate models were tested for intention to treat and per protocol data, to

find the best fitting models.

After finding a well-fitting growth model for the three different HAM-D scores,

we evaluated the influence of specific covariates (baseline OSA severity, indexed by

AHI; BMI; age; antidepressant use: no antidepressant use = 0, antidepressant use = 1;

education in years, gender: male = 1, female = 0), and CPAP use: mean daily use, and


proportion of nights CPAP use was 4 hours, separately, on these three latent growth

curve models.

Overall model fit was evaluated using χ2, along with the Comparative Fit Index

(CFI; Bentler, 1990), Tucker-Lewis Index (TLI; Tucker & Lewis, 1973) Root-Mean-

Square Error of Approximation (RMSEA; Steiger & Lind, 1980), Standardized Root

Mean Squared Residual (SRMR; Jöreskog & Sörbom, 1993), Akaike Information

Criterion (AIC; Akaike, 1974) and parsimony ratio. The parsimony ratio represents the

complexity of the model relative to the number of observed variables, where values

closer to 1 indicate a more parsimonious model (Raykov, Tomer, & Nesselroade,

1991). Recommended cutoffs for the fit indexes were: CFI: .95, TLI: .95, RMSEA:

.050, and SRMR: .060 (Hu & Bentler, 1999), and an AIC that is smaller was

favoured (Akaike, 1974). Significance was indicated by p < .05.

To compare the baseline scores (intercept) and decline in symptoms over time

(slope) between overlapping and non-overlapping symptom proportion scores

(Hypotheses 1 and 3), chi square tests were conducted by running the models

(overlapping vs. non-overlapping) in parallel. To examine if there was a significant

difference in the intercepts, the intercepts were constrained to be equal, and the slopes

were freed. In contrast, to examine if there was a significant difference in the slope, the

slopes were constrained, and the intercepts were freed. A significant difference in

intercepts and slopes would be indicated by a significant worsening in chi-square model

fit, relative to the unconstrained model.

Results

Descriptive statistics

Participant characteristics by per protocol and intention to treat are in Table 3.2

and the HAM-D scores (total, overlapping and non-overlapping symptom scores) for


each visit are outlined in Table 3.3. The proportion of individuals (per protocol data

only) in each HAM-D severity classification category at visit 1 (pre-treatment) and visit

4 (post-treatment), is in Table 3.4.

Table 3.2

Participant characteristics for per protocol and intention to treat data

Participant

characteristics

Per protocol data

(N = 135)

Intention to treat data

(N = 176)

M SD Range M SD Range

Age 56.23 13.57 23.00 – 83.00 55.48 13.96 20.00 – 83.00

Number of males (%) 67 (49.60) 86 (48.90)

Education

(years) 11.94 2.71 6.00 – 19.50 11.94 2.72 5.00 - 19.50

BMI (kg/m2) 34.32 8.00 22.02 – 80.15 34.13 7.74 19.99 - 80.15

Antidepressant use

(number, %) 49 (36.30) 71 (40.30)

Baseline AHI

(events/hr) 35.80 22.56 6.40 – 107.40 34.53 22.02 5.10 - 107.40

Mean CPAP usage/hr

per night 5.36 1.75 1.20 – 9.85 4.96 2.05 0.01 – 9.85

Proportion (%) of

nights CPAP use was

4 hours

69.60 22.17 8.53 - 100.00 63.78 28.10 0 - 100.00


Table 3.3

Descriptive statistics for HAM-D total scores, overlapping and non-overlapping

symptom percentage scores, for per protocol and intention to treat analyses

Note. N = Number of people who attended a visit; Ns differ between visits as patients did not attend all visits. Overlapping depressive symptoms and non-overlapping depressive symptom subscale scores were converted to percentages. Data were missing

at random across all visits for both per protocol and intention to treat analyses (Little’s MCAR: p > .05).

Visit

Per protocol data

(N = 135)

Intention to treat data

(N = 176)

N M SD Range N M SD Range

HAM-D total scores

Visit 1 scores 135 10.97 6.52 0 – 30 176 11.56 6.78 0 - 30

Visit 2 scores 134 7.71 5.83 0 - 26 174 8.74 6.32 0 - 26

Visit 3 scores 129 6.64 5.76 0 - 29 169 7.95 6.44 0 - 29

Visit 4 scores 134 6.84 6.05 0 - 29 174 8.07 6.60 0 - 29

Overlapping symptom percentage scores

Visit 1 scores 135 30.68 17.01 0 - 73.08 176 31.58 17.17 0 – 73.08

Visit 2 scores 134 20.58 14.80 0 – 61.54 174 23.17 16.00 0 – 73.08

Visit 3 scores 129 17.92 14.55 0 – 65.38 169 21.19 16.25 0 – 73.08

Visit 4 scores 134 18.23 15.82 0 – 69.23 174 21.33 17.05 0 – 73.08

Non-overlapping symptom percentage scores

Visit 1 scores 135 11.51 10.00 0 - 46.15 176 12.87 11.06 0 – 46.15

Visit 2 scores 134 9.07 9.32 0 – 42.31 174 10.46 10.04 0 – 42.31

Visit 3 scores 129 7.60 8.87 0 - 46.15 169 9.38 9.93 0 – 46.15

Visit 4 scores 134 8.09 8.80 0 – 42.31 174 9.70 9.79 0 – 42.31


Table 3.4

Percentage of individuals in each HAM-D severity classification category at visit 1 and

visit 4, for per protocol data

HAM-D severity classification Visit 1

N (%)

Visit 4

N (%)

No depression 46 (34.10) 87 (64.90)

Mild depression 60 (44.40) 33 (24.60)

Moderate depression 26 (19.30) 13 (9.70)

Severe depression 3 (2.20) 1 (0.70)

Note. HAM-D severity classification cut-offs: No depression (0-7), mild depression (8-16), moderate

depression (17-23), and severe depression (≥24) (Zimmermann et al., 2013).

Total HAM-D score severity classification. Based on Table 3.4, the majority

of individuals scored in the mild or more severe range (65.90%) at visit 1, whereas, at

visit 4, the majority of individuals scored in the no depression severity range (64.90%).

An exact McNemar’s test determined that there were significantly more individuals in

the no depression range at visit 4, compared to visit 1 (p < .001).

HAM-D total scores: Identification of single growth model and change

model. A quadratic fit (with the variance of the quadratic change term set to 0 due to a

very small variance) was chosen as the best fitting single growth model for HAM-D

total scores, for per protocol (Table 3.5; Model 1a) and for intention to treat (Table 3.5;

Model 1b) data, based on a combination of fit indices, consistency of fit, and

interpretability. See Appendix B (Table S3.1 and Table S3.4) for the different

functional forms of the change models (e.g. linear, quadratic, and logarithmic growth

trends) that were considered for the intention to treat and per protocol HAM-D total

scores data.

For the per protocol data, higher total HAM-D scores (Table 3.5; Model 1a) at

baseline (intercept term) were associated with a greater decline (slope term) in

depressive symptoms over time (correlation coefficient: -.36, p =.019). At visit 1, the


average total HAM-D score was 10.91, which is in the mild depression severity

category. Significant variance in the intercept term indicated significant individual

variability in HAM-D scores at visit 1. As indicated by the slope term, the average

decrease in HAM-D scores was 3.87 per visit but this, too, varied significantly between

participants. As indicated by the combination of the linear and quadratic terms, there

was an overall decrease from visit 1 to visit 4 (by 0.84 units), with a slight increase in

scores from visit 3 to visit 4. There was significant variability in scores between

individuals. See Figure 3.1 for the trajectory of change of the HAM-D total symptom

scores over time, for the per protocol data. A very similar pattern of results was found

for the intention to treat data (Table 3.5; Model 1b).

Figure 3.1. Change model for HAM-D total scores for the per protocol data.

Identification of single growth model fit for overlapping and non-

overlapping symptom percentage scores: Quadratic change models (with the variance

of the quadratic change term set to 0 due to a small, negative variance) were the best

fitting single growth models for both the overlapping and non-overlapping symptom

percentage scores, for per protocol (Table 3.5; Models 2a and 3a) and intention to treat

0

2

4

6

8

10

12

Visit 1 Visit 2 Visit 3 Visit 4

HA

M-D

tota

l sc

ore

Clinic visit

Change model for HAM-D total scores for the per protocol data


(Table 3.5; Models 2b and 3b) data, based on a combination of fit indices, consistency

of fit, and interpretability.

See Appendix B (Tables S3.2 – S3.3 and S3.5 – S3.6) for the different

functional forms of change models that were considered for the non-overlapping and

overlapping symptom percentage scores (linear, quadratic, and logarithmic growth

trends) for the intention to treat and per protocol data.

Comparison between overlapping and non-overlapping symptom

percentage scores at baseline (visit 1): For per protocol data, comparison of the

intercept terms (baseline scores) for the non-overlapping (Table 3.5, Model 2a) and

overlapping symptom (Table 3.5, Model 3a) models revealed that, at visit 1,

individuals reported a significantly greater percentage of overlapping symptoms than

non-overlapping symptoms: 30.34 vs 11.59; 2(8) = 96.92, p <.010 (intercepts

constrained to be the same and freeing the slopes; difference between the constrained

and unconstrained models). For both the non-overlapping symptoms (Table 3.5, Model

2a) and overlapping symptom change models (Table 3.5, Model 3a) there was

individual variability in the percentage of symptoms endorsed at baseline. This pattern

of results was substantively the same for the intention to treat data (Table 3.5, Models

2b and 3b).

Comparison between overlapping and non-overlapping symptoms in the

decline of symptom scores over time: For per protocol, there was a significant decline

(slope term) for both the non-overlapping (Table 3.5, Model 2a) and overlapping

symptoms (Table 3.5, Model 3a). There were differences in the rate of change (slope

terms) between the two models, with a greater reduction in overlapping symptoms than

non-overlapping symptoms from visit 1 to 4: -11.53 vs -3.21; 2(8) = 400.42, p<.01

(slopes constrained to be the same and freeing the intercepts; difference between the


constrained and unconstrained models ).

Individuals who had more severe overlapping symptoms at baseline had a

greater decrease in their severity of symptoms over time (correlation coefficient: -0.36,

p = .015). Whereas, there was no relationship between baseline non-overlapping

symptom scores and rate of symptom decline over time (correlation coefficient: -0.14; p

= .117).

There was no variability between individuals in the rate of decline of non-

overlapping symptoms (Table 3.5, Model 2a). Whereas, for the overlapping symptom

change model (Table 3.5, Model 3a), there was variability in the rate of decline between

individuals.

There was variability in the quadratic terms between individuals for both the

non-overlapping symptom (Table 3.5, Model 2a), and overlapping symptom change

models (Table 3.5, Model 3a). Figure 3.2 illustrates the change models for the

overlapping and non-overlapping HAM-D symptoms for the per protocol data.

These patterns of results were substantively the same for the intention to treat

data (Table 3.5, Models 2b and 3b). As there were substantially consistent findings for

intention to treat and per protocol data, only the results using the per protocol data are

reported for subsequent analyses. See Appendix B (Tables S3.7 and S3.8) for the results

using the intention to treat data for the subsequent analyses.


Figure 3.2. Change model for overlapping and non-overlapping HAM-D symptoms for

the per protocol data.

0

4

8

12

16

20

24

28

32

Visit 1 Visit 2 Visit 3 Visit 4

Per

centa

ge

of

sym

pto

ms

endors

ed

Clinic visit

Change models for overlapping and non-overlapping HAM-D

symptoms

Overlapping symptoms Non-overlapping symptoms


Table 3.5

Best fitting single growth curve models for the total HAM-D scores, non-overlapping symptom percentage scores, and overlapping

symptom percentage scores, for per protocol and intention to treat data (time modelled as 0, 1, 2, 3)

Note. N = 135 for per protocol and N = 176 for intention to treat. 1 = q@0. †p < .05, ‡ p < .01. AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as

df / [.5k(k+1)] where k is the number of observed variables: higher parsimony ratios are better Recommended cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and

RMSEA: .050 (Hu & Bentler, 1999). Growth parameter estimates are unstandardized values.

Model

number Scale Group Model 2 df p AIC

Parsimony

Ratio

RMSEA

(90% CI) CFI TLI SRMR

Growth parameter estimates

Intercept Slope Quadratic

Mean Var Mean Var Mean Var

1a Total HAM-D Per

Protocol

Quadratic

change1

2.62

4 .623 3091.13 .40 0

(0 -.11) 1.00 1.00 .04 10.91‡ 30.25‡ -3.87‡ 1.30† 0.84‡ 0

1b Total HAM-D Intention

to treat

Quadratic

change1

3.62 4 .460 3962.29 .40 0

(0 - .11) 1.00 1.00 .03 11.48‡ 35.58‡ -3.23‡ 1.21‡ 0.71‡ 0

2a

Non-

overlapping

symptoms

Per

Protocol

Quadratic

change1

4.08 4 .395 3553.65 .40 .01

(0 - .13) 1.00 1.00 .04 11.59‡ 73.15‡ -3.21‡ 0.66 0.67‡ 0

2b

Non-

overlapping

symptoms

Intention

to treat

Quadratic

change1

3.04 4 .552 4577.42 .40 0

(0 - .10) 1.00 1.00 .02 12.90‡ 92.39‡ -3.00‡ 0.61 0.65‡ 0

3a Overlapping

symptoms

Per

Protocol

Quadratic

change1

3.75 4 .442 4169.16 .40 0

(0 - .13) 1.00 1.00 .04 30.34‡ 181.96‡ -11.53‡ 12.01‡ 2.52‡ 0

3b Overlapping

symptoms

Intention

to treat

Quadratic

change1

5.77 4 .217 5373.33 .40 .05

(0 - .13) .99 .99 .05 31.20‡ 205.87‡ -9.28‡ 11.38‡ 2.03‡ 0


Covariates associated with total HAM-D score change models.

For HAM-D total scores (Table 3.6; Model 1a), individuals who were on

antidepressant medication and who had a higher BMI, had higher total HAM-D scores

at baseline (intercept term). In regards to the rate of change, higher mean CPAP use per

night and greater BMI were associated with a significantly faster decline of HAM-D

total scores, over time.

Covariates associated with overlapping and non-overlapping symptoms

change models. For the non-overlapping symptoms change model (Table 3.6; Model

2a), individuals who were younger and on antidepressants reported higher non-

overlapping symptom scores at visit 1. Notably, none of the covariates was associated

with the rate of change.

For the overlapping symptoms change model (Table 3.6; Model 3a), individuals

who had a higher BMI, or were on antidepressant medication, reported higher

overlapping symptom scores at visit 1. In regards to the rate of change (slope term),

higher average CPAP use per night and greater BMI were associated with a faster

decline in overlapping symptoms scores over time, which contrasts with the change in

non-overlapping depressive symptoms where CPAP use was not associated with the

reduction of symptoms.


Table 3.6

Models of total HAM-D scores, non-overlapping symptom percentage scores, and non-overlapping symptom percentage scores, predicted

by age, sex, baseline AHI, antidepressant use, education years, BMI, and average CPAP use, by per protocol analysis

Model Outcome 2 df p AIC Parsimony

Ratio

RMSEA

(90% CI) CFI TLI SRMR Predictor

Relationship with growth parameter estimates

Intercept on predictor

Slope on predictor

Mean Mean

`1a Total HAM-D

scores 19.93 18 .337 3055.89 .27

.03

(0 - .09) .99 .99 .03

Age -0.07 0.01

Sex 0.51 0.16

Baseline AHI 0.00 0.01 Antidepressant use 4.54‡ -0.39

Education years 0.01 0.02

BMI 0.11† -0.04†

Average CPAP use -0.42 -0.20†

2a

Non-

overlapping

symptom percentage

scores

16.59 18 .552 3522.86 .27 0

(0 - .07) 1.00 1.00 .03

Age -0.15† 0.01

Sex 0.51 0.12

Baseline AHI 0.00 0.01

Antidepressant use 5.77‡ -0.24


BMI 0.13 -0.02

Average CPAP use -0.63 -0.09

3a

Overlapping

symptom percentage

scores

22.04 18 .230 4122.36 .27 .04

(0 - .09) .99 .98 .03

Age -0.12 0.02

Sex 1.44 0.50


Antidepressant use 11.70‡ -1.28 Education years -0.05 0.01

BMI 0.31† -0.13†

Average CPAP use -0.97 -0.69†

Note. N = 134 (N = 1 not included in the analysis as BMI was not calculated). AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)]

where k is the number of observed variables. Recommended cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999). †p < .05, ‡ p<.01.

Growth parameter estimates are unstandardized values.


When proportion of nights CPAP use was 4 hours was entered as a predictor,

instead of mean daily CPAP usage (Table 3.7, Models 1a, 2a and 3a) in these change

models, this covariate was a significant predictor only in the total HAM-D and

overlapping symptoms change models. Higher CPAP use was associated with a faster

decrease in HAM-D total scores and overlapping symptom scores. CPAP use had no

effect on the rate of decline in the non-overlapping symptom change model.


Note. N = 134 (N = 1 not included in the analysis as BMI was not calculated). AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)]

where k is the number of observed variables. Recommended cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999). †p < .05, ‡ p<.01.

Growth parameter estimates are unstandardized values.

Table 3.7

Models of total HAM-D scores, non-overlapping symptoms percentage scores, and non-overlapping symptom percentage scores,

predicted by age, sex, baseline AHI, antidepressant use, education years, BMI, and proportion of CPAP use 4 hours, by per protocol

analysis

Model Outcome 2 df p AIC Parsimony

Ratio

RMSEA



Intercept on

predictor

Slope on

predictor

Mean Mean

1a Total HAM-

D scores 15.43 18 .632 3054.66 .27

.03

(0 - .07) 1.00 1.00 .02

Age -0.06 0.01

Sex 0.51 0.19


BMI 0.12† -0.04†



Proportion of 4 hours of CPAP

-0.04 -0.02†

2a

Non-

overlapping

symptom percentage

scores

14.40 18 .703 3521.96 .27 0

(0 - .06) 1.00 1.00 .02

Age -0.14† 0.01

Sex 0.50 0.14

Baseline AHI 0.00 0.01 BMI 0.14 -0.02




-0.06 -0.01

3a

Overlapping

symptom

percentage

scores

17.85 18 .466 4121.08 .27 0

(0 - .08) 1.00 1.00 .03

Age -0.10 0.02

Sex 1.46 0.58

Baseline AHI 0.00 0.01 Antidepressant use 11.61‡ -1.40

Education years -0.05 -0.02

BMI 0.32† -0.12†

Proportion of 4

hours of CPAP

-0.08 -0.05†


Discussion

The present study examined the effect of CPAP in ameliorating overlapping and

non-overlapping depressive symptoms using the HAM-D, in a 12-week, single arm

study. It was hypothesized that: (1) the severity of overlapping depressive symptom

scores would be higher than non-overlapping depressive symptoms, at baseline, (2)

HAM-D total scores, and both overlapping and non-overlapping symptom scores would

significantly decrease over time, (3) there would be a faster rate of decline of

overlapping symptoms than non-overlapping symptoms, over time, and (4) greater

CPAP use would be associated with a greater decrease in HAM-D total scores, and in

both overlapping and non-overlapping symptoms, in a “dose-dependent” manner,

however, the strength of this relationship would be weaker for the non-overlapping

symptoms.

Depressive symptoms in OSA

Prior to treatment, the majority of individuals scored in the mild or more severe

range of depression on the HAM-D. As predicted, the severity of overlapping

depressive symptoms was higher than non-overlapping depressive symptoms, at

baseline, therefore Hypothesis 1 was supported. Although it was anticipated that

individuals would endorse more severe overlapping symptoms, it is plausible that there

are differences in how overlapping symptoms and non-overlapping symptoms are

endorsed by individuals with OSA, which may be attributable to differential item

functioning (DIF). As discussed in the thesis introduction (Chapter 1), an individual

may be endorsing an overlapping symptom because of genuine depression, or due to

their OSA, which may be inflating scores. Therefore, until we statistically examine DIF

in a depression questionnaire, we do not know the extent to which this hypothesis is

true. This intriguing idea is explored in the preceding chapters of this thesis.


CPAP and depressive symptoms in OSA

As anticipated (Hypothesis 2), HAM-D total scores, overlapping symptom

scores and non-overlapping symptom scores, all decreased from visit 1 to visit 4.

Indeed, nearly two thirds of individuals with OSA at baseline (visit 1) met the criteria

(based on HAM-D cut-offs) for mild or more severe depression, whereas at post-

treatment (visit 4), almost half of these (leaving 35.10% with mild or more severe

depression) had remitted. There was a significantly higher proportion of individuals at

post-treatment who were in the no depression severity range, compared to baseline.

Based on the latent growth curve models, total HAM-D, overlapping and non-

overlapping symptom scores all improved significantly over time. Furthermore,

overlapping depressive symptom severity improved faster than non-overlapping

depressive symptom severity, supporting Hypothesis 3. The overall decline in

depressive symptom scores over time is consistent with previous research (e.g. Edwards

et al., 2015; El-Sherbini et al., 2009).

Hypothesis 4 was partially supported in that greater CPAP use was only

associated with a faster decline in total HAM-D scores and in overlapping depressive

scores, but not with the decline in non-overlapping symptom scores. The results suggest

that CPAP is causally related to the change, and suggest a “dose-dependent”

relationship between CPAP use and improvement in overlapping depressive symptoms,

only. Therefore, overlapping depressive symptoms appear to be more responsive to

OSA treatment.

This is the first study to examine if there is a different effect of CPAP in

reducing overlapping and non-overlapping depressive symptoms, by modeling

longitudinal trajectories to examine the rate (and variability) of change over time,

however, conclusions should be considered in the context of the study’s limitations.


Although the decrease in overlapping depressive symptoms was a function of

CPAP use, its not clear what the improvement in non-overlapping symptoms over time

represents. In the absence of a control treatment condition, we cannot be certain if the

decline of these symptoms over time is a genuine effect of treatment, or due to

regression to the mean (Bland & Altman, 2014) and/or placebo effects. The placebo

effect may be the result of patient expectations, mediated primarily by interaction with

the healthcare system for OSA treatment (Crawford et al., 2012). One critical limitation

was that control treatment condition was not used. Due to clinical equipoise, a SHAM

CPAP was not appropriate in the project’s clinical setting; this is discussed in greater

detail in the thesis discussion (Chapter 6).

Another possibility why there was an effect of CPAP use on overlapping and

not the non-overlapping depressive symptoms is the way we divided the symptoms.

Although dividing the HAM-D into overlapping and non-overlapping symptoms and

exploring the effect of CPAP on symptom change was one solution to the problem of

symptom overlap, it is not ideal. We imposed a post-hoc (based on theory) division of

the HAM-D symptoms. The classification of the HAM-D symptoms divided into either

one of these symptom categories was challenging. For example, item 7: work and

activities, assesses both overlapping symptoms (e.g. fatigue or weakness), and non-

overlapping symptoms (e.g. loss of interest). Based on our work (Chapter 2; Study 1),

items that consisted of both overlapping and non-overlapping symptoms, within the one

item, were classified as overlapping symptoms. Therefore, despite the possibility that

an individual’s score on that item may have been driven entirely by a non-overlapping

symptom (e.g. anhedonia), their score on that item counted towards their overlapping

symptom percentage score. Although this only applied to four items, this may have

inflated an individual’s overlapping symptom percentage score and perhaps influenced


the results. Using a measurement invariant questionnaire (i.e. a questionnaire that is not

affected by DIF) would answer whether this was driving the effect. This is further

explored in Chapter 5 (Study 4) and Chapter 6 (thesis discussion).

Despite not using a measurement invariant questionnaire and a treatment control

condition, the most parsimonious explanation of the results is that CPAP treats

overlapping depressive symptoms, which may suggest that CPAP is just treating OSA,

rather than depression, per se. This intriguing idea is further discussed in the thesis

discussion (Chapter 6). Importantly, as the non-overlapping depressive symptoms are

less responsive to CPAP and do not improve as much as overlapping symptoms, over

time, other types of treatment strategies, such as psychological interventions or

antidepressant medication, may be required to further ameliorate the severity of non-

overlapping depressive symptoms (Luyster, Buysse, & Strollo, 2010). This idea is

discussed in greater detail in the thesis discussion (Chapter 6).

Although depressive symptoms significantly decreased over time, from a

clinical perspective, a question that is also of importance is whether a course of CPAP

sufficiently improves depressive symptoms so individuals’ symptom scores are no

longer in the dysfunctional population range (treatment-seeking population). That is,

whether CPAP leads to clinically significant improvement in depressive symptoms.

Clinical significance methodology (Jacobson, Follette, & Revenstorf, 1984) provides a

way of quantitatively of conceptualizing changes individuals make between two time

periods (e.g. pre and post treatment). This method takes in consideration both, a)

whether an individual has made an improvement that is statistically reliable, and b)

whether an individuals’ post-treatment score resembles a member of the functional,

healthy population, or the dysfunctional treatment-seeking population (Jacobson and

Truax, 1991). Although clinical significance methodology could have been examined in


this study, the issue of DIF would have confounded the results. Due to this limitation,

even if we examined clinical significance in this study, we cannot be certain that the

proportion of individuals who made a clinical significant improvement at post-

treatment is a true representation in their score severity, or due to DIF.

Demographic patient characteristics

In relation to patient characteristics, individuals who were on antidepressant

medication reported more severe depressive scores at baseline, which is consistent with

previous research (e.g. Schwartz, Kohler, and Karatinos, 2005), whereby individuals

who are on antidepressant medication reported greater severity of depressive symptoms.

One possible explanation is that some individuals may have been misdiagnosed as

experiencing depression, instead of OSA, and treated accordingly.

Higher BMI was associated with more severe total HAM-D scores and

overlapping depressive symptoms, at baseline, which may be due to obesity being

strongly associated with the development of OSA, leading to the presence of

overlapping depressive symptoms (e.g. fatigue, sleepiness). This is consistent with why

those with a higher BMI improved in terms of the severity of their total HAM-D and

overlapping depressive symptoms faster, than those with a lower BMI.

In relation to age, the results suggest that individuals who are younger endorsed

a greater percentage of non-overlapping symptom scores, which contrasts to our meta-

analysis results (Chapter 2), where there was no relationship between age and

depressive symptom categories, albeit, the meta-analysis used group averages for ages,

which may have obscured the results.

Summary of results and future studies

The study findings demonstrate that: 1) individuals with OSA endorse a greater

proportion of overlapping depressive symptoms than non-overlapping depressive


symptoms, however, it is not clear yet if this is due to DIF or is a genuine difference in

the way individuals experience overlapping and non-overlapping depressive symptoms

in OSA, 2) non-overlapping symptoms reduce over time, however it is unclear if this is

a genuine improvement, or due to DIF, placebo or regression to the mean effects, and 3)

the decrease in overlapping depressive symptoms is a function of CPAP use.

The next steps in this field are to: 1) examine the effect of CPAP on depressive

symptoms, by using a measurement invariant (i.e. not affected by DIF) depression

questionnaire, and 2) examine the proportion of individuals with OSA that make a

clinical significant improvement during the course of treatment. These ideas will be

explored in Chapter 5 (Study 4) of this thesis.


CHAPTER 4: EXAMINING THE USE OF THE DASS-21 IN OSA

Preface

Chapter 2 (Study 1) found that questionnaires that consist of a higher proportion

of overlapping symptoms are associated with higher prevalence of depression.

As the assessment of depression is confounded by symptom overlap, it is

unknown if continuous positive airway pressure (CPAP) ameliorates both overlapping

and non-overlapping depressive symptoms equally. Chapter 3 (Study 2) reported the

impact of CPAP treatment on symptoms of depression using the HAM-D, dividing the

symptoms into those that overlap with OSA (overlapping depressive symptoms) and

those that do not (non-overlapping depressive symptoms). As noted in the chapter’s

discussion, this is not an ideal way to address the problem of differential item

functioning (DIF) in depression measures, albeit it revealed that overlapping symptoms

were more severe, responded better to CPAP treatment than non-overlapping

symptoms, and the response was related to the dose of CPAP used by each individual.

To address the problem of symptoms overlap, it is important to find a

depression questionnaire that is not affected by DIF so that it can be used in confidence

This chapter was published in the journal Psychological Assessment:

Nanthakumar, S., Bucks, R. S., Skinner, T. C., Starkstein, S., Hillman,

D., James, A., & Hunter, M. (2017). Assessment of the Depression,

Anxiety, and Stress Scale (DASS-21) in untreated obstructive sleep

apnea (OSA). Psychological Assessment, 29(10), 1201-1209.

http://dx.doi.org/10.1037/pas0000401.

It is presented below, as published, but formatted, for consistency, as the

rest of the thesis.


in an OSA population. Therefore, this study examined whether the Depression,

Anxiety, and Stress scale (DASS-21) is an appropriate measure to assess depressive

symptoms in OSA.


Abstract

The assessment of depression in obstructive sleep apnoea (OSA) is confounded

by symptom overlap. The Depression, Anxiety, and Stress Scale–short form (DASS-21)

is a commonly used measure of negative affect, but it not known whether the DASS-21

is suitable for use in an OSA sample. This study compared the fit of Lovibond and

Lovibond’s (1995) correlated 3-factor structure of the DASS-21 and measurement

invariance between a non-OSA and an OSA sample using confirmatory factor analysis.

As measurement invariance was not found, to determine the source of non-invariance

differential item functioning (DIF) was examined using dMACS. The correlated 3-

factor structure (with correlated errors) of the DASS-21 was a better fit in the non-OSA

sample. dMACS indicated that there was a degree of DIF for each of the subscales,

especially for the Anxiety subscale, in which 2 symptoms (that are also physiological

symptoms of OSA) produced lower severity scores in the OSA sample compared with

the non-OSA sample. However, the degree of DIF for each of the subscales is not

sufficient to cause concern when using the DASS-21; therefore, the total DASS-21 is

suitable for use in an OSA sample. Interestingly, the impact of symptom overlap in

anxiety symptoms may be reducing anxiety scores because of DIF, which contrasts with

the proposed effect of symptom overlap in depression, where it leads to the inflation of

depression scores in OSA. This deserves greater consideration in relation to OSA and

other clinical disorders or chronic illness conditions with different patterns of



Assessment of the Depression, Anxiety, and Stress Scale (DASS-21) in Untreated

Obstructive Sleep Apnoea (OSA)

Obstructive sleep apnoea (OSA) is a common and often under-diagnosed

condition (Motamedi, McClary, & Amedee, 2009). It is characterised by recurrent

episodes of partial (hypopnoea) or a near-complete (apnoea) obstruction of the

pharyngeal airway during sleep (Spicuzza, Caruso, & Di Maria, 2015). These

obstructive events are associated with drops in oxygen saturation and are usually

terminated by arousals, resulting in fragmentation of sleep (Schröder & O’Hara, 2005;

Eckert & Malhotra, 2008). Epidemiological studies have shown that 13% of men and

6% of women have OSA of at least moderate severity (Apnoea-Hypopnoea Index

(AHI) ≥ 15; Peppard, Young, Barnet, Palta, Hagen, & Hla, 2013). The condition

manifests as snoring, nocturnal gasping and choking, excessive daytime sleepiness,

reduced quality of life, and fatigue (Gupta, Chandra, Verma, & Kumar, 2010). OSA

also increases the risk of a constellation of health-related issues, including

cardiovascular problems, such as risk of hypertension (Marin et al., 2012), stroke

(Yaggi et al., 2005), and myocardial infarction (Marin, Carrizo, & Kogan., 1998), type-

2 diabetes (Tasali, Mokhlesi, & Van Cauter, 2008), and metabolic syndromes

(Vgontzas, Bixler, & Chrousos, 2005).

Previous research (e.g. Harris, Glozier, Ratnavadivel, & Grunstein, 2009;

Schröder & O’Hara, 2005; Nanthakumar, Bucks, & Skinner, 2016) has highlighted the

association between OSA and depression. Cardinal symptoms of depression include

low mood, anhedonia, hypersomnia/insomnia, fatigue/loss of energy, feelings of

worthlessness/guilt, concentration difficulty/ indecisiveness, psychomotor

agitation/retardation, suicidal ideation, appetite changes, and weight gain/loss (DSM-5,

American Psychiatric Association, 2013). Additionally, depression frequently presents


with symptoms of anxiety and stress, such as excessive worry, phobia, and irritability

(DSM-5, American Psychiatric Association, 2013). Depression questionnaires are

commonly used to assess these different features of depression in clinical and research

settings (Coyne, Thompson, Klinkman, & Nease Jr, 2002). The prevalence of current

depression in OSA when using depression questionnaires (using standardized cut-offs),

in clinical-based and population-based studies have ranged from 8% to 68%

(Nanthakumar et al., 2016). The assessment of depression in OSA is confounded by the

overlap in symptoms between the two conditions, such as insomnia, fatigue, decreased

libido, irritability, weight changes, and cognitive difficulties (Harris et al., 2009;

Nanthakumar et al., 2016). Different depression questionnaires have varying

proportions of shared symptoms, and increasing symptom overlap is associated with

increasing prevalence of depression in OSA (Nanthakumar et al., 2016). This overlap in

symptomatology makes it difficult to determine which symptoms reliably identify

depression in the presence of OSA.

The Depression, Anxiety, and Stress scale (DASS; Lovibond and Lovibond,

1995) is a frequently used measure of overall negative affect in clinical and non-clinical

populations. The DASS contains three subscales that assess depression, anxiety, and

stress symptoms, and is available in the original 42 item (14 items per subscale), and a

shortened 21 item form (DASS-21; 7 items per subscale). As noted above, anxiety and

stress symptoms are associated features of depression. Previous studies have shown that

the scale has sound psychometric properties of a correlated, three-factor structure

(depression, anxiety, and stress) across clinical and non-clinical samples (Antony,

Bieling, Cox, Enns, & Swinson, 1998; Henry & Crawford, 2005), including other

chronic disease samples (e.g. persistent pain; Wood, Nicholas, Blythe, Asghari, &

Gibson, 2010).


Measurement invariance assesses the equivalence of the factor structure

parameters (i.e. factor structure, factor loadings, factor means, item loadings) across

different samples (Bryne & Watkins, 2003; Nye, Roberts, Saucier & Zhou, 2008;

Steenkamp & Baumgartner, 1998). In the absence of measurement invariance,

comparisons on a measure are problematic and ambiguous. That is, one cannot

determine if an observed difference between samples represent a true difference in the

underlying construct, or systematic biases in the way people from different samples

perceive and respond to items (Horn & McArdle, 1992; Cheung & Rensvold, 2002;

Byrne & Watkins, 2003). When examining the symptoms assessed in the DASS-21 in

an OSA sample, there are two underlying factors predicting a person’s score, 1) the

degree to which there are true differences in the way that the person is genuinely

depressed/anxious/stressed, and 2) the degree to which they are also endorsing items as

a function of their OSA. Therefore, questionnaires that consist of a high proportion of

overlapping symptoms with OSA may not be assessing genuine

depression/anxiety/stress in OSA.

In a recent meta-analysis by Nanthakumar et al., (2016), symptoms of

depression in commonly-used depression scales were categorised into overlapping and

non-overlapping symptoms. These categories can also be used to divide the DASS-21

items. Non-overlapping symptoms of depression reflect anhedonia (e.g. I felt that I had

nothing to look forward to), or depressive cognitions (e.g. I felt I wasn’t worth much as

a person). Non-overlapping symptoms of anxiety assess anxious cognitions (e.g. I felt

scared without any good reason), and non-overlapping stress symptoms assess stress

cognitions (e.g. I felt that I was using a lot of nervous energy) and agitation/restlessness

symptoms (e.g. I found myself getting agitated). By contrast, the Anxiety subscale

assesses overlapping somatic symptoms (e.g. I was aware of dryness of my mouth, and,


I experienced breathing difficulty (e.g. excessively rapid breathing, breathlessness in

the absence of physical exertion)), and the Stress subscale assesses overlapping

irritability symptoms (e.g. I felt that I was rather touchy, and, I was intolerant of

anything that kept me from getting on with what I was doing), whereas the Depression

subscale contains no overlapping symptoms. No study to date has examined the impact

of overlapping symptoms on the suitability of the total DASS-21 in an OSA sample,

compared to a non-OSA sample.

The present study examined, a) if Lovibond & Lovibond’s (1995) correlated

three-factor structure holds in an OSA sample, compared with a non-OSA sample, and

b) measurement invariance across these samples. We hypothesized that there may be a

lack of measurement invariance across samples, primarily due to differential fit in the

Anxiety and/or Stress subscales, due to symptom overlap.

Method

Participants and protocol

Participants were selected from the Busselton Healthy Ageing Study. Busselton

is a coastal community in the South-West of Western Australia with a relatively stable

population of predominantly European descent (James et al., 2013). All, non-

institutionalised adults born from 1946 – 1964, aged 45 – 66 years at the time of the

baseline survey, who currently live in the Shire, and are listed on the electoral roll are

eligible to participate. Data were obtained from Phase 1 of this longitudinal study.

Informed consent was obtained from each participant, who completed a comprehensive

medical and lifestyle questionnaire (e.g. demographics, questions about mood, sleep

history, smoking history, alcohol consumption, presence of chronic diseases), and

physical health assessments. Sleep apnoea screening devices were offered to

participants to use over one night at home. Participants who received current CPAP or


mandibular advancement splint therapy for sleep apnoea were excluded from the sleep

apnoea screening study. Sleep studies were conducted using dual channel (oximetry and

nasal pressure) ApneaLinkTM devices (Resmed, Sydney, Australia). Participants were

instructed on how to fit and operate the units, and the monitors were returned to the

study centre the following day. Data were downloaded and automatically analysed

using ApneaLink software version 8.00. Scores of respiratory disturbances including

AHI, number of snoring events, and oxygen desaturation were recorded. The

ApneaLink has been found to be a good screening tool for OSA (Ng et al., 2009;

Erman, Stewart, Einhorn, Gordon, & Casal, 2007; Gantner et al., 2010; Crowley et al.,

2013). An AHI of 15 on an ApneaLink has adequate sensitivity (66.7 - 100.0%) and

specificity (88.5 - 100.0%) in indicating the presence of OSA compared to full

polysomnography (see Crowley et al., 2013; Ng et al., 2009; Erman et al., 2007;

Ragette, Wang, Weinreich, & Teschler., 2010). The study received approval from the

University of Western Australia Human Research Ethics Committee.

OSA sample. There were 2129 individuals recruited at baseline who completed

an overnight sleep study, of whom 285 (13.39%) had an AHI of 15 using the

ApneaLink, and reported no other sleep disorder (based on self-reported medical

history). Mean age was 58 years (SD: 5.26; range: 45 to 68 years), mean BMI was

31.03 (SD: 6.12; range: 17.70 to 65.70), mean AHI was 24.68 (SD: 11.75; range 15 to

89). 109 (38.25%) individuals were female.

Age and gender matched controls (Non-OSA sample). From the remaining

individuals whose AHI was <15 and who reported no sleep disorder, 285 (15.46%) age-

and gender-matched individuals were selected. These individuals were selected by

pairing an OSA participant to a non-OSA participant who was of the same gender and

age +/- 5 years of the age of the OSA participant. Mean age was 58 years (SD: 5.24;


range: 45 to 68 years), mean BMI was 30.64 (SD: 4.20; range: 21.87 to 44.42), and

mean AHI was 5.16 (SD: 3.99; range: 0 to 14). The AHI was 5 in 59.30% and 10 in

88.10% of individuals, and 109 (38.25%) were female. There was no significant

difference in BMI, sleepiness (based on the Epworth Sleepiness Scale (ESS) scores),

and age between the two samples (p > .05). As expected, AHI was significantly lower

in the age-and gender-matched controls (p < .01).

Materials

Depression, Anxiety, and Stress Scale-short form (DASS-21) (Lovibond &

Lovibond, 1995). The DASS-21 is a 21 item, self-report questionnaire designed to

measure depression, anxiety, and stress symptoms. It is the short form of Lovibond and

Lovibond’s (1995) 42-item version from which the 42 highest loading items were

selected (7 from each of the 3 subscales; Lovibond & Lovibond, 1995). Participants

indicate the degree to which they felt negative affect symptoms over the past week,

from 0 (did not apply to me at all) to 3 (applied to me very much, or most of the time).

Scale scores range from 0 to 21, with higher scores indicating greater depression,

anxiety, and stress, respectively. The subscales have good internal reliability

(Depression subscale: cronbach’s α = .82 – .84; Anxiety subscale: cronbach’s α = .74 –

.81; Stress subscale: cronbach’s α = .74 – .88; Norton, 2007), good convergent and

discriminant validity (Henry & Crawford, 2005), and the full-scale has a reliable

correlated three-factor structure in clinical and non-clinical samples (Henry &

Crawford, 2005; Antony et al., 1998).

Data analysis Plan. Data screening and descriptive analyses were performed

using SPSS 19.0 for Windows, IBM (2015B). CFA and measurement invariance

analyses were performed using MPlus, version 3.0 (Muthén & Muthén, 1998-2011) and

dMacs (Nye and Drasgow, 2011).


Confirmatory Factor Analysis (CFA). CFA was used to examine Lovibond &

Lovibond’s (1995) correlated three-factor structure of the DASS-21 in the OSA and

non-OSA samples. As recommended by Muthén and Muthén (1998-2011), we used

maximum likelihood estimation with robust standard errors as this is robust to non-

normality. χ2 was used to evaluate overall model fit, along with the Comparative Fit

Index (CFI; Bentler, 1990), Tucker-Lewis Index (TLI; Tucker & Lewis, 1973) Root-

Mean-Square Error of Approximation (RMSEA; Steiger & Lind, 1980), and

Standardized Root Mean Squared Residual (SRMR; Jöreskog & Sörbom, 1993).

Recommended cutoffs for the fit indexes were: CFI: .95, TLI: .95, RMSEA: .050,

and SRMR: .060 (Hu & Bentler, 1999).

Measurement invariance. Provided that the correlated three-factor structure of

the DASS-21 was an acceptable fit in the non-OSA sample, the multi-group

confirmatory factor analysis (MGCFA) approach for testing measurement invariance

was conducted by performing configural, metric, and then scalar invariance testing

(Vandenberg & Lance, 2000; Cheung & Rensvold, 2002). A chi-square difference test

using scaling correction factors for MLR across unconstrained and a series of

constrained measurement models was performed to examine the different levels of

measurement invariance (Muthén & Muthén, 1998-2011). The unconstrained model is

estimated without any conditions, whereas the constrained models are estimated with

the condition that one or more specified factor parameters would have the same value

for both samples (Muthén & Muthén, 1998-2011). A lack of measurement invariance is

indicated by a statistically significant chi-square difference between the respective pair

of models (Muthén & Muthén, 1998-2011) and a ΔCFI of >.002 (Meade, Johnson, &

Braddy, 2008).

Configural invariance was tested first, which evaluates whether the same overall


factor structure holds in the two samples (Vandenberg & Lance, 2000; Milfont &

Fischer, 2010; Cheung & Rensvold, 2002). Configural invariance is a prerequisite for

higher levels of equivalence testing (i.e. metric and scalar invariance). If configural

invariance was met, metric invariance and scalar invariance were tested. Metric

invariance evaluates the degree to which the factor loadings are equal across samples,

and scalar invariance assesses the equality of intercept terms (i.e., means/thresholds)

between the samples (Vandenberg & Lance, 2000; Milfont & Fischer, 2010; Cheung &

Rensvold, 2002).

If measurement non-invariance was found, confirmatory factor analytic (CFA)

mean and covariance structure (MACS) analysis to examine the source of non-

invariance by assessing differential item functioning (DIF), was conducted using

dMACS (see Nye and Drasgow, 2011), evaluating: 1) which items are being responded

to in a different way between the samples, and 2) the magnitude of these differences

between the samples (i.e. DIF). Effect sizes were compared to Cohen’s (1988)

guidelines (0.20 = small, 0.50 = medium, 0.80 = large).

Results


Means, SDs, ranges, alpha values, and difference in means for the three subscale

scores for each sample are shown in Table 4.1.


Table 4.1

Characteristics of the DASS-21 subscales for the Non-OSA and OSA samples

Subscale Non-OSA sample

(N = 285)

OSA sample

(N = 285)

Difference

in means

between

samples

M SD Range Alpha M SD Range Alpha

Depression 3.36 4.59 0-24 .83 4.17 5.91 0-34 .87 p >.05

Anxiety 2.25 3.44 0-26 .68 2.84 3.94 0-18 .66 p >.05

Stress 5.75 6.16 0-34 .86 6.74 7.00 0-34 .88 p >.05

Note. Scores on the DASS-21 are multiplied by 2 to calculate the final subscale scores, producing a maximum of 42

points.

Confirmatory Factor Analysis (CFA)

The correlated three-factor structure (Model 1a) had fairly poor fit (Table 2).

Theoretically, the requirement for any item to load exclusively on a single latent

construct has been problematic in CFA because items within a measure can sometimes

be multidimensional (see Morin, Arens, & March, 2016), which may account for

inadequate fit of the correlated three-factor model. Given that work from Nanthakumar

et al., (2016) and Fried (2015) suggests that depression, anxiety and stress symptoms

are multidimensional and may cluster into subfacets (e.g. anhedonia, cognition,

cognitive symptoms), we allowed error terms within subscales to correlate. As in Szabó

(2010), correlations were based on modification indices supported by a conceptual

rationale, restricting correlated errors to item pairs within a subscale. Correlating error

terms to improve model fit was recommended by Henry and Crawford (2005), and has

been applied in other studies (e.g. Asghari, Saed, & Dibajnia, 2008; Wood et al., 2010;

Szabó, 2010) that have examined the factor structure of the DASS-21. Correlated pairs

included four Depression item pairs reflecting anhedonia or cognition symptoms (range

of modification indices (MI): 4.03 – 47.20), six item pairs from the Anxiety subscale


reflecting consequences of panic/anxiety (range of MI: 6.12 – 13.54), and one item pair

from the Stress subscale reflecting restlessness (range of MI: 9.32). Permitting

correlated errors (see Table 4.3) produced a well-fitting model (Model 1b, Table 4.2).

When this model was then tested with the OSA sample (Model 2, Table 4.2), fit

statistics indicated less than ideal fit, according to the significant χ2, TLI, and CFI, so

we proceeded to invariance testing.

Table 4.2

Lovibond and Lovibond’s (1995) correlated three- factor structure (with correlated

errors) of the DASS-21 CFA in the Non-OSA and OSA samples

Note. Recommended cutoffs for the fit indexes: SRMR: .060; CFI: .95; TLI: .95; and RMSEA: .050 (Hu &

Bentler, 1999). LO 90 = Lower boundary of 90% confidence interval; HI 90 = Upper boundary of 90% confidence

interval.

Measurement invariance

Model df χ2 p SRMR CFI TLI RMSEA LO 90 HI 90

1a. Non-OSA

sample:

Correlated three-

factor structure

186 314.31 <.001 .060 .90 .88 .049 .040 .058

1b. Non-OSA

sample:

Correlated three-

factor structure

(with correlated

errors)

175 197.55 .117 .049 .98 .98 .021 0 .035

2. OSA sample:

Correlated three-

factor structure

(with correlated

errors)

175 271.22 <.001 .049 .94 .93 .044 .033 .054


Model 1b served as the baseline model in evaluating measurement invariance

across samples. Results from configural invariance analysis indicated that the ∆ χ2 was

significant (p < .001) and ∆ CFI = .003. Therefore, further invariance analyses were not

conducted.

Given the lack of configural invariance, dMACS was used to investigate the

source of measurement non-invariance by assessing DIF. Mplus estimates of the item

parameters that were used to calculate the effect sizes for the Depression, Anxiety and

Stress subscales are provided in Table 4.4. The first column shows the effect sizes of

non-equivalence for each of the items (dMACS), and the last four columns show the

factor loadings and the intercepts for the OSA and non-OSA samples.

A range of non-equivalence was found in the subscales. The broadest range of

effect sizes was observed for the Anxiety subscale, where effects ranged from very

small, 0.12, for the item, Scared without reason, to medium, 0.57, for the item, Action

of my heart in the absence of physical exertion. Inspection of the loadings and

intercepts indicated that the items, Dryness in mouth and Action of my heart in the

absence of physical exertion produced the largest differences between the two samples,

particularly at the upper end of the latent trait continuum (see Nye and Drascow, 2011).

Figure 4.1 illustrates the impact of DIF on the scores for these two items (1a. Dryness

of mouth; 1b. Action of my heart) contrasted with a third item (1c. Worry about panic

and making a fool of self), which had an effect size of 0.31. For the first two items, at

higher levels of anxiety (the latent trait), individuals with OSA endorsed a lower

severity score than those without OSA. That is, DIF produced lower responses on the

subscale item for similar levels of the latent trait of anxiety for the OSA sample than the

non-OSA sample. For the third item illustrated, the effect was the opposite, OSA

individuals with higher levels of anxiety scored higher on that item than those without


OSA. Inspection of the pattern of loadings across the anxiety items revealed that 4 had

higher loadings for the non-OSA sample and 2 for the OSA sample. A small range of

non-equivalence was identified in the Depression and Stress subscales, but there were

no meaningful differences in the way DIF affected the way these items were being

responded to between the samples.

Collapsing across all the items within each subscale revealed only small effects

of DIF on the scales’ properties (see Table 4.3). The largest difference in mean was -

0.49 points for the Stress subscale and the smallest, despite the effects described above,

was -0.29 points for the Anxiety subscale. The negative values in this column suggest

that DIF results in higher means for the OSA sample. However, these differences were

in raw score points, therefore, compared with the non-OSA sample, the OSA sample

had means that were 0.39 points higher for the Depression subscale, 0.29 points higher

for the Anxiety subscale, and 0.49 points higher for the Stress subscale, due to DIF. As

these differences were at the scale level, they must be interpreted relative to a 42-point

maximum score (i.e. 7 items with 4 response options coded 0 to 3, doubled to 42),

which indicates that they are small. DIF accounted for 97.5%, 100%, and 98% of the

observed differences in means between the samples on the Depression, Anxiety, and

Stress subscales, respectively. Therefore, the majority of the observed differences

between the means of the two samples were due to DIF, rather than true mean

differences, but the overall effect on scale means scores was negligible

.


Table 4.3

Item parameters for the Depression, Anxiety, and Stress subscales

Note. MACS = mean and covariance structure. a = referent item. dMACS values are effect sizes: 0.20 = small, 0.50 = medium, 0.80 = large. *Latent means were set to 0 on dMACS (see Nye & Drascow, 2011). Factor loadings are unstandardized.

Subscript numbers that are the same within each subscale indicate that the error terms for these item pairs were permitted to

correlate.

dMACS Factor loading Intercept

Subscale, (item number), and

item name

Non-OSA

sample OSA sample

Non-OSA

sample OSA sample

Depression subscale

(3) Positive feelings 0.30 0.76 0.91 0.22 0.31

(5) Initiative 1,2 0.09 0.88 0.88 0.51 0.56

(10) Look forward to 1 0.25 0.89 0.73 0.16 0.18

(13) Down-hearted and blue a 0.20 1.00 1.00 0.29 0.36

(16) Enthusiastic 2,3 0.22 0.98 1.01 0.31 0.39

(17) Self-worth 4 0.21 0.66 0.70 0.13 0.19

(21) Life was meaningless 3,4 0.11 0.45 0.42 0.07 0.09

Latent mean* 0 0

Latent variance 0.14 0.23

Anxiety subscale

(2) Dryness of mouth 0.47 1.41 0.45 0.37 0.49

(4) Breathing difficulty 5, 6, 7,8, 9 0.18 1.12 0.92 0.13 0.12

(7) Trembling 5 0.18 0.47 0.66 0.09 0.12

(9) Worry about panic and

making a fool of self 6, 10 0.31 1.10 1.45 0.19 0.24

(15) Close to panic 7 0.23 0.94 0.91 0.07 0.14

(19) Action of my heart 8 0.57 1.58 0.71 0.20 0.21

(20) Scared a, 9, 10 0.12 1.00 1.00 0.08 0.11

Latent mean* 0 0


Stress subscale

(1) Wind down 11 0.16 0.79 0.92 0.53 0.60

(6) Over-react 0.17 1.06 0.95 0.52 0.59

(8) Nervous energy 0.20 0.62 0.70 0.22 0.29

(11) Agitated 0.22 1.04 0.92 0.42 0.48

(12) Difficult to relax a, 11 0.21 1.00 1.00 0.40 0.50

(14) Intolerant 0.20 0.83 0.93 0.40 0.48

(18) Touchy 0.15 0.95 0.93 0.38 0.44

Latent mean* 0 0



In regards to the variance of the subscales, in the most extreme case, non-

equivalence accounted for variance 1.67 points higher on the Anxiety subscale in the

OSA sample than the non-OSA sample. Additionally, DIF accounted for 0%, 135.77%,

and 9.34% of the observed differences in variance between the samples on the

Depression, Anxiety, and Stress subscales, respectively. Therefore, the observed

differences between the variances of the two samples on the Anxiety subscale were

likely due to DIF, albeit the variances were not large to begin with.

Figure 4.1. Comparing the items (1a: Dryness of mouth, 1b: Action of my heart, and

1c: Worry about panic and making a fool of self, across the OSA and non-OSA

samples.


Table 4.4

Measurement Equivalence of the Depression, Anxiety and Stress subscales between

the non-OSA sample (referent group) and OSA sample (focal group)

Subscale ∆ Mean

Observed

mean

difference

% of

observed

mean

difference

∆ Var

Differences

in observed

variance

% of

observed

differences

dMACS

min - max

Depression -0.39 -0.40 97.5 0.03 -3.37 0 0.09 - 0.30

Anxiety -0.29 -0.29 100 -1.67 -1.23 135.77 0.11 - 0.57

Stress -0.49 -0.50 98 0.24 -2.57 9.34 0.15 - 0.22

Note. DIF = differential item functioning. ∆Mean = amount of observed mean differences that can be attributed to

DIF (referent group –focal group). Observed mean difference = DIF + true mean differences. ∆Var = difference

in the variances of the subscale due to DIF. Differences in observed variance = DIF + true variances. MACS =

mean and covariance structure. For all difference metrics, negative values indicate larger values in the OSA sample than in the non-OSA sample.

Discussion

The present study explored Lovibond and Lovibond’s (1995) correlated three

factor structure of the DASS-21 in OSA and non-OSA samples by means of CFA and

examined measurement invariance. The results showed that Lovibond and Lovibond’s

(1995) correlated three factor structure did not fit the OSA sample as well as the non-

OSA sample. Additionally, configural invariance was not found, suggesting that both

samples may be interpreting the constructs of the DASS-21 differently (Cheung &

Rensvold, 2002; Milfont & Fischer, 2010). Inspection of the individual items within the

subscales to determine the source of this difference indicated that overall, DIF led to

slightly higher factor means and variances for each of the subscales in the OSA sample,

compared with the non-OSA sample. However, the actual difference that this produced

in mean subscale scores was small (between just 0.29 and 0.49 of a point, i.e. less than

half a point across any subscale, overall). On balance, therefore, we take the view that

the degree of DIF for each of the Depression, Anxiety and Stress subscales of the

DASS-21 is not sufficient to cause concern when using the full-scale DASS-21 to

assess negative affect in an OSA sample.


Interestingly, there was no impact of the symptom overlap within the Stress

subscale. However, inspection of the problematic items within the Anxiety subscale

does challenge our understanding of the impact of symptom overlap on the

measurement of negative affect in clinical disorders or chronic illness. Whilst the mean

of the Anxiety subscale differed only by 0.29 of a point, there were modest problems

with two symptoms within this subscale. For two physiological symptoms of anxiety

(items: dryness in mouth, and action of my heart in the absence of physical exertion),

small to medium-sized differences in the way these items were endorsed were evident,

particularly at the upper end of the latent trait continuum. That is, highly anxious

respondents in the OSA and non-OSA samples responded differently to these items.

Whilst dryness of the mouth was identified as an overlapping symptom with OSA, heart

action was not categorised as an overlapping symptom in Nanthakumar et al., (2016).

On reflection, however, OSA is associated with increased risk of atrial fibrillation (see

Gami et al., 2007), which may be experienced as heart beat changes in the absence of

exertion. Strikingly, the effect of symptom overlap for these items produced lower

severity scores in the OSA sample, despite higher mean subscale scores. This contrasts

markedly with the evidence that overlapping symptoms of depression and OSA leads to

inflation of depression scores, thus producing higher prevalence estimates of depression

in OSA (Nanthakumar et al., 2016). Therefore, the impact of symptom overlap on

depressive symptomology may have separate, and opposing effects, depending on the

nature of the symptom and the clinical disorder or chronic illness. From the point of

view of each DASS-21 subscale, these opposing effects appear to cancel each other out,

making the DASS-21 suitable for use in OSA in its entirety.

The present study is the first to examine whether the full-scale of the DASS-21

is suitable for use in OSA, and examine the impact of symptom overlap. Although


individuals were recruited from an epidemiological sample, there was a limited age

range as the sample consisted of baby boomers who were on the electoral roll. It is

possible that there may be differential effects in measurement invariance in younger or

older age groups (Gabbay & Lavie, 2012). Further, although an ApneaLink has been

found to have adequate sensitivity and specificity as a screening tool for diagnosing

OSA, overnight polysomnography (PSG), which is the gold standard for diagnosing

OSA, was not available (Ragette et al., 2010; Mansfield, Antic, McEvoy, 2013).

Samples did not differ in terms of sleepiness (based on ESS scores), which might

suggest that the findings are not generalizable, as excessive daytime sleepiness is

thought to be a cardinal symptom of OSA (Harris et al., 2009). However, previous

studies have argued that there is no clear association between the severity of OSA and

sleepiness (e.g. Vgontzas, 2008; Young et al., 1993), therefore, this lack of a group

difference in sleepiness does not affect the generalisability of the findings. Additionally,

as individuals in the non-OSA sample were matched to the OSA sample on age, gender,

and as close as possible to BMI, we had insufficient participants to select non-OSA

controls with an AHI of 10 or even 5. That said, 88% of the non-OSA sample had

an AHI of 10 and 59% had an AHI of 5. Future studies may wish to recruit a wide

age-range, diagnose OSA with PSG, and use a lower AHI cut-off (e.g. AHI 5) to

select individuals in the non-OSA sample, to further examine any difference between

OSA and non-OSA samples, and to examine measurement invariance.

Whilst this study suggests that the full-scale DASS-21 is suitable for use in

OSA, this is clearly an area that warrants greater consideration both in relation to OSA,

with respect to other negative affect measures which may suffer greater DIF, and other

clinical disorders or chronic illness conditions such as Insomnia (Carney et al., 2009),

Parkinson’s disease (Schrag et al., 2007), and diabetes (Anderson, Freedland, Clouse, &


Lustman, 2001), as DIF may impact the suitability of other questionnaires used to

assess negative affect in these conditions by inflating or reducing the degree to which

symptoms are endorsed, in unexpected or confounding ways.


CHAPTER 5: EFFECT OF CPAP ON THE DASS-21 SYMPTOMS IN OSA

Preface

Study 1 (Chapter 2) indicated through meta-analytic techniques that symptom

overlap potentially confounds the assessment of depression in OSA.

Study 2 (Chapter 3) reported the impact of CPAP treatment on symptoms of

depression using the HAM-D, dividing the symptoms into those that overlap with OSA

and those that do not. As noted in the chapter’s discussion, this is not an ideal way to

address the problem of symptom overlap, albeit it revealed that overlapping symptoms

were more severe and responded better to CPAP treatment than non-overlapping

symptoms. Additionally, although there was improvement in depressive symptoms at

post-treatment, it is unclear if CPAP leads to clinical significant improvement in

depressive symptoms for OSA individuals.

Study 3 (Chapter 4) then explored whether the Depression Anxiety and Stress

Scale – Short Form (DASS-21) was suitable for use in OSA by explicitly testing

whether it was measurement invariant. Comparing samples with and without OSA, the

study revealed that the DASS-21 was not substantially impacted by symptom overlap,

and is suitable in an OSA sample.

As the DASS-21 is substantially measurement invariant in OSA, any change in

the three subscale (depression, anxiety and stress) scores, represents a “true change” in

depression, anxiety and stress, over time, rather than the effect of DIF. This study

(Chapter 5) examined the effect of CPAP on the DASS-21 using the per protocol data

from Chapter 3 (Study 2) by using latent change modelling and clinical significance

analyses.


Abstract

The present study examined the effect of CPAP on the DASS-21 subscales

(depression, anxiety, and stress) for the per protocol participants from Chapter 3 (Study

2), using latent change modelling and clinical significance methodology. Based on

latent change modelling, depression and anxiety symptom scores significantly

decreased from pre- to post-treatment, albeit the decrease was very small, and the

decline in symptoms was not a function of CPAP use. Stress symptom scores did not

significantly change. In regards to clinical significance improvement, across all

subscales, the majority of individuals did not make a clinically significant improvement

with treatment, albeit most of the individuals were in the normal range for anxiety,

depression, and stress, at baseline. A higher proportion of individuals’ depression,

anxiety and stress scores were in the functional population range than the dysfunctional

population range at post-treatment, than not. Implications are discussed.


Effect of Continuous Positive Airway Pressure (CPAP) on the Depression,

Anxiety and Stress scale (DASS-21) symptoms in OSA using latent change

modelling and clinical significance analysis

As the assessment of depression in OSA is potentially confounded by symptom

overlap, it is unclear if CPAP ameliorates both overlapping and non-overlapping

depressive symptoms in OSA and/or if it does so to the same extent. To examine this

question, Chapter 3 (Study 2) of this thesis compared the effect of CPAP in

ameliorating overlapping and non-overlapping depressive symptoms, assessed with the

Hamilton Rating Scale for Depression (HAM-D), and examined if the decline in

depressive symptoms was related to CPAP use, in a 12-week, single arm study

(Obstructive Sleep Apnoea and Related Symptoms (OSARS) study).

Both overlapping (e.g. genital symptoms, psychomotor retardation, stress, sleep

difficulties) and non-overlapping (negative affect, suicidal ideation, anhedonia)

depressive symptoms significantly decreased over time, with a greater decrease in the

severity of overlapping depressive symptoms. Critically, greater CPAP use was

associated with a faster rate of decline in overlapping depressive symptoms, alone.

These findings suggest that overlapping symptoms are more responsive to CPAP

treatment.

Although dividing the HAM-D into overlapping and non-overlapping symptoms

and examining the effect of CPAP on symptom change was one solution to exploring

the problem of symptom overlap, it has its limitations. The categorisation of symptoms

as overlapping and non-overlapping was determined theoretically, rather than by

exploring the ways in which individuals with and without OSA respond to items

differently, statistically.

Chapter 4 (Study 3) of this thesis revealed that the Depression, Anxiety, Stress


Scale (DASS-21; Lovibond & Lovibond, 1995) is substantially measurement invariant

in OSA, therefore, individuals with OSA do not respond differently to the items within

the DASS-21 because of their OSA diagnosis (that is, differential item functioning

(DIF) was not an issue). The DASS-21 assesses anxiety and stress symptoms, which are

associated features of depression (American Psychiatric Association, 2013). Arguably,

therefore, the DASS-21 can be used in an OSA population with some confidence, and

any change in subscale scores can be regarded as a change in depressive symptoms,

rather than a change in OSA symptoms. They are the parallel symptoms to the non-

overlapping HAM-D depressive symptoms, in Chapter 3 (Study 2).

Latent change modelling (LCM) is a way to examine the effect of CPAP on

depressive symptoms. LCM produces a latent change score (the change between pre-

and post-treatment scores) for each individual, and explicitly, models the variance in

the change scores, which allows us to explore whether all individuals experience similar

change, over time, having accounted for measurement error (McArdle, Ferrer-Caja,

Hamagami, & Woodcock, 2002; Geiser, 2012; Steyer, Eid, & Schwenkmezger, 1997).

Additionally, it allows us to examine the association between covariates (such as CPAP

use) on the latent change score. LCM are sometimes referred to “true change models”.

However, an issue with LCM is it does not allow us to examine if the difference in

scores over time for an individual, is clinically-significant.

Although, it is important to examine the mean change in symptom scores with

CPAP treatment, group mean changes obscure individual differences in treatment

outcome. There could be a significant improvement in symptoms between pre- and

post-treatment (at group a level), it is possible that there are individuals who have

worsened or do not change with treatment (Guyatt, Osoba, Wu, Wyrich, & Normal,

2002). Therefore, a treatment effect may be statistically significant in cases where the


actual change is small and not clinically meaningful (Nelson & Allred, 2005).

Assessing an individual’s change in symptoms during treatment is very important when

evaluating the effectiveness of the treatment and also helps clinicians to guide their

decisions regarding an individual’s future health care. As the gold standard treatment

for depression is psychotherapeutic intervention and/or antidepressant medication, it is

possible that some individuals with OSA may also require psychotherapeutic

intervention and/or antidepressant medication to ameliorate depressive symptoms that

do not respond to CPAP treatment (Misri, Reebye, Corral, & Mills, 2004).

Clinical significance methodology (Jacobson, Follette, & Revenstorf, 1984)

provides a valuable way of evaluating the symptomatic changes an individual makes

between two time-periods (e.g. pre-treatment and post-treatment). This methodology

takes into consideration: a) whether an individual experiences a change that is

considered statistically reliable, and b) whether an individual’s post-treatment score

resembles the score of an individual in the functional, healthy population, or the

dysfunctional, treatment-seeking population (Jacobson & Truax, 1991). Although

clinical significance methodology has been used in clinical trials to assess the

effectiveness of treatments (e.g. Asarnow et al., 2005), this methodology has never been

examined in regards to the impact of CPAP on depressive symptoms, in an OSA

sample.

As patient characteristics, such as BMI, age, antidepressant use, gender, and

baseline AHI have been found to influence the presentation of depressive symptoms in

OSA (Nanthakumar, Bucks, & Skinner, 2016; Harris, Glozier, Ratnavadivel, &

Grunstein, 2009), these characteristics were examined in the latent change modelling

analyses to determine whether they predicted any change in depression, anxiety, and

stress subscale scores with treatment.


Therefore, the present study investigated the effect of CPAP on the DASS-21

using the per protocol data from Chapter 2 (sample recruited from the Obstructive Sleep

Apnoea and Related Symptoms (OSARS) study; 12-week, single arm study), by

examining the effect CPAP has on depression, anxiety and stress subscale scores, from

pre-treatment to post-treatment, using latent change modelling. Subsequently, clinical

significance methodology was used to examine the proportion of individuals who made,

and did not make, a clinically significant improvement over time.

It was hypothesized that, 1) there would be a significant reduction in depression,

anxiety, and stress scores, 2) and this change in subscale scores would be associated

with greater CPAP use. In regards to clinical significance, it was hypothesized that

CPAP treatment would lead to a greater proportion of OSA individuals making

clinically-significant improvement than not, across all subscales.

Method

The data used for this study were from the Obstructive Sleep Apnoea and

Related Symptoms (OSARS) study. The OSARS study was a longitudinal, 12-week,

single arm trial of CPAP. As outlined in Chapter 3 (Study 2), all participants provided

written, informed consent. Participants, male and female, aged 18 years old and over,

were referred between January 2014 and March 2016 to the Sleep Disorders Clinic at

the Queen Elizabeth II Medical Centre for diagnosis of suspected OSA. Patients with

OSA (AHI >5; derived from overnight PSG), and who met the inclusion criteria (e.g.

AHI < 5, central (non-obstructive) sleep apnoea (>50% of apnoeic/hypnoeic events are

non-obstructive); see methodology outlined in Chapter 3 (Study 2) for the full list of

exclusion criteria), were invited to participate in the study. The study spanned 12

weeks, consisting of 5 clinic visits (visit 1, visit 1a, visit 2, visit 3, and visit 4), and any

unscheduled sleep visits. The DASS-21 was collected at visits 1 and 4 only.


Participants

OSA sample. As outlined in Chapter 3 (Study 2), 176 participants were

successfully screened. Of these, 135 completed the trial per protocol. As latent change

modelling and clinical significance methodology both require complete pre - and post-

treatment scores, only the per protocol data were used in this study (N = 134; 1

individual completed the trial but did not complete the DASS-21 and was excluded

from analyses). Although “last observation carried forward (LoCF)” could be used for

the individuals who withdrew from the trial (N = 41) or did not complete the post-

treatment DASS-21 measure (N = 1), this would inflate the number of individuals with

zero change from visit 1 to visit 4 and violate assumptions of normality.

Normative sample. The normative sample data (N = 395) were derived from

Crawford, Cayley, Lovibond, Wilson and Hartley (2011) and were used for the clinical

significance analyses within this study. The normative sample they recruited was

recruited to be broadly representative of the general, Australian, adult population.

Recruitment took place in Adelaide, South Australia, between 1995 and 2000 and

participants were asked to complete relevant questionnaires (e.g. DASS-21)

anonymously. For the purposes of this study, the norms used were from the age band of

25-90 years old so that the age range overlapped with the OSA sample (23 – 83 years

old) in this study. No information was provided regarding the mean age, gender

proportions and BMI for this normative sample.

Measures

Demographic details. As outlined in Chapter 3 (Study 2), participants

completed questionnaires related to demographic information and their general health

(e.g. smoking history, medical history, alcohol consumption). Height and weight

measurements were completed at Visit 1 and were used to calculate BMI.


Overnight sleep study. As described in Chapter 3 (Study 2), all individuals

underwent a standard, 12-channel PSG. Data were analysed by qualified sleep

technologists using standard criteria to identify apnoeas, hypopnoeas, and desaturations

to obtain an apnoea/hypopnoea index (AHI), which was derived using the American

Academy of Sleep Medicine scoring criteria (AASM; 2012).

Depression, Anxiety, Stress Scale (DASS-21). The DASS-21 (Lovibond &

Lovibond, 1995) was assessed at visit 1 (pre-treatment) and visit 4 (post-treatment)

only. The DASS-21 is a 21 item, self-report questionnaire designed to measure

depression, anxiety, and stress symptoms. Participants indicate the degree to which they

felt negative affect symptoms over the past week, from 0 (did not apply to me at all) to

3 (applied to me very much, or most of the time). Scale scores range from 0 to 21

(doubled to equal a sum score of 42), with higher scores indicating greater depression,

anxiety, and stress, respectively. The total scores for each subscale was used for the

clinical significance analysis.

CPAP device: usage and compliance. As outlined in Chapter 3 (Study 2),

participants were provided with a CPAP device at Visit 1. Data were downloaded from

the CPAP device/smart card at each visit (including 1A and unscheduled visits). CPAP

use was defined as 1) the mean, daily usage of CPAP in hours/per night, and 2) the

proportion of days that CPAP use was 4 hours per night (CPAP adherent according to

clinical guidelines; Weaver & Grunstein, 2008). Where there were missing CPAP use

data, e.g. due to technical faults (N = 7 visits; 4 (2.27%) participants), CPAP usage was

based on the CPAP data that were available for the other visits.

Data analysis. Data screening and descriptive analyses were performed using

SPSS 19.0 for Windows, IBM (2015B). Latent Change Modelling was performed using

MPlus, version 3.0 (Muthén & Muthén, 1998-2011).


Latent Change Models. Analyses were conducted using Maximum Likelihood

Estimation with robust standard errors (MLR), as this approach is robust to non-

normality (the DASS-21 data at visit 1 and visit 4 were positively skewed; Muthén &

Muthén, 1998-2011). Whilst linear regression or repeated measures analysis of

covariance could be used to explore change between two repeated measures and the

impact of covariates, these models cannot take account of measurement error, which

may confound results (Coman et al., 2013). Additionally, such methods do not allow for

direct testing of the significance of mean changes between pre and post treatment,

therefore, cannot explain inter-individual and group differences in change over time

(Coman et al., 2013).

In latent change models (McArdle et al., 2002; Geiser, 2012; Steyer et al.,

1997), change is measured directly through latent difference variables (Geiser, 2012).

The latent difference variable is corrected for random measurement error, as they

represent the “interindividual differences in true intraindividual change over time”

(Geiser, 2012). Therefore, latent change models allow: 1) the examination of whether

the mean of the latent difference variable significantly changes over time, and 2)

exploration of correlations between covariates and the latent difference variable.

To create latent change models for visit 1 data, each DASS-21 subscale was

divided into two parts, one part consisted of the mean of 3 items from the subscale, i.e.

(Item 1 + Item 2 + Item 3)/3 = d11, and the other part consisted of the mean of the

remaining 4 items from the subscale, i.e. (Item 4 + Item 5 + Item 6 + Item 7)/4 = d12.

Variables d11 and d12 were used as indicators for a latent variable, State 1. This

process was replicated for the visit 4 DASS-21 data (d21 and d22) to generate State 2

(see Figure 5.1). A latent difference variable (Diff: State 2 - State 1) was then created to

examine the change in error-free scores over time. This process was replicated for the


three subscales to create three, latent change models for depression, anxiety, stress

symptoms.

After creating the three latent change models, covariates: baseline AHI; BMI;

age; antidepressant use (no antidepressant use = 0, antidepressant use = 1); education in

years; gender (male = 1, female = 0); and, CPAP use (mean daily use and proportion of

4 hours of CPAP) were used to predict latent change difference scores (state 2 – state

1) to examine if the change in scores between time 1 (pre-treatment) and time 2 (post-

treatment) could be explained by the covariates. These covariates were examined for

each of the depression, anxiety and stress latent change models.

Overall model fit was evaluated using χ2, along with the Comparative Fit Index

(CFI; Bentler, 1990), Tucker-Lewis Index (TLI; Tucker & Lewis, 1973) Root-Mean-

Square Error of Approximation (RMSEA; Steiger & Lind, 1980), Standardized Root

Mean Squared Residual (SRMR; Jöreskog & Sörbom, 1993), Akaike Information

Criterion (AIC; Akaike, 1974) and parsimony ratio. Recommended cutoffs for the fit

Figure 5.1. Latent change models for two indicators Y ik (i = indicator, k = time point)

and two measurement occasions. i = time-invariant state factor loads; ik = measurement

error variable. The variable names used in this input appear in parentheses. Figure

replicated from Geiser (2012).


indexes were: CFI: .95, TLI: .95, RMSEA: .050, and SRMR: .060 (Hu & Bentler,

1999), and an AIC that is smaller was favoured (Akaike, 1974). Significance was

indicated by p < .05.

Clinical significance methodology. In addition to evaluating whether there was

a statistically significant latent change, at the group level, the degree of change was also

evaluated at the level of the individual. The current study used the Jacobson-Truax

method for calculating clinical significance (Jacobson et al., 1984; Jacobson & Truax,

1991). To calculate clinical significance, a cut-off point separating the dysfunctional

from the functional population is proposed. Jacobson and Truax (1991) defined the

functional population as individuals not involved in therapy, whereas the dysfunctional

represents those individual who are. The Cut-off C is the mid-point between the two

distributions (Hsu, 1996; Jacobson & Truax, 1991). Cut-off C is recommended when

normative values are available (Jacobson & Truax, 1991).

The formula is:

C = (𝑆0× 𝑀1)+(𝑆1× 𝑀0)

𝑆0+ 𝑆1

Where

𝑀0 = mean of normal population

𝑆0 = standard deviation of normal population

𝑀1= mean of pre-treatment scores

𝑆1 = standard deviation of pre-treatment scores


To classify clinical significance, a Reliable Change Index (RCI) for each

individual is created, which takes into account the change made by an individual over

time, on the outcome measure, in the context of the standard error of measurement

(Jacobson et al., 1984). It is calculated using the following formula:

and

where

Positive reliable change is indicated by an RCI score greater than 1.96 when

using positive measurement instruments (corresponding to an alpha level of .05). The

four possible outcomes for classifying clinically significant change are: recovered, in

which an individual has moved reliably from the dysfunctional population to the

functional population; improved, in which an individual has made a reliable change in a

positive direction, but does not yet resemble an individual of the functional population;

unchanged, in which no reliable change has been made; and deteriorated, in which an

individual has made a reliable change in a negative direction. The proportion of

individuals assigned to each category based on their responses on the DASS-21

following treatment can be used to evaluate the extent to which a treatment outcome is

successful (Jacobson & Truax, 1991).


Results


Participant characteristics are presented in Table 5.1. Table 5.2 shows the pre

and post scores and severity range percentages across the depression, anxiety and stress

subscales. The majority of the individuals at baseline across all subscales were in the

normal range for Depression (68.70%), Anxiety (67.20%), and Stress (85.80%).

Table 5.1

Characteristics of the OSA sample (N = 134)

Participant characteristics M (SD)

Age 56. 38 (13.51)

Number of males (%) 67 (50.0)

BMI (kg/m2) 34.35 (8.02)

Baseline AHI 35. 63 (22.56)

Antidepressant use (number, %) 49 (36.6)

Mean CPAP usage/hr per night 5.35 (1.75)

Proportion (%) of nights CPAP use was 4

hours

69.59 (22.26)


Table 5.2

Pre and post subscale scores and severity range percentages across the depression, anxiety and stress subscale

Note. N (%). The data for the depression, anxiety and stress subscales are positively skewed. The raw scores on the DASS-21 are doubled to 42 (Lovibond & Lovibond,

1995). Cut off ranges (Lovibond & Lovibond, 1995): Depression subscale: Normal = 0 – 9, Mild = 10-13, Moderate = 14-20, Severe = 21-27, Extremely Severe = 28+;

Anxiety subscale: Normal = 0 – 7, Mild = 8-9, Moderate = 10-14, Severe = 15-19, Extremely Severe = 20+; Stress subscale: Normal = 0-14, Mild = 15-18, Moderate = 19-

25, Severe = 26-33, Extremely Severe = 34+.

Visit N M SD Range Normal Mild Moderate Severe Extremely

Severe

Depression subscale

Visit 1

(pre-treatment) 134 7.21 8.44 0 – 38

92 (68.70)

18 (13.40) 12 (9.01) 6 (4.50) 6 (4.50)

Visit 4

(post-treatment) 134 4.45 6.99 0 - 28

111 (82.80)

5 (3.70) 10 (7.50) 5 (3.70) 3 (2.20)

Anxiety subscale

Visit 1

(pre-treatment) 134 6.31 6.91 0 - 36 90 (67.20) 10 (7.50) 21 (15.70) 7 (5.20) 6 (4.50)

Visit 4

(post-treatment) 134 4.94 5.68 0 – 28 102 (76.10) 9 (6.70) 13 (9.70) 5 (3.70) 5(3.70)

Stress subscale

Visit 1

(pre-treatment) 134 7.61 8.44 0 - 40

115 (85.80)

5 (3.70) 6 (4.50) 4 (3.00) 4 (3.00)

Visit 4

(post-treatment) 134 5.28 6.47 0 – 30

123 (91.80)

4 (3.00) 4 (3.00) 3 (2.20) 0


Latent change modelling

Table 5.3 outlines the descriptive statistics of the DASS-21 subscales from the

latent change modelling analysis.

Table 5.3

Descriptive statistics for the DASS-21 data from latent change modelling (N = 134)

Note. Latent change models are derived from the difference (State 2 – State 1) between latent variables (State

1 and State 2) based on 2 indicators using the average scores from i) items 1 to 3, and ii) items 4 to 7. Values

for i) and ii) range from 0 to 3 (the maximum severity score for any DASS-21 item).

Visit M Variance Median

Depression subscale

Visit 1

(pre-treatment)

State 1

Mean items 1 to 3 1.21 1.87

0.67


0.67

Visit 4

(post-treatment)

State 2 Mean items 1 to 3 0.90 1.34 0.50

Mean items 4 to 7 0.57 1.10 0

Anxiety subscale

Visit 1

(pre-treatment)

State 1


1.33


0.67

Visit 4

(post-treatment)

State 2 Mean items 1 to 3 0.60 0.96 0

Mean items 4 to 7 0.47 0.72 0

Stress subscale

Visit 1

(pre-treatment)

State 1


0.67


0.50

Visit 4

(post-treatment)

State 2 Mean items 1 to 3 0.69 0.92 0

Mean items 4 to 7 0.81 0.95 0.50


Latent change models and the relationships between the covariates and the

change scores from pre-treatment (visit 1) to post-treatment (visit 4)

Depression subscale latent change models. The depression data fit the latent

change model well (Table 5.4; Model 1a). As indicated by a significant, mean latent

change there was a significant decline in depression scores from pre- to post- treatment.

Additionally, there was significant variability in mean latent change between

individuals.

When covariates (age, sex, baseline AHI, antidepressant use, education years,

and CPAP use were entered into the models (proportion of 4 hours of CPAP and

mean daily use entered in separate models), no covariates predicted latent change and,

based on the fit indices, the fit of the models was not adequate (Tables 5.5 and 5.6;

Models 1b and 1c). Therefore, the decrease in depression scores over time was not a

function of any of the covariates, including CPAP use.

Anxiety subscale latent change models. The anxiety data fit the latent change

model well (Table 5.4; Model 2a). As indicated by a significant, mean latent change,

there was a significant decline in anxiety scores from from pre- to post- treatment.

Additionally, there was significant variability in the latent change between individuals

in the sample.

As with depression, none of the covariates (age, sex, baseline AHI,

antidepressant use, education years, and CPAP use (either proportion of 4 hours of

CPAP or mean daily use, separate models) predicted latent change and all of the

covariate models were poorly fitting (Tables 5.5 and 5.6; Models 2b and 2c). Therefore,

the decrease in anxiety scores over time was not a function of any of the covariates,

including CPAP use.

Stress subscale latent change models. The stress data fit the latent change


model well (Table 5.4; Model 3a). Stress scores did not significantly decline over time,

however there was a decline on trend level. There was significant variability in mean

latent change between individuals. There were there no significant covariates of

individual latent change, and the data were a poor fit to the covariate models (Tables

5.5 and 5.6; Models 3b and 3c).


Table 5.4

Latent change models for Depression, Anxiety and Stress mean subscale scores

Note. N = 134. †p < .05, ‡ p<.01. a significant variance (p < .05) AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)] where k is

the number of observed variables. A significant negative value for the mean of latent change variable indicates a significant decline in scores from visit 1 to visit 4. Recommended

cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999). Growth parameter estimates are unstandardized values.

Model number

Subscale 2 df p AIC Parsimony

Ratio RMSEA (90% CI)

CFI TLI SRMR

Latent change

Mean Variance

1a Depression

0.31

2 .855 1260.39 .13 .0

(0 - .09) 1.00 1.00 .01

-0.31‡

0.22†

2a Anxiety 2.34 2 .327 1382.23 .13 .03

(0 - .18) 1.00 .99 .04

-0.76‡

0.62‡

3a Stress 2.66 2 .264 1305.98 .13 .05

(0 - .19) 1.00 .99 .02

-0.13

0.39†


Table 5.5

Relationships between the covariates (age, sex, baseline AHI, antidepressant use, education years, BMI, and average CPAP use) and the

mean latent change variables for the Depression, Anxiety and Stress subscales

Model Subscale Latent

change 2 df p AIC

Parsimony

Ratio

RMSEA


Relationship

with latent

change

1b Depression

subscale -0.31 66.78 24 <.001 1237.30 .31

.12

(.08 - .15) .89 .84 .09

Age 0

Sex 0.05

Baseline AHI 0

Antidepressant use -0.02

Education years 0.02

BMI 0

Average CPAP use 0.04

2b Anxiety

subscale

-0.76‡

58.93 24 <.001 1395.24 .31 .11

(.07 - .14) .81 .73 .09

Age -0.01

Sex 0.07

Baseline AHI 0

Antidepressant use 0.07


BMI 0

Average CPAP use 0.06

3b Stress

subscale -0.13 56.47 24 <.001 1301.42 .31

.10

(.07 - .14) .91 .88 .09

Age 0

Sex -0.20

Baseline AHI 0



BMI 0

Average CPAP use -0.04

Note. N = 133 as one individual’s BMI was not available; AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)] where k is the

number of observed variables. Recommended cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999); †p < .05, ‡ p<.01. Growth

parameter estimates are unstandardized values.


Table 5.6

Relationships between the covariates (age, sex, baseline AHI, antidepressant use, education years, BMI, and proportion of CPAP use 4

hours) and the mean latent change variables for the Depression, Anxiety and Stress subscales

Model Subscale Latent change 2 df p AIC Parsimony

Ratio

RMSEA


Relationship with latent

change variable

1c Depression

subscale -0.31 65.42 24 <.001 1236.81 .31

.11

(.08 - .15) .89 .85 .09

Age 0.00

Sex 0.05

Baseline AHI 0.00

BMI 0.00

Antidepressant use -0.02


Proportion of 4

hours of CPAP 0.00

2c Anxiety

subscale

-0.76

58.02 24 <.001 1390.76 .31 .10

(.07 - .14) .81 .73 .09

Age -0.01

Sex 0.06

Baseline AHI 0.00

BMI 0.00



Proportion of 4

hours of CPAP 0.00

3c Stress

subscale -0.12 55.97 24 <.001 1301.42 .31

.10

(.07 - .13) .91 .88 .09

Age 0.00

Sex -0.19

Baseline AHI 0.00



BMI 0.00

Proportion of 4

hours of CPAP -0.03

Note. N = 133 as one individual’s BMI was not available; AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)] where k is the

number of observed variables. Recommended cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999); †p < .05, ‡ p<.01. Growth

parameter estimates are unstandardized values.


Clinical significance methodology

The data used in the clinical significance analysis are reported in Table 5.7.

Table 5.7

Information relevant to calculating clinical significance and reliable change indexes

based on the Jacobson-Truax Method (1991) for the DASS-21 subscales

Note. Cutoff scores are the scores that indicate whether an individual’s post-treatment score is in the dysfunctional or functional range. For the purposes of clinical-significance, normative DASS-21 data

were multiplied by 2 to render them comparable to the DASS-21 data from the OSARS study.

The first step was to examine the proportion of individuals who reported a

positive reliable change, negative reliable change, or no reliable change, and then

consider clinical significance classifications. As previously stated, the majority of

individuals scored in the normal range for depression, anxiety and stress. Therefore, we

report the percentages for the subset of individuals who scored above the mild or more

severe depression (N = 42), anxiety (N = 45) and stress (N = 19) cut offs.

Depression Anxiety Stress

Cronbach’s alpha 0.90 0.79 0.88

Mean of Normative data

(𝑀0) 4.42 2.96 7.58

Std. dev of Normative data

(𝑆0) 7.20 5.20 8.20

Mean of pre-treatment scores

(𝑀1) 7.21 6.31 7.61

Std. dev of pre-treatment

scores (𝑆1) 8.44 6.91 8.44

Cut off 5.70 4.40 7.59


Table 5.8

Number of individuals (%) from whole sample and subsample with more severe

negative affect, who recovered, improved, were unchanged, or declined over time

Classification

category Subscales

Depression subscale Anxiety subscale Stress subscale

Whole

sample

(N = 134)

Subsample

with mild or

more severe

depression

(N = 42)

Whole

sample

(N=134)

Subsample

with mild or

more severe

anxiety

(N = 44)

Whole

sample

(N = 134)

Subsample

with mild or

more severe

stress

(N = 19)

Recovered

N (%) 18 (13.4) 16 (38.1) 7 (5.2) 7 (15.9) 8 (6.0) 5 (26.3)

Improved

N (%) 7 (5.2) 7 (16.7) 3 (2.2) 3 (6.8) 9 (6.7) 9 (47.4)

Unchanged

N (%) 104 (77.6) 17 (40.5) 120 (89.6) 33 (75.0) 112 (83.6) 5 (26.3)

Deteriorated

N (%) 5 (3.7) 2 (4.8) 4 (3.0) 1 (2.3) 5 (3.7) 0

Post-

treatment

score in the

functional

range

N (%)

102 (76.1) 20 (47.6) 134 (100.0) 44 (100.0) 134 (100.0) 19 (100.0)

Note: Classification categories are based on Jacobson & Truax (1991). Recovered: Individual has made a positive

reliable change and their post-treatment score places them in the functional range ; Improved: individual had made a positive reliable change and their score at post-treatment places them in the dysfunctional range; Unchanged:

Individuals have not made a reliable change (score in the functional or dysfunctional range at post-treatment is

irrelevant); Deteriorated: Individuals have made a negative reliable change (score in the functional or dysfunctional

range at post-treatment is irrelevant).

Depression subscale. From the total sample (N = 134), the majority of

individuals did not make clinically-significant change across treatment; 18.6% of

individuals either recovered or improved. Significantly more individuals either did not

make a clinically-significant change or deteriorated, compared to individuals who either

recovered or improved (81.4% vs 18.6%; 2(1) = 52.66, p < .001). Only a minority of

individuals deteriorated. At post-treatment, a significantly higher number of

individuals’ depression scores were in the functional range, compared to the

dysfunctional range (76.1% vs. 23.9%; 2(1) = 36.57, p < .001).


In the more severe sample (N = 42), the majority of individuals either improved

or recovered (54.8%) , however there was no significant difference between the

proportion of individuals who either improved or recovered and those who did not

change or deteriorated (45.2% vs. 54.8%; 2(1) = .381, p = .537). At post-treatment,

there was no significant difference between the proportion of individuals who were in

the functional range compared to the dysfunctional range (47.6% vs. 52.4%; 2(1) =

.095, p = .758).

Anxiety subscale. From the total sample (N = 134), the majority of individuals

did not make clinically-significant change across treatment; 7.4% of individuals either

recovered or improved, but significantly more individuals did not make a clinically-

significant change or deteriorated, compared to individuals who either recovered or

improved (92.6% vs 7.4%; 2(1) = 96.99, p < .001). Only a minority of individuals’

anxiety scores deteriorated.

In the more severe sample (N = 42), the majority did not make a clinically

significant change with treatment. A significantly higher proportion of individuals

either deteriorated or did not make a change, compared to either improving or

recovering (77.3% vs. 22.7%; 2(1) = 13.09, p < .001). Although the majority of the

individuals did not improve, recover, or get worst at post-treatment, all of the

individuals’ anxiety scores at post-treatment were in the functional range.

Stress subscale. From the total sample (N = 134), the majority of individuals

did not make clinically-significant change across treatment: 12.7% of individuals either

recovered or improved. Significantly more individuals did not make a clinically-

significant change or deteriorated, compared to individuals who either recovered or

improved (87.3% vs 12.7%; 2(1) = 74.63, p < .001). Only a minority of individuals’

stress scores deteriorated.


In the more severe sample (N = 42), the majority of the individuals recovered or

improved (73.7%). A significantly higher proportion of individuals improved or

recovered compared to those who deteriorated or did not make a change (73.7 vs.

26.3% vs. ; 2(1) = 4.26, p < .050). Although the majority of the individuals did not

improve, recover, or get worst at post-treatment, all of the individuals’ stress scores at

post-treatment were in the functional range.

Discussion

The present study examined the effect of CPAP on the DASS-21 using the per

protocol data from Chapter 2 (sample recruited from the Obstructive Sleep Apnoea and

Related Symptoms (OSARS) study; 12-week, single arm study), by examining change

in depression, anxiety and stress subscale scores, from pre-treatment to post-treatment,

using latent change modelling and clinical significance analysis. In relation to latent

change modelling, it was hypothesized that, 1) there would be a significant reduction in

depression, anxiety, and stress scores, 2) and this change in subscale scores would be

associated with greater CPAP use. In regards to clinical significance, it was

hypothesized that CPAP treatment would lead to a greater proportion of OSA

individuals making clinically-significant improvement than not, across all subscales.

Latent change modelling revealed significant improvement in depression and

anxiety symptoms, but not in stress, which may be due to the majority of individuals’

stress scores being in the normal range at baseline (albeit, there was a trend for a

decline in scores). Thus, the first hypothesis was partially supported. The improvement

in depressive symptoms (which include anxiety symptoms) from pre-treatment to post-

treatment, is consistent with other studies (e.g. Edwards et al., 2015; Means et al., 2003;

El-Sherbini, Bediwy, & El-Mitwalli, 2009) and the findings discussed in Chapter 3. No

individual characteristics (age, gender, education years, BMI, antidepressant use, and


baseline AHI) were associated with the decline of depressive symptoms. Importantly,

the decline in anxiety and stress scores over time was not a function of CPAP use.

Despite this significant improvement in symptoms, the majority of individuals

did not experience a clinically-significant change over time, which may be due to the

majority of individuals being in the normal range of depression, anxiety and stress at

baseline. To address this issue, a subset of individuals with more severe depression,

anxiety and stress were examined. For the depression symptoms, there was no

significant difference between the proportion of individuals who either improved or

recovered and those who did not change or deteriorated. For the stress symptoms, the

majority of individuals either improved or recovered at post-treatment, whereas for the

anxiety symptoms, the majority did not improve or recover.

Taken together, these findings suggest, that despite low mean subscale scores at

baseline, only depression and anxiety symptoms improved significantly over time.

Critically, in the subset of individuals who reported more marked depressive symptoms

at pre-treatment, many scored in the functional range, at post-treatment. Given the

claims that CPAP is useful for treating depression (e.g. Edwards et al., 2015), there was

no association between the amount of CPAP used and decline in depressive symptoms,

based on latent change modelling. This was the case whether CPAP use was considered

as a continuous predictor, or CPAP use was categorized into more or less than the

minimum therapeutic dose (commonly 4 hours, e.g. Weaver & Sawyer, 2010).

These data can be interpreted a few ways: The improvement of depressive

symptoms in OSA may be a placebo effect, which may explain why there was a similar

effect across the subscales. A placebo effect may be the result of patient expectations,

mediated primarily by interaction with the healthcare system for OSA treatment

(Crawford et al., 2012). However, if it was a placebo effect, then it is quite surprising


that the majority of individuals’ scores were in the functional range at post-treatment.

Without a treatment control intervention, we cannot be certain that the results do not

represent a placebo effect.

Additionally, there was no great change in depression, anxiety and stress

because the sample had low levels of depression, anxiety, and stress, to begin with, thus

restricting the range of possible improvements, therefore the results could be due to

regression to the mean (Bland & Altman, 1994). To overcome this issue, we looked at

those who were above the subscale cut-offs for the clinical significant analysis, and

found that the majority of individuals improved, so there was movement, which should

have been associated with CPAP use rates, if CPAP was causing the improvement in

depressive symptoms. Albeit, perhaps participants did not use enough CPAP. Studies

(e.g. Weaver & Grunstein, 2008) state that 4+ hours of CPAP per night is good, and

neither above or below 4+ hours groups nor CPAP use as a continuous predictor,

predicted improvement, so this is unlikely. Additionally, the average hours per night

was 5 in this study, which compares favorably with other CPAP trials (Weaver &

Sawyer, 2010; McEvoy et al., 2011). Although we did not find a correlation between

CPAP use and depression, anxiety and stress subscale scores, it is plausible that

individuals respond to CPAP in different ways: individuals using less than 4 hours per

night may see a large benefit to their depressive symptom scores, whereas another

person may use CPAP for 8 hours a night, but see a little response, therefore it would be

interesting to examine whether individuals who presented with higher symptom scores

used less CPAP. Unfortunately, we were not able to conduct this analysis in those

participants with severe and extreme severe depressive symptoms only, given the

insufficient numbers in these groups. Albeit, this warrants consideration in future

research.


As discussed in Chapter 3 (Study 2), a possible explanation for the difference in

the relationship between CPAP use and the non-overlapping and overlapping symptoms

was due to imposing a post-hoc split (based on theory) to determine which symptoms

were overlapping and non-overlapping depressive symptoms. It was demonstrated

statistically, in Chapter 4 (Study 3), that the DASS-21 is measurement invariant.

Therefore, the DASS-21 symptoms are the parallel symptoms to the non-overlapping

HAM-D symptoms. Based on the two studies, regardless of using a measurement

invariant measure or imposing post-hoc split, the same result was found: non-

overlapping HAM-D symptoms and DASS-21 symptoms declined over time, and this

was unrelated to CPAP use. An overall summary of the findings from the treatment

studies in this thesis are discussed in the thesis discussion (Chapter 6).

Although a moderate proportion of individuals made a clinically-significant

improvement and subscales scores were in the functional range, a proportion of

individual either did not change or deteriorated, at post-treatment. For these individuals,

another sort of intervention, such as antidepressant medication or psychotherapy, to

reduce their depressive symptoms may be required (Luyster, Buysse, & Strollo, 2010).

This idea is further discussed in the thesis discussion (Chapter 6).

This is the first study to examine the effect of CPAP on depression, anxiety and

stress symptoms, assessed by a measurement invariant questionnaire, using latent

change modelling and clinical significance analysis. Overall, a low proportion of

individuals endorsed mild or more severe depressive symptoms, prior to CPAP

treatment. Depression and anxiety symptom scores significantly decreased over time,

however this was not a function of CPAP use. Stress symptom scores did not

significantly change over time, albeit there was a trend. Across all subscales, the

majority of individuals did not make a clinically-significant improvement, albeit most


of the individuals were in the normal range for anxiety, depression, and stress, at

baseline. When a subset of individuals with more severe symptoms were examined,

there was no significant difference between the proportion of individuals who either

improved or recovered and those who did not change or deteriorated, for the depressive

symptoms. For the stress symptoms, the majority of individuals either improved or

recovered at post-treatment, whereas for the anxiety symptoms, the majority did not

improve or recover. As we did not use a treatment control treatment condition, the

decline of these depressive symptoms over time could be a genuine effect of treatment,

or may be due to a placebo effect or regression to the mean.


CHAPTER 6: GENERAL DISCUSSION

There are multiple processes related to the development of mood dysfunction in

OSA, such as sleep fragmentation, blood gas abnormalities (e.g. intermittent

hypoxemia), and chemical changes to bodily systems (e.g. serotonergic system; Harris,

Glozier, Ratnavadivel, Grunstein, 2009). Due to the disruption of these underlying

processes, depression is prevalent in OSA. Symptom overlap potentially confounds the

relationship between depression and OSA. Therefore, this thesis aimed to examine the

impact symptom overlap potentially has on the way we understand the presentation of

depression in OSA, and the treatment of depression in OSA.

Prevalence of depression in OSA and depression questionnaires

Due to potential symptom overlap, it was unknown if the prevalence of

depression in OSA is impacted by the proportion of overlapping symptoms in validated

depression questionnaires. The meta-analysis outlined in Chapter 2 (Study 1) found that

the aggregated prevalence of depression in OSA is 27%, whilst prevalence estimates

ranged from 8-68%. More critically, questionnaires that consist of a higher proportion

of shared symptoms are associated with higher depression prevalence estimates,

compared to questionnaires that have a lower proportion of overlapping symptoms.

Therefore, using questionnaires that consist of a lower proportion of overlapping

depressive symptoms may reduce the likelihood of overestimating the prevalence of

depression in OSA. Based on these findings, the assessment of depression in OSA

appears to be confounded by symptom overlap.

This idea is consistent with our findings from Chapter 3 (Study 2) where

individuals with OSA endorsed a greater proportion of OSA-related depressive

symptoms (overlapping symptoms), compared to non-overlapping symptoms

(symptoms independent of OSA). Taken together, these results suggest that individuals


with OSA may be responding to overlapping symptoms differently, compared to

individuals without OSA: perhaps reflecting differential item functioning; DIF.

Due to this issue, the type of questionnaire used to assess depressive symptoms

in OSA is very important. One way to address the issue of symptom overlap is to use

questionnaires that consist of a low proportion of overlapping symptoms. The findings

from the Study 1 meta-analysis revealed that the Hospital Anxiety and Depression –

Depression subscale (HADS-D; Snaith & Zigmond, 1986) and the Geriatric Depression

Scale (GDS; 15-item short form; Yesavage & Sheikh, 1986) consist of a low proportion

of overlapping symptoms, therefore these measures may be appropriate when assessing

depressive symptomatology in OSA, albeit these measures would require invariance

testing.

Another way, which is the recommended option, is statistically to examine

whether a depression questionnaire is measurement invariant (i.e. not impacted by DIF)

in OSA. We demonstrated in Chapter 4 (Study 3) that although the DASS-21 consists

of a proportion of overlapping symptoms, it is substantially measurement invariant in

OSA. That is, the symptoms assessed in the DASS-21 are not attributable to OSA.

Therefore, the DASS-21(Lovibond & Lovibond, 1995) is also a recommended

questionnaire to examine depressive symptoms in OSA1.

The published meta-analysis in Chapter 2 (Study 1) conceptually categorized

items within commonly-used depression questionnaires into symptom categories (e.g.

anhedonia, somatic/behavioural consequences of depression, sleep disturbance, loss of

energy or fatigue, cognitive, cognition, negative affect, anhedonia, and OSA-related

depression), which can be used by any researcher or clinician to divide questionnaires

1 None of the studies reported in our meta-analysis (Study 1, Chapter 2) had used the DASS (either in its

42 or its 21-item forms) so we were unable to explore prevalence estimates with the DASS.


into overlapping and non-overlapping depressive symptoms. This point will be

considered further in the clinical implications section of this thesis.

The effect of CPAP on depressive symptoms in OSA

Chapter 3 (Study 2), revealed that both overlapping and non-overlapping HAM-

D depressive symptoms decreased over time, with a greater decrease in the severity of

overlapping depressive symptoms. Notably, only the decrease of overlapping

depressive symptoms was a function of CPAP use. In Chapter 5 (Study 4) where the

DASS-21 was used, latent change modelling showed that only the depression and

anxiety symptom scores declined significantly over time, however this change was not

a function of CPAP use. Stress scores decreased, but only at trend level. Based on the

clinical significance analysis, the majority of individuals’ post-treatment scores across

all subscales were in the functional range at post-treatment, even those who began in

the mild or more severe range at baseline.

Both treatment studies tell a consistent story. The non-overlapping HAM-D

symptoms, and the DASS-21 symptoms, which are reflective of symptoms that are not

OSA-related (DASS symptoms being measurement invariant), improve over time, but

not in a way that is related to CPAP use. In the absence of a control treatment condition,

we cannot be certain if the decline of these symptoms over time is a genuine effect of

treatment, or due to regression to the mean and/or placebo effects. In contrast, the

decline of overlapping HAM-D symptoms over time was a function of CPAP use.

Without a treatment control condition, the most parsimonious explanation for the

results is that CPAP treats overlapping depressive symptoms, which may suggest that

CPAP is just treating OSA, rather than depression, per se.

A SHAM CPAP was not used in the OSARS treatment trial due to clinical

equipoise. Untreated OSA may lead to hypertension, inflammation, hyperlipidemia,


obesity (Young, Skatrud, & Peppard, 2004) and cognitive impairment (Bucks, Olaithe,

& Eastwood, 2013). The longer it remains untreated the more likely that some of these

changes will be irreversible (Bédard, Montplaisir, Malo, Richer, & Rouleau, 1993).

Additionally, CPAP improves daytime sleepiness and reaction time and steering

performing in driving simulations, therefore, failure to provide CPAP to individuals

with OSA, who are at high risk of motor vehicle crashes, is likely to lead to an

unacceptable level of risk to the individual with OSA, and others (George, Boudreau, &

Smiley, 1997; Sassani et al., 2004). Thus, it would be unethical to withhold a

therapeutic CPAP, which would be the equivalent to using a SHAM CPAP for

individuals with OSA.

A possible alternative is a wait-list condition where individuals with OSA are on

a wait-list to commence the trial; these individuals would also receive CPAP. If the

non-overlapping depression scores improve for these individuals, then this would

suggest that the improvement of the non-overlapping symptoms and DASS-21

symptoms found in Studies 2 and 4 are due to regression to the mean. Unfortunately,

this alternative does not answer the question about the whether the improvement of the

non-overlapping symptoms and DASS-21 symptoms represent a placebo effect. In the

absence of a placebo control, this is a tricky problem to solve. Alternatively, another

way to examine the symptomatology of depression in OSA, and examine if CPAP treats

depression in OSA, is by using a symptom-based approach.

Symptom-based approach of depression in OSA

This thesis aimed to establish the nature of depression in OSA by examining

symptom severity scores derived by summing severity ratings across symptoms.

Another approach, which may enhance our understanding regarding the

phenomenology of depression in OSA, and which might overcome the difficulty of not


being able to use a non-CPAP treatment condition, is to focus on the individual

depression symptoms, rather than symptom severity. A symptom-based approach offers

an alternative to understanding the phenomenology and heterogeneity of depression in

OSA. Two promising symptom-based approaches are Latent Profile Analysis (LPA)

and Network Analysis. LPA assumes there are distinct patterns of observed scores

(indicators) and that individuals can be grouped into a small number of

subtypes/phenotypes based on their similar pattern of profile of scores (Gibson, 1959;

Lanza, Collins, Lemmon, & Schafer, 2007). Unlike traditional cluster analysis, LPA is

model-based, that is, cases are assigned to the subtype phenotype for which they have

the highest probability of membership (Gibson, 1959; Lanza et al., 2007). LPA

provides a useful technique for examining the heterogeneity of depression in OSA.

LPA has been used in understanding the phenomenology of depression in

Parkinson’s Disease (Starkstein et al., 2011). The authors found evidence for a 4-class

solution: a major depression-anxiety class (high probabilities for all anxiety and

depression symptoms, using the PHQ-9), a minor depression-anxiety class (low

probabilities of anxiety and depression symptoms), a sub-syndromal anxiety class

(higher probabilities of anxiety, compared to depressive symptoms), and a class with no

depression or anxiety. They also found significant demographic and clinical differences

(e.g. age, gender proportion, education) between the classes. Similarly, Starkstein et al.,

(2014) examined the syndromal pattern of depression in Type-2 diabetes, and also

found 4 classes: major anxious depression, minor anxious depression, subclinical

anxiety, and no anxious depression. They also found significant differences between the

symptom classes in terms of demographic and clinical differences. To the best of my

knowledge, LPA analyses of depression in OSA has not been published.


However, an unpublished analysis of data from a clinical sample of patients

with untreated OSA (N = 655) where LPA was used empirically to identify clusters of

depressive symptoms revealed three latent profiles. Using the Centre for Epidemiologic

Studies Depression Scale (CESD-10), 121 (18%) participants had high, partial

conditional probabilities for poor concentration, sad mood, loss of energy, social

withdrawal and loss of motivation, which we labelled ‘severe depression’. 226 (34%)

had low partial conditional probabilities for all symptoms of depression except for

hopelessness and insomnia and were labelled as ‘no depression’. The third group of 318

(48%) had scores intermediate between classes 1 and 2 (above) which were labelled as

‘moderate depression’ (See Figure 6.1).

Figure 6.1. Subgroups of individuals with OSA, based on LPA.

Based on the findings of this thesis and the studies that have examined the

heterogeneity of depression in other chronic conditions, it is likely that there are distinct

subgroups of individuals with OSA, who have different types of depressive symptoms.


Possible subgroups may include: 1) individuals who have a high probability of only

overlapping depressive symptoms (e.g. fatigue, loss of libido, sleepiness), which may

suggest no depression, just OSA, 2) individuals who endorse a very high proportion of

non-overlapping symptoms (e.g. suicidal ideation, anhedonia), suggesting that these

individuals have co-morbid depression, independent of OSA, and 3) individuals endorse

a proportion of of both non-overlapping (e.g. suicidal ideation) and overlapping

depressive symptoms (e.g. fatigue, physical anxiety symptoms (e.g. dry mouth, rapid

heartbeat)), suggesting that these individuals have co-morbid depression and OSA, or

possibly, subsyndromal depression or minor depression. These profiles are likely to

significantly differ in terms of demographic and clinical differences. Therefore, LPA

will help clarify the diagnostic heterogeneity of depression in OSA, and might show

which depressive symptoms differentiate those with high levels of depression in OSA,

from those depressive symptoms that are common to all individuals with OSA.

A further step related to LPA is Latent Transition Analysis (LTA), which is an

extension of LCA that uses longitudinal data to identify movement between the classes

over time (Lanza, Flaherty & Collins, 2003). This methodology would allow the

exploration of which depressive symptom cluster(s) ameliorate(s) with CPAP

treatment, and which do not. No studies to date have examined LTA in the field of

depression and chronic illness. This analysis would be very beneficial to our

understanding of the type of depressive symptom cluster that ameliorate with CPAP,

and which are responsive to OSA treatment.

Another methodology is called Network Analysis (NA; Fried, 2015), which

examines how related certain symptoms are to a clinical disorder (e.g. depression). The

network approach to psychopathology allows the study of interactions between

symptoms (Fried, 2015). Network models estimate the relationships between


symptoms, including how symptoms can trigger other symptoms (e.g. insomnia can

cause fatigue, which can lead to psychomotor problems, which can then lead to

anhedonia; Fried & Nesse, 2015). From a network perspective, the syndrome (e.g.

depression) does not exist apart from its symptoms, thus, the emphasis is on

understanding the individual symptoms of the syndrome (Fried & Nesse, 2015). NA has

been highlighted as a very useful technique for comorbidity research (Fried, 2015).

Robinaugh, LeBlanc, Vuletich, & McNally (2014) used NA to understand the

etiology of Persistent Complex Bereavement Disorder (PCBD). They found that

emotional pain is the core symptom of PCBD. Additionally, emotional pain was an

important predictor of other symptoms such as avoidance, whereas yearning did not,

indicating that emotional pain may be the affective experience that underpins grief-

related avoidance, commonly associated with PCBD. Further, they found that low

emotional stability and greater interpersonal dependence are symptoms that may

maintain the syndrome. They argued that addressing emotional pain is an important

target for the treatment PCBD. Therefore, network analysis can be used to inform the

treatment intervention.

In relation to OSA, a network analysis could be used to identify the causal

relationships of depressive symptoms in OSA, and answer important questions such as

whether there are multiple networks of depressive symptoms (e.g. non-overlapping vs.

overlapping symptoms) in OSA, and which depressive symptoms are related to each

other and which are not. It can be used to identify which symptom increases the

likelihood of experiencing one symptom, when experiencing the other. This will have

implications for understanding the phenomenology of depression in OSA and provide

information about which depressive symptoms may need to be treated first, to have a

beneficial effect on the other symptoms within the network (e.g. do overlapping


symptoms need to be treated first to reduce the severity of non-overlapping depressive

symptoms). A review of the network analysis approach is outlined by Fried (2015) and

Borsboom and Cramer (2013).

Clinical Depression in OSA

This thesis was based on the assessment of depressive symptoms (overlapping

vs. non-overlapping depressive symptoms) in OSA and the relationship with CPAP.

However, to further understand the phenomenology of depression in OSA, examining

clinical depression (based on a SCID diagnosis), would be clinically useful.

Inter-individual covariates

A number of key, inter-individual variables have been found to relate to the

expression and treatment of depressive symptoms in OSA. These variables are

addressed in many studies, with contrasting results. This thesis examined the

relationship between OSA, depression, and covariates such as study type, age, BMI,

antidepressant use, sex, and baseline AHI.

We found that opportunity samples were associated with higher depression

prevalence estimates than population-based studies (38.60% vs 9.70%). This may be

attributable to differences in depression severity, biases in subject selection, quality of

life, or comorbid disease in individuals with OSA in these two types of samples (Harris

et al., 2009).

In relation to age, the results suggest that individuals who are younger endorsed

a greater percentage of non-overlapping symptom scores, which contrasts to the meta-

analysis results (Chapter 2), where there was no relationship between age and

depressive symptom categories, albeit, the meta-analysis used group averages for ages,

which may have obscured the results. Few studies have examined the relationship

between age and depressive symptoms in OSA but, in those that have, results (e.g.


Asghari, Mohammadi, Kamrava, Tavakoli, & Farhadi, 2012) suggest no relationship

between age and depressive symptoms in OSA. A limitation to these studies is that they

did not compare the effects of age on overlapping and non-overlapping symptoms,

separately. Therefore, based on our results, younger individuals endorse a greater

proportion of non-overlapping symptoms, than older individuals, albeit, this effect may

also be due to younger adults experiencing less overlapping symptoms (e.g., loss of

libido, psychomotor retardation) compared to older adults (Meston, 1997; Alexopoulos,

2005), in general. Age also was not a significant predictor of CPAP treatment trajectory

for either overlapping or non-overlapping symptoms, which suggests that the same

effect occurs over time across all age groups (>18 years old).

AHI did not moderate the prevalence of depression in OSA, nor influence the

amelioration of depressive symptoms in OSA. A limitation of our meta-analysis is that

we used group averages to examine the effects of AHI, which may have obscured the

effects, albeit when assessing AHI as a continuous variable (Studies 2 and 4), OSA-

severity was not associated with depressive symptoms either. This is not surprising as

the majority of studies (e.g. Lee, Lee, Chung, & Kim, 2015; Macey, Woo, Kumar,

Cross & Harper, 2010; Asghari et al., 2012; El Sherbini, Bediwy, El-Mitwalli, 2011)

that have examined the relationship between AHI and depressive symptoms have not

found a relationship. Additionally, there was no relationship between AHI and the

trajectory of decline in either overlapping or non-overlapping depressive symptoms.

This is interesting because we would expect a relationship between depressive

symptoms and AHI, if depression is a function of OSA. This suggests that the presence

of depressive symptoms in OSA may not be due to changes in the number and

frequency of hypoxic events. Although AHI is a perfectly sensible measure of OSA

severity (i.e. for diagnosing OSA), it may not capture the severity of OSA in a way that


is specific to the proposed mechanisms by which OSA produces depression. Other

predictors that were not examined in this study, such as daytime sleepiness, frequency

of arousal, or degree of oxygen desaturation may be better measures for future studies

to examine this relationship (Harris et al., 2009; Olaithe et al., 2015).

Gender was a significant predictor of prevalence estimates in the meta-analysis

in Chapter 2. Contrary to the general population (Piccinelli & Wilkinson, 2000), males

have a higher prevalence estimates of depression symptoms, than females. In addition,

males generally have more severe OSA than females, thus they may be more likely to

endorse symptoms on depression questionnaires due to their OSA, than females, which

may account for this result (Wahner-Roedler et al., 2007). However, if this was the

case, we would expect that perhaps gender moderated the change in overlapping

depressive symptoms over time, as these symptoms are OSA-related. Interestingly,

gender did not moderate the decline in either overlapping or non-overlapping

depressive symptoms over time. Therefore, our findings suggest that the same change

occurred in depressive symptoms over time, in both females and males.

Higher BMI was associated with more severe total HAM-D scores and

overlapping depressive symptoms, at baseline, which may be due to obesity being

strongly associated with the development of OSA, leading to the presence of

overlapping depressive symptoms (e.g. fatigue, sleepiness). This is consistent with why

those with a higher BMI improved in terms of the severity of their total HAM-D and

overlapping depressive symptoms faster, than those with a lower BMI.

In relation to antidepressant medication use, similar to Schwartz, Kohler, and

Karatinos (2005), we found that individuals who took antidepressant medication prior

to the commencement of CPAP had higher depression severity scores (using the HAM-

D) at baseline. Despite being on antidepressant medication, however, these individuals


did not show a steeper decrease in depressive symptoms over time. One possible

explanation for this is that some individuals may have been misdiagnosed as

experiencing depression, instead of OSA, and treated accordingly, but without a clinical

need. Antidepressant use was a categorical variable (antidepressant use vs. no

antidepressant use), which may have obfuscated the results as the specific dosage an

individual takes may influence the results. Future studies may wish to examine the

dosage of antidepressant medication as a continuous variable to examine the influence

of antidepressant medication on the treatment of depressive symptoms in OSA.

Implications for clinicians

Due to the high comorbidity between depression and OSA, as well as the

negative health, cognitive, and social implications of both conditions, a greater

awareness about both these conditions to assist with the detection and treatment of both

these conditions, is warranted.

Identifying Depressive Symptoms and OSA in clinical settings. The

evaluation of depression in OSA can be challenging. An individual may present to a

General Practitioner complaining of physical symptoms of OSA, such as snoring,

frequent nocturnal awakenings, feeling unrefreshed upon awakening, or feeling tired

during the day (Netzer, Stoohs, Netzer, Clark, & Strohl, 1999). If such symptoms are

reported, then measures can be used to assess if these symptoms are indicative of OSA.

For example, the Berlin Questionnaire (BQ) is a 10 minute, self-administered

questionnaire that identifies the level of risk an individual has of sleep disordered

breathing (Netzer et al., 1999). If there is suspected OSA, they should be referred to a

Sleep Disorders Clinic for evaluation by nocturnal PSG to confirm the diagnosis of

OSA.


As discussed throughout this thesis, individuals with possible OSA may also

experience depressive symptoms, independent of their undiagnosed OSA. As structured

clinical interviews are not often used in clinical settings, a convenient way of assessing

an individual’s depressive symptoms is to administer a depression questionnaire

(Coyne, Thompson, Klinkman, & Nease, 2002). As demonstrated in Chapter 2 (Study

1), not all depression questionnaires are equally appropriate for assessing depressive

symptomatology in OSA. The thesis findings suggest the development of a specific

questionnaire to assess depression in OSA is warranted. At the most basic level, such a

questionnaire could assess depressive symptoms that do not overlap with OSA,

however this might ignore important features of depression as a result. A new

questionnaire will need to be validated against gold standard diagnostic interview

measures, such as the SCID (First, Spitzer, Gibbon, & Williams, 2012). Until this has

been done, and based on the findings from this thesis, we recommend the use of GDS,

HADS-D, or DASS-21. Alternatively, Chapter 2 (Study 1) outlines the different

categories of depressive symptoms, which can be applied to any depression

questionnaire to categorise symptoms (e.g. anhedonia, cognitive, cognition items).

These symptom categories could be used by health-care professionals to guide their

assessment and the care that they provide to the individual (discussed below).

The data from the studies within this thesis suggest that individuals who report

high levels of anhedonia, negative affect, and cognition (i.e. non-overlapping

depressive symptoms), are individuals who probably have OSA plus co-morbid

depression, as a separate disease entity. If an individual endorses a high proportion of

non-overlapping symptoms (e.g. anhedonia, suicidal ideation, cognitive items), then a

referral to a Clinical Psychologist/Psychiatrist, or anti-depressant medication, may be

warranted. If suicidal ideation is present, a comprehensive risk assessment by the


healthcare professional is necessary. Furthermore, for individuals who are diagnosed

with OSA, an assessment of their depressive symptoms every few months is warranted

as OSA is a possible risk factor for developing depression (Peppard, Szklo-Coxe, Hla,

& Young, 2006).

As symptom overlap potentially confounds the detection of OSA and

depression, it is likely that Clinical Psychologists and Clinical Neuropsychologists will

be referred individuals with undiagnosed OSA (Olaithe et al., 2015). If an individual

visits a Psychologist complaining of largely physiological symptoms of depression (e.g.

fatigue, sexual dysfunction, irritability etc; i.e. overlapping depressive symptoms),

consideration should be given to the idea that this individual may have untreated OSA,

and this should be investigated. These individuals should be referred to a GP, who can

administer the BQ and subsequently refer them to a Sleep Disorders Clinic for

evaluation for OSA. Failure to recognise and treat OSA in a timely manner may lead to

the persistence of depressive symptoms and increase the risk of development of other

medical issues (e.g. Type-2 diabetes, cardiovascular disease; Young et al., 2004). A

collaborative approach between health professionals is important for managing

individuals with OSA and psychological disturbances.

Treatment of depressive symptoms in OSA in clinical settings. Based on

the findings of this thesis, the type of depressive symptoms an individual has prior to

the commencement of treatment can inform the intervention approach adopted. If an

individual presents with a high proportion of overlapping depressive symptoms (e.g.

fatigue, loss of libido, psychomotor retardation, cognitive difficulty), a course of CPAP

therapy and a review at three months to examine if their depressive symptoms have

improved, is recommended. The data from Studies 2 and 4 suggest that their depressive


symptoms would be reduced and, if this is the case, then CPAP therapy should be

continued.

If an individual reports a high proportion of non-overlapping depressive

symptoms, prior to CPAP, a course of CPAP is still recommended, as the findings from

Study 2 and Study 4 suggest that non-overlapping depressive symptoms would still

reduce over time, however to a lesser extent than overlapping depressive symptoms. If

an individual’s non-overlapping depressive symptoms have not substantially decreased,

or their symptoms have not decreased to the extent that an individual has noticed an

improvement to their overall daily functioning, then these individuals may need to: 1)

commence anti-depressant medication; anti-depressant medication has been found to

successfully ameliorate depressive symptoms in individuals with mild to severe

depression (Fournier et al., 2013; Cipriani et al., 2009), or 2) engage in psychotherapy

intervention, such as Cognitive Behaviour Therapy (CBT), contemporaneously, with

CPAP treatment.

CBT has been found to significantly reduce depressive symptoms in individuals

with chronic conditions (e.g. Parkinson’s Disease, Dobkin et al., 2011; Diabetes,

Rustad, Musselman & Nemeroff, 2011; Insomnia, Manber et al., 2011). Notably, CBT

addresses negative cognitions, anhedonia, and suicidal ideation (Jacobson, Dobson &

Truax, 1996), all of which are non-overlapping depressive symptoms in OSA.

Furthermore, if individuals present with depressive symptoms such as suicidal

ideation, anhedonia, or cognition items (e.g. low self-esteem, worthlessness), they may

need to be started on antidepressant medication or engage in CBT, at the same time

they are provided with a CPAP machine, as these are “red flags” to a clinician that their

depressive symptoms may be indicative of depression symptoms that are not

attributable to OSA.


The role of Psychologists with CPAP use. Depressive symptoms have been

found to predict CPAP use, whereby more severe depression is a risk factor for poorer

CPAP use (Engleman, 1999; Law, Naughton, Ho, Roebuck, & Dabscheck, 2014).

CPAP improves quality of life, reduces car accidents and, to some extent, assists with

deficits in cognition (Sassani et al., 2004). It is plausible that poorer CPAP compliance

in depressed OSA individuals may not only reflect their affective symptoms, but also

their executive dysfunction (as cognitive deficits are evident in OSA individuals;

Olaithe et al., 2015). Importantly, between 48 - 83% of individuals with OSA do not

use their CPAP for a minimum of 4 hours a night (Weaver & Grunstein, 2008).

Interestingly, the average CPAP use for both the treatment studies within this thesis was

5.35 hours, which compares favorably to other CPAP treatment studies where averages

have ranged from 3.00 - 4.40 (Weaver & Sawyer, 2010; McEvoy et al., 2011). It is

estimated that over 50% of individuals who are on CPAP do not use their CPAP 1 year

later, and of those using it 1 year later, most are not using CPAP for the whole night, as

prescribed (Stepnowsky, Marler & Ancoli-Israel, 2002).

Research has found that psychological correlates such as self-efficacy, illness

beliefs, social support, locus of control, problem-solving skills, and engagement coping

strategies, are related to CPAP use (Edinger, Carwile, Miller, Hope, & Mayti, 1994;

Wild, Engleman, Douglas, & Espie, 2004; Olsen, Smith, Oei, & Douglas, 2008).

Additionally, depression may impede an individuals’ use of CPAP; depression is

associated with reduced adherence to medical treatment regimens in many different

patient populations (DiMatteo, Giordani, Lepper, & Croghan, 2002).

Psychological-based interventions have been found to assist with CPAP use

(Olsen, Smith, Oei, & Douglas, 2012). Motivational interviewing has been found to

increase CPAP uptake (Olsen et al., 2012; Aloia, Arnedt, Riggs, Hecht & Borrelli,


2004). Richards, Bartlett, Wong, Malouff, and Grunstein (2007) found that a CBT

intervention resulted in increased CPAP compliance. The CBT approach uses

components related to social cognitive theory (increased perceived self-efficacy,

outcome expectations and social support), and aims to change distorted beliefs about

using a CPAP machine (Richards et al., 2007). Additionally, Psychologists can explore

illness-related beliefs during the course of therapy and use self-administered

questionnaires (e.g. the Illness Perception Questionnaire-Revised, Moss-Morris et al.,

2002) as an outcome measure to monitor improvement. Furthermore, individuals with

little social support use less CPAP (Stepnowsky, Marler, Palau, & Brooks, 2006).

Therefore, providing social support or helping clients to develop peer relationships may

improve CPAP use, as well as improve their depressive symptoms. Therefore,

Psychologists play a pivotal role in the overall management of depression and OSA in

individuals.

Psychology training. Due to the comorbidity of depression and OSA, to

accomplish greater awareness, detection, and support for individuals, clinicians need to

be informed and taught about the relationship between OSA and depression. However,

the processes underlying the relationship between sleep disorders and mental health are

not routinely taught in clinical or neuropsychology courses (Waters & Bucks, 2011).

The Australian Psychological Society (APS) does not outline that sleep disorders are a

necessary area for teaching in Psychology courses. It is important that universities teach

Clinical Psychology and Clinical Neuropsychology postgraduate trainees about how to

work with individuals who present with sleep disorders in clinical settings.

Additionally, sleep physicians would also benefit from learning about the association

between depression and sleep disorders, such as OSA, to help guide the detection and

intervention approach to both conditions.


Conclusion

This thesis examined depressive symptomatology in OSA, demonstrating that

symptom overlap potentially impacts upon the presentation and treatment of depression

in OSA. When identifying depressive symptoms in OSA, the type of depression

questionnaire used should be taken into consideration; we recommend the use of

measures that are measurement invariant, such as the DASS-21, or questionnaires that

are less inflated by overlapping depressive symptoms, such as HADS-D or GDS. Both

overlapping and non-overlapping depressive symptoms decline over the treatment

interval, however only the decline of overlapping symptoms is a function of CPAP use.

As we were not able to use a control treatment condition, the change in non-overlapping

symptoms over time may be due to placebo effects or regression to the mean. A

parsimonious explanation for the results is: 1) CPAP only reduces depressive symptoms

that are OSA related, and 2) CPAP may only be treating OSA, and not depression.

Future research is required to understand the phenomenology of depressive symptoms

in OSA and whether CPAP treats depression in OSA, by using a symptoms-based

approach. The goal of the studies within this thesis is to encourage researchers and

clinicians to start thinking about the importance of individual depressive symptoms in

OSA better to understand the phenomenology of depression in OSA.


References

AASM. (2012). The AASM manual for the scoring of sleep and associated events;

rules, terminology and technical specifications. Illinois: American Academy of

Sleep Medicine.

Akaike, H. (1974). A new look at the statistical model identification. IEEE

Transactions on Automatic Control, 19(6), 716–723.

https://doi.org/10.1109/TAC.1974.1100705.

Akashiba, T., Kawahara, S., Akahoshi, T., Omori, C., Saito, O., Majima, T., & Horie,

T. (2002). Relationship between quality of life and mood or depression in

patients with severe obstructive sleep apnea syndrome. Chest Journal, 122(3),

861-865.

Al Lawati, N. M., Patel, S. R., & Ayas, N. T. (2009). Epidemiology, Risk Factors, and

Consequences of Obstructive Sleep Apnea and Short Sleep Duration. Progress

in Cardiovascular Diseases, 51(4), 285–

293.https://doi.org/10.1016/j.pcad.2008.08.001.

Alchanatis, M., Zias, N., Deligiorgis, N., Liappas, I., Chroneou, A., Soldatos, C., &

Roussos, C. (2008). Comparison of cognitive performance among different age

groups in patients with obstructive sleep apnea. Sleep & Breathing = Schlaf &

Atmung, 12(1), 17–24. https://doi.org/10.1007/s11325-007-0133-y.

Alexopoulos, G. S. (2005). Depression in the elderly. The lancet, 365(9475), 1961-

1970.

Aloia, M. S., Arnedt, J. T., Davis, J. D., Riggs, R. L., & Byrd, D. (2004).

Neuropsychological sequelae of obstructive sleep apnea-hypopnea syndrome: A

critical review. Journal of the International Neuropsychological Society, 10(5),

772–785. http://dx.doi.org/10.1017/S1355617704105134.

https://doi.org/10.1109/TAC.1974.1100705

https://doi.org/10.1016/j.pcad.2008.08.001

https://doi.org/10.1007/s11325-007-0133-y

http://dx.doi.org/10.1017/S1355617704105134


Aloia, M. S., Arnedt, J. T., Riggs, R. L., Hecht, J., & Borrelli, B. (2004). Clinical

management of poor adherence to CPAP: motivational enhancement.

Behavioral Sleep Medicine, 2(4), 205–222.

https://doi.org/10.1207/s15402010bsm0204_3.

Aloia, M. S., Arnedt, J. T., Smith, L., Skrekas, J., Stanchina, M., & Millman, R. P.

(2005). Examining the construct of depression in obstructive sleep apnea

syndrome. Sleep Medicine, 6(2), 115–121.

https://doi.org/10.1016/j.sleep.2004.09.003.

Altevogt, B. M., & Colten, H. R. (Eds.). (2006). Sleep disorders and sleep deprivation:

an unmet public health problem. National Academies Press.

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental

disorders (5th ed.). Arlington, VA: American Psychiatric Publishing

Anderson, R. J., Freedland, K. E., Clouse, R. E., Lustman, P. J., McGlynn, M.,

Shubrook, J., … Oordt, M. (2001). The prevalence of comorbid depression in

adults with diabetes: a meta-analysis. Diabetes Care, 24(6), 1069–78.

https://doi.org/10.2337/diaspect.23.1.15.

Andresen, E. M., Malmgren, J. a, Carter, W. B., & Patrick, D. L. (1994). Screening for

depression in well older adults: evaluation of a short form of the CES-D (Center

for Epidemiologic Studies Depression Scale). American Journal of Preventive

Medicine, 10(2), 77–84. https://doi.org/10.4236/health.2013.53A078.

Antony, M. M., Bieling, P. J., Cox, B. J., Enns, M. W., & Swinson, R. P. (1998).

Psychometric properties of the 42-item and 21-item versions of the Depression

Anxiety Stress Scales in clinical groups and a community sample. Psychological

Assessment, 10(2), 176–181. https://doi.org/10.1037/1040-3590.10.2.176

https://doi.org/10.1207/s15402010bsm0204_3

https://doi.org/10.1016/j.sleep.2004.09.003

https://doi.org/10.2337/diaspect.23.1.15

https://doi.org/10.4236/health.2013.53A078

https://doi.org/10.1037/1040-3590.10.2.176


Asarnow, J. R., Jaycox, L. H., Duan, N., LaBorde, A. P., Rea, M. M., Murray, P., …

Wells, K. B. (2005). Effectiveness of a quality improvement intervention for

adolescent depression in primary care clinics: a randomized controlled trial.

JAMA: The Journal of the American Medical Association, 293(3), 311–319.

https://doi.org/10.1001/jama.293.3.311.

Asghari, A., Mohammadi, F., Kamrava, S. K., Tavakoli, S., & Farhadi, M. (2012).

Severity of depression and anxiety in obstructive sleep apnea syndrome. Eur

Arch Otorhinolaryngol. https://doi.org/10.1007/s00405-012-1942-6.

Asghari, A., Saed, F., & Dibajnia, P. (2008). Psychometric properties of the

Depression Anxiety Stress Scales-21 (DASS-21) in a non-clinical Iranian

sample. International Journal of Psychology, 2(2), 82–102.

Bachour, A., & Maasilta, P. (2004). Mouth breathing compromises adherence to nasal

continuous positive airway pressure therapy. Chest, 126(4), 1248–1254.

https://doi.org/10.1378/chest.126.4.1248.

Ballard, R. D., Gay, P. C., & Strollo, P. J. (2007). Interventions to improve compliance

in sleep apnea patients previously non-compliant with continuous positive

airway pressure. Journal of Clinical Sleep Medicine, 3(7), 706–712.

Ballester, E., Solans, M., Vila, X., Hernandez, L., Quint??, L., Bolivar, I., …

Montserrat, J. M. (2000). Evaluation of a portable respiratory recording device

for detecting apnoeas and hypopnoeas in subjects from a general population.

European Respiratory Journal, 16(1), 123–127. https://doi.org/10.1034/j.1399-

3003.2000.16a22.x.

Bardwell, W. (2001). Types of coping strategies are associated with increased

depressive symptoms in patients with obstructive sleep apnea. Sleep: Journal of

Sleep, 24(8), 905–9.

https://doi.org/10.1001/jama.293.3.311

https://doi.org/10.1007/s00405-012-1942-6

https://doi.org/10.1378/chest.126.4.1248

https://doi.org/10.1034/j.1399-3003.2000.16a22.x

https://doi.org/10.1034/j.1399-3003.2000.16a22.x


Bardwell, W. A., Ancoli-Israel, S., & Dimsdale, J. E. (2007). Comparison of the effects

of depressive symptoms and apnea severity on fatigue in patients with

obstructive sleep apnea: A replication study. Journal of Affective Disorders,

97(1–3), 181–186. https://doi.org/10.1016/j.jad.2006.06.013.

Bardwell, W. A., Moore, P., Ancoli-Israel, S., & Dimsdale, J. E. (2000). Does

obstructive sleep apnea confound sleep architecture findings in subjects with

depressive symptoms? Biological psychiatry, 48(10), 1001-1009.

Beebe, D. W. (2005). Neurobehavioral effects of obstructive sleep apnea: an overview

and heuristic model. Current opinion in pulmonary medicine, 11(6), 494-500.

Beebe, D. W., & Gozal, D. (2002). Obstructive sleep apnea and the prefrontal cortex:

towards a comprehensive model linking nocturnal upper airway obstruction to

daytime cognitive and behavioral deficits. Journal of sleep research, 11(1), 1-

16.

Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An

inventory for measuring depression. Archives of General Psychiatry, 4(6), 561–

571. https://doi.org/10.1001/archpsyc.1961.01710120031004.

Bédard, M. A., Montplaisir, J., Malo, J., Richer, F., & Rouleau, I. (1993). Persistent

neuropsychological deficits and vigilance impairment in sleep apnea syndrome

after treatment with continuous positive airways pressure (CPAP). Journal of

Clinical and Experimental Neuropsychology, 15(2), 330-341.

Benca, R. M., Obermeyer, W. H., Thisted, R. A., & Gillin, J. C. (1992). Sleep and

psychiatric disorders: a meta-analysis. Archives of general psychiatry, 49(8),

651-668.

Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological

Bulletin, 107(2), 238–246. https://doi.org/10.1037/0033-2909.107.2.238.

https://doi.org/10.1016/j.jad.2006.06.013

https://doi.org/10.1001/archpsyc.1961.01710120031004

https://doi.org/10.1037/0033-2909.107.2.238


Benton, T., Staab, J., & Evans, D. L. (2007). Medical co-morbidity in depressive

disorders. Annals of Clinical Psychiatry: Official Journal of the American

Academy of Clinical Psychiatrists, 19(4), 289–303.

https://doi.org/10.1080/10401230701653542.

Bland, J. M., & Altman, D. G. (1994). Statistics notes: some examples of regression

towards the mean. Bmj, 309(6957), 780.

Borak, J., Cieslicki, J. K., Koziej, M., Matuszewski, A., & Zielinski, J. (1996). Effects

of CPAP treatment on psychological status in patients with severe obstructive

sleep apnoea. Journal of Sleep Research, 5(2), 123–127.

Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2011).

Introduction to meta-analysis. New York, NY: Wiley.

Borsboom, D., & Cramer, A. O. J. (2013). Network Analysis: An Integrative Approach

to the Structure of Psychopathology. Annual Review of Clinical Psychology,

9(1), 91–121. https://doi.org/doi:10.1146/annurev-clinpsy-050212-18560.

Bucks, R. S., Olaithe, M., & Eastwood, P. (2013). Neurocognitive function in

obstructive sleep apnoea: A meta-review. Respirology.

https://doi.org/10.1111/j.1440-1843.2012.02255.x.

Byrne, B. M., & Watkins, D. (2003). The Issue Of Measurement Invariance Revisited.

Journal of Cross-Cultural Psychology, 34(2), 155–175.

https://doi.org/10.1177/0022022102250225.

Carney, C. E., Ulmer, C., Edinger, J. D., Krystal, A. D., & Knauss, F. (2009).

Assessing depression symptoms in those with insomnia: An examination of the

Beck Depression Inventory 2nd ed. (BDI-II). Journal of Psychiatric Research,

43(5), 576–582. https://doi.org/10.1016/j.jpsychires.2008.09.002.

https://doi.org/10.1080/10401230701653542

https://doi.org/doi:10.1146/annurev-clinpsy-050212-18560

https://doi.org/10.1111/j.1440-1843.2012.02255.x

https://doi.org/10.1177/0022022102250225

https://doi.org/10.1016/j.jpsychires.2008.09.002


Carskadon, M.A., & Dement, W.C. (2011). Monitoring and staging human sleep. In

M.H. Kryger, T. Roth & W.C. Dement (Eds.), Principles and Practice of Sleep

Medicine (5th edition). St. Louis: Elsevier Saunders.

Castro, L. S., Castro, J., Hoexter, M. Q., Quarantini, L. C., Kauati, A., Mello, L. E., …

Bittencourt, L. (2013). Depressive symptoms and sleep: A population-based

polysomnographic study. Psychiatry Research, 210(3), 906–912.

https://doi.org/10.1016/j.psychres.2013.08.036.

Chang, L., Connelly, B. S., & Geeza, A. A. (2012). Separating method factors and

higher order traits of the Big Five: A meta-analytic multitrait–multimethod

approach. Journal of Personality and Social Psychology, 102(2), 408–426.

https://doi.org/10.1037/a0025559

Chen, Y.-H., Keller, J. K., Kang, J.-H., Hsieh, H.-J., & Lin, H.-C. (2013). Obstructive

sleep apnea and the subsequent risk of depressive disorder: a population-based

follow-up study. Journal of Clinical Sleep Medicine: JCSM:: Official

Publication of the American Academy of Sleep Medicine, 9(5), 417–23.

https://doi.org/10.5664/jcsm.2652.

Chervin, R. D. (2000). Sleepiness, fatigue, tiredness, and lack of energy in obstructive

sleep apnea. Chest, 118(2), 372–379. https://doi.org/10.1378/chest.118.2.372.

Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for

testing measurement invariance. Structural Equation Modeling, 9, 233–255.

http://dx.doi.org/10.1207/S15328007SEM0902_5.

Cipriani, A., Furukawa, T. A., Salanti, G., Geddes, J. R., Higgins, J. P., Churchill, R., ...

& Tansella, M. (2009). Comparative efficacy and acceptability of 12 new-

generation antidepressants: a multiple-treatments meta-analysis. The

Lancet, 373(9665), 746-758.

https://doi.org/10.1016/j.psychres.2013.08.036

https://doi.org/10.1037/a0025559

https://doi.org/10.5664/jcsm.2652


http://dx.doi.org/10.1207/S15328007SEM0902_5


Cohen-Zion, M., Stepnowsky, C., Marler, Shochat, T., Kripke, D. F., & Ancoli-Israel,

S. (2001). Changes in cognitive function associated with sleep disordered

breathing in older people. Journal of the American Geriatrics Society, 49(12),

1622–1627. https://doi.org/10.1111/j.1532-5415.2001.49270.x.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.).

Hillsdale, NJ: Erlbaum.

Coman, E. N., Picho, K., McArdle, J. J., Villagra, V., Dierker, L., & Iordache, E.

(2013). The paired t-test as a simple latent change score model. Frontiers in

psychology, 4, 738.

Coyne, J. C., Thompson, R., Klinkman, M. S., & Nease, D. E. J. (2002). Emotional

disorders in primary care. Journal of Consulting and Clinical Psychology, 70(3),

798–809. https://doi.org/10.1037/0022-006X.70.3.798

Crawford, J., Cayley, C., Lovibond, P. F., Wilson, P. H., & Hartley, C. (2011).

Percentile Norms and Accompanying Interval Estimates from an Australian

General Adult Population Sample for Self-Report Mood Scales (BAI, BDI,

CRSD, CES-D, DASS, DASS-21, STAI-X, STAI-Y, SRDS, and SRAS).

Australian Psychologist, 46(1), 3–14. https://doi.org/10.1111/j.1742-

9544.2010.00003.x.

Crawford, M. R., Bartlett, D. J., Coughlin, S. R., Phillips, C. L., Neill, A. M., Espie, C.

a, … Marshall, N. S. (2012). The effect of continuous positive airway pressure

usage on sleepiness in obstructive sleep apnoea: real effects or expectation of

benefit? Thorax, 67(10), 920–4. https://doi.org/10.1136/thoraxjnl-2012-201622.

Crowley, K. E., Rajaratnam, S. M., Shea, S. A., Epstein, L. J., Czeisler, C. A., &

Lockley, S. W., & the Harvard Work Hours, Health and Safety Group. (2013).

https://doi.org/10.1111/j.1532-5415.2001.49270.x

https://doi.org/10.1037/0022-006X.70.3.798

https://doi.org/10.1111/j.1742-9544.2010.00003.x

https://doi.org/10.1111/j.1742-9544.2010.00003.x

https://doi.org/10.1136/thoraxjnl-2012-201622


Evaluation of a single-channel nasal pressure device to assess obstructive sleep

apnea risk in laboratory and home environments.

Curran, P. J., Obeidat, K., & Losardo, D. (2010). Twelve frequently asked questions

about growth curve modeling. Journal of Cognitive Development, 11(2), 121–

136. https://doi.org/10.1080/15248371003699969.

Cusin, C., Yang, H., Yeung, A., & Fava, M. (2009). Rating scales for depression.

In Handbook of clinical rating scales and assessment in psychiatry and mental

health (pp. 7-35). Humana Press.

Detke, M. J., Lu, Y., Goldstein, D. J., Hayes, J. R., & Demitrack, M. A. (2002).

Duloxetine, 60 mg once daily, for major depressive disorder: a randomized

double-blind placebo-controlled trial. The Journal of clinical psychiatry, 63(4),

308-315.

DiMatteo, M. R., Giordani, P. J., Lepper, H. S., & Croghan, T. W. (2002). Patient

adherence and medical treatment outcomes: a meta-analysis. Medical Care,

40(9), 794–811. https://doi.org/10.1097/01.MLR.0000024612.61915.2D.

Dobkin, R. D., Menza, M., Allen, L. A., Gara, M. A., Mark, M. H., Tiu, J., ... &

Friedman, J. (2011). Cognitive-behavioral therapy for depression in Parkinson's

disease: a randomized, controlled trial. American Journal of

Psychiatry, 168(10), 1066-1074.

Douglas, N. J., & Engleman, H. M. (2000). Effects of CPAP on vigilance and related

functions in patients with the sleep apnea/hypopnea syndrome. Sleep, 23, S147-

9.

Eckert, D. J., & Malhotra, A. (2008). Pathophysiology of Adult Obstructive Sleep

Apnea. Proceedings of the American Thoracic Society, 5(2), 144–153.

http://doi.org/10.1513/pats.200707-114MG.

https://doi.org/10.1080/15248371003699969

https://doi.org/10.1097/01.MLR.0000024612.61915.2D

http://doi.org/10.1513/pats.200707-114MG


Eckert, D. J., Jordan, A. S., Merchia, P., & Malhotra, A. (2007). Central sleep apnea:

pathophysiology and treatment. Chest Journal, 131(2), 595-607.

Edinger, J. D., Carwile, S., Miller, P., Hope, V., & Mayti, C. (1994). Psychological

status, syndromatic measures, and compliance with nasal CPAP therapy for

sleep apnea. Perceptual and motor skills, 78(3 suppl), 1116-1118.

Edwards, C., Mukherjee, S., Simpson, L., Palmer, L. J., Almeida, O. P., & Hillman, D.

R. (2015). Depressive symptoms before and after treatment of obstructive sleep

apnea in men and women. Journal of clinical sleep medicine: JCSM: official

publication of the American Academy of Sleep Medicine, 11(9), 1029.

El-Sherbini, A. M., Bediwy, A. S., & El-Mitwalli, A. (2011). Association between

obstructive sleep apnea (OSA) and depression and the effect of continuous

positive airway pressure (CPAP) treatment. Neuropsychiatr Dis Treat, 7, 715-

21.

Engleman, H. M., Kingshott, R. N., Wraith, P. K., Mackay, T. W., Deary, I. J., &

Douglas, N. J. (1999). Randomized placebo-controlled crossover trial of

continuous positive airway pressure for mild sleep apnea/hypopnea

syndrome. American journal of respiratory and critical care medicine, 159(2),

461-467.

Engleman, H., Martin, S., Douglas, N., & Deary, I. (1994). Effect of continuous

positive airway pressure treatment on daytime function in sleep

apnoea/hypopnoea syndrome. The Lancet, 343(8897), 572–575.

https://doi.org/10.1016/S0140-6736(94)91522-9.

Epstein, L. J., Kristo, D., Strollo, P. J. J., Friedman, N., Malhotra, A., Patil, S. P., …

Weinstein, M. D. (2009). Clinical guideline for the evaluation, management and

long-term care of obstructive sleep apnea in adults. Journal of Clinical Sleep

https://doi.org/10.1016/S0140-6736(94)91522-9


Medicine: JCSM: Official Publication of the American Academy of Sleep

Medicine, 5(3), 263–276.

Erman, M. K., Stewart, D., Einhorn, D., Gordon, N., & Casal, E. (2007). Validation of

the ApneaLink for the screening of sleep apnea: A novel and simple single-

channel recording device. Journal of Clinical Sleep Medicine, 3, 387–392.

First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (2012). Structured

Clinical Interview for DSM–IV Axis I Disorders (SCID-I), clinician version,

administration booklet. Arlington, VA: American Psychiatric Publishing.

Fournier, J. C., DeRubeis, R. J., Hollon, S. D., Gallop, R., Shelton, R. C., &

Amsterdam, J. D. (2013). Differential change in specific depressive symptoms

during antidepressant medication or cognitive therapy. Behaviour research and

therapy, 51(7), 392-398.

Fried, E. I. (2015). Problematic assumptions have slowed down depression research:

Why symptoms, not syndromes are the way forward. Frontiers in Psychology.

https://doi.org/10.3389/fpsyg.2015.00309.

Fried, E. I., & Nesse, R. M. (2015). Depression sum-scores don’t add up: why

analyzing specific depression symptoms is essential. BMC Medicine, 13(1), 72.

https://doi.org/10.1186/s12916-015-0325-4.

Fried, E. I., van Borkulo, C. D., Epskamp, S., Schoevers, R. A., Tuerlinckx, F., &

Borsboom, D. (2016). Measuring depression over time . . . Or not? Lack of

unidimensionality and longitudinal measurement invariance in four common

rating scales of depression. Psychological Assessment.

https://doi.org/10.1037/pas0000275.

Frommberger, U. H., Bauer, J., Haselbauer, P., Fräulin, a, Riemann, D., & Berger, M.

(1997). Interleukin-6-(IL-6) plasma levels in depression and schizophrenia:

https://doi.org/10.3389/fpsyg.2015.00309

https://doi.org/10.1186/s12916-015-0325-4

https://doi.org/10.1037/pas0000275


comparison between the acute state and after remission. European Archives of

Psychiatry and Clinical Neuroscience, 247(4), 228–233.

https://doi.org/10.1007/BF02900219.

Gabbay, I. E., & Lavie, P. (2012). Age- and gender-related characteristics of obstructive

sleep apnea. Sleep and Breathing, 16(2), 453–460.

https://doi.org/10.1007/s11325-011-0523-z.

Gami, A. S., Hodge, D. O., Herges, R. M., Olson, E. J., Nykodym, J., Kara, T., &

Somers, V. K. (2007). Obstructive Sleep Apnea, Obesity, and the Risk of

Incident Atrial Fibrillation. Journal of the American College of Cardiology,

49(5), 565–571. https://doi.org/10.1016/j.jacc.2006.08.060.

Gantner, D., Ge, J. Y., Li, L. H., Antic, N., Windler, S., Wong, K. . . .McEvoy, D.

(2010). Diagnostic accuracy of a questionnaire and simple home monitoring

device in detecting obstructive sleep apnoea in a Chinese population at high

cardiovascular risk. Respirology, 15, 952–960. http://dx.doi.org/10.1111/j.1440-

1843.2010.01797.x.

Geiser, C. (2012). Data analysis with Mplus. Guilford Press.

George, C. F., Boudreau, A. C., & Smiley, A. (1997). Effects of nasal CPAP on

simulated driving performance in patients with obstructive sleep

apnoea. Thorax, 52(7), 648-653.

Gibson, W. A. (1959). Three multivariate models: Factor analysis, latent structure

analysis, and latent profile analysis. Psychometrika, 24(3), 229-252.

Giles, T. L., Lasserson, T. J., Smith, B. J., White, J., Wright, J., & Cates, C. J. (2006).

Continuous positive airways pressure for obstructive sleep apnoea in adults

(Review). Cochrane Database of Systematic Reviews.

https://doi.org/10.1002/14651858.CD001106.pub3.

https://doi.org/10.1007/BF02900219

https://doi.org/10.1007/s11325-011-0523-z

https://doi.org/10.1016/j.jacc.2006.08.060

http://dx.doi.org/10.1111/j.1440-1843.2010.01797.x

http://dx.doi.org/10.1111/j.1440-1843.2010.01797.x

https://doi.org/10.1002/14651858.CD001106.pub3


Goldstein, D. J., Mallinckrodt, C., Lu, Y., & Demitrack, M. a. (2002). Duloxetine in the

treatment of major depressive disorder: a double-blind clinical trial. The Journal

of Clinical Psychiatry, 63(3), 225–31.

Gupta, R. K., Chandra, A., Verm, A. K., & Kumar, S. (2010). Obstructive sleep apnoea:

A clinical review. The Journal of the Association of Physicians of India, 58,

438–441.

Guyatt, G. H., Osoba, D., Wu, A. W., Wyrwich, K. W., & Norman, G. R. (2002).

Methods to explain the clinical significance of health status measures. Mayo

Clinic Proceedings, 77(4), 371–83. https://doi.org/10.1016/S0025-

6196(11)61793-X.

Hamilton, M. (1960). A rating scale for depression. Journal of Neurology,

Neurosurgery and Psychiatry, 23, 56–62. https://doi.org/10.1136/jnnp.23.1.56.

Harris, M., Glozier, N., Ratnavadivel, R., & Grunstein, R. R. (2009). Obstructive sleep

apnea and depression. Sleep Medicine Reviews.

https://doi.org/10.1016/j.smrv.2009.04.001.

Hartling, L., Hamm, M. P., Milne, A., Vandermeer, B., Santaguida, P. L.,Ansari, M., . .

. Dryden, D. M. (2013). Testing the risk of bias tool showed low reliability

between individual reviewers and across consensus assessments of reviewer

pairs. Journal of Clinical Epidemiology, 66, 973–981.

http://dx.doi.org/10.1016/j.jclinepi.2012.07.005.

Hashmi, A., Giray, N., & Hirshkowitz, M. (2006). Sleep-Related Breathing Disorders

and Mood Disorders. Sleep Medicine Clinics, 1(2), 247–262.

https://doi.org/10.1016/j.jsmc.2006.11.001.

Henry, J. D., & Crawford, J. R. (2005). The short-form version of the Depression

Anxiety Stress Scales (DASS-21): Construct validity and normative data in a

https://doi.org/10.1016/S0025-6196(11)61793-X

https://doi.org/10.1016/S0025-6196(11)61793-X

https://doi.org/10.1136/jnnp.23.1.56

https://doi.org/10.1016/j.smrv.2009.04.001

http://dx.doi.org/10.1016/j.jclinepi.2012.07.005

https://doi.org/10.1016/j.jsmc.2006.11.001


large non-clinical sample. British Journal of Clinical Psychology, 44, 227–239.

http://dx.doi.org/10.1348/014466505X29657.

Hillman, D. R., Murphy, A. S., & Pezzullo, L. (2006). The economic cost of sleep

disorders. Sleep, 29(3), 299–305.

Horn, J. L., & McArdle, J. J. (1992). A practical and theoretical guide to measurement

invariance in aging research. Experimental aging research,18(3), 117-144.

http://dx.doi.org/10.1080/03610739208253916.

Hozo, S. P., Djulbegovic, B., & Hozo, I. (2005). Estimating the mean and variance

from the median, range, and the size of a sample. BMC Medical Research

Methodology, 5, 13. http://dx.doi.org/10.1186/1471-2288-5-13.

Hsu, L. M. (1996). On the identification of clinically significant client changes:

Reinterpretation of Jacobson’s cut scores. Journal of Psychopathology and

Behavioral Assessment, 18, 371–385. http://dx.doi.org/10.1007/BF02229141.

Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure

analysis: Conventional criteria versus new alternatives. Structural Equation

Modeling: A Multidisciplinary Journal, 6(1), 1–55.

https://doi.org/10.1080/10705519909540118.

Jacobson, Dobson, & Truax. (1996) A component analysis of cognitive behavioral

treatment for depression. J Consult Clin Psychol, 64, 295-304.

Jacobson, N. S., & Truax, P. (1991). Clinical significance: A statistical approach to

defining meaningful change in psychotherapy research. Journal of Consulting

and Clinical Psychology, 59(1), 12–19. https://doi.org/10.1037/0022-

006X.59.1.12.

Jacobson, N. S., Follette, W. C., & Revenstorf, D. (1984). Psychotherapy outcome

research: Methods for reporting variability and evaluating clinical significance.

http://dx.doi.org/10.1348/014466505X29657

http://dx.doi.org/10.1080/03610739208253916

http://dx.doi.org/10.1186/1471-2288-5-13

http://dx.doi.org/10.1007/BF02229141

https://doi.org/10.1080/10705519909540118

https://doi.org/10.1037/0022-006X.59.1.12

https://doi.org/10.1037/0022-006X.59.1.12


Behavior Therapy, 15(4), 336–352. https://doi.org/10.1016/S0005-

7894(84)80002-7.

James, A., Hunter, M., Straker, L., Beilby, J., Bucks, R., Davis, T. . . . the Busselton

Healthy Ageing Study Investigator Group. (2013). Rationale, design and

methods for a community-based study of clustering and cumulative effects of

chronic disease processes and their effects on ageing: The Busselton healthy

ageing study. BMC Public Health, 13, 936. http://dx.doi.org/10.1186/1471-

2458-13-936.

Janssens, J. P., Pautex, S., Hilleret, H., & Michel, J. P. (2000). Sleep disordered

breathing in the elderly. Aging (Milan, Italy), 12, 417–429.

Jöreskog, K. G., & Sörbom, D. (1998). LISREL 8: Structural equation modeling with

the SIMPLIS command language. Hillsdale, NJ: Lawrence Erlbaum. Lovibond,

S. H., & Lovibond, P. F. (1995). Manual for the depression anxiety stress scales.

Sydney, Australia: Psychology Foundation.

Kapsimalis, F., & Kryger, M. H. (2002). Gender and obstructive sleep apnea syndrome,

part 1: Clinical features. Sleep, 25(4), 412–419.

Kapur, V. K., Auckley, D. H., Chowdhuri, S., Kuhlmann, D. C., Mehra, R., Ramar, K.,

& Harrod, C. G. (2017). Clinical practice guideline for diagnostic testing for

adult obstructive sleep apnea: an American Academy of Sleep Medicine clinical

practice guideline. J Clin Sleep Med, 13(3), 479-504.

Karacan, I., & Karatas, M. (1995). Erectile dysfunction in sleep apnea and response to

CPAP. Journal of Sex & Marital Therapy, 21, 239–247.

http://dx.doi.org/10.1080/00926239508414643.

https://doi.org/10.1016/S0005-7894(84)80002-7

https://doi.org/10.1016/S0005-7894(84)80002-7

http://dx.doi.org/10.1186/1471-2458-13-936

http://dx.doi.org/10.1186/1471-2458-13-936

http://dx.doi.org/10.1080/00926239508414643


Katon, W. J. (2003). Clinical and health services relationships between major

depression, depressive symptoms, and general medical illness. Biological

Psychiatry, 54, 216–226. http://dx.doi.org/10.1016/S0006-3223(03)00273-7.

Kerner, N. A., & Roose, S. P. (2016). Obstructive sleep apnea is linked to depression

and cognitive impairment: evidence and potential mechanisms. The American

Journal of Geriatric Psychiatry, 24(6), 496-508.

http://dx.doi.org/10.1016/j.jagp.2016.01.134

Kessler, R. C., Berglund, P., Demler, O., Jin, R., Koretz, D., Merikangas, K. R., ... &

Wang, P. S. (2003). The epidemiology of major depressive disorder: results

from the National Comorbidity Survey Replication (NCS-R). Jama, 289(23),

3095-3105.

Kingshott, R. N., Vennelle, M., Hoy, C. J., Engleman, H. M., Deary, I. J., & Douglas,

N. J. (2000). Predictors of improvements in daytime function outcomes with

CPAP therapy. American Journal of Respiratory and Critical Care Medicine,

161(3 I), 866–871. https://doi.org/10.1164/ajrccm.161.3.9905053.

Kripke, D. F., Ancoli-Israel, S., Klauber, M. R., Wingard, D. L., Mason, W. J., &

Mullaney, D. J. (1997). Prevalence of sleep-disordered breathing in ages 40-64

years: a population-based survey. Sleep, 20(1), 65–76. https://doi.org/9130337.

Kristiansen, H. A., Kvaerner, K. J., Akre, H., Overland, B., Sandvik, L., & Russell, M.

B. (2012). Sleep apnoea headache in the general population. Cephalalgia, 32,

451– 458. http://dx.doi.org/10.1177/0333102411431900.

Kryger, M.H., Roth, T., & Dement, W.C. (2016). Principles and Practice of Sleep

Medicine (sixth edition). St. Louis: Elsevier Sanders.

Kuo, T. F. (2000). Sleep-disordered breathing (SDB) and neuropsychological function:

A community sample study. Doctoral dissertation, University of Arizona.

http://dx.doi.org/10.1016/S0006-3223(03)00273-7

https://dx.doi.org/10.1016%2Fj.jagp.2016.01.134

https://doi.org/10.1164/ajrccm.161.3.9905053

https://doi.org/9130337

http://dx.doi.org/10.1177/0333102411431900


Lanza, S. T., Collins, L. M., Lemmon, D. R., & Schafer, J. L. (2007). PROC LCA: A

SAS procedure for latent class analysis. Structural Equation Modeling, 14(4),

671-694. http://dx.doi.org/10.1080/10705510701575602.

Lanza, S. T., Flaherty, B. P., & Collins, L. M. (2003). Latent class and latent transition

analysis. Handbook of psychology.

Law, M., Naughton, M., Ho, S., Roebuck, T., & Dabscheck, E. (2014). Depression may

reduce adherence during CPAP titration trial. Journal of clinical sleep medicine:

JCSM: official publication of the American Academy of Sleep Medicine, 10(2),

163.

Lee, I. S., Bardwell, W., Ancoli-Israel, S., Loredo, J. S., & Dimsdale, J. E. (2012).

Effect of 3 weeks of continuous positive airway pressure treatment on mood in

patients with obstructive sleep apnoea: a randomized placebo-controlled

study. Sleep Medicine, 13(2), 161.

Lee, W., Lee, S. A., Chung, Y. S., & Kim, W. S. (2015). The relation between apnea

and depressive symptoms in men with severe obstructive sleep apnea:

mediational effects of sleep quality. Lung, 193(2), 261-267.

Liu, Y., Su, C., Liu, R., Lei, G., Zhang, W., Yang, T., . . . Li, Z. (2011). NREM-AHI

greater than REM-AHI versus REM-AHI greater than NREM-AHI in patients

with obstructive sleep apnea: Clinical and polysomnographic features. Sleep and

Breathing, 15, 463–470. http://dx.doi.org/10.1007/s11325-010-0358-z.

Lovibond, P. F., & Lovibond, S. H. (1995). The structure of negative emotional states:

Comparison of the Depression Anxiety Stress Scales (DASS) with the Beck

Depression and Anxiety Inventories. Behaviour research and therapy, 33(3),

335-343. http://dx.doi.org/10.1016/0005-7967(94)00075-U.

http://dx.doi.org/10.1080/10705510701575602

http://dx.doi.org/10.1007/s11325-010-0358-z

http://dx.doi.org/10.1016/0005-7967(94)00075-U


Luyster, F. S., Buysse, D. J., & Strollo, P. J. (2010). Comorbid insomnia and

obstructive sleep apnea: challenges for clinical practice and research. Journal of

Clinical Sleep Medicine: JCSM: Official Publication of the American Academy

of Sleep Medicine, 6(2), 196–204.

Macey, P. M., Woo, M. A., Kumar, R., Cross, R. L., & Harper, R. M. (2010).

Relationship between obstructive sleep apnea severity and sleep, depression and

anxiety symptoms in newly-diagnosed patients. PLoS ONE, 5(4).

https://doi.org/10.1371/journal.pone.0010211.

Manber, R., Bernert, R. A., Suh, S., Nowakowski, S., Siebern, A. T., & Ong, J. C.

(2011). CBT for insomnia in patients with high and low depressive symptom

severity: adherence and clinical outcomes. J Clin Sleep Med, 7(6), 645-652.

Mansfield, D. R., Antic, N. A., & McEvoy, R. D. (2013). How to assess, diagnose, refer

and treat adult obstructive sleep apnoea: A commentary on the choices. The

Medical Journal of Australia, 199, 21–26.

http://dx.doi.org/10.5694/mja13.10909.

Marin, J. M., Agusti, A., Villar, I., Forner, M., Nieto, D., Carrizo, S. J. . . .Jelic, S.

(2012). Association between treated and untreated obstructive sleep apnea and

risk of hypertension. JAMA: Journal of the American. Medical Association, 307,

2169–2176. http://dx.doi.org/10.1001/jama.2012.3418.

Marin, J. M., Carrizo, S. J., & Kogan, I. (1998). Obstructive sleep apnea and acute

myocardial infarction: clinical implications of the association. Sleep, 21(8),

809–15.

Mathieu, A., Mazza, S., Décary, A., Massicotte-Marquez, J., Petit, D., Gosselin, N., . . .

Montplaisir, J. (2008). Effects of obstructive sleep apnea on cognitive function:

https://doi.org/10.1371/journal.pone.0010211

http://dx.doi.org/10.5694/mja13.10909

http://dx.doi.org/10.1001/jama.2012.3418


A comparison between younger and older OSAS patients. Sleep Medicine, 9,

112–120. http://dx.doi.org/10.1016/j.sleep.2007.03.014.

McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R. W. (2002).

Comparative longitudinal structural analyses of the growth and decline of

multiple intellectual abilities over the life span. Developmental

psychology, 38(1), 115. http://dx.doi.org/10.1037/0012-1649.38.1.115.

McCall, W. V., Harding, D., & O’Donovan, C. (2006). Correlates of depressive

symptoms in patients with obstructive sleep apnea. Journal of Clinical Sleep

Medicine, 2, 424–426.

McEvoy, R. D., Anderson, C. S., Antic, N. A., Chen, B., He, Q., Heeley, E., ... &

Zhong, N. (2011). The sleep apnea cardiovascular endpoints (SAVE) trial:

Rationale and start-up phase. Journal of thoracic disease, 2(3), 138-143.

Meade, A. W., Johnson, E. C., & Braddy, P. W. (2008). Power and sensitivity of

alternative fit indices in tests of measurement invariance. Journal of Applied

Psychology, 93, 568–592. http://dx.doi.org/10.1037/0021-9010.93.3.568.

Means, M. K., Lichstein, K. L., Edinger, J. D., Taylor, D. J., Durrence, H. H., Husain,

A. M., ... & Radtke, R. A. (2003). Changes in depressive symptoms after

continuous positive airway pressure treatment for obstructive sleep apnea. Sleep

and Breathing, 7(01), 031-042.

Meston, C. M. (1997). Aging and sexuality. Western Journal of Medicine, 167(4), 285.

Milfont, T. L., & Fischer, R. (2010). Testing measurement invariance across groups:

Applications in cross-cultural research. International Journal of Psychological

Research, 3, 111–130.

http://dx.doi.org/10.1016/j.sleep.2007.03.014

http://dx.doi.org/10.1037/0012-1649.38.1.115

http://dx.doi.org/10.1037/0021-9010.93.3.568


Millman, R. P., Fogel, B. S., McNamara, M. E., & Carlisle, C. C. (1989). Depression as

a manifestation of obstructive sleep apnea: Reversal with nasal continuous

positive airway pressure. Journal of Clinical Psychiatry, 50, 348–351.

Misri, S., Reebye, P., Corral, M., & Milis, L. (2004). The use of paroxetine and

cognitive-behavioral therapy in postpartum depression and anxiety: a

randomized controlled trial. The Journal of Clinical Psychiatry, 65(9), 1236–41.

https://doi.org/http://dx.doi.org/10.4088/JCP.v65n0913.

Morin, A. J., Arens, A. K., & Marsh, H. W. (2016). A bifactor exploratory structural

equation modeling framework for the identification of distinct sources of

construct-relevant psychometric multidimensionality. Structural Equation

Modeling, 23, 116 –139. http://dx.doi.org/10.1080/10705511.2014.961800.

Moss-Morris, R., Weinman, J., Petrie, K. J., Horne, R., Cameron, L. D., & Buick, D.

(2002). The revised Illness Perception Questionnaire (IPQ-R). Psychology &

Health, 17(1), 1–16. https://doi.org/10.1080/08870440290001494.

Motamedi, K. K., McClary, A. C., & Amedee, R. G. (2009). Obstructive sleep apnea: a

growing problem. The Ochsner Journal, 9(3), 149-153.

Munoz, A., Mayoralas, L. R., Barbe, F., Pericas, J., & Agusti, A. G. (2000). Long-term

effects of CPAP on daytime functioning in patients with sleep apnoea

syndrome. The European Respiratory Journal, 15(4), 676–681.

Muthén, L. K., & Muthén, B. O. (1998 –2011). MPlus user’s guide (6th ed.). Los

Angeles, CA: Muthén & Muthén.

Mysliwiec, V., Gill, J., Lee, H., Baxter, T., Pierce, R., Barr, T. L., . . . Roth, B. J.

(2013). Sleep disorders in US military personnel: A high rate of comorbid

insomnia and obstructive sleep apnea. Chest, 144, 549–557.

http://dx.doi.org/10.1378/chest.13-0088.

https://doi.org/http:/dx.doi.org/10.4088/JCP.v65n0913

http://dx.doi.org/10.1080/10705511.2014.961800

https://doi.org/10.1080/08870440290001494

http://dx.doi.org/10.1378/chest.13-0088


Nadeem, R., Molnar, J., Madbouly, E. M., Nida, M., Aggarwal, S., Sajid, H., …

Loomba, R. (2013). Serum inflammatory markers in obstructive sleep apnea: A

meta-analysis. Journal of Clinical Sleep Medicine, 9(10), 1003–1012.

https://doi.org/10.5664/jcsm.3070.

Nanthakumar, S., Bucks, R. S., & Skinner, T. C. (2016). Are we overestimating the

prevalence of depression in chronic illness using questionnaires? Meta-analytic

evidence in obstructive sleep apnoea. Health Psychology, 35, 423–432.

http://dx.doi.org/10.1037/hea0000280.

Nanthakumar, S., Bucks, R. S., Skinner, T. C., Starkstein, S., Hillman, D., James, A., &

Hunter, M. (2017). Assessment of the Depression, Anxiety, and Stress Scale

(DASS-21) in untreated obstructive sleep apnea (OSA). Psychological

Assessment, 29(10), 1201-1209. http://dx.doi.org/10.1037/pas0000401.

Nelson, T. S., & Allred, D. D. (1996). Survey research in marriage and family

therapy. Research methods in family therapy, 447-468.

Netzer, N. C., Stoohs, R. A., Netzer, C. M., Clark, K., & Strohl, K. P. (1999). Using the

Berlin Questionnaire to identify patients at risk for the sleep apnea

syndrome. Annals of internal medicine, 131(7), 485-491.

Ng, S. S. S., Chan, T. O., To, K. W., Ngai, J., Tung, A., Ko, F. W. S., &Hui, D. S. C.

(2009). Validation of a portable recording device (ApneaLink) for identifying

patients with suspected obstructive sleep apnoea syndrome. Internal Medicine

Journal, 39, 757–762. http://dx.doi.org/10.1111/j.1445-5994.2008.01827.x.

Norton, P. J. (2007). Depression Anxiety and Stress Scales (DASS-21): Psychometric

analysis across four racial groups. Anxiety, Stress, & Coping: An International

Journal, 20, 253–265. http://dx.doi.org/10.1080/10615800701309279.


http://dx.doi.org/10.1037/hea0000280

http://psycnet.apa.org/doi/10.1037/pas0000401

http://dx.doi.org/10.1111/j.1445-5994.2008.01827.x

http://dx.doi.org/10.1080/10615800701309279


Nye, C. D., & Drasgow, F. (2011). Effect size indices for analyses of measurement

equivalence: Understanding the practical importance of differences between

groups. Journal of Applied Psychology, 96(5), 966.

http://dx.doi.org/10.1037/a0022955.

Nye, C. D., Roberts, B. W., Saucier, G., & Zhou, X. (2008). Testing the measurement

equivalence of personality adjective items across cultures. Journal of Research

in Personality, 42, 1524–1536. http://dx.doi.org/10.1016/j.jrp.2008.07.004.

Olaithe, M., Nanthakumar, S., Eastwood, P. R., & Bucks, R. S. (2015). Cognitive and

mood dysfunction in adult obstructive sleep apnoea (OSA): Implications for

psychological research and practice. Translational Issues in Psychological

Science, 1(1), 67–78. https://doi.org/10.1037/tps0000021.

Olsen, S., Smith, S. S., Oei, T. P. S., & Douglas, J. (2012). Motivational interviewing

(MINT) improves continuous positive airway pressure (CPAP) acceptance and

adherence: A randomized controlled trial. Journal of Consulting and Clinical

Psychology, 80(1), 151–63. https://doi.org/10.1037/a0026302.

Olsen, S., Smith, S., Oei, T., & Douglas, J. (2008). Health belief model predicts

adherence to CPAP before experience with CPAP. European Respiratory

Journal, 32(3), 710–717. https://doi.org/10.1183/09031936.00127507.

Pagel, J. F., & Pandi-Perumal, S. R. (Eds.). (2014). Primary Care Sleep Medicine: A

Practical Guide. Springer.

Palmer, S., Vecchio, M., Craig, J. C., Tonelli, M., Johnson, D. W., Nicolucci, A., . . .

Strippoli, G. F. (2013). Prevalence of depression in chronic kidney disease:

Systematic review and meta-analysis of observational studies. Kidney

International, 84, 179–191. http://dx.doi.org/10.1038/ki.2013.77.

http://dx.doi.org/10.1037/a0022955

http://dx.doi.org/10.1016/j.jrp.2008.07.004

https://doi.org/10.1037/tps0000021

https://doi.org/10.1037/a0026302

https://doi.org/10.1183/09031936.00127507

http://dx.doi.org/10.1038/ki.2013.77


Patel, S. K., Hanly, P. J., Smith, E. E., Chan, W., & Coutts, S. B. (2015). Nocturnal

Hypoxemia Is Associated with White Matter Hyperintensities in Patients with a

Minor Stroke or Transient Ischemic Attack. Journal of Clinical Sleep Medicine:

JCSM: Official Publication of the American Academy of Sleep Medicine,

11(12), 1417–24. https://doi.org/10.5664/jcsm.5278.

Pelletier-Fleury, N., Rakotonanahary, D., & Fleury, B. (2001). The age and other

factors in the evaluation of compliance with nasal continuous positive airway

pressure for obstructive sleep apnea syndrome. A Cox's proportional hazard

analysis. Sleep medicine, 2(3), 225-232.

Peppard, P. E., Szklo-Coxe, M., Hla, K. M., & Young, T. (2006). Longitudinal

association of sleep-related breathing disorder and depression. Archives of

Internal Medicine, 166(16), 1709–1715.

https://doi.org/10.1001/archinte.166.16.1709.

Peppard, P. E., Young, T., Barnet, J. H., Palta, M., Hagen, E. W., & Hla, K. M. (2013).

Increased prevalence of sleep-disordered breathing in adults. American Journal

of Epidemiology, 177(9), 1006–1014. https://doi.org/10.1093/aje/kws342.

Phillips, B. G., Kato, M., Narkiewicz, K., Choe, I., & Somers, V. K. (2000). Increases

in leptin levels, sympathetic drive, and weight gain in obstructive sleep apnea.

American Journal of Physiology. Heart and Circulatory Physiology, 279,

H234–H237.

Piccinelli, M., & Wilkinson, G. (2000). Gender differences in depression. The British

Journal of Psychiatry, 177(6), 486-492.

Pillar, G., & Lavie, P. (1998). Psychiatric symptoms in sleep apnea syndrome: effects

of gender and respiratory disturbance index. Chest, 114(3), 697–703.

https://doi.org/10.1378/chest.114.3.697.


https://doi.org/10.1001/archinte.166.16.1709

https://doi.org/10.1093/aje/kws342



Porta, N., Bonet, C., & Cobo, E. (2007). Discordance between reported intention-to-

treat and per protocol analyses. Journal of Clinical Epidemiology.

https://doi.org/10.1016/j.jclinepi.2006.09.013.

Povitz, M., Bolo, C. E., Heitman, S. J., Tsai, W. H., Wang, J., & James, M. T. (2014).

Effect of Treatment of Obstructive Sleep Apnea on Depressive Symptoms:

Systematic Review and Meta-Analysis. PLoS Medicine, 11(11).

Radloff, L. S. (1977). The CES-D scale a self-report depression scale for research in the

general population. Applied Psychological Measurement, 1, 385–401.

http://dx.doi.org/10.1177/014662167700100306.

Ragette, R., Wang, Y., Weinreich, G., & Teschler, H. (2010). Diagnostic performance

of single airflow channel recording (ApneaLink) in home diagnosis of sleep

apnea. Sleep and Breathing, 14, 109–114. http://dx.doi.org/10.1007/s11325-

009-0290-2.

Raykov, T., Tomer, A., & Nesselroade, J. R. (1991). Reporting structural equation

modeling results in Psychology and Aging: some proposed

guidelines. Psychology and aging, 6(4), 499. http://dx.doi.org/10.1037/0882-

7974.6.4.499.

Reyes-Zúñiga, M., Castorena-Maldonado, A., Carrillo-Alduenda, J. L., Pérez-Padilla,

R., Martínez-Estrada, A., Gómez-Torres, L., & Torre-Bouscoulet, L. (2012).

Anxiety and depression symptoms in patients with sleep-disordered breathing.

The open respiratory medicine journal, 6, 97.

Richards, D., Bartlett, D. J., Wong, K., Malouff, J., & Grunstein, R. R. (2007).

Increased adherence to CPAP with a group cognitive behavioral treatment

intervention: a randomized trial. Sleep, 30(5), 635–640.

https://doi.org/10.1037/0278-6133.23.2.207.

https://doi.org/10.1016/j.jclinepi.2006.09.013

http://dx.doi.org/10.1177/014662167700100306

http://dx.doi.org/10.1007/s11325-009-0290-2

http://dx.doi.org/10.1007/s11325-009-0290-2

http://dx.doi.org/10.1037/0882-7974.6.4.499

http://dx.doi.org/10.1037/0882-7974.6.4.499

https://doi.org/10.1037/0278-6133.23.2.207


Robinaugh, D. J., LeBlanc, N. J., Vuletich, H. A., & McNally, R. J. (2014). Network

analysis of persistent complex bereavement disorder in conjugally bereaved

adults. Journal of abnormal psychology, 123(3), 510.

http://dx.doi.org/10.1037/abn0000002.

Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bulletin, 118, 183–

192. http://dx.doi.org/10.1037/0033-2909.118.2.183.

Roytblat, L., Rachinsky, M., Fisher, a, Greemberg, L., Shapira, Y., Douvdevani, a, &

Gelman, S. (2000). Raised interleukin-6 levels in obese patients. Obesity

Research, 8, 673–675. https://doi.org/10.1038/oby.2000.86.

Rush, A. J., Trivedi, M. H., Ibrahim, H. M., Carmody, T. J., Arnow, B., Klein, D. N., .

. . Keller, M. B. (2003). The 16-Item Quick Inventory of Depressive

Symptomatology (QIDS), clinician rating (QIDS-C), and self-report (QIDS-

SR): A psychometric evaluation in patients with chronic major depression.

Biological Psychiatry, 54, 573–583. http://dx.doi.org/10.1016/S0006-

3223(02)01866-8.

Rustad, J. K., Musselman, D. L., & Nemeroff, C. B. (2011). The relationship of

depression and diabetes: pathophysiological and treatment

implications. Psychoneuroendocrinology, 36(9), 1276-1286.

Rutledge, T., Reis, V. A., Linke, S. E., Greenberg, B. H., & Mills, P. J. (2006).

Depression in heart failure a meta-analytic review of prevalence, intervention

effects, and associations with clinical outcomes. Journal of the American

College of Cardiology, 48, 1527–1537.

http://dx.doi.org/10.1016/j.jacc.2006.06.055.

Sassani, A., Findley, L. J., Kryger, M., Goldlust, E., George, C., & Davidson, T. M.

(2004). Reducing motor-vehicle collisions, costs, and fatalities by treating

http://dx.doi.org/10.1037/abn0000002

http://dx.doi.org/10.1037/0033-2909.118.2.183

https://doi.org/10.1038/oby.2000.86

http://dx.doi.org/10.1016/S0006-3223(02)01866-8

http://dx.doi.org/10.1016/S0006-3223(02)01866-8

http://dx.doi.org/10.1016/j.jacc.2006.06.055


obstructive sleep apnea syndrome. Sleep, 27(3), 453–458.

https://doi.org/10.1164/rccm2102028.

Saunamäki, T., & Jehkonen, M. (2007). Depression and anxiety in obstructive sleep

apnea syndrome: a review. Acta Neurologica Scandinavica, 116(5), 277-288.

https://doi.org/10.1111/j.1600-0404.2007.00901.x.

Schrag, A., Barone, P., Brown, R. G., Leentjens, A. F., McDonald, W. M., Starkstein,

S. . . . Goetz, C. G. (2007). Depression rating scales in Parkinson’s disease:

Critique and recommendations. Movement Disorders, 22, 1077–1092.

http://dx.doi.org/10.1002/mds.21333.

Schröder, C. M., & O'Hara, R. (2005). Depression and obstructive sleep apnea

(OSA). Annals of general psychiatry, 4(1), 1-13. https://doi.org/10.1186/1744-

859X-4-13.

Schwab, R. J., Gefter, W. B., Hoffman, E. a, Gupta, K. B., & Pack, A. I. (1993).

Dynamic Upper Airway Imaging during Awake Respiration in Normal Subjects

and Patients with Sleep Disordered Breathing. American Review of Respiratory

Disease, 148(5), 1385–1400. https://doi.org/10.1164/ajrccm/148.5.1385

Schwartz, D. J., Kohler, W. C., & Karatinos, G. (2005). Symptoms of depression in

individuals with obstructive sleep apnea may be amenable to treatment with

continuous positive airway pressure. Chest, 128(3), 1304–1309.

https://doi.org/128/3/1304.

Sforza, E., De Saint Hilaire, Z., Pelissolo, A., Rochat, T., & Ibanez, V. (2002).

Personality, anxiety and mood traits in patients with sleep-related breathing

disorders: Effect of reduced daytime alertness. Sleep Medicine, 3(2), 139–145.

https://doi.org/10.1016/S1389-9457(01)00128-9.

https://doi.org/10.1164/rccm2102028

https://doi.org/10.1111/j.1600-0404.2007.00901.x

http://dx.doi.org/10.1002/mds.21333

https://doi.org/10.1186/1744-859X-4-13

https://doi.org/10.1186/1744-859X-4-13

https://doi.org/10.1164/ajrccm/148.5.1385

https://doi.org/128/3/1304

https://doi.org/10.1016/S1389-9457(01)00128-9


Sharafkhaneh, A., Giray, N., Richardson, P., Young, T., & Hirshkowitz, M. (2005).

Association of psychiatric disorders and sleep apnea in a large cohort. Sleep,

28(11), 1405–1411. https://doi.org/10.1016/S8756-3452(08)70453-5.

Sin, D. D., Mayers, I., Man, G. C., & Pawluk, L. (2002). Long-term compliance rates to

continuous positive airway pressure in obstructive sleep apnea: a population-

based study. CHEST Journal, 121(2), 430-435.

Snaith, R. P., & Zigmond, A. S. (1986). The hospital anxiety and depression scale.

British Medical Journal (Clinical Research Ed.), 292, 344.

http://dx.doi.org/10.1136/bmj.292.6516.344.

Spicuzza, L., Caruso, D., & Di Maria, G. (2015). Obstructive sleep apnoea syndrome

and its management. Therapeutic Advances in Chronic Disease, 6(5), 273–285.

http://doi.org/10.1177/2040622315590318.

Starkstein, S. E., Davis, W. A., Dragovic, M., Cetrullo, V., Davis, T. M., & Bruce, D.

G. (2014). Diagnostic criteria for depression in type 2 diabetes: a data-driven

approach. PloS one, 9(11), e112049.

Starkstein, S., Dragovic, M., Jorge, R., Brockman, S., Merello, M., Robinson, R. G., ...

& Wilson, M. (2011). Diagnostic criteria for depression in Parkinson's disease: a

study of symptom patterns using latent class analysis. Movement

Disorders, 26(12), 2239-2245.

Steenkamp, J. B. E., & Baumgartner, H. (1998). Assessing measurement invariance in

cross‐national consumer research. Journal of consumer research, 25(1), 78-107.

http://doi.org/10.1086/209528.

Steiger, J. H., & Lind, J. C. (1980). Statistically based tests for the number of common

factors. In Proceedings of the Annual Meeting of the Psychometric Society

(Vol. 758, pp. 424–453). Iowa City, IA.

https://doi.org/10.1016/S8756-3452(08)70453-5

http://dx.doi.org/10.1136/bmj.292.6516.344

http://doi.org/10.1177/2040622315590318

http://doi.org/10.1086/209528


Stepnowsky, C. J., Marler, M. R., & Ancoli-Israel, S. (2002). Determinants of nasal

CPAP compliance. Sleep medicine, 3(3), 239-247.

Stepnowsky, C. J., Marler, M. R., Palau, J., & Brooks, J. A. (2006). Social-cognitive

correlates of CPAP adherence in experienced users. Sleep medicine, 7(4), 350-

356. http://doi.org/10.1016/j.sleep.2005.11.004.

Stepnowsky, C. J., Palau, J. J., Zamora, T., Ancoli-Israel, S., & Loredo, J. S. (2011).

Fatigue in sleep apnea: The role of depressive symptoms and self-reported sleep

quality. Sleep Medicine, 12, 832–837.

http://dx.doi.org/10.1016/j.sleep.2011.07.004.

Stepnowsky, C., Palau, J. J., Marler, M. M., & Gifford, A. L. (2006). Effect of wireless

monitoring on CPAP adherence and treatment efficacy: a pilot study [Abstract].

Sleep. Retrieved from

http://www.mrw.interscience.wiley.com/cochrane/clcentral/articles/626/CN-

00591626/frame.html.

Steyer, R., Eid, M., & Schwenkmezger, P. (1997). Modeling true intraindividual

change: True change as a latent variable. Methods of Psychological Research

Online, 2(1), 21-33.

Svaldi, J. J., & Mackinger, H. F. (2003). Obstructive sleep apnea syndrome:

Autobiographical memory predicts the course of depressive affect after nCPAP

therapy. Scandinavian Journal of Psychology, 44, 31–37.

http://dx.doi.org/10.1111/1467-9450.00318.

Szabó, M. (2010). The short version of the Depression Anxiety Stress Scales (DASS-

21): Factor structure in a young adolescent sample. Journal of Adolescence, 33,

1–8. http://dx.doi.org/10.1016/j.adolescence.2009.05.014.

http://doi.org/10.1016/j.sleep.2005.11.004

http://dx.doi.org/10.1016/j.sleep.2011.07.004

http://www.mrw.interscience.wiley.com/cochrane/clcentral/articles/626/CN-00591626/frame.html

http://www.mrw.interscience.wiley.com/cochrane/clcentral/articles/626/CN-00591626/frame.html

http://dx.doi.org/10.1111/1467-9450.00318

http://dx.doi.org/10.1016/j.adolescence.2009.05.014


Tasali, E., Mokhlesi, B., & Van Cauter, E. (2008). Obstructive sleep apnea and type 2

diabetes: Interacting epidemics. Chest, 133, 496–506.

http://dx.doi.org/10.1378/chest.07-0828.

Tucker, L. R., & Lewis, C. (1973). A reliability coefficient for maximum likelihood

factor analysis. Psychometrika, 38, 1–10.

http://dx.doi.org/10.1007/BF02291170.

Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement

invariance literature: Suggestions, practices, and recommendations for

organizational research. Organizational Research Methods, 3, 4–70.

http://dx.doi.org/10.1177/109442810031002.

Vandeputte, M., & de Weerd, A. (2003). Sleep disorders and depressive feelings: A

global survey with the Beck depression scale. Sleep Medicine, 4, 343–345.

http://dx.doi.org/10.1016/S1389-9457(03)00059-5.

Veasey, S. C. (2003). Serotonin agonists and antagonists in obstructive sleep apnea:

therapeutic potential. American Journal of Respiratory Medicine: Drugs,

Devices, and Other Interventions, 2(1), 21–9.

https://doi.org/10.1007/BF03256636.

Vgontzas, A. N. (2008). Excessive daytime sleepiness in sleep apnea: it’s not just apnea

hypopnea index. Sleep medicine, 9(7), 712.


Vgontzas, A. N., Bixler, E. O., & Chrousos, G. P. (2005). Sleep apnea is a

manifestation of the metabolic syndrome. Sleep Medicine Reviews, 9, 211–224.

http://dx.doi.org/10.1016/j.smrv.2005.01.006.

Wahner-Roedler, D. L., Olson, E. J., Narayanan, S., Sood, R., Hanson, A. C., Loehrer,

L. L., & Sood, A. (2007). Gender-specific differences in a patient population

http://dx.doi.org/10.1378/chest.07-0828

http://dx.doi.org/10.1007/BF02291170

http://dx.doi.org/10.1177/109442810031002

http://dx.doi.org/10.1016/S1389-9457(03)00059-5

https://doi.org/10.1007/BF03256636

https://dx.doi.org/10.1016%2Fj.sleep.2008.05.001

http://dx.doi.org/10.1016/j.smrv.2005.01.006


with obstructive sleep apnea-hypopnea syndrome. Gender Medicine, 4, 329 –

338. http://dx.doi.org/10.1016/S1550-8579(07)80062-3.

Waters, F., & Bucks, R. S. (2011). Neuropsychological effects of sleep loss: implication

for neuropsychologists. Journal of the International Neuropsychological

Society, 17(04), 571-586. http://dx.doi.org/10.1017/S1355617711000610.

Weaver, T. E., & Grunstein, R. R. (2008). Adherence to continuous positive airway

pressure therapy: the challenge to effective treatment. Proceedings of the

American Thoracic Society, 5(2), 173-178.

Weaver, T. E., & Sawyer, A. M. (2010). Adherence to continuous positive airway

pressure treatment for obstructive sleep apnea: implications for future

interventions. The Indian journal of medical research, 131, 245.

Wells, R. D., Freedland, K. E., Carney, R. M., Duntley, S. P., & Stepanski, E. J. (2007).

Adherence, reports of benefits, and depression among patients treated with

continuous positive airway pressure. Psychosomatic medicine, 69(5), 449-454.

http://dx.doi.org/10.1097/psy.0b013e318068b2f7.

Wild, M. R., Engleman, H. M., Douglas, N. J., & Espie, C. A. (2004). Can

psychological factors help us to determine adherence to CPAP? A prospective

study. European Respiratory Journal, 24(3), 461–465.

https://doi.org/10.1183/09031936.04.00114603.

Wood, B. M., Nicholas, M. K., Blyth, F., Asghari, A., & Gibson, S. (2010). The utility

of the short version of the Depression Anxiety Stress Scales (DASS-21) in

elderly patients with persistent pain: Does age make a difference? Pain

Medicine, 11, 1780–1790. http://dx.doi.org/10.1111/j.1526-4637.2010.01005.x.

http://dx.doi.org/10.1016/S1550-8579(07)80062-3

http://dx.doi.org/10.1017/S1355617711000610

http://dx.doi.org/10.1097/psy.0b013e318068b2f7

https://doi.org/10.1183/09031936.04.00114603

http://dx.doi.org/10.1111/j.1526-4637.2010.01005.x


World Health Organization. (1990). Composite International Diagnostic Interview

(CIDI). Geneva, Switzerland: World Health Organisation, Division of Mental

Health.

Yaffe, K., Laffan, A. M., Harrison, S. L., Redline, S., Spira, A. P., Ensrud, K. E., . . .

Stone, K. L. (2011). Sleep-disordered breathing, hypoxia, and risk of mild

cognitive impairment and dementia in older women. JAMA: Journal of the

American Medical Association, 306, 613–619.

http://dx.doi.org/10.1001/jama.2011.1115.

Yaggi, H. K., Concato, J., Kernan, W. N., Lichtman, J. H., Brass, L. M., & Mohsenin,

V. (2005). Obstructive sleep apnea as a risk factor for stroke and death. The New

England Journal of Medicine, 353, 2034–2041.

http://dx.doi.org/10.1056/NEJMoa043104.

Yamamoto, H., Akashiba, T., Kosaka, N., Ito, D., & Horie, T. (2000). Long-term

effects nasal continuous positive airway pressure on daytime sleepiness, mood

and traffic accidents in patients with obstructive sleep apnoea. Respir Med,

94(1), 87–90. https://dx.doi.org/S0954-6111(99)90698-4.

Yap, W. S., & Fleetham, J. A. (2001). Central sleep apnea and hypoventilation

syndrome. Current treatment options in neurology, 3(1), 51-55.

Ye, L., Liang, Z. A., & Weaver, T. E. (2008). Predictors of health‐related quality of life

in patients with obstructive sleep apnoea. Journal of advanced nursing, 63(1),

54-63. https://dx.doi.org/10.1111/j.1365-2648.2008.04652.x.

Ye, L., Pien, G. W., & Weaver, T. E. (2009). Gender differences in the clinical

manifestation of obstructive sleep apnea. Sleep Medicine.


http://dx.doi.org/10.1001/jama.2011.1115

http://dx.doi.org/10.1056/NEJMoa043104

https://dx.doi.org/S0954-6111(99)90698-4

https://dx.doi.org/10.1111/j.1365-2648.2008.04652.x

https://doi.org/10.1016/j.sleep.2009.02.006


Ye, L., Pien, G. W., Ratcliffe, S. J., & Weaver, T. E. (2009). Gender differences in

obstructive sleep apnea and treatment response to continuous positive airway

pressure. J Clin Sleep Med, 5(6), 512-518.

Yesavage, J. A., & Sheikh, J. I. (1986). Geriatric Depression Scale (GDS): Recent

evidence and development of a shorter violence. Clinical Gerontologist, 5, 165–

173. http://dx.doi.org/10.1300/J018v05n01_09.

Young, T., Palta, M., Dempsey, J., Skatrud, J., Weber, S., & Badr, S. (1993). The

occurrence of sleep-disordered breathing among middle aged adults. The New

England Journal of Medicine, 328, 1230–1235.

http://dx.doi.org/10.1056/NEJM199304293281704.

Young, T., Peppard, P. E., & Gottlieb, D. J. (2002). Epidemiology of obstructive sleep

apnea: A population health perspective. American Journal of Respiratory and

Critical Care Medicine. https://doi.org/10.1164/rccm.2109080.

Young, T., Shahar, E., Nieto, F. J., Redline, S., Newman, A. B., Gottlieb, D. J., . . . the

Sleep Heart Health Study Research Group. (2002). Predictors of sleep-

disordered breathing in community-dwelling adults: The Sleep Heart Health

Study. Archives of Internal Medicine, 162, 893–900.

http://dx.doi.org/10.1001/archinte.162.8.893.

Young, T., Skatrud, J., & Peppard, P. E. (2004). Risk factors for obstructive sleep apnea

in adults. JAMA: The Journal of the American Medical Association, 291(16),

2013–6. https://doi.org/10.1001/jama.291.16.2013.

Zimmerman, M., Martinez, J. H., Young, D., Chelminski, I., & Dalrymple, K. (2013).

Severity classification on the Hamilton depression rating scale. Journal of

Affective Disorders, 150(2), 384–388. https://doi.org/10.1016/j.jad.2013.04.028.

http://dx.doi.org/10.1300/J018v05n01_09

http://dx.doi.org/10.1056/NEJM199304293281704

https://doi.org/10.1164/rccm.2109080

http://dx.doi.org/10.1001/archinte.162.8.893

https://doi.org/10.1001/jama.291.16.2013

https://doi.org/10.1016/j.jad.2013.04.028


Zung, W. W. (1965). A self-rating depression scale. Archives of General Psychiatry,

12, 63–70. http://dx.doi.org/10.1001/archpsyc.1965.01720310065008.

http://dx.doi.org/10.1001/archpsyc.1965.01720310065008


Appendix A

For Chapter 2 (Study 1)

Appendix A was online supplementary material for Study 1

Figure S2.1. Forest plot for the effect of sample type on prevalence estimates (event

rate) using a random effects model.

Figure S2.2. Moderated Regression plot of the proportion of OSA items against the




Figure S2.3. Moderated Regression plot of the proportion of loss of energy/fatigue

items against the prevalence of depression (as a logit event rate). Note: Larger logit

event rate = higher prevalence estimate.

Figure S2.4. Moderated Regression plot of the proportion of sleep items against the




Figure S2.5. Moderated Regression plot of the proportion of anhedonia items against

the prevalence of depression (as a logit event rate). Note: Larger logit event rate =

higher prevalence estimate.

Figure S2.6. Moderated Regression plot of the proportion of males against the




Appendix B

For Chapter 3 (Study 2)

Table S3.1

Single growth curve model building for total HAM-D scores, by per protocol analysis

Note: N = 135. Bold font: best fitting model. 1 = q@0. AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)] where k is the number

of observed variables. Recommended cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999). †p < .05, ‡ p<.01. Growth parameter

estimates are unstandardized values.

Model 2 df p AIC Parsimony

Ratio

RMSEA





1.Intercept only 70.49 8 <.001 3178.28 .80 .24 (.19 - .29) .69 .77 .21 7.55‡ 25.62‡ 0 0 N/A N/A 2.Linear change 28.28 5 <.001 3128.05 .50 .19 (.12 - .26) .89 .86 .09 9.60‡ 28.69‡ -1.14‡ 0.83 N/A N/A

3. Quadratic

change 0.60 1 .439 3093.71 .10 0 (0 - .21) 1.00 1.00 .01 10.97‡ 41.90‡ -3.89‡ 16.81 0.85‡ 0.61

4. Quadratic

change1 2.62 4 .623 3091.13 .40 0 (0 -.11) 1.00 1.00 .04 10.91‡ 30.25‡ -3.87‡ 1.30† 0.84‡ 0

5. Logarithmic

change 14.18 5 .015 3105.68 .50 .12 (.05 - .19) .96 .95 .05 10.47‡ 31.94‡ -3.09‡ 6.10† N/A N/A


Table S3.2

Single growth curve model building for overlapping symptom percentage scores, by per protocol analysis


Ratio

RMSEA





1.Intercept only 79.60 8 <.001 4260.16 .80 .26 (.21 - .31) .57 .68 .27 20.06‡ 153.24‡ 0 0 N/A N/A

2.Linear change 34.00 5 <.001 4209.78 .50 .21 (.15 - .28) .83 .79 .11 26.27‡ 165.29‡ -3.30‡ 7.78 N/A N/A

3. Quadratic change 2.17 1 .141 4171.85 .10 .09 (0 - .27) .99 .96 .02 30.61‡ 267.12‡ -11.64‡ 121.95 2.54‡ 4.86

4. Quadratic change1 3.75 4 .442 4169.16 .40 0 (0 - .13) 1.00 1.00 .04 30.34‡ 181.96‡ -11.53‡ 12.01‡ 2.52‡ 0

5. Logarithmic change 17.39 5 .004 4185.56 .50 .14 (.07 - .21) .93 .91 .06 29.03‡ 195.76‡ -9.18‡ 55.83† N/A N/A

Note. N = 135. Bold font: best fitting model. 1 = q@0. AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)] where k is the number




Table S3.3

Single growth curve model building for non-overlapping symptom percentage scores, by per protocol analysis

Note. N = 135. Bold font: best fitting model. 1 = q@0. AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)]

where k is the number of observed variables. Recommended cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999); †p < .05, ‡ p <.01. Growth parameter estimates are unstandardized values.


Ratio

RMSEA






2.Linear change 10.83 5 .055 3561.40 .40 .09 (0 - .17) .97 .97 .06 10.64‡ 72.74‡ -1.08‡ 0.38 N/A N/A

3. Quadratic change 0.42 1 .520 3553.92 .40 0 (0 - .20) 1.00 1.00 .01 11.50‡ 101.23‡ -3.17‡ 47.61 0.66‡ 2.08

4. Quadratic change1 4.08 4 .395 3553.65 .40 .01 (0 - .13) 1.00 1.00 .04 11.59‡ 73.15‡ -3.21‡ 0.66 0.67‡ 0

5. Logarithmic change 7.62 5 .179 3556.50 .40 .06 (0 - .15) .99 .99 .05 11.20‡ 74.50‡ -2.65‡ 3.85 N/A N/A


Table S3.4

Single growth curve model building for total HAM-D scores, by intention to treat analysis





Ratio

RMSEA






2.Linear change 33.60 5 <.001 4003.81 .50 .18 (.13 - .24) .91 .89 .09 10.24‡ 34.64‡ -0.86‡ 0.88† N/A N/A

3. Quadratic change 1.23 1 .268 3964.22 .10 .04 (0 - .21) 1.00 1.00 .01 11.54‡ 43.53‡ -3.25‡ 14.25 0.71 0.54

4. Quadratic change1 3.62 4 .460 3962.29 .40 0 (0 - .11) 1.00 1.00 .03 11.48‡ 35.58‡ -3.23‡ 1.21‡ 0.71‡ 0

5. Logarithmic change 17.41 5 .004 3979.66 .50 .12 (.06 - .18) .96 .95 .05 11.02‡ 36.19‡ -2.46‡ 5.69† N/A N/A


Table S3.5

Single growth curve model building for non-overlapping symptom percentage scores, by intention to treat analysis





Ratio

RMSEA





1.Intercept only 37.82 8 <.001 4612.26 .80 .15

(.10 - .20) .91 .93 .08 10.36‡ 82.09‡ 0 0 N/A N/A

2.Linear change 13.34 5 .020 4589.10 .50 .10

(.04 -.16) .97 .97 .05 11.83‡ 91.94‡ -0.89‡ 0.37 N/A N/A

3. Quadratic change 0.05 1 .854 4557.49 .10 0

(0 - .11) 1.00 1.00 .00 12.88‡ 109.14‡ -3.00‡ 31.05 0.65‡ 1.67

4. Quadratic change1 3.04 4 .552 4577.42 .40 0

(0 - .10) 1.00 1.00 .02 12.90‡ 92.39‡ -3.00‡ 0.61 0.65‡ 0

5. Logarithmic change 8.10 5 .151 4581.96 .50 .06

(0 - .13) .99 .99 .03 12.41‡ 93.97‡ -2.30‡ 2.93 N/A N/A


Table S3.6

Single growth curve model building for overlapping symptom percentage scores, by intention to treat analysis





Ratio

RMSEA





1.Intercept only 89.12 8 <.001 5470.25 .80 .24

(.20 - .29) .68 .76 .25 22.87‡ 209.80‡ 0 0 N/A N/A

2.Linear change 37.34 5 <.001 5414.93 .50 .19

(.14 - .25) .87 .85 .11 27.49‡ 196.97‡ -2.40‡ 8.49† N/A N/A

3. Quadratic

change 2.79 1 .095 5373.74 .10

.10

(0 - .25) .99 .96 .01 31.49‡ 269.13‡ -9.38‡ 120.58 2.06‡ 4.29

4. Quadratic

change1 5.77 4 .217 5373.33 .40

.05

(0 - .13) .99 .99 .05 31.20‡ 205.87‡ -9.28‡ 11.38‡ 2.03‡ 0

5. Logarithmic

change 19.55 5 .002 5389.89 .50

.13

(.07 - .19) .94 .93 .06 29.86‡ 208.19‡ -7.05‡ 54.39‡ N/A N/A


Table S3.7

Individual models of HAM-D total scores, non-overlapping symptom percentage scores, and non-overlapping symptom percentage scores,

predicted by age, sex, baseline AHI, antidepressant use, education years, BMI and average CPAP use, by intention to treat analysis

Outcome 2 df p AIC Parsimony

Ratio RMSEA



Intercept on

predictor

Slope on

predictor

Mean Mean

Total HAM-D

scores 25.70 18 .107 3708.75 .27

.05

(0 - .09) .99 .97 .02

Age -0.08† 0.01

Sex 1.40 0.11



Education years -0.17 0.03

BMI 0.06 -0.02

Average CPAP use -0.65† -0.20‡

Non-

overlapping

symptom

percentage

scores

15.66 18 .616 4303.26 .27 0

(0 - .06) 1.00 1.00 .02

Age 0.15‡ 0.01

Sex 1.49 -0.01


Antidepressant use 5.63‡ 0.11


BMI 0.09 -0.01

Average CPAP use -0.99† -0.10

Overlapping

symptom percentage

scores

28.97 18 .049 5024.29 .27 .06

(0 - .10) .98 .96 .03

Age -0.13 0.01

Sex 2.86 0.51




BMI 0.14 -0.08

Average CPAP use -1.51† -0.67‡

Note. N = 165. AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)] where k is the number of observed variables. Recommended

cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999). †p < .05, ‡ p<.01. Growth parameter estimates are unstandardized values.


Table S3.8

Individual models of HAM-D total scores, non-overlapping symptom percentage scores, and non-overlapping symptom percentage scores,

predicted by age, sex, baseline AHI, antidepressant use, education years, BMI, and proportion of CPAP use 4 hours, by intention to treat

analysis

Note. N = 165. AIC = Akaike Information Criterion (lower score = better fit). Parsimony ratio is defined as df / [.5k(k+1)] where k is the number of observed variables.

Recommended cutoffs for the fit indexes: SRMR: .060, CFI: .95, TLI: .95, and RMSEA: .050 (Hu & Bentler, 1999). †p < .05, ‡ p<.01. Growth parameter estimates are unstandardized values.

Outcome 2 df p AIC Parsimony

Ratio

RMSEA



Intercept on predictor

Slope on predictor

Mean Mean

Total HAM-D

scores 20.09 18 .328 3706.63 .27

.03

(0 - .08) 1.00 .99 .02

Age -0.07† 0.01

Sex 1.14 0.14



BMI 0.07 -0.02

Proportion of 4 hours of

CPAP -0.05‡ -0.01‡

Non-

overlapping

symptom

percentage scores

13.70 18 .749 4301.22 .27 0

(0 - .05) 1.00 1.00 .02

Age -0.15‡ 0.01

Sex 1.49 0.00



BMI 0.11 -0.01


-0.08† -0.01

Overlapping

symptom percentage

scores

23.87 18 .159 5022.49 .27 .04

(0 - .09) .99 .98 .03

Age -0.12 0.01

Sex 2.86 0.55

Baseline AHI -0.01 0.01 Antidepressant use 10.40‡ -0.64


BMI 0.16 -0.07

Proportion of 4 hours of

CPAP -0.12† -0.05‡

Documents

Depressive symptomatology in Obstructive Sleep Apnoea ... fileDepressive symptomatology in Obstructive Sleep Apnoea