SLEEP, Vol. 28, No. 12, 2005

INTRODUCTION

SEVERAL STUDIES HAVE SUGGESTED A CHRONIC STATE OF HYPERAROUSAL AS THE UNDERLYING MECHANISM OF CHRONIC INSOMNIA.1-3 IN A SIMILAR VEIN, insomni-acs’ complaints of elevated daytime levels of anxiety and fatigue have been hypothesized to be attributable to chronic hyperarous-al, rather than being a consequence of the previous night’s poor sleep.4-6 Such hyperarousal could result from a combination of internalization of significant stressful events (e.g., divorce, poor job satisfaction) and certain predisposing factors.2,5,7 In the con-text of insomnia, hyperarousal is often conceptually differentiated in cognitive, emotional, and physiologic hyperarousal. In experi-mental studies of insomnia, it has been difficult to discriminate these concepts,8 suggesting that they may have the same etiol-ogy.

Bonnet and Arand9 assessed cardiac autonomic activity in in-somniacs and good sleepers as a marker of physiologic arousal. They reported that insomniacs showed increased sympathetic and decreased parasympathetic nervous system activity during sleep, which the authors attributed to the influence of hyperarous-al. Domitrovich et al10 presented preliminary data of decreased parasympathetic activity specifically in non-rapid eye movement (NREM) sleep in a group of insomniacs. The increase in physi-ologic arousal that would accompany this change in sympathova-gal influence might be a reason why insomniacs do not sleep well.11 Other indexes have been employed to assess the presence of physiologic arousal as well. For instance, in a study by Vgont-zas et al,3 an increase in the insomniacs’ cortisol level during the nocturnal period was found, particularly around the sleep-onset period. The authors related the increase in nighttime cortisol se-cretion to hyperarousal rather than to the previous night’s poor sleep, based on the finding that sleep loss does not disrupt the usual pattern of cortisol levels in healthy young adults.12 Similar findings were shown in other studies measuring hypothalamic-pituitary-adrenal axis activity.13,14 Further studies have shown an increase in alpha and beta frequencies in the electroencepha-lograms of insomniacs during sleep,15,16 an increase in 24-hour whole-body metabolism17 and brain metabolism,18 and an eleva-tion in body temperature.1,19,20 These effects have been interpreted as converging evidence of a relationship between insomnia and hyperarousal. Despite these consistent findings, there is still some doubt

Physiologic Indexes in Chronic Insomnia During a Constant Routine: Evidence for General Hyperarousal?Michael Varkevisser, PhD1,2; Hans P.A. Van Dongen, PhD3; Gerard A. Kerkhof, PhD1,4

1Department of Psychology, University of Amsterdam, Amsterdam, The Netherlands; 2Cognitive Psychology Unit, Department of Psychology, Leiden University, Leiden, The Netherlands; 3Unit for Experimental Psychiatry, Division of Sleep and Chronobiology, Department of Psychiatry, and Center for Sleep and Respiratory Neurobiology, University of Pennsylvania School of Medicine, Philadelphia, PA; 4Centre for Sleep and Wake Disorders, MCH Westeinde, The Hague, The Netherlands

Physiological Indexes in Chronic Insomnia—Varkevisser et al

Disclosure StatementThis was not an industry supported study. Dr. Van Dongen has received research support from Cephalon, Inc. Drs. Varkevisser and Kerkhof have indicated no financial conflicts of interest.

Submitted for publication August 2004Accepted for publication August 2005Address correspondence to: Michael Varkevisser, Department of Psychol-ogy, University of Amsterdam, Roetersstraat 15, 1018 WB Amsterdam, The Netherlands; Telephone: 31 15 2784640; Fax: 31 15 2787316; E-mail: M.Varkevisser@uva.nl

1588

Study Objectives: It has been hypothesized that general hyperarousal, present during both sleep and wakefulness, may underlie chronic in-somnia. The present study explored, under strictly controlled conditions, whether chronic insomnia is associated with altered physiologic markers of arousal, both in absolute levels and in terms of circadian rhythmicity, relative to controls. Design: A 24-hour constant-routine protocol was implemented to assess physiologic measures.Setting: The study was conducted in an isolated, temperature- and light-controlled, sound-attenuated sleep laboratory.Participants: Eleven subjects with clinically diagnosed chronic insomnia were compared with 13 healthy matched controls.Interventions: The subjects underwent physiologic parameter recordings and cognitive performance testing during 24 hours of total sleep depriva-tion under strictly controlled circumstances. Measurements and Results: Cardiovascular parameters, free cortisol, and body temperature were subjected to mixed-model analysis of vari-ance and mixed-model harmonic regression. Overall, no differences were found in either the absolute level or the circadian parameters (amplitude,

phase) of these variables between the insomniacs and the control sub-jects. Conclusions: Although physiologic indexes of arousal were slightly el-evated in the insomnia group relative to the controls, the differences be-tween the groups were not statistically significant. This could have been due to a lack of statistical power or could reflect the actual absence of arousal in our sample of chronic insomniacs. Systematic interindividual level differences overwhelmed any differences between the 2 groups, making it unlikely that general hyperarousal was a critical underlying factor in our sample. Earlier findings of hyperarousal in insomnia during studies that allowed sleep may have been specifically related to the sleep state.Keywords: Chronic insomnia, general hyperarousal, constant routine, sleep deprivation, sympathovagal activity, free cortisol, core body tem-perature, interindividual differencesCitation: Varkevisser M; Van Dongen HPA; Kerkhof GA. Physiologic in-dexes in chronic insomnia during a constant routine: evidence for general hyperarousal? SLEEP 2005;28(12): 1588-1596.

SLEEP, Vol. 28, No. 12, 2005

whether physiologic hyperarousal is a primary cause of chronic insomnia. A study by Riemann and colleagues21 measuring noc-turnal plasma cortisol levels did not replicate the findings of Vgontzas et al.3 Furthermore, in a study of Stepanski et al,22 no evidence of hyperarousal was found when insomniacs were sleep deprived. In the latter study, the influence of hyperarousal was expected to moderate sleepiness as induced by sleep loss. How-ever, the level of physiologic sleepiness after a night of sleep de-privation was comparable in the insomnia group and in a healthy control group. Finally, in a laboratory study by Lichstein et al,23 daytime pupillometry was applied as an index for arousal in the sympathetic nervous system. The results showed no differences in pupil diameter between insomniacs and healthy controls, indicat-ing no daytime differences in sympathetic nervous system activ-ity. While the majority of published studies offer support for some type of hyperarousal, these contradictory findings suggest that the role of arousal as a major contributor to poor sleep and daytime symptoms (i.e., sleepiness) in insomnia should be considered with some caution. Most studies to date have not controlled for exogenous or physiologic factors that could influence the level of arousal, such as posture, light exposure, etc. It is noteworthy that laboratory studies in which these so-called masking factors were not con-trolled well failed to find any distinct differences in cognitive per-formance between insomniacs and controls.24 Under the strictly controlled conditions of a constant-routine study,25 however, we have found that cognitive performance is consistently impaired in chronic insomniacs compared with healthy controls.26 Precise experimental control over masking factors may be equally crucial for measuring hyperarousal in insomnia. The aim of the present study was to assess, in a constant-routine protocol employed to neutralize masking factors, whether chronic insomnia is associated with changes in sympathovagal activity and changes in cortisol and temperature levels across the 24-hour day. Our research hypothesis was that, throughout the 24 hours, chronic insomniacs would display an increased level of arousal as measured by these physiologic indexes, relative to age- and sex-matched healthy controls. We expected to observe 1 of 3 possible outcomes for the insomniacs: (a) a systematically elevated level of arousal (i.e., general hyperarousal across the 24-hour day), (b) a specifically nocturnal increase in arousal (around the normal time for sleep), or (c) no differences in the level of arousal relative to healthy controls.

METHODS

Subjects

The patient group of this study consisted of 13 individuals diag-nosed with chronic psychophysiologic insomnia (7 men, 6 wom-en; mean age ± SD: 43.8 ± 8.9 years; range 31-54). Patients had been referred to the Centre for Sleep and Wake Disorders at the Westeinde Hospital (The Hague) by their physician. All patients underwent 48-hour ambulatory polysomnography according to the standard protocol of the Centre for Sleep and Wake Disorders. Inclusion criteria were taken from the International Classification of Sleep Disorders,27 according to which the insomnia complaint should (a) have a duration of more than 6 months, (b) occur 3 nights or more per week, and (c) be confirmed by polysomnogra-phy (sleep efficiency less than 85%; sleep latency more than 30 minutes and/or wake time during sleep more than 45 minutes).

Since circadian rhythm problems are not part of the psychophysi-ologic insomnia disorder, shift workers were not included in the study. None of the insomniacs had evidence of psychiatric disease, as indicated by the Symptom Checklist-90.28 To further ensure that the participants were properly classified as having psychophysi-ologic insomnia, 2 subjects (1 man, one woman) of the original 13 subjects were excluded from data analysis due to daily intake of an antidepressant medication. Three subjects of the remain-ing 11 patients occasionally took central nervous system-active medication, as prescribed by their physician: a benzodiazepine (1 subject, on average once a week) and an antidepressant (2 sub-jects, on average 3 times a week). Medication was taken primarily to improve sleep.29 During the laboratory component of the study, medication intake was not allowed. The control group of this study consisted of 13 healthy vol-unteers (7 men, 6 women; mean age ± SD: 44.9 ± 7.7; range 33-53 years) who were recruited through advertisements. History of medical or psychiatric illness and sleep times were obtained in a standardized telephone interview and questionnaires. No psychi-atric or medical illnesses in recent years were reported.25 All con-trol subjects were free from medication and were excluded if they reported excessive consumption of alcohol or other drugs. The self-reported sleep times of the control subjects were between 10:00 PM and 1:00 AM (in bed) and 6:00 AM and 9:00 AM (out of bed), indicating a normal sleep range. The control subjects were individually age- and sex-matched to the insomniacs. After thorough explanation of the experiment, subjects gave consent to participate. This work was approved by the Medical Ethics Committee of the MCH Westeinde hospital in The Hague, The Netherlands.

Experimental Procedures

A constant-routine study25 was conducted in the Sleep Re-search Laboratory of the University of Leiden to record 24-hour physiology and performance, controlling for the masking effects of sleep, changes in light exposure, postural changes, and physi-cal activity. Subjects remained awake in a semirecumbent posi-tion for a period of 26 hours under constant supervision (the first 2 hours served to acclimatize). Lighting conditions (30-50 lux, as measured in the horizontal angle of gaze) and ambient tem-perature (18°C) were kept constant throughout the laboratory ex-periment. Food intake was standardized by means of isocaloric snacks (100-120 kcal) provided every hour. Before each snack, subjective sleepiness and subjective fatigue were assessed. The experiment started at 9:00 AM; recordings started at 11:00 AM after the 2-hour acclimatization period. Throughout the study, subjects were engaged in nonstrenuous activities in bed. Cogni-tive performance tests were administered every hour; details of the test battery are presented elsewhere.30 Between test bouts, par-ticipants were allowed to read, play board games, study, etc., but no vigorous activities were allowed. Subjects were instructed to wear a wrist actigraph and keep a sleep diary during the 3 days prior to the experimental period. In this preexperimental period, subjects slept at home. A brief over-view of the entire study protocol is provided in Table 1.

Measurements

Core body temperature was used as a biologic marker of

1589 Physiological Indexes in Chronic Insomnia—Varkevisser et al

SLEEP, Vol. 28, No. 12, 2005

circadian rhythmicity. It was measured with a rectal thermistor (ASKEY, Leiden, The Netherlands) and recorded at 6-minute in-tervals on an Escort Junior Data Logger (EDL Systems Ltd., Syd-ney, Australia). Outliers in the body-temperature data (defined as acute temperature changes that exceeded 0.3°C) were marked as missing values. Due to equipment failure, the temperature data from 1 subject in the insomnia group and 2 subjects in the control group could not be used for analysis. Temperature values were averaged per hour prior to analysis. Cardiac activity was continuously recorded with the VU-AMS v4.6 ambulatory system (Free University, Amsterdam, The Neth-erlands), and preprocessed with dedicated software. The VU-AMS device used six Ag/AgCl electrodes to record both electrocardiog-raphy and impedance cardiography. Details on electrode place-ment, R-peak detection, and thoracic impedance (dZ) assessed by this device can be found in De Geus et al.31 The sampling rate for the interbeat-interval (IBI) time series was 1000 Hz. At 3-hour intervals, cardiovascular data were extracted from 30-minute periods preceding performance testing. Artifacts in these data were removed by rejecting outlier IBI values (<300 milliseconds or >1800 milliseconds). From the R-peak time se-ries, heart rate average (HRA) was analyzed for each participant in 30-second segments. The root mean of the squared successive differences (rMSSD) of the IBIs was computed as an estimate of the short-term compo-nent of heart rate variability, which is an index of vagal tone.32,33 The rMSSD for consecutive beats in each 30-second segment was calculated as: rMSSD = √(1/n Σi (IBIi - IBIi-1)

2). Values exceeding 200 milliseconds were considered outliers and discarded. The preejection period (PEP), an index of sympathetic cardi-ac regulation,34 was assessed with the following procedure. The maximum velocity of ejection (dZ/dt), obtained from the thoracic impedance data and sampled at 250 Hz around each R-wave, was ensemble-averaged over each 30-second segment. The onset of the dZ/dt upstroke (opening of the aortic valves) was manually determined for each ensemble-averaged complex, and the PEP values were determined by adding a fixed Q-wave-to-R-wave in-terval of 48 milliseconds to the R-B interval time.31 From each selected 30-minute period, one 5-minute continuous and artifact-free block was extracted. The mean HRA, rMSSD,

and PEP were averaged per subject over the 30-second segments in each 5-minute block. To minimize invasive procedures during the experiment, saliva samples were used to measure free cortisol. Saliva was collected every 3 hours, prior to performance testing, in noncoated, absor-bent swabs (Salivettes™; Sarstedt, Nümbrecht, Germany) placed under the tongue for 2 minutes. After centrifugation (4 minutes at 2500 rpm) the saliva was stored at -20°C until analysis. The concentration of free cortisol was determined by enzyme-linked immunosorbent assay (DSL, Etten-Leur, the Netherlands). Sensi-tivity of the assay was 1 ng/mL. During the 3 preexperimental days, general motor activity was assessed with a wrist actigraph (Actiwatch; Cambridge Neuro-technology Ltd., UK). The activity data were analyzed with dedi-cated Actiwatch Sleep Analysis software (version 1.06). Habitual sleep times (total sleep time, time in bed, intermittent wake time), sleep efficiency, and sleep quality were calculated (for details see Mills and Waterhouse25). Finally, the Dutch versions of the Symptom Check List28 and Checklist Individual Strength were used in this study.35 The Symp-tom Checklist-90 was applied to evaluate possible psychiatric co-morbidity, while the Checklist Individual Strength questionnaire was used to determine the coping strength of the insomniacs. The subscales of the Checklist Individual Strength questionnaire (fa-tigue, concentration, motivation, physical activity) are known to be sensitive to the impact of insomnia.6,24

Analyses Descriptive statistics were performed for actigraphy measures and questionnaire scores. Group comparisons for these data were done by means of Wilcoxon signed ranks tests. Primary analyses of the physiologic indexes of arousal were performed using SAS version 8.0 (SAS Institute Inc., Cary, NC). Body temperature data, cardiovascular parameters, and free cor-tisol measurements were subjected to mixed-model analysis of variance (ANOVA),36 which is a statistically more powerful anal-ysis procedure for time series than conventional repeated-mea-sures ANOVA (because it accounts for systematic interindividual variability). In mixed-model ANOVA, both fixed effects (reflect-ing changes over time) and random effects (reflecting interindi-vidual variability around the fixed effects) are involved. Mixed-model ANOVA is robust to missing values, so there was no need to replace missing data points (no more than 6% of the data were missing). For the HRA, PEP, rMSSD, and cortisol variables, which were determined at 3-hour intervals across the 24-hour recording pe-riod, the ANOVA model consisted of a fixed Time of Day effect (9 levels), a fixed Group effect (insomniacs vs controls), a fixed interaction term (Time of Day by Group), and an intercept with a random effect. The Time of Day and interaction effects were tested with an F test and the Group term effects were tested with a t test. For body temperature, which was assessed at 1-hour inter-vals across the 24 hours, the model had 24 levels for Time of Day but was otherwise the same. In the case of significant Group or in-teraction effects, planned contrasts (t tests; 1-tailed) were imple-mented to detect differences between the groups on the specific levels of Time of Day. The magnitude of systematic interindivid-ual differences was quantified by means of the between-subjects

1590 Physiological Indexes in Chronic Insomnia—Varkevisser et al

Table 1—Overview of the Protocol

Stage Event Time of TestingScreening Ambulatory 2 x 24 h polysomnography (insomnia patients only) Preexperimental Baseline questionnaires Sometime during the 3 days(3 days) Wrist activity Continuous Sleep diary Upon awakening each dayExperimental Laboratory adaptation 9:00 AM (2 h)(26 h) Temperature Continuous Cardiac activity Continuous (analyzed for every 3 h starting at 11:00 AM) Cortisol (saliva) Every 3 h starting at 11:00 AM Cognitive testing Every 3 h starting at (not reported here) 11:00 AM Questionnaires Every hour starting at (not reported here) 11:00 AM

SLEEP, Vol. 28, No. 12, 2005

standard deviation (i.e., the square root of the between-subjects variance) for the random effect on the intercept. Contingent of the presence of a Time of Day main effect or interaction, 24-hour sinusoidal curves were fitted to the data by mixed-model harmonic regression36,37 to estimate the contribution of circadian rhythmicity in the physiologic indexes. This analy-sis considered the data from all subjects at once while explicitly considering group differences (fixed effect) and systematic inter-individual differences (random effect). Planned contrasts (t tests) were implemented for overall level, circadian amplitude, and cir-cadian phase to test differences between the 2 groups. Finally, multivariate ANOVA (MANOVA) was carried out in SPSS version 9.0 (SPSS Inc., Chicago, IL) to assess overall dif-ferences in cardiac activity between insomniacs and controls dur-ing the daytime (11:00 AM-8:00 PM), during the nighttime (11:00 PM-8:00 AM), and during the whole 24-hour period. For this analy-sis, HRA, rMSSD and PEP were averaged for each time interval, so that 1 value per parameter remained for each subject. Homo-geneity was tested by means of Box test of equality. Wilk’s λ was used to determine statistical significance of the omnibus test.

Power Calculations

Based on the results of comparing the insomniacs with the con-trol group in terms of performance and fatigue in this study, as reported earlier,25 we performed power calculations to assess the probability that any differences in physiologic variables might be overlooked due to a lack of statistical power. Effect sizes were calculated for the overall level results from mixed-model har-monic regression. Effect sizes for performance variables varied from 0.966 (motor control accuracy) to 1.431 (working memory response latency); for subjective fatigue, the effect size was 2.069. Assuming a type I error threshold of 0.05 and using 2-sided t tests to compare groups, statistical power for the present investigation was calculated, using nQuery.38

RESULTS

Table 2 displays the sleep and personality characteristics of the

2 groups. Sleep efficiency showed lower values for the insomnia group, and almost all personality scores differed between groups as well. More details on the characteristics of sleep, subjective well-being, and performance for both groups have been published elsewhere.25 Figures 1 to 3 illustrate the 24-hour patterns of body tempera-ture, cardiac activity, and cortisol, respectively. Higher values of rMSSD represent more vagal activity (Figure 2b), and lower val-ues of PEP represent more sympathetic activity (Figure 2c). Table 3 shows the results of the mixed-model ANOVAs, and Table 4 shows the results of the mixed-model harmonic regressions. Vi-sual inspection of the figures suggested that, with the exception of core body temperature, all physiologic parameters in the in-somnia group were changed in the direction of elevated arousal, but the differences with the control group were not statistically significant.

Temperature

Figure 1 shows the temperature data for the insomniacs and the controls. Mixed-model ANOVA revealed a significant Time of Day effect in the temperature data. No significant Group effect or interaction was found (Table 3). There was circadian rhyth-

1591 Physiological Indexes in Chronic Insomnia—Varkevisser et al

Table 2—Descriptive Variables

Insomniacs Controls P value*Actigraphy Time in bed, min 454 (50) 478 (50) NS Total sleep time, min 339 (84) 393 (57) 0.093 Sleep onset latency, min 34 (17) 19 (10) NS Mean wake time, min 48 (37) 43 (28) NS Sleep efficiency, % 73.3 (3.6) 82.3 (6.7) 0.036Symptom Check List Anxiety 16.6 (6.5) 11.3 (1.9) 0.017 Depression 26.8 (8.1) 20.2 (4.1) 0.028 Psychoneuroticism 145.8 (23.0) 109.4 (12.0) 0.008Checklist Individual Strength Subjective fatigue 40.4 (9.6) 12.1 (6.1) 0.005 Reduction in concentration 19.6 (6.3) 11.5 (9.0) 0.074 Reduction in motivation 14.6 (6.6) 6.9 (3.5) 0.025

Data are presented as mean ± SD *P values are derived from Wilcoxon signed ranks tests. NS denotes nonsignificance (P > .10)

Figure 1a—Group mean core body temperature profiles (top panel) and group mean difference (bottom panel). In the top panel, the sol-id curve (open symbols) represents the insomniacs; the dotted curve (closed symbols) represents the controls. The curve in the bottom panel represents the difference between the 2 curves in the top panel; the error bars denote the standard errors of the differences between the 2 groups as derived from mixed-model analysis of variance.

SLEEP, Vol. 28, No. 12, 2005

micity in the temperature data of both groups, as indicated by the significance of the amplitude parameters in the mixed-model har-monic regression (Table 4). The 2 groups were not significantly different in circadian amplitude or phase. Minimum temperature in both groups occurred at around 4:30 AM. There was no signifi-cant difference in overall level of the circadian pattern between the 2 groups, which is in agreement with the absence of a Group effect in the mixed-model ANOVA.

Cardiac Activity

Average Heart Rate

Mixed-model ANOVA indicated a Time of Day effect in the HRA data (Figure 2a). No statistical significance was reached in Group or interaction effects (Table 3). Mixed-model harmonic re-gression revealed that the amplitude parameters within the groups reached significance, indicating circadian rhythmicity (Table 4). The 2 groups did not differ significantly with respect to overall level, amplitude, or phase.

Root Mean of the Squared Successive Differences

Figure 2b indicates peak values during the nocturnal period in the rMSSD patterns of both groups. This is corroborated by the results of mixed-model ANOVA, which revealed a Time of Day effect in the data of the 2 groups. Neither the Group effect nor the Time of Day by Group interaction reached statistical signifi-cance (Table 3). A circadian rhythm was found in the HRA data of both groups, as indicated by the significant amplitude parameters observed in the mixed-model harmonic regression (Table 4). No differences were detected in overall level, amplitude, or phase between the groups.

Preejection Period

The results of mixed-model ANOVA showed no significant effects in the PEP data (Figure 2c) for Time of Day, Group, or interaction (Table 3). Because of the absence of a Time of Day

1592 Physiological Indexes in Chronic Insomnia—Varkevisser et al

Table 3—Interactions of Time of Day and Group on Temperature, Heart Rate, and Cortisol Levels

Time of Day Group Interaction F(df1,df2) P t(df) P F(df1,df2) PTemperature,°C 19.96(23,429) .00 0.46 (429) .64 0.62(23,429) .92 HRA, bpm 7.67(8,162) .00 1.18 (162) .24 0.68(8,162) .71rMSSD, ms 2.76(8,162) .01 0.91 (162) .37 1.29(8,162) .25PEP, ms 0.78(8,162) .62 0.56 (162) .58 0.79 (8,162) .61Cortisol ng/mL* 7.27(8,167) .00 0.86 (167) .39 0.50(8,167) .86

The F columns show the F-values for the effect of Time of Day and for the interaction with Group, respectively, with degrees of freedom (df1, df2). The t column shows the t values for the effect of Group, with degrees of freedom (df). The P columns show the P values for each test. Mixed-model analysis of variance results. HRA refers to heart rate average; rMSSD, root mean of the squared successive differences; PEP, preejection period.*Degrees of freedom differ from those for cardiac parameters due to a small difference in the number of missing data points (no more than 6%).

Figure 2—Group mean profiles of the cardiac variables (top panels) and group mean differences (bottom panels). Plotted are (a) heart rate average (HRA); (b) root mean of the squared successive differences (rMSSD); and (c) preejection period (PEP). In the top panels, the solid curves (open sym-bols) represent the insomniacs; the dotted curves (closed symbols) represent the controls. The curves in each of the bottom panels represent the differ-ences between the 2 curves in the corresponding top panels; the error bars denote standard errors of the differences between the 2 groups as derived from mixed-model analysis of variance.

SLEEP, Vol. 28, No. 12, 2005

main effect and interaction, no mixed-model harmonic regression was performed.

Overall Cardiac Activity

The results of MANOVA showed no significant differences in overall cardiac activity (composite of HRA, rMSSD, PEP) be-tween the 2 groups for the full 24-hour interval (Wilk’s λ = 0.92; F3,20 = 0.60; P = .62), for the daytime (Wilk’s λ = 0.91; F3,20 = 0.64; P = .60), and for the nighttime (Wilk’s λ = 0.95; F3,20 = 0.37; P = .77). Homogeneity of variance was met in each case (P > .05).

Cortisol

Although visual inspection (Figure 3) suggested an elevation in cortisol values of about 15% to 20% in the insomnia group, mixed-model ANOVA showed no significant differences between the cortisol patterns of the 2 groups (Table 3). A distinct Time of Day effect was found. Mixed-model harmonic regression re-vealed a significant circadian amplitude in both the insomnia and control groups, with the cortisol maximum in both groups occur-ring at around 9:00 AM (Table 4). No differences were detected between the circadian amplitudes, phases, or overall levels of the 2 groups.

Statistical Power

Power calculations showed that with the present sample size (11 insomniacs and 13 controls), any effect of insomnia in the physiologic variables with a standardized effect size equal to or greater than what we previously observed for subjective fatigue—i.e., the largest standardized effect size observed in Varkevisser and Kerkhof25—would have been detected with more than 99% power (at the conventional type I error threshold of .05). Even if standardized effect sizes in the present data set would have been equivalent to those previously observed for working memory and vigilance performance,25 the present statistical power would have

1593 Physiological Indexes in Chronic Insomnia—Varkevisser et al

Table 4—Mixed-Model Harmonic Regression Results for Overall Level, Circadian Amplitude, and Circadian Phase Insomniacs Controls Difference t P ES R2

Overall Level Temperature, ºC 36.89 ± 0.08 36.94 ± 0.08 0.05 ± 0.12 0.58 .57 0.38 0.80 HRA, bpm 69.5 ± 2.4 65.4 ± 2.2 4.1 ± 3.3 1.22 .24 1.04 0.82 rMSSD, ms 33.9 ± 5.3 40.6 ± 4.9 6.7 ± 7.2 0.92 .37 0.62 0.72 Cortisol, ng/mL 4.78 ± 0.69 3.97 ± 0.64 0.81 ± 0.94 0.86 .40 0.26 0.35Amplitude Temperature, ºC 0.20 ± 0.01* 0.17 ± 0.01* 0.03 ± 0.02 1.38 .18 HRA, bpm 2.3 ± 0.6* 2.2 ± 0.6* 0.1 ± 0.8 0.06 .95 rMSSD, ms 3.1 ± 1.5* 3.4 ± 1.6* 0.3 ± 2.2 0.11 .91 Cortisol, ng/mL 1.54 ± 0.45* 1.52 ± 0.42* 0.02 ± 0.61 0.03 .98Phase Temperature, ºC 04:47 ± 15 04:22 ± 17 25 ± 23 1.09 .29 HRA, bpm 01:31 ± 59 03:05 ± 55 94 ± 81 1.17 .25 rMSSD, ms 23:14 ± 121 04:08 ± 96 293 ± 154 1.91 .07 Cortisol, ng/mL 08:59 ± 70 08:52 ± 66 7 ± 96 0.07 .94

Mixed-model harmonic regression results (estimate ± standard error) for overall level; circadian amplitude; and circadian phase (hh:mm ± min). The t column shows the t test values for the difference between the 2 groups. The P column shows the corresponding P values. The ES column shows the standardized effect size for the overall level difference between groups. The R2 column in the overall level section shows explained variance as an index for goodness of fit. No harmonic regression analyses were performed for the preejection period because of the absence of a significant Time of Day effect in the mixed-model analysis of variance (Table 3). HRA refers to heart rate average; rMSSD, root mean of the squared successive differences.*Statistically significant circadian amplitude within group (P < .05)

Figure 3—Group mean profiles of salivary cortisol concentrations (top panel) and group mean difference (bottom panel). In the top panels, the solid curve (open symbols) represents the insomniacs; the dotted curve (closed symbols) represents the controls. The curve in the bottom panel represents the difference between the 2 curves in the top panel; the error bars denote the standard errors of the differences between the 2 groups as derived from mixed-model analysis of variance.

SLEEP, Vol. 28, No. 12, 2005

been at least 80%. To get an indication of statistical power from independent sources, we looked for quantitative data from previous laboratory studies of insomniacs and controls, as reported in the literature. However, most laboratory studies investigating autonomic car-diac control either did not provide enough quantitative detail to calculate effect sizes or reported nonsignificant results. We found only 1 constant-routine study39 that could be used for an inde-pendent assessment of power in the present study. This study by Lushington and colleagues involved 8 elderly insomniacs who were compared with 8 control subjects. Lushington et al documented a statistically significant differ-ence in the overall level of core body temperature compared with controls (mean ± SEM: 36.84ºC ± 0.06ºC for the insomniacs; 36.66ºC ± 0.06ºC for the controls).39 The difference in the overall temperature level between the 2 groups was 0.18ºC, which was modest relative to the reported standard errors (and accordingly, it is somewhat remarkable that the difference was found to be statistically significant). The temperature difference corresponded to an estimated standardized effect size of 1.06. Power calcula-tions for conventional repeated-measures ANOVA revealed that, with the present sample size, we would have had 69% power to detect such temperature effects with statistical significance. How-ever, since we employed the mixed-model ANOVA approach for time-series analysis, which takes into account systematic interin-dividual differences, our statistical power may have been greater than that.40

These power calculations suggest that the consistent absence of statistically significant differences between insomniacs and con-trols across all measures considered in the present study may not have been due to a mere lack of statistical power.

Systematic Interindividual Differences

Table 5 shows the magnitude of systematic interindividual dif-ferences as compared with the (nonsignificant) Group differences in the mixed-effects ANOVAs for all 5 physiologic measures. It is clear that the interindividual differences were considerably larger than the differences between the 2 groups in all cases.

DISCUSSION

In the published literature, it has remained unclear whether hy-perarousal is a primary contributor to the symptoms experienced by insomniacs, although the balance would appear to be in favor

of such an explanation. We studied physiologic indexes of hyper-arousal, all of which are known to be sensitive to fluctuations in the level of arousal20,32 in subjects with clinically defined chronic psychophysiologic insomnia27 and matched healthy controls. The subjects were exposed to the rigorous experimental control of the constant-routine protocol for 24 hours, during which time they were not allowed to sleep. Under these unmasking circum-stances,25 we found no statistically significant evidence for over-all level differences or circadian rhythm differences between the insomniacs and the normal sleepers. The physiologic indexes of arousal measured under constant-routine conditions were not significantly different from those ob-served in the control group. Even so, there were subtle, nonsignif-icant level differences between the 2 groups for most parameters, which pointed in the direction of relative hyperarousal for the insomniacs. The absence of significant findings could have been due to inadequate power, although power calculations did cast some doubt on this explanation. We used statistically powerful mixed-model analysis techniques, which accounted for system-atic interindividual differences.36,40 We thus observed that any dif-ferences between the 2 groups were overwhelmed by the magni-tude of systematic interindividual differences within the groups. These interindividual differences would make the physiologic measures of arousal, taken by themselves, highly unreliable as a diagnostic tool for insomnia because of a lack of (sensitivity and) specificity.41 Because of this limited specificity, it would appear to be unlikely that general hyperarousal was a critical underlying factor in our sample of chronic insomniacs. Limitations of the present study include the moderate sample size (11 insomniacs and 13 controls), which limited our ability to investigate relationships among the different outcome vari-ables. In addition, prior to enrollment in the study, 3 insomni-acs occasionally took central nervous system-active medication (benzodiazepines and antidepressants) to improve their sleep. Hypothalamic-pituitary-adrenal axis activity, sleep quality, and performance can be affected by such medications.29,42 Although the 3 individuals were within the same physiologic range as the medication-free insomniacs and medication intake was not allowed during the study proper, it cannot be ruled out entirely that these specific insomniacs had some differential effect on the study outcomes. A few previous studies have measured core body temperature in a group of insomniacs under strictly controlled conditions.1,20,39 In these studies, the level of body temperature was higher in the in-somnia group than in the control group. Also, the circadian phase of body temperature was (somewhat) advanced in the insomnia group. These results were not supported by the present outcomes. A possible explanation for this discrepancy may involve the se-lection criteria used in the other studies, which recruited their subjects by means of public announcements. This may have re-sulted in heterogeneous groups of insomniacs ranging from those with mild to severe symptoms and with different etiologies and, possibly, also subjects with sleep-state misperception or circadian rhythm disorders.47 In the present study, the patient group con-sisted exclusively of individuals diagnosed with chronic psycho-physiologic insomnia by the criteria defined in the International Classification of Sleep Disorders.27 With this sample, the earlier evidence for hyperarousal from core body temperature data was not replicated. To our knowledge, this was the first study to measure cardiac

1594 Physiological Indexes in Chronic Insomnia—Varkevisser et al

Table 5—The Magnitude of Systematic Interindividual Differences as Estimated by Mixed-Model Analysis of Variance.

σbs ΔbgTemperature ºC 0.28 0.06 HRA, bpm 8.30 4.06rMSSD, ms 17.98 6.80PEP, ms 17.55 4.03Cortisol, ng/mL 2.16 0.84

σbs denotes the between-subjects SD for interindividual level differ-ences. For comparison, Δbg shows the corresponding overall level difference between the 2 groups. HRA refers to heart rate average; rMSSD, root mean of the squared successive differences; PEP, pre-ejection period.

SLEEP, Vol. 28, No. 12, 2005

autonomic activity in a group of subjects with chronic insomnia under strictly controlled constant-routine conditions with con-tinuous wakefulness. In earlier studies during which sleep was allowed, autonomic cardiac control was shown to differ in in-somniacs compared with good sleepers during sleep, possibly due to hyperarousal.9,10 We originally hypothesized that, in our study, when both groups were not allowed to sleep, sympathetic activity would still be higher and vagal tone lower in the insom-niacs relative to the normal controls. Indeed, vagal tone seemed to show a modest decrease in the insomnia group throughout the experiment. Also, sympathetic control and heart rate appeared to be somewhat elevated. Yet, these differences did not reach statisti-cal significance. Furthermore, the 2 groups responded similarly to the experimental protocol in terms of circadian rhythmicity. There were statistically significant circadian rhythms in HRA and rMS-SD, which were comparable in the insomniacs and the controls and resembled those found in other studies.43,44 No clear circadian variation was present in either group for sympathetic influence, as measured by the PEP, in accordance with results of other studies measuring healthy subjects in a constant-routine setting.45,46 When the cardiac data were averaged over time and combined in a multivariate analysis, neither overall nor daytime- and night-time-specific cardiac output differed significantly between the groups. It should be noted that we used a fixed time interval to define nighttime (11:00 PM to 8:00 AM) in this analysis. This was within the individual ranges of self-reported habitual bedtimes, but it is recognized that interindividual differences in the timing of subjective night may have contributed some error variance in this analysis. Even so, there were no convincing indications of hyperarousal in the cardiac measurements of the insomniacs re-gardless of how the data were analyzed. In the insomnia group, there was a minor overall increase of approximately 1 ng/mL for free cortisol measured in saliva. In a study by Vgontzas et al with a similar sample size,3 a comparable increase was found in plasma cortisol. Considering the range of circadian variation (circa 6 ng/mL), the elevation of about 1 ng/mL in the insomnia group would not seem to be of physiologic importance. Yet, in the study by Vgontzas et al, the difference with control subjects was statistically significant (i.e., in the eve-ning and in 24-hour pulses). In the present study, there was no statistically significant evidence either for overall 24-hour hy-persecretion or specifically for a nocturnal or diurnal elevation of cortisol. The lack of statistically significant differences in our study could be due to the fact that cortisol data are noisier when measured in saliva than when measured in plasma. On the other hand, the differences between the outcomes of the 2 studies could also be explained by differences in experimental design. In the study by Vgontzas et al,3 subjects were allowed to sleep. The hypersecretion of cortisol around the sleep period in the insomnia group studied by Vgontzas et al3 could have been due to an anticipatory reaction to sleep.48 Indeed, insomniacs are characterized by worrying about sleep, especially during the pre-sleep period.8,47 Conditioned arousal to the bedroom and perfor-mance anxiety (i.e., trying too hard to sleep) are typical features in insomnia.27 Not allowing subjects to sleep, as in the present study, may have prevented the subjects from worrying about sleep, which may explain the absence of cortisol hypersecretion. In addition, it is possible that the insomniacs in the present study were more relaxed, even sleepy, because they were in a standard-ized, monotonous bedroom, whereas, in the Vgontzas study,3 the

subjects were ambulatory during the daytime. Hence, the subjects in that study were exposed to changing or stimulating environ-ments, which could have had a marked effect on cortisol secretion patterns. In this line of reasoning, while it is possible that hyperarousal was not present at all in our sample of clinical insomniacs, it is also possible that the patients normally did experience arousal but solely as a result of presleep worrying. Thus, insomniacs may experience hypersensitivity to specific sleep-related matters.11 In-somniacs might also react differently to sleep-related instructions (e.g., “try to fall asleep”) than do control subjects. As such, sleep-iness assessment tools like the Multiple Sleep Latency Test49,50 and Multiple Relaxation Test51 would seem to be of limited value in insomniacs—which could explain the inconsistent and poten-tially overinterpreted findings of reduced daytime sleepiness by sleep-latency criteria in these patients.52 If the present findings under strictly controlled unmasking conditions can be replicated, they may have important clinical implications. Based on the study outcomes, we would not expect general relaxation therapy to be an effective treatment in insomni-acs, as we found no compelling evidence of general hyperarousal. Rather, the focus of treatment should perhaps be on the sleep-on-set period (sleep expectations, stimulus control, beliefs, etc.), as has also been suggested by others.53 Future research is needed to explore whether the hypothesized relationship between presleep worrying and insomnia is causal. When causality—or the absence thereof—is established, better inferences can be made with re-spect to treatment alternatives.

ACKNOWLEDGMENTS

The authors are grateful to Dr. Eco de Geus at the Department of Biological Psychology of the Free University in Amsterdam, the Netherlands, for his advice on the processing of the cardiovas-cular data. The Netherlands Organization for Scientific Research (NWO) is gratefully acknowledged for funding this project. The research was conducted while M. Varkevisser was supported by grant 580-02.401 of the Foundation for Behavioural and Educa-tional Sciences of this organization, awarded to G.A. Kerkhof. H.P.A. Van Dongen was supported by NIH grant HL70154 and by the Institute for Experimental Psychiatry Research Foundation.

REFERENCES

1. Benoit O, Aguirre A. Homeostatic and circadian aspects of sleep regulation in young poor sleepers. Neurophysiol Clin 1996;26:40-50.

2. Bonnet MH, Arand DL. Hyperarousal and insomnia. Sleep Med Rev 1997;1:97-108.

3. Vgontzas AN, Bixler EO, Lin H, et al. Chronic insomnia is as-sociated with nyctohemeral activation of the hypothalamic-pitu-itary-adrenal axis: clinical implications. J Clin Endocrinol Metab 2001;86:3787-94.

4. Bonnet MH, Arand DL. Caffeine use as a model of acute and chron-ic insomnia. Sleep 1992;15:526-36.

5. Drake C, Richardson G, Roehrs T, Scofield H, Roth T. Vulner-ability to stress-related sleep disturbance and hyperarousal. Sleep 2004;27:285-91.

6. Morin CM, Rodrigue S, Ivers H. Role of stress, arousal, and coping skills in primary insomnia. Psychosom Med 2003;65:259-67.

7. Hauri PJ. Psychological and psychiatric issues in the etiopatho-genesis of insomnia. Primary Care Companion J Clin Psychiatry 2002;4(suppl 1):17-20.

1595 Physiological Indexes in Chronic Insomnia—Varkevisser et al

SLEEP, Vol. 28, No. 12, 2005

8. Lundh LG, Broman JE. Insomnia as an interaction between sleep-interfering and sleep-interpreting processes. J Psychosom Res 2000;49:299-310.

9. Bonnet MH, Arand DL. Heart rate variability in insomniacs and matched normal sleepers. Psychosom Med 1998;60:610-5.

10. Domitrovich PP, Pigeon WR, Stein PK, Perlis ML. Heart rate vari-ability in patients with insomnia and good sleeper controls. Sleep 2004;27(suppl):A269.

11. Harvey AG. A cognitive model of insomnia. Behav Res Ther 2002;40:869-93.

12. Vgontzas AN, Mastorakos G, Bixler EO, Kales A, Gold PW, Chrou-sos GP. Sleep deprivation effects on the activity of the hypothalam-ic-pituitary-adrenal and growth axis: potential clinical implications. Clin Endocrinol 1999;51:205-15.

13. Rodenbeck A, Hajak G. Neuroendocrine dysregulation in primary insomnia. Rev Neurol 2001;157:S57-61.

14. Rodenbeck A, Huether G, Rüther E, Hajak G. Interactions between evening and nocturnal cortisol secretion and sleep parameters in patients with severe chronic primary insomnia. Neurosci Lett 2002;324:159-63.

15. Perlis ML, Merica H, Smith MT, Giles DE. Beta EEG activity and insomnia. Sleep Med Rev 2001;5:363-74.

16. Freedman RR. EEG power spectra in sleep-onset insomnia. Electro-enceph Clin Neurophysiol 1986;63:408-13.

17. Bonnet MH, Arand DL. 24-hour metabolic rate in insomniacs and matched normal sleepers. Sleep 1995;18:581-8.

18. Nofzinger EA, Buysse DJ, Germain A, Price JC, Miewald JM, Kup-fer DJ. Functional neuroimaging evidence for hyperarousal in in-somnia. Am J Psychiatry 2004;161:2126-9.

19. Adams K, Tomeny M, Oswald I. Physiological and psychologi-cal differences between good and poor sleepers. J Psychiatr Res 1986;20:301-16.

20. Morris M, Lack L, Dawson D. Sleep-onset insomniacs have delayed temperature rhythms. Sleep 1990;13:1-14.

21. Riemann D, Klein T, Rodenbeck A, et al. Nocturnal cortisol and mel-atonin secretion in primary insomnia. Psychiatry Res 2002;113:17-27.

22. Stepanski E, Zorick F, Roehrs T, Roth T. Effects of sleep deprivation on daytime sleepiness in primary insomnia. Sleep 2000;23:215-9.

23. Lichstein KL, Johnson RS. Pupillometric discrimination of insom-niacs. Behav Res Ther 1994;32:123-9.

24. Riedel BW, Lichstein KL. Insomnia and daytime functioning. Sleep Med Rev 2000;4:277-98.

25. Mills JN, Waterhouse JM. The use of constant routines in unmask-ing the endogenous component of human circadian rhythms. Chro-nobiol Int 1984;1:205-16.

26. Varkevisser M, Kerkhof GA. Chronic insomnia and performance in a 24-hour constant routine study. J Sleep Res 2005;14:49-49.

27. The International Classification of Sleep Disorders, Revised: Diag-nostic and Coding Manual. Chicago: American Academy of Sleep Medicine; 2001.

28. Derogatis LR, Rickels K, Rock AF. The SCL-90 and the MMPI: a step in the validation of a new self-report scale. Br J Psychiatry 1976;128:280-9.

29. Rodenbeck A, Cohrs S, Jordan W, Huether G, Rüther E, Hajak G. The sleep-improving effects of doxepin are paralleled by a normal-ized plasma cortisol secretion in primary insomnia. Psychopharma-cology 2003;170:423-8.

30. Varkevisser M, Kerkhof GA. 24-hour assessment of performance on a palmtop computer: validating a self-constructed test-battery. Chronobiol Int 2003;20:109-21.

31. De Geus EJC, Willemsen GHM, Klaver CHAM, van Doornen LJP. Ambulatory measurement of respiratory sinus arrhythmia and respi-ratory rate. Biol Psychol 1995;41:205–27.

32. Malik M, Task Force of the European Society of Cardiology, North American Society of Pacing and Electrophysiology. Heart rate vari-ability: standards of measurement, physiological interpretation, and

clinical use. Eur Heart J 1996;17:354-81.33. Penttilä J, Helminen A, Jarrti T, et al. Time domain, geometrical

and frequency domain analysis of cardiac vagal outflow: effects of various respiratory patterns. Clin Physiol 2001;21:365-76.

34. Berntson GG, Cacioppo JT, Binkley PF, Uchino BN, Quigley KS, Fieldstone A. Autonomic cardiac control III: psychological stress and cardiac response in autonomic space as revealed by pharmaco-logical blockades. Psychophysiology 1994;31:599-608.

35. Vercoulen JHMM, Hommes OR, Swanink CMA, et al. The mea-surement of fatigue in multiple sclerosis: a multidimensional com-parison with, patients with chronic fatigue syndrome and healthy subjects. Arch Neurol 1996;53:642-9.

36. Van Dongen HPA, Olofsen E, Dinges DF, Maislin G. Mixed-model regression analysis and dealing with interindividual differences. In: Johnson ML, Brand L, eds. Numerical computer methods, part E. Methods in Enzymology 384. Amsterdam: Academic Press, 2004: 139-71.

37. Mikulich SK, Zerbe GA, Jones RH, Crowley TJ. Comparing lin-ear and nonlinear mixed model approaches to cosinor analysis. Stat Med 2003;22:3195-211.

38. Elashoff JD. nQuery Advisor Version 4.0 User’s Guide. Los Ange-les; 2000.

39. Lushington K, Dawson D, Lack L. Core body temperature is el-evated during constant wakefulness in elderly poor sleepers. Sleep 2000;23:504-10.

40. Burton P, Gurrin L, Sly P. Extending the simple linear regression model to account for correlated responses: An introduction to gen-eralized estimating equations and multi-level mixed modelling. Stat Med 1998;17:1261-91.

41. Van Dongen HPA, Vitellaro KM, Dinges DF. Individual differences in adult human sleep and wakefulness: Leitmotif for a research agenda. Sleep 2005;28:479-96.

42. Bastien CH, Fortier-Brochu E, Rioux I, LeBlanc M, Daley M, Mo-rin CM. Cognitive performance and sleep quality in the elderly suf-fering from chronic insomnia: relationship between objective and subjective measures. J Psychosom Res 2003;54:39-49.

43. Van Eekelen APJ, Houtveen JH, Kerkhof GA. Circadian variation in cardiac autonomic activity. Eur J Appl Physiol 2004;92:39-46.

44. Kräuchi K, Wirz-Justice A. Circadian rhythm of heat production, heart rate, and skin and core temperature under unmasking condi-tions in men. Am J Physiol 1994;267:R819-29.

45. Van Eekelen APJ, Houtveen JH, Kerkhof GA. Circadian variation in cardiac autonomic activity: reactivity measurements to different types of stressors. Chronobiol Int 2004;21:107-29.

46. Burgess HJ, Trinder J, Kim Y, Luke D. Sleep and circadian influ-ences on cardiac autonomic nervous system activity. Am J Physiol 1997;273:H1761-8.

47. Harvey AG. Pre-sleep cognitive activity in insomnia: a compari-son of sleep-onset insomniacs and good sleepers. Br J Clin Psychol 2000;39:275-86.

48. Shaver JLF, Johnston SK, Lentz MJ, Landis CA. Stress exposure, psychological distress, and physiological stress activation in midlife women with insomnia. Psychosom Med 2002;64:793-802.

49. Chambers MJ, Keller B. Alert insomniacs: are they really sleep de-prived? Clin Psychol Rev 1993;13:649-66.

50. Stepanski E, Zorick F, Roehrs T, Young D, Roth T. Daytime alert-ness in patients with chronic insomnia compared with asymptom-atic control subjects. Sleep 1988;11:54-60.

51. Schneider-Helmert D. Twenty-four-hour sleep-wake function and personality patterns in chronic insomniacs and healthy controls. Sleep 1987;10:452-62.

52. Bonnet MH, Arand DL. Clinical effects of sleep fragmentation ver-sus sleep deprivation. Sleep Med Rev 2003;7:297-310.

53. Morin CM, Espie CA, eds. Insomnia: a Clinical Guide to Assess-ment and Treatment. New York: Kluwer Academic/Plenum Pub-lishers; 2003.

1596 Physiological Indexes in Chronic Insomnia—Varkevisser et al

The complete picture.

Compact.

Comfort.

Compliance.

Big on features, small on size. The ResMed

S8 is the smallest CPAP on the market and

the Mirage Swift is the most comfortable

nasal interface. Put them together for com-

plete comfort and compliance.

ResMed S8and

Mirage Swift

™

™

Global leaders in sleep and respiratory medicine www.resmed.com

800.424.0737

S8swift ad 8.5x11 10/11/05 2:43 PM Page 1