Upload
pennyrobinsonmonash
View
45
Download
3
Embed Size (px)
Citation preview
Medicine, Nursing and Health Sciences
Moderate/severe obesity is associated with recurrence after breast cancer: A “real life” example of survival analysis
Penny Robinson, Robin Bell, Susan DavisBBioMed Sc (Hons), M Biostat, GCAPWomen’s Health Research Program, SPHPM, Monash University
YSC 2015 Conference, 5 – 6 February 2015
Acknowledgements and Funding BUPA study coordinator: Maria La China,
Marijana Lijovic
Pam Fradkin, Jo Bradbury (design of questionnaires)
Study Advisory Group:Dr Jacquie Chirgwin, A/Prof John Collins, Prof Graham Giles, Mr Peter Gregory, Mr Stewart Hart, Miss Suzanne Neil and Mrs Avis McPhee.
No conflict of interest
BUPA Health Foundation Novartis Australia L.E.W. Carty Trust Connie and Craig Kimberley NHMRC (Grant no. 219279 and 490938) Jack and Robert Smorgon Families
Foundation Roy Morgan Research. Victorian Cancer Agency Research
Fellowship (Robin Bell) NHMRC (Prof Davis NHMRC Prinicipal
Research Fellow)
5–6 February 2015Moderate/severe obesity is associated BC recurrence 2
Background There are factors that increase the risk of getting breast cancer (BC). There are also factors that affect the outcome after a diagnosis of BC.
Improvements in treatment Survival after BC is improving.(More women surviving longer after BC.)
– 5 year survival (2006 - 2010) was 89%.– Nodes negative: survival 96.5%– Nodes positive: survival 80.2%.
AIHW 2012. BC in Australia. Cancer series no 71.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 3
http://www.cancer.gov/
Moderate/severe obesity is associated BC recurrence 4
Important terminology
Active disease definition (an “event”):– Recurrence in the same breast.– Metastatic disease (BC spread elsewhere in the body).– Cancer involving the other breast (spread or new BC).– The initial BC (reported in EQ) doesn’t count.
Stage 1 definition:– Small cancer (<2 cm) confined to the breast– No lymph node involvement
Types of surgery:– Lumpectomy: removal of cancerous lump– Mastectomy: removal of entire breast
5–6 February 2015
http://healthsciencedegree.info/stages-of-breast-cancer-1-4/
Research question Does moderate-severe obesity (BMI 30 to <40 kg/m2)
contribute to BC recurrence and mortality?
Population of interest:– Women whose cancer was hormone receptor positive (HR +) and
human epidermal growth factor receptor (HER2) negative.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 5
Fat cells Fat/obese
Insulin resistance
Estrogen
HR+ Estrogen +/-Progesterone
Insulin levels
Recurrence / Death
Endocrine therapy
Body Mass Index (BMI)
Used World Health Organisation (WHO) BMI classification.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 6
BMI classification category BMI range n (%) Comments
Underweight < 18.5 kg/m2 14 (1.2%) Excluded
Normal weight 18.5 to < 25 kg/m2 528 (44.0%) Reference
Overweight 25 to < 30 kg/m2 374 (31.2%)
Moderate-severe obesity 30 to < 40 kg/m2 253 (21.1%)
Morbid obesity ≥ 40 kg/m2 30 (2.5%) Excluded
Methods - Recruitment BUPA Health Foundation Health and Wellbeing after BC study.
Recruited < 1 year after initial invasive breast cancer diagnosis.Recruitment from June 2004 to December 2006.“Enrolment questionnaire” (EQ).
Followed up annually for 5 years.“Follow-up questionnaire” (FQ1 to FQ5)
Recruitment mainly (78%) through the Victorian Cancer Registry (VCR).
Representative of all Victorian women with BC.
Lijovic M et al. BMC Cancer 2008;8:126–30
Cohort design use survival analysis for research question
5–6 February 2015Moderate/severe obesity is associated BC recurrence 7
Methods – Data collected Enrolment questionnaire (EQ)
– Demographics (including age, BMI)– Initial & subsequent treatment
• Surgery• Radiotherapy• Chemotherapy• Oophorectomy
– Use of OAET
Follow-up questionnaires (FQ1 to FQ5)– Any active disease
• Location, Date• Check against change in treatment
Victorian Cancer Registry (VCR)– Oestrogen / Progesterone
receptor status– Tumour diameter– Number of nodes taken,
positive– Stage of disease
– Monthly death checks
5–6 February 2015Moderate/severe obesity is associated BC recurrence 8
Survival analysis is about events
We are interested in two types of events.
1. First report of active disease (recurrence, cancer spread or new BC)
2. Death due to BC.
We want to know:
Woman experience event (active disease / death due to BC) – Y / N
When?
Survival analysis answers this.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 9
Calculating dates of events
Date of BC diagnosis = entry into survival analysis.
Exit dates:
Date of active disease = date of event.
Completed FQ5 (no active disease): Date of FQ5 = date of censoring.
Follow-up questionnaires sent out every 1 year.
95% of women returned questionnaires within 1.2 years of previous one.– “Withdrawals” classified at the 1.2 year mark.
If death within 1.2 years of last completed questionnaire– If death from BC (no recurrence). Date of death = event.– If death not BC. Date of death = date of censoring.
If no questionnaire returned within 1.2 years (and no death in that time) withdrawal classification (active / passive withdrawal, loss to follow-up) Date of last completed questionnaire = date of censoring.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 10
All exit dates one variable for survival analysis set-up
5–6 February 2015Moderate/severe obesity is associated BC recurrence 11
Flow chart describing “events” and censoring
5–6 February 2015Moderate/severe obesity is associated BC recurrence 12
1199 EQ
1139 FQ1
1081 FQ2
1053 FQ3
1009 FQ4
954 FQ5
60 did not complete FQ1
58 did not complete FQ2
28 did not complete FQ3
44 did not complete FQ4
55 did not complete FQ5
1121 No active disease up to FQ1
18 Active disease up to FQ1
Completed EQ
1051 No active disease up to FQ2
30 Active disease up to FQ2
1013 No active disease up to FQ3
40 Active disease up to FQ3
961 No active disease up to FQ4
48 Active disease up to FQ4
898 No active disease up to FQ5
56 Active disease up to FQ5
1121
1051
1013
961
89816
47
16
36
14
24
18
52
1860
6 (33%)12
26 4 (13%)
32 8 (20%)
40 8 (17%)
Died BC, n = 7
Died BC, n = 2
Died BC, n = 4
Died BC, n = 2
Died BC, n = 5
non-BC, n = 2Withdrew n = 51
non-BC, n = 5Withdrew n = 51
non-BC, n = 3Withdrew n = 21
non-BC, n = 7Withdrew n = 35
non-BC, n = 6Withdrew n = 44
Statistical analysis (Stata 12.1) Preliminary descriptive statistics
– Cross-tabs and chi-square tests (categorical variables)– Median & quartiles (continuous variables)
Survival analysis
stset BCdeathrecur_eventdate_19May2014 if(activeEQ==1), origin(time DateDiagnosis) failure(BCdeathrecur_eventtype_19May2014==22 21) scale(365.25)
– Cox proportional hazard ratio modelling• Univariable regression first.• Manual stepwise regression done.• Obesity “forced in” first.• Obesity + Stage 1 had the highest log-likelihood chi-square
value, so that was the next model, etc.– Kaplan-Meier survival curves
• By “BMI category”, adjusting for other factors in the final Cox regression model.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 13
Results – Descriptive statistics 1155 included in analysis.
Age at diagnosis 58.4 ± 11.6 years. (57.49% were 50 – 69 yrs.)
71.08% (n=821) had a lumpectomy; 28.57% (n=330) had a mastectomy.
600 (51.95%) had Stage 1 disease, 517 (44.76%) were beyond Stage 1.
Majority (88.66%, n=1024) received OAET by FQ1.
Many had radiotherapy (75%, n = 867) or chemotherapy (41%, n = 476).
Median follow-up time: 5.6 years.
98 events (event rate 8.48%) – 78 active disease, 20 deaths from BC.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 14
Cox regression (n = 1155) 98 events (event rate 8.48%)
Univariable Cox regression Multivariable Cox Regression
HR (95% CI) p-value HR (95% CI) p-value
Body Mass Index
≥ 18.5 - < 25 (Normal) 1.00 1.00
≥ 25 - < 30 (Overweight) 1.35 (0.83 to 2.18) 0.226 1.00
≥ 30 - < 40 (Obese) 2.05 (1.27 to 3.33) 0.003 1.71 (1.12 to 2.62) 0.014
Mastectomy surgery at EQ 2.80 (1.88 to 4.17) <0.001 3.28 (1.98 to 5.44) <0.001
HRT at diagnosis 0.87 (0.48 to 1.55) 0.632
Beyond Stage 1 at diagnosis 3.30 (2.10 to 5.17) <0.001 2.87 (1.73 to 4.75) <0.001
On OAET by FQ1 0.52 (0.30 to 0.88) 0.015 0.26 (0.14 to 0.46) <0.001
Radiotherapy by FQ1 0.81 (0.52 to 1.26) 0.355 2.12 (1.24 to 3.63) 0.006
Chemotherapy by FQ1 2.34 (1.56 to 3.51) <0.001
5–6 February 2015Moderate/severe obesity is associated BC recurrence 15
Cox regression – Stage 1 only26 events (event rate 4.3%), n = 600
Univariable Cox regression Multivariable Cox Regression
HR (95% CI) p-value HR (95% CI) p-value
Body Mass Index
≥ 18.5 - < 25 (Normal) 1.00 1.00
≥ 25 - < 30 (Overweight) 1.05 (0.37 to 2.94) 0.930 1.00
≥ 30 - < 40 (Obese) 3.28 (1.36 to 7.91) 0.008 3.23 (1.48 to 7.03) 0.003
Mastectomy surgery at EQ 1.94 (0.81 to 4.61) 0.135
HRT at diagnosis 0.72 (0.22 to 2.39) 0.588
On OAET by FQ1 0.41 (0.17 to 0.99) 0.047 0.41 (0.17 to 0.98) 0.046
Radiotherapy by FQ1 0.45 (0.20 to 1.02) 0.055
Chemotherapy by FQ1 1.39 (0.53 to 3.69) 0.506
5–6 February 2015Moderate/severe obesity is associated BC recurrence 16
Final model survival curve, by obesity – adjusted for other factors in the model.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 17
Source: Robinson PJ, Bell RJ, Davis SD. Maturitas 79 (2014) 279 - 286.
Conclusion - Biostatistics
“Real life” survival analysis can be more complicated than examples taught in survival analysis courses.
Survival analysis can be difficult when the total number of deaths/ recurrences/ other undesirable outcomes is small.
– In our study, only 98 (8.48%) experienced an “event”– 864 (74.81%) completed FQ5– Others withdrew/died from non-BC.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 18
Conclusion – Clinical significance
Moderate-severe obesity associated with a higher BC recurrence rate (in HR+, HER2- disease).
This result remained in women who had Stage 1 disease (good prognosis).
Obesity is a modifiable risk factor.
Further research is required to determine if reducing weight after diagnosis improves prognosis.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 19
Questions?
Publication from this work:
Robinson PJ, Bell RJ, Davis SD. Maturitas 79 (2014) 279 - 286.
Contact:
Twitter: @PennyRobaus (me)
@Monash_WHRP (Women’s Health Research Program)
@Monash_SPHPM (Monash Public Health)
5–6 February 2015Moderate/severe obesity is associated BC recurrence 20
Moderate/severe obesity is associated BC recurrence 21
Additional slides for reference/information
5–6 February 2015
Types of withdrawals
Active withdrawal (AW)– Woman (or family member) advised us they were withdrawing
Passive withdrawal (PW)– Did not return a questionnaire, despite follow-up.– Not an active withdrawal,
Lost contact (LC)– Questionnaire “return to sender” (to us).– Unable to contact woman by post, home/work/mobile or
alternative contact.
Administrative error (admin)– Woman inadvertently not sent a questionnaire.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 22
Flow chart – different withdrawals1683 EQ
1588 FQ1
1496 FQ2
1444 FQ3
1377 FQ4
1305 FQ5
56 deaths post FQ5
20Deaths prior to FQ1
32Deaths prior to FQ2
19Deaths prior to FQ3
25Deaths prior to FQ4
25Deaths prior to FQ5
19Active withdrawals
20Active withdrawals
12Active withdrawals
21Active withdrawals
15Active withdrawals
54Passive Withdrawals
38Passive withdrawals
18Passive withdrawals
18Passive withdrawals
27Passive withdrawals
2Lost to follow-up
1Lost to follow-up
3Lost to follow-up
2Lost to follow-up
2Lost to follow-up
1Forced withdrawal
3Forced exclusions
5 6
7 8
0 1
4 0
6 5
Total 220 deaths up to June 2013
1Admin error
1
95 did not complete FQ1
31
92 did not complete FQ2
52 did not complete FQ3
67 did not complete FQ4
72 did not complete FQ5
48
20
29
36
Getting to stset – generating “Event date”gen BCdeathrecur_eventdate_19May2014 = .replace BCdeathrecur_eventdate_19May2014 = datedeath21May2013 if deaths21May2013==1replace BCdeathrecur_eventdate_19May2014 = datedeath21May2013 if deaths21May2013==1 & deathcause21May2013==1
replace BCdeathrecur_eventdate_19May2014 = dateFQ5 if activeFQ5==1
replace BCdeathrecur_eventdate_19May2014 = dateEQ if newAW==1replace BCdeathrecur_eventdate_19May2014 = dateFQ1 if newAW==2replace BCdeathrecur_eventdate_19May2014 = dateFQ2 if newAW==3replace BCdeathrecur_eventdate_19May2014 = dateFQ3 if newAW==4replace BCdeathrecur_eventdate_19May2014 = dateFQ4 if newAW==5
replace BCdeathrecur_eventdate_19May2014 = dateEQ if newPW==1replace BCdeathrecur_eventdate_19May2014 = dateFQ1 if newPW==2replace BCdeathrecur_eventdate_19May2014 = dateFQ2 if newPW==3replace BCdeathrecur_eventdate_19May2014 = dateFQ3 if newPW==4replace BCdeathrecur_eventdate_19May2014 = dateFQ4 if newPW==5
replace BCdeathrecur_eventdate_19May2014 = dateFQ3 if newPW==6replace BCdeathrecur_eventdate_19May2014 = dateFQ4 if newPW==7replace BCdeathrecur_eventdate_19May2014 = dateFQ1 if newPW==8
replace BCdeathrecur_eventdate_19May2014 = dateEQ if Lost_Contact_Maria=="LC 1"replace BCdeathrecur_eventdate_19May2014 = dateFQ1 if Lost_Contact_Maria=="LC 2"replace BCdeathrecur_eventdate_19May2014 = dateFQ2 if Lost_Contact_Maria=="LC 3"replace BCdeathrecur_eventdate_19May2014 = dateFQ3 if Lost_Contact_Maria=="LC 4"replace BCdeathrecur_eventdate_19May2014 = dateFQ4 if Lost_Contact_Maria=="LC 5"
replace BCdeathrecur_eventdate_19May2014 = recurr_date if recurrFQ5==1
5–6 February 2015Moderate/severe obesity is associated BC recurrence 24
Getting to stset – generating “Event type”gen BCdeathrecur_eventtype_19May2014 = .replace BCdeathrecur_eventtype_19May2014 = 1 if deaths21May2013==1 Non-BC deathreplace BCdeathrecur_eventtype_19May2014 = 22 if deaths21May2013==1 & deathcause21May2013==1 BC death
replace BCdeathrecur_eventtype_19May2014 = 2 if activeFQ5==1 Complete FQ5
replace BCdeathrecur_eventtype_19May2014 = 6 if newAW==1 (active withdrawal) AW1replace BCdeathrecur_eventtype_19May2014 = 7 if newAW==2 AW2replace BCdeathrecur_eventtype_19May2014 = 8 if newAW==3 AW3replace BCdeathrecur_eventtype_19May2014 = 9 if newAW==4 AW4replace BCdeathrecur_eventtype_19May2014 = 10 if newAW==5 AW5
replace BCdeathrecur_eventtype_19May2014 = 11 if newPW==1 (passive withdrawal) PW1replace BCdeathrecur_eventtype_19May2014 = 12 if newPW==2 PW2replace BCdeathrecur_eventtype_19May2014 = 13 if newPW==3 PW3replace BCdeathrecur_eventtype_19May2014 = 14 if newPW==4 PW4replace BCdeathrecur_eventtype_19May2014 = 15 if newPW==5 PW5
replace BCdeathrecur_eventtype_19May2014 = 3 if newPW==6 Admin FQ4replace BCdeathrecur_eventtype_19May2014 = 4 if newPW==7 Admin FQ5replace BCdeathrecur_eventtype_19May2014 = 5 if newPW==8 Admin FQ2
replace BCdeathrecur_eventtype_19May2014 = 16 if Lost_Contact_Maria=="LC 1“ (lost contact) LC1replace BCdeathrecur_eventtype_19May2014 = 17 if Lost_Contact_Maria=="LC 2“ LC2replace BCdeathrecur_eventtype_19May2014 = 18 if Lost_Contact_Maria=="LC 3“ LC3replace BCdeathrecur_eventtype_19May2014 = 19 if Lost_Contact_Maria=="LC 4“ LC4replace BCdeathrecur_eventtype_19May2014 = 20 if Lost_Contact_Maria=="LC 5“ LC5
replace BCdeathrecur_eventtype_19May2014 = 21 if recurrFQ5==1 Recurrence
5–6 February 2015Moderate/severe obesity is associated BC recurrence 25
Survival analysis set-upstset BCdeathrecur_eventdate_19May2014 if(activeEQ==1), origin(time DateDiagnosis) failure(BCdeathrecur_eventtype_19May2014==22 21) scale(365.25)
Entry point = Date of diagnosis.
Exit date– Date of event
• Recurrence or BC death
– Date of censoring• Non-BC death• Withdrawal (AW, PW, LC or admin)• Complete FQ5
Scale = 1 year
5–6 February 2015Moderate/severe obesity is associated BC recurrence 26
Descriptive statistics of BC
5–6 February 2015Moderate/severe obesity is associated BC recurrence 27
CharacteristicPrevalence Event rate
n % n %
Body Mass Index ≥ 18.5 to < 25 528 45.71% 34 6.4%
≥ 25 to < 30 374 32.38% 32 8.6%
≥ 30 to < 40 253 21.90% 32 12.7%
Age (years) < 50 289 25.02% 30 10.4%
50 to <70 664 57.49% 54 8.1%
≥70 202 17.49% 14 6.9%
Type of surgery at EQ Mastectomy 330 28.57% 50 15.2%
Lumpectomy 821 71.08% 47 5.7%
HRT at diagnosis Yes 167 14.46% 13 7.8%
No 988 85.54% 85 8.6%
Stage at diagnosis Beyond Stage 1 517 44.76% 71 13.7%
Stage 1 600 51.95% 26 4.3%
Cancer type (51 missing) Lobular 201 17.40% 21 10.5%
Ductal 907 78.53% 73 8.0%
OAET by FQ1 Yes 1024 88.66% 82 8.0%
No 131 11.34% 16 12.2%
Radiotherapy by FQ1 Yes 867 75.06% 71 8.2%
No 288 24.94% 27 9.4%
Chemotherapy by FQ1 Yes 476 41.21% 60 12.6%
No 679 58.79% 38 5.6%
Strengths and Weaknesses of studyStudy strengths
Prospective design, cohort study with high retention rate.
Postal survey – Women from metro & country areas could participate.
Recruitment representative of BC community
Study weaknesses
Used BMI (obese/non-obese) at EQ for follow-up period.
5–6 February 2015Moderate/severe obesity is associated BC recurrence 28