Duration of Adulthood Overweight, Obesity - PowerPoint Presentation

1. Duration of Adulthood Overweight, Obesity, and Cancer Risk – Never too late to lose weight?Melina Arnold, PhDLuohua Jiang, PhDHoda Anton-Culver, PhDUniversity of California, Irvine

2. Overweight and ObesityDefinitionsOverweight: Body Mass Index (BMI) of 25 or higherObesity: Body Mass Index (BMI) of 30 or higherBody Mass Index (BMI): A measure of an adult’s weight in relation to his or her height, calculated by using the adult’s weight in kilograms divided by the square of his or her height in meters.

3. Prevalence* of Self-Reported Obesity Among U.S. Adults by State and Territory, BRFSS, 2011*Prevalence estimates reflect BRFSS methodological changes started in 2011. These estimates should not be compared to prevalence estimates before 2011. 15%–<20% 20%–<25% 25%–<30% 30%–<35% ≥35% CAMTIDNVUTAZNMWYWAORCONENDSDTXOKKSIAMNARMOLAMIINKYILOHTNMSALWIPAWVSCVANCGAFLNYVTMEHIAKPRGUAMNHMARICTNJDEMDDCSource: CDC

4. Prevalence* of Self-Reported Obesity Among U.S. Adults by State and Territory, BRFSS, 2013*Prevalence estimates reflect BRFSS methodological changes started in 2011. These estimates should not be compared to prevalence estimates before 2011.CAMTIDNVUTAZNMWYWAORCONENDSDTXOKKSIAMNARMOLAMIINKYILOHTNMSALWIPAWVSCVANCGAFLNYVTMEHIAKNHMARICTNJDEMDDCPRGUAM 15%–<20% 20%–<25% 25%–<30% 30%–<35% ≥35% Source: CDC

5. Global trends in obesity prevalenceObesity prevalence doubled from 6.4% in 1980 to 12.0% in 2008Overweight prevalence increased from 24.6% in 1980 to 34.4% in 2008Acceleration of increase in most countriesPreviously observed increases in some high-income countries seem to level offIn 2010, more deaths were attributable to overweight and obesity than to malnutritionStevens et al. Pop Health Metric 2012Flegal et al. JAMA 2012Lim et al. Lancet 2012

6. Obesity and major chronic diseasesObesity has been associated with:Cardiovascular disease (mainly heart disease and stroke)DiabetesMusculoskeletal disorders (especially osteoarthritis)Some cancers

7. Obesity and cancerRenehan et al. Lancet 2008Confounding by smoking?

8. Obesity and cancerRenehan et al. Lancet 2008Protective for premenopausal BC?Gallstones?Estrogen?

9. Obesity and cancer: pathways

10. Attributable cases by region(in thousands)Global burden of cancer attributable to high body-mass index in 2012BMI>25 kg/m2 associated with increased risk of cancer. Population attributable fractions (PAFs) derived using relative risks and BMI estimates in adults in 184 countries.Assuming a 10-year lag-period between high BMI and cancer occurrence, numbers of new cancer cases attributable to high BMI estimated. Arnold et al., Lancet Onc 2014

11. More than 1 in 10 cancers attributable to high BMI in developed countriesMalesFemalesArnold et al., Lancet Onc 2014

12. Obesity and cancer: methodological issuesEffect modification by hormone useConfounding by smoking Difference by tumor histology Variation according to tumor stage Choice of body size measureReverse causalityTiming in life of adiposity (dose-response?)Estimation of overweight/obesity duration

13. Obesity during the life course and cancer risk: questionsDoes it matter how long you have been overweight/obese in your life?Does it matter when you are overweight/obese during your life?Does it matter at which point in your life you accumulate overweight/obese years?Does it matter if your weight/BMI has changed a lot during your life?

14. ObjectivesObjective 1: To estimate the cancer risk associated with overweight and obesity duration during life courseObjective 2: To compare different approaches in estimating overweight and obesity duration during life course Objective 3: To assess the preventability of cancer (in particular breast cancer) in terms of maintaining a healthy BMI

15. Methods: study populationWomen’s Health Initiative Observational Study (n=93,676)Exclusion criteriaCancer history at baseline (n=12,827)Incomplete Follow-up information (n=411)Information on BMI on less than 3 occasions (n=3,523)Missing in covariates included in growth model (n=3,002)Final sample for analyses: 73,913 (78.8%)

16. Methods: overweight and obesity assessmentObservational study:Retrospective self-reports of height and weight at ages 18, 35 and 50Measured height and weight at baseline and at 3-yr follow-upSelf-reported weight during follow-up years 4-8

17. Methods: case ascertainmentSites with convincing evidence of a positive relationship with BMIInvasive cancers:Postmenopausal breastColorectumEndometriumGallbladderKidneyLiverPancreasOvaryThyroid

18. Methods: statistical analyses (1)Modeling of BMI across ages using a quadratic growth model with random intercept and random slopeAdjusted in a stepwise manner for ethnicity, education, energy intake, physical activity and smoking statusUsing the predicted age-specific BMI data, calculate:Overweight/obesity duration in yearsWeighted cumulative overweight/obese years (OWY/OBY)

19. prospectiveretrospective

20.

21.

22. Methods: statistical analyses (2)Cox proportional hazard models to quantify the relationship between overweight/obesity duration and cancer riskStep-wise adjustments for important risk factors/confounders3-knot restricted cubic splines to model non-linear relationships of obesity duration with cancer riskA priori interactions by stratifying the analyses by hormone use, hysterectomy, ethnicity, smoking, diabetes, BMI information (retrospective vs. prospective, measured vs. self-reported)

23. Results: descriptive

24. Results (1): OVERWEIGHT/obesity duration and cancer risk

25. Results: life time overweight/obesity

26. Results: Overweight duration and Cancer*Adjusted for age, ethnicity, education, smoking, physical activity, energy intake, parity, hormone use and age at first birthHR* = 1.05 (1.03-1.07) per 10 years increase in overweight durationHR* = 1.07 (1.04-1.10) per 10 years increase in obesity duration

27. *Adjusted for age, ethnicity, education, smoking, physical activity, energy intake, parity, hormone use and age at first birthResults: Overweight duration and CancerHR* = 1.17 (1.12-1.22) per 10 years increase in overweight durationHR* = 1.23 (1.18-1.28) per 10 years increase in obesity duration

28. *Adjusted for age, ethnicity, education, smoking, physical activity and energy intakeResults: Overweight duration and CancerHR* = 1.07 (1.06-1.09) per 10 years increase in overweight durationHR* = 1.10 (1.08-1.12) per 10 years increase in obesity duration

29. Results: Overweight duration and Cancer

30. Results: Overweight duration and Cancer

31. Results: OWY yearsBreast cancerHR* = 1.08 (1.05-1.12) per 100 units increase in cumulative overweight years (OWY)HR* = 1.07 (1.02-1.12) per 100 units increase in cumulative obese years (OBY)*Adjusted for age, ethnicity, education, smoking, physical activity, energy intake, parity, hormone use and age at first birth

32. Results: OWY yearsEndometrial cancerHR* = 1.37 (1.29-1.46) per 100 units increase in cumulative overweight years (OWY)HR* = 1.29 (1.22-1.36) per 100 units increase in cumulative obese years (OBY)*Adjusted for age, ethnicity, education, smoking, physical activity, energy intake, parity, hormone use and age at first birth

33. Results: OWY yearsAll cancersHR* = 1.12 (1.09-1.15) per 100 units increase in cumulative overweight years (OWY)HR* = 1.12 (1.08-1.15) per 100 units increase in cumulative obese years (OBY)*Adjusted for age, ethnicity, education, smoking, physical activity and energy intake

34. Results: overweight duration HORMONE USE*Adjusted for age, ethnicity, education, smoking, physical activity, energy intake, parity and age at first birth

35. Results: overweight duration HORMONE USE*Adjusted for age, ethnicity, education, smoking, physical activity, energy intake, parity and age at first birth

36. Results (2): Comparison of different approaches estimating overweight and obesity during life course

37. Results: Patterns & Rates of Unknown BMI AllOSCTAgeN% unknownN% unknownN% unknown1876,26243.576,2620.90100.03576,46643.376,4660.70100%5077,85842.374,8890.51,24697.8515,98095.61,48695.62,62995.5528,05494.01,68894.94,13392.95311,40891.52,02592.65,70790.25414,78389.03,56390.37,34687.35519,59885.55,06487.29,77883.15624,47281.97,33783.912,11379.15728,53878.89,30781.514,30075.35832,26676.110,94479.016,07372.35934,59574.411,90677.517,30970.16037,86271.912,88175.418,93167.36139,89170.414,11074.220,03065.46241,05869.614,25674.021,03063.76341,93368.914,33273.221,33663.26442,47268.514,78172.821,49962.96542,56268.515,01072.521,41163.16642,88768.215,28772.121,37763.1

38. Methods: statistical analysesModeling of BMI across ages using a quadratic growth model with random intercept and random slopeAdjusted in a stepwise manner for ethnicity, education, energy intake, physical activity and smoking statusUsing the predicted age-specific BMI data, calculate:Overweight/obesity duration in yearsWeighted cumulative overweight/obese years (OWY/OBY)Compare with two simpler methods:LOCF: Last Observation Carry ForwardLI: Linear Interpolation

39. Results: HRs Based on Different MethodsS – LOCF, OS+CT; L – Linear Interpolation, OS+CT; P – Prediction, OS+CT; SO – LOCF, OS; LO – Linear Interpolation, OS; PO – Prediction, OS.

40. Methods: Monte Carlo SimulationsGenerated data based on estimated parameters from the final growth curve model Assumed that BMI was only associated with two baseline covariates: Smoking status ~ Bernoulli, p = 0.1Dietary energy intake ~ N (1626, 708)500 simulated datasets with 1000 pts eachGen BMI and OW duration for each pt firstEvent times conditional on OW duration with a pre-specified HRSet up missing BMI according to different scenariosCalculate OW duration using different methods

41. Results: Simulated vs. Observed BMISimulatedObserved

42. Results: MSPE of Different Imputation MethodsOS+CT participantsOS participants

43. Results: MSPE of Different MethodsEvery 5 YearsEvery 3 Years

44. Results: Performance of Different Methods  % Bias95% CI CoverageMissing Pattern LOCFLIPredLOCFLIPredLOCFLIPredOS+CT, MCARHR          1.1-14.0-15.4-10.30.4870.4460.3970.940.930.95 1.2-10.8-12.9-8.10.4820.4780.4170.960.950.97 1.5-20.1-17.9-10.30.8930.8050.5910.560.610.80 2.0-29.9-25.3-12.02.0631.7560.9200.000.010.61OS, MCAR           1.1-7.8-13.2-10.40.4270.3980.3850.950.930.96 1.2-6.5-10.5-8.20.4320.4280.4160.940.910.93 1.5-12.8-13.1-10.00.6380.6490.5740.870.850.90 2.0-19.2-15.8-11.41.3581.1140.8860.120.300.62Every 5 years, 50% Missing           1.1-9.3-8.8-10.60.4020.3900.3830.960.970.96 1.2-10.4-8.3-8.40.4640.4320.4260.930.930.92 1.5-13.8-10.7-10.20.6700.5840.5680.790.840.88 2.0-19.6-13.3-11.21.3921.0030.8730.140.500.59Every 5 years, 75% Missing           1.1-21.1-20.7-12.40.4370.4260.3970.920.940.96 1.2-22.2-21.1-12.90.6400.5930.4780.800.810.90 1.5-28.9-26.3-15.41.2031.1170.7420.460.570.81 2.0-35.9-32.3-17.72.4552.2271.2760.000.080.49

45. Results: wrap-up Overweight/obesity durationLonger overweight/obesity duration associated with increased risk of all obesity-related cancers, but also breast, colorectal and endometrial cancer Threshold effects for some sites (endometrium after 30 yrs of overweight, kidney after 10 years)Non-linear relationships between overweight/obesity duration and risk for almost all included sitesDifferent methods to estimate overweight durationPrediction method performed similarly or better than the other two methods Its advantage was most pronounced when the true hazard ratio was large and missing rates were high.

46. Strengths & LimitationsStrengthsFirst study assessing this relationshipLarge number of study participants with repeated weight/height measurementsLimitationsHeight/weight information: Self-reports vs. measured Prospective vs. retrospectiveBMI as a measure of overweight/obesity

47. To-do listObjective/project 3:BMI trajectories and risk of breast cancer BMI stability vs. BMI fluctuation during lifeImpact of age at onset of obesityPrevention scenario: how long do you have to maintain a normal BMI in order to reduce your cancer risk?special focus on the impact of stage at diagnosis, estrogen/progesterone receptor status, hormone use, hysterectomy, age at menarche and other reproductive risk factors

48. Merci beaucoup!Marcia Stefanick, Stanford UniversityRoss Prentice, University of WashingtonIsabelle Soerjomataram, IARCUC Irvine, GERIIARCWHIUICC