Abstract
Background
Previously we showed that adulthood body mass index (BMI) trajectories that result in obesity were associated with elevated risks of fatal prostate cancer (PCA). To further explore this relationship, we conducted a study within the NIH-AARP Diet and Health Study.
Methods
Among 153 730 eligible men enrolled in the NIH-AARP cohort from 1995 to 1996 (median follow-up = 15.1 years), we identified 630 fatal PCA cases and 16 896 incident cases. BMI was assessed for ages 18, 35 and 50 and at study entry, enabling examination of latent class-identified BMI trajectories. Hazard ratios (HRs) and 95% confidence intervals (CI) were estimated using Cox proportional hazards regression.
Results
BMI at study entry (mean age = 63, HR = 1.12; 95% CI = 1.01, 1.24, per 5-unit increase) and maximum BMI during adulthood (HR = 1.12; 95% CI = 1.02, 1.24, per 5-unit increase) shared modest associations with increased risk of fatal PCA. Smoking status likely modified the relationship between BMI trajectories and fatal PCA (Pinteraction = 0.035 via change-in-estimate variable section, P = 0.065 via full a priori model). Among never-smokers, BMI trajectory of normal weight to obesity was associated with increased risk of fatal disease (HR = 2.37; 95% CI = 1.38, 4.09), compared with the maintained normal weight trajectory, whereas there was no association among former or current-smokers. Total and non-aggressive PCA exhibited modest inverse associations with BMI at all ages, whereas no association was observed for aggressive PCA.
Conclusions
Increased BMI was positively associated with fatal PCA, especially among never-smokers. Future studies that examine PCA survival will provide additional insight as to whether these associations are the result of biology or confounding.
Keywords: Body mass index, body weight, disease progression, fatal outcome, prostatic neoplasms, weight change
Key Messages
This study provides evidence that mid-to-late adulthood BMI and maximum BMI during adulthood are associated with increased risks of fatal PCA.
The relationship between obesity across adulthood and fatal PCA was complex. There was no overall association; however among never-smokers, BMI trajectories during adulthood that resulted in obesity were associated with a 2-fold greater risk of fatal PCA compared with men who maintained a stable normal BMI. Our findings here that suggest effect modification by smoking are an important step in helping to clarify these complex relationships between obesity and fatal PCA risk.
Future studies are needed for a better aetiological and mechanistic understanding of how BMI affects fatal PCA and PCA survival, particularly with regards to interactions with smoking.
Introduction
In the USA, 161 360 men will be diagnosed with prostate cancer (PCA) in 2017, and 26 730 men are expected to die from their PCA.1 Meanwhile, two-thirds of the adult male population is overweight or obese.2,3 Thus, understanding the biological mechanisms which link obesity with potentially greater risk of PCA, and death from the disease, has notable clinical and public health implications.4 Studies have found evidence that obesity deregulates multiple hormonal and metabolic pathways,5,6 alters circulating sex hormones concentrations and modifies insulin receptor signalling resulting in chronic inflammation,7,8 each of which has been implicated in prostate carcinogenesis and progression.9–13
In a previous study of the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO), we provided evidence that BMI trajectories during adulthood that resulted in obesity were associated with an approximate doubling in the risk of fatal PCA.14 Fatal PCA is quickly becoming the ‘gold standard’ endpoint in PCA studies, as it represents a subset of truly aggressive (‘clinically-relevant’) PCA that is essential for aetiological studies in an era in which a majority of PCA cases are indolent.15 In our earlier study, we also investigated effect measure modification (EMM), but we had limited statistical power due to accrual of—a respectable but modest—255 fatal PCA cases.14 Meanwhile, evidence has accumulated for strong interactions between BMI and cigarette smoking in relation to cancer and disease risks.16,17 Therefore, to further examine the relationship between BMI trajectories during adulthood and fatal PCA, including assessment of EMM, we conducted study within the prospective Naional Institutes of Health-American Association of Retired Persons (NIH-AARP) Diet and Health Study which has accrued among the highest number of fatal PCA cases of any epidemiological cohort.
Methods
Study population
This prospective cohort is based in the NIH-AARP Diet and Health Study which recruited individuals from 1995 to 1996 in two metropolitan areas and six states, and has been described in detail previously.18 Briefly, eligible individuals were AARP members, aged 50–71 years at the time of study recruitment, who completed a baseline questionnaire. Further information on physical activity and anthropometric history during adulthood was collected 6 months later via a risk factor questionnaire (RFQ).
Of the 334 905 participantwhocompleted the risk factor questionnaire, we excluded: females (n = 136 407); men with a proxy respondent (n = 10 383); men diagnosed with cancer before RFQ (9634); and men with cancer reported only by death certificate (n = 1580). We further excluded men: diagnosed with carcinoma in situ (n = 14); diagnosed with PCA as their second or later cancer (n = 1419); without any follow-up (n = 14); missing BMI at any age (n = 12 703); with an out of range BMI (<15 or >60 kg/m2, n = 2696) or height (<1.52 or >1.98 meters, n = 1279); and with implausible energy intakes (<0.5th percentile or >99.5th percentile, n = 5046). Our final analytical cohort consisted of 153 730 individuals.
The study protocol was approved by the Institutional Review Board (IRB) of the National Cancer Institute.
Exposure ascertainment
Participants provided their current height and body weight at study entry, and anthropometric history (weight at ages 18, 35 and 50 years, and maximum weight) at RFQ. To examine excess adiposity, BMI (kg/m2) was assessed at each specific age, and was modelled as a continuous and categorical metric [underweight (<18.5), normal weight (18.5–24.9), overweight (25.0–29.9), obese (≥30.0)].19 Maximum BMI among the time points provided was also assessed, as there is evidence that it circumvents the problem of reverse causality.20,21
To examine BMI trajectories, we used latent-class fixed-effect trajectory models to define longitudinal patterns of BMI change using the four timepoints available during adulthood.22 We used previously published criteria to select the model with optimum fit,14 which included the Bayesian Information Criterion, the mean posterior probability of each trajectory greater than 70%, and trajectory class membership of at least 1% of available individuals.23,24
Outcome ascertainment
Our primary outcome of interest was fatal PCA, defined as PCA being the underlying cause of death. Determination of cause of death were obtained via linkage to the National Death Index Plus (>95% of the cohort was available for linkage).25 We also conducted analyses of PCA incidence, with follow-up conducted through linkage to state cancer registries, and vital status ascertained by annual linkage to the Social Security Administration Death Master File. The case definition for these analyses was men diagnosed with histologically verified malignant PCA [defined as International Classification of Disease for Oncology, 3rd edition (ICD-O-3) code C61) diagnosed as their first incident cancer. Aggressive PCA was defined as biopsy Gleason score >7, clinical stage (AJCC, TNM Group) ≥III, or PCA being the underlying cause of death (a proxy of true aggressiveness, using the date of PCA incidence for these analyses); remaining cases were classified as non-aggressive.
Statistical analysis
Cox proportional hazards regression analysis,26 with age in months as the time metric and hazards stratified by 5-year birth cohorts,27 was used to calculate adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for associations of age-specific and longitudinal BMI measures with PCA risks. For analyses of fatal PCA, follow-up began at RFQ and ended at date of PCA death or right-censoring (loss to follow-up, non-PCA death or end of follow-up on 31 December 2011), whichever occurred first. A similar analytical strategy was used for secondary analyses of incident PCA, but the event date was date of PCA diagnosis and subjects were also right-censored for at the time of other primary cancer diagnosis (excluding non-melanoma skin cancer). We tested the proportional hazards assumption by examining interactions between BMI and log (time) using the Wald test.
Potential confounders were chosen a priori based on literature review, and variables remained in the adjusted model if variable elimination changed the log HR by greater than or equal to 10% for any given outcome.28 Variables which met these criteria were the following: age; race (White, Black, other); smoking status (never, former, current); family history of PCA; diabetes; myocardial infarction; and state of residence. All analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, NC).
We tested for potential effect modifiers chosen a priori, including smoking status (never, former or current), diabetes and race, using likelihood ratio tests that compared regression models with and without a multiplicative term.29 Given the non-independence of exposures and non-independence of outcomes, adjustment for multiple comparisons was not made. In sensitivity analyses, we examined three distinct, but related, definitions of aggressive PCA, and whether exclusion of PCAs diagnosed in the first 2 years of follow-up altered risk estimates. We also assessed if the additional inclusion of physical activity, marital status, education and total energy intake materially affected the results, and whether weight change during adulthood (adjusted for height) exhibited similar associations with PCA risks. Last, we forced the latent-class trajectory models into longitudinal patterns of BMI change that were found for men in the PLCO, to assess if the PLCO trajectory classes of BMI14 and associations with PCA observed in the PLCO exhibited similar associations among men in this study.
Results
Among the 153 730 men included in the analytical cohort, the mean age was 63.0 years at study entry. We identified 630 fatal PCA cases and 16 896 incident PCA cases (2185 aggressive) during a median follow-up of 15.1 years. The cohort was predominately White (94%), married (86%) and college educated (49%), and there was a high prevalence of current/former smokers (67%) (Table 1). Compared with men who did not develop PCA, fatal and non-fatal PCA cases were more likely to be: older; Black; have a family history of PCA; and have no history of diabetes (Supplementary Table 1, available as Supplementary data at IJE online).
Table 1.
Baseline characteristics of male participants in the NIH-AARP Diet and Health Cohort by BMI category at study entry
| Body mass index (BMI) at study entry |
|||||
|---|---|---|---|---|---|
| All men | Underweight (<18.5 kg/m2) | Normal weight (18.5-24.9 kg/m2) | Overweight (25.-29.9 kg/m2) | Obese (≥30.0 kg/m2) | |
| Characteristic | n = 153 730 | n = 380 | n = 46 336 | n = 76 022 | n = 30 992 |
| Person-years | 1 775 651 | 3857 | 537 109 | 882 050 | 352 635 |
| Age at entry, mean (SD) | 62.95 (5.28) | 64.16 (5.33) | 63.35 (5.28) | 62.99 (5.24) | 62.21 (5.28) |
| Race, n (%) | |||||
| White | 144696 (94.1%) | 343 (90.3%) | 43469 (93.8%) | 71715 (94.3%) | 29169 (94.1%) |
| Black | 3095 (2.0%) | 5 (1.3%) | 714 (1.5%) | 1516 (2.0%) | 860 (2.8%) |
| Othera | 4613 (3.0%) | 24 (6.3%) | 1764 (3.8%) | 2151 (2.8%) | 488 (2.2%) |
| Missing | 1326 (0.9%) | 8 (2.1%) | 389 (0.8%) | 640 (0.8%) | 289 (0.9%) |
| Cigarette smoking, n (%) | |||||
| Never | 46205 (30.1%) | 118 (31.1%) | 16019 (34.6%) | 22042 (29.0%) | 8026 (25.9%) |
| Former | 88498 (57.6%) | 157 (41.3%) | 23529 (50.8%) | 45245 (59.5%) | 19567 (63.1%) |
| Current | 14023 (9.1%) | 99 (26.1%) | 5359 (11.6%) | 6241 (8.2%) | 2324 (7.5%) |
| Missing | 5004 (3.3%) | 6 (1.6%) | 1429 (3.1%) | 2494 (3.3%) | 1075 (3.5%) |
| Family history of | |||||
| prostate cancer, n (%) | |||||
| Yes | 12947 (8.4%) | 25 (6.6%) | 3875 (8.4%) | 6489 (8.5%) | 2558 (8.3%) |
| Diabetes, n (%) | |||||
| Yes | 14514 (9.4%) | 27 (7.1%) | 2620 (5.7%) | 6568 (8.6%) | 5299 (17.1%) |
| Myocardial infarction, n (%) | |||||
| Yes | 27004 (17.6%) | 71 (18.7%) | 7396 (16.0%) | 13220 (17.4%) | 6317 (20.4%) |
| Physical activity ≥20 min (in the past 12 months) | |||||
| 5+ times per week | 34319 (22.3%) | 77 (20.3%) | 13000 (28.1%) | 16631 (21.9%) | 4611 (14.9%) |
| Education, n (%) | |||||
| College graduate | 74771 (48.6%) | 177 (46.6%) | 25556 (55.2%) | 36363 (47.8%) | 12675 (40.9%) |
| Marital status, n (%) | |||||
| Married or cohabiting | 131514 (85.5%) | 280 (73.7%) | 38767 (83.7%) | 65998 (86.8%) | 26469 (85.4%) |
| Total energy (kcal/day), mean (SD) | 2005 (767) | 2026 (827) | 1964 (725) | 1992 (759) | 2099 (836) |
State of residence is not provided.
SD, standard deviation.
Other race includes Asian, Native Hawaiian or other Pacific Islander, American Indian or Alaskan Native.
BMIs at ages 18, 35 and 50 were not associated with risk of fatal PCA, but BMI at study entry was (HR = 1.12; 95% CI = 1.01, 1.24, per 5-unit increase), with the strongest association seen among obese men at study entry compared with normal-weight men (HR = 1.27; 95% CI = 1.01, 1.60) (Table 2). There was some evidence that the association between BMI at study entry interacted with smoking status to affect fatal PCA risk (Pinteraction = 0.06)—among never-smokers, men who were obese at study entry were at an increased risk of fatal PCA (HR = 1.99; 95% CI = 1.28, 3.10) compared with men who were normal weight (Table 3). Maximum BMI during adulthood was positively associated with fatal PCA (HR = 1.12; 95% CI = 1.02, 1.24), particularly among men whose maximum BMI was obese compared with men with a maximum normal BMI (HR = 1.34; 95% CI = 1.06, 1.69)—and this association was magnified among never-smokers.
Table 2.
Adjusted hazard ratios (HR) and 95% confidence intervals (CI) for associations of age-specific and longitudinal measures of body mass index in relation to fatal prostate cancer in the NIH-AARP Diet and Health Cohort
| Mortality |
||||
|---|---|---|---|---|
| Fatal prostate | ||||
| cancer |
||||
| Characteristic | Non-cases | Cases | MRa | HR (95% CI) |
| BMI, age 18, kg/m2 | ||||
| <18.5 | 17060 | 102 | 44.1 | 1.29 (1.04, 1.60) |
| 18.5–25 | 100744 | 458 | 33.4 | Reference |
| 25–30 | 16698 | 62 | 27.8 | 0.93 (0.72, 1.22) |
| 30+ | 2332 | 8 | 26.8 | 1.00 (0.50, 2.01) |
| Continuous, per 5 kg/m2 | 0.95 (0.83, 1.08) | |||
| BMI, age 35, kg/m2 | ||||
| <18.5 | 2560 | 12 | 35.7 | 1.00 (0.57, 1.78) |
| 18.5–25 | 76879 | 379 | 36.1 | Reference |
| 25–30 | 48757 | 208 | 27.4 | 0.98 (0.82, 1.16) |
| 30+ | 8638 | 31 | 19.4 | 1.02 (0.70, 1.48) |
| Continuous, per 5 kg/m2 | 0.94 (0.83, 1.06) | |||
| BMI, age 50, kg/m2 | ||||
| <18.5 | 1265 | 4 | 24.8 | 0.73 (0.27, 1.96) |
| 18.5–25 | 51218 | 241 | 34.5 | Reference |
| 25–30 | 65916 | 317 | 35.2 | 1.13 (0.95, 1.33) |
| 30+ | 18435 | 68 | 28.1 | 1.13 (0.85, 1.49) |
| Continuous, per 5 kg/m2 | 1.06 (0.94, 1.19) | |||
| BMI, at baseline, kg/m2 | ||||
| <18.5 | 348 | 3 | 75.9 | 2.24 (0.71, 7.01) |
| 18.5–25 | 40943 | 176 | 31.6 | Reference |
| 25–30 | 67505 | 325 | 35.2 | 1.17 (0.98, 1.40) |
| 30+ | 28038 | 126 | 33.8 | 1.27 (1.01, 1.60) |
| Continuous, per 5 kg/m2 | 1.12 (1.01, 1.24) | |||
| Maximum BMI, kg/m2, all ages | ||||
| <18.5 | 92 | 0 | – | – |
| 18.5–25 | 34004 | 145 | 31.2 | Reference |
| 25–30 | 70791 | 336 | 34.7 | 1.15 (0.95, 1.40) |
| 30+ | 31947 | 149 | 35.3 | 1.34 (1.06, 1.69) |
| Continuous, per 5 kg/m2 | 1.12 (1.02, 1.24) | |||
| BMI trajectory | ||||
| Stable normal | 52825 | 249 | 34.6 | Reference |
| Normal to overweight | 61345 | 296 | 35.4 | 1.13 (0.95, 1.34) |
| Normal to obese | 12017 | 48 | 29.9 | 1.16 (0.84, 1.58) |
| Stable overweight | 7910 | 32 | 30.7 | 1.11 (0.77, 1.61) |
| Overweight to obese | 2737 | 5 | 14.4 | 0.70 (0.29, 1.70) |
Empty cells (–) are missing HRs and 95% CIs due to 0 case counts for the respective category. All multivariable Cox proportional hazards regression models were conducted with age (month) as the underlying time metric and were adjusted for race, smoking status, history of diabetes, history of heart disease, family history of prostate cancer and state of residence.
Age-standardized mortality rates (MR) per 100 000 person-years.
Table 3.
Adjusted hazard ratios (HR) and 95% confidence intervals (CI) for associations of age-specific and longitudinal measures of body mass index in relation to fatal prostate cancer in the NIH-AARP Diet and Health Cohort by smoking status
| Prostate cancer-specific mortality |
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Never-smokers |
Former smokers |
Current smokers |
||||||||||
| Characteristic | Non-cases | Cases | MRa | HR (95% CI) | Non-cases | Cases | MRa | HR (95% CI) | Non-cases | Cases | MRa | HR (95% CI) |
| BMI, age 18, kg/m2 | ||||||||||||
| <18.5 | 4795 | 23 | 33.9 | 1.30 (0.83, 2.04) | 10110 | 67 | 49.1 | 1.43 (1.10, 1.87) | 1609 | 9 | 45.4 | 0.81 (0.40, 1.63) |
| 18.5–25 | 30032 | 112 | 26.3 | Reference | 58093 | 266 | 33.8 | Reference | 9354 | 61 | 52.9 | Reference |
| 25–30 | 5015 | 22 | 31.6 | 1.36 (0.86, 2.16) | 9510 | 30 | 23.7 | 0.78 (0.54, 1.14) | 1583 | 8 | 41.5 | 0.99 (0.47, 2.06) |
| 30+ | 647 | 1 | 11.7 | 0.56 (0.08, 4.04) | 1340 | 4 | 23.3 | 0.85 (0.32, 2.27) | 258 | 3 | 98.1 | 2.67 (0.83, 8.59) |
| Continuous, per 5 kg/m2 | 1.05 (0.81, 1.37) | 0.89 (0.74, 1.06) | 1.13 (0.79, 1.61) | |||||||||
| BMI, age 35, kg/m2 | ||||||||||||
| <18.5 | 709 | 1 | 10.258 | 0.39 (0.05, 2.81) | 1450 | 8 | 42.4 | 1.20 (0.59, 2.44) | 306 | 3 | 80.1 | 1.37 (0.43, 4.38) |
| 18.5–25 | 22461 | 88 | 27.428 | Reference | 44431 | 222 | 36.8 | Reference | 7551 | 56 | 60.2 | Reference |
| 25–30 | 14613 | 58 | 28.24 | 1.10 (0.79, 1.53) | 28364 | 121 | 31.6 | 0.95 (0.76, 1.19) | 4135 | 18 | 35.3 | 0.71 (0.42, 1.20) |
| 30+ | 2706 | 11 | 30.383 | 1.43 (0.76, 2.71) | 4808 | 16 | 25.8 | 0.91 (0.54, 1.52) | 812 | 4 | 41.7 | 1.09 (0.39, 3.06) |
| Continuous, per 5 kg/m2 | 1.12 (0.89, 1.42) | 0.91 (0.77, 1.08) | 0.72 (0.51, 1.02) | |||||||||
| BMI, age 50, kg/m2 | ||||||||||||
| <18.5 | 349 | 1 | 20.9 | 0.86 (0.12, 6.25) | 682 | 3 | 34.7 | 1.05 (0.33, 3.30) | 176 | 0 | – | – |
| 18.5–25 | 16167 | 55 | 23.8 | Reference | 27930 | 131 | 34.6 | Reference | 5512 | 47 | 70.1 | Reference |
| 25–30 | 18777 | 83 | 31.3 | 1.38 (0.98, 1.95) | 39439 | 194 | 36.2 | 1.16 (0.93, 1.45) | 5564 | 28 | 40.3 | 0.69 (0.43, 1.10) |
| 30+ | 5196 | 19 | 26.9 | 1.50 (0.88, 2.56) | 11002 | 39 | 27.1 | 1.07 (0.74, 1.55) | 1552 | 6 | 31.8 | 0.76 (0.31, 1.82) |
| Continuous, per 5 kg/m2 | 1.10 (0.88, 1.39) | 1.06 (0.91, 1.25) | 0.84 (0.59, 1.20) | |||||||||
| BMI, at baseline, kg/m2 | ||||||||||||
| <18.5 | 106 | 2 | 140.7 | 6.39 (1.54, 26.49) | 141 | 1 | 63.7 | 2.02 (0.28, 14.48) | 95 | 0 | – | – |
| 18.5–25 | 13974 | 44 | 22.1 | Reference | 20820 | 89 | 31.6 | Reference | 4876 | 38 | 64.7 | Reference |
| 25–30 | 19215 | 75 | 27.6 | 1.29 (0.89, 1.87) | 40393 | 202 | 36.7 | 1.23 (0.96, 1.58) | 5679 | 32 | 44.9 | 0.75 (0.47, 1.21) |
| 30+ | 7194 | 37 | 37.5 | 1.99 (1.28, 3.10) | 17699 | 75 | 32.1 | 1.19 (0.87, 1.63) | 2154 | 11 | 41.7 | 0.85 (0.43, 1.69) |
| Continuous, per 5 kg/m2 | 1.27 (1.05, 1.54) | 1.12 (0.98, 1.27) | 0.86 (0.63, 1.18) | |||||||||
| Maximum BMI, kg/m2, all ages | ||||||||||||
| <18.5 | 37 | 0 | – | – | 30 | 0 | – | – | 25 | 0 | – | – |
| 18.5–25 | 11815 | 36 | 21.3 | Reference | 17015 | 72 | 31.1 | Reference | 4147 | 33 | 65.6 | Reference |
| 25–30 | 20369 | 82 | 28.4 | 1.33 (0.90, 1.98) | 42015 | 206 | 36.1 | 1.21 (0.92, 1.58) | 6092 | 33 | 43.5 | 1.72 (0.44, 1.16) |
| 30+ | 8268 | 40 | 35.3 | 1.91 (1.21, 3.02) | 19993 | 89 | 33.8 | 1.28 (0.93, 1.75) | 2540 | 15 | 48.6 | 1.01 (0.54, 1.89) |
| Continuous, per 5 kg/m2 | 1.23 (1.02, 1.49) | 1.11 (0.97, 1.26) | 1.00 (0.74, 1.34) | |||||||||
| BMI trajectory | ||||||||||||
| Stable normal | 16468 | 54 | 22.9 | Reference | 29000 | 141 | 35.9 | Reference | 5695 | 44 | 63.5 | Reference |
| Normal to overweight | 17706 | 76 | 30.3 | 1.41 (0.99, 2.00) | 36395 | 181 | 36.6 | 1.12 (0.90, 1.40) | 5227 | 30 | 46.0 | 0.85 (0.53, 1.35) |
| Normal to obese | 3124 | 18 | 41.4 | 2.37 (1.38, 4.09) | 7558 | 22 | 21.9 | 0.81 (0.51, 1.27) | 916 | 3 | 26.3 | 0.64 (0.19, 2.09) |
| Stable overweight | 2398 | 10 | 30.7 | 1.64 (0.83, 3.24) | 4471 | 19 | 32.2 | 1.09 (0.67, 1.77) | 748 | 3 | 33.7 | 0.82 (0.25, 2.66) |
| Overweight to obese | 793 | 0 | – | – | 1629 | 4 | 19.5 | 0.87 (0.32, 2.37) | 218 | 1 | 37.9 | 1.15 (0.16, 8.46) |
Empty cells (–) are missing HRs and 95% CIs due to 0 case counts for the respective category. All multivariable Cox proportional hazards regression models were conducted with age (month) as the underlying time metric and were adjusted for race, smoking status, history of diabetes, history of heart disease, family history of prostate cancer and state of residence.
Age-standardized mortality rates (MR) per 100 000 person-years.
In the BMI trajectory analysis, the optimal latent-class model was quadratic over-time with five BMI trajectories (stable normal BMI, normal BMI to overweight, normal BMI to obese, stable overweight, and overweight to obese; Figure 1). Overall, no association was seen for risk of fatal PCA, yet smoking status appeared to modify these relationships (Pinteraction = 0.035). Among never-smokers, men with a trajectory of normal weight at age 20 progressing to obesity by study entry were associated with increased risk of fatal disease (HR = 2.37; 95% CI = 1.38, 4.09), compared with men who maintained a normal weight, whereas no association was seen in former (HR = 0.81; 95% CI = 0.51, 1.27) or current-smokers (HR = 0.64; CI = 0.19, 2.09) (Table 3). Age-specific and maximal BMIs and positive BMI trajectories shared modest inverse associations with risk of total and non-aggressive PCA, and no associations were seen for aggressive PCA (Supplementary Table 2, available as Supplementary data at IJE online). None of the putative effect modifiers altered associations between BMI and incident PCA risk, and smoking was the only effect modifier of the relationship between BMI and fatal PCA.
Figure 1.
Prediagnostic body mass index trajectories among men in the NIH-AARP Diet and Health Study. Each trajectory was modeled using quadratic polynomials.
In sensitivity analyses, exclusion of PCAs diagnosed in the first 2 years of follow-up found similar risk estimates, and use of alternative clinical definitions of aggressive PCA did not materially affect the results (data not shown). Similarly, additional adjustment for potential confounders (physical activity, energy intake, education and marital status), which were eliminated based on variable selection criteria, exhibited minimal effects on estimated associations; however, the interaction between smoking status and men with a trajectory of normal weight at age 20 progressing to obesity by study entry (compared with men who maintained a normal weight) in relation to fatal prostate cancer did slightly weaken (Pinteraction = 0.065). (Supplementary Tables 3 and 4, available as Supplementary data at IJE online). Analyses of weight change (adjusted for height) yielded overall null associations in relation to PCA risks, and stratified analyses by smoking status were similar to the main results, although there was no strong evidence of evidence of interaction (P >0.10) (Supplementary Table 5, available as Supplementary data at IJE online). Finally, models which forced AARP men into BMI trajectory classes derived from men in the PLCO study found associations with PCA risks that were similar to the main results (Supplementary Table 6, available as Supplementary data at IJE online). In all Cox models, the proportional hazards assumption was not violated (P >0.05).
Discussion
From this large prospective cohort study, we provide evidence that smoking was an effect modifier of the relation between BMI trajectories and fatal PCA; among never-smokers, BMI trajectories that resulted in obesity more than doubled the risk of fatal PCA compared with never-smokers who maintained a stable weight. BMI at study entry (mean age = 63) and maximum BMI during adulthood were also associated with slightly increased risks of fatal PCA. There were slight inverse associations between BMI measures and risk of total and non-aggressive PCA, whereas no association was seen for aggressive PCA.
The link between obesity and PCA is still poorly understood and evidence remains mixed.30–33 Since the latest International Agency for Research on Cancer (IARC) report concluded that there was limited evidence for an association between excess body fatness and fatal PCA,33 further evidence has been published linking obesity with increased risks of advanced and fatal PCA.14,34 In our recent study in which we used a trajectory approach among men in the prospective PLCO cohort,14 we found strong associations between BMI trajectories which resulted in obesity and fatal PCA, compared with men whose BMI remained stable. Whereas no statistical interaction was found between smoking and BMI trajectories in the PLCO (Pinteraction = 0.14), the 2-fold increased risk of fatal PCA among men who progressed from normal weight to obesity was driven by the never-smokers (HR = 4.35; CI = 2.15, 8.77) compared with current (HR = 1.81; 95% CI = 0.39, 8.44) or former smokers (HR = 1.09; 95% CI = 0.52, 2.25), and this agrees with the results presented in this study of the NIH-AARP cohort. Other earlier studies of BMI and PCA have either been too small (especially for fatal PCA) to detect smoking as an effect modifier,35 or did not evaluate any such.36–41
In the Health Professionals Follow-up Study (HPFS),42 self-reported body shape evolution during the life course was not associated with advanced sStage ≥III) PCA, and there was no evidence of effect modification by smoking (P = 0.18). However, among never-smokers there was tentative evidence that ‘lean-to moderate-increase’ body shape evolution was positively associated with increased risk among never-smokers (HR = 1.40; 95% CI = 0.97, 2.03), but no association was observed for men with a ‘lean to marked-increase’. This HPFS study did not examine body shape evolutions for risk of fatal PCA, but another HPFS study of localized PCA (stage ≤II) cases found that smoking was an effect modifier between long-term weight gain (>13.61 kgs) and progression to lethal PCA (stage ≥III or fatal disease) among never-smokers (HR = 1.59; 95% CI = 1.01, 2.50) when compared with stable weight (±4.54 kgs) men.34 Our study supports and extends these findings by providing evidence that smoking status is an effect modifier of BMI trajectories and risk of fatal PCA.
Potential biological mechanisms that could underlie an interaction between smoking and BMI in relation to cancer risks are complex and elusive, despite the high prevalence of these exposures and the strong effects they confer.16,43 BMI is a proxy for excess adiposity which, in men, typically accumulates in the abdomen and includes highly metabolic visceral adipose fat that has far-reaching systemic effects.44,45 Smoking has a complex relationship with body weight; it generally increases energy expenditure, suppresses appetite, is associated with lighter weight in lighter smokers and heavier weight in heavier smokers and leads to weight gain following cessation.16,43,46 Given the recency of the evidence associating BMI trajectories with fatal PCA, we agree with Samet that further mechanistic studies to help elucidate these complex relationships are required to complement and enhance our epidemiological findings.43
Our results for an increased risk of fatal PCA among men whose maximum BMI was obese (HR = 1.34)—which was heightened among never-smokers (HR = 1.91)—is consistent with the increased risk (HR = 1.59; 95% CI = 1.10, 2.31) for men with a maximum BMI of obese during the same period of adulthood that was seen in the PLCO.14 Maximum BMI has been examined more often in recent studies, as it is hypothesized to minimize reverse causality associated with recent BMIs that may be affected by the disease process. In age-specific results, our study found obesity at study entry (mean age = 63), was associated with increased risks of fatal PCA (HR = 1.27; or HR = 1.12 per 5 kg/m2 increase), which were again heightened among never-smokers (HR = 1.99). A recent meta-analysis of BMI during mid-to-late adulthood and fatal PCA reports a 15% (95% CI = 7, 23%) increase in PCA mortality per 5 kg/m2 increase in BMI.47 However, we are unaware of any meta-analysis stratified by smoking status for risk between obesity and fatal PCA.
The inverse associations between BMI and risks of total and non-aggressive PCA are consistent with previous studies, both for age-specific BMI,30,48 and for longitudinal measures of weight gain and BMI trajectories.14,42,49,50 Potential factors that may explain the inverse relationship between BMI and total and non-aggressive PCA include reduced levels of circulating androgens in obese men, which is thought to drive PCA progression,51 and greater likelihood of false-negative diagnostic testing, given PSA haemodilution and larger prostate volumes.30
The study may have limited generalizability to some groups as the cohort is 94% White and well educated, with high rates of previous PSA testing. The detection of PCA in obese men may differ from non-obese men, which may have led to diagnostic bias, but we expect any bias to be minimal as no effect modification by PSA testing history was detected in a post hoc analysis. Trajectory groups were based on four BMI time points and self-reported, possibly limiting pattern sensitivity. However our earlier study of the PLCO population found similar BMI trajectories.14 Self-reported weight and height are the largest limitations of this study, as they can lead to misclassification of obesity status due to recall (particularly for recalled weight decades in the past) and this would usually cause attenuation of any true effect towards the null.52 Conversely, cohorts with similar population characteristics have found strong correlations between measured anthropometrics and current and historical self-reported anthropometrics.51–56 A final limitation is that of residual confounding, which is a feature of all observational studies.
Strengths of our study include one of the largest sample sizes to date to examine associations between BMI and fatal PCA. Our use of longitudinal BMI trajectories enabled us to classify men into groups with distinct BMI patters during adulthood. Research has established that trajectory modelling offers a robust risk prediction method when examining the impact of repeated measures of BMI, compared with cumulative measures of body weight change.57 Finally, over 95% of the cohort has complete follow-up, and all cancer diagnoses had histological verification and, among men who died, causes of death were accurately determined.
This study provides evidence that BMI at study entry and maximum BMI during adulthood are associated with increased risks of fatal PCA. In addition, it suggests that, among never-smokers, substantial weight gain during adulthood that results in obesity is associated with a 2-fold greater risk of fatal PCA compared with maintaining a stable normal weight. For adulthood BMI changes to have clinical utility, a greater understanding of how BMI affects PCA incidence, survival and mortality is needed. Our findings here of effect modification by smoking is an important step in helping to clarify these complex relationships.
Funding
This work was supported by the Intramural Program of the National Cancer Institute at the National Institutes of Health and Department of Health and Human Services. The study sponsor had no role in the design of the study, data collection, the analysis or interpretation of the data, the writing of the article or the decision to submit for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Conflict of interest: There are no financial disclosures from any of the authors.
Supplementary Material
References
- 1. Siegel RL, Miller KD, Jemal A.. Cancer statistics, 2017. CA Cancer J Clin 2017;67:7–30. [DOI] [PubMed] [Google Scholar]
- 2. Flegal KM, Carroll MD, Kit BK. et al. Prevalence of obesity and trends in the distribution of body mass index among us adults, 1999-2010. JAMA 2012;307:491–97. [DOI] [PubMed] [Google Scholar]
- 3. Finucane MM, Stevens GA, Cowan MJ. et al. National, regional, and global trends in body-mass index since 1980: systematic analysis of health examination surveys and epidemiological studies with 960 country-years and 9.1 million participants. Lancet 2011;377:557–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Park J, Morley TS, Kim M. et al. Obesity and cancer - mechanisms underlying tumour progression and recurrence. Nat Rev Endocrinol 2014;10:455–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Renehan AG, Roberts DL, Dive C.. Obesity and cancer: pathophysiological and biological mechanisms. Arch Physiol Biochem 2008;114:71–83. [DOI] [PubMed] [Google Scholar]
- 6. Calle EE, Kaaks R.. Overweight, obesity and cancer: epidemiological evidence and proposed mechanisms. Nat Rev Cancer 2004;4:579–91. [DOI] [PubMed] [Google Scholar]
- 7. Renehan AG, Zwahlen M, Egger M.. Adiposity and cancer risk: new mechanistic insights from epidemiology. Nat Rev Cancer 2015;15:484–98. [DOI] [PubMed] [Google Scholar]
- 8. Hursting SD, Berger NA.. Energy balance, host-related factors, and cancer progression. J Clin Oncol 2010;28:4058–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Albanes D, Weinstein SJ, Wright ME. et al. Serum insulin, glucose, indices of insulin resistance, and risk of prostate cancer. J Natl Cancer Inst 2009;101:1272–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. De Marzo AM, Platz EA, Sutcliffe S. et al. Inflammation in prostate carcinogenesis. Nat Rev Cancer 2007;7:256–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Platz EA, De Marzo AM.. Epidemiology of inflammation and prostate cancer. J Urol 2004;171:S36–40. [DOI] [PubMed] [Google Scholar]
- 12. Platz EA, Giovannucci E.. The epidemiology of sex steroid hormones and their signaling and metabolic pathways in the etiology of prostate cancer. J Steroid Biochem Mol Biol 2004;92:237–53. [DOI] [PubMed] [Google Scholar]
- 13. Giovannucci E, Michaud D.. The role of obesity and related metabolic disturbances in cancers of the colon, prostate, and pancreas. Gastroenterology 2007;132:2208–25. [DOI] [PubMed] [Google Scholar]
- 14. Kelly SP, Graubard BI, Andreotti G. et al. Prediagnostic body mass index trajectories in relation to prostate cancer incidence and mortality in the PLCO cancer screening trial. J Natl Cancer Inst 2017;109:djw225. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Attard G, Parker C, Eeles RA. et al. Prostate cancer. Lancet 2016;387:70–82. [DOI] [PubMed] [Google Scholar]
- 16. Song M, Giovannucci E.. Estimating the influence of obesity on cancer risk: stratification by smoking is critical. J Clin Oncol 2016;34:3237–39. [DOI] [PubMed] [Google Scholar]
- 17. Berrington de Gonzalez A, Hartge P, Cerhan JR. et al. Body-mass index and mortality among 1.46 million white adults. N Engl J Med 2010;363:2211–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Schatzkin A, Subar AF, Thompson FE. et al. Design and serendipity in establishing a large cohort with wide dietary intake distributions - The National Institutes of Health-American Association of Retired Persons Diet and Health Study. Am J Epidemiol 2001;154:1119–25. [DOI] [PubMed] [Google Scholar]
- 19.World Health Organization (WHO). Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000;894:i–xii, 1–253. [PubMed] [Google Scholar]
- 20. Preston SH, Mehta NK, Stokes A.. Modeling obesity histories in cohort analyses of health and mortality. Epidemiology 2013;24:158–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Stokes A. Using maximum weight to redefine body mass index categories in studies of the mortality risks of obesity. Popul Health Metrics 2014;12:7.. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Nagin DS, Odgers CL.. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol 2010;6:109–38. [DOI] [PubMed] [Google Scholar]
- 23. Nagin DS. Analyzing developmental trajectories: a semiparametric, group-based approach. Psychol Methods 1999;4:139–57. [DOI] [PubMed] [Google Scholar]
- 24. Jones BL, Nagin DS.. Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociol Methods Res 2007;35:542–71. [Google Scholar]
- 25. Miller AB, Yurgalevitch S, Weissfeld JL.. Death review process in the prostate, lung, colorectal and ovarian (PLCO) cancer screening trial. Control Clin Trials 2000;21:400s–06s. [DOI] [PubMed] [Google Scholar]
- 26. Cox DR. Regression models and life-tables. J R Stat Soc Ser B Stat Methodol 1972;34:187–220. [Google Scholar]
- 27. Korn EL, Graubard BI, Midthune D.. Time-to-event analysis of longitudinal follow-up of a survey: choice of the time-scale. Am J Epidemiol 1997;145:72–80. [DOI] [PubMed] [Google Scholar]
- 28. Rothman KJ, Greenland S, Lash TL.. Modern Epidemiology. Philadelphia, PA: Wolters Kluwer Health/Lippincott Williams & Wilkins, 2008. [Google Scholar]
- 29. Kleinbaum DG, Klein M.. Survival Analysis: A Self-Learning Text, 3rd edn New York, NY: Springer, 2012. [Google Scholar]
- 30. Allott EH, Masko EM, Freedland SJ.. Obesity and prostate cancer: weighing the evidence. Eur Urol 2013;63:800–09. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Freedland SJ, Platz EA.. Obesity and prostate cancer: making sense out of apparently conflicting data. Epidemiol Rev 2007;29:88–97. [DOI] [PubMed] [Google Scholar]
- 32.World Cancer Research Fund International. Continuous Update Project Report: Diet, Nutrition, Physical Activity, and Prostate Cancer 2014. http://www.wcrf.org/sites/default/files/Prostate-Cancer-2014-Report.pdf (16 October 2017, date last accessed).
- 33. Lauby-Secretan B, Scoccianti C, Loomis D. et al. Body fatness and cancer—viewpoint of the IARC Working Group. N Engl J Med 2016;375:794–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Dickerman BA, Ahearn TU, Giovannucci E. et al. Weight change, obesity and risk of prostate cancer progression among men with clinically localized prostate cancer. Int J Cancer 2017;141:933–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Rodriguez C, Freedland SJ, Deka A. et al. Body mass index, weight change, and risk of prostate cancer in the Cancer Prevention Study II Nutrition Cohort. Cancer Epidemiol Biomarkers Prev 2007;16:63–69. [DOI] [PubMed] [Google Scholar]
- 36. Moller E, Wilson KM, Batista JL. et al. Body size across the life course and prostate cancer in the Health Professionals Follow-up Study. Int J Cancer 2016;138:853–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Chamberlain C, Romundstad P, Vatten L. et al. The association of weight gain during adulthood with prostate cancer incidence and survival: a population-based cohort. Int J Cancer 2011;129:1199–206. [DOI] [PubMed] [Google Scholar]
- 38. Hernandez BY, Park SY, Wilkens LR. et al. Relationship of body mass, height, and weight gain to prostate cancer risk in the multiethnic cohort. Cancer Epidemiol Biomarkers Prev 2009;18:2413–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Discacciati A, Orsini N, Andersson SO. et al. Body mass index in early and middle-late adulthood and risk of localised, advanced and fatal prostate cancer: a population-based prospective study. Br J Cancer 2011;105:1061–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Bassett JK, Severi G, Baglietto L. et al. Weight change and prostate cancer incidence and mortality. Int J Cancer 2012;131:1711–19. [DOI] [PubMed] [Google Scholar]
- 41. Moller E, Adami HO, Mucci LA. et al. Lifetime body size and prostate cancer risk in a population-based case-control study in Sweden. Cancer Causes Control 2013;24:2143–55. [DOI] [PubMed] [Google Scholar]
- 42. Song M, Willett WC, Hu FB. et al. Trajectory of body shape across the lifespan and cancer risk. Int J Cancer 2016;138:2383–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Samet JM. Lung cancer, smoking, and obesity: it's complicated. J Natl Cancer Inst 2018; doi:10.1093/jnci/djy019 [DOI] [PubMed] [Google Scholar]
- 44. McGown C, Birerdinc A, Younossi ZM.. Adipose tissue as an endocrine organ. Clin Liver Dis 2014;18:41–58. [DOI] [PubMed] [Google Scholar]
- 45. Walker GE, Marzullo P, Ricotti R. et al. The pathophysiology of abdominal adipose tissue depots in health and disease. Horm Mol Biol Clin Investig 2014;19:57–74. [DOI] [PubMed] [Google Scholar]
- 46. Chiolero A, Faeh D, Paccaud F. et al. Consequences of smoking for body weight, body fat distribution, and insulin resistance. Am J Clin Nutr 2008;87:801–09. [DOI] [PubMed] [Google Scholar]
- 47. Zhong S, Yan X, Wu Y. et al. Body mass index and mortality in prostate cancer patients: a dose-response meta-analysis. Prostate Cancer Prostatic Dis 2016;19:122–31. [DOI] [PubMed] [Google Scholar]
- 48. Discacciati A, Orsini N, Wolk A.. Body mass index and incidence of localized and advanced prostate cancer - a dose-response meta-analysis of prospective studies. Ann Oncol 2012;23:1665–71. [DOI] [PubMed] [Google Scholar]
- 49. Chen Q, Chen T, Shi W. et al. Adult weight gain and risk of prostate cancer: a dose-response meta-analysis of observational studies. Int J Cancer 2016;138:866–74. [DOI] [PubMed] [Google Scholar]
- 50. Keum N, Greenwood DC, Lee DH. et al. Adult weight gain and adiposity-related cancers: a dose-response meta-analysis of prospective observational studies. J Natl Cancer Inst 2015;107:djv088. [DOI] [PubMed] [Google Scholar]
- 51. Williams G. Aromatase up-regulation, insulin and raised intracellular oestrogens in men, induce adiposity, metabolic syndrome and prostate disease, via aberrant ER-alpha and GPER signalling. Mol Cell Endocrinol 2012;351:269–78. [DOI] [PubMed] [Google Scholar]
- 52. Gorber SC, Tremblay M, Moher D. et al. Diagnostic in obesity comorbidities—a comparison of direct vs. self-report measures for assessing height, weight and body mass index: a systematic review. Obes Rev 2007;8:307–26. [DOI] [PubMed] [Google Scholar]
- 53. Rimm EB, Stampfer MJ, Colditz GA. et al. Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1990;1:466. [DOI] [PubMed] [Google Scholar]
- 54. Dahl AK, Hassing LB, Fransson EI. et al. Agreement between self-reported and measured height, weight and body mass index in old age - a longitudinal study with 20 years of follow-up. Age Ageing 2010;39:445–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55. Dahl AK, Reynolds CA.. Accuracy of recalled body weight - a study with 20 years of follow-up. Obesity (Silver Spring) 2013;21:1293–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56. Spencer EA, Appleby PN, Davey GK. et al. Validity of self-reported height and weight in 4808 EPIC-Oxford participants. Public Health Nutr 2002;5:561–65. [DOI] [PubMed] [Google Scholar]
- 57. Zheng H, Tumin D, Qian Z.. Obesity and mortality risk: new findings from body mass index trajectories. Am J Epidemiol 2013;178:1591–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.