Prospective study of adiposity and weight change in relation to prostate cancer incidence and mortality

Abstract

BACKGROUND.

Adiposity has been linked inconsistently with prostate cancer, and few studies have evaluated whether such associations vary by disease aggressiveness.

METHODS.

The authors prospectively examined body mass index (BMI) and adult weight change in relation to prostate cancer incidence and mortality in 287,760 men ages 50 years to 71 years at enrollment (1995–1996) in the National Institutes of Health-AARP Diet and Health Study. At baseline, participants completed questionnaires regarding height, weight, and cancer screening practices, including digital rectal examinations and prostate-specific antigen tests. Cox regression analysis was used to calculate relative risks (RR) and 95% confidence intervals (95% CIs).

RESULTS.

In total, 9986 incident prostate cancers were identified during 5 years of follow-up, and 173 prostate cancer deaths were ascertained during 6 years of follow-up. In multivariate models, higher baseline BMI was associated with significantly reduced total prostate cancer incidence, largely because of the relationship with localized tumors (for men in the highest BMI category [≥40 kg/m²] vs men in the lowest BMI category [<25 kg/m²]: RR, 0.67; 95% CI, 0.50–0.89; P = .0006). Conversely, a significant elevation in prostate cancer mortality was observed at higher BMI levels (BMI <25 kg/m²: RR, 1.0 [referent group]; BMI 25–29.9 kg/m²: RR, 1.25; 95% CI, 0.87–1.80; BMI 30–34.9 kg/m²: RR, 1.46; 95% CI, 0.92–2.33; and BMI ≥35 kg/m²: RR, 2.12; 95% CI, 1.08–4.15; P = .02). Adult weight gain from age 18 years to baseline also was associated positively with fatal prostate cancer (P = .009), but not with incident disease.

CONCLUSIONS.

Although adiposity was not related positively to prostate cancer incidence, higher BMI and adult weight gain increased the risk of dying from prostate cancer. Cancer 2007. Published 2007 by the American Cancer Society.

The prevalence of overweight and obesity in most developed countries has been increasing at an alarming rate. In 2000, nearly 66% of adults in the United States were classified as overweight or obese.1 It is widely accepted that obesity increases the risk of several chronic diseases, including type II diabetes, cardiovascular disease, and several types of cancer.2

The relation between adiposity and prostate cancer risk has been studied extensively but remains inconclusive. Most previous epidemiological studies do not support adiposity as a predictor of prostate cancer risk, although a recent meta-analysis demonstrated a weak, positive association between body mass index (BMI) and prostate cancer.3 In contrast, 1 prospective study showed that obesity may be associated inversely with prostate cancer risk, but only in specific subsets of men defined by age and family history.4

Although results from epidemiological studies have been inconclusive regarding adiposity and the development of prostate cancer, some investigators have reported that higher BMI is linked more consistently to increased risk of advanced5–7 and fatal prostate cancer.8–11 Furthermore, increased biochemical progression after radical prostatectomy has been noted in obese men compared with normal-weight men, which supports a biologic role for adiposity in more aggressive disease.12

It also is possible that adult weight gain influences prostate cancer independent of initial body mass. Although most studies have reported null results pertaining to weight change and prostate cancer,5, 7, 13–16 the results from 1 prospective study indicated that radical prostatectomy patients who gained weight at the greatest rate (>1.5 kg per year) between age 25 years and diagnosis experienced more rapid biochemical failure after treatment than patients who gained weight more slowly, suggesting a correlation with disease progression.17

It remains unclear whether and how obesity differentially affects localized versus aggressive prostate cancers.18, 19 We therefore examined the associations of BMI and adult weight change separately with total, localized, extraprostatic, and fatal prostate cancer in a large, prospective cohort of men participating in the National Institutes of Health (NIH)-AARP Diet and Health Study.

MATERIALS AND METHODS

Study Population

The NIH-AARP Diet and Health Study was initiated in 1995–1996, when an extensive baseline questionnaire was mailed to 3.5 million AARP members ages 50 years to 71 years who resided in 1 of 6 states (California, Florida, Pennsylvania, New Jersey, North Carolina, and Louisiana) or in 2 metropolitan areas (Atlanta, Georgia, and Detroit, Michigan).20 This questionnaire ascertained information on usual dietary intake, current height and weight, and other risk factors. In total, 617,119 individuals returned the baseline questionnaire, with 567,169 determined to be satisfactorily completed.20 In late 1996, a supplementary questionnaire was mailed to all participants who had successfully completed the baseline questionnaire. There were specific questions regarding prostate-specific antigen (PSA) testing and digital rectal examination (DRE) during the 3 years prior to baseline, and weight at age 18 years. In total, 334,910 individuals responded to the supplementary questionnaire.

Among the 567,169 individuals who returned the baseline questionnaire, we excluded those with duplicate questionnaires (n = 179), individuals who had died or moved out of the study area prior to baseline (n = 582), individuals who withdrew from the study (n = 1), individuals who had questionnaires completed by proxy respondents (n = 15,760), women (n = 225,471), and men who had been previously diagnosed with cancer, except for nonmelanoma skin cancer (n = 27,269), had missing information on height or weight (n = 5610), or had extreme values for total energy consumption (n = 2505), height (n = 1456), weight (n = 402), or BMI (n = 174). After these exclusions, 287,760 men were available for the current analysis, including 172,961 men who had available information from the more detailed supplementary questionnaire.

The NIH-AARP Diet and Health Study was approved by the Special Studies Institutional Review Board of the National Cancer Institute. All participants provided informed consent.

Assessment of BMI and Adult Weight Change

Information on height and weight was obtained by self-report. BMI was categorized according to World Health Organization guidelines for underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), obese Class I (30.0–34.9 kg/m²), obese Class II (35.0–39.9 kg/m²), and obese Class III/morbid obesity (≥40.0 kg/m²). For the referent category, we used a BMI <25 kg/m² because there were insufficient numbers of men in the underweight category. The exclusion of men with a BMI <18.5 kg/m² from the reference group had no effect on risk estimates. We collapsed obesity categories when examining extraprostatic and fatal prostate cancers to ensure sufficient statistical power. BMI at age 18 years was categorized as <18.5 kg/m², 18.5–20.9 kg/m², 21–22.9 kg/m², 23–24.9 kg/m², and ≥25 kg/m².

Weight change between age 18 years and baseline was calculated by subtracting self-reported weight at age 18 years from baseline weight and was divided into the following 7 groups: weight loss >4 kg, stable weight (gain or loss <4 kg), and weight gains of 4–9.9 kg, 10–19.9 kg, 20–29.9 kg, 30–39.9 kg, or ≥40 kg.

Ascertainment and Classification of Prostate Cancer Cases and Deaths

Men with incident, first primary prostate cancer (International Classification of Diseases 9th version, rubric 185 or 10th version, rubric C61) were identified through December 31, 2000 through linkage of the NIH-AARP cohort database with the state cancer registries. Deaths from prostate cancer as the underlying cause were ascertained through December 31, 2001 using the National Death Index. We defined localized (organ-confined) prostate cancer as a clinical or pathologic tumor classification of T1a to T2b and N0M0 according to the American Joint Committee on Cancer 1997 Tumor-Lymph Node-Metastasis classification system.21 Men with clinical or pathologic tumor classifications of T3 or T4, N1 status, or M1 status, and men who died of prostate cancer from 1995 to 2000, were considered to have extraprostatic tumors. Men with prostate cancer who were identified first through linkage with a state cancer registry and subsequently died from prostate cancer from 1995 to 2000 were classified as fatal cases, regardless of their stage at diagnosis. Men with prostate cancer who were identified through linkage with a state cancer registry and subsequently died from prostate cancer from 2000 to 2001 (outside of the follow-up period for incident cancers) were classified As incident cases based on their stage at diagnosis. The exclusion of prostate cancers that were classified as T1a (n = 89) or T3a (n = 219) from the localized disease and extraprostatic disease groups, respectively, did not alter the results; these patients were retained in all analyses. Information on Gleason grade was unavailable.

Statistical Analysis

Follow-up for each participant accrued from the date of return of the baseline questionnaire to the date of prostate cancer diagnosis, death, or the end of the study period. Cox proportional hazards models with age as the underlying time metric were used to estimate relative risks (RRs) and 95% confidence intervals (95% CIs) of prostate cancer. Tests for trend were conducted using the median value of each BMI category modeled as a continuous variable. The proportional hazards assumption was tested and upheld in all analyses. Covariates that were selected for evaluation as confounders included those that consistently were associated with prostate cancer risk in the literature or those that represented significant predictors of total prostate cancer in our dataset. All multivariate models were adjusted for race, smoking status, educational attainment, personal history of diabetes, and family history of prostate cancer. Further adjustment for height, physical activity, and dietary intakes of energy, red meat, fish, tomatoes, alcohol, α-linolenic acid, calcium, α-tocopherol, and selenium, individually and collectively, did not alter the BMI-prostate cancer association by >10%, and these factors were not included in the final multivariate models. In addition, a subanalysis was performed in the subset of individuals who completed the supplementary questionnaire to evaluate confounding by self-reported PSA and DRE screening practices.

Adjusted RRs and 95% CIs for prostate cancer endpoints were estimated for continuous values of baseline BMI in proportional hazards models that included terms for a 4-knot cubic spline. The reference value for these curves (RR, 1.00) was set at 23.5 kg/m².

Effect modification was evaluated in stratified analyses and was tested statistically by adding multiplicative interaction terms to the appropriate multivariate model. We examined whether the association between BMI and prostate cancer risk varied according to subgroups defined by age, race, smoking status, personal history of diabetes, family history of prostate cancer, and screening history of PSA and DRE. Cubic splines were used to determine how to collapse BMI into groups for stratified analyses. For all comparisons, P values were 2-sided, and α values <.05 indicated statistical significance.

RESULTS

During up to 5 years of follow-up (from 1995 to 2000), 9986 men with incident prostate cancer were identified. Of these, 8541 men had localized (organ-confined) disease and 1445 men had extraprostatic disease. There were 173 prostate cancer deaths that occurred between 1995 and 2001. The mean age was 62.1 years, and the average BMI at baseline was 27.3 kg/m². Approximately 29% of men were classified as normal weight (BMI <25 kg/m²), 50% were classified as overweight (BMI 25–29.9 kg/m²), and 21% were classified as obese (BMI ≥30 kg/m²).

Men in the higher categories of BMI at baseline were younger, were less physically active, were more likely to be black, were more likely not to currently smoke, and were less educated than leaner individuals (Table 1). In addition, a larger proportion of heavy men reported a personal history of diabetes, whereas they were less likely to have received a PSA or DRE screening test within 3 years prior to baseline. BMI had a positive association with energy consumption and energy-adjusted intakes of red meat, fish, tomatoes, α-linolenic acid, and selenium and was inversely related to the consumption of alcohol, calcium, and vitamin E.

Table 1. Demographic and Lifestyle Characteristics According to Body Mass Index at Baseline*

Characteristic	BMI categories, kg/m^2
	<25	25–29.9	30–34.9	35–39.9	≥40
BMI indicates body mass index; DRE, digital rectal examination; PSA, prostate specific antigen. * Means and proportions. † Available for individuals (n = 172,961) who completed the supplemental questionnaire that inquired about screening history and weight at age 18 years. ‡ Defined as physical activity for ≥20 minutes that caused increases in breathing or heart rate or worked up a sweat. § Includes Hispanics, Asians, Pacific Islanders, and American Indians/Alaska Natives. ‖ PSA blood test within 3 years prior to baseline. ¶ DRE within 3 years prior to baseline. # Adjusted for total energy intake with the exception of alcohol. ** Includes intakes from diet and supplements.
No. of participants	83,771	142,749	47,749	10,506	2985
Baseline BMI, kg/m^2	23.1	27.2	31.9	36.9	43.1
Baseline height, m	1.78	1.78	1.78	1.78	1.77
Baseline weight, kg	73.5	86.2	102	117	136
Weight at age 18 y, kg†	64.8	68.5	73.3	77.4	81.2
Age, y	62.6	62.2	61.6	61.0	60.2
Physical activity, times per wk‡	2.90	2.63	2.19	1.80	1.44
Race, %
White	93.2	94.0	93.9	93.9	92.7
Black	2.08	2.60	3.46	3.70	4.76
Other§	4.68	3.40	2.63	2.41	2.55
Smoking status, %
Never	33.4	28.8	26.1	25.4	27.1
Former	53.2	61.6	65.2	66.8	65.2
Current	13.5	9.6	8.7	7.8	7.7
Education, %
<12 y	4.8	5.6	6.9	7.8	8.3
12 y	13.2	15.8	17.8	18.5	18.1
Some college	28.6	31.8	34.2	35.4	36.5
≥College degree	53.3	46.7	41.0	38.3	37.2
Family history of prostate cancer: Yes, %	8.1	8.5	8.3	8.2	8.5
Personal history of diabetes: Yes, %	6.3	9.2	15.6	23.0	29.7
Screening history: Yes, %†
PSA‖	72.4	72.1	69.6	66.6	63.5
DRE¶	86.4	86.6	85.4	83.3	80.2
Daily dietary intake#
Total energy, kcal	1975	1996	2078	2161	2286
Red meat, g	61.9	71.7	80.6	86.5	91.0
Fish, g	20.3	20.7	21.0	21.3	22.7
Tomatoes, pyramid servings	0.61	0.63	0.65	0.66	0.68
Alcohol, g	17.1	17.0	16.1	14.2	11.1
α-Linolenic acid, g	1.27	1.31	1.34	1.37	1.39
Calcium, mg**	930	908	899	901	900
α-Tocopherol, mg**	79.4	74.9	69.1	66.7	66.3
Selenium, μg**	105	105	107	109	111

In age- and multivariate-adjusted analyses, men in the 2 highest baseline BMI categories (Class II and morbid obesity) had substantially lower risks of total prostate cancer compared with men in the referent group (Table 2). This relation was due primarily to the inverse association with localized disease, which comprised >85% of prostate cancer cases in this cohort. We noted an inverse association with extraprostatic tumors that was limited to men who had a BMI ≥35 kg/m². In contrast, BMI was related positively with fatal prostate cancer. Men who had a BMI ≥35 kg/m² had a >2-fold increased risk of dying from prostate cancer, with a significant dose-response trend.

Table 2. Relative Risks for Prostate Cancer Incidence and Mortality According to Body Mass Index at Baseline

	BMI categories, kg/m^2
	<25	25–29.9	30–34.9	35–39.9	≥40	P trend
BMI indicates body mass index; RR, relative risk; 95% CI, 95% confidence interval; Ref, reference category. * Adjusted for age, race, smoking status, education, personal history of diabetes, and family history of prostate cancer. † For the analyses of extraprostatic disease and prostate cancer mortality, the BMI categories 35–39.9 kg/m^2 and ≥40 kg/m^2 categories were combined into a single category of ≥35 kg/m^2.
Incidence
Person-years	360,359	616,744	206,100	45,230	12,770
Total no. of cases	3076	5054	1532	269	55
Age-adjusted RR (95% CI)	1.0 (Ref)	0.99 (0.94–1.03)	0.94 (0.88–1.00)	0.79 (0.70–0.89)	0.61 (0.47–0.80)	<.0001
Multivariate RR (95% CI)*	1.0 (Ref)	1.00 (0.95–1.04)	0.97 (0.91–1.03)	0.84 (0.74–0.95)	0.65 (0.50–0.85)	.0008
Localized disease, no.	2652	4328	1277	236	48
Age-adjusted RR (95% CI)	1.0 (Ref)	0.98 (0.94–1.03)	0.91 (0.85–0.97)	0.81 (0.71–0.92)	0.63 (0.47–0.84)	<.0001
Multivariate RR (95% CI)*	1.0 (Ref)	0.99 (0.94–1.04)	0.94 (0.88–1.01)	0.86 (0.75–0.98)	0.67 (0.50–0.89)	.0006
Extraprostatic disease, no.	424	726	255	40†
Age-adjusted RR (95% CI)	1.0 (Ref)	1.01 (0.90–1.14)	1.09 (0.93–1.27)	0.63 (0.46–0.87)†		.26
Multivariate RR (95% CI)*	1.0 (Ref)	1.03 (0.91–1.16)	1.14 (0.97–1.33)	0.68 (0.49–0.94)†		.64
Mortality
Person-years	458,242	785,391	261,923	73,176†
No. of deaths	44	87	31	11†
Age-adjusted RR (95% CI)	1.0 (Ref)	1.22 (0.85–1.76)	1.44 (0.91–2.28)	2.07 (1.07–4.02)†		.02
Multivariate RR (95% CI)*	1.0 (Ref)	1.25 (0.87–1.80)	1.46 (0.92–2.33)	2.12 (1.08–4.15)†		.02

The exclusion of men who were diagnosed within the first year of follow-up did not alter the BMI-prostate cancer associations, suggesting that weight change caused by the presence of asymptomatic disease at baseline did not bias the results: The RR of total prostate cancer for men with a BMI ≥40 kg/m² versus a BMI <25 kg/m² was 0.69 (95% CI, 0.52–0.93), the RR of extraprostatic disease for men with a BMI ≥35 kg/m² versus a BMI <25 kg/m² was 0.78 (95% CI, 0.54–1.13), and the RR of fatal prostate cancer for men with a BMI ≥35 kg/m² versus a BMI <25 kg/m² was 2.45 (95% CI, 1.11–5.43). We repeated all of the aforementioned analyses in the subset of individuals who completed the supplementary questionnaire on cancer screening practices and observed similar associations overall: The RR of fatal prostate cancer was 1.0 for men with a BMI <25 kg/m², 1.55 for men with a BMI from 25–29.9 kg/m², 1.88 for men with a BMI from 30–34.9 kg/m², and 3.63 for men with a BMI ≥35 kg/m² (95% CI, 1.52–8.71; P = .003). Additional adjustment for PSA and DRE screening did not alter the findings: The RR of fatal prostate cancer was 1.0 for men with a BMI <25 kg/m², 1.56 for men with a BMI from 25–29.9 kg/m², 1.88 for men with a BMI from 30–34.9 kg/m², and 3.58 for men with a BMI ≥35 kg/m² (95% CI, 1.49–8.60; P = .004).

To further characterize the observed associations, we entered baseline BMI as a continuous variable in a restricted, 4-knot, cubic spline regression analysis (Fig. 1). The patterns of risk were similar to those observed in the categorical analyses. Specifically, the risks of total and localized prostate cancer declined as BMI exceeded 28 kg/m², and the risk of fatal prostate cancer increased monotonically above a BMI of 25 kg/m².

Figure 1. Cubic spline regression for prostate cancer incidence and mortality according to body mass index (BMI). Solid line indicates the predicted relative risk (RR); dashed lines, 95% confidence intervals. The reference value (RR, 1.00) is 23.5 kg/m².

A marginally significant interaction between BMI and personal history of diabetes was observed for prostate cancer mortality (P for interaction = .05) (Table 3). There were no quantitative interactions between BMI and age, family history of prostate cancer, race, or smoking status for any prostate cancer endpoint (data not shown), nor were there any interactions between BMI and PSA screening or DRE examination (Table 3).

Table 3. Relative Risks for Prostate Cancer Incidence and Mortality According to Body Mass Index at Baseline Stratified by Selected Factors*

Factor	Incidence						Mortality
	Total		Localized		Extraprostatic
	BMI ≥28 vs <28 kg/m^2†		BMI ≥28 vs <28 kg/m^2†		BMI ≥30 vs <30 kg/m^2†		BMI ≥25 vs <25 kg/m^2†
	No. of cases ≥28/<28	RR (95% CI)‡	No. of cases ≥28/<28	RR (95% CI)‡	No. of cases ≥30/<30	RR (95% CI)‡	No. of deaths ≥25/<25	RR (95% CI)‡
BMI indicates body mass index; RR, relative risk; 95% CI, 95% confidence interval; DRE, digital rectal examination; PSA, prostate-specific antigen;. * All P values for interaction were >.05 with the exception of diabetes and prostate cancer mortality (P = .05). † Cubic splines were used to determine appropriate cut-off points for BMI in stratified analyses. ‡ Adjusted for age, race, smoking status, education, personal history of diabetes (except in models stratified by diabetes), and family history of prostate cancer. § PSA blood test within 3 years prior to baseline. Models also were adjusted for DRE screening. ‖ Available for individuals (n = 172,961) who completed the supplemental questionnaire. ¶ DRE within 3 years prior to baseline. Models also were adjusted for PSA screening.
Diabetes
No	2990/6237	0.95 (0.91–0.99)	2527/5364	0.93 (0.89–0.98)	262/1074	1.06 (0.92–1.21)	117/38	1.49 (1.03–2.15)
Yes	402/357	1.01 (0.87–1.17)	345/305	1.02 (0.87–1.19)	33/76	0.79 (0.52–1.19)	12/6	0.52 (0.19–1.41)
PSA§‖
No	286/553	0.96 (0.83–1.10)	226/460	0.91 (0.77–1.07)	32/121	1.05 (0.71–1.55)	22/6	1.86 (0.74–4.64)
Yes	1479/3075	0.98 (0.92–1.05)	1278/2709	0.96 (0.90–1.03)	121/446	1.23 (1.00–1.51)	45/13	1.72 (0.92–3.21)
DRE‖¶
No	197/381	0.93 (0.78–1.11)	163/320	0.92 (0.73–1.15)	19/76	0.97 (0.58–1.61)	14/6	1.11 (0.42–2.96)
Yes	1693/3454	0.98 (0.93–1.04)	1445/3027	0.96 (0.90–1.02)	147/528	1.23 (1.02–1.48)	55/14	1.99 (1.10–3.59)

We examined BMI at age 18 years, weight at age 18 years, baseline weight, and weight change between age 18 years and baseline in relation to prostate cancer incidence and mortality (Table 4). BMI at age 18 years, weight at age 18 years, and baseline weight were inversely associated with localized cancers. In contrast, baseline weight and weight gain from age 18 years to baseline both had a positive association with prostate cancer mortality, with a strong dose-response trend.

Table 4. Relative Risks for Prostate Cancer Incidence and Mortality According to Body Mass Index at Age 18 Years, Weight at Age 18 Years, Weight at Baseline, and Weight Change Between Age 18 Years and Baseline*

Variable	Incidence						Mortality
	Total		Localized		Extraprostatic
	No. of cases	RR (95% CI)	No. of cases	RR (95% CI)	No. of cases	RR (95% CI)	No. of deaths	RR (95% CI)
BMI indicates body mass index; RR indicates relative risk; 95% CI, 95% confidence interval; Ref, reference category. * Based on individuals (n = 172,961) who completed the supplementary questionnaire. † Adjusted for age, race, smoking status, education, personal history of diabetes, family history of prostate cancer, BMI at baseline, and height. ‡ Adjusted for age, race, smoking status, education, personal history of diabetes, family history of prostate cancer, weight at baseline, and height. § Cut points were based on quintiles. ‖ Adjusted for age, race, smoking status, education, personal history of diabetes, family history of prostate cancer, weight at age 18 years, and height.
BMI at age 18 y, kg/m^2†
<18.5	723	0.95 (0.87–1.04)	633	0.95 (0.86–1.04)	90	0.98 (0.77–1.26)	13	1.67 (0.82–3.42)
18.5–20.9	1787	1.0 (Ref)	1570	1.0 (Ref)	217	1.0 (Ref)	18	1.0 (Ref)
21–22.9	1510	1.01 (0.95–1.09)	1317	1.01 (0.94–1.09)	193	1.04 (0.86–1.27)	25	1.65 (0.90–3.02)
23–24.9	775	0.90 (0.83–0.98)	653	0.87 (0.80–0.96)	122	1.11 (0.88–1.39)	16	1.71 (0.86–3.39)
≥25	641	0.93 (0.84–1.02)	535	0.89 (0.80–0.99)	106	1.15 (0.90–1.47)	11	1.35 (0.62–2.95)
P trend		.17		.04		.18		.73
Weight at age 18 y, kg‡§
≤58.6	1004	1.0 (Ref)	881	1.0 (Ref)	123	1.0 (Ref)	18	1.0 (Ref)
58.7–64.5	1338	1.01 (0.93–1.10)	1185	1.02 (0.94–1.12)	153	0.95 (0.74–1.20)	13	0.52 (0.26–1.07)
64.6–69.9	1043	0.99 (0.91–1.09)	903	0.98 (0.89–1.08)	140	1.08 (0.84–1.38)	15	0.73 (0.36–1.48)
70–76.7	1138	0.99 (0.91–1.09)	988	0.99 (0.90–1.09)	150	1.03 (0.80–1.33)	17	0.73 (0.36–1.47)
>76.7	1071	0.92 (0.84–1.02)	891	0.88 (0.80–0.98)	180	1.18 (0.91–1.54)	23	0.87 (0.43–1.78)
P trend		.08		.01		.13		.86
Weight at baseline, kg§‖
≤74.5	1126	1.0 (Ref)	989	1.0 (Ref)	137	1.0 (Ref)	14	1.0 (Ref)
74.6–81.3	1224	1.02 (0.93–1.11)	1077	1.01 (0.93–1.11)	147	1.03 (0.81–1.32)	11	0.77 (0.35–1.72)
81.4–87.2	1204	1.01 (0.92–1.10)	1046	1.00 (0.91–1.10)	158	1.09 (0.85–1.40)	19	1.32 (0.63–2.75)
87.3–97.2	1157	1.00 (0.91–1.09)	984	0.96 (0.87–1.06)	173	1.26 (0.98–1.62)	19	1.49 (0.70–3.17)
>97.2	1014	0.91 (0.82–1.00)	859	0.87 (0.78–0.97)	155	1.16 (0.88–1.52)	26	2.19 (1.00–4.78)
P trend		.99		.50		.08		.01
Weight change from age 18 y to baseline, kg‖
<−4	161	1.00 (0.83–1.19)	139	1.03 (0.85–1.25)	22	0.81 (0.50–1.31)	3	1.18 (0.29–4.74)
−4 to 3.9	430	1.0 (Ref)	363	1.0 (Ref)	67	1.0 (Ref)	6	1.0 (Ref)
4–9.9	936	1.04 (0.93–1.17)	812	1.06 (0.94–1.20)	124	0.92 (0.68–1.24)	12	1.06 (0.40–2.83)
10–19.9	1896	1.12 (1.00–1.24)	1648	1.13 (1.01–1.27)	248	1.03 (0.78–1.35)	23	1.17 (0.47–2.92)
20–29.9	1425	1.12 (1.00–1.26)	1245	1.14 (1.01–1.29)	180	1.03 (0.77–1.39)	24	1.74 (0.69–4.40)
30–39.9	469	0.99 (0.87–1.14)	400	0.98 (0.85–1.14)	69	1.08 (0.76–1.53)	10	2.05 (0.72–5.90)
≥40	277	1.03 (0.88–1.20)	241	1.03 (0.87–1.22)	36	0.99 (0.65–1.51)	8	2.98 (0.99–9.04)
P trend		.81		.92		.37		.009

DISCUSSION

In this large, prospective study, obesity was inversely related to localized prostate cancer and possibly to extraprostatic tumors and was positively associated with fatal disease. Increasing weight gain from age 18 years to baseline also was associated with greater risk of dying from prostate cancer. These associations did not vary by age, family history of prostate cancer, race, or screening history.

More than 20 prospective studies have examined the relation between adult BMI and incident prostate cancer: Most demonstrated no association,5, 8, 13–16, 22–35 although some studies reported positive6, 7, 36–39 or inverse4 findings. Several studies have evaluated localized and advanced prostate cancer separately,4–7, 13, 16, 24, 27 and 3 studies5–7 indicated that BMI was related more strongly to increased risk of aggressive rather than localized disease, although that relation was not always significant.5 We observed little evidence of a positive association between BMI and extraprostatic disease. In fact, an inverse relation was noted among men with a BMI ≥35 kg/m², although there was no dose-response trend across categories. One previous prospective study demonstrated an inverse association between BMI and total, organ-confined, and advanced prostate cancer, although that finding was restricted to men aged ≤60 years at diagnosis or men who had a positive family history of prostate cancer.4 We observed no interaction between BMI and age or family history in our cohort. Although BMI has been linked variably with incident prostate cancer, including advanced disease, a consistent increase in prostate cancer mortality in obese men has been noted in 4 cohort studies8–11 and is supported by our current findings. We are unable to explain why we observed divergent associations of obesity with extraprostatic and fatal prostate cancer, because both are considered to reflect aggressive disease.

We are unaware of any previous study suggesting that adult weight gain increases the risk of prostate cancer mortality. In the Health Professionals Follow-Up Study, there was no association between weight change from age 21 years to adulthood with total or advanced prostate cancer.13 Similarly, weight change or percent change in BMI from early adulthood to baseline was not associated with prostate cancer in 5 additional prospective cohort studies.5, 7, 14–16 In a cohort of prostate cancer patients who had undergone radical prostatectomy, however, men who rapidly gained weight between age 25 years and diagnosis experienced a higher rate of biochemical failure after treatment than men who had experienced slower weight gain.17

Several biologic mechanisms have been proposed to explain the relation between adiposity and the risk of prostate cancer. The predominant hypotheses concern hormonal and metabolic alterations sustained in obese men. Some of these perturbations may lower prostate cancer risk (eg, decreased circulating levels of testosterone), whereas others could increase risk (eg, enhanced levels of insulin, free or bioactive insulin-like growth factor-1 [IGF-1], and leptin).

Testosterone is involved in normal prostatic growth, and high testosterone levels stimulate proliferation in this organ.40 Thus, depressed testosterone concentrations in obese men may explain an inverse association between BMI and indolent prostate cancer. Morbid obesity, in fact, often is accompanied by overt testosterone deficiency caused by hypogonadism.41 In contrast, testosterone helps maintain differentiation in prostatic epithelium,19 and recent studies have shown that total and free testosterone are inversely associated with high-grade disease. This suggests that low testosterone may increase the risk of developing poorly differentiated and hormone-insensitive prostate tumors.42

IGF-1 is involved in angiogenesis, in the development of bone metastases, and in androgen-independent progression of prostate cancer43, and leptin induces cellular migration and growth-factor expression in hormone-resistant prostate cancer cells.44, 45 Thus, increased levels of IGF-1 and leptin in obese men may explain an apparent adverse effect of obesity on prostate cancer progression that leads to death. The observation of a positive association between obesity and risk of fatal prostate cancer suggests that, in this select group of particularly aggressive cancers, the adverse effects of excess body fat outweigh hormonal perturbations that potentially may decrease the risk of prostate cancer in obese men.

We observed a marginally significant interaction between BMI and diabetes in relation to prostate cancer mortality, with a positive BMI-mortality association evident only in men with no history of diabetes. The biologic mechanism underlying this interaction is speculative but suggests that any potentially positive association between BMI and fatal prostate cancer that is mediated by higher insulin levels may be masked in diabetics, who experience decreasing insulin levels over time. However, our finding also may be explained by chance because of the small numbers of men with fatal disease among those who had a history of diabetes.

We considered whether detection bias may have accounted for the observed associations in our cohort. Our study was initiated after the widespread introduction of PSA testing in the United States. We observed that the proportion of men who reported PSA and/or DRE testing prior to baseline declined across increasing BMI levels, suggesting lower prostate cancer detection rates due to less frequent screening among obese men compared with lean men. Furthermore, studies have shown that prostate cancer is more difficult to detect in obese men because of a combination of lower serum PSA levels, fewer abnormal DRE findings, and larger prostate sizes than lean individuals.18 In our study, adjustment for DRE and PSA screening variables did not alter the associations and screening practices did not modify the relations between BMI and prostate cancer. However, the lack of comprehensive information on prostate cancer screening practices may not have allowed us to control fully for the influence of screening throughout the study period. We also explored the possibility that survival bias may explain our findings of increased prostate cancer mortality among increasingly obese men. However, we did not observe any differences in the length of survival across BMI categories in our study, making such bias an unlikely alternative explanation for our results.

There are several notable strengths of this study. The large number of men with prostate cancer afforded us ample power to detect modest associations and to evaluate effect modification by several important factors. In addition, a relatively wide distribution of BMI values enabled us to examine associations at more extreme ranges than had been possible previously. We also were able to test whether a large number of variables confounded the BMI-prostate cancer associations.

Limitations of our study include the short period of follow-up, the possibility of undiagnosed prostate cancer at baseline because of slow tumor growth rates, and the lack of information on Gleason score, which is an important clinical variable that predicts survival. Although height and weight were based on self-reports in our study, it has been shown previously that these assessments are highly valid compared with measured values (correlation coefficient, >0.9).46 It also has been demonstrated that recalled weight from several decades earlier among elderly patients is correlated highly with measured weight at that time.47 Finally, although BMI is a widely used indicator of obesity, it is an indirect measure of total adiposity and is limited by its inability to differentiate between body fat and lean body mass.48 Analyses of body fat distribution, including waist-to-hip ratio or waist circumference, may provide additional insight into the role of obesity in prostate carcinogenesis. Furthermore, studies of adiposity in childhood and adolescence would contribute to our understanding of the relevant period in life during which adiposity affects the risk for prostate cancer.

In summary, we observed that BMI was inversely associated with incident prostate cancer, which may be attributed in part to detection bias. In contrast, BMI and adult weight gain each were linked with higher prostate cancer mortality, strongly suggesting that adiposity is related adversely to prostate cancer progression leading to death. Public health efforts should continue to address the health consequences of the growing prevalence of obesity in Western countries and worldwide.

Acknowledgements

Supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute.