Volume 125, Issue 16 p. 2877-2885
Original Article
Free Access

Body fat distribution on computed tomography imaging and prostate cancer risk and mortality in the AGES-Reykjavik study

Barbra A. Dickerman PhD

Corresponding Author

Barbra A. Dickerman PhD

Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts

Corresponding author: Barbra A. Dickerman, PhD, Department of Epidemiology, Harvard T. H. Chan School of Public Health, 677 Huntington Avenue, Boston, MA 02115; [email protected]

Search for more papers by this author
Johanna E. Torfadottir PhD

Johanna E. Torfadottir PhD

Centre for Public Health Sciences, University of Iceland, Reykjavik, Iceland

Icelandic Cancer Registry, Reykjavik, Iceland

Search for more papers by this author
Unnur A. Valdimarsdottir PhD

Unnur A. Valdimarsdottir PhD

Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts

Centre for Public Health Sciences, University of Iceland, Reykjavik, Iceland

Department of Medical Epidemiology and Biostatistics, Karolinska Institute, Stockholm, Sweden

Search for more papers by this author
Edward Giovannucci MD, ScD

Edward Giovannucci MD, ScD

Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts

Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts

Department of Nutrition, Harvard T. H. Chan School of Public Health, Boston, Massachusetts

Search for more papers by this author
Kathryn M. Wilson ScD

Kathryn M. Wilson ScD

Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts

Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts

Search for more papers by this author
Thor Aspelund PhD

Thor Aspelund PhD

Centre for Public Health Sciences, University of Iceland, Reykjavik, Iceland

Icelandic Heart Association, Kopavogur, Iceland

Search for more papers by this author
Laufey Tryggvadottir MSc

Laufey Tryggvadottir MSc

Icelandic Cancer Registry, Reykjavik, Iceland

Faculty of Medicine, University of Iceland, Reykjavik, Iceland

Search for more papers by this author
Lara G. Sigurdardottir MD, PhD

Lara G. Sigurdardottir MD, PhD

Centre for Public Health Sciences, University of Iceland, Reykjavik, Iceland

Faculty of Medicine, University of Iceland, Reykjavik, Iceland

Department of Education and Prevention, Icelandic Cancer Society, Reykjavik, Iceland

Search for more papers by this author
Tamara B. Harris MD

Tamara B. Harris MD

Laboratory of Epidemiology and Population Sciences, Intramural Research Program, National Institute on Aging, Bethesda, Maryland

Search for more papers by this author
Lenore J. Launer PhD

Lenore J. Launer PhD

Laboratory of Epidemiology and Population Sciences, Intramural Research Program, National Institute on Aging, Bethesda, Maryland

Search for more papers by this author
Vilmundur Gudnason MD, PhD

Vilmundur Gudnason MD, PhD

Icelandic Cancer Registry, Reykjavik, Iceland

Icelandic Heart Association, Kopavogur, Iceland

Search for more papers by this author
Sarah C. Markt ScD

Sarah C. Markt ScD

Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts

Department of Population and Quantitative Health Sciences, Case Western Reserve University, Cleveland, Ohio

The last 2 authors contributed equally to this article.Search for more papers by this author
Lorelei A. Mucci ScD

Lorelei A. Mucci ScD

Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts

Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts

The last 2 authors contributed equally to this article.Search for more papers by this author
First published: 10 June 2019
Citations: 33
See editorial on pages 2730-1, this issue.

Abstract

Background

The World Cancer Research Fund classifies as “strong evidence” the link between obesity and the risk of advanced prostate cancer. In light of the different hormonal profiles associated with where adipose is stored, this study investigated the role of objectively measured body fat distribution and the risk of clinically relevant prostate cancer.

Methods

This was a prospective study of 1832 men in the Age, Gene/Environment Susceptibility–Reykjavik study. From 2002 to 2006, participants underwent baseline computed tomography imaging of fat deposition, bioelectric impedance analysis, and measurement of body mass index (BMI) and waist circumference. Men were followed through linkage with nationwide cancer registries for the incidence of total (n = 172), high-grade (Gleason grade ≥8; n = 43), advanced (≥cT3b/N1/M1 at diagnosis or fatal prostate cancer over follow-up; n = 41), and fatal prostate cancer (n = 31) through 2015. Cox regression was used to evaluate the association between adiposity measures and prostate cancer outcomes.

Results

Among all men, visceral fat (hazard ratio [HR], 1.31 per 1–standard deviation [SD] increase; 95% confidence interval [CI], 1.00-1.72) and thigh subcutaneous fat (HR, 1.37 per 1-SD increase; 95% CI, 1.00-1.88) were associated with risk of advanced and fatal disease, respectively. Among men who were leaner based on BMI, visceral fat was associated with both advanced and fatal disease. BMI and waist circumference were associated with a higher risk of advanced and fatal disease. No adiposity measures were associated with total or high-grade disease.

Conclusions

Specific fat depots as well as BMI and waist circumference were associated with the risk of aggressive prostate cancer, which may help to elucidate underlying mechanisms and target intervention strategies.

Introduction

Obesity, as measured by body mass index (BMI) or waist circumference, has been consistently associated with a higher risk of advanced prostate cancer and a poorer prognosis after diagnosis.1 Emerging evidence suggests that the specific distribution of body fat may be an important prognostic factor for prostate cancer outcomes.2-5 Body fat distribution is of interest because it may be a marker for different metabolic, hormonal, and inflammatory milieus that play a role in prostate carcinogenesis.3, 6-13 For example, visceral fat is inversely associated with bioavailable testosterone7, 8 and is more strongly associated with insulin resistance and proinflammatory cytokines than subcutaneous fat.6 Greater intermuscular thigh fat has been associated with poorer glucose tolerance,9 whereas subcutaneous thigh fat has been associated with more favorable metabolic factors.10, 11 The identification of adiposity phenotypes at highest risk of aggressive prostate cancer may, therefore, help to elucidate the mechanisms linking obesity with aggressive disease and target corresponding intervention strategies.

To date, few studies have investigated directly measured body fat distribution and prostate cancer risk. Cross-sectional and retrospective studies have reported associations between computed tomography (CT) measures of visceral fat and total4 and high-grade prostate cancer.5 However, these studies were limited by small samples and the potential that the disease or its treatment may have influenced adiposity.4, 5 Furthermore, the association between other fat depots and prostate cancer outcomes remains unclear.

Here we undertook the first prospective study of CT-measured fat distribution and the risk of prostate cancer and measures of aggressive disease.

Materials and Methods

Study Population

We leveraged data from the Age, Gene/Environment Susceptibility–Reykjavik (AGES-Reykjavik) study, a longitudinal, population-based study in Iceland described in detail elsewhere.14 Briefly, AGES-Reykjavik originates from the Reykjavik study, a cohort of 19,381 Reykjavik residents that was established in 1967 to prospectively investigate cardiovascular disease in Iceland. From 2002 to 2006, a random sample of 5764 participants (42% men) were reexamined as part of the AGES-Reykjavik study and underwent a comprehensive baseline examination involving a medical history, a physical examination, imaging studies, and questionnaires on health-related behaviors. At baseline, we excluded those with a history of cancer (n = 453), those missing CT data (n = 136), and those with a BMI <18.5 kg/m2 (n = 17); this left 1832 men in our analysis. Men who were excluded were similar to those included with respect to all baseline characteristics in Table 1. The study was approved by the Icelandic Ethical Review Board and the Icelandic Data Protection Authority.

Table 1. Age-Standardized Characteristics of 1832 Men at Entry Into the AGES-Reykjavik Study by Fat Depot Measures, 2002
Characteristic Abdominal Visceral Abdominal Subcutaneous Thigh Intermuscular Thigh Subcutaneous
<Median (n = 908) ≥Median (n = 924) <Median (n = 907) ≥Median (n = 925) <Median (n = 887) ≥Median (n = 945) <Median (n = 901) ≥Median (n = 931)
Follow-up time, mean (SD), ya, b 8.6 (3.8) 8.8 (3.5) 8.6 (3.8) 8.8 (3.5) 8.9 (3.7) 8.5 (3.6) 8.7 (3.7) 8.6 (3.6)
Age at entry, mean (SD), yb 76.7 (5.4) 75.9 (5.2) 76.9 (5.5) 75.6 (5.1) 76.0 (5.4) 76.5 (5.3) 76.6 (5.4) 76.0 (5.3)
Height, mean (SD), cm 175.0 (6.0) 176.0 (6.2) 175.0 (6.0) 176.0 (6.2) 175.1 (6.0) 175.9 (6.3) 175.2 (6.0) 175.8 (6.3)
BMI, mean (SD), kg/m2 24.9 (2.8) 28.9 (3.4) 24.5 (2.4) 29.3 (3.1) 25.2 (3.1) 28.6 (3.6) 25.2 (2.9) 28.7 (3.6)
Waist circumference, mean (SD), cm 96.6 (7.7) 108.3 (9.2) 95.6 (6.8) 109.2 (8.5) 98.0 (8.9) 106.7 (9.9) 97.5 (8.0) 107.3 (9.9)
Total fat mass, mean (SD), kg 14.8 (4.9) 22.3 (6.2) 14.3 (4.2) 22.9 (5.9) 15.6 (5.6) 21.7 (6.7) 15.4 (5.2) 21.8 (6.6)
Percent body fat, mean (SD) 19.1 (4.7) 24.7 (4.2) 18.8 (4.3) 25.1 (4.1) 19.7 (4.9) 24.2 (4.8) 19.5 (4.7) 24.3 (4.7)
Highest education, %
Primary 16.9 14.9 16.2 15.6 16.3 15.7 15.4 16.6
Secondary 53.4 53.3 54.0 52.7 53.0 53.2 53.5 52.7
College 12.1 12.1 12.5 11.6 12.0 12.2 12.7 11.5
University 16.8 18.4 16.6 18.6 17.7 17.8 17.6 17.9
Smoking status, %
Never 32.0 27.0 29.9 28.9 30.0 28.3 31.9 26.9
Formerc 54.0 64.2 56.5 61.9 56.2 62.6 55.5 63.2
Current 13.9 8.7 13.5 9.2 13.7 9.0 12.6 9.7
Frequency of moderate/vigorous physical activity ≥4 h/wk, % 38.9 34.7 39.0 34.6 39.5 33.7 37.7 35.7
Family history of prostate cancer, % 9.6 9.3 10.0 8.6 8.7 9.9 10.1 8.9
Physician visit over past 12 mo, % 78.2 83.5 79.5 82.3 77.0 84.3 79.9 82.0
Type 2 diabetes, %d 12.0 20.0 13.1 19.1 13.5 18.6 16.6 15.8
  • Abbreviations: AGES-Reykjavik, Age, Gene/Environment Susceptibility–Reykjavik; BMI, body mass index; SD, standard deviation.
  • Values are standardized to the age distribution of the study population. Adiposity measures are dichotomized at the median: 195 cm2 for abdominal visceral fat, 193 cm2 for abdominal subcutaneous fat, 33 cm2 for thigh intermuscular fat, and 71 cm2 for thigh subcutaneous fat. Percentages may not sum to 100% due to rounding.
  • a The time from enrollment to prostate cancer diagnosis, death, or end of follow-up.
  • b The values are not age-adjusted.
  • c Men who regularly smoked at least 100 cigarettes or 20 cigars in their lifetime.
  • d Type 2 diabetes was determined by self-report or a fasting glucose level ≥7 mmol/L.

Adiposity Measures and Covariates

Adiposity was assessed at baseline. Participants underwent CT imaging for the assessment of fat area in the abdomen (visceral and subcutaneous) and thigh (intermuscular and subcutaneous). CT imaging is the gold standard for measuring fat distribution,15 and the internal reliability of this measure was excellent (coefficient of variation <5% for all fat depot measures). CT imaging was performed with a 4-row detector system (Sensation; Siemens Medical Systems, Erlangen, Germany). Abdominal visceral and subcutaneous fat areas (cm2) were measured from a single 10-mm transaxial section at the L4/L5 vertebrae. Visceral fat was distinguished from subcutaneous fat by tracing along the fascial plane defining the internal abdominal wall. Thigh intermuscular and subcutaneous fat areas (cm2) were measured from a single 10-mm transaxial section with a 120-kV peak at the femoral midpoint by manually drawing a line along the deep fascial plane surrounding the thigh muscles.16 The analysis of the CT images was performed with specialized software developed at the University of California, San Francisco. Total body fat was assessed by bioelectrical impedance. Height, weight, and waist circumference were measured by trained technicians. BMI was calculated as weight (kg) divided by height (m) squared. We obtained information on lifestyle and clinical covariates from the baseline questionnaire.

Outcome Ascertainment

Record linkage to the nationwide Icelandic Cancer Registry through unique identification numbers was used to identify prostate cancer diagnosed from study entry through December 31, 2015. Cancer registration is mandatory, and the estimated completeness is very high (99.2%).17 More than 98% of prostate cancer diagnoses were morphologically verified.17 Incident prostate cancer was categorized as total, high-grade (Gleason grade ≥8), advanced (≥cT3b or N1 or M1 at diagnosis or fatal prostate cancer over follow-up), or fatal (which was also included in the advanced category). We were missing data on stage and grade for 13 cases (7.6%). Linkage to the Cause of Death Registry held by the Directorate of Health was used to identify all-cause and prostate cancer–specific deaths over the study period. The cause of death (International Classification of Diseases, Tenth Revision) was coded from death certificates by a trained physician. The reported validity of death certificates for identifying prostate cancer as the underlying cause of death is high (96%).18

Statistical Analysis

We estimated the correlation between adiposity measures by calculating Spearman correlation coefficients (rs). We also conducted a partial correlation analysis adjusted for age.

We used Cox regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for total, high-grade, advanced, and fatal prostate cancer. To verify that the proportional hazards assumption was not violated, we included product terms between each adiposity measure and time and tested whether the coefficients for those terms were statistically significantly different from 0 (Wald test χ2 with 1 df). Time since study entry was the underlying time scale. We followed men from the date of the baseline examination until the incident prostate cancer outcome of interest, death, or the administrative end of follow-up (December 31, 2015), whichever happened first. We adjusted for the baseline covariates: age, family history of prostate cancer, smoking status, education, frequency of moderate/vigorous physical activity during youth and midlife, and presence of a physician visit over the past year. Our primary analyses did not adjust for alcohol consumption because of inconsistent findings for a link with prostate cancer; however, estimates were qualitatively similar with adjustment for alcohol (data not shown). Models for fat depots and waist circumference were additionally adjusted for height (continuous). In sensitivity analyses for the fat depot models, we additionally adjusted for BMI and mutually adjusted for all fat depots. Men missing data on covariates and men with complete data on all covariates (93%) were similar with respect to their baseline characteristics. Missing data for categorical covariates were assigned to the most populous group (smoking status, n = 2; education, n = 19, physical activity, n = 111; physician visit, n = 9); estimates were similar in analyses restricted to men with complete data for all covariates.19

We further conducted prespecified stratified analyses to evaluate whether the association between fat distribution and prostate cancer varied by BMI (dichotomized at the median: <27 vs ≥27 kg/m2). This cutoff was selected to optimize the case distribution and power for analyses in each stratum. Tests for heterogeneity were performed with likelihood ratio tests comparing models with and without a product term between the exposure of interest and BMI. Finally, we conducted sensitivity analyses 1) excluding men older than 80 years at study entry and 2) excluding the first 5 years of follow-up to address potential reverse causation.

Analyses were conducted with SAS 9.3 (SAS Institute, Inc, Cary, North Carolina).

Results

Among 1832 men, there were 172 prostate cancer diagnoses, including 31 prostate cancer–specific deaths, during the study period. Only 1 man was diagnosed with prostate cancer at the time of fatal prostate cancer. Of the incident prostate cancer diagnoses, 41 were advanced, and 43 were high-grade tumors. The median follow-up time was 10.1 years (range, 0.1-13.3 years) until prostate cancer diagnosis and 10.4 years (range, 0.1-13.3 years) until prostate cancer death.

Table 1 shows baseline characteristics of the men by fat depot measures dichotomized at the median. Those with higher visceral fat had higher BMIs and waist circumferences, had less physical activity during youth and midlife, and were less likely to be current smokers. Similar associations were seen for higher levels of other fat depots.

Supporting Table 1 shows the distribution of adiposity measures. At baseline, the median BMI was 27 kg/m2, and the median waist circumference was 102 cm. Table 2 shows the Spearman correlation coefficients between adiposity measures. BMI was highly correlated with waist circumference (rs = 0.87) and total body fat (rs = 0.84). Of the fat depots, abdominal subcutaneous fat was highly correlated with BMI and waist circumference (rs = 0.82 and rs = 0.83, respectively); visceral fat was correlated with BMI and waist circumference to a lesser extent (rs = 0.69 and rs = 0.73, respectively). Estimates were similar after adjusting for age (data not shown).

Table 2. Spearman Correlations Between Adiposity Measures at Baseline for Men in the AGES-Reykjavik Study, 2002
BMI Waist Circumference Total Body Fat Percent Body Fat Abdominal Visceral Abdominal Subcutaneous Thigh Intermuscular Thigh Subcutaneous
BMI 1.00
Waist circumference 0.87 1.00
Total body fat 0.84 0.87 1.00
Percent body fat 0.71 0.77 0.95 1.00
Abdominal visceral 0.69 0.73 0.70 0.65 1.00
Abdominal subcutaneous 0.82 0.83 0.81 0.75 0.55 1.00
Thigh intermuscular 0.57 0.55 0.57 0.53 0.38 0.53 1.00
Thigh subcutaneous 0.60 0.58 0.59 0.56 0.36 0.69 0.34 1.00
  • Abbreviations: AGES-Reykjavik, Age, Gene/Environment Susceptibility–Reykjavik; BMI, body mass index.

Visceral fat was associated with the risk of advanced prostate cancer (HR, 1.31 per 1–standard deviation [SD] increase; 95% CI, 1.00-1.72; Table 3). Thigh subcutaneous fat was associated with the risk of fatal prostate cancer (HR, 1.37 per 1-SD increase; 95% CI, 1.00-1.88; Table 3). Mutual adjustment for all fat depots did not qualitatively change these results (HR for advanced disease, 1.31 per 1-SD increase in visceral adiposity; 95% CI, 0.96-1.80; HR for fatal disease, 1.42 per 1-SD increase in thigh subcutaneous adiposity; 95% CI, 0.90-2.25). Additional adjustment for BMI attenuated the estimates, particularly for the other fat depots (Supporting Table 2). Results for total fat mass and percent fat were similar; a 1-SD increase in either was associated with a nonstatistically significantly higher risk of advanced and fatal disease (Table 3). The association between visceral fat and advanced and fatal disease was stronger and statistically significant among men with a BMI <27 kg/m2 and weaker and not significant among men with a BMI ≥27 kg/m2; however, CIs were wide, and tests for heterogeneity by BMI were not significant (Table 4).

Table 3. Association Between Adiposity Measures and the Risk of Prostate Cancer Among Men in the AGES-Reykjavik Study, 2002-2015
Total Prostate Cancer High-Grade Prostate Cancer Advanced Prostate Cancer Fatal Prostate Cancer
Events/Total Age-Adjusted HR (95% CI)a Fully Adjusted HR (95% CI)b Events/Total Age-Adjusted HR (95% CI)a Fully Adjusted HR (95% CI)b Events/Total Age-Adjusted HR (95% CI)a Fully Adjusted HR (95% CI)b Events/Total Age-Adjusted HR (95% CI)a Fully Adjusted HR (95% CI)b
CT imaging of fat depots
Abdominal visceral per 1-SD increase 172/1832 1.02 (0.88-1.19) 1.02 (0.88-1.19) 43/1832 1.01 (0.75-1.37) 0.98 (0.72-1.33) 41/1832 1.31 (0.99-1.74) 1.31 (1.00-1.72) 31/1832 1.21 (0.86-1.71) 1.24 (0.89-1.73)
Abdominal subcutaneous per 1-SD increase 172/1832 0.96 (0.82-1.12) 0.97 (0.83-1.13) 43/1832 1.02 (0.76-1.38) 1.02 (0.76-1.38) 41/1832 1.17 (0.87-1.57) 1.22 (0.91-1.63) 31/1832 1.18 (0.84-1.66) 1.26 (0.89-1.78)
Thigh intermuscular per 1-SD increase 172/1832 0.91 (0.78-1.07) 0.91 (0.78-1.08) 43/1832 0.92 (0.67-1.27) 0.92 (0.66-1.27) 41/1832 1.00 (0.73-1.36) 1.02 (0.75-1.40) 31/1832 1.22 (0.88-1.70) 1.27 (0.91-1.78)
Thigh subcutaneous per 1-SD increase 172/1832 1.01 (0.87-1.18) 1.02 (0.88-1.19) 43/1832 1.14 (0.87-1.50) 1.14 (0.86-1.50) 41/1832 1.21 (0.92-1.59) 1.25 (0.95-1.64) 31/1832 1.29 (0.94-1.77) 1.37 (1.00-1.88)
Bioelectric impedance analysis
Total fat mass per 1-SD increase 132/1425 1.00 (0.84-1.19) 0.98 (0.83-1.18) 35/1425 1.01 (0.72-1.41) 0.98 (0.69-1.40) 32/1425 1.17 (0.83-1.65) 1.17 (0.83-1.67) 25/1425 1.15 (0.77-1.72) 1.17 (0.78-1.75)
Percent fat per 1-SD increase 132/1425 1.00 (0.84-1.19) 0.99 (0.83-1.18) 35/1425 1.05 (0.75-1.49) 1.03 (0.73-1.46) 32/1425 1.20 (0.84-1.71) 1.19 (0.83-1.69) 25/1425 1.20 (0.79-1.81) 1.20 (0.80-1.81)
Anthropometric measurements
BMI per 5 kg/m2 increase 172/1832 1.01 (0.82-1.24) 1.01 (0.82-1.24) 43/1832 1.05 (0.71-1.58) 1.02 (0.67-1.53) 41/1832 1.45 (0.98-2.16) 1.52 (1.02-2.27) 31/1832 1.46 (0.91-2.34) 1.56 (0.97-2.53)
BMI <25 kg/m2 56/579 1 1 12/579 1 1 10/579 1 1 7/579 1 1
25 kg/m2 ≤ BMI <30 kg/m2 81/899 0.86 (0.61-1.21) 0.84 (0.59-1.19) 22/899 1.05 (0.52-2.12) 0.96 (0.47-1.96) 18/899 1.14 (0.53-2.48) 1.19 (0.54-2.61) 16/899 1.53 (0.63-3.73) 1.68 (0.68-4.14)
BMI ≥30 kg/m2 35/354 0.94 (0.61-1.44) 0.95 (0.62-1.46) 9/354 1.06 (0.44-2.53) 1.00 (0.42-2.43) 13/354 2.18 (0.95-5.03) 2.54 (1.08-6.00) 8/354 2.11 (0.76-5.87) 2.59 (0.90-7.45)
WC per 1-SD increase 172/1832 1.01 (0.87-1.17) 1.02 (0.87-1.19) 43/1832 0.97 (0.72-1.32) 0.95 (0.69-1.31) 41/1832 1.32 (0.98-1.77) 1.40 (1.04-1.89) 31/1832 1.31 (0.92-1.86) 1.45 (1.01-2.07)
  • Abbreviations: AGES-Reykjavik, Age, Gene/Environment Susceptibility–Reykjavik; BMI, body mass index; CI, confidence interval; CT, computed tomography; HR, hazard ratio; SD, standard deviation; WC, waist circumference.
  • Unless otherwise noted, continuous adiposity measures were modeled per 1-SD increase. The adiposity measures (and corresponding 1-SD increments) are abdominal visceral fat (85.7 cm2), abdominal subcutaneous fat (85.6 cm2), thigh intermuscular fat (16.0 cm2), thigh subcutaneous fat (39.2 cm2), total fat mass (6.8 kg), percent fat (5.3%), and waist circumference (10.3 cm).
  • a Adjusted for the age at study entry (continuous).
  • b Additionally adjusted for the following variables measured at study entry: family history of prostate cancer (yes or no), smoking status (never, former, or current), education (primary/secondary or college/university), physical activity (≤3 or ≥4 h/wk), and physician visit over the past 12 months (yes or no). Models for fat depots and waist circumference were additionally adjusted for height (continuous).
Table 4. Association Between Fat Depots and the Risk of Prostate Cancer Among Men in the AGES-Reykjavik Study by BMI, 2002-2015
Total Prostate Cancer High-Grade Prostate Cancer Advanced Prostate Cancer Fatal Prostate Cancer
Events/Total Fully Adjusted HR (95% CI)a Events/Total Fully Adjusted HR (95% CI)a Events/Total Fully Adjusted HR (95% CI)a Events/Total Fully Adjusted HR (95% CI)a
Abdominal visceral per 1-SD (85.7-cm2) increase
BMI <27 kg/m2 99/981 1.20 (0.91-1.57) 26/981 1.31 (0.76-2.24) 20/981 1.95 (1.07-3.54) 15/981 2.13 (1.12-4.05)
BMI ≥27 kg/m2 73/851 1.05 (0.83-1.34) 17/851 0.97 (0.59-1.61) 21/851 1.11 (0.73-1.68) 16/851 0.83 (0.47-1.48)
P b .82 .92 .67 .41
Abdominal subcutaneous per 1-SD (85.6-cm2) increase
BMI <27 kg/m2 99/981 0.97 (0.69-1.37) 26/981 1.60 (0.82-3.11) 20/981 1.02 (0.48-2.16) 15/981 1.34 (0.56-3.21)
BMI ≥27 kg/m2 73/851 1.12 (0.88-1.42) 17/851 1.14 (0.70-1.85) 21/851 1.33 (0.87-2.04) 16/851 1.27 (0.75-2.17)
P b .51 .67 .31 .72
Thigh intermuscular per 1-SD (16.0-cm2) increase
BMI <27 kg/m2 99/981 0.90 (0.65-1.22) 26/981 0.86 (0.46-1.63) 20/981 0.59 (0.28-1.25) 15/981 0.65 (0.28-1.52)
BMI ≥27 kg/m2 73/851 0.99 (0.79-1.24) 17/851 1.11 (0.70-1.74) 21/851 1.11 (0.73-1.69) 16/851 1.56 (0.96-2.53)
P b .78 .63 .67 .08
Thigh subcutaneous per 1-SD (39.2-cm2) increase
BMI <27 kg/m2 99/981 1.12 (0.84-1.51) 26/981 1.66 (0.99-2.79) 20/981 1.07 (0.55-2.08) 15/981 1.16 (0.55-2.44)
BMI ≥27 kg/m2 73/851 1.08 (0.88-1.33) 17/851 1.19 (0.80-1.78) 21/851 1.34 (0.95-1.90) 16/851 1.50 (0.96-2.33)
P b .59 .45 .17 .14
  • Abbreviations: AGES-Reykjavik, Age, Gene/Environment Susceptibility–Reykjavik; BMI, body mass index; CI, confidence interval; HR, hazard ratio; SD, standard deviation.
  • a Adjusted for the following variables measured at study entry: age (continuous), height (continuous), family history of prostate cancer (yes or no), smoking status (never, former, or current), education (primary/secondary or college/university), physical activity (≤3 or ≥4 h/wk), and physician visit over past 12 months (yes or no).
  • b Likelihood ratio test for heterogeneity of the HRs from the 2 strata.

Each 5 kg/m2 increase in BMI was associated with a 50% higher risk of advanced (HR, 1.52; 95% CI, 1.02-2.27) and fatal prostate cancer (HR, 1.56; 95% CI, 0.97-2.53; Table 3). Those who were obese (BMI ≥30 kg/m2) had a higher risk of advanced (HR, 2.54; 95% CI, 1.08-6.00) and fatal disease (HR, 2.59; 95% CI, 0.90-7.45) compared with those with a healthy BMI (Table 3). Each 1-SD (10.3-cm) increase in waist circumference was associated with a 40% higher risk of advanced (HR, 1.40; 95% CI, 1.04-1.89) and fatal disease (HR, 1.45; 95% CI, 1.01-2.07; Table 3).

No adiposity measures were associated with the risk of total or high-grade prostate cancer (Table 3). Results for all adiposity measures were qualitatively similar in sensitivity analyses excluding men older than 80 years at study entry and excluding the first 5 years of follow-up (data not shown).

Discussion

In this prospective cohort of Icelandic men with objective measures of adiposity, visceral fat and thigh subcutaneous fat were associated with the risk of advanced and fatal prostate cancer, respectively. Among men with a lower BMI, visceral fat was associated with both advanced and fatal disease. BMI and waist circumference were also associated with a higher risk of advanced and fatal disease. No adiposity measures were associated with total or high-grade disease.

To our knowledge, this is the first prospective study of directly measured fat distribution and the risk of advanced prostate cancer. Previous retrospective and cross-sectional studies incorporating CT measures of adiposity have reported mixed findings.4, 5, 20 A case-control study (63 prostate cancer cases) reported a positive association between visceral fat and total prostate cancer.4 In contrast, we found an association between prospectively measured visceral fat and the risk of advanced and fatal disease but not total prostate cancer. In cross-sectional studies of men undergoing radiotherapy for prostate cancer (with sample sizes ranging from 276 to 308 men), higher visceral fat and abdominal subcutaneous fat were associated with a higher National Comprehensive Cancer Network prostate cancer risk group,20 and abdominal subcutaneous adiposity was also associated with high-grade prostate cancer.5 One of these studies found that visceral fat and high-grade (Gleason grade ≥7) disease were positively associated among black men but not associated among nonblack men, and this was similar to our results for a population of white men.5 These previous studies differed in their design, size, participant characteristics (eg, age, race, and adiposity measures), modeling of adiposity measures, and analytic approach.

BMI has been associated with a higher risk of advanced and fatal, but not total, prostate cancer,3 which is in agreement with our findings. A meta-analysis showed an 8% higher risk of advanced prostate cancer (relative risk, 1.08; 95% CI, 1.04-1.12; 23 studies) and an 11% higher risk of prostate cancer–specific mortality (relative risk, 1.11; 95% CI, 1.06-1.17; 12 studies) per 5 kg/m2 increase in BMI.3 In the current study, we found that each 5 kg/m2 increase in BMI was associated with a 50% higher risk of advanced and fatal disease. Different estimates across studies may be related to the timing of the BMI measurement, the length of follow-up, and patient characteristics. For example, studies suggest that the association between BMI and prostate cancer risk may differ according to age21, 22 and race.23 Because age and race are key determinants of fat distribution,24 the heterogeneity of findings for BMI may be partly explained by variation in fat distribution patterns differentially associated with prostate cancer.

Findings for waist circumference, a surrogate of central adiposity, and prostate cancer have been mixed. Some studies have found a higher waist circumference to be associated with a higher risk of advanced and high-grade disease,2, 3 whereas other studies have been null.25 This is in line with our findings of a positive association between waist circumference and advanced disease but null results for high-grade disease. Waist circumference is limited by the inability to differentiate visceral adipose from subcutaneous adipose, which may partly explain heterogeneous findings.

Percent body fat, assessed with bioelectric impedance, has been associated with high-grade prostate cancer in case-control studies.26, 27 In contrast, we found no association between prospectively measured percent body fat and high-grade disease. A prospective analysis of 10,564 initially cancer-free men in the Malmö Diet and Cancer cohort similarly found no association between percent body fat and the risk of aggressive prostate cancer (≥cT3 or N1 or M1, Gleason grade ≥8, or pretreatment prostate-specific antigen [PSA] level ≥50 ng/mL).25

A prospective study among 129,502 men in the European Prospective Investigation into Cancer and Nutrition (EPIC) reported that central adiposity, assessed by waist circumference, was associated with a higher risk of advanced and high-grade prostate cancer, particularly among men with a healthy BMI.2 We similarly found that the association between visceral fat and advanced disease was stronger among men with a lower BMI versus a higher BMI, although the CIs were wide. Further exploration of metabolically unhealthy, normal-weight phenotypes with respect to prostate cancer outcomes is needed.

Fat distribution may be an important prognostic factor for prostate cancer outcomes by serving as a marker for metabolic, hormonal, and inflammatory milieus that play a role in prostate carcinogenesis.3, 6-13 For example, visceral fat is inversely associated with bioavailable testosterone7, 8 and adiponectin4 and is more strongly associated with insulin resistance and proinflammatory cytokines than subcutaneous fat6— factors that may influence prostate cancer progression.12, 13, 28, 29

Further studies are needed to investigate whether the fat depots themselves exert systemic or local effects in ways that promote aggressive disease or whether they are markers for a physical activity pattern or underlying hormonal milieu that influences both fat distribution and aggressive disease.30 For example, fat may be preferentially deposited in the visceral depot among leaner men in the presence of a particular hormonal milieu. If this hormonal milieu is also a prognostic factor for advanced prostate cancer, this could partially explain the results of our analyses stratified by BMI.

These findings should be considered in light of potential limitations and strengths. Exposures were measured once at cohort entry, so we were unable to assess changes in fat depots over time. However, given the follow-up time, we were able to assess adiposity in a reasonable etiologic time window of exposure.31, 32 It has been hypothesized that obese men may experience delayed detection (due to lower PSA values and biopsy accuracy) and, therefore, more advanced disease at diagnosis than leaner men, and this might partially explain our findings of a higher risk of aggressive disease for men with higher overall obesity.33, 34 However, we found that higher visceral fat was associated with a higher risk of aggressive disease even among leaner men based on BMI. We did not have data on PSA testing and cannot rule out the possibility that our findings might be partially explained by this factor. However, our population was not subject to routine PSA testing, and we adjusted for a measure of recent health care utilization to account for varying degrees of diagnosis opportunity. The number of advanced and fatal cancers was small, and thus power was reduced. Lastly, our study population consisted of older white men, so the results may not be generalizable to younger, more diverse groups of men.

The major strength of this study is that it is the first prospective analysis of CT-quantified fat depots and prostate cancer risk. Our prospective design minimizes the likelihood of reverse causation, whereby the disease or its treatment influences fat distribution. Furthermore, the use of gold-standard measures of fat distribution enabled us to examine the obesity–prostate cancer link with higher resolution than studies of BMI and waist circumference. This provides more insight into potential underlying mechanisms. The misclassification of fat distribution is a risk in studies relying on surrogate measures and may contribute to the variability in epidemiologic findings on obesity and prostate cancer. Precise measures of fat distribution are particularly important among older individuals because BMI becomes a less reliable measure of adiposity with age due to the loss of lean body mass and the redistribution of adipose toward the visceral compartment.35 Additional strengths of this study include its population-based sample, the long duration of follow-up, the complete and reliable outcome data obtained through registry linkage, and the availability of comprehensive questionnaire data.

In summary, we found that specific fat depots as well as BMI and waist circumference were associated with the risk of advanced and fatal prostate cancer. Studies of BMI or waist circumference alone may not capture important subphenotypes, and this may explain the heterogeneity of previous findings for obesity and prostate cancer. Further studies are needed to prospectively investigate fat distribution and prostate cancer outcomes, with attention to changes in fat depots over time, biological pathways, and potential heterogeneity by BMI. The identification of the adiposity phenotypes at highest risk of clinically relevant prostate cancer may help to elucidate the mechanisms linking obesity with aggressive disease and target intervention strategies.

Funding Support

This work was supported by funding from the National Cancer Institute of the National Institutes of Health (T32 CA 009001 to Barbra A. Dickerman and P50 CA 090381-15 to Sarah C. Markt and Lorelei A. Mucci), an ASISA Fellowship (to Barbra A. Dickerman), and the Prostate Cancer Foundation (to Lorelei A. Mucci and Kathryn M. Wilson). The AGES-Reykjavik study is supported by the Intramural Research Program of the National Institute on Aging (contract N01 AG 12100), the Hjartavernd (the Icelandic Heart Association), and the Althingi (the Icelandic Parliament). In addition, this study was supported by funding from the Harvard Catalyst Award and the Icelandic Cancer Society.

Conflict of Interest Disclosures

The authors made no disclosures.

Author Contributions

Barbra A. Dickerman: Conceptualization, formal analysis, methodology, writing–original draft, and writing–review and editing. Johanna E. Torfadottir: Writing–review and editing. Unnur A. Valdimarsdottir: Writing–review and editing. Edward Giovannucci: Writing–review and editing. Kathryn M. Wilson: Writing–review and editing. Thor Aspelund: Resources and writing–review and editing. Laufey Tryggvadottir: Writing–review and editing. Lara G. Sigurdardottir: Writing–review and editing. Tamara B. Harris: Data curation, funding acquisition, and writing–review and editing. Lenore J. Launer: Data curation, funding acquisition, and writing–review and editing. Vilmundur Gudnason: Data curation, funding acquisition, resources, and writing–review and editing. Sarah C. Markt: Conceptualization, methodology, supervision, and writing–review and editing. Lorelei A. Mucci: Conceptualization, methodology, supervision, and writing–review and editing.