The Effects of Body Mass Index on Changes in Prostate-Specific Antigen Levels and Prostate Volume Over 15 Years of Follow-up: Implications for Prostate Cancer Detection

Prostate cancer is the most common noncutaneous cancer in U.S. men, with an estimated 192,280 new cases diagnosed in 2009 (1). Prostate-specific antigen (PSA) currently is the most common screening test for prostate cancer, with 58% of Caucasian men receiving an annual test (2). The low specificity of PSA testing and questionable benefits of PSA screening on prostate cancer mortality highlight the need for better detection strategies for prostate cancer (3). Knowledge of the influence of concomitant comorbidities on serum PSA concentrations may improve the discriminant value of this test for predicting prostate cancer and reduce the number of unnecessary biopsies and subsequent overdiagnoses of indolent cancers.

Obesity is a growing global epidemic, with more than half of the world's adults categorized as being overweight and up to 30% categorized as obese [body mass index (BMI) ≥ 30 kg/m²; ref. 4]. In the United States, 68% of adults are overweight or obese (5). Obesity may influence the detection of prostate cancer through prostate cancer screening (6, 7), specifically by influencing PSA levels and prostate volumes. Obese men have lower PSA levels than do nonobese men (7–9), which may be due to decreased testosterone concentrations (6) and/or plasma hemodilution, that is, the dilution of soluble tumor markers by increased plasma volumes (9). In addition, previous studies suggest that obese men are more likely to have a diagnosis of later-stage disease and less likely to have a diagnosis of early-stage disease (10–12), implying that the consequences of the role of prostate volume in this detection issue among obese men have important implications in terms of prostate cancer aggressiveness. Therefore, it is possible that obese men are less likely to be recommended for biopsy so that diagnoses may be delayed, resulting in later-stage disease at diagnosis and thus detection of fewer cancers in this group.

Results from previous studies suggesting that obesity may be associated with lower serum PSA concentrations are limited by their cross-sectional designs, their failure to simultaneously account for other aspects of this potential detection bias, including prostate volume, and their inability to elucidate why this association exists. Furthermore, the estimated effects of baseline obesity and weight change on change in serum PSA level and prostate volume in a population-based setting are limited. Understanding these longitudinal associations is relevant because change in serum PSA level may be better than a single serum PSA measurement for prostate cancer screening (13) and therefore has important implications for prostate cancer detection. The increasing prevalence of obesity coupled with the increasing speculation regarding the reliable detection of prostate cancer with serum PSA levels makes it crucial to elucidate how obesity influences the multiple facets of prostate cancer detection over time. Therefore, the goal of this study was to use longitudinal data to investigate the associations between BMI and change in BMI with change in PSA level and to assess the possible roles of PSA hemodilution and prostate volume in explaining the associations of obesity with PSA level and prostate cancer diagnosis.

The Olmsted County Study (OCS) of Urinary Symptoms and Health Status among Men is a longitudinal study of Caucasian men residing in Olmsted County, Minnesota (14, 15). In 1990, a random sample of men aged 40 to 79 years, as enumerated by the Rochester Epidemiology Project, was screened for inclusion (16). Men with a history of prostate or bladder surgery, urethral surgery or stricture, or medical or neurologic conditions that affect normal urinary function were excluded. Eligible men (n = 3,874) were invited to take part in the study, and 2,115 (55%) agreed to participate. Participants completed a previously validated baseline questionnaire that ascertained information on urinary symptoms, medical histories, and various demographic and behavioral characteristics. A 25% random subset of the total cohort was invited to participate in a detailed urologic clinical examination, which included prostate volume and serum PSA measurements. Of the 537 randomly selected men, 476 (89%) agreed to participate in the clinical portion of the study.

Since 1990, the cohort has been actively followed biennially by using a questionnaire similar to the one used at baseline. During the second and third rounds of visits, men who did not participate in the follow-up were replaced by randomly selected eligible men from the community (n = 332 total cohort; n = 159 clinic cohort; Fig. 1). After the third round, the study has been maintained as a fixed cohort. Of the 2,447 men in the OCS, 634 men participated in the clinic cohort with up to 8 rounds of follow-up. Overall, 552 (87%) men had at least 2 PSA, BMI, and prostate volume measurements. All data up to the point of diagnosis or initiation of surgical or medical therapy for benign prostatic hyperplasia (BPH) were included in the analyses. Measurements of PSA level, prostate volume, plasma volume, and PSA mass were censored after diagnosis of prostate cancer, prostate surgeries, and procedures and use of any medications (prescription and herbal) for treatment of BPH. A total of 122 men had at least 1 observation censored. As a result, 545 men were included in this analysis.

Measurements of BMI, prostate volume, and PSA were collected at each round of follow-up during the clinic examination. A trained research assistant measured height and weight, and BMI was calculated by dividing the weight in kilograms by the height in meters squared. Men with a BMI of 30 kg/m² or more were considered obese, based on the definition established by the World Health Organization (WHO; ref. 17). Prostate volume was measured via transrectal sonographic imaging, and serum PSA levels were determined with the Tandem-R PSA assay (Hybritech Inc.). To investigate what role plasma hemodilution plays in influencing PSA levels among obese men, body surface area, plasma volume, and PSA mass were estimated using the following established formulas: body surface area (m²) = body weight (kg)^0.425 × height (m)^0.725 × 0.2025 (18), plasma volume (L) = body surface area (m²) × 1.670 (19), and PSA mass = PSA (ng/mL) × plasma volume (9). Demographic information including salary, years of education, and age at baseline blood draw were collected using the questionnaire.

The cross-sectional associations between baseline obesity and participant demographics at baseline were tested using χ² tests for association and 2-sided Student's t tests where appropriate. The distributions of baseline values and predicted rates of change of PSA, BMI, and prostate volume were determined overall and across 10-year age groups, using medians and interquartile ranges. The associations of both the baseline values and predicted rates of change with age were tested using the test for trend from linear regression.

A longitudinal 2-step analytic approach was used to examine the associations of the individual intercepts and rates of change of BMI and prostate volume with the annual percent change in PSA. First, the rates of change in PSA, BMI, and prostate volume were estimated by individually regressing each measure on time from initial blood draw and age (10-year categories), using linear mixed-effects regression models. Interaction (product) terms were included to allow for different slopes across these age groups. Fixed and random effects were included both to reflect the mean effect and to allow for individual variation in the baseline intercept and change over time. An overall annual change in each measure for each man was estimated by combining the average change over time (fixed effects) with the individual changes (random effects). Similarly, both fixed and random effects allowed determination of an overall baseline intercept for each age decade and for offsets for individual variation. Because of their skewed distributions, PSA level and prostate volume were natural log-transformed and therefore parameter estimates represent percent changes per year. The change in BMI reflects annual absolute changes.

The second step of this approach was to estimate the effects of predicted intercepts of PSA, BMI, and prostate volume and the predicted annual changes of BMI and prostate volume on the predicted annual percent change in PSA (all derived from the mixed model in step 1), using linear regression models adjusting for age.

Additional models of changes in PSA and prostate volume included a categorical measure for obesity (BMI < 30 kg/m²; BMI ≥ 30 kg/m²) and an interaction (product) term between obesity and time to compare the slope of PSA among those who were obese and those who were not obese. Similar models considered baseline BMI (treated as continuous), baseline BMI based on the WHO classifications, and repeated measures of BMI. Finally, to assess a possible delayed effect of BMI on PSA change, these additional models were rerun lagging PSA measures. Because measurements were collected biennially, PSA measures were lagged 2 years after the BMI measures.

The adjusted predicted values of the intercepts and annual changes of plasma volume and PSA mass were also estimated using linear mixed-effects regression models. Because of the skewed distribution, PSA mass was natural log-transformed and therefore parameter estimates represent percent changes per year. The change in plasma volume reflects annual absolute changes. The means and SDs of the predicted intercepts and annual changes of PSA mass and plasma volume from the mixed models were then compared across levels of age and baseline obesity. All statistical analyses were done using SAS 9.2 software (SAS Institute).

Baseline characteristics of the study participants are shown in Table 1. Age was inversely associated with obesity at baseline (P_trend = 0.03), but marital status, education, salary, and prostate volume differed little by baseline obesity status (Table 1). Obese men more often reported a history of type 2 diabetes and a history of hypertension than did nonobese men at baseline (P = 0.006 and 0.008, respectively). Men who were obese had a lower mean PSA level (1.1 ng/mL) than did men who were not obese at baseline (1.4 ng/mL; P = 0.06; Table 1).

Table 2 displays the overall and age-specific distributions of the observed baseline values and predicted annual rates of change for BMI, PSA, and prostate volume (derived from the mixed models). At baseline, the median BMI in the total sample was 27.03 kg/m² and it was inversely associated with age (P_trend = 0.07). Both PSA and prostate volume at baseline were strongly and positively associated with age (P_trend < 0.0001). The median rate of change in BMI was 0.14 kg/m² per year, and it was inversely, though not monotonically, associated with age (P_trend = 0.001). The median rate of change of PSA was 3.82% per year, and this rate increased with age (P_trend < 0.0001). The median rate of change in prostate volume overall was 2.19% per year, and it was minimally associated with age (Table 2).

Table 3 summarizes the means and SDs of the predicted intercepts and annual percent changes in PSA and prostate volume by baseline obesity status. Baseline obesity (BMI ≥ 30 kg/m²) was weakly and inversely associated with baseline PSA levels adjusting for age (P = 0.25). Similar associations were observed with continuous baseline BMI and BMI categorized according to the WHO classifications (Table 3). Men who were obese at baseline had a slightly lower age-adjusted mean annual percent change in PSA level than did men who were not obese (3.04 vs. 3.79; P = 0.33). Baseline obesity was positively associated with baseline prostate volume (P = 0.002) and this association was similar when assessed using baseline BMI and categorical BMI based on WHO classifications (Table 3). Men who were obese at baseline had slightly higher mean annual percent change in prostate volume than did men who were not obese (2.48 vs. 2.20; P = 0.27). The association between baseline obesity and annual percent change in PSA did not change when repeated measurements of BMI were modeled and when PSA measurements were lagged 2 years after the measurement of obesity (β = −0.05; 95% CI: −0.17 to 0.08).

In multivariable models considering the association of predicted annual percent change in PSA with the predicted intercepts of PSA, prostate volume, and BMI and the rates of change of prostate volume and BMI, baseline BMI was inversely associated with the rate of change in PSA (P = 0.03). Age, baseline PSA level, and annual percent change in prostate volume were also positively associated with annual percent in PSA when mutually adjusted for the other covariates (P = 0.04, < 0.0001, and < 0.0001, respectively). Further adjustment for baseline statin and nonsteroidal anti-inflammatory drug use did not change the results (data not shown).

Table 4 summarizes the means and SDs of the predicted intercepts of and rates of change in plasma volume and PSA mass, by 10-year age category and baseline obesity status. Men who were obese at baseline had a higher age-adjusted mean plasma volume level at baseline (3.69 L) than did men who were not obese (3.33 L) (P < 0.001). Men who were obese at baseline also experienced a smaller annual change in plasma volume (0.001 L/y) than did men who were not obese (0.005 L/y; P < 0.001). Neither baseline levels nor the rate of change in PSA mass differed across baseline obesity status, adjusting for age (both P > 0.20).

In this longitudinal study of Caucasian men aged 40 to 79 years, those with higher baseline BMIs had lower baseline PSA levels and less rapid increases in PSA over time. Baseline obesity was also associated with increased plasma volume at baseline and the age-adjusted rate of change in plasma volume over time. Our results suggest that higher BMI is associated with lower PSA levels and less rapid increases in PSA level, purportedly through hemodilution effects. These findings suggest that men with higher BMIs may be less likely to be screened for prostate cancer because of consistently lower PSA levels over time.

Previous cross-sectional studies have shown that obese men have lower PSA levels than do nonobese men (9, 20, 21), which may result in fewer prostate biopsies among obese men. Our finding that baseline BMI was inversely associated with baseline PSA level is similar to those previous cross-sectional findings (8, 21, 22). Furthermore, our results suggest that baseline obesity is also inversely associated with how rapidly PSA levels increase over time. This finding is important, as the rate of change in PSA level or PSA velocity is currently the preferred measure used to diagnose prostate cancer (13, 23).

Our findings also suggest that the association between BMI and PSA is at least in part due to the hemodilution of PSA because baseline obesity was associated with increased plasma volume but not with baseline PSA mass (the product of PSA concentration and plasma volume). Similar associations were seen longitudinally, as baseline obesity was associated with the rate of change in plasma volume but not with the annual percent change in PSA mass over time. Our cross-sectional results are similar to previous studies that found plasma volume to increase with increasing BMI but found no relation between BMI and PSA mass (9, 24). Our longitudinal findings are similar to the results from the placebo arm of the Prostate Cancer Prevention Trial that suggest that increased weight gain is inversely associated with change in PSA (25). Our findings further suggest that the hemodilution of PSA is in part responsible for the association of baseline BMI and changes in PSA over time, which is similar to those simulated previously that suggest both static obesity and weight gain attenuate PSA velocity over time (26). Taken together, these findings have important implications for prostate cancer screening and detection, as current prostate cancer screening practices that use cross-sectional or longitudinal measures of PSA without taking into account BMI may result in missed or delayed diagnoses due to obese men being less likely to be referred for biopsy.

Our results also support the notion that prostate cancer detection issues among obese men could be partly due to the influence of BMI on prostate volume. In this cohort, men with higher baseline BMIs had higher baseline prostate volumes. Our results are similar to several previous studies that found obese men to have larger prostates (7, 27). It is therefore possible that fewer cancers are then detected in obese men because of their larger prostate size or that the diagnoses are delayed, resulting in more aggressive disease at diagnosis.

Overall, the findings from our study, taken together with findings from previous studies, suggest that prostate cancer may be less likely to be detected in obese men partly due to the influence of BMI on the rate of prostate volume growth and the hemodilution of PSA. (7, 28) This bias, coupled with the findings that obese men present with later-stage disease (12) and have worse outcomes after treatment, including greater risk of recurrence (29, 30), suggests that it may be appropriate to take into consideration the obesity status of a man when interpreting PSA screening results and recommending prostate biopsies. To minimize this detection bias, it may be prudent to lower PSA cutoffs for overweight or obese men and/or take more cores in these men at biopsy.

The strengths of this study include the use of longitudinal data with 15 years of follow-up that contained rigorously collected repeated measures of BMI, prostate volume, and serum PSA levels. Censoring outcome values collected after prostate cancer diagnosis, BPH treatment, or prostate surgery yields a disease-free, asymptomatic population. However, this study also has potential limitations that need to be considered. This cohort is composed entirely of Caucasian men, limiting our inferences to other racial and ethnic groups. The long follow-up period makes attrition inevitable; however, previous work in this cohort found that participant dropout was not associated with chronic diseases or serum PSA levels adjusted for age (31), thus suggesting that bias resulting from attrition may be small. Finally, there may have been unmeasured time-dependent confounders that we were unable to account for in our analyses.

In conclusion, baseline obesity was associated with baseline PSA and prostate volume and the rate of changes in PSA. Our results suggest that obesity may delay detection of prostate cancer due in part to larger prostate volumes and PSA hemodilution in obese men. Clinically, it may be prudent to lower PSA cutoff points for overweight or obese men and/or take more cores in these men at biopsy.

No potential conflicts of interest were disclosed.

AG034676, The Rochester Epidemiology Project, DK058859- Natural History of Prostatism: The Olmsted County Study. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.