Associations of Lifestyle and Physiologic Factors with Prostate-Specific Antigen Concentrations: Evidence from the National Health and Nutrition Examination Survey (2001-2004)

Screening for prostate cancer with the use of serum prostate-specific antigen (PSA) concentrations has been commonplace in the United States. Previous evidence suggests that obesity, which has reached epidemic proportions in the United States (1), may be inversely associated with PSA (2-7), likely due to lower testosterone concentrations that influence PSA production (8-10) and/or the hemodilution of PSA among obese men (11). However, there is little evidence whether lifestyle and physiologic factors related to obesity, including energy, fat, and carbohydrate intake, physical activity, insulin resistance, and metabolic syndrome, influence PSA concentrations. One previous large clinical trial (N = 3,341) evaluated energy, fat, and carbohydrate intake in relation with PSA concentrations and found no associations (12). Another study that examined physical activity in relation to PSA reported null findings (13). To our knowledge, the associations of PSA with insulin resistance and metabolic syndrome have not been examined previously. Although one study, making use of national data, has reported inverse associations between diabetes and PSA (14), it must be noted that not all diabetics are insulin resistant or have metabolic syndrome.

Because the relationships of PSA, obesity, physical activity, diet, insulin resistance, and metabolic syndrome are complex and not well defined, there is a need to understand these further. Investigating this area may aid in the more accurate interpretation of PSA and has potential implications on prostate cancer screening and detection in the population. Therefore, making use of a large national database, we extend the previous obesity-PSA hypothesis to evaluate whether PSA varies with the above physiologic and lifestyle factors associated with obesity among American men. Figure 1 presents our hypothetical framework for asking these questions.

We combined data from the National Health and Nutrition Examination Survey (NHANES) 2001-2002 and 2003-2004 study populations, two nationally representative cross-sectional surveys of the civilian noninstitutionalized population of the United States. The details of the procedures involved in sampling and data collection have been published elsewhere (15). All procedures were approved by the National Center for Health Statistics Institutional Review Board; a written, informed consent was obtained from all participants.

Of 3,108 men ages >40 y, those who underwent a recent rectal digital examination in the past week, prostate biopsy in the past 30 d, cytoscopy in the past 30 d, or had a history of prostate cancer or current prostate inflammation (n = 203) were ineligible for PSA testing in the NHANES and were excluded from this study (16) because these conditions cause PSA levels to be elevated. We also excluded those participants for whom there were no data available for the PSA eligibility criteria (n = 359) or with missing body mass index (BMI; n = 88), leaving 2,458 men in the analyses data set. These exclusion criteria are consistent with previously published data from the NHANES 2001-2004 study population (7).

Diet was estimated from the 24-h dietary recall during an in-person interview for both NHANES study populations. For the study participants of the NHANES 2003-2004, a second 24-h recall was done over the phone as a follow-up. However, to ensure consistency across the two surveys, we assessed the energy intake for these participants from the in-person 24-h recall. We determined the level of leisure-time physical activity from the responses to a series of questions asked in the interview about physical activity done during the past 30 d. The participants were considered to be physically active if they mentioned any leisure-time physical activity. To ascertain whether the men were vigorously active, they were queried whether they did any activity that caused heavy sweating or large increases in breathing or heart rate (e.g., swimming, aerobics, or fast cycling). For moderate physical activity, the participants were queried whether they did any activities that caused light sweating or a moderate increase in the heart rate, such as playing golf, dancing, bicycling for pleasure, or walking. The individuals who did not report participating in either moderate or vigorous physical activity were considered to be sedentary.

Participant age, race/ethnicity, smoking history, years of education, and history of doctor-diagnosed medical conditions, including hypertension, were self-reported in a structured personal interview (15). Serum PSA, triglycerides, and high-density lipoprotein concentrations were estimated from blood samples collected in the Mobile Examination Center exam. Fasting plasma glucose and insulin concentrations, measured in a subpopulation of participants >12 y examined in the morning session after an overnight fast of at least 8 h, were used to compute insulin resistance through the homeostasis model assessment algorithm, in which a value of 2.4 or higher was considered insulin resistant (17). Insulin concentrations were assayed with different laboratory methods Pharmacia methods in the NHANES 2001-2002 and Tosoh methods, in the NHANES 2003-2004, and respectively. Therefore, we used the regression coefficient [Y (Tosoh) = (1.0027 × Pharmacia values) − 2.2934] recommended by the NHANES to make the values comparable when combining the data from the two surveys. Weight, height, and waist circumference were measured in the Mobile Examination Center with the use of standardized procedures. Based on the Adult Treatment Panel III (18), metabolic syndrome was defined as the presence of three of five criteria: (a) waist circumference >102 cm in men, (b) self-reported physician-diagnosed hypertension, (c) high fasting glucose >100 mg/dL or 6.1 mmol/L, (d) hypertriglyceridemia >150 mg/dL or 1.695 mmol/L, and (e) low high-density lipoprotein cholesterol <40 mg/dL or 1.036 mmol/L in men. For this study, BMI [weight (kg)/height² (m)] was categorized as lean (<25), overweight (25-30), and obese (>30) per WHO definitions (19).

The serum PSA concentrations were not normally distributed; therefore, we log-transformed the PSA. We calculated the percent of total energy consumed from dietary fats and carbohydrate. The total energy, percent fat, and carbohydrate intakes were divided into tertiles, and physical activity was defined as a dichotomous variable (sedentary/moderate to intense activity). We computed the geometric mean PSA by the intake levels of energy, fat, and carbohydrate and the level of physical activity. We then computed the geometric mean PSA concentration for two physiologic factors associated with obesity: status of insulin resistance (dichotomous; yes/no) and metabolic syndrome (dichotomous; yes/no). We adjusted the geometric mean PSA concentrations for age, race, and BMI because these variables were associated with PSA in this sample (7, 20). Additionally, we included a variable for the survey year (NHANES 2001-2002 or NHANES 2003-2004) in the models. P values comparing the PSA for the highest and lowest tertiles of energy, carbohydrate, and fat intake, level of physical activity, and PSA among men with versus without insulin resistance and metabolic syndrome were calculated based on two-sample, unequal-variance t tests with Satterthwaite approximation of degrees of freedom. Further, for metabolic syndrome, we explored the association of PSA with the number of metabolic abnormalities as a continuous variable (with values ranging from 0 to 5) to investigate whether the cumulative effect of the biochemical derangements that are components of metabolic syndrome influence PSA concentrations. In separate models, we repeated all analyses for the three racial subgroups (non-Hispanic whites, non-Hispanic blacks, and Mexican Americans). Additionally, we explored the effect of the joint exposure of BMI categorized as lean (<25), overweight (25-30), obese (30-35), and very obese (>35) and each lifestyle or physiologic factor on PSA. Because the results were in a similar direction for the obese (BMI = 30-35) and very obese men (BMI = >35), we combined the two groups to increase power. The NHANES 2001-2004 sample weights were applied in all the analyses to account for individual selection probabilities, nonresponse, and poststratification that resulted from the complex survey design. All analyses were done with the use of SAS version 9 (SAS Institute, Inc.).

Nearly 80% of the study population was non-Hispanic white, with a median age of 52 years (Table 1). Three fourths of the population was either obese or overweight (BMI >25), with the median BMI being 28. At least half of the participants had PSA concentrations of 0.9 ng/mL, consumed ∼2,130 kcal/d, and reported being either moderately or intensely active. Further, in the subpopulation in which we were able to evaluate metabolic syndrome, 41% of the participants had metabolic syndrome, and about half the population had insulin resistance.

Higher energy intake was inversely associated with PSA concentrations after adjustment for age, race, and BMI (P = 0.0002). The tertile of dietary fat and carbohydrate intake (measured as percent energy) and level of physical activity, either when evaluated in three categories (sedentary, moderate, and intense; data not shown) or in two categories (sedentary versus moderately to intensely active) were not associated with PSA (Table 2). When all the analyses were repeated in racial subgroups, there was no change in the direction of the associations within each group (data not shown). The additional adjustment for the survey year (NHANES 2001-2002 or NHANES 2003-2004) did not change these associations.

Although the interactions between BMI and the lifestyle factors were not significant, because BMI is related with lifestyle, we explored the joint exposure of BMI and the three lifestyle factors to better understand their associations with PSA. We observed that, for men with BMI of 25 to 30, higher versus lower energy intake was associated with significantly lower PSA concentrations (P = 0.001). The inverse trend of PSA concentration and energy intake was similar for men with BMI >30 but not statistically significant (Table 3). Differently, the PSA concentrations were similar across the BMI categories when evaluated within each level of fat intake, carbohydrate intake, and physical activity (data not shown). Further, when these relationships were reevaluated separately within each racial subgroup, the directions of the associations were similar for each race (data now shown).

The PSA concentrations were determined among people with and without insulin resistance and metabolic syndrome (n = 1,200). The PSA concentrations were lower among men with insulin resistance (P = 0.04) after adjustment for age, race, and BMI but not among men with metabolic syndrome. The additional adjustment for the survey year (NHANES 2001-2002 or NHANES 2003-2004) did not change these associations. For exploratory purposes, when we investigated the relation of PSA with the two physiologic factors within three BMI categories (<25, 25-30, >30), we noted that the PSA concentrations did not differ by insulin resistance or metabolic syndrome (data not shown). Further, when the number of metabolic abnormalities, treated as a continuous variable, was evaluated in relation with PSA, we noted no associations in the whole population (P = 0.12). These relationships were consistent when evaluated within each racial subgroup (data not shown).

The purpose of this study was to identify physiologic and lifestyle factors associated with obesity that may explain in part the previously observed inverse associations between obesity and PSA (2-7). Werny et al. (7) reported that PSA concentrations were lower among men with higher BMI in the NHANES 2001-2004 population. We now extend these investigations in the same nationally representative population to evaluate the association of PSA concentrations with obesity-related lifestyle factors (diet and physical activity) and physiologic factors (metabolic syndrome and insulin resistance).

The data from this large sample of American men who had not undergone a recent digital rectal examination, prostate biopsy, or cytoscopy, or had a history of prostate cancer or prostatic inflammation suggest that higher energy intake was inversely associated with PSA concentrations. Furthermore, when we explored the joint exposure of higher body weight and higher energy intake, we noted that the PSA concentrations were lower among men with a BMI of 25 to 30 who reported higher energy intakes (n = 1,067; P = 0.001), possibly suggesting that energy and PSA are associated independently of BMI. A similar association that was not statistically significant was observed among men with BMI >30 (n = 746; P = 0.09). This specific result, implying that the PSA levels of obese or overweight men consuming higher energy may be lower than the actual value, has potential public health significance because the majority of American men are overweight or obese (1), as reflected in this present study sample (median BMI = 28 and waist size is 102 cm). However, we cannot rule out the possibility that the inverse association between energy and PSA among overweight men may be obtained due to chance. It must be noted that, in this data set, energy intake was measured through a 24-hour recall and, like any other dietary assessment tool, this may contribute to measurement error in the estimation of energy intake. Therefore, the inverse associations that we note between energy and PSA concentration in this population may be attenuated due to the misclassification of energy intake.

Our finding of an inverse association between PSA and self-reported energy intake differs from the results of the large Prostate Cancer Prevention Trial, a volunteer-based trial that noted no associations between PSA and quintile of energy intake estimated from a food frequency questionnaire in the past year (12).

A plausible explanation for the lower PSA concentrations among men with higher energy intakes in this sample may be due to inverse associations of energy and testosterone concentrations, which modulate PSA production in the prostate gland. In laboratory animals, long-term energy restriction enhanced the production of serum testosterone compared with rats fed ad libitum, thereby suggesting that higher energy intake may lower testosterone concentrations (21). In a small study (n = 36), Fontana et al. (22) reported a nonsignificant suggestion of lower serum testosterone in men on typical high-caloric western diets. Further, it has been shown that PSA gene expression is closely linked with free testosterone concentrations (23). However, there may be unidentified alternate biochemical mechanisms through which higher energy intakes may relate to PSA concentration.

We also note that men with insulin resistance had significantly lower PSA concentrations compared with men without insulin resistance after adjusting for age, race, and BMI. The underlying mechanisms are unclear at this time; however, insulin resistance may also be associated with lower testosterone concentrations (24), which, in turn, may relate to lower PSA production as outlined in Fig. 1. When we investigated the PSA concentrations and insulin resistance within BMI categories, PSA was not related to insulin resistance. This could possibly be due to reduced power because insulin resistance was only estimated in a subsample of the NHANES participants (n = 1,200). Differently from the associations observed between insulin resistance and PSA, we did not find any difference in the PSA concentrations among men with or without metabolic syndrome, or the number of metabolic abnormalities present among these men.

We also observed no associations between PSA and percent energy from fat or carbohydrate intake and level of physical activity in this data set. These observations were consistent with the lack of association between fat and carbohydrate intake and PSA in the Prostate Cancer Prevention Trial (12). Similarly, data from two other prospective dietary intervention trials (25, 26) among men who were randomized to low-fat, high-fruit, and high-vegetable diets showed no effect of such diets on PSA concentrations compared with control diets higher in fats. Furthermore, consistent with other studies, the PSA concentrations did not differ with level of physical activity in this population (12, 13).

Some limitations need to be considered when interpreting the results of our study. The NHANES 2001-2004 is a cross-sectional survey that ascertained PSA concentrations at the same time as diet, physical activity, and blood levels. As in all dietary assessment tools used in nutritional epidemiology studies, there is likely to be the unavoidable measurement error of dietary energy, fat, and carbohydrate intake. However, we minimize this bias in our analyses by drawing conclusions based on the comparison of extreme groups of energy, fat, and carbohydrate intakes. Despite our analytic strategy, residual confounding relating to measuring diet imperfectly or over only short periods of time (24 hours) is likely to remain so that the associations of dietary intake to PSA concentrations are likely to reflect this fact, at least in part.

Although the NHANES anthropometric measurements were standardized and field tested, the BMI may be subject to random measurement error, biasing results toward the null. The levels of physical activity in this data set were self-reported. The lack of association between physical activity levels and PSA may be due to the misclassification of physical activity levels. Further, the level of physical activity was queried for the past 30 days; therefore, we were unable to determine the longer-term physical activity level among these participants. Another limitation of this study is that there might be about 24% men with undiagnosed prostate cancer, as suggested by the Prostate Cancer Prevention Trial study, a large clinical trial (27). Lastly, there may also be undetected prostate infection and inflammation in this population. The inability to exclude men with these conditions might cause the PSA concentrations to be higher in our study population.

In conclusion, our findings of inverse associations between two factors associated with obesity, energy intake and insulin resistance, and PSA concentrations are provocative and may have important implications on prostate cancer detection among American men. PSA is a common test used to screen for prostate cancer, and obesity is a public health concern in the United States, with about 74% to 78% of the men, 40 years or older, being either overweight or obese (1). These results suggest that insulin-resistant men and men who are on higher energy intakes are less likely to have elevated PSA and may not reach the biopsy threshold. Accordingly, PSA velocity may be more valuable in identifying progressive prostate cancer among these men. The biochemical mechanisms of how higher energy intake modulates a decrease in PSA production is unclear at this time, and further studies are required to investigate this issue. The interrelationships of PSA, obesity, diet, insulin resistance, and metabolic syndrome require clarity in other populations, specifically in studies making use of a prospective design with better measures of diet and physical activity. The confirmation of these relationships in additional studies may lead to a more accurate interpretation of this important screening analyte.

No potential conflicts of interest were disclosed.

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.

We thank the National Center of Health Statistics, the participants of the National Health and Nutrition Examination Survey (2001-2004), and Thanusha Puvananayagam for her research and technical support.