Serum Retinol and Prostate Cancer Risk: a Nested Case-Control Study in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial

Vitamin A is a fat-soluble vitamin that plays an essential role in the visual cycle and is required in the normal growth of bone, reproduction, embryonic development, and in differentiation of epithelial tissues (1). There has also been a great deal of interest in the cancer-preventive potential of vitamin A (retinol), much of which stems from in vitro and animal studies showing a key role for retinol in regulating the growth, differentiation, and apoptosis of normal and malignant cells.

Several observational epidemiologic studies and one randomized clinical trial have examined the associations of vitamin A and prostate cancer, although no consistent association with prostate cancer risk has been established. Prospectively designed nested case-control studies have reported no association between circulating concentrations of retinol at baseline and subsequent prostate cancer risk (2-6), whereas others have reported an increased (7) or decreased (8, 9) risk associated with higher retinol concentrations (Fig. 1). However, many of the previous studies had small sample sizes, and few examined the associations of retinol concentration and prostate cancer risk by stage or grade of disease.

This brief report examined whether serum retinol concentrations were associated with prostate cancer incidence in a large nested case-control study within the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial, based on healthy men ages 55 years and older who were screened for prostate cancer regularly, following a standardized protocol.

The study setting and population have been described previously (10). In brief, this nested case-control study was limited to men randomized to the screening arm of the PLCO trial, who were offered prostate cancer screening by serum prostate-specific antigen (PSA) at entry and annually for 5 y, and digital rectal examination (DRE) at entry and annually for 3 y. Men with a positive screening result (PSA ≥4 ng/mL or DRE suspicious for prostate cancer) were referred to their medical care providers for diagnostic evaluation, and follow-up for diagnosis of cancer was conducted by annually mailed questionnaires. Medical and pathologic records related to prostate cancer diagnoses were acquired for all men suspected for prostate cancer by screening examination or annual questionnaire.

The men in this study had no prior history of prostate cancer and at least one blood collection and one valid prostate cancer screen prior to October 1, 2001 (the censor date). All men were followed from their initial prostate cancer screen (PSA and/or DRE), to the occurrence of prostate cancer, loss-to-follow-up, death (by National Death Index), or censor date, whichever came first. Cases were men diagnosed with adenocarcinoma of the prostate. Aggressive cancers were defined as clinical stages III or IV (regionally invasive or distant metastatic disease) or biopsy Gleason Sum ≥7 based on the pathologic report.

A total of 1,320 prostate cancers were identified, from which non-Hispanic Black cases and cases diagnosed in the first year after blood draw were excluded, leaving a total of 803 cases. Controls were frequency-matched by age at entry (5-year intervals), time since initial screening (1-year intervals), year of blood draw, and race/ethnicity using incidence-density sampling (11) with a case-control ratio of 1:1.2. Serum collected at study entry was available for 692 (86.2%) cases and 844 (88.9%) controls.

At enrollment, all participants were asked to complete a questionnaire about sociodemographic factors, medical history, and risk factors for cancer. Usual dietary intake over the 12 mo prior to enrollment was assessed with a 137-item food frequency questionnaire including 14 additional questions about intake of vitamin and mineral supplements and 10 additional questions on meat cooking practices (12). Daily dietary nutrient intake was calculated by multiplying the daily frequency of each consumed food item by the nutrient value of the sex-specific portion size, using the nutrient database from the U.S. Department of Agriculture (13). Total vitamin and mineral intakes were calculated using the sum of dietary and supplement intakes.

Nonfasting blood specimens collected at the clinical centers were processed and frozen within 2 h of blood draw and stored at −70°C. Serum concentrations of total retinol (bound and unbound) were determined using reversed-phase high-performance liquid chromatography, with UV detection (14). Cases and their matched controls were analyzed in the same batch. Blinded quality control samples were randomly inserted into each batch. The coefficient of variation estimated from 171 blinded duplicates was 5.1%. To investigate the reproducibility of serum retinol concentrations over time we also included a second serum sample drawn 1 year after study entry in the same sample batch as the subject-paired serum sample collected at study entry for a subset of 46 controls. Serum concentrations of selenium and cholesterol were measured for another nested case-control study, and the methods have been described previously (15). Serum cholesterol was measured enzymatically by a standard procedure at 37°C on a Hitachi 912 analyzer.

Conditional logistic regression was used to estimate odds ratios and 95% confidence intervals of prostate cancer, with serum retinol concentrations categorized as quintiles, based on the distribution among controls. Tests for linear trend were based on quintile-specific median values expressed as a continuous variable. Analyses were conditioned on matching factors (age, time since initial screening, and year of blood draw), and adjusted for study center. In addition, we evaluated potential confounders based on a priori hypotheses for prostate cancer risk factors [PSA; DRE screening; family history of prostate cancer; educational attainment; physical activity; body mass index; aspirin and ibuprofen use; history of diabetes; smoking; intakes of alcohol, energy, fat, red meat, heterocylic amines from meat (2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine or PhIP), fruits, vegetables, cruciferous vegetables, vitamin E, calcium, serum selenium, serum cholesterol; month of blood draw], but none were included in the analyses because the factors, either separately or together, did not change the risk estimates by more than 10%. To assess effect modification we did stratified analyses and evaluated the statistical significance of multiplicative interaction terms by comparing the −2 log-likelihood statistics of the main effect model and the model including the interaction term. Spearman correlation coefficients were calculated to measure the correlation between serum retinol concentrations obtained 1 y apart in 46 controls and between serum concentrations and dietary vitamin A intake in all controls. Correlation coefficients were adjusted for month of blood draw, serum cholesterol concentration, smoking, body mass index, age, and energy intake. All P values are two-sided and significance is set at the 0.05 level.

The distribution, overall and by retinol concentrations, of demographic and health-related covariates possibly linked to prostate cancer risk is shown in Table 1. The average age at study entry among controls was 64.7 years and did not vary by retinol concentration. Compliance with the PLCO screening protocol was very high and did not differ by serum retinol concentrations, as shown by the average number of cancer screens (PSA or DRE; Table 1). Men with higher serum retinol concentrations were less likely to have a personal history of diabetes, and had higher serum cholesterol concentrations, were less obese, more likely to smoke, more physically active, and more likely to use aspirin. Men with higher serum retinol concentrations also had a lower intake of fat and red meat, and higher intakes of PhIP, vitamin D, supplemental vitamin E, and alcohol. The average time between blood collection and prostate cancer diagnosis (among cases) was 2.4 + 1.4 years. The correlation between serum retinol concentrations measured in 46 men at study entry and 1 year later was modest (r = 0.38; P = 0.009), and the correlation between serum retinol and vitamin A intake from diet and supplements among controls was weak (r = 0.08, P = 0.06).

Median serum retinol concentrations did not differ significantly between cases and controls (67.11 versus 68.52 ug/dL; P = 0.22). In multivariate analyses comparing men in the highest to the lowest quintiles, serum retinol was associated with a nonsignificant 20% reduction in the risk of prostate cancer with no indication of a linear trend (P_trend = 0.11; Table 2). When restricted to aggressive prostate cancers (Gleason Sum >7 or clinical stage III or IV), high concentrations of serum retinol were associated with a significant reduction in risk (odds ratio, 0.58; 95% confidence interval, 0.36-0.92 for highest versus lowest quintile; P_trend = 0.02; Table 2), with the strongest association for high-grade disease (Gleason Sum >7; odds ratio, 0.52; 95% confidence interval, 0.32-0.84 for highest versus lowest quintile; P_trend = 0.01; Table 2). For both aggressive and high-grade prostate cancers, the greatest decrease in risk occurs between the first and the second quintiles, with risk being similar for quintiles 2 through 5.

Similar associations between serum retinol and overall prostate cancer and aggressive disease were observed for subgroups characterized by year of diagnosis (<4 and > 4 years), age (<65 and >65 years), smoking (ever versus never), family history of prostate cancer (yes versus no), alcohol (0-2.9, 3-14.9, and ≥15 g/day), and serum vitamin D (<49.9, 50.0-68.7, and ≥68.8 nmol/L; data not shown). Furthermore, odds ratios did not vary when these analyses were limited to the years of active screening (years 1-5) or when the second year of follow-up was excluded (the first year of follow-up was already excluded by design; data not shown).

In this prospective analysis of 692 incident cases, we observed a decreased risk of aggressive prostate cancer, particularly high-grade disease, with increased concentrations of serum retinol. This association was not modified by age, smoking, family history, alcohol intake, or vitamin D.

Our findings of a decreased risk of aggressive prostate cancer with increased concentrations of serum retinol are supported by the results from two earlier cohort studies; in the National Health and Nutrition Examination Survey I Epidemiologic Follow-up Study, high retinol concentrations were associated with a decreased risk of prostate cancer (8), and in CLUE I a nonsignificant inverse association between serum retinol and prostate cancer was reported (9), although this finding was not confirmed on further follow-up (6).

Analyses in many of these studies were limited to total prostate cancer, and therefore, differences in associations according to stage or grade of disease may have been obscured (2, 4-6, 8, 9). For prostate cancer, this is particularly important because the disease is heterogeneous and the significance of early-stage (PSA-detected) disease is unclear (16). Two larger prospective studies (n = 578 and n = 966 cases) analyzed nonaggressive and aggressive cancers separately, although neither found an association between retinol concentrations and the risk of aggressive (stage III or IV or Gleason sum ≥7 (7)) or advanced (T₃ or T₄, N₁₊, M₁ or metastatic (3)) disease, and one reported an elevated risk for high levels of retinol that was limited to nonaggressive (stage I or II and Gleason sum <7) disease (7). In the current study, the association of a decreased risk with increasing retinol concentrations was limited to aggressive disease and was strongest for high-grade disease. This association did not differ by year of diagnosis (<4 versus > 4), suggesting that a reverse-causality effect on retinol is unlikely. These findings could indicate that retinol has a specific effect on disease progression (17), although it is also possible that this association is due to chance as our study is the only one reporting an inverse association with aggressive disease specifically.

Our findings are supported by several in vitro and animal models that show a key role for retinol in regulating the growth, differentiation, and apoptosis of normal and malignant prostate cells (18, 19). In experimental models, retinoids suppress the transforming effects of carcinogens, inhibit the growth of premalignant cells, enhance differentiation of malignant cells, and induce apoptosis (20). These regulatory effects are mediated through the activation of two families of retinoid nuclear receptors, retinoic acid receptors and retinoid X receptors. Prostate cancer tissue has lower levels of retinoic acid (21), and low retinol levels could also directly alter the expression of retinoic acid and retinoid X receptors (22), which have been shown to be selectively reduced in prostate cancers (23).

It is unlikely that these findings can be solely attributed to differences in dietary or supplemental retinol intake, given that serum retinol concentrations are under tight homeostatic control (24), and neither dietary intake nor supplements are strongly correlated with retinol concentrations (25, 26). It should be noted, however, that the range of retinol concentration in this study population is wide (27.4 to 262.6 ug/dL), and other factors such as body mass index, physical activity, and dietary fat and alcohol intake are correlated with retinol concentrations, suggesting that in a well-nourished population, these factors influence retinol concentrations to some degree. Lower concentrations of retinol may reflect insufficient vitamin A stores in the liver, although this is unlikely given the healthy study population. Alternatively, low circulating retinol concentrations could result from underlying aberrations in cellular retinol binding protein metabolism (27). Cellular retinol binding protein plays an essential role in the maintenance of hepatic retinol stores and the synthesis of retinoic acid from retinol (24), and the reduction or loss of cellular retinol binding protein 1 expression has been shown in prostate carcinomas (28).

A limitation of this study and previous observational studies is the use of a single baseline measure of serum retinol. Although one examination of the repeatability of serum retinol concentrations 15 years apart showed moderate correlation (r = 0.58; 95% confidence interval, 0.46-0.67 (29)), the correlation between serum retinol concentrations 1 year apart in this study was somewhat lower (r = 0.38; P = 0.009), and thus a single measure of serum retinol may not be representative of long-term retinol exposure. Nonetheless, such nondifferential measurement error in serum retinol generally leads to an attenuation of the estimated association, and the true protective effect should be stronger than that found in our study. A further limitation is the relatively short follow-up of 8 years.

The strengths of this study include standardized procedures for prostate cancer screening and very high compliance with the screening protocol, availability of prediagnostic serum samples, a large sample size, and detailed diagnostic data which allowed stratification by stage and grade and simultaneous adjustment for multiple confounders.

In summary, our results support a protective association between serum retinol concentrations and the risk of aggressive prostate cancer. Further research to better understand the significance of elevations in serum retinol concentrations and the possible effects of aberrations in cellular retinol binding protein metabolism on prostate cancer risk would be useful.

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.