Associations of Subtype and Isomeric Plasma Carotenoids with Prostate Cancer Risk in Low-Income African and European Americans

The biological connection of retinoids and carotenoids with cancer development has been associated with their roles in immune system stimulation, gap junction communication (1), cell-cycle regulation and apoptosis, modulation of growth factors, cell differentiation (2), and various receptors and adhesion molecules (3), although results from epidemiologic studies remain controversial and inconsistent for some cancers, including prostate cancer. These inconsistencies could be due to characteristic differences in the population under study, such as race, multivitamin supplement usage, tumor stage, smoking status, and/or subtypes of retinoids and carotenoids.

One large pooled analysis, based on 15 prospective and intervention trial studies (4), found positive associations between circulating levels of lycopene and retinol with risk of aggressive prostate cancer. However, the associations were primarily driven by studies of Caucasians, including Finns, Norwegians, Australians, and European-Americans, with the exception of 1 cohort, which included a respective 218 and 438 African-American cases and controls. In the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) trial, consisting of 29,133 Finnish male smokers, β-carotene supplementation was not associated with prostate cancer incidence (5). The associations of retinoids and carotenoids with risk of prostate cancer have not been well-studied in low-income African-American and European-American populations. In a study of 1,079 European-Americans (164 high- and 915 low-aggressive prostate cancer) and 1,023 African-Americans (206 high- and 817 low-aggressive prostate cancer), Antwi and colleagues (6) reported that dietary lycopene and β–cryptoxanthin were inversely associated with aggressive status of prostate cancer in both races.

Carotenoid subtypes and isomeric arrangements may explain the reported inconsistent associations between circulating carotenoid levels and risk of prostate cancer. However, to our knowledge, no previous study has examined the associations by molecular configuration and/or isomeric occurrence. Overloading on antioxidant multivitamins, especially hepatic-storable lipid-soluble retinol and carotenoid antioxidants, can lead to prooxidant properties (7). Provitamin A carotenoids, including α-carotene, β-carotene, and β-cryptoxanthin, can be converted to retinol (5) but cannot lead to vitamin A toxicity due to enzymatic control over retinol synthesis in the body. Nonprovitamin A carotenoids, including lycopene, lutein, and lutein's isomer zeaxanthin, are not converted to retinol and have biological effects independent of vitamin A activity, particularly in the eyes, offering an advantage of chemoprevention without the associated toxicity from retinoids (6). In addition, the major circulating carotenoids in humans consist of hydrocarbons with (β-cryptoxanthin and lutein/zeaxanthin, which contain hydroxyl groups) or without (β-carotene, α-carotene, and lycopene) oxygen groups. Oxygen-containing carotenoids are more polar than carotenes, generally located on the surface of lipoproteins, and more likely to undergo rapid surface transfer (8). Most carotenoids occur in large varieties of cis-trans (or E/Z) geometrical configurations. However, it is not completely understood whether carotenoid configurations have specific biological properties (9).

This study is aimed at evaluating associations between plasma retinol and carotenoid levels with prostate cancer risk according to their bioactive forms, including α and β subtypes and cis and trans isomers, and exploring whether the associations differ by ethnicity, smoking status, and multivitamin supplementation.

We conducted a case–control study nested within the Southern Community Cohort Study (SCCS), a prospective cohort study investigating health disparities among predominantly low-income African-Americans and European-Americans across 12 Southeastern states. Details about the SCCS have been published previously (10). All participants provided written informed consent, and the study was approved by the Institutional Review Boards (IRB) at Vanderbilt University (Nashville, TN) and Meharry Medical College. Blood samples were collected at baseline recruitment in 10-mL EDTA tubes and 10-mL serum tubes. The vast majority (>95%) of blood samples were processed within 24 hours after collection and stored at −80°C. The plasma samples were used for all assays in this study.

Incident prostate cancer cases from nearly 34,000 male participants, aged 40 to 79 years at enrollment between 2002 and 2009, were identified through linkages with state tumor registries and/or National Death Index mortality records through January 2016. A total of 879 incident prostate cancer cases were identified from cohort participants. Among them, 347 had blood samples collected at baseline and were included for this study. Two controls per case were randomly selected from SCCS participants who provided a blood sample at the baseline survey. Matching criteria included race (African-American/European-American/others), enrollment site, age at enrollment +/− 2 years (relaxed to 5 years, when necessary), and date of blood sample collection +/- 6 months (relaxed to 9 months, when necessary). A total of 653 controls were included in the study. After exclusion of subjects without race information (n = 2) and other ethnicities (n = 15), a total of 343 cases and 640 controls (293 cases and 541 controls for African-Americans, and 50 cases and 99 controls for European-Americans) participated in this study. The average time between cohort entry and prostate cancer diagnosis was 4.4 years.

Plasma concentrations of retinol and carotenoids were determined by high-pressure liquid chromatography (HPLC) with photo-diode array detection, as applied previously (11). Briefly, 0.15 mL plasma was mixed with an ethanolic solution of 3 internal standards (one each for retinoids, carotenoids, and tocopherols), followed by partitioning into hexane, drying, and redissolving in 0.15 mL of the HPLC mobile phase, as detailed previously (12, 13). Ten microliters were injected onto a Spherex C18 analytic column (150 mm × 3.2 mm, 3 μm), coupled to a Spherex C18 precolumn (4 mm × 3.0 mm, 10 μmol/L; Phenomenex), using isocratic elution with a mobile phase of 650 mL methanol, 250 mL dichloromethane, 100 mL acetonitrile, and 2 mL aqueous bis-tris propane (0.5 mol/L, pH 6.8; Sigma-Aldrich), and containing 0.25 g/L butylated hydroxytoluene (BHT; 1.14 mmol/L; Sigma-Aldrich), at an elution rate of 0.23 mL/minute. Retinol and carotenoid concentrations were determined by monitoring at 325 nm and 450 nm, respectively. Detection limits were approximately 1 ng/mL. Inter-assay variability was between 4% and 10% for the major analytes, at levels of 100 to 500 ng/mL. The assay was validated by participating in the quality assurance program for fat soluble vitamins, organized by the National Institute for Standards and Technologies (Gaithersburg, MD). The intraassay and interassay coefficients of variability were <7% and 12%, respectively, for all measurements.

All information on covariates was based on self-reported data gathered at baseline. Education was categorized across five groups as <9 years, 9–11 years, 12 years, some college or training, and ≥graduated college. Annual household income was divided across three groups as <$15,000, $15,000–$24,999, and >$25,000. Family history of prostate cancer referred to first degree relatives, i.e., father and brother(s). The Nam-Powers-Boyd (NPB) occupational index, derived based on participants' longest-held occupation, was used to determine occupational status. For chronic disease history, the comorbidity index, a modified Charlson index (14) ranging from 0 to 12, was adopted. Scores were grouped as 0, 1, 2, 3, and ≥4. Body mass index (BMI) was categorized based on World Health Organization (WHO) criteria as <18.5, 18.5–24.9, 25–29.9, 30–34.9, and ≥35 kg/m². Lifestyle factors included smoking (never, former, and current), alcohol consumption (never and 3 groups by amount of daily use), total daily activity (TDA; quartile), and healthy eating index (HEI; quartile) scores. Dietary information was collected using a validated food frequency questionnaire (FFQ; ref. 15). To calculate HEI scores, SCCS FFQ data was linked to the MyPyramid Equivalents Database (version 2.0) to generate equivalent intakes (cup or ounce equivalent per 1,000 kcal) for food groups listed in the U.S. Dietary Guidelines for Americans. The equivalent intake for each food group was then scored using the standard HEI-2010 or HEI-2005. The reference SAS code used for calculating HEI scores was provided by the USDA (16). Multivitamin supplementation was based on any multivitamin use. To avoid a reduced number of informative covariates, we assigned missing values with the high frequency (for categorical variables) or median (for continuous variables).

After adjustment for potential confounders including education, family income, history of diabetes, family history of prostate cancer, NPB score, comorbidity index, BMI, smoking, alcohol consumption, TDA, HEI, and retinol and total carotenoid levels, associations of retinol and carotenoids with prostate cancer risk were assessed and presented by two approaches. We first compared geometric means of retinol and carotenoid plasma levels between cases and controls with adjustment for potential confounders using the general linear model (GLM), followed by conditional logistic regression to compute OR and 95% confidence intervals (CI). Trend test was conducted by assigning an ordinal score to each exposure level of a categorical variable and treating the ordinal score as continuous scale in the model. To minimize potential false positive findings due to multiple comparisons, we made a Bonferroni correction based on the number of items examined. In total, we assessed 8 categories of retinol and carotenoids (carotene, cryptoxanthin, lycopene, dihydrolycopene, lutein, anhydrolutein, and zeaxanthin) and set P values of <0.006 (0.05/8) as statistically significant. Stratified analyses were also carried out to explore whether the associations varied by race (African-American vs. European-American), smoking status (never and former vs. current), and ever-users of multivitamins (yes vs. no). For analyses evaluating potential modifications of smoking status and multivitamin supplement use, we included an interaction term between each plasma vitamin level and smoking status or multivitamin supplement use (yes/no) by alternating the codes 0 and 1 for the stratifiers to generate stratum-specific estimates, to avoid the breakdown of matched sets in the conditional logistic regression model. Additional analyses were conducted among cases with stage II–IV prostate cancer, cases with stage III–IV prostate cancer, and by Gleason score, separately. All analyses were performed using SAS version 9.3 (SAS Institute).

For prostate cancer cases included in the study, the distributions of tumor–node–metastasis (TNM) stages were: T1_a-c, 26.9%; T2_a-b, 7.9%; T2_a-b, 1.5%; N0, 31.2%; N1, 0.9%; M0, 32.6%; and M1, 2.0%. Gleason score percentages were: <6, 30.7%; 7, 31.5%; ≥8, 9.6%; and missing, 28.3%. There were only 39 stage III and IV prostate cancer cases, 33 of whom had a Gleason score ≥8, which is considered to be aggressive prostate cancer (17).

Sociodemographic factors including education, family income, occupational status, and comorbidity were not associated with prostate cancer risk. The lowest BMI (<18.5 kg/m²) was associated with increased prostate cancer risk; however, the statistics were based on a small number of participants in this category. Smoking, alcohol drinking, and sleep duration were not significantly associated with prostate cancer risk. Higher scores for HEI (OR _{Q1 vs. Q3} = 0.63; 95% CI, 0.42–0.95) were inversely associated with prostate cancer risk (Table 1). Associations of demographic and lifestyle factors with retinol and total carotenoids are shown in Supplementary Table S1.

Table 2 displays comparisons of case–control differences across 8 major categories of retinol and carotenoids, as well as in α, β subtypes and cis or trans isomeric occurrence within 4 carotenoid subcategories. Some case–control differences in geometric means were noted; however, not all reached multiple comparison adjusted significance (P < 0.006). Plasma levels of β-carotene and lutein, occurring predominantly as trans isomers (P_adj = 0.001 and 0.017, respectively), were higher in cases than controls and were positively associated with the risk of prostate cancer (OR _{T1 vs. T3} = 2.13; 95% CI, 1.32–3.43; P_trend = 0.002 and OR _{T1 vs. T3} = 1.86; 95% CI, 1.20–2.88; P_trend = 0.006, respectively). A similar positive association was observed for β-cis-carotene (OR _{T1 vs. T3} = 1.65; 95% CI, 1.06–2.59; P_trend = 0.030). On the other hand, plasma levels of lycopene, both trans and cis isomers, were inversely associated with prostate cancer risk (OR _{T1 vs. T3} = 0.42; 95% CI, 0.18–0.96; P_trend = 0.040; and OR _{T1 vs. T3} = 0.23; 95% CI, 0.09–0.56; P_trend = 0.001, respectively). Dihydrolycopene levels were inversely associated with prostate cancer risk (OR _{T1 vs. T3} = 0.37; 95% CI, 0.18–0.74; P_trend = 0.006), with average plasma levels of 54.4 ng/mL for cases and 57.8 ng/mL for controls (P_adj = 0.016). A similar inverse association was observed for cis-anhydrolutein (OR _{T1 vs. T3} = 0.57; 95% CI, 0.33–0.96; P_trend = 0.037). The positive associations of β-carotene and lutein and inverse associations of lycopene and dihydrolycopene with prostate cancer risk persisted after excluding cases diagnosed within 1 or 2 years (Supplementary Table S2). In the analysis restricted to stage II to IV prostate cancer, positive associations were found for plasma levels of β-carotene (P_trend = 0.008), in both trans and cis isomers (P_trend = 0.009 and 0.032, respectively), and trans-lutein (P_trend = 0.005). Inverse associations were found for lycopene (P_trend = 0.003), in both trans and cis isomers (P_trend = 0.013 and 0.001, respectively), and dihydrolycopene (P_trend = 0.004; Table 2). We conducted additional analyses restricted to stage III and IV cases and found similar association patterns; however, none of the risk estimates reached statistical significance due to a small sample size (Supplementary Table S3). Results stratified by Gleason score are shown in Supplementary Table S4.

Subgroup analyses stratified by race showed that mean β–trans-carotene (P_adj = 0.001) and trans-lutein (P_adj = 0.019) levels were higher in African-American cases than controls, with OR _{T1 vs. T3} = 2.10 (95% CI, 1.25–3.52) for β–trans-carotene and OR _{T1 vs. T3} = 1.89 (95% CI, 1.18 to 3.02) for trans-lutein. In contrast, dihydrolycopene levels were inversely associated with prostate cancer risk in African-Americans (OR _{T1 vs. T3} = 0.27; 95% CI, 0.12–0.59; P_trend = 0.001; ref. Table 3). No associations were observed among European-Americans. However, the sample size for European-Americans was small (n = 50 case–control sets). Tests for interaction were not significant.

Results of additional analyses stratified by smoking status and multivitamin supplement use are presented in Tables 4 and 5. β–Trans- and β–cis-carotene plasma levels were higher in cases than controls in never/former smokers (P_adj = 0.019 and 0.043, Table 4). Positive associations were observed for β-trans-carotene in both never/former smokers (OR _{T1 vs. T3} = 2.01; 95% CI, 1.12–3.60; P_trend = 0.017) and current smokers (OR _{T1 vs. T3} = 2.44; 95% CI, 1.30–4.56; P_trend = 0.008) and for β-cis-carotene and trans-lutein among never/former smokers (OR _{T1 vs. T3} = 2.00; 95% CI, 1.11–3.60; P_trend = 0.020 and OR _{T1 vs. T3} = 2.33; 95% CI, 1.32–4.12; P_trend = 0.003). Inverse associations were observed for cis-lycopene (P_trend = 0.022 and 0.001 in never/former and current smokers, respectively) and dihydrolycopene (P_trend = 0.007 and 0.032 in never/former and current smokers, respectively), regardless of smoking status (Table 4). Cis-anhydrolutein was negatively associated with the risk of prostate cancer in current smokers only (OR _{T1 vs. T3} = 0.52; 95% CI, 0.27–1.02; P_trend = 0.035).

β-trans-carotene levels were higher in multivitamin users than in subjects without multivitamin supplementation, and a positive association of β-trans-carotene with risk of prostate cancer was observed regardless of multivitamin supplementation status (OR _{T1 vs. T3} = 1.94; 95% CI, 1.13–3.32; P_trend = 0.018 in nonusers, OR _{T1 vs. T3} = 2.60; 95% CI, 1.28–5.29; P_trend = 0.005 in users; Table 5). Plasma levels of cis-lycopene (OR _{T1 vs. T3} = 0.19; 95% CI, 0.07–0.51; P_trend = 0.001), dihydrolycopene (OR _{T1 vs. T3} = 0.30; 95% CI, 0.14–0.64; P_trend = 0.002), and cis-anhydrolutein (OR _{T1 vs. T3} = 0.50; 95% CI, 0.28–0.90; P_trend = 0.023) were inversely associated, while plasma trans-lutein levels were positively associated with the risk of prostate cancer in subjects with no multivitamin supplementation (OR _{T1 vs. T3} = 1.97; 95% CI, 1.16–3.33; P_trend = 0.014; Table 5).

In this nested case–control study within the SCCS, we found associations of plasma β-trans-carotene, cis-lycopene, dihydrolycopene, and trans-lutein with prostate cancer risk after adjusting for multiple comparisons. Trans isomers for β-carotene were found to be nominally and positively associated with prostate cancer risk, predominantly in African-Americans and regardless of smoking status and multivitamin use. The positive association of plasma β-trans-lutein was observed in African-Americans, current and former smokers, and subjects without multivitamin supplement use. The inverse associations of plasma lycopene and dihydrolycopene with the risk of prostate cancer were more evident among African-Americans, never or former smokers, and subjects without multivitamin-supplement use. The inverse association of cis-anhydrolutein was restricted in subjects without multivitamin-supplement use. To our knowledge, this study is the first to examine associations between subtypes (α and β) and carotenoid isomers (cis and trans) on the risk of prostate cancer.

Retinol, the active form of vitamin A, has been proposed as a candidate for prostate cancer prevention because of its effect on regulating growth, differentiation, and apoptosis of normal and malignant cells through regulation of DNA transcription and DNA polymerase activity, or increasing levels of other antioxidants, such as selenium and α-tocopherol (18). However, retinol can also stimulate growth and dedifferentiation of prostate cells, thus negating any anticancer activity (19). Although some studies have reported positive (19, 20) or inverse (21) associations between circulating retinol and prostate cancer, we failed to find these associations after accounting for race, smoking status, and multivitamin use in a low socioeconomic and predominantly African-American population. Our findings are supported by results from a recent pooled analysis, showing a null association (4). Shiels and colleagues (22) found a smoking–retinol interaction on sex hormone levels, in which male smokers were shown to have higher levels of sex steroid hormones, such as testosterone, and suggested retinol may act synergistically with smoking through this mechanism. In our study, however, we did not observe an interaction between plasma retinol level and smoking status on prostate cancer risk. It is worth noting that circulating retinol levels are highly regulated by the liver, and thus, retinol may not be a good biomarker for vitamin A status. This may explain the large null results found in our and previous studies.

A pooled analysis reported positive associations for α- and β-carotene subtypes with prostate cancer risk in unadjusted analyses, but the associations were attenuated by sociodemographic and lifestyle factors (4). In our study, total carotene levels were higher, rather than lower, among prostate cancer cases compared with controls, consistent with results from the Multiethnic Cohort (MEC) study (23), which included less than 47% African-Americans, in which significant positive associations were found for overall and advanced prostate cancer. The significant positive associations we found for carotene, both trans and cis isomers, on prostate cancer risk were predominantly observed in the β but not α subtype. Positive associations for β-carotene were predominately seen in stage II–IV prostate cancer, consistent with findings from the Prostate, Lung, Colorectal, and Ovarian (PLCO) cancer screening trial (24). Our findings support the claim that high carotene levels, particularly high β-carotene, may promote prostate cancer growth (25). When we examined isomers for β-carotene, the trans isomer was significantly higher among cases than controls in African-Americans. Trans isomers are more likely to be broken down into vitamin A compared with cis isomers, a possible explanation for the positive associations observed between provitamin A on prostate cancer risk (26, 27).

No association of cryptoxanthin with prostate cancer risk was observed in our study, regardless of subtype or isomer. Several prospective cohort studies, including the PLCO (24), European Prospective Investigation into Cancer and Nutrition (EPIC; ref. 28), and “Give Us a Clue to Cancer” (CLUE I and II; ref. 29) also reported null associations between plasma β-cryptoxanthin and prostate cancer risk, consistent with our results. Those studies, however, only analyzed the β-cryptoxanthin trans isomer, and not its cis isomer (30).

Among the various carotenoids, lycopene has been shown to be among those with the strongest antioxidant activities (31) and is regarded as potentially the most potent agent against risk of prostate cancer, especially the more lethal forms (32). The inhibitory effects of lycopene on carcinogenesis could involve free radical scavenging, upregulation of detoxification systems, interference with cell proliferation, induction of gap-junction communication, inhibition of cell-cycle progression, modulation of signal transduction pathways, and interaction with growth factors (33). The EPIC (27) and Physicians' Health Study (PHS; ref. 23), as well as a recent pooled analysis (4), showed inverse associations, especially with risk of aggressive prostate cancer, supporting the findings of our study. Approximately 80% of total lycopene occurs in the cis form in prostate tissue, compared with 50% in blood (34) and 5% to 10% in foods (35), although the biological significance of this isomeric composition is not known. The cis-lycopene isomer differs slightly in molecular shape than the all-trans form, and, therefore, may have a different pharmacologic effect and metabolism (36). In the PHS, the inverse association of lycopene was particularly apparent in aggressive prostate cancer and in men who did not consume β-carotene supplements (37). In our study, total lycopene was associated with the risk of prostate cancer, in both trans and cis isomers, particularly among African-Americans. Cis-lycopene plasma levels were associated with prostate cancer risk in never-users of multivitamin supplements and in both never/former and current smokers. Vogt and colleagues suggested that serum lycopene concentrations were significantly lower in African-Americans than in European-Americans, suggesting that differences in lycopene exposure may contribute to incident prostate cancer racial disparities (28). Our study, including 83% African-American and 17% European-American participants, also showed lower lycopene levels among African-American controls than European-American controls.

The plasma lycopene metabolite 5,6-dihydroxy-5,6-dihydrolycopene, resulting from in vivo oxidation of lycopene to an intermediate lycopene epoxide product, undergoes metabolic reduction to form the above metabolite. It has been shown that tomatoes and tomato-based food products are major dietary sources of this compound (38). The correlation between 5,6-dihydroxy-5,6-dihydrolycopene and lycopene was as high as 0.70 in this study, suggesting that the variation in 5,6-dihydroxy-5,6-dihydrolycopene level was largely due to lycopene levels. Plasma dihydrolycopene showed an inverse association with prostate cancer risk in our study, predominantly for stage II–IV prostate cancer. This association was mainly seen in African-Americans, never or former smokers, and subjects without supplementation.

In vivo oxidation of lutein and zeaxanthin is a key reaction in the metabolism of these nonprovitamin A active dihydroxycarotenoids (8). When free-radical mechanisms are involved in the initiation and promotion of carcinogenesis, carotenoids such as lutein and zeaxanthin may participate in quenching peroxides and protecting cells from oxidative damage (39). Our study showed a significant positive association with plasma trans-lutein on prostate cancer risk, predominantly among African-Americans, in subjects without multivitamin supplementation, and in advanced prostate cancer, inconsistent with the reduction in risk claimed for these carotenoid compounds. On the other hand, anhydrolutein, or 3′-hydroxy-3,4-dehydro-ss-carotene, is lutein's dehydrated metabolite. The presence of anhydrolutein I and II in human plasma is potentially due to acid catalyzed dehydration of dietary lutein as it passes through the stomach (40). In our study, cis-type anhydrolutein concentrations differed between prostate cancer cases and controls and were inversely associated with prostate cancer risk, particularly among those with no vitamin supplementation. No association was observed between plasma zeaxanthin and prostate cancer risk in the present study, consistent with findings from most previous cohort studies (24, 28, 29, 41). A pooled analysis (4) suggested heterogeneity in lutein- and zeaxanthin–prostate cancer associations; an inverse association between lutein and prostate cancer risk was only observed for prostate cancer diagnosed before the age of 60 years, and zeaxanthin was positively associated with prostate cancer risk only among men with a BMI of less than 25 kg/m².

Our study has many strengths. First, this prospective study included prediagnostic plasma samples, a large sample size of African-Americans, and detailed baseline data, which allowed for simultaneous adjustment for multiple confounders and evaluation of effect modifiers. Second, circulating plasma vitamin concentrations were measured using standardized procedures at one laboratory, with extensive experience in measuring lipid-phase micronutrients, especially carotenoids, in only a few batches and in a relatively short timeframe (thus, attenuating any potential technical measurement errors). Third, our report is the first study to consider various subtypes and isomers for each carotenoid simultaneously among the same population.

Our study also has some potential limitations. First, single measurements of plasma retinol and carotenoids may not be sufficient in detecting meaningful differences in long-term exposures between prostate cancer cases and controls. However, several reports (19, 42, 43) have suggested that single measurements of these biomarkers provide moderately reliable information on usual concentrations over a few years, although the true associations with prostate cancer risk are likely to be underestimated. Second, we did not have adequate power to investigate the associations for late-stage or aggressive-type of prostate cancers. Because of regular PSA screening, most prostate cancer cases are diagnosed at an early stage, so assembling large numbers of advanced cases has become increasingly difficult (our study had 11% stage III–IV cases), which hinged analyses for late-stage or aggressive prostate cancer in our study. Third, despite adjusting for HEI, we could not completely rule out confounding from healthier diets, as this is associated with higher screening rates. Fourth, the relatively short follow-up time raises a concern of reverse causation. However, excluding cases diagnosed within 1 or 2 years after blood sample collection showed a similar association pattern. Fifth, our study had low statistical power for subgroup analyses, particularly for aggressive prostate cancer and white participants. Sixth, 45.8% of prostate cancer patients were included in the current study, raising a potential concern on selection bias. We found that patients included in our study and those excluded had a similar Gleason score but differed significantly in several sociodemographic characteristics. Patients included in this study were less educated, more likely to have a lower family income, to be in lower NPB category and to have a lower HEI score, and less likely to take multivitamin supplement than patients excluded (Supplementary Table S5). However, all these factors were adjusted in the multivariate analyses. Finally, the generalizability of these results may be restricted, as study participants were predominantly of low socioeconomic status. Considering that risk factors for prostate cancer may differ by race, future studies investigating the effects of carotenoid subtypes and isomeric arrangements should incorporate men of different ethnicities and sociodemographic backgrounds into their study design.

In conclusion, in a low-income study population, we found positive associations of trans isomers, particularly β-carotene and lutein, and inverse associations of lycopene, dihydrolycopene, and cis-anhydrolutein with prostate cancer risk, even after adjusting for multiple comparisons. The associations were restricted by subgroups of individuals defined by race, smoking status, and multivitamin supplement use. Nevertheless, it should be noted that case–control differences in mean measured concentrations for all carotenoids were relatively small. The clinical significance for these findings remains to be determined.

S.A. Lee reports nonfinancial support from Kangwon National University during the conduct of the study. Q. Cai reports grants from NIH during the conduct of the study. W.J. Blot reports grants from NIH during the conduct of the study. X.O. Shu reports grants from NCI during the conduct of the study. No disclosures were reported by the other authors.

The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the CDC or the Mississippi Cancer Registry.

S.A. Lee: Conceptualization, formal analysis, writing–original draft. Q. Cai: Data curation, writing–review and editing. A.A. Franke: Data curation, writing–review and editing. M. Steinwandel: Data curation, writing–review and editing. J. Wu: Data curation, writing–review and editing. W. Wen: Formal analysis, writing–review and editing. W. Zheng: Resources, funding acquisition, methodology, writing–review and editing. W.J. Blot: Resources, funding acquisition, methodology, writing–review and editing. X.O. Shu: Conceptualization, supervision, methodology, writing–original draft, writing–review and editing.

The authors thank the study participants and research staff of the SCCS for their contribution to this study. We thank Regina Courtney for laboratory assistance and Dr. Mary Shannon Byers for assistance with editing and manuscript preparation. The SCCS is funded by a grant from the NCI (grant numbers: R01CA092447 and U01CA202979). Sample preparation was conducted at the Vanderbilt Survey and Biospecimen Shared Resource, which is supported in part by the Vanderbilt-Ingram Cancer Center (P30CA068485). HPLC analyses were supported by a Cancer Center Support grant to the University of Hawaii Cancer Center (P30 CA71789). S.A. Lee is supported by a research grant of Kangwon National University, 2017–2018. Data on SCCS cancer cases used in this publication were provided by the Alabama Statewide Cancer Registry; Kentucky Cancer Registry (Lexington, KY); Tennessee Department of Health, Office of Cancer Surveillance; Florida Cancer Data System; North Carolina Central Cancer Registry, North Carolina Division of Public Health; Georgia Comprehensive Cancer Registry; Louisiana Tumor Registry; Mississippi Cancer Registry; South Carolina Central Cancer Registry; Virginia Department of Health, Virginia Cancer Registry; Arkansas Department of Health, Cancer Registry. The Arkansas Central Cancer Registry is fully funded by a grant from National Program of Cancer Registries, Centers for Disease Control and Prevention (CDC). Data on SCCS cancer cases from Mississippi were collected by the Mississippi Cancer Registry, which participates in the National Program of Cancer Registries (NPCR) of the CDC.

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.