|Home | About | Journals | Submit | Contact Us | Français|
Red meat, processed and unprocessed, has been considered a potential prostate cancer (PCA) risk factor; epidemiological evidence, however, is inconclusive. An association between meat intake and PCA may be due to potent chemical carcinogens that are generated when meats are cooked at high temperatures. We investigated the association between red meat and poultry intake and localized and advanced PCA taking into account cooking practices and polymorphisms in enzymes that metabolize carcinogens that accumulate in cooked meats. We analyzed data for 1096 controls, 717 localized and 1140 advanced cases from the California Collaborative Prostate Cancer Study, a multiethnic, population-based case–control study. We examined nutrient density-adjusted intake of red meat and poultry and tested for effect modification by 12 SNPs and 2 copy number variants in 10 carcinogen metabolism genes: GSTP1, PTGS2, CYP1A2, CYP2E1, EPHX1, CYP1B1, UGT1A6, NAT2, GSTM1 and GSTT1. We observed a positive association between risk of advanced PCA and high intake of red meat cooked at high temperatures (trend P = 0.026), cooked by pan-frying (trend P = 0.035), and cooked until well-done (trend P = 0.013). An inverse association was observed for baked poultry and advanced PCA risk (trend P = 0.023). A gene-by-diet interaction was observed between an SNP in the PTGS2 gene and the estimated levels of meat mutagens (interaction P = 0.008). Our results support a role for carcinogens that accumulate in meats cooked at high temperatures as potential PCA risk factors, and may support a role for heterocyclic amines (HCAs) in PCA etiology.
Other than age, the only well-established risk factors for prostate cancer (PCA) are race/ethnicity, family history of prostate cancer and low-penetrance genetic variants that have emerged from genome-wide association studies. However, migrant studies and comparisons of ethnically similar populations across different countries suggest that environmental and/or lifestyle factors play an important role in the etiology of PCA (1). Among these factors, dietary factors such as intake of meat have been considered as a potential PCA risk factor. Given that few modifiable risk factors are known for PCA, understanding the role of dietary factors in PCA etiology is of high public health relevance.
A review by the World Cancer Research Fund concluded that, although evidence was limited, consumption of processed meat may increase the risk of PCA (2). Several cohort studies have reported positive associations (3–6); however, a recent review and meta-analysis of prospective studies (3), and a few additional recent cohort studies (7,8) provide limited evidence for an association between total red meat intake and PCA risk. Similarly, several case–control studies around the world reported positive associations with high intake of red meat (4,9–12), whereas others did not (4,13–15). Some of these studies also examined processed meats as a separate food category and, similar to total red meat, found inconsistent results, with several cohort (3,16) and case–control (13,15,17,18) studies reporting positive associations, whereas others did not (3,11,13,19,20). In addition, an ecological study of data from 1930 to 2000 reported that meat intake was the strongest dietary correlate of the temporal trend of increasing PCA mortality (21). Overall, the role of high intake of red meat, processed and unprocessed, remains inconclusive.
Several explanations have been put forward for the association between red meat and PCA. A prominent one is the relationship between meat and dietary fat, which has long been proposed as a PCA risk factor. However, studies of dietary fat and PCA are even less consistent than those of red meat (4), and some studies have found positive associations with red meat in the absence of associations with dietary fat (22). Other potential explanations include the presence of zinc in meats, which is essential for testosterone synthesis (23), the presence of heme iron in red meats, which may catalyze oxidative reactions (24), and the inverse relationship between consumption of meat and vegetables, which are known to contain many potentially anti-carcinogenic factors. Finally, an association between PCA and meat intake may be due to potent chemical carcinogens generated during cooking and/or processing of red meat and poultry, such as heterocyclic amines (HCAs), polycyclic aromatic hydrocarbons (PAHs) and N-nitroso compounds. The prostate gland is able to metabolize these chemicals into activated carcinogens (25); therefore, meat carcinogens are plausible PCA risk factors.
The relative proportion of each of these compounds in cooked meats depends on the amount and type of meat consumed, cooking and processing method, and cooking temperature and time. PAHs are deposited on the surface of smoked or grilled meat due to pyrolysis of fat droppings (26). HCAs are formed from the interaction of creatine/creatinine, amino acids and reducing sugars, which occurs at high temperatures (27). HCA formation increases with increasing cooking temperature and duration. Cooking methods that produce high levels of mutagens are broiling, grilling and pan-frying (28), with pan-frying yielding higher mutagenic activity when compared with grilling at a similar temperature (29). Since PAHs and HCAs have been shown to induce damage to prostatic epithelium cells (30), and were associated with formation of DNA adducts in prostatic tissue (31,32), it is biologically plausible that consumption of red and white meats cooked in conditions that favor PAH and HCA formation may increase the risk of PCA.
Among all epidemiological studies of meat and PCA conducted to date, few have considered level of doneness and cooking methods. Some studies found support for an association with high intake of meat cooked with high temperature cooking methods (5,13) or well-done meat (33–35), whereas others did not (7,8). Among studies that estimated levels of carcinogens, one cohort study reported a positive association with the HCA 2-amino-1-methyl1-6-phenylimidazo[4,5-b]pyridine (PhIP) (33) and another reported an association with the PAH benzo-a-pyrene (BaP) (5). Two other studies reported no associations (8,34).
The amount of DNA-damaging carcinogens in the prostate is determined by the amount and type of meat consumed, the cooking method used and the level of activity of key metabolism enzymes that activate and detoxify carcinogens. Therefore, it is plausible that genetic variation in key enzymes that activate or detoxify HCAs and PAHs may modify the association between diets high in red or white meat and PCA risk.
Recently, using data from the San Francisco Bay area component of the California Collaborative Prostate Cancer Study, we reported positive associations between consumption of hamburgers, processed meat, grilled red meat and well-done or very well-done red meat and advanced, but not localized, PCA risk (36). Moreover, we reported an association between PhIP intake and advanced PCA, although a dose–response relationship was lacking, since increased risk was associated with intermediate, but not high, intake. We now extend these analyses to the entire California Collaborative Prostate Cancer Study, which includes non-Hispanic white (NHW), African-American (AA) and Hispanic cases and controls from the San Francisco Bay area and from Los Angeles County (LAC). We investigated the associations of different red meats, processed meats and poultry with risk of localized and advanced PCA, taking into account cooking methods, level of doneness, estimated levels of carcinogens and the potential modifying role of selected polymorphisms in enzymes that metabolize meat mutagens.
The California Collaborative Prostate Cancer Study is a multiethnic, population-based case–control study conducted in Los Angeles County and in the San Francisco Bay area (SFBA) (37–39). Incident cases of PCA were identified through two regional cancer registries (Los Angeles County Registry and the Greater Bay Area Cancer Registry) that participate in the Surveillance, Epidemiology, and End Result (SEER) program and the California Cancer Registry. In both study sites, PCA was classified as advanced if the tumor extended beyond the prostatic capsule or into the adjacent tissue or involved regional lymph nodes or metastasized to distant locations (SEER 1995 clinical and pathologic extent of disease codes 41–85). At the Los Angeles site, controls were identified through a standard neighborhood walk algorithm, as we previously described (39). At the San Francisco site, controls were identified through random-digit dialing and, for those aged 65–79 years, through random selections from the rosters of beneficiaries of the Health Care Financing Administration. At both sites, controls were matched to cases on race/ethnicity and the expected 5-year age distribution of cases.
Characteristics of this study, including eligibility criteria and ascertainment rates, have been previously reported (36,38). Briefly, incident cases aged 40–79 years diagnosed with a first primary localized prostate cancer from 1997 to 1998 were randomly sampled (15% of NHW cases and 60% of African-American cases). Cases with a first primary advanced prostate cancer included all NHW and African-American men diagnosed from 1997 to 2000. In-person interviews were completed by 208 (73 African American and 135 NHW) localized cases, 568 (118 African American and 450 NHW) advanced cases and 545 (90 African American and 455 NHW) controls. Blood or mouthwash samples were collected for controls and advanced cases only. The genotype analyses were based on 378 advanced cases (77 African American and 301 NHW) and 252 controls (35 African American and 217 NHW) with DNA from blood and reliable dietary information from food frequency questionnaire (FFQ).
Characteristics of this study, including eligibility criteria and ascertainment rates, have been previously reported (39). Briefly, African-American, Hispanics and NHW males diagnosed with incident PCA from 1999 to 2003 were identified using the rapid case ascertainment protocol of the Los Angeles County Cancer Surveillance Program and the Los Angeles County cancer registry. A total of 594 controls (163 African American, 122 Hispanics and 309 NHW) and 1184 cases (351 African American, 333 Hispanics and 500 NHW) completed the in-person interview and provided a blood sample. The analyses in this study were based on 559 advanced cases (126 African American, 157 Hispanic and 276 NHW), 499 localized cases (194 African American, 112 Hispanic and 193 NHW) and 511 controls (143 African American, 81 Hispanic and 287 NHW) with DNA from blood and reliable dietary information from FFQ.
Written informed consent was obtained from all study participants, and the study was approved by the institutional review boards at the Cancer Prevention Institute of California and the University of Southern California.
Trained professional interviewers administered a structured questionnaire at the participants’ home that asked about demographic background, medical history and various lifestyle factors, and obtained measurements of height and weight. Usual dietary intake was assessed for the reference year, defined as the calendar year prior to diagnosis for cases and the calendar year prior to selection into the study for controls, using a modified version of the Block FFQ (40). Questions on cooking methods and degree of doneness and browning were adapted from a commonly used cooking module developed by Sinha et al. (41). An aggregate level socioeconomic status (SES) variable was derived from 2000 census data as previously described (36,38). Body mass index (BMI) was calculated as reported weight (in kilograms) in the reference year divided by measured height (in meters) squared. For subjects with missing information on self-reported weight (1 case and 1 control), the BMI calculation was based on measured weight. Similarly, for subjects who declined the measurement of height (4.9% of cases and 4.8% of controls), BMI was based on self-reported height. Even though weight was measured at the time of diagnosis, we chose to use self-reported weight given the concern of potential bias introduced by weight loss among cancer patients at the time of interview. Whereas using self-reported weight for the reference year may be associated with misclassification, we speculated that this approach would lead to less misclassification than the one introduced by case weight loss at the time of interview. BMI was categorized as normal weight (BMI <25), overweight (BMI 25–29.9) and obese (BMI ≥30). Underweight men (BMI <18.5, n = 15) were grouped with normal-weight men.
The FFQ assessed the usual portion size and frequency of consumption of red meat (all types of beef and pork, hamburgers and steak), poultry (chicken and turkey) and processed meat (sausages made from red or white meat, bacon and cold cuts). Information was obtained on usual method of preparation (e.g., pan-frying, oven-broiling, grilling and other methods) and level of doneness (by choosing from a series of color photographs showing increasing degrees of doneness and browning) was obtained for hamburgers, steak, poultry, bacon, sausages and hot dogs.
We utilized the multivariate nutrient density approach (42) to adjust for energy intake of red meat, poultry, processed meat and meat mutagen variables. These nutrient density variables were categorized into quintiles based on the distribution among controls. Variables for specific red meats (hamburger, steak), processed meats (bacon, sausage), homemade gravy, meats cooked by specific cooking methods (pan-frying, oven-broiling, baking and grilling) or well-done meats (browned or charred on the surface) were categorized into four levels of consumption: never/rarely, low (quintiles 2 and 3), medium (quintile 4) and high (quintile 5). The categorization between the second and third quintile was highly influenced by the variation in the denominator of the nutrient density variable (inverse calorie intake). To minimize spurious associations due to this effect, second and third quintile were collapsed to form the low intake category. Given that oven-broiling, grilling and pan-frying expose foods to high cooking temperatures, we combined these methods into a single category termed “high temperature cooking methods”, and nutrient density variables for specific meats cooked with high temperature methods were created and categorized into four levels as described above.
We collected data on 12 SNPs in 8 genes, which were selected given previous reports on their impact on enzyme function and/or associations with PCA, as we previously described (43): GSTP1 Ile105Val (rs1695), PTGS2 −765 G/C (rs20417), CYP1A2 −154 A/C (rs762551), EPHX1 Tyr113His (rs1051740), CYP1B1 Leu432Val (rs1056836), UGT1A6 Thr181Ala and Arg184Ser (rs1105879, rs2070959), NAT2 Ile114Thr, Arg197Gln, Gly286Glu and Arg64Gln (rs1799930, rs1799931, rs1801279, rs180120) and CYP2E1 −1054 C>T (rs2031920). In addition we genotyped two copy number variants in GSTM1 and GSTT1. We used the information on the four NAT2 polymorphisms to construct haplotypes using haplo.stats package in R. These polymorphisms define different NAT2 alleles, which have been characterized for their impact on protein function, consistent with the existing classification (44), we classified carriers of two copies of the slow haplotype as ‘”slow” and carriers of all other haplotypes as “fast” phenotype. All genotypes were obtained using Taqman assays, available “on demand” from ABI (Applied Biosystems, Foster City, CA), following manufacturer's instructions. Call rates were >97%. No differences were found between observed genotypic frequencies and those expected assuming Hardy–Weinberg equilibrium.
After excluding 43 controls, 44 localized cases and 59 advanced cases with dietary data considered unreliable (i.e., daily caloric intake <600 kcal or >6000 kcal), the analyses of questionnaire data were based on 1096 controls, 717 localized cases and 1140 advanced cases. Analyses of genotype data were based on subjects with DNA from blood, including 800 controls, 535 localized cases and 988 advanced cases. These individuals did not differ from those without DNA with regards to age, calorie intake, family history, SES and BMI in either study site (data not shown).
To best correct for differences in race/ethnicity, SES and the case–control ratio across the two study sites, we created a variable that classified subjects according to study site (SFBA or LAC), SES (five-level variable from low to high) and race/ethnicity (African American, Hispanic, NHW), and used it to group individuals into 18 categories in conditional logistic regression models. Five quintiles of SES were derived using 2000 census data at the level of the block group, based on levels of education, income and occupation as described previously (36,45). At the LAC site, the three ethnic groups were combined with four SES categories (quintiles 1, 2, 3, 4–5) to form 12 groups. SES was collapsed into three categories (quintiles 1–2, 3, 4–5) at the SFBA site and combined with the two ethnic groups (AA and NHW) to form six groups. SES quintiles were collapsed to maintain sufficient number of individuals in each group, thus allowing for stable variance estimates in regression models after adjusting for multiple covariates. The fourth and the fifth quintiles of SES were collapsed together because among Hispanic and African-American individuals there were low numbers in the high SES category. Similarly for the SFBA site, which was predominantly composed of non-Hispanic white participants, the first and second quintile groups were collapsed together.
Differences in sociodemographic, lifestyle and dietary characteristics between cases and controls were computed within each study site, using Pearson’s χ2 test for categorical variables and t-test for continuous variables. For the exposure variables, odds ratios (ORs) and 95% confidence intervals (CIs) were estimated from conditional logistic regression, with separate analyses performed for localized and advanced cases. In multivariate model 1, analyses were adjusted for age (years, continuous), BMI (<25, 25.0–29.9, ≥30), total calorie intake (kcal/day, continuous) and family history of PCA in first-degree relatives (yes or no). In multivariate model 2, we further adjusted for potentially confounding dietary and lifestyle factors such as total fat intake (g/day), total fruit consumption (g/day), total vegetable consumption (g/day) and alcohol consumption (g/day) during the reference year and lifetime cigarette smoking (pack-years).
To examine the potential modifying role of the selected polymorphisms on the associations between red meat/poultry and PCA risk, we performed SNP × red meat or poultry interaction tests coding each SNP as log-additive and each meat variable as a three-level variable (using the median level of exposure among controls at each of the levels). We conducted both 2-df interaction tests by treating the three-level meat exposure variable as categorical 1-df tests by treating meat exposures as ordinal. For these gene × environment analyses we evaluated whether genotypes for eight SNPs and two null variants in nine carcinogen metabolism genes (CYP1A2, CYP1B1, CYP2E1, EPHX1, GSTP1, PTGS2, UGT1A6, GSTT1 and GSTM1), and the predicted phenotype for NAT2, modified the associations with the following dietary variables: high temperature-cooked red meat, well-done red meat, high temperature-cooked poultry, well-done poultry, processed meat, well-done processed meat and estimated level of carcinogens. These variables were selected because they capture meat dietary patterns that have been associated with the accumulation of chemical carcinogens that require metabolism by proteins coded by the selected genes. The main effects of these SNPs have been previously reported (43). Due to strong LD between the UGT1A6 Thr181Ala and UGT1A6 Arg184Ser SNPs, only the former was utilized for the purpose of interaction analyses.
For all analyses, tests for trend were performed by assigning median values to each quartile and modeling these categories as a continuous variable. All hypothesis tests were two-sided and all analyses were done using the statistical software Stata S/E 11.0 for Windows (STATA Corporation, College Station, TX).
Table I shows the sociodemographic characteristics and lifestyle/dietary patterns of cases and controls. Cases were more likely to be African American and less likely to be non-Hispanic white, as compared with controls in both the study sites; at the LAC site, cases also had an overrepresentation of Hispanic men. In both the sites the frequency of individuals with at least one first-degree relative with prostate cancer was higher among cases. Cases were older than controls at the LAC site, but not in the SFBA site. Differences between distributions of SES were observed at both the sites between cases and controls. Cigarette smoking, total calorie intake and total fat consumption were higher among cases as compared with controls in the LAC site, but no such differences were seen at the SFBA site. BMI was lower among cases in the SFBA study site, but not in the LAC site. Other potential risk factors of prostate cancer were not observed to be different between cases and controls in both the study sites (Table I). Energy-adjusted red meat, poultry and processed meat consumption patterns by race/ethnicity and prostate cancer defined by stage at diagnosis are summarized in Supplementary Table I, available at Carcinogenesis Online. NHW male controls had a lower intake of red meat (mean = 13.9g/1000 kcal/day) and poultry (mean = 21.0g/1000 kcal/day) than African-American (mean = 16.3 and 27.6g/1000 kcal/day, respectively) or Hispanic (mean = 16.8 and 26.5g/1000 kcal/day, respectively) controls. Hispanic controls had a lower intake of hamburger, but a higher intake of steak. African-American controls had a higher intake of bacon and sausage as well as total processed meats as compared with NHW and Hispanic controls. Differences were also observed in red meat cooking preferences; African-American controls had a higher intake of oven-broiled meats than NHW or Hispanic men, whereas Hispanic controls had a higher consumption of grilled and pan-fried meats, which contributed to an overall higher consumption of high temperature-cooked red meats and poultry as compared with the remaining two racial/ethnic groups. There were 72% African-American, 52% Hispanic and 62% NHW controls who did not consume any high temperature-cooked poultry; baking was the most preferred way of cooking poultry. No racial/ethnic differences in consumption of well-done poultry were observed. African-American controls had the highest consumption of pan-fried and well-done processed meats. The mean consumption in g/1000 kcal/day of pan-fried processed meat was 3.6 among African-American controls, 1.6 among Hispanic controls and 1.7 among NHW controls, whereas that of well-done processed meat was 4.1, 1.7 and 1.8 among African-American, Hispanic and NHW controls, respectively.
Intake of red meat or processed meat was not associated with localized or advanced PCA risk (Table II). Poultry intake was inversely associated with advanced PCA risk only (P trend = 0.009), with an OR of 0.7 (95% CI = 0.6–1.0) for highest versus lowest quartile of intake. The associations with specific meat types did not vary significantly across the three racial/ethnic groups (Supplementary Table I, available at Carcinogenesis Online).
As shown in Table III, increased risks of advanced PCA were associated with high consumption (highest versus lowest quartile) of pan-fried red meat (OR = 1.3; 95% CI = 1.0–1.8, P trend = 0.035) and red meat cooked by high temperature cooking methods such as grilling, oven-broiling or pan-frying (OR = 1.4, 95% CI = 1.0–1.9, P trend = 0.026). When individual meat types were considered, high intake (highest versus lowest quartile) of hamburgers cooked with high temperature methods was positively associated with both localized and advanced PCA risk, with a stronger association for advanced PCA (OR = 1.7, 95% CI = 1.3–2.2, P trend < 0.001). Steak cooked at high temperature was not associated with localized or advanced PCA. High intake (highest versus lowest quartile) of well-done red meat was also positively associated with advanced PCA (OR = 1.4, 95% CI = 1.1–1.8, P trend = 0.013). However, no individual meat type contributed predominantly to this association.
Baked poultry was inversely associated with both localized (P trend = 0.073) and advanced (P trend = 0.028) PCA, whereas pan-fried poultry consumption was marginally associated with increased risk of advanced PCA (P trend = 0.069) (Table IV). No associations were observed with poultry cooked by grilling or oven broiling or with well-done poultry.
We found no evidence of association between sausage, bacon or total processed meat intake and PCA risk when considering different cooking methods and doneness levels (Supplementary Table II, available at Carcinogenesis Online). Similarly, for both localized and advanced PCA, we found no evidence of association with energy-adjusted estimated levels of HCA mutagens, BaP and total meat mutagens (revertant colonies) (Table V).
The associations with various meat types when considering cooking practices did not vary significantly across the three racial/ethnic groups (data not shown). An exception was the association between hamburgers cooked at high temperature and advanced PCA, which was observed in NHW (P trend < 0.001) and Hispanic (P trend = 0.041) men, but not African-American men (Supplementary Table III, available at Carcinogenesis Online). Although no significant heterogeneity was detected by race/ethnicity in the associations between estimated meat mutagens and localized or advanced PCA, elevated ORs were observed for high intake of meat mutagens and localized PCA risk among Hispanic men (Supplementary Table IV, available at Carcinogenesis Online). These findings, however, were based on a very small sample size. No such associations were observed for advanced PCA.
Using a P-value cutoff of 0.005 (Bonferroni multiple testing corrected P-value = 0.05 for testing nine polymorphisms and one predicted phenotype), for heterogeneity of trends, we observed that the −765G>C promoter SNP in PTGS2, modified the association between estimated meat mutagens and advanced PCA risk (Table VI). Estimated mutagenic activity (actual and predicted revertant colonies) was associated with advanced PCA in carriers of the G/G genotype at the PTGS2 promoter locus. However, this association was not observed among carriers of at least one C allele (P Heterogeneity of trend = 0.004). There was no other evidence of effect modification of the association between any of the meat cooking/processing variables and advanced or localized PCA by the selected carcinogen metabolism SNPs when considering a P-value cutoff of 0.005.
In this multiethnic population-based case–control study we observed that high intake of red meat per se was not associated with PCA risk. Only when cooking practices were considered, did we find an association between high intake of red meats cooked at high temperature, especially pan-fried red meats, and advanced PCA. When considering specific red meats, intake of hamburgers, but not steak, was associated with risk. Intake of well-done red meat was also associated with advanced PCA risk. Whereas intake of baked poultry showed an unexpected inverse association with advanced PCA, intake of pan-fried poultry showed a positive association.
Pan-frying has been consistently implicated in the formation of meat mutagens (28). The oil used in pan-frying acts as an efficient heat transfer medium between the pan and the surface of the meat, and therefore high surface temperatures are reached. Pan-frying does not expose meats to open flames and fats from the meats do not have an opportunity to drip on the flames undergoing incomplete combustion. Thus, pan-frying is typically not associated with accumulation of PAHs (46). Similarly, there is no evidence for formation of N-nitroso compounds during pan-frying. Therefore, the most relevant carcinogens generated by pan-frying seem to be HCAs, which may drive the observed associations between pan-fried red meat and poultry and PCA risk.
We observed an association between consumption of high temperature-cooked red meats, especially hamburgers, and the risk of advanced prostate cancer. The association for steak cooked at high temperatures or to well-done levels was weaker in magnitude and not statistically significant. We speculate that these observations may be due to the formation of different amounts and types of HCAs in these specific meat types and also because hamburgers attain high internal and external temperatures faster than steak (47).
Contrary to our expectation that HCAs drive the observed associations, the estimated levels of HCAs were not associated with PCA risk. Interestingly, one other study has also reported an association with well-done red meat intake and PCA risk, but no evidence of association with estimated meat mutagens (34). There are several potential explanations for this discrepancy. First, of more than 15 HCAs identified to date that are known to accumulate in cooked meat (48), only PhIP, MeIQx and DiMeIQx are captured by the CHARRED database that all epidemiological studies to date, including the current study, have used to estimate meat mutagens. Compounds formed at lower concentrations, and hence not measured, or not-yet identified HCAs, may play a more relevant role in cancer causation. For example, in a case–control study of PCA in New Zealand (49), intake of well-cooked red meat (beef and steak) was associated with increased PCA risk, whereas no associations were found with estimated PhIP, MeIQx and DiMeIQx. A fourth HCA (IFP), however, was weakly associated with PCA risk. Second, although the assessment of meat mutagens using the cooking module and CHARRED, is vastly better than previously used methods, it is subject to measurement error. Interpolation from laboratory measurements of different meat samples cooked under different conditions, and adaptation of the cooking module to different FFQs, add additional approximations, thereby contributing to exposure misclassification and biasing risk estimates towards the null, assuming non-differential misclassification between cases and controls.
Alternative explanations for our findings include the absorption of culinary fat, free radical formation from fats and exposure to cookware-related substances such as perfluorooctanoate (50). Since red meats naturally have a high fat content, they are less likely to imbibe culinary fats. However, the fatty acid content in red meat, particularly corn fed meats, may be higher in n-6 polyunsaturated fatty acids, which are substrate to oxidation and free radicals formation (51). It is also possible that pan-fried red meat tags other intangible aspects of lifestyle, not considered in our analyses, which may confer increased PCA risk.
There is insufficient evidence to draw firm conclusions about the role of poultry in PCA risk. Studies have reported that while red meats predominantly accumulate higher levels of MeIQx-like HCAs, high temperature-cooked poultry accumulates much higher levels of PhIP-like HCAs, as compared with red meat (52,53). Our finding of a positive association between pan-fried poultry and PCA risk supports this fact. However, we also report an inverse association of baked poultry intake with advanced and localized PCA risk. This inverse association may be explained by a form of vitamin K2, menaquinone, present in dark poultry meat that has been associated with reduced risk of advanced PCA in a European cohort (54). Further studies need to test this hypothesis.
We did not observe any association between processed meat intake and PCA risk. According to the WCRF report, associations between processed meats and PCA risk are stronger in prospective studies than in case–control studies (55). Case–control studies typically assess dietary intake for the period 1–2 years before diagnosis, whereas cohort studies obtain dietary information at baseline. It is possible that the stronger association found in prospective studies may be due to the capturing of exposures during a more relevant time period in cohort studies.
Although overall, the findings in this study were in agreement with those reported previously within the San Francisco Bay area component (36), certain differences were also observed. In both reports there were null associations between all types of meat and cooking variables and localized prostate cancer risk. For advanced prostate cancer risk we could validate the associations of well-done/very well-done red meats observed in the previous study. However, the previously reported positive main effect associations of hamburgers, steak and processed meats with advanced prostate cancer could not be replicated in this combined study. These discrepancies could potentially be due to differences in ethnic, socioeconomic and life-style patterns in the LAC site as compared with the SFBA site, and the fact that the previous study was more homogeneous than this study. However, the pooled analysis of the two sites allowed us to improve power, while statistically adjusting for these heterogeneous effects. Therefore, we were able to test for associations for cooking methods within specific meat types that may have previously been missed, such as associations of pan-fried hamburgers and advanced prostate cancer risk. Both studies reported no associations between estimated meat mutagens and risk of localized or advanced prostate cancer.
When we considered genes that play key roles in the metabolism of meat mutagens, the strongest genetic modifier identified was a PTGS2 promoter polymorphism (−765 G>C, rs20417) (56). The PTGS2 gene codes for the COX-2 protein, which is overexpressed in prostate cancer cells and is induced by HCAs and PAHs, which it can activate into carcinogens (57–60). PAH activation in the prostate can also be carried out by CYP1A1 and CYP1B1 (61); therefore, if COX-2 levels are reduced, other enzymes might be able to compensate for this. In contrast, COX-2 activity levels may be more critical for HCA activation in the prostate given the absence of other enzymes capable of carrying out this step. CYP1A2 is a critical enzyme for HCA activation in the liver; however, it is not expressed in the prostate (62). Given our observation of a role of pan-frying in PCA risk, it is plausible that the estimated mutagenic activity modified by a polymorphism in PTGS2 is capturing potential HCAs that may be activated by COX-2 in the prostate. Since COX-2 is only expressed during periods of inflammatory stress, it is possible that meat derived HCAs may be relevant in prostate carcinogenesis during periods of inflammation in the prostate. In support of this, animal experiments have shown that the HCA PhIP can induce inflammation in regions of the prostate that coincide with areas of tumor development, which suggests that in addition to bulky adducts and cell proliferation, HCAs may also cause inflammation and thus promote cancer development (63). This dual role for HCAs suggest that red meat cooked at high temperature might contribute to inflammation in the prostate and in this way contribute to progression of existing tumors that otherwise might remain localized. If this were true, it would explain why our study, along with many others, found an association between red meat intake and PCA risk exclusively, or stronger, for advanced than localized disease.
We observed that mutagenic activity was associated with PCA risk only among individuals carrying the PTGS2 −765 G>C (rs20417) G/G genotype. The effects of this polymorphism, which is located in a binding site upstream of the translation initiation site, on COX-2 protein function are still unclear. In vitro studies done with different cell types have reported lower expression of the protein coded by the C allele than the one coded by the G allele or vice versa, depending on which cell type was used (56,64). These inconsistencies make it difficult to interpret our findings. If the C allele conferred reduced PTGS2 gene expression, as reported by some (56,65), this may lead to reduced activation of HCAs or PAHs in the prostate during inflammatory stress. This would be consistent with our finding of an association between higher mutagenic activity and PCA risk among carriers of the G allele. However, recently we reported that high intake of white fish, in particular white fish cooked at high temperature, was associated with PCA risk (66), and that this association was stronger among carriers of the PTGS2 (−765 G>C, rs20417) C allele (43). Given the dual role of COX-2, in both carcinogen metabolism and fatty acid metabolism, it is a possibility that the impact of the -765 G>C (rs20417) polymorphism may be different for each of these two functions. If that were the case, our finding of effect modification of PTGS2 on the association between mutagenic activity and PCA risk would be pointing to a role of chemical carcinogens that accumulate in red meat and poultry, and require COX-2 activation, whereas the association between high intake of white fish could be pointing to a role of fatty acids, that also require COX-2 metabolism. We note that the estimated levels of mutagenic activity obtained with the CHARRED database do not include fish intake, and that in our data set correlations between red meat, poultry and fish variables were low. To our knowledge, no previous study has assessed interactions between PTGS2 variants and estimates of meat mutagens from the CHARRED database in relation to PCA risk. Therefore, further studies that include a more comprehensive coverage of this gene may help clarify these opposite findings.
Strengths of our study include its population-based design with cases ascertained using two SEER cancer registries, a large sample size, the inclusion of a diverse study sample, enrichment with advanced cases and the ability to distinguish between different meat types, cooking methods and level of doneness ascertained using colored photographs. The questionnaire used in this study allowed us to estimate intake of specific nutrients and meat mutagens. However, we acknowledge potential weaknesses, such as possible misclassification of meat intake and cooking practices due to measurement error via questionnaire. Although differential misclassification due to recall bias among cases is unlikely given that at the time of interview there was no widespread knowledge that consumption of meat or certain cooking methods may be associated with PCA risk, we cannot exclude the possibility of non-differential misclassification. Moreover, in our study we did not consider the effect of using marinades in the cooking of meats, which have been shown to decrease the formation of HCAs (67). Therefore, lack of information on marinating practices may have contributed to exposure misclassification, potentially biasing the results towards the null. Lastly, even though our study included a larger number of men from racial/ethnic minority populations as compared with other studies, the sample size for specific exposure categories was limited. It is therefore possible that we may have missed associations present in certain subgroups only.
In summary, our findings support the hypothesis that mutagenic compounds formed during the process of cooking meats may contribute to prostate carcinogenesis. Our findings may have implications for public health recommendations. Exposure to HCAs from pan-fried meats may be reduced by cooking meats for longer time using lower surface temperatures, frequent meat flipping and pretreatment of meat in the microwave before pan-frying, which alters concentrations of precursors (e.g., creatine, sugars) (68). If these findings can be replicated in independent studies, and pan-frying of meats can indeed be established as a relevant risk factor for PCA through an HCA mediated mechanism, then guidelines can be given to the public on how to modify meat cooking practices to reduce the formation of meat mutagens.
Supplementary materials can be found at http://carcin.oxfordjournals.org/
We are grateful to the men who participated in this study, without whom this research would not be possible. We also thank Mr. Andre Kim for help with genotyping and Mr. Chris Yoon for assistance with data cleaning.
The project described was supported in part by the Prostate Cancer Foundation (to M.C.S). M.C.S. received support from grant RSF-09-020-01-CNE from the American Cancer Society and from award number 5P30 ES07048 from the National Institute of Environmental Health Sciences and award number P30CA014089 from the National Cancer Institute. The Northern and Southern California studies were funded by grants 99-00527V-10182 (to E.M.J.) and 99-00524V-10258 (to S.A.I.) from the Cancer Research Fund, under Interagency Agreement #97-12013 (University of California contract #98-00924V) with the Department of Health Services Cancer Research Program, and by grant R01CA84979 (to S.A.I.) from the National Cancer Institute, National Institutes of Health. The collection of cancer incidence data used in this study was supported by the California Department of Health Services as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the National Cancer Institute’s Surveillance, Epidemiology, and End Results Program under contract HHSN261201000140C awarded to the Cancer Prevention Institute of California, contract HHSN261201000035C awarded to the University of Southern California, and contract HHSN261201000034C awarded to the Public Health Institute; and the Centers for Disease Control and Prevention’s National Program of Cancer Registries, under agreement #1U58 DP000807-01 awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the authors, and endorsement by the State of California, the California Department of Health Services, the National Cancer Institute, or the Centers for Disease Control and Prevention or their contractors and subcontractors is not intended nor should be inferred.