Cooked Meat and Risk of Breast Cancer—Lifetime Versus Recent Dietary Intake

Abstract

Background: 

Polycyclic aromatic hydrocarbons (PAHs) and heterocyclic amines (HCAs) are carcinogens formed in or on the surface of well-done meat, cooked at high temperature.

Methods: 

We estimated breast cancer risk in relation to intake of cooked meat in a population-based, case-control study (1508 cases and 1556 controls) conducted in Long Island, NY from 1996 to 1997. Lifetime intakes of grilled or barbecued and smoked meats were derived from the interviewer-administered questionnaire data. Dietary intakes of PAH and HCA were derived from the self-administered modified Block food frequency questionnaire of intake 1 year before reference date. Unconditional logistic regression was used to estimate adjusted odds ratios (ORs) and 95% confidence intervals (CIs).

Results: 

Modest increased risk was observed among postmenopausal, but not premenopausal, women consuming the most grilled or barbecued and smoked meats over the life course (OR = 1.47; CI = 1.12–1.92 for highest vs. lowest tertile of intake). Postmenopausal women with low fruit and vegetable intake, but high lifetime intake of grilled or barbecued and smoked meats, had a higher OR of 1.74 (CI = 1.20–2.50). No associations were observed with the food frequency questionnaire-derived intake measures of PAHs and HCAs, with the possible exception of benzo(α)pyrene from meat among postmenopausal women whose tumors were positive for both estrogen receptors and progesterone receptors (OR = 1.47; CI = 0.99–2.19).

Conclusions: 

These results support the accumulating evidence that consumption of meats cooked by methods that promote carcinogen formation may increase risk of postmenopausal breast cancer.

Polycyclic aromatic hydrocarbons (PAHs) and heterocyclic amines (HCAs) are 2 classes of carcinogens that are found in the human diet. PAHs can appear on or near the surface of foods from the smoke created by incomplete combustion of carbon and hydrogen in fat that has fallen onto hot coals (such as in grilling or barbecuing), or by contamination from air or water pollutants.¹ PAHs are found in a variety of food products, including grilled or barbecued meat, smoked meat, vegetables, fruits, yogurt, margarine, grains, and cereals.¹ HCAs are formed when amino acids pyrolyze in meat juice, and are particularly high in pan-fried, grilled and, to a lesser extent, broiled meat.² The method, temperature, and duration of cooking greatly affect the amount of HCAs found in meat products, with greater doneness associated with higher concentrations of HCAs.^3,4 HCAs and PAHs are known carcinogens in animals, and are involved in the development of mammary tumors.^5,6 Dietary intake of these compounds has been consistently linked to colorectal cancer,⁷ and has been linked to breast cancer in a few epidemiologic studies,^8,9 but not all.¹⁰

Several epidemiologic studies have observed positive associations between recent intake of well-done cooked meat and breast cancer,^11–13 whereas other studies have not.^14–16 None of the previous studies assessed lifetime intake of cooked meat, which may be the more relevant time frame of exposure in carcinogenesis. Additionally, many of the studies were limited in their ability to estimate dietary intake of specific carcinogens from cooked meat because the exposure assessment method relied on food frequency questionnaires (FFQs) without specific questions related to meat preparation techniques or doneness preference.

This large population-based study was undertaken to address the hypothesis that breast cancer risk may be associated with dietary intake of PAHs, HCAs, and meat, according to cooking methods and level of doneness. We assessed intake both recently and throughout the lifetime. Further, we looked for possible interactions with fruit and vegetable intake, based on the possibility that these foods may mitigate the carcinogenic effects of PAHs and HCAs, as shown in animal models.^17,18

METHODS

Institutional Review Board approval was obtained by all collaborating institutions. The study was conducted in accordance with national and institutional guidelines for the protection of all study participants.

Study Design and Population

The Long Island Breast Cancer Study Project is a population-based case-control study funded by the National Cancer Institute and the National Institute of Environmental Health Sciences in response to a federal mandate (Public Law 103-43, June 10, 1993) that a case-control study be conducted on Long Island, New York to examine the relationship between environmental exposures and breast cancer. Methods have been described in detail previously.¹⁹ Cases (n = 1508; 82% of eligible cases) were identified through pathology/cytology records of 33 institutions; they included residents of Nassau and Suffolk counties, newly diagnosed with breast cancer between 1 August 1996 and 31 July 1997. Population-based controls (n = 1556; 63% of eligible controls) were identified using random digit dialing for women under the age of 65 years and by Center for Medicare and Medicaid Services (formerly known as Health Care Financing Administration) rosters for women 65 years and older. Controls were frequency-matched to the expected age distribution of cases by 5-year age group. All participants signed informed consent forms before enrolling in the study.

Exposure Assessment

An in-home questionnaire that lasted on average 101 minutes was administered to each subject by a trained interviewer. The questionnaire elicited information on demographic factors, current and past residences, occupational history, environmental exposures, reproductive history, menstruation and menopause history, contraceptive and hormone use, medical history, body size, physical activity, family history of disease, alcohol consumption, and smoking history (questionnaire available at http://epi.grants.cancer.gov/LIBCSP/projects/Questionnaire.html). This main questionnaire also included assessment of intake of 4 categories of grilled/barbecued and smoked meats over each decade of life since the teenage years (Section C. Residential History, pages 19–20 of the questionnaire on the Web site). Among the respondents who completed the main questionnaire, 98% of breast cancer cases and 98% of controls also completed a self-administered Block FFQ (detailed below), including approximately 100 food items that assessed diet in the previous year.

Exposure Indices Calculated

Lifetime Intake of Grilled or Barbecued and Smoked Foods

As previously described,²⁰ variables quantifying lifetime consumption of grilled/barbecued and smoked foods were constructed based on 1) data collected in the main questionnaire that focused on past lifetime consumption and 2) a checklist that focused on consumption in the 4 weeks before the interview. In the main questionnaire, women were queried about their consumption patterns over 6 decades of life (<20 years, 20–29 years, 30–39 years, 40–49 years, 50–59 years, 60+ years) for 4 different groups of PAH-containing foods: smoked beef, lamb, and pork; grilled/barbequed beef, lamb, and pork; smoked poultry or fish; and, grilled/barbequed poultry or fish. The average of the 6 (or fewer) decades was calculated to derive an average lifetime consumption of these 4 PAH-containing food groups.

Values were missing for <2% of respondents for each of the 24 groups (6 decades of consumption × 4 food groups), which translated to 9%–10% of respondents missing the values for the 4 derived averages. Most of missing values were restricted to a single decade of reported consumption. Imputations for the missing values were derived by multiple regression²¹ using data from women with complete data to predict the missing variables in each of the 24 groups. In addition, regression analyses were performed separately by the decade of age at interview because of concerns about possible cohort effects. For example, to predict average consumption of smoked beef, lamb, and pork between the ages of 20 and 29 years, for women in their 30s at reference, multiple regression was conducted using subjects in their 30s at reference to construct a model:

These regression coefficients, β₁ and β₂, were then used to impute beef consumption for all women in their 30s who were missing only beef consumption during the decade 20 to 29 years of age. These steps were repeated for other women missing only 1 interval of consumption within each food category. To correct for artificially minimized standard errors for the odds ratios (ORs) produced when using imputations, the standard errors obtained using imputed data were inflated back to the lower sample-size level. This imputation strategy reduced the amount of missing intake data in the following manner: lifetime grilled/barbecued beef, lamb, or pork consumption reduced from 9% to 3%; lifetime intake of grilled/barbecued poultry or fish from 10% to 3%; lifetime smoked beef, lamb, or pork consumption from 9% to 3%; lifetime intake of smoked poultry or fish from 10% to 5%; and lifetime consumption of all 4 types of food combined from 15% to 5%.

The results obtained from this imputed data set were not materially different from those obtained from the data set in which missing consumption values were simply dropped, although, as expected, confidence intervals were wider for the latter data set (data not shown). Sensitivity analyses revealed that substitution of more crudely derived imputations (for example the highest or lowest observed values) did not substantially affect the observed ORs (data not shown). Thus, the results based on models with the more precisely estimated regression coefficients imputed for the missing values are shown.

HCA and Benzo(α)pyrene Indices

Calculation of these exposure indices was based on data collected as part of the Block FFQ,²² which was modified to include a detailed assessment of frequency, preparation techniques, and doneness levels for 5 items (hamburgers, steak, pork, poultry, and fish).²³ Frequency of intake was categorized as never, <1 per month, 1 per month, 2–3 per month, 1–2 per week, 3–4 per week, 5–6 per week, and at least 1 time per day. For each meat and fish item, participants were asked, “how often during the last 12 months, did you eat meat … grilled or barbequed, pan-fried, (not deep-fat-fried), oven-broiled, oven-baked, microwaved, and other (specify: __________).” Levels of doneness were categorized as rare/medium, well-done, or very well-done. Frequency of cooking methods for each food item was converted to the common denominator of times per week. Responses to the frequency and portion sizes of food items were translated into daily intakes (in grams) for each food item using the National Cancer Institute DietSys program, version 3.²⁴

We examined 1 PAH [benzo(α)pyrene] and 3 HCAs [2-amino-1-methyl-6-phenyl-imidazo[4,5-b]pyridine (PhIP), 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx), and 2-amino-3,4,8-trimethylimidazo[4,5-f]quinoxaline (DiMeIQX,)]. The values per gram of food item were each applied to consumption (in grams) of hamburger, steak, pork, bacon, sausage, fried chicken, and other types of poultry using a modification of the method developed by Sinha.^3,4,25,26

PAH and HCA values for meats vary with cooking method, level of doneness, and consumption of the skin for poultry items.¹ To translate participants’ responses into daily intake in grams for each meat item, cooking method weights were created based on the percent of time a women used each specific cooking method. These weights were then multiplied to the HCA and benzo(α)pyrene values unique to that cooking method and doneness level. Weighted HCA and benzo(α)pyrene levels for each cooking method were then summed across cooking methods for each meat item. The resulting number of nanograms of HCA or benzo(α)pyrene per 1 g for each meat item was multiplied by the woman's intake (in grams) of that meat item.

Women who did not respond to a meat item's doneness level and frequency of intake were assumed to be nonconsumers of that item (assigned to 0). Among the minimal number of remaining women with missing data for the cooking method section, HCA and benzo(α)pyrene values for the missing meat item were set to missing; this was done for 28 hamburger responses [14 cases (0.9%), 14 controls (0.9%)], 28 steak responses [14 cases (0.9%), 14 controls (0.9%)], 44 pork responses [24 cases (1.6%), 20 controls (1.3%)], and 36 chicken responses [15 cases (1.0%), 21 controls (1.4%)]. The FFQ did not specifically ask cooking methods for steaks or roasts, so the frequency of cooking methods for “beef, including hamburger” was assigned to steaks or roasts. Cooking methods for bacon and sausage were also not specifically queried, so cooking methods reported for pork were assigned to bacon and sausage. In the control population, the categories “Don't know” and missing values for doneness of meat items were assumed to be the most common doneness response. For hamburger [22 (1.6%) cases, 35 (2.4%) controls] and steak [44 (1.7%) cases, 29 (2.0%) controls], missing values were assigned the medium doneness level that was selected by greater than 50% of controls. For pork, 41 (3.3%) cases and 51 (2.9%) controls, and for poultry, 28 (1.9%) cases and 40 (2.7%) controls were assigned the “well” doneness selected by greater than 50% of controls.

Covariate Assessment

The following variables were studied with regard to confounding or effect modification: 1) from the main questionnaire, reference age (defined as date of diagnosis for cases and date of identification for controls), menopausal status, race, education, age at first birth, parity, age at menarche, history of breast-feeding, use of oral contraceptives, use of hormone replacement therapy, family history of breast cancer, smoking status, physical activity, body mass index, alcohol intake; and 2) from the FFQ, consumption of energy, fruits and vegetables, and single and multiple vitamin use. Stratified analyses were performed to test for heterogeneity by stage of disease and estrogen receptor/progesterone receptor (ER/PR) status as previously described.¹⁹

Statistical Methods

The main exposure variables were total and average lifetime intake of grilled/barbecued and smoked meats estimated from the main questionnaire, and estimated intake of total benzo(α)pyrene, benzo(α)pyrene from meat only, benzo(α)pyrene from sources other than meat, and each of the 3 HCAs, all based on FFQ responses as described above. These variables were categorized into quantiles (deciles, quintiles, quartiles, tertiles, or dichotomous) based on menopause-specific distributions among the controls to assess the best representation of the data. Because the findings were similar regardless of the quantile chosen, we used tertiles to improve power during stratified analyses. Unconditional logistic regression was used to calculate ORs and 95% confidence intervals (CIs) of breast cancer in relation to tertiles of the main exposure variables. Crude ORs and corresponding 95% CIs adjusting only for age were calculated, as well as adjusted ORs controlling for potential confounders.

Effect modification on the multiplicative scale was evaluated by comparing the ORs for meat intake variables across strata of covariates and testing for heterogeneity. We created interaction terms between meat intake variables and covariates, and conducted a likelihood ratio test to compare the model with the interaction terms to the reduced model containing only the individual variables. We assessed additive interaction by computing interaction contrast ratios using the hypothesized lowest risk category of low meat intake and high fruit and vegetable intake as the common referent and comparing all other joint categories of meat and fruit and vegetable intake to this common referent. In these data, menopausal status was not an effect modifier on a multiplicative scale. Nonetheless, because we²⁷ and others²⁸ have found breast cancer risk in relation to diet to vary with menopause, and because there was some indication that the lifetime cooked meat results varied with menopausal status, we present all analyses stratified by menopausal status.

We assessed confounding with backward elimination, beginning with a full model. A covariate remained in the model if the OR of the reduced model (without the covariate) changed by more than 10% when compared with the OR of the full model. Tests for trend were calculated by setting each tertile equal to the median value and treating these 3 median tertile values as a continuous variable.

RESULTS

Breast cancer risk among postmenopausal women was elevated in relation to measures of both total and average lifetime intake of grilled/barbecued and smoked meats (as assessed in the main questionnaire; Table 1). Compared with women in the lowest tertile, ORs for total lifetime intake among women in the second and third tertile were both 1.47 (95% CI = 1.11–1.95 and 1.12–1.92, respectively); the corresponding results for average lifetime intake were similar (OR = 1.48 [CI = 1.13–1.93] and 1.35 [CI = 1.02–1.79], respectively). Modest increased risks among postmenopausal women were also observed in relation to average and total lifetime intake of grilled/barbecued beef, pork, and lamb, and in relation to total lifetime intake of smoked ham, pork, and lamb. No associations were observed for any of the lifetime cooked meat intake variables among premenopausal women (Table 1). Further, no associations were observed for intakes of grilled/barbecued poultry or smoked fish among pre- or postmenopausal women (data not shown).

TABLE 1: Adjusted ORs and 95% CIs for Breast Cancer by Intake of Grilled/Barbecued and Smoked Meats Over the Woman's Lifetime (Assessed From Main Questionnaire) in the Long Island Breast Cancer Study Project, 1996–1997

We also examined the association between breast cancer and intake of cooked meat by decade of life (data not shown). For postmenopausal women, although some individual associations were noted, no clear patterns emerged to suggest differential risk by intake in specific decades of life. In general, associations between the various cooked meat variables and breast cancer risk were spread throughout the lifetime. When examining associations by decade of life in premenopausal women, a few associations were observed in the second but not the third tertiles of intake.

As shown in Table 2, intakes of benzo(α)pyrene and HCAs in the previous year (as assessed by the FFQ) were not associated with increased risk of breast cancer. Contrary to our hypothesis, for premenopausal women, a reduced risk of breast cancer was observed in women in the third tertile of intake of 2 HCAs (MeIQx and DiMeIQx) compared with women in the first tertiles. Among postmenopausal women, no associations were observed with regard to benzo(α)pyrene and HCA indices.

TABLE 2: Adjusted ORs and 95% CIs for Breast Cancer by Intake of Polycyclic Aromatic Hydrocarbons and Heterocyclic Amines in the Previous Year (Assessed From FFQ)

As shown in Table 3, when the results among postmenopausal women were further stratified by fruit and vegetable intake (as assessed in the FFQ), the increased risk of breast cancer associated with the highest tertile of lifetime consumption of grilled/barbecued and smoked meats was limited to women in the lowest category of fruit and vegetable intake (P value for multiplicative interaction = 0.25). The increased risk associated with intake of smoked ham, pork, and lamb was also limited to consumers of low fruit and vegetable diets. No associations between benzo(α)pyrene or HCA intake and breast cancer were observed in either high or low consumers of fruits and vegetables for postmenopausal women (Table 4). There was no evidence for additive interaction for any of the meat variables with fruit and vegetable intake. Additionally, we examined the interactions shown in Tables 3 and 4 for premenopausal women separately, and found no joint effects between any of the meat intake variables and fruit and vegetable intake (data not shown).

TABLE 3: Adjusted ORs and 95% CIs for Postmenopausal Breast Cancer by Intake of Grilled/Barbecued and Smoked Meats Over the Women's Lifetime (Assessed From Main Questionnaire) Stratified by Fruit and Vegetable Intake

TABLE 4: Adjusted ORs and 95% CIs for Postmenopausal Breast Cancer by Intake of Polycyclic Aromatic Hydrocarbons and Heterocyclic Amines in the Previous Year (Assessed From FFQ) Stratified by Fruit and Vegetable Intake

Our large sample size allowed for examination of effects within joint categories of estrogen- and progesterone-receptor status in postmenopausal women. The risk of breast cancer among postmenopausal women consuming high amounts of grilled/barbecued meats over their lifetime was similar for ER+/PR+ and ER−/PR− tumors (data not shown). We found no strong associations for any of the benzo(α)pyrene or HCA intake variables assessed from the FFQ across categories of ER/PR status (data not shown), with the exception of a positive association observed in ER+/PR+ cases (OR = 1.47; CI = 0.99–2.19; P for trend = 0.07) in the highest tertile of benzo(α)pyrene from meats intake when compared with those in the lowest tertile. No consistent associations were observed for meat intake variables within ER+/PR− and ER−/PR+ tumors, though the small numbers of cases within these categories limited our power to detect associations within these groups (data not shown).

Analyses stratified by stage of diagnosis (in situ versus invasive) revealed slight differences in effect estimates for lifetime intake variables, although these were inconsistent across different intake variables (data not shown). For example, the increased risk associated with total grilled/barbecued beef, lamb, or pork was slightly higher for in situ cases, compared with invasive cancers, in postmenopausal women (for third tertile compared with the first tertile, the OR for in situ cancer was 1.41 [95% CI = 0.79–2.52] and for invasive cancers it was 1.23 [0.92–1.66]). In contrast, for total smoked ham, lamb, or pork, the increased risk was higher for invasive cancer (OR = 1.43; CI = 1.01–2.03) than for in situ cases (1.04; 0.53–2.05). No differences in effect estimates between in situ and invasive cancer were observed in the levels of benzo(α)pyrene and HCA intake in the year before interview (data not shown).

DISCUSSION

Using data from a large population-based case-control study, we examined associations of breast cancer risk with lifetime intake of grilled/barbecued and smoked meats, and also with intake of benzo(α)pyrene and HCAs in the previous year. Modest positive associations were found for high lifetime intake of grilled/barbecued and smoked meats in postmenopausal women. These associations were strongest among women with low consumption of fruits and vegetables. The associations did not appear to differ substantially by tumor stage or hormone receptor status.

We found no consistent associations between breast cancer and high dietary intake of HCAs or benzo(α)pyrene in the previous year among pre- or postmenopausal women. Two other studies of recent diet and breast cancer risk found no associations with cooked or processed meat,^14,29 and 2 large prospective studies of up to 18 years of follow-up found no associations between meat intake and breast cancer, although detailed information related to cooking methods and doneness preferences were not collected.^15,16 In contrast, intake of very well-done cooked red meat was associated with increased risk of breast cancer in 1 case-control study,¹² and intake of well-done deep-fried red meat in the previous 5 years was highly associated with increased risk of breast cancer, particularly in women at high body mass indexes in another study.¹¹

We found high lifetime intake of grilled/barbecued and smoked meats was associated with increased risk of breast cancer among women consuming few fruits and vegetables. We previously reported an inverse association between fruit and vegetable intake and postmenopausal breast cancer in this population.²⁷ Experimental studies suggest that chemopreventive constituents of fruits and vegetables, such as isothiocyanates and chlorophyllin, may protect against PAH- and HCA-induced genotoxicity.^17,18 Our results support laboratory findings in animal models that fruits and vegetables may confer some protection against the harmful effects of cooked meat intake.

In animal models, benzo(α)pyrene and HCAs are potent mammary carcinogens. Their mechanism of action is believed to be through direct damage to DNA (formation of DNA adducts).⁵ However, more recently, there is evidence from cell culture experiments that 1 HCA (PhIP, but not MeIQx) may be estrogenic.³⁰ The majority of epidemiologic and animal model evidence supports a causal relationship between estrogen levels and breast cancer risk. Thus, it is surprising that we did not observe increased risk for breast cancer with increasing intake of HCAs, and in particular, PhIP. One possible explanation may be that intake in the year before diagnosis is not representative of a woman's lifetime intake of HCA-related foods. Most cancers are believed to be slow-growing, developing over a lifetime with initiating events occurring possibly as early as adolescence. A few studies have attempted to retrospectively assess adolescent diet and examine associations with breast cancer,^31–34 with few significant findings. One study by Baer et al³⁵ reported positive associations of intake of total meat, red meat, and hot dogs with risk of proliferative benign breast disease. The numbers of questions relating to diet are usually limited in these studies, and none has attempted to assess meat cooking methods in adolescence or over the lifetime.

Also surprising were the inverse associations between the third tertile of MeIQx and DiMeIQX intake as compared with the first tertiles and breast cancer among premenopausal women. Another study found decreased risk of breast cancer for the highest quartile of PhIP intake as compared with the lowest quartile (which was attenuated after adjustment for chicken intake), and nonsignificant decreased risk for the highest quartiles of MeIQx and DiMeIQx as compared with the lowest quartile in premenopausal and postmenopausal women combined.¹⁰ Chicken is a major contributor to PhIP intake. The authors suggest this may partially explain the inverse associations for PhIP because white meat may help support proper immune function, or may be a surrogate for other healthy lifestyle factors. However, 2 other studies found either no effect of DiMeIQX and MeIQX and an increased risk with PhIP intake in women aged 55–69 years,⁸ or increased risks in postmenopausal women with high intakes of MeIQx and PhIP and no associations in premenopausal women.⁹ The ranges of estimated HCA intake in these studies were similar to ours. Foods high in MeIQx are pan-fried hamburger, sausage, and steak cooked well-done or very well-done, whereas DiMeIQx has been detected in small quantities in pan-fried very well-done steak and in pan-fried and grilled/barbecued chicken.^4,36 Thus, the inverse associations in premenopausal women in our study are unexplained and may be due to chance.

Using an in-person, interviewer-administered questionnaire, we indirectly assessed dietary intake of PAHs by asking about intake of 4 types of meat prepared either as grilled/barbecued or smoked during each decade of life. Our results based on these measures must be interpreted with care because the lifetime meat intake questionnaire has not been validated. Although the validity of recall of lifetime intake of cooked meats has not been directly addressed in the literature, we can extrapolate from other studies that have attempted to validate long-term recall of dietary intake. These studies have reported modest correlations between dietary assessments at one point in time, and recall of the same diet 11–24 years later (average correlations for food and nutrient intakes ranged from 0.23 to 0.59).^26,37–41 Recall of adolescent diet in adulthood as compared with maternal reporting has also been examined in the Nurses’ Health Study II, and moderate correlations were observed (correlations ranged from 0.13 to 0.59).⁴² Misclassification of exposure, when the variable is multilevel, may result in bias toward or away from the null. An additional concern is that the accuracy of long-term recall may have differed by case or control status; however, this is unlikely for the following reasons. First, previous validation studies of long-term recall of diet have not found differential recall between cancer cases and controls.^40,41,43 Second, to assess whether preconceived notions of breast cancer etiology would bias a women's recall of meat intake, we examined the participants’ responses to what they believed caused breast cancer. Few women listed meat intake as a causal factor, and among these women there was no case-control difference. Therefore, recall bias is unlikely to have a strong influence on our results.

Given the known measurement error in FFQs, nondifferential misclassification of cooked meat intake variables may be another explanation for our findings of no association between recent intake of benzo(α)pyrene and HCAs and breast cancer. Unlike some studies showing positive associations,^8,12 we did not use color photographs of meat cooked to varying levels of doneness, but instead relied on written descriptions of doneness preference, which may be a less accurate means of collecting these data. Additionally, some meat items that do not have similar HCA or PAH values were grouped together in the food questionnaire. For example, steaks and roasts were grouped in “beef (steaks, roasts, etc., including on sandwiches).” However, the content of HCAs and PAHs in meats is highly dependent on cooking methods, so the fact that we used reported frequency of the cooking methods to assign HCA and PAH values to the meat items should have improved our estimation of intake of these carcinogens over the basic FFQ. We also did not specifically ask cooking methods for bacon and sausage, but instead assigned the reported cooking methods for pork to that of bacon and sausage. As the true cooking methods for these different meat items may vary considerably (eg, pan-frying for bacon and sausage vs. oven-broiling or baking for pork) and the corresponding HCA and benzo(α)pyrene content of bacon is different depending on whether the meat is oven-broiled or pan-fried,^1,3 this assumption may have led to misclassification of exposure.

Due to the lack of validation of methods to assess long-term dietary recall, it is possible that current diet, if correlated with recall of long-term diet, could confound the relationship between lifetime intake and breast cancer.³⁵ We examined the correlation between intake of grilled/barbecued beef, pork, and lamb from the lifetime intake questionnaire with intake of grilled and barbecued beef and pork on the FFQ. The Spearman correlation coefficient between intake before age 20 and intake in the year before interview was weak (r = 0.25), whereas the correlation between intake in the previous decade and intake in the year before interview was stronger (r = 0.53). Similar results were found for poultry and fish, and the results did not differ greatly between cases and controls. These correlations support the findings from Baer et al³⁵ that current diet is not strongly correlated with long-term recall of diet.

Given that we did not collect data on overall lifetime dietary intake (especially lifetime total meat intake irrespective of the cooking methods), we cannot exclude the possibility that our observed results may be due to other potentially correlated dietary factors, such as lifetime total meat intake or lifetime total fat intake, rather than the specific cooking methods of meat that were queried. We examined confounding by all known risk factors for breast cancer. However, there may have been residual confounding by exposures measured with error, such as other dietary intake variables, lifetime physical activity, or alcohol use. It is unlikely, though, that residual confounding could fully account for our findings.

Our study supports a role in postmenopausal breast cancer etiology for lifetime intake, but not more recent intake, of meats cooked in ways that enhance PAH formation, particularly among women consuming few fruits and vegetables. Future investigation into this area in larger studies may provide useful information on the contribution of these dietary exposures to breast cancer etiology.

ACKNOWLEDGMENTS

For their valuable contributions to the Long Island Breast Cancer Study Project, the authors thank: members of the Long Island Breast Cancer Network; the 31 participating institutions on Long Island and in New York City, NY; our National Institutes of Health collaborators, Gwen Collman, National Institutes of Environmental Health Sciences; G. Iris Obrams, formerly of the National Cancer Institute; members of the External Advisory Committee to the population-based case-control study: Leslie Bernstein, (Committee chair) Gerald Akland, Barbara Balaban, Blake Cady, Dale Sandler, Roy Shore, and Gerald Wogan, as well as other collaborators who assisted with various aspects of our data collection efforts including Steve Stellman, Maureen Hatch, Mary Wolff, Geoff Kabat, Gail Garbowski, H. Leon Bradlow, Martin Trent, Ruby Senie, Carla Maffeo, Pat Montalvan, Gertrud Berkowitz, Margaret Kemeny, Mark Citron, Freya Schnabel, Allen Schuss, Steven Hajdu, and Vincent Vinciguerra.