Background: Heterocyclic amines, mutagens formed in meats cooked at high temperatures, have been demonstrated as mammary carcinogens in animals. We conducted a nested, case-control study among 41 836 cohort members of the Iowa Women's Health Study to evaluate the potential role of heterocyclic amines and intake of well-done meat in the risk for human breast cancer. Methods: A questionnaire was mailed to individuals in the cohort who had breast cancer diagnosed during the period from 1992 through 1994 and a random sample of cancer-free cohort members to obtain information on usual intake of meats and on meat preparation practices. Color photographs showing various doneness levels of hamburger, beefsteak, and bacon were included. Multivariate analysis was performed on data from 273 case subjects and 657 control subjects who completed the survey. Results: A dose- response relationship was found between doneness levels of meat consumed and breast cancer risk. The adjusted odds ratios (ORs) for very well-done meat versus rare or mediumdone meat were 1.54 (95% confidence interval [CI] = 0.96-2.47) for hamburger, 2.21 (95% CI = 1.30-3.77) for beef steak, and 1.64 (95% CI = 0.92- 2.93) for bacon. Women who consumed these three meats consistently very well done had a 4.62 times higher risk (95% CI = 1.36-15.70) than that of women who consumed the meats rare or medium done. Risk of breast cancer was also elevated with increasing intake of well-done to very well-done meat. Conclusions: Consumption of well-done meats and, thus, exposures to heterocyclic amines (or other compounds) formed during high-temperature cooking may play an important role in the risk of breast cancer.
High consumption of meat has been reported to be associated with an increased risk of breast cancer in some epidemiologic studies ( 1 ) , including several large cohort studies ( 2–4 ) . Dietary fat has long been suspected to be responsible for the meat-breast cancer association. This hypothesis, however, has not been supported by most prospective cohort studies ( 5 ) . Accumulating evidence from recent animal studies and some human studies has implicated heterocyclic amines in the pathogenesis of breast cancer ( 6 , 7 ) .
Heterocyclic amines are a group of mutagenic compounds identified in cooked foods, particularly in well-done meats and fish ( 7–9 ) . These compounds are formed as pyrolysis products of amino acids and proteins and are some of the most potent mutagens detected by use of the Ames test ( 8–10 ) . In animal studies, heterocyclic amines have been shown to increase the occurrence of various tumors, including those of the mammary glands ( 10–12 ) . Oral administration of 2-amino- 1-methyl-6-phenylimidazo[4,5b]pyridine (PhIP), the most abundant carcinogenic heterocyclic amine in cooked meats, has been shown to induce mammary tumors in rats ( 12 , 13 ) in a dose-response manner. Furthermore, PhIP-induced mammary carcinomas exhibit an unusually high frequency of guanine to adenine transitions at the second base of codon 12 of the Ha-ras oncogene in F344 rats ( 11 ) , an unusual type of mutation observed in tumors induced by other mammary carcinogens, such as N-methyl-N-nitrosourea ( 11 ) , strongly suggesting that heterocyclic amines are potent mammary carcinogens.
Despite evidence from animal studies and in vitro experiments ( 6–14 ) , the potential role of heterocyclic amines as a risk factor for human breast cancer has not been appropriately investigated. Epidemiologic studies have attempted to measure the association between heterocyclic amines and risk of breast cancer, using dietary questionnaires that include consumption levels of fried or broiled meats as surrogate measures of heterocyclic amine exposure ( 15–19 ) . However, because the levels of heterocyclic amines in foods depend, to a large extent, on the duration and temperature of cooking, it is important to obtain information on the degree of meat doneness to estimate levels of heterocyclic amine exposure ( 8 , 20 , 21 ) . No epidemiologic study of diet and breast cancer has incorporated this information into the exposure assessment. To evaluate the role of well-done meat intake as a risk factor for breast cancer, we conducted a nested, case-control study of cases of breast cancer occurring during the period from 1992 through 1994 among participants in the Iowa Women's Health Study.
Subjects and Methods
Iowa Women's Health Study
Detailed descriptions of this cohort study have been published elsewhere ( 22–24 ) . Briefly, 41 836 Iowa women, aged 55-69 years, who completed a mailed questionnaire in January 1986, have been followed for mortality and for cancer incidence. The follow-up was accomplished through computer linkage of study participants with Iowa death certificate files, the National Death Index, and cancer diagnosis data collected by the Iowa State Health Registry, part of the Surveillance, Epidemiology, and End Results (SEER) Program1 of the National Cancer Institute. The self-administered questionnaire used in the 1986 baseline survey included information on diet and other major risk factors for cancer. Diet was assessed by use of a semiquantitative foodfrequency questionnaire almost identical to that used in the 1984 Nurses' Health Study ( 25 ) . Usual intake of specified portions of 127 food items was ascertained, and nutrient intake was estimated by use of the nutrient database developed for the Nurses' Health Study. Because no information was collected at the baseline survey on usual intake of meats by cooking method and usual doneness levels of meat, a case-control study with a supplementary survey was conducted during the period from 1995 through 1996 in a subset of cohort members.
Case-Control Study of Breast Cancer
Eligible case subjects for this study included all cohort members who had breast cancer diagnosed during the period from 1992 through 1994 (n = 453). A control sample of 900 women was randomly selected from 27 186 cohort members who were alive and free of cancer on January 1, 1992, and participated in the 1992 follow-up survey. Of these 900 women, 24 were excluded from the control group because they were later found either to have a breast cancer diagnosis during the period from 1992 through 1994 (n = 3) or to have been selected to participate in other Iowa Women's Health Study ancillary projects (n = 21) (eligible control subjects, n = 876). The three breast cancer patients were included in the case group.
All eligible subjects were asked to complete a self-administered food-frequency questionnaire on meat intake habits during the “reference” year (defined below). This questionnaire, which was developed on the basis of a questionnaire provided by Dr. Sinha for the use of in-person interviews ( 26–28 ) , included questions on usual intake and preparation of 15 meats. The meats are hamburgers or cheeseburgers; beefsteaks; pork chops; bacon; breakfast sausage links; breakfast sausage patties; other sausages that are cooked before eating; bratwurst; hot dogs or franks; chicken or turkey breast, thighs, legs, or wings; fish (including fried fish sandwiches, tuna); venison; game poultry; and smoked meats and fish. For each meat item, intake according to the following cooking methods was assessed: grilled or barbecued, fried, oven broiled, and prepared in other ways, as well as from “fast food,” if applicable. In addition, participants reported their usual preference for level of meat doneness by use of a series of color photographs that represented increasing levels of doneness of hamburger (four photographs), beefsteak (four photographs), and bacon (three photographs). Photographs were labeled only with number and represented a range of doneness levels from extremely rare to very well done. Information related to the doneness level of other 12 food items was not obtained in this study.
To minimize the potential effect of a breast cancer diagnosis on dietary intake, we obtained information for case subjects on usual dietary habits 1 year before cancer diagnosis. Because breast cancer cases were diagnosed during the period from 1992 through 1994 and dietary assessment was conducted during the period from 1995 through 1996, three reference years (1991, 1992, and 1993), corresponding to the years immediately before breast cancer diagnosis, were identified. Women in the control group of the study were also divided randomly into three corresponding groups with approximately equal sample sizes to obtain their dietary habits during these three reference years. A list of national and international events that occurred during the reference year was provided in the questionnaire to help participants recall their eating habits during that year.
Two hundred seventy-three (60.3%) of selected case subjects and 657 (75.0%) of the control subjects participated. The major reasons for nonparticipation were refusal (29.1% of case subjects and 18.7% of control subjects), inability to locate (4.9% of case subjects and 3.8% of control subjects), and death before contacting (5.7% of case subjects and 2.5% of control subjects).
Odds ratios (ORs) were used to measure the strength of the association between exposures and cancer risk ( 29 ) . Unconditional logistic regression was used to control for potential confounders, assessed mostly at the baseline survey, and to derive adjusted ORs and 95% confidence intervals (CIs). Dietary intakes were adjusted for total energy intake to reduce potential variation that may have arisen from overreporting or underreporting of food intake. Trend tests for dose-response relationships across levels of each dietary variable were performed by treating ordinal-score variables (with values of 1, 2, 3, . . .) as continuous variables in logistic regression. Reported P values (two-sided) were from x2 tests (for categorical variables), Wilcoxon rank-sum tests (for continuous variables), and age-adjusted linear regression models using log-transferred data.
Meats were categorized by type (red versus white) and preparation method to facilitate statistical analyses. The “all meat” category included all 15 meats listed in the questionnaire. The “red meat” category included hamburgers, cheeseburgers, beefsteaks, pork chops, bacon, breakfast sausage links, breakfast sausage patties, other sausages, bratwurst, and hot dogs or franks. The “white meat” group included chicken or turkey breast; chicken or turkey thighs, legs, or wings; and fish. Individual meats and meat groupings were further classified by preparation method. Fried, grilled, and broiled red meats referred to the preparation methods of the aforementioned red meat category. Intake in grams was estimated for each food item included in the questionnaire. Thus, analyses using meat groups were based on the summation of the individual gram weights of the meats consumed in each category. A doneness score was also calculated to describe the eating preferences of participants on the basis of their responses to the color photographs. Doneness levels of rare or medium, well done, and very well done were given scores of 1, 2, and 3, respectively. The doneness score was defined as the sum of the doneness preferences for each of the three meat photographs. For example, a person who reported usually consuming rare or medium-done hamburger, beefsteak, and bacon received a score of 3, whereas a person reporting usual intake of all three meats as very well done received a score of 9.
The distributions of demographic and risk factors from the baseline survey conducted in 1986 are shown in Table 1 for all eligible subjects (left three columns) and for those who completed the supplementary questionnaire (right three columns). Among study participants, a family history of breast cancer, use of hormone replacement therapy, and waistto- hip ratio were positively associated with risk of breast cancer ( P <.05). Therefore, all analyses were adjusted for these variables. No apparent association with other variables, including dietary fat and total energy intake, was observed among study participants. Similar findings were observed from the analyses that included all eligible subjects selected for the study, suggesting that study participants represent reasonably well the eligible subject cohort in the distribution of major factors for breast cancer.
The median intake levels of major meat categories by case-control status are presented in Table 2 . Total meat intake was higher among breast cancer case subjects than among control subjects. This difference was apparently due to a higher intake of red meats (17.7% case-control difference) but not white meats (2.1% difference). Because heterocyclic amines are formed primarily in fried, grilled, and broiled meats and human exposure to heterocyclic amines is primarily through intake of meats prepared by these three cooking methods, these meats were combined into one group for data analysis. There was an 18.8% ( P = .06) difference in the median intake of fried, grilled, and broiled red meats between case subjects (26.5 g/day) and control subjects (22.3 g/day). Case subjects also consumed more red meats prepared in other ways than did control subjects (20.9% difference, P 4 .16), but intake of these meats was low, accounting for less than 15% of total red meats consumed. Compared with control subjects, case subjects had a slightly higher intake of fried, grilled, and broiled white meats (7.6%) but a lower intake of white meats prepared by other cooking methods (-5.6%). These case-control differences were not statistically significant. Intake of smoked meats and fish was low, and no apparent association with risk of breast cancer was observed. The P values from age-adjusted linear regression models using natural log-transformed data were similar to those obtained from the Wilcoxon rank-sum tests.
To evaluate the potential dose-response relationship between risk of breast cancer and intake levels of various meats, case subjects and control subjects were categorized into four groups for each meat variable listed in Table 2 , according to the quartile distribution of the variable among controls; ORs for the upper three quartiles were estimated, as compared with the lowest quartile. There was a statistically significant positive association between intake of red meat and risk of breast cancer ( P for trend, .02), with a 78% elevated risk observed for the highest versus the lowest intake quartile group (data not shown). High intake of fried, grilled, or broiled red meats and of other red meat preparations was associated with increased risk of breast cancer, but the trend tests were not statistically significant. No apparent association was observed for other meat groups listed in Table 2 .
Hamburger, beefsteak, and bacon accounted for more than 60% of red meat intake in this study population. Information on doneness preferences for these meats was obtained, and their association with risk of breast cancer is shown in Table 3 . There was a clear, positive, dose-response association between doneness levels of each of the three meats and risk of breast cancer. Women who usually consumed all three of these meats at a very well-done level had a 4.62-fold elevated risk (95% CI = 1.36-15.70) of breast cancer compared with those who usually ate these meats at a rare or medium doneness level.
Information on doneness levels and intake levels of hamburger, beefsteak, and bacon was combined to further classify the exposure status of case subjects and control subjects ( Table 4 ). The risks of breast cancer were elevated with increasing doneness of meats in all intake groups. As previously mentioned, a high intake of red meat was related to an increased risk of breast cancer, but this positive association is less notable in comparison with that for meat doneness level.
To our knowledge, this is one of the most comprehensive epidemiologic studies to date to evaluate the hypothesis that the intake of well-done meat may be related to the risk of breast cancer. We found that a preference for consuming well-done meats was associated with an elevated risk of breast cancer in a dose- response manner. In contrast, high intake of dietary fat was not associated with risk of breast cancer in the cohort of the Iowa Women's Health Study ( 22 ) , and intake of red meat was only weakly associated with the risk of breast cancer in this subcohort. These findings suggest that heterocyclic amines and possibly other compounds, such as polycyclic aromatic hydrocarbons, formed during hightemperature cooking of animal foods may be related to the risk of breast cancer.
Heterocyclic amine exposure, as measured indirectly in epidemiologic studies by levels of fried-food intake and a preference for heavily browned (well-done) meat, has been repeatedly shown to be related to an increased risk of colorectal cancer ( 30 , 31 ) . Although many epidemiologic studies have been conducted to evaluate dietary hypotheses for breast cancer, only a few of them have investigated the relationship between the intake of well-done meat and the risk of breast cancer. In a prospective cohort study in Finland ( 15 ) , a significant 80% increase in risk of breast cancer was found for high intake of fried meats, while intake of nonfried meat was not related to risk. High intake of fried meat was also reported to be positively associated with breast cancer in a recently published hospital-based, case-control study ( 19 ) . In that study, a 2.7-fold elevated risk (95% CI = 1.61-4.55) of breast cancer was observed among women in the uppermost quartile of fried meat intake, whereas intake of boiled meat was inversely related to breast cancer risk. These studies suggest that the risk of breast cancer may be more closely related to meat-cooking methods than to the level of meat intake per se. A positive relationship between intake of fried meat and breast cancer was also reported from several early case-control studies ( 16–18 ) . As with the two more recent studies ( 15 , 19 ) , however, information on other measures of heterocyclic amine exposures, such as degree of meat doneness and intake of grilled, barbecued, or broiled meats, was usually not obtained. Therefore, considerable misclassification in the assessment of heterocyclic amine exposure may exist in these previous studies.
In contrast to the strong positive association of doneness levels of meat with breast cancer, we found only a weak positive association for intake of fried, grilled, and broiled meats. It is likely that an individual's intake preference for meat doneness level is more consistent over time than preference for cooking method, especially when one considers all settings (e.g., restaurant or home) where a person consumes food. Thus, recall of doneness level may be more accurate than for preparation type. It is also possible that the use of photographs that represented various doneness levels of meats facilitated the ascertainment of information for this set of variables. Alternatively, doneness level may be a stronger predictor for cancer risk than intake levels of fried, grilled, and broiled meat with all doneness levels combined. Indeed, it has been shown that meats that have been fried, grilled, or broiled to “just until done” contain very low levels of heterocyclic amines ( 21 ) . Therefore, it is important in epidemiologic studies to obtain information on the level of meat doneness in the assessment of heterocyclic amine exposure.
As with any epidemiologic study of diet and cancer, the measurement error in assessing dietary intake may be a concern in this study, particularly because the questionnaire used in this study was relatively new. By use of a similar questionnaire, well-done meat was found to increase the risk of colorectal adenomas in a recent study ( 32 ) , providing some assurance to the validity of this questionnaire. Furthermore, there is no reason to speculate that breast cancer case subjects would differentially recall intake of well-done to very well-done meats versus control subjects, particularly because the hypothesis investigated in this study was relatively new, and consumption of well-done hamburger is recommended in news media to reduce the risk of Escherichia coli infection. Therefore, the measurement errors in dietary assessment are most likely to be random, which may attenuate the exposure- disease association in most situations. We have adjusted for all other risk factors for breast cancer in the analyses; however, residual confounding, particularly from other dietary variables that may have been measured with error, may remain a concern. The participation rate (60%–75%) in this study was comparable to most case-control studies. Still, about 40% of case subjects and 25% of control subjects did not participate, and this may have caused selection bias if participation were related to exposure. As shown in Table 1 , however, participants were, in general, similar to the subcohort of women who were eligible for the study in virtually all baseline risk factors and dietary habits, indicating that the potential for a selection bias may be minimal.
In summary, this case-control study found that intake preference for well-done meats and high intake of well-done, particularly very well-done, meats were associated with an increased risk of breast cancer. These findings, along with evidence accumulated from previous epidemiologic and laboratory studies, strongly suggest that heterocyclic amines and possibly other compounds formed during high-temperature cooking may be breast carcinogens in humans. It has been shown that several metabolic enzymes are involved in the activation or detoxification of heterocyclic amines in humans ( 33 , 34 ) , and investigation into the associations of these enzymes with breast cancer risk may help shed light on causal pathways for this common cancer. Therefore, future epidemiologic studies exploring the heterocyclic amine-breast cancer hypothesis should focus on the improvement of exposure assessment of heterocyclic amines and evaluate the potential effect of genetic variability in metabolic activities of these enzymes in the association of heterocyclic amine exposure and breast cancer risk.