Journal of Exposure Analysis and Environmental Epidemiology (2001) 11, 155±168 # 2001 Nature Publishing Group All rights reserved 1053-4245/01/$17.00

www.nature.com/jea

U.S. dietary exposures to heterocyclic amines1

KENNETH T. BOGEN2 AND GARRETT A. KEATING

Lawrence Livermore National Laboratory, University of California, Livermore, California 94550

Heterocyclic amines (HAs) formed in fried, broiled or grilled meats are potent mutagens that increase rates of colon, mammary, prostate and other cancers in bioassay rodents. Studies of how human dietary HA exposures may affect cancer risks have so far relied on fairly crudely defined HA-exposure categories. Recently, an integrated, quantitative approach to HA-exposure assessment (HAEA) was developed to estimate compound-specific intakes for particular individuals based on corresponding HA-concentration estimates that reflect their meat-type, intake-rate, cooking-method and meat-doneness preferences. This method was applied in the present study to U.S. national Continuing Survey of Food Intakes by Individuals (CSFII) data on meats consumed and cooking methods used by >25,000 people, after adjusting for underreported energy intake and conditional on meat-doneness preferences estimated from additional survey data. The U.S. population average lifetime time-weighted average of total HAs consumed was estimated to be 9 ng/kg/day, with 2-amino-1- methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) estimated to comprise about two thirds of this intake. Pan-fried meats were the largest source of HA in the diet and chicken the largest source of HAs among different meat types. Estimated total HA intakes by male vs. female children were generally similar, with those by (0- to 15-year-old) children 25% greater than those by (16+ -year-old) adults. Race-, age- and sex-specific mean HA intakes were estimated to be greatest for African American males, who were estimated to consume 2- and 3-fold more PhIP than white males at ages <16 and 30+ years, respectively, after considering a relatively greater preference for more well-done items among African Americans based on national survey data. This difference in PhIP intakes may at least partly explain why prostate cancer (PC) kills 2-fold more African American than white men, in view of experimental data indicating that PhIP mutates prostate DNA and causes prostate tumors in rats. Journal of Exposure Analysis andEnvironmental Epidemiology (2001) 11, 155±168.

Keywords: age, cancer, cooked-food mutagens, energy, meat, prostate, race, sex, variability .

Introduction Felton and Knize, 1990a,b; Sinha et al., 1995). Dietary exposure to PhIP has been shown to induce colon, intestinal Heterocyclic amines (HAs) are potent mutagens formed at and mammary adenocarcinomas in rats (Ohgaki et al., 1986; particularly elevated concentrations in well-done meats, Ochiai et al., 1991; Ito et al., 1991, 1997; Ghoshal et al., chicken and fish (Keating et al., 1999). HAs also cause 1994), as well as prostate cancer (PC) in rats (Shirai et al., cancer at a variety of sites in multiple bioassay animal 1997,1999). Another predominant HA found in cooked species/strains/sexes, as well as at multiple sites within meats, 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline many of species/strains/sexes tested (Bogen, 1994). A (MeIQx), causes tumors at multiple sites in rats and mice predominant HA found in cooked and particularly in well- (Kato et al., 1986; Bogen, 1994). HAs are metabolically done chicken and beef is 2-amino-1-methyl-6-phenyli- activated by P450 and N-acetyltranserase enzymes, and midazo[4,5-b]pyridine (PhIP) (Felton et al., 1984, 1986; activated forms are observed to bind covalently to DNA at sites including tissues (including prostate) in which HAs induce cancer in rats (Thorgeirsson, 1984; Thorgeirsson et 1. Abbreviations: A C, 2-amino-9H-pyrido[2,3-b]indole; 2-amino- al., 1983; Rosenkranz and Mermeistein, 1985; Sato et al., -carboline (26148-68-5); BMR, basal metabolic rate; BW, body weight; cmf, cumulative probability mass function; CSFII, Continuing 1986; Kato and Yamazoe, 1987; Snyderwine and Battula, Survey of Food Intakes by Individuals; DiMeIQx, 2-amino-3,4,8- 1989; McManus et al., 1990; Turesky et al., 1991; Davis et trimethylimidazo[4,5 -f]quinoxaline (95896- 78-9); HA, heterocyclic al., 1993; Kaderlik et al., 1994a,b; Takahashi et al., 1998; amine; HAEA, HA-exposure assessment; IQ, 2-amino-3-methylimida- Schut and Snyderwine, 1999). zo[4,5-f]quinoline (76180-96-6); IT, maximum internal temperature; Increased risks of colorectal adenoma and adenocarci- LBM, lean body mass; MeIQx, 2-amino-3,8-dimethylimidazo[4,5- f]quinoxaline (77500-04-0); PC, prostate cancer; PhIP, 2-amino-1- noma, and of stomach, breast, lung and PCs have been methyl-6-phenylimidazo[4,5-b]pyridine (105650-23-5); SE, standard weakly or clearly associated with potential or estimated HA error. exposure in most (Shiffman and Felton, 1990; Gerhardsson 2. Address all correspondence to: Kenneth T. Bogen, Health & Ecological de Verdier et al., 1991; Lang et al., 1994; Probst-Hensch et Assessment Division (L-396), Lawrence Livermore National Laboratory, al., 1997; Ward et al., 1997; Kampman et al., 1999; Norrish 7000 East Avenue, Livermore, CA 94550-9900. Tel.: +1-925-422- 0902. Fax: +1-925-424-3255. E-mail: [email protected] et al., 1999; Sinha et al., 1998a, 1999; Zheng et al., 1998, Received 2 October 2000; accepted 19 January 2001. 1999), but not all (Lyon and Mahoney, 1988; Muscat and Bogen andKeating Dietary exposures to heterocyclic amines

Wynder, 1994; Augustsson et al., 1999), case±control Methods studies investigating possible associations between in- creased human cancer risk and dietary HA-exposure IntegratedHA-Exposure Assessment categories based on self-reported meat-cooking methods Integrated quantitative HAEA was done using a method and meat-doneness preferences. Positive studies have combining HA-concentration information from laboratory- tended to be those in which interactions among meat types, cooking studies all of which involved: (1) cooking methods intake frequencies, cooking methods and/or cooking consistently found to generate HAs in meat (oven broiling, doneness preferences were reflected in the categories used pan-frying and grilling/barbecuing), (2) common HA- to classify relative HA-exposure levels (Probst-Hensch et forming meat/fish types (beef hamburger, other beef, al., 1997; Ward et al., 1997; Sinha et al., 1998b, 1999; chicken, nonbacon pork/ham, bacon and/or fish) and (3) Zheng et al., 1998). measures of at least two of five predominant HAs Potential cancer risks associated with human HA considered (namely, PhIP, MeIQx, DiMeIQx, IQ and exposures are hypothesized to reflect the level of chronic A C) (Keating and Bogen, 2001). This method applies dietary intakes in exposed populations. Dietary HA regression equations fitted to these data to estimate total HA exposures from cooked-food sources typically include concentrations in each meat-type (Meat) and cooking- exposures dominated by five compounds: 2-amino-9H- method (Method) combination, at values of maximum pyrido[2,3-b]indole (A C), 2-amino-3-methylimida- internal temperature (IT) equal to 71.6, 76.6, 82.2 and zo[4,5-f]quinoline (IQ), 2-amino-3,4,8-trimethylimida- 87.78C. The first three IT values reflect medium (M), well zo[4,5-f]quinoxaline (DiMeIQx), MeIQx and PhIP (W) and very well (VW) levels of meat doneness according (Hatch et al., 1986; Sugimura et al., 1986, 1988; Felton to USDA guidelines (USDA, 1997), while 87.78C and Knize, 1990a,b; Layton et al., 1995). Meat type, represents a charred/blackened/``extra-well'' (XW) done- cooking duration and cooking temperature are among ness level. Conversion of estimated total HA concentrations factors known to influence the extent to which these (and (C total ) to HA-specific ones (C i, i=1,...,5) for the five other) HAs form in cooked meats (Knize et al., 1994; Skog HAs listed above (respectively) was done using Meat- / et al., 1995; Sinha et al., 1998a,c; Keating et al., 1999). Method-specific models C i =f iC total independent of IT, Estimates of population-average dietary HA intake vary involving the HA-specific fractions f i obtained from considerably, ranging from 1 to >20 ng/kg/day (Layton additional regression equations fitted to HA-specific data, et al., 1995; Thomson et al., 1996; Augustsson et al., 1997; as summarized in Table 1 (Keating and Bogen, 2001). Byrne et al., 1998). Reasons for this variability include different assignment of HA concentrations to cooked meats U.S. Food-Intake Data consumed by the public, differences in estimated meat U.S. dietary consumption of cooked meats and fish was consumption, and differences in assumed cooking practices estimated from 1989 to 1991 CSFII data on food intakes (Keating et al., 1999). over three consecutive days by 11,912 people in 48 An integrated approach to HA-exposure assessment conterminous states, and from 1994 to 1996 CSFII data (HAEA) was developed recently for estimating dietary HA on food, energy and nutrient intakes on two nonconsecutive exposures in a way that reflects individual meat-specific days by 15,303 people in 50 states, corresponding to a total consumption preferences, intake rates, cooking methods of 27,215 people (USDA, 1993, 1998). Both surveys used and doneness preferences, as well as a large database of national probability samples of households stratified to measured HA concentrations from laboratory meat-cook- account for geographic location, degree of urbanization and ing studies (Keating and Bogen, 2001). When applied to socioeconomic factors, with age- and sex-specific data meat-intake and cooking-method information extracted obtained by telephone from all strata (USDA, 1993, 1998). from 6 years of Continuing Survey of Food Intakes by Summary CSFII data on individual sex, age (A, in years), Individuals (CSFII) data on >25,000 individuals (USDA, body weight (BW, in kilograms), height (H, in meters),

1993, 1998), application of this new approach yielded an total food-energy intake (E in, in kilocalories), and related estimated mean U.S. dietary HA intake of about 10 ng/kg/ summary data available jointly for only 25,895 people were day, which is somewhat lower than previous estimates classified into sex-specific Infant (0 years), Child (0±15 (Layton et al., 1995). However, that preliminary analysis years), Young adult (16±29 years), and Older adult (30+ did not address substantial food-energy intake under- years) age bins, and then were used to estimate correspond- reporting that generally occurs in dietary surveys (Briefel ing missing BWs. Missing Child weights were estimated as et al., 1997; Layton, 1993; NRC, 1986), nor did it examine BW=10.86 kg+1.110A (kg/year)+0.1605A 2 (kg/year2), HA exposures by age/sex/race-ethnicity to identify based on the corresponding least-squares fit obtained for 2 possible high-risk populations. To address these two noninfant children (A>0; R =0.999; F 12,7055 [goodness- issues, integrated HAEA was reapplied to the same CSFII of-fit] =1.88, p=0.032; F 2,7067 [regression] =1.11, p0), data in the present study. and BW for other age/sex groups were estimated using

156 Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) Dietary exposures to heterocyclic amines Bogen andKeating

Table 1. Estimated HA concentrations by meat type, cooking method, and doneness.a

Meat Cooking method Total HA (ng/g meat) at donenessb Fraction of total HA comprisingc M WD VWD XWD PhIP MeIQx DiMeIQx A CIQ Bacon BR 0.867 0.867 43.2 43.2 0.986 0.112 0 0 0 F 0.635 0.293 0.073 0 0 Beef BR 3.99 6.40 10.2 18.8 0.784 0.112 0 0 0 F 0.635 0.293 0.052 0.121 0.013 G 0.716 0.246 0 0.327 0.013 Burger F 0 2.71 7.24 12.7 0.760 0.188 0.052 0 0.013 G 0.716 0.246 0.004 0.327 0 Chicken BR 0 9.04 31.6 58.8 0.986 0 0 0 0 F 0.938 0.013 0.052 0 0 G 0.978 0.018 0.004 0.327 0 Fish BR 0.152 12.3 68.4 202. 0 0.433 0 0 0.536 F 0.815 0.106 0 0.565 0 G 0.432 0 0 0.565 0 Pork F 0.60 1.7 2.8 3.9 0.635 0.293 0.052 0 0.013 BR,G 0.534 0.383 0 0 0 aBeef=nonground beef, Burger=ground beef, BR=broiled, F=fried, G=grilled. Total HA concentrations of 0 ng/g were assumed for the following meats (and cooking methods): bacon (F), burger (BR), pork (BR), and all lamb (see Keating and Bogen, 2001). bM=medium, WD=well-done, VWD=very well-done, XWD=extra well-done. Standard errors for estimated total HA concentrations predicted for the primary source of dietary HAs (namely, from VW and XW beef, hamburger, and chicken) are all <25% of the corresponding estimated predicted values (see Keating and Bogen, 2001). cThe sum of total HA fractions for meats containing IQ and/or A C exceeds 1 because the fractions for these compounds were calculated independently from those for PhIP, MeIQx and DiMeIQx (see Keating and Bogen, 2001). sample-weighted mean values listed in Table 2. CSFII data fried, baked/fried, deep-fried, barbecued, ``cooked Ð not on 47 breast-fed infants for whom E in =0 was reported were specified'', or ``totally not specified''; and (2) a CSFII meat excluded. Using estimated BW values, summary CFSII data type comprising or including bacon, beef, chicken, ham, including BW were thus available for a total of 27,168 pork, lamb, hamburger, fish, or any of 28 specific types of individuals, each coded by a unique identifier (ID). fish, and excluding uncommon meat cuts (e.g., brain, neckbone). Definitions for CSFII ``mixture'' codes typically Meat-Consumption Analysis did not contain these phrases, so for these codes the CSFII By examining definitions and corresponding recipes for all recipe database was searched for the phrases. Accounting 2006 CSFII food codes for meat/fish items (including for 114 codes involving 2 meat types and/or cooking ``mixtures''), 1636 unique food codes (Code) were found to methods, a total of 1750 ``meat-code records'' were used to designate items that comprise or contain ``meat'' capable of identify potentially HA-forming CSFII food items, as forming HAs upon cooking. HA-forming codes were previously summarized (see Table 3 in Keating and Bogen, defined as those reflecting: (1) a CSFII-defined cooking 2001). Each meat-code record included a CSFII food method specified as broiled, broiled/baked, broiled/fried, identifier (Code), meat type (Meat), an indicated or

Table 2. Summary information for 25,895 individuals based on 6 years of CSFII survey data.a

Age bin Sex Number surveyed Age (years) BW (kg) Ratio (unitless) Scale (unitless) Infant (0±12 months) M&F 450b 0 7.6 ± Child (0±15 years) M 3829 7.6 32.9 1.60 1 F 3691 7.7 31.7 1.56 1 Young adult (16±29 years) M 1991 22.6 78.9 1.41 1.26 F 2261 22.5 63.7 1.26 1.38 Older adult (30+ years) M 6622 49.8 84.2 1.46 1.22 F 7501 51.3 68.9 1.24 1.44 aAll individuals surveyed in 1989±1991 and 1994±1996 for whom weight/height and energy-intake data were acquired (excluding CSFII data on 47 breast-fed infants). Sample-weighted mean values are shown for age and BW, and for the variates Ratio and Scale defined in the Methods section. bIncluded in totals listed for Child.

Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) 157 Bogen andKeating Dietary exposures to heterocyclic amines

Table 3. Meat type and cooking-method fractions estimated from CSFII survey data.

Meat type (age) Fraction of all meat consumeda Fraction broiled/grilled Fraction pan-fried Fraction otherb Beefsteak 0.371 0.339 0.224 0.417 Hamburger 0.119 0.629 0.353 0.018 Chicken 0.233 0.124 0.560 0.316 Pork 0.175 0.162 0.350 0.488 Bacon 0.020 0.050c 0.950c 0 Fish ( <16 years) 0.082 0.244 0.517 0.239 Fish (16+ years) 0.082 0.387 0.542 0.071 aFor consumption of those HA-bearing meats considered (which do not include fast food items), measured in grams meat per kilogram BW. bNon±HA-forming cooking methods (including all methods involving lamb). cAssumed; cooking method was not specified in CSFII data.

estimated mass fraction (F HA ) comprising meat potentially BMR. The reference BMR values correspond to the capable of forming HAs, the CSFII-specified cooking following age- and sex-specific regression models for method (MethodC), a corresponding summary HA-related BMR as functions of age (A) and sex (with units method (Method), and a (Source) variate indicating suppressed): BMR=61.29±2.278A for males <16 years, whether the food item was obtained at a sit-down 26.33±0.09337A for males=16 years, 58.74±2.298A for restaurant. MethodC values assumed to form no or females <15.47 years, and 24.99±0.09337 for fe- 2 negligible HA levels (i.e., for which Method=False was males=15.47 years (R =0.998, F 2,6 [unisex vs. sex- assumed) were: baked, braised, roasted, steamed/poached, specific models] =8.25, p=0.019). These models were stewed, steamed/stewed, and uncooked. Method=Broiled used to estimate BMR values corresponding to additional

(or broiled-medium, or broiled-well-done) was assigned values of E x estimated by Layton (1993, in his Table 7) for all MethodC values including the term ``broiled'' for four reference male and female adults of various ages (together with doneness level if specified), Method=Fried based on assumed reference activity patterns. These data was assigned for all MethodC values including the term imply 20 corresponding R x values that have a mean‹1 ``fried'', and Method=NS was assigned to all unspecified standard error (SE) equal to Rx ˆ 1:6 Æ 0:027; the male CSFII cooking methods. It was assumed that Method=False and female reference R x values do not differ significantly for all foods from fast-food sources (Knize et al., 1995; (p=0.26, by Welch's t test) and together do not correlate Keating et al., 1999). with age (R 2=0.0044, p=0.78). For the present study, the

The CSFII data were then searched by Code and ID to ratio R m =(E in /BW)/BMR, representing BW-normalized identify potentially HA-related ``meat-intake records'', for food-energy intake divided by BMR for the (nonreference) each of which a corresponding aggregate food-item mass CSFII survey population, was therefore assumed to have a (M, in grams) and Intake (I) in grams per kilogram per day (metabolically consistent) population-average value of was obtained, defined as I=MÂF HA /(nBW) where HA Rx ˆ 1:6, independent of age, sex and race/ethnicity, as fraction F HA was defined above, and where n=3 and n=2 follows. Implied ID-specific R m values, each reexpressed in for records from the 1989±1991 and 1994±1996 CSFII terms of lean body mass (LBM, in kilograms) as R m =(E in / surveys, respectively. Each meat-intake record listing an LBM)/(BMRÂ[BW/LBM] ), were estimated from corre- unspecified Meat and/or Method was replaced by a set of sponding CSFII data on E, BW and H (defined above) Meat- /Method-specific records that were assigned corre- reported for 25,895 subjects, assuming that {LBM, BW/ sponding default values of relative frequency based on LBM}{0.806BW, 1.24}, {20.4H 2, 1.19} and {17.5 intake-weighted summary CSFII data characteristics listed H 2, 1.34} for children ( <16 years), adult (16+ years) in Table 3. males and adult (16+ years) females, respectively (ICRP, 1975, pp. 43±44). (Age- and sex-specific sample-

Adjustment of Meat Intakes to Account for Underreported weighted mean values of R m listed in Table 4 were used Energy Intake for 932 IDs for which a height (H) value was missing.) Values of basal metabolic rate (BMR, in kilocalories per From the resulting empirical age- and sex-specific R m kilogram BW) at three childhood and three adult ages for distributions, adjustment factors F m were determined reference males and females were obtained from Table 6 numerically to solve Max Rmin;Fm  Rm†ˆ1:6, where of Layton (1993) (based on Schoeller and van Santen, the overbar again denotes a sample-weighted population 1985), together with corresponding reference values of average, and where R min was assumed to be 1, 1.3 and 1.3 energy expended (E x, in kilocalories per kilogram BW) for children ( <16 years), adult (16+ years) males and adult and corresponding reference values of the ratio R x =E x / (16+ years) females, respectively. For comparison, this

158 Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) Dietary exposures to heterocyclic amines Bogen andKeating

Table 4. Comparison of food-energy intake vs. BMR in CSFII subjects.a

Group, Sex Age (years) R min All subjects Nondieting subjects

nRÅ CVR (%) F m nRÅ CVR (%) F m Child 0±15 1.0b 7520 1.58 37 1.00 5556 1.60 36 1.00 Young adults, M 16±29 1.3 1991 1.38 39 1.07 1268 1.39 37 1.06 Young adults, F 16±29 1.3 2261 1.18 37 1.25 1409 1.22 34 1.23 Older adults, M 30+ 1.3 6622 1.31 36 1.13 4908 1.34 35 1.12 Older adults, F 30+ 1.3 7501 1.21 35 1.24 5198 1.26 33 1.20 Total 25,895 ( =100%) 18,339 ( =70.8%) aCSFII subjects restricted to those who participated in all study days: 1989±1991 (3 days), 1994±1996 (2 days). Age-group- /sex-specific food-intake adjustment factors (F m ) were numerically obtained that increase subject-specific food-energy intakes and corresponding values of the ratio R m =(E in / LBM)/ [BMR/(BW/LBM)] such that corresponding sample-weighted population average values of Max(R min, F mÂR m )=1.6, for food-energy intakes (E) and basal metabolic rates (BMR) in kilocalories per kilogram estimated lean body mass (LBM). CVR =(standard deviation of R)/RÅ , where RÅ =arithmetic mean of R m. Note that R, RÅ , and F are unitless. bConstraint applied only for analysis of Child data among ``All subjects''. analysis was repeated using data pertaining only to 18,339 fractions are 0 for bacon, 1/2 for beef/steak, and 2/3 for subjects not reported as being on a restricted-calorie diet. all other meats and fish. These distributions were applied

The extent to which the F m values obtained (listed in Table regardless of race/ethnicity, in view of limited meat- 4) are >1 therefore corresponds to the extent of estimated specific survey doneness-preference data for U.S. sub-

E in underreporting among CSFII subjects. Because esti- populations. A recent national survey involving >15,000 mated F m values for dieters and nondieters were quite U.S. residents indicated an 3-fold lower intake of ``pink'' similar, age- and sex-specific intake adjustment factors F m hamburger by African American vs. white adults, which obtained for all (dieting+nondieting) subjects were applied appears to reflect the fact that income correlates well with to corresponding intakes I by converting them to corre- ingestion of relatively more ``pink'' (less well-done) sponding adjusted intakes I*, where I*=F mÂI. hamburger (Yang et al., 1998). To explore the potential impact of this difference in doneness preference, under the Doneness Preference Distributions assumption that it applies to all HA sources (not just Distributions reflecting interindividual variability in meat- hamburgers), additional analyses were done for CSFII- specific doneness preferences were estimated from available sampled African Americans using corresponding doneness- survey data pertaining to beefsteak, hamburgers and bacon preference distributions modified as follows (where aster- (Muscat and Wynder, 1994; Reagan and Buyck, 1995; Cox isks denote modified values): f M*=f M /3, and et al., 1997; Probst-Hensch et al., 1997; Ward et al., 1997; f L*=f L [(1Àf M*)/(1Àf M )] for L={W,VW,XW}. Zheng et al., 1998). These distributions each had the form

{f M, f W, f VW, f XW}, where relative frequency f L denotes the Data Analysis fraction of people who eat meat cooked to doneness level L, Intakes I (defined above) from meat-intake records such that f M +f W +f VW +f XW=1. The survey-based dis- common to each {ID, Meat, Method, Source} combina- tributions were as follows: {0.62, 0.25, 0.12, 0.01} for tion were summed and then converted to a cumulative hamburger, {0.09, 0.53, 0.30, 0.08} for bacon, and {0.566, probability mass function (cmf) reflecting interindividual 0.317, 0.105, 0.012} for beefsteak. Additional distributions variability in each corresponding HA-specific concentra- were obtained under the assumptions that: (1) 10% of beef/ tion Ci by an applicable doneness-preference distribution steak characterized as VW in surveys is cooked to an XW discussed above. All C i -specific cmfs common to each level of doneness (which was not addressed in the survey ID were then used (analytically) to sum C i from all ID- data); (2) 20% of bacon characterized as VW in surveys is specific {Meat, Method, Source} combinations assuming cooked to an XW level of doneness (also not addressed in that the C i contributions are 100% correlated (i.e., a the survey data); (3) 100% of hamburger and 50% of other single doneness preference was assumed to apply to all meat/fish cooked in restaurants is cooked to doneness levels meat/fish consumed by each individual), thus yielding a distributed as nonrestaurant items but (to reflect current single C i -specific cmf corresponding to each ID. trends in restaurant cooking) truncated below a W- Expected values Ci† of ID-specific C i cmfs were used doneness level; and (4) fish and meats other than beef/ to obtain corresponding ID- and HA-specific intakes (H i ) steak/hamburger have doneness-preference distributions in nanograms per kilogram per day, where Hi ˆ I ÃÂCi. similar to beefsteak. Absent specific data on the grilled For each analysis done by age, sex, and/or race, mean HA fractions of meats cooked by a Method corresponding to the intakes were estimated using sample weights supplied for CSFII ``broiled'' category, it was assumed that these each ID and CSFII survey period (1989±1991 vs. 1994±

Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) 159 Bogen andKeating Dietary exposures to heterocyclic amines

1996), after normalizing each 3-year subset of age/sex/ F ratio regression statistics (F df1,df2) are provided with race-specific sample weights by the corresponding mean subscripts indicating degrees of freedom (df), where sample-weight value (i.e., all sample weights used to obtain df1=numerator df and df2=denominator df, and corre- each reported HA-intake estimate were normalized to have sponding unadjusted values of R 2 =1À [(sum of squared a mean value of 1). residuals)/(sum of squared ordinate values)]. Approxi- Approximately 125,000 potentially HA-related meat- mate significance of differences in mean HA-intake rates intake records were identified, comprising 74,978 unique were compared by t tests, using Welch's t test in case of ID- /Code-specific records 13% of which pertained to unequal variances as assessed by corresponding F tests fast-food items. The remaining 65,292 ID- and Code- (Kendall and Stuart, 1979). Significance p values=10 À 10 specific records pertained to 24,790 ``non±fast-food'' (a are reported as being 0. All calculations were done using sample-weighted 91.6% of all 27,168 qualifying) IDs and Mathematica1 4.0 (Wolfram, 1999) and RiskQ 4.0 to 1384 Codes. About a third of the ``fast-food only'' IDs (Bogen, 2000). represent children aged 0±5 years, but the rest were found to have age/sex distributions similar to ``non±fast-food'' IDs. Among the latter IDs, Method=False pertained to all meat-intake records for 5310 (a sample-weighted 19.3%) of the surveyed population, which most likely reflects the relatively brief 2- or 3-day CSFII sample periods used (i.e., consistent with HA intakes occurring typically on 3 out of every 4 days for most people), but which may also reflect subpopulations that remain virtually unexposed to HAs due to lifelong dietary patterns. U.S. dietary HA intakes were estimated in this study using only data involving the 21,858 HA-positive IDs identified. Uncertainties associated with the integrated HAEA method used were discussed by Keating and Bogen (2001), are noted in Table 1 (footnote b), and are reasonably assumed to pertain systematically to analyses reported below. Interindividual variability in H i attributable to variability in all HA sources addressed in this study except variability in meat consumption per se (see Discussion), conditional on expected values of cumulative HA-specific intakes Ci for each ID, were characterized within each of the three age categories used by the corresponding sample- weighted functional average of ID-specific cmfs that each reflect variability in normalized ID- and HA-specific intakes, defined as I Â Ci=Ci. These averaged cmfs were then re-averaged using corresponding fraction-of-lifetime weights to obtain a single cmf characterizing lifetime HA- intake variability attributable specifically to all HA sources addressed in this study except variability in meat consump- tion per se. Unweighted general linear model regression for BW vs. A, the analysis of variance done for BMR models, and analyses of product±moment correlations between intakes of major HAs (PhIP and MeIQx) and other variates on which data were obtained in the CSFII surveys (by sex and child/nonchild age group), were all done using standard methods (Draper and Smith, 1981; Selvin, 1995). Addi- tional variates considered in correlations analyses include Figure 1. Empirical cumulative relative frequency distributions (solid sex and weight-normalized intakes of total fat, saturated curves) characterizing interindividual variability in the ratio R m of fat, total fiber, carotene, vitamin C, grains, vegetables consumedfood to basal energy intakes per unit LBM, calculated for (excluding potatoes), and fruit. Substantial significant each of 25,895 CSFII participants. Separate plots are shown for each age group and sex; dashed and dotted vertical lines indicate R m values correlations identified are those with r=0.10 and p=10À 6. of 1 and 1.6, respectively.

160 Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) Dietary exposures to heterocyclic amines Bogen andKeating

Table 5. Estimated mean HA intakes by all 21,858 CSFII participants who consumed HAs, by age group.

Agea n Sex Estimated HA intake (ng/kg/day) PhIP MeIQx DiMeQx A CIQ Child ( <16 years) 3,359 M 7.2 1.4 0.28 1.6 0.17 3,173 F 7.2 1.3 0.27 1.6 0.15 Young (16±29 years) 1,691 M 5.8 1.1 0.22 1.2 0.12 1,875 F 5.8 1.0 0.20 1.3 0.14 Older (30+ years) 5,609 M 5.9 1.2 0.20 1.6 0.21 6,151 F 5.7 1.1 0.18 1.8 0.30 Lifetime TWA 10,659 M 6.2 1.2 0.22 1.5 0.18 11,199 F 6.0 1.1 0.20 1.6 0.23 aTWA=time-weighted average, M=males, F=females.

Results 1. Corresponding estimated ID-specific adjustment frac-

tions F m required to obtain corresponding sample-weighted Empirical distributions of the ratio R m of food-energy to population-average R m values all equal to the metabolically basal-energy intakes per unit LBM calculated for each of consistent value of 1.6, conditional on reasonable values of

25,895 CSFII subjects who consumed HAs during CSFII Min(R m ), are listed in Table 4. As indicated in Figure 1 and survey periods are shown by age group and by sex in Figure Table 4, the mean and approximate median R m values for

Table 6. Estimated mean HA intakes by all 21,858a CSFII participants who consumed HAs, by race±ethnicity, age group and sex.

Race/ethnicityb Ageb Sex n Estimated HA intake (ng/kg/day) PhIP MeIQx DiMeQx A CIQ White Child M 2478 6.1 1.2 0.23 1.3 0.16 F 2233 6.5 1.2 0.24 1.4 0.14 White Young M 1359 5.5 1.1 0.20 1.1 0.12 F 1399 5.6 0.96 0.19 1.1 0.13 White Older M 4734 5.6 1.1 0.18 1.5 0.20 F 4955 5.3 1.0 0.17 1.6 0.28 African American Child M 520 12. 1.8 0.51 2.6 0.24 F 608 8.9 1.5 0.33 1.9 0.11 African American Young M 183 8.3 1.3 0.35 1.8 0.093 F 312 7.3 1.2 0.27 2.0 0.12 African American Older M 546 8.1 1.6 0.30 2.3 0.16 F 858 7.4 1.2 0.23 2.4 0.23 African American* Child M 520 20. 2.6 0.80 4.7 0.43 F 608 14. 2.0 0.51 3.2 0.19 African American* Young M 183 13. 1.7 0.52 2.7 0.16 F 312 12. 1.7 0.41 3.2 0.18 African American* Older M 546 12. 2.2 0.44 3.8 0.25 F 858 12. 1.8 0.35 4.2 0.42 Asian/Pacific Islander Child M 79 10. 1.9 0.27 3.4 0.29 F 70 9.2 1.9 0.22 3.8 0.90 Asian/Pacific Islander Young M 43 5.8 1.0 0.22 1.1 0.33 F 39 5.9 1.1 0.16 1.9 0.32 Asian/Pacific Islander Older M 98 8.2 1.7 0.20 3.2 0.69 F 91 9.6 1.9 0.21 4.4 1.2 aThis summary excludes data on 243 Indian/Eskimo participants, among whom those in the Child and Older age bins (see footnote b) were estimated to have HA intakes similar to those of the African American/Female and about 15% less than those of the White/Older groups indicated above, respectively. bAsterisks indicate analyses done assuming a threefold lesser-than-average preference for items cooked to a medium or lesser doneness level (e.g., associated with redness or pinkness in cooked hamburger), based on national survey data. Child= <16 years, Young (adult)=16±29 years, Older (adult)=30+ years.

Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) 161 Bogen andKeating Dietary exposures to heterocyclic amines

children ( =15 years) are 1.6 without adjustment, and R m PhIP) by African Americans and by Asian/Pacific Islanders values <1 were estimated for <20% of these subjects. In than by whites, in analyses done without considering any contrast, R m values for adult males and females were <1 for racial±ethnic differences involving doneness preference. 75±80% of participants, with corresponding mean values This difference between mean white vs. African American clearly less than 1.6, indicating substantial E in under- HA consumption was more pronounced in those analyses reporting among CSFII adults, particularly among younger (indicated by an asterisk in Table 6) that reflected survey (16- to 29-year-old, i.e., ``courtship''-age) females. evidence of an African American preference for relatively

Corresponding R m distributions for 18,339 nondieting more well-done meat. After considering this doneness CSFII participants were found to be similar to those (for preference, mean HA intakes for African Americans were all participants regardless of dieting status) shown in Figure estimated to be roughly 1.5- to 2-fold greater than those of

1, as indicated by similar F m estimates obtained for all vs. whites or Asian/Pacific Islanders, and PhIP consumption in nondieting participants listed in Table 4. particular was estimated to be >3-fold greater among Estimated HA-specific intakes by age, sex, racial±ethnic African American vs. white male children (20 vs. 6.1 ng/ group, meat/fish source, and cooking method for CSFII kg/day, p0). Overall, the relative differences between participants studied are summarized in Tables 5±8. The mean female vs. male HA-specific intakes are fairly small estimated intakes reported in Table 5 show that on average for each HA within each age and racial±ethnic group, with over a lifetime, PhIP is clearly the major HA consumed (at the exception that estimated intakes by male African 6 ng/kg/day) among all five HAs examined. These American children are 30±50% greater than those by estimates indicate that PhIP represents 66% of the total females (40% greater in the case of PhIP, p<10À 5 ). input of HAs considered, while MeIQx and A C each The mean lifetime time-weighted average HA intakes by comprise 15% and DiMeIQx and IQ each 2% of this CSFII participants summarized by cooking method, meat total input. type, and sex in Table 7 indicate that, whereas about twice as Estimated mean HA intakes reported in Table 6 by race- much consumed PhIP is derived from pan-frying than from ethnicity, age group and sex indicate about 1.5- to 2-fold broiling/grilling, similar amounts of MeIQx and of A C are greater estimated consumption of most HAs (including derived from these two different cooking methods, nearly all

Table 7. Estimated mean values of lifetime time-weighted average HA intake by all 21,858 CSFII participants who consumed HAs, by cooking method, meat type, and sex.

Cooking methoda Estimate type/unit Sexa Estimated fraction or valueb PhIP MeIQx DiMeQx A CIQ Broiled/grilledc fraction M 0.27 0.28 0.0078 0.38 0.83 F 0.28 0.32 0.0089 0.41 0.85 Pan-fried fraction M 0.51 0.37 0.63 0.5 0.081 F 0.50 0.34 0.63 0.47 0.057 NS fraction M 0.22 0.35 0.36 0.12 0.088 F 0.22 0.34 0.36 0.12 0.095 Bacon fraction M 0.072 0.16 0.22 0. 0. F 0.068 0.16 0.22 0. 0. Beefsteak fraction M 0.31 0.50 0.27 0.24 0.14 F 0.27 0.44 0.23 0.19 0.098 Hamburger fraction M 0.033 0.049 0.038 0.027 0.012 F 0.031 0.049 0.039 0.024 0.0094 Chicken fraction M 0.36 0.025 0.44 0.071 0. F 0.38 0.027 0.48 0.073 0. Fish fraction M 0.20 0.21 0. 0.66 0.84 F 0.23 0.27 0. 0.72 0.88 Pork fraction M 0.020 0.053 0.038 0. 0.012 F 0.018 0.051 0.036 0. 0.0095 Total ng/kg per day M 6.2 1.2 0.22 1.5 0.18 F 6.0 1.1 0.20 1.6 0.23 aNS=not specified; M=males (n=10,659), F=females (n=11,199). bListed fractions sum vertically across cooking methods and across meat types to 1.00 (except for rounding error) within each HA/sex combination. cNote that 1/2 and 2/3 of all ``broiled'' beef and hamburger, respectively, were assumed to be grilled.

162 Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) Dietary exposures to heterocyclic amines Bogen andKeating

Table 8. Estimated mean HA intakes by white vs. African American male CSFII children who consumed HAs, by cooking method and meat type.

Cooking methoda Estimate type/unit Race±ethnicitya Estimated fraction or valueb PhIP MeIQx DiMeQx A CIQ Broiled/grilledc fraction W 0.25 0.26 0.0074 0.38 0.79 AA* 0.077 0.16 0.00083 0.17 0.77 Pan-fried fraction W 0.45 0.31 0.57 0.42 0.082 AA* 0.76 0.47 0.75 0.71 0.061 NS fraction W 0.3 0.43 0.43 0.19 0.12 AA* 0.17 0.36 0.25 0.13 0.17 Bacon fraction W 0.082 0.18 0.23 0. 0. AA* 0.068 0.22 0.18 0. 0. Beefsteak fraction W 0.34 0.52 0.28 0.28 0.15 AA* 0.11 0.24 0.09 0.065 0.054 Hamburger fraction W 0.047 0.068 0.049 0.047 0.016 AA* 0.028 0.054 0.043 0.0055 0.02 Chicken fraction W 0.36 0.024 0.42 0.075 0. AA* 0.49 0.051 0.66 0.012 0. Fish fraction W 0.15 0.16 0. 0.6 0.82 AA* 0.29 0.37 0. 0.92 0.91 Pork fraction W 0.018 0.049 0.028 0. 0.01 AA* 0.017 0.063 0.032 0. 0.015 Total ng/kg per day W 6.1 1.2 0.23 1.3 0.16 AA* 20. 2.6 0.8 4.7 0.43 aNS=not specified; W=white male children (n=2,478), AA*=African American male children (n=520). Asterisk indicates analyses that reflect an assumed threefold lesser-than-average preference for items cooked to a medium or lesser doneness level (e.g., associated with redness or pinkness in cooked hamburger), based on national survey data. bListed fractions sum vertically across cooking methods and across meat types to 1.00 (except for rounding error) within each HA/sex combination. cNote that 1/2 and 2/3 of all ``broiled'' beef and hamburger, respectively, were assumed to be grilled.

DiMeIQx intake is due to pan-frying, and nearly all IQ differ significantly (2.2 vs. 2.1 ng/kg/day, respectively, intake is due to broiling/grilling. The sources (and p=0.41), African American male children evidently con- corresponding approximate relative contributions) of total sume a 5-fold greater amount PhIP from chicken and fish PhIP intake were estimated to be: chicken (35±40%), than do white male children (16 vs. 3.1 ng/kg/day, p0). beefsteak (30%), fish (20%), bacon (7%), hamburger Cumulative relative frequency distributions were ob- (3%), and pork/ham (2%). For MeIQx, these approximate tained (Figure 2) characterizing interindividual variability relative contributions are: 40±50% (beefsteak), 20±30% in lifetime time-weighted average values of individual- (fish), 15% (bacon), 5% (hamburger, pork), 3% (chicken). mean-normalized HA-specific intakes. These variability Most (70±90%) of the estimated A C and IQ intakes are distributions were based on estimates made for all 21,895 attributable to fish consumption. CSFII participants who consumed HAs during CSFII survey Factors contributing to the up to >3-fold difference in periods. These distributions each have an expected value of mean HA intakes by African American vs. white male 1 and reflect different preferences for meat/fish, cooking children revealed in Table 6 are revealed by corresponding methods, and doneness. Corresponding standard deviation estimated intakes summarized by cooking method and by () estimates obtained are 1.2, 1.1, 1.3, 1.3 and 0.95 for meat type in Table 8. This analysis indicates roughly 30± PhIP, MeIQx, DiMeIQx, A C and IQ, respectively. For this 70% greater relative contributions of each HA except IQ via a analysis, doneness-preference variability was modeled at a pan-frying cooking method in African American vs. white national level (see Methods), without reference to racial± male children (e.g., 76% vs. 45% in the case of PhIP, p0). ethnic differences in doneness preference like those White male children obtained about 3-fold more of their addressed in Tables 6 and 8. A similar analysis done using relative PhIP intakes from beefsteak as did African American CSFII data on 1128 African American children resulting in male children (p0), who in turn obtained about 1.4- and 2- similar normalized variability distributions with somewhat fold more of their relative PhIP intakes from chicken lower  estimates: 0.89, 0.88, 1.0, 0.94 and 0.77 for PhIP, (p<10À 5) and fish (p0), respectively, as did the white MeIQx, DiMeIQx, A C and IQ, respectively. male children. While the corresponding absolute beefsteak Correlation between PhIP and MeIQx intakes by race- intakes by African American vs. white male children do not ethnicity and by age group were strongly correlated

Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) 163 Bogen andKeating Dietary exposures to heterocyclic amines

group, revealed relatively few significant correlations that are summarized in Table 9, all of which are fairly small (r<0.25). Dietary total fat and saturated fat were most consistently correlated with PhIP and MeIQx intakes for white and African American children and adults. Among whites, these correlations involving total fat (r=0.08±0.13) are slightly larger than those involving saturated fat (r=0.05±0.11). This pattern also appears for African Americans after accounting for a greater preference for well-done items (r=0.12±0.19 vs. 0.09±0.12 for total vs. saturated fat, respectively). For both of these CSFII groups and for both fat types, these correlations are slightly larger for MeIQx than for PhIP, and tend to be larger for African Americans than for whites. Estimated correlations between Figure 2. Cumulative relative frequency distributions (solid curves) total fat and PhIP intake are largest for African American characterizing interindividual variability in lifetime time-weighted adults and for Asian/Pacific Islander children (r0.2). average values of individual-mean-normalized intakes of five HAs, based on HA-specific intake estimates made for all 21,895CSFII participants who consumed HAs during CSFII survey periods. These variability distributions each have an expected value of 1, and they reflect different preferences for meat/fish, cooking methods, and Discussion doneness. For this analysis, doneness-preference variability was modeled at a national level (see Methods section), without reference The estimated daily total HA intake for adults, and as a to racial±ethnic differences in doneness preference like those addressed in Tables 6 and 8. lifetime time-weighted average, implied by results summar- ized in Table 5 (9 ng/kg/day) fall within the range of those (r=0.65±0.75, p0), with MeIQx intakes being approxi- reported by others. Layton et al. (1995) estimated an average mately uniformly distributed between 0% and 25% of adult daily HA intake for the U.S. population to be 26 ng/ total PhIP+MeIQx intakes (data not shown). In contrast, kg/day, whereas Augustsson et al. (1997) estimated a analyses of correlations between PhIP or MeIQx intakes and corresponding value of 2.3 ng/kg/day for a Swedish sex and other dietary intakes, by race-ethnicity and by age population, assuming a 70-kg BW in both cases. Compar-

Table 9. Correlations between major estimated HA intakes by 21,858a HA-consuming CSFII participants and other dietary intakes and sex, by race- ethnicity and age group.

Race/ethnicitya Agea n HAa Values of 100r between PhIP/MeIQx andb Sex TotFat SatFat Grains VegXP White Child 4,177 PhIP ± 7.8 5.6 À8.2 ± MeIQx ± 12 11 À7.8 ± White Adult 12,447 PhIP ± 10 5.3 ± 4.7 MeIQx À4.1 13 9.9 ± 5.5 African American Child 1,128 PhIP ± 13 ± ± ± MeIQx ± 13 12 ± ± African American Adult 1,899 PhIP ± 20 21 ± ± MeIQx - 11 15 ± ± African American* Child 1,128 PhIP ± 12 ± ± ± MeIQx ± 14 12 ± ± African American* Adult 1,899 PhIP ± 18 8.8 ± ± MeIQx ± 19 12 ± ± Asian/Pacific Islander Child 149 PhIP ± 22 24 ± ± MeIQx ± ± ± ± ± Asian/Pacific Islander Older 271 PhIP ± ± ± ± ± MeIQx ± ± ± ± 18 aThis summary excludes data on 243 Indian/Eskimo participants (see Table 6). Asterisks indicate analyses done assuming a 3-fold lesser-than-average preference for items cooked to a medium or lesser doneness level (e.g., associated with redness or pinkness in cooked hamburger), based on national survey data. Child= <16 years, Adult=16+ years. bFor estimated values of product±moment correlation (r), listed as ``±'' if the corresponding two-tailed p value is >10 À 2 for Asian/Pacific Islanders or is >10À 3 otherwise. Sex=1 (male) or 2 (female), Totfat=total fat, SatFat=saturated fat, VegXP=total vegetables excluding potatoes.

164 Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) Dietary exposures to heterocyclic amines Bogen andKeating ison of our estimate with that of Layton et al. (1995) is identification of fat intake per se as a contributing risk factor particularly relevant as that study also used selected in previous case±control studies could very well reflect laboratory-derived HA values and the 1989±1991 CSFII effects due largely or entirely to PhIP rather than fat survey to estimate HA intake. By estimating HA concentra- exposure. This hypothesis is supported by the relatively tions at conditions directly related to actual cooking greater estimated PhIP exposures among African American practices and incorporating meat doneness into dietary vs. white men indicated in Table 6, in view of greater intake, our estimate of daily HA intake for the adult U.S. increased PC risks observed to be associated with dietary population is 2.5-fold less than that of the Layton et al. intake of animal/saturated fats by African American vs. (1995). More recent data (Keating et al., 2000) indicate that white men (Whittemore et al., 1995; Hayes et al., 1999). HA concentrations in home-cooked pan-fried meats may be HA-exposure differences by themselves, however, are considerably less than those previously estimated based on unlikely to explain a large fraction of observed racial±ethnic laboratory cooking studies (Keating and Bogen, 2001), so differences in PC occurrence. Indeed, the Los Angeles urine the HA intake rates we report here are likely to be reduced as analysis study referred to above also reported similar PhIP more data on HAs in cooked meats become available. levels in urine sampled from African vs. Asian Americans The greatest racial±ethnic difference in estimated HA (Kidd et al., 1999), whereas Japanese and Chinese intake identified in this study is the roughly 50±100% Americans have among the lowest PC rates in the U.S., greater intake of PhIP by African American males compared while African Americans have the highest (Miller et al., to white males Ð a difference that reflects a >3-fold 1996). Asian/Pacific Islander males were also estimated in greater intake of PhIP by African American compared to the present study to have PhIP exposures substantially larger white boys (0±15 years) when a greater preference for than white males, though not quite as large as those more well-done items among African Americans is predicted for African Americans after dietary doneness considered, based on available national survey data. These preferences were considered (Table 6). Other factors are results are consistent with survey data indicating that therefore likely to explain part or even most of the African Americans tend to eat somewhat more HA sources substantial racial±ethnic PC-rate differences that continue (beef, pork, chicken, fish) than whites (USDA, 1998), and to occur. Rates of PC incidence rise very sharply as a tend to eat more well-done beef than whites (Yang et al., function of age compared to other epithelial tumors, 1998). The results are also consistent with higher levels of suggesting that comparatively more intermediate stages of PhIP/PhIP metabolite(s) in urine from African American clonal progression toward malignancy may be operative in than white adults detected in a study done recently in Los the prostate (Armitage and Doll, 1957). Detailed histologi- Angeles (Kidd et al., 1999). cal evidence of multifocality and genetic heterogeneity Our finding that, on a national basis, African American within excised prostate tumors supports this hypothesis males, and boys in particular, are exposed to substantially (Ruijter et al., 1996; Macintosh et al., 1998; Djavan et al., more PhIP than white males is particularly interesting not 1999). Testosterone- and insulin-related hormones are only in view of PhIP's capacity to damage and mutate known to regulate prostate cell proliferation, and racial± prostate DNA and cause prostate tumors in rats (see ethnic differences in genetic markers that modulate the Introduction), but also in view of studies in which as few as levels and/or effectiveness of these hormones may help 8±10 doses of PhIP during the peak period of end-bud cell explain observed racial±ethnic differences in PC occurrence proliferation in young/adolescent mammary epithelium (Ross et al., 1998; Pettaway, 1999; Signorello et al., 1999). (which like the prostate is a hormonally regulated tissue) Nevertheless, exposure to environmental carcinogenic are sufficient to induce subsequent mammary cancer in mutagens such as dietary PhIP may contribute to PC risk female rats (Ghoshal et al., 1994; El-Bayoumy et al., by acting independently from or perhaps even synergisti- 1995). In humans, race-ethnicity is the best predictor of PC cally with underlying genetic susceptibilities to hormonal occurrence besides age, and African Americans have the PC promotion. Such a PhIP effect would be consistent with highest age-specific PC rates worldwide-incidence rates evidence that environmental (e.g., dietary) factors explain a >30% to 70% higher and mortality rates >2-fold higher substantial fraction of human cancers, including PC (Doll than those among U.S. white men (Miller et al., 1996; and Peto, 1981; Lichtenstein et al., 2000), and would likely Robbins et al., 1998; Hsing et al., 2000). Dietary intake of be greatest (at least in absolute terms) among African saturated/animal fat has been the environmental factor most Americans compared to other racial±ethnic groups. consistently linked to significantly increased PC risk, albeit Reevaluation of cooking-method-specific predictions of one that at most explains only a small fraction of observed individual and total HA concentrations is warranted as more racial±ethnic differences in risk for this disease (Whitte- data become available. The HAEA method used in this more et al., 1995; Hayes et al., 1999; Kolonel et al., 1999). study did not assess whether its underlying cooking-study However, because PhIP intake is significantly positively database reflects cooking conditions typically used by the correlated with total and saturated fat intakes (Table 9), the public (Keating and Bogen, 2001). Our analyses may

Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) 165 Bogen andKeating Dietary exposures to heterocyclic amines therefore be skewed by higher-than-normal HA concentra- Augustsson K., Skog K., Jagerstad M., Dickman P.W., and Steineck G. tions obtained in laboratory cooking studies. In particular, Dietary heterocyclic amines and cancer of the colon, rectum, bladder, and kidney: a population based study. Lancet 1999: 353: 703±707. there were limited data for A C and IQ concentrations and Augustsson K., Skog K., Jagerstad M., and Steineck G. Assessment of the for total HA concentrations in fish, which may have induced human exposure to heterocyclic amines. Carcinogenesis 1997: 18: a bias due to extremely high values from laboratory studies 1931±1935. conducted simply to detect these compounds in cooked Bogen K.T. Cancer potencies of heterocyclic amines found in cooked other meats, rather than to evaluate their formation under foods. FoodChem Toxicol 1994: 32: 505±515. Bogen K.T. RiskQ: an Interactive Approach to Probability, Uncertainty, and normal cooking conditions. The HAEA method used Statistics for Use with Mathematica1 4.0., UCRL-MA-110232 Rev. 1. estimates neither HA levels in pan residues, which can be Lawrence Livermore National Laboratory, Livermore, CA, 2000. considerable (Skog et al., 1995; Sinha et al., 1998c), nor Briefel R.R., Sempos C.T., McDowell M.A., Chien S.C.-Y., and Alaimo consumption of these residues in the diet. Finally, our HA K. Dietary methods research in the third National Health and Nutrition intake estimates were based on a 2- or 3-consecutive-day Examination Survey: underreporting of energy intake. Am J Clin Nutr CSFII survey data which may not be representative of the 1997: 65 (suppl): 1203S±1209S. Byrne C., Sinha R., Platz E.A., Giovannucci E., Colditz G.A., Hunter D.J., habitual consumption patterns. The CSFII data represent a Speizer F.E., and Willett W.C. Predictors of dietary heterocyclic amine random ``snapshot'' of U.S. dietary intakes from which intake in three prospective cohorts. Cancer Epidemiol Biomark Prev average rates of food-specific intake may be estimated, but 1998: 7: 523±529. not corresponding interindividual variabilities. The present Cox R.J., Thomson J.M., Cunial C.M., Winter S., and Gordon A.J. The effect of degree of doneness of beef steaks on consumer acceptability study thus characterized interindividual variability in HA of meals in restaurants. Meat Sci 1997: 45: 75±85. intakes due only to meat/fish, cooking-method, and (for Davis C.D., Schut H.A.J., Adamson R.H., Thorgeirsson U.P., Thorgeirsson African Americans) doneness preferences, and not to S.S., and Snyderwine E.G. Mutagenic activation of IQ, PhlP and MelQx variability in meat/fish intake per se. Variability in HA by hepatic microsomes from rat, monkey and man: low mutagenic intakes due to the latter factor would more appropriately be activiation of MelQx in cynomolgous monkeys in vitro reflects low DNA adduct levels in vivo. Carcinogenesis 1993: 14: 61±65. addressed by analyzing data from a study such as the third Djavan B., Susani M., Bursa B., Basharkhah A., Simak R., and Marberger National Health and Nutrition Examination Survey M. Predictability and significance of multifocal prostate cancer in the (NHANES III), in which 3-month recall data on meat radical prostatectomy specimen. Tech Urol 1999: 5: 139±142. intakes were obtained (DHHS, 1998). Doll R., and Peto R. The causes of cancer: quantitative estimates of In conclusion, we used methods for estimating HA avoidable risks of cancer in the United States. J Natl Cancer Inst 1981: concentrations in cooked meats based on the internal 66: 1191±1308. Draper N.R., and Smith H. Applied Regression Analysis, 2nd ed. Wiley, temperature to which the meats are cooked to estimate New York, NY, 1981. daily HA intake based on a national survey of food El-Bayoumy K., Chae Y.-H., Upadhyaya P., Riveson A., Kurtzke C., consumption and internal temperatures selected as repre- Reddy B., and Hecht S.S. Comparative tumorigenicity of benzo[a]- sentative of meat doneness levels consumed by the public. pyrene, 1-nitropyrene, and 2-amino-1-methyl-6-phenylimida- zo [ 4,5- b ]pyridine administered by gavage to female CD rats. We believe these estimates can be improved using similar Carcinogenesis 1995: 16: 431±434. methods as more data become available on HA concentra- Felton J.S., and Knize M.G. Heterocyclic-amine mutagens/carcinogens in tions in cooked meats and dietary information pertaining to foods. In: Cooper C.S., and Grover P.L. (eds.), Handbook of Experimen- cooking practices and preferences. tal Pharmacology, 94/1. Springer-Verlag, Berlin, 1990a, pp. 471±502. Felton J.S., and Knize M.G. New mutagens from cooked food. In: Pariza M., Aeschbacher H.-U., Felton J.S., and Sato S. (eds.), Mutagens and Carcinogens in the Diet. Wiley-Liss, New York, 1990b, pp. 19±38. Felton J.S., Knize M.G., Shen N.H., Andresen B.D., Bjeldanes L.F., and Acknowledgments Hatch F.T. Identification of the mutagens in cooked beef. Environ Health Perspect 1986: 67: 17±24. We thank our LLNL colleagues Jim Felton and Mark Knize Felton J.S., Knize M.G., Wood C., Wuebbles B.J., Healy S.K., Stuermer for their comments on a draft of this manuscript. This work D.H., Bjeldanes L.F., Kimble B.J., and Hatch F.T. Isolation and was performed under the auspices of the U.S. Department of characterization of new mutagens from fried ground beef. Carcinogen- esis 1984: 5: 95±102. Energy by University of California Lawrence Livermore Gerhardsson de Verdier M., Hagman U., Peters R.K., Steineck G., and National Laboratory under contract No. W-7405-Eng-48, OÈ vervik E. Meat, cooking methods and colorectal cancer: a case- with funding provided by the National Cancer Institute referent study in Stockholm. Int J Cancer 1991: 40: 1±6. (National Institutes of Health Grant No. P01 CA55861-01). Ghoshal A., Preisegger K.-H., Takayama T., Thorgeirsson S., and Snyderwine E.G. Induction of mammary tumors in female Sprague± Dawley rats by the food-derived carcinogen 2-amino-1-methyl-6- phenylimidazo[4,5,-b]pyridine and effect of dietary fat. Carcinogen- References esis 1994: 15: 2429±2433. Hatch F.T., MacGregor J.T., and Zeiger E. Review: putative mutagens and Armitage P., and Doll R. A two-stage theory of carcinogenesis in relation to carcinogens in foods: VII. Genetic toxicology of the diet. Environ the age distribution of human cancer. Br J Cancer 1957: 11: 161±169. Mutagen 1986: 8: 467±484.

166 Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) Dietary exposures to heterocyclic amines Bogen andKeating

Hayes R.B., Ziegler R.G., Gridley G., Swanson C., Greenberg R.S., Lang N.P., Butler M.A., Massengill J., Lawson M., Stotts R.C., Hauer- Swanson G.M., Schoenberg J.B., Silverman D.T., Brown L.M., Pottern Hensen M., and Kadlubar F.F. Rapid metabolic phenotypes for L.M., Liff J., Schwartz A.G., Fraumeni Jr. J.F., and Hoover R.N. acetyltransferase and cytochrome P4501A2 and putative exposure to Dietary factors and risks for prostate cancer among blacks and whites food-borne heterocyclic amines increase the risk for colorectal cancer in the United States. Cancer Epidemiol Biomark Prev 1999: 8: 25±34. or polyps. Cancer Epidemiol Biomark Prev 1994: 8: 675±682. Hsing A.W., Tsao L., and Devesa S.S. International trends and patterns of Layton D.W. Metabolically consistent breathing rates for use in dose prostate cancer incidence and mortality. Int J Cancer 2000: 85: 60±67. assessments. Health Phys 1993: 64: 23±36. International Commission on Radiation Protection (ICRP). Reference Layton D.W., Bogen K.T., Knize M.G., Hatch F.T., Johnson V.M., and Felton Manual ICRP Report No. 25. Permagon, Cambridge, UK, 1975. J. Cancer risk assessment of heterocyclic amines in cooked foods: an Ito N., Hasegawa R., Imaida K., Tamano S., Hagiwara A., Hirose M., and analysis and implications for research. Carcinogenesis 1995: 16: 39±52. Shirai T. Carcinogenicity of 2 - amino - 1- methyl - 6- phenylimida- Lichtenstein P., Holm N.V., Verkasalo P.K., Iliadou A., Kaprio J., zo[4,5-b]pyridine (PhIP) in the rat. Mutat Res 1997: 376: 107±14. Koskenvuo M., Pukkala E., Skytthe A., and Hemminki K. Environ- Ito N., Hasegawa R., Sano M., Tamano S., Esumi H., Takayama S., and mental and heritable factors in the causation of cancer. N Engl J Med Sugimura T. A new colon and mammary carcinogen in cooked food, 2000: 343: 78±85. 2-amino-1-methyl-6-phenylimidazo- [4,5-b]pyridine (PhlP). Car- Lyon J.L., and Mahoney A.W. Fried foods and the risk of colon cancer. Am cinogenesis 1991: 12: 1503±1506. J Epidemiol. 1988: 128: 1000±1006. Kaderlik K.R., Minchin R.F., Mulder G.J., Ilett K.F., Daugaard-Jenson Macintosh C.A., Stower M., Reid N., and Maitland N.J. Precise M., Teitel C.H., and Kadlubar F.F. Metabolic activation pathway for microdissection of human prostate cancers reveals genotypic hetero- the formation of DNA adducts of the carcinogen 2-amino-1-methyl- geneity. Cancer Res 1998: 58: 23±28. 6-phenylimidazo[4,5-b]pyridine (PhIP) in rat extrahepatic tissues. McManus M.E., Burgess W.M., Veronese M.E., Hugget A., Quattrochi Carcinogenesis 1994a: 15: 1703±1709. L.C., and Tukey R.H. Metabolism of 2-acetylaminofluorene and Kaderlik K.R., Mulder G.J., Shaddock J.G., Casciano D.A., Teitel C.H., benzo(a)pyrene and activation of food-derived heterocyclic amine and Kadlubar F.F. Effect of glutathione depletion and inhibition of mutagens by cytochromes P450. Cancer Res 1990: 50: 3367±3376. glucuronidation and sulfation on 2-amino-1-methyl-6-phenylimida- Miller B.A., Kolonel L.N., Bernstein L., Young J.L. Jr, Swanson G.M., West zo[4,5-b]pyridine (PhIP) metabolism, PhIP±DNA adduct formation D., Key C.R., Liff J.M., Glover C.S., Alexander G.A., et al. (eds.). and unscheduled DNA synthesis in primary rat hepatocytes. Racial/Ethic Patterns of Cancer in the United States 1988±1992. NIH Carcinogenesis 1994b: 15: 1711±1716. Publ. No. 96-4104, National Cancer Institute, Bethesda, MD, 1996. Kampman E., Slattery M.L., Bigler J., Leppert M., Samowitz W., Caan Muscat J.E., and Wynder E.L. The consumption of well-done red meat and B.J., and Potter J.D. Meat consumption, genetic susceptibility, and the risk of colorectal cancer. Am J Public Health 1994: 84: 856±858. colon cancer risk: a United States multicenter case±control study. National Research Council (NRC). Nutrient Adequacy: Assessment Using Cancer Epidemiol Biomark Prev 1999: 8: 15±24. Food Consumption Surveys. National Academy Press, Washington, Kato R., and Yamazoe Y. Metabolic activation and covalent binding to DC, 1986. nucleic acids of carcinogenic heterocyclic amines from cooked food and Norrish A.E., Ferguson L.R., Knize M.G., Felton J.S., Sharpe S.J., and amino acid pyrolysates. Jpn J Cancer Res (Gann) 1987: 81: 10±14. Jackson R.T. Heterocyclic amine content of cooked meat and risk of Kato T., Kikugawa K., and Hayatsu H. Occurence of the mutagens 2- prostate cancer. J Natl Cancer Inst 1999: 91: 2038±2044. amino - 3,8 - dimethylimidazo [ 4,5- f ]quinoxaline (MeIQx) and 2- Ochiai M., Ogawa K., Wakabayashi K., Sugimura T., Nagase S., Esumi amino-3,4,8-trimethylimidazo[4,5,-f]quinoxaline( 4,8-Me2IQx ) in H., and Nagao M. Induction of intestinal adenocarcinomas by 2- some Japanese smoked, dried fish products. J Agric FoodChem amino-1-methyl-6-phenylimidazo[4,5-b]pyridine in Nagase ana- 1986: 34: 810±814. lbuminemic rats. Jpn J Cancer Res (Gann) 1991: 82: 363±366. Keating G.A., and Bogen K.T. Methods to estimate heterocyclic amine Ohgaki H., Hasegawa H., Kato T., Suenaga M., Ubakata M., Sato S., concentrations in cooked meats in the U.S. diet. FoodChem Toxicol Takayama S., and Sugimura T. Carginogenicity in mice and rats of 2001: 39: 29±43. heterocyclic amines in cooked foods. Environ Health Perspect 1986: Keating G.A., Layton D.W., and Felton J.S. Factors determining dietary 67: 129±134. intakes of heterocyclic amines in cooked foods. Mutat Res 1999: 443: Pettaway C.A. Racial differences in the adrogen/androgen receptor 149±156. pathway in prostate cancer. J Natl MedAssoc 1999: 91: 652±660. Keating G.A., Sinha, R., Layton, D., Salmon, C.P., Knize, M.G., Bogen, Probst-Hensch N.M., Sinha R., Longnecker M.P., Witte J.S., Ingles S.A., K.T., Lynch, C.F., and Alavanja, M. Comparison of heterocyclic amine Frankl H.D., Lee E.R., and Haile R.W. Meat preparation and colorectal levels in home-cooked meats with exposure indicators. Cancer, adenomas in a large sigmoidoscopy-based case±control study in Causes Control 2000: 11: 731±739. California (United States). Cancer, Causes Control 1997: 8: 175±183. Kendall M., and Stuart A. The Advanced Theory of Statistics, Vol. 2. Reagan J.O., and Buyck M.J. Beef Customer Satisfaction. National Live MacMillan Publishing, New York, NY, 1979, pp. 159±160. Stock and Meat Board, Chicago, IL, 1995. Kidd L.R., Stillwell W.G., Yu M.C., Wishnok J.S., Skipper P.L., Ross Robbins A.S., Whittemore A.S., and Van Den Eden S.K. Race, prostate R.K., Henderson B.E., and Tannenbaum S.R. Urinary excretion of 2- cancer survival, and membership in a large health maintenance amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP in White, organization. J Natl Cancer Inst 1998: 90: 986±990. African-American, and Asian-American men in Los Angeles County. Rosenkranz H.S., and Mermeistein R. The genotoxicity, metabolism and Cancer Epidemiol Biomark Prev 1999: 8: 439±445. carcinogenicity of nitrated polycyclic aromatic hydrocarbons. J Knize M.G., Dolbeare F.A., Carroll K.L., and Felton J.S. Effect of cooking Environ Sci Health 1985: C3: 221±272. time and temperature on the heterocyclic amine content of fried-beef Ross R.K., Pike M.C., Coetzee G.A., Reichhardt J.K.V., Yu M.C., patties. FoodChem Toxicol 1994: 32: 595±603. Feigelson H., Stanczyk F.Z., Kolonel L.N., and Henderson B.E. Knize M.G., Sinha R., Rothman N., Brown E.D., Salmon C.P., Levander Androgen metabolism and prostate cancer: establishing a model of O.A., Cunningham P.L., and J.S. F. Heterocyclic amine content in genetic susceptibility. Cancer Res 1998: 58: 4497±4504. fast-food meat products. FoodChem Toxicol 1995: 33: 545±551. Ruijter E.T., van de Kaa C.A., Schalken J.A., Debruyne F.M., and Ruiter Kolonel L.N., Nomura A.M.Y., and Cooney R.V. Dietary fat and prostate D.J. Histological grade heterogeneity in multifocal prostate cancer. cancer: current status. J Natl Cancer Inst 1999: 91: 414±428. Biological and clinical implications. J Pathol 1996: 180: 295±299.

Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3) 167 Bogen andKeating Dietary exposures to heterocyclic amines

Sato S., Negishi C., Umemoto A., and Sugimura T. Metabolic aspects of demonstration of carcinogen±DNA adducts in tissues of rats given 2- pyrolysis mutagens in food. Environ Health Perspect 1986: 67: 105± amino -1- methyl -6- phenylimidazo[4,5 -b ]pyridine (PhIP): detec- 109. tion in paraffin-embedded sections and tissue distribution. Cancer Schoeller D.A., and van Santen E. Predicting basal metabolic rate, new Res 1998: 58: 4307±4313. standards and review of previous work. Hum Nutr Clin Nutr 1985: 39C Thomson B.M., Lake R.J., Cressey P.J., and Knize M.G. Estimated cancer (suppl. 1): 5±41. risk from heterocyclic amines in cooked meat Ð a New Zealand Schut H.A.J., and Snyderwine E.G. DNA adducts of heterocyclic amine perspective. Proc Nutr Soc 1996: 21: 106±115. food mutagens: implications for mutagenesis and carcinogenesis. Thorgeirsson S.S. Metabolic determinants in the carcinogenicity of Carcinogenesis 1999: 20: 353±368. aromatic amines. In: Greim H., Jung R., Kramer M., Marquardt H., Selvin S. Practical Biostatistical Methods. Duxbury Press, New York, NY, and Oesch F. (eds.) Biochemical Basis of Chemical Carcinogenesis. 1995. Raven Press, New York, 1984, pp. 47±56. Shiffman M.H., and Felton J.S. Re Fried foods and the risk of colon Thorgeirsson S.S., Glowinski I.B., and McManus M.E. Metabolism, cancer [letter]. Am J Epidemiol 1990: 131: 376±378. mutagenicity and carcinogenicity of aromatic amines. In: Hodgson E., Shirai T., Cui L., Takahashi S., Futakuchi M., Asamoto M., Kato K., and Bend J.R., and Philpot R.M. (eds.), Reviews in Biochemical Ito N. Carcinogenicity of 2-amino-1-methyl-6-phenylimidazo[4,5- Toxicology 5. Elsevier Biomedical, New York, Amsterdam, Oxford, b]pyridine (PhIP) in the rat prostate and induction of invasive 1983, pp. 349±386. carcinomas by subsequent treatment with testosterone propionate. Turesky R.J., Lang N.P., Butler M.A., Teitel C.H., and Kadlubar F.F. Cancer Lett 1999: 143: 217±221. Metabolic activation of carcingenic heterocyclic aromatic amines by Shirai T., Sano M., Tamano S., Takahashi S., Hirose M., Futakushi M., human liver and colon. Carcinogenesis 1991: 12: 1839±1845. Hasegawa R., Imaida K., Matsumoto K., Wakabayashi K., Sugimora U.S. Department of Agriculture (USDA). 1989±91 Continuing Survey of T., and Ito N. The prostate: a target for carcinogenicity of 2-amino-1- Food Intakes by Individuals and 1989±91 Diet and Health Knowledge methyl-6-phenylimidazo[4,5-b ]pyridine (PhIP) derived from Survey ( CSFII / DHKS 1989 ± 91 Data Set ). CD - ROM. USDA cooked foods. Cancer Res 1997: 57: 195±198. Agricultural Research Service, Beltsville Human Nutrition Research Signorello L.B., Brismar K., Bergstrom R., Andersson S.-O., Wolk A., Center, Food Surveys Research Group, 1993. Trichopoulos D., and Adami H.-O. Insulin-like growth factor-binding U.S. Department of Agriculture (USDA). Use a Meat Thermometer. Internet protein 1 and prostate cancer. J Natl Cancer Inst 1999: 91: 1965±1966. electronic publication. USDA Food Safety Inspection Service, http:// Sinha R., Chow W.H., Kulldorff M., Denobile J., Butler J., Garcia-Closas www.fsis.usda.gov/OA/pubs/cithermo.htm 1997. M., Weil R., Hoover R.N., and Rothman N. Well-done, grilled red U.S. Department of Agriculture (USDA). 1994±96 Continuing Survey of meat increases the risk of colorectal adenomas. Cancer Res 1999: 59: Food Intakes by Individuals and 1994±96 Diet and Health Knowledge 4320±4324. Survey ( CSFII / DHKS 1994 ± 96 Data Set ). CD - ROM. USDA Sinha R., Knize M.G., Salmon C.P., Brown E.D., Rhodes D., Felton J.S., Agricultural Research Service, Beltsville Human Nutrition Research Levander O.A., and Rothman N. Heterocyclic amine content of pork Center, Food Surveys Research Group, 1998. products cooked by different methods and to varying degrees of U.S. Department of Health and Human Services (DHHS). Third National doneness. FoodChem Toxicol 1998a: 36: 289±297. Health and Nutrition Examination Survey, 1988±1994, NHANES III Sinha R., Kulldorff M., Curtin J., Brown C.C., Alavanja M.C., and Second Laboratory Data File (CD-ROM, Series 11, No. 2A). Centers Swanson C. Fried, well-done red meat and risk of lung cancer in for Disease Control and Prevention, National Center for Health women (United States). Cancer Causes Control 1998b: 9: 621±630. Statistics, Hyattsville, MD, 1998. Sinha R., Rothman N., Brown E.D., Salmon C.P., Knize M.G., Swanson WardM.H.,SinhaR.,HeinmanE.F.,RothmanN.,MarkinR., C.A., Rossi S.C., Mark S.D., Levander O.A., and Felton J.S. High Weisenburger D.D., Correa P., and Zahm S.H. Risk of adenocarcinoma concentrations of the carcinogen 2-amino-1-methyl-6-phenylimida- of the stomach and esophagus with meat cooking method and zo[4,5-b]pyridine occur in chicken but are dependent on the cooking doneness preference. Int J Cancer 1997: 71: 14±19. method. Cancer Res 1995: 55: 4516±4519. Whittemore A.S., Kolonel L.N., Wu A.H., John E.M., Gallager R.P., Sinha R., Rothman N., Salmon C.P., Knize M.G., Brown E.D., Swanson Howe G.R., Burch J.D., Hankin J., Dreon D.M., West D.W., Teh C.A., Rhodes D., Rossi S., Felton J.S., and Levander O.A. C.-Z., and Paffenbarger J., R.S. Prostate cancer in relation to diet, Heterocyclic amine content in beef cooked by different methods and physical activity, and body size in Blacks, Whites, and Asians in to varying degrees of doneness and gravy made from meat drippings. the United States and Canada. J Natl Cancer Inst 1995: 87: 652± FoodChem Toxicol 1998c: 36: 279±287. 661. Skog K., Steineck G., Augustsson K., and Jagerstad M. Effect of cooking Wolfram S. Mathematica Book, 4th ed. Cambridge University Press, temperature on the formation of heterocyclic amines in fried meat Cambridge, UK, 1999. products and pan residues. Carcinogenesis 1995: 16: 861±867. Yang S., Leff M.G., McTague D., Horvath K.A., Jackson-Thompson J., Snyderwine E.G., and Battula N. Selective mutagenic activation by Marayi T., Boeselager G.K., Melnick T.A., Gildemaster M.C., cytochrome P3-450 of carcinogenic arylamines found in foods. J Natl Ridings D.L., Altekruse S.F., and Angulo F.J. Multistate surveillance Cancer Inst 1989: 81: 223±227. for food-handling, preparation, and consumption behaviors asso- Sugimura T., Sato S., Ohgaki H., Takayama S., Nagao M., and ciated with foodborne diseases: 1995 and 1996 BRFSS Food-Safety Wakabayashi K. Mutagens and carcinogens in cooked food. In: Questions. Morb Mortal Wkly Rev 1998: 47 (SS-4): 33±57. Knudsen I. (ed.), Genetic Toxicology of the Diet. Alan R. Liss, New Zheng W., Deitz A.C., Campbell D.R., Wen W.Q., Cerhan J.R., Sellers York, 1986, pp. 85±107. T.A., Folsom A.R., and Hein D.W. N-acetyltransferase 1 genetic Sugimura T., Sato S., and Wakabayashi K. Mutagens/carcinogens in polymorphism, cigarette smoking, well-done meat intake, and pyrolysates of amino acids and proteins in cooked foods: heterocyclic breast cancer risk. Cancer Epidemiol Biomark Prev 1999: 8: 233± aromatic amines. In: Woo Y.T., Lai D.Y., Arcos J.C., and Argus M.F. 239. (eds.), Chemical Induction of Cancer, Structural Bases and Biological Zheng W., Gustafson D.R., Sinha R., Cerhan J.R., Moore D., Hong C.P., Mechanisms. Academic Press, New York, NY, 1988, pp. 681±710. Anderson K., Kushi L.H., Sellers T.A., and Folsom A.R. Well-done Takahashi S., Tamano S., Hirose M., Kimoto N., Ikeda Y., Sakakibara M., meat intake and the risk of breast cancer. J Natl Cancer Inst 1998: 90 Tada M., Kadlubar F.F., Ito N., and Shirai T. Immunohistochemical 1724±1729.

168 Journal of Exposure Analysis and Environmental Epidemiology (2001) 11(3)