Content uploaded by Sabine Rohrmann
Author content
All content in this area was uploaded by Sabine Rohrmann on Jan 20, 2014
Content may be subject to copyright.
Development of a short questionnaire to assess the dietary intake
of heterocyclic aromatic amines
Sabine Rohrmann
1,2,
* and Nikolaus Becker
1
1
Deutsches Krebsforschungszentrum, Division of Clinical Epidemiology, Im Neuenheimer Feld 280, D-69120
Heidelberg, Germany:
2
Present address: John Hopkins Bloomberg School of Public Health, Department of
Epidemiology, 615 North Wolfe Street, Room E6133, Baltimore, MD 21205, USA
Submitted 13 August 2001: Accepted 22 January 2002
Abstract
Objective: Development and validation of a short instrument to assess the dietary
intake of heterocyclic aromatic amines (HCA).
Design: At first, a longer instrument asking for the consumption of 11 meat and fish
items and different preparation methods was developed. The degree of browning of
these foods was assessed by means of photos. This questionnaire was sent to 500
participants of the European Prospective Investigation into Cancer and Nutrition
(EPIC) in Heidelberg, Germany, in June 1999. Using 385 completed questionnaires, a
short questionnaire was developed covering just seven food items, which was sent to
the participants again. Of these, 344 were returned within four months. Total dietary
intake of HCA as well as the intake of different HCA were calculated and compared
between both versions.
Results: Median dietary intake of total HCA was 103 ng day
21
as assessed with the
short version; the intakes of 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine
(PhIP), 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx) and 2-amino-3,4,8-
trimethylimidazo[4,5-f]quinoxaline (DiMeIQx) were 63, 34 and 2 ng day
21
, respect-
ively. These results did not differ significantly from those obtained with the longer
version. Spearman rank correlation coefficients between the long and the short
version ranged from 0.46 to 0.6. In quartile cross-classification, 70–78% of the
participants were assigned into the same or an adjacent quartile while categorisation
into opposite quartiles was #3.5%.
Conclusion: The short version of the HCA questionnaire demonstrates good validity
compared with the longer version. The intake of HCA as assessed with the short
questionnaire is comparable to that found in other studies using a short questionnaire.
Keywords
Heterocyclic aromatic amines
Nutrition assessment
Questionnaire
Intake
Epidemiology
EPIC
Thirty years ago, mutagenic and carcinogenic compounds
were found in fried meat and fish that were called
heterocyclic aromatic amines (HCA) because of their
chemical structures
1
. HCA are formed from precursors in
meat and fish (creatinine, sugar, amino acids) at
temperatures exceeding 1308C
2,3
. The amount of HCA
production depends mainly on cooking method, tem-
perature, and the meat or fish itself. The highest amounts
have been found in foods cooked at high temperatures by
methods like barbecuing, grilling and frying
2
. Apart from
the meat itself, meat drippings and gravy made from these
drippings also contain considerable amounts of HCA
4
.
Although the carcinogenicity of different HCA has been
proved in animal studies
5,6
, conflicting results arose from
epidemiological studies. In spite of some case –control
studies conducted in Uruguay and the USA showing a
possible relationship between the intake of HCA and the
risk of breast, colon, lung and gastric cancer
7–13
, other
groups were not able to detect any association between
HCA intake and cancer risk at different sites
14 – 17
. Only in
some of these studies
8,9,14 – 16
was a questionnaire
designed to assess the intake of HCA used. In the study
of Augustsson et al.
16
, a questionnaire was applied that
assessed the degree of browning of foods with the help of
photographs, showing the food with different degrees of
browning. Voskuil et al.
18
examined the ability of this HCA
questionnaire to assess the intake when the number of
items included in the questionnaire was reduced. Their
results were encouraging, indicating that the loss of
accuracy was small when the number of food items
decreased.
It was the purpose of the present study to develop a
short questionnaire for the assessment of dietary HCA
appropriate for meat and fish consumption habits in
Germany. This was done by reducing the list of items of a
longer version. In a second step this short questionnaire
qThe Authors 2002*Corresponding author: Email srohrman@jhsph.edu
Public Health Nutrition: 5(5), 699–705 DOI: 10.1079/PHN2002331
was sent to the participants again and results of both
versions were compared.
Methods
Subjects
Participants of the European Prospective Investigation into
Cancer and Nutrition (EPIC) in Heidelberg were included
in this study. EPIC is a multi-centre study focusing on the
relation between diet, nutritional and metabolic charac-
teristics, various lifestyle factors and the risk of cancer
19
.In
Heidelberg, 25 544 persons, aged between 35 and 65 years
at the time of recruitment, take part in this study. They
were recruited between 1994 and 1998
20,21
and partici-
pated in the first follow-up between 1998 and 2000
22
.In
June 1999, 500 HCA questionnaires were sent to randomly
selected participants of EPIC–Heidelberg. Four months
later, 385 HCA questionnaires had been returned. After
developing the short version of the questionnaire, this was
sent again to these 385 participants.
Development of the questionnaire
The development of the questionnaire was done in two
steps. In the first step, the meat and fish items that
contributed most to meat and fish intake, as computed
from the food-frequency questionnaire (FFQ) used in
EPIC–Heidelberg during the recruitment
23
, were chosen
for the HCA questionnaire. In addition, food items that
were not eaten in high amounts but may contribute much
to HCA intake because of high HCA concentration
2
were
included. For each of these food items, the questionnaire
addressed different cooking methods usually used in
Germany to prepare these foods. For inclusion, the
cooking methods had to be relevant for HCA formation.
Steaming, boiling, micro-waving and deep-frying do not
cause relevant HCA concentrations
2,24 – 26
, thus only pan-
frying, broiling and grilling were included in the
questionnaire. In total, 26 combinations of food items
and cooking methods were considered in the question-
naire (Table 1). The amount of each food item consumed
per day was not assessed in the HCA questionnaire,
because these data could be obtained from the FFQ
applied in the recruitment of EPIC–Heidelberg
23,27
. The
amounts of the items consumed were computed by
combining consumption frequency and portion size.
Since HCA intake depends on the degree of browning of
the foods
2
, participants could indicate which degree of
browning they prefer for each of the 11 food items. This
was done by means of photographs showing the food
items with four different degrees of browning (lightly
browned, moderately browned, strongly browned,
extremely browned)
28
.
The HCA concentrations in different food items for
different cooking methods and degrees of browning were
taken from the literature
4,29 – 33
. By combining information
about degree of browning, cooking method and the
amount of meat and fish intake, the dietary intake of HCA
per day was calculated. HCA intake from gravy was
assessed by asking for the use of meat or fish drippings to
prepare gravy. When these questions were answered
positively, the intake was calculated by combining
information on gravy consumption from the FFQ with
HCA concentration in gravy of the corresponding meat or
fish item and cooking method.
We calculated the HCA intake from this questionnaire
for each combination of food item and preparation
method, and used these results to develop a short version.
A variance-based method, Max_r, was used to reduce the
list of food items and preparation methods
34,35
. From a
sample of kfood items (in this study k¼26), Max_r selects
a subset of Lfoods (L!k) that best preserves the
between-person variance in nutrient intake. The adequacy
of the chosen subset of foods is measured by the Pearson
correlation coefficient rbetween the nutrient intake
calculated from the complete list of items and the intake
estimated from the subset of items. For selecting the subset
of food items, Max_r first chooses the food item that
maximises the Pearson correlation coefficient rbest; then,
excluding this item, the best item from the remaining list is
chosen, etc. The results of Max_r were used to reduce the
HCA questionnaire. The shorter version thus developed
was sent to those participants from whom we had received
the first HCA questionnaire.
Statistics
HCA intake was calculated from the longer and short
versions of the HCA questionnaire. The Wilcoxon matched
pairs signed rank sum test was used to test for differences
in HCA intake calculated from the long and short versions
of the questionnaire since data were not normally
distributed. Two-tailed P-values ,0.05 were considered
to indicate statistical significance. For the total HCA intake
as well as for the single HCA, Spearman rank correlation
coefficients and their 95% confidence limits were
calculated between both versions of the questionnaire
using the SAS procedures PROC FREQ and the options
Table 1 Food items and cooking methods included in the first,
longer version of the HCA questionnaire used in a pilot study in
EPIC–Heidelberg
Food item Cooking method
Beef steak, fillet, loin Pan-fried, broiled, grilled
Roast beef, beef roulade, goulash Pan-fried
Pork chop, cutlet, steak, fillet, loin Pan-fried, broiled, grilled
Pork roast, goulash Pan-fried
Bacon, pork belly Pan-fried, grilled
Smoked ham, ribs of pork Pan-fried, broiled, grilled
Meat balls, meat loaf Pan-fried, broiled
Liver loaf (‘leberkaese’) Pan-fried, broiled, grilled
Bratwurst Pan-fried, grilled
Fried chicken, fried turkey Pan-fried, broiled, grilled
Fish Pan-fried, broiled, grilled
S Rohrmann and N Becker700
MEASURES. This was additionally done for subgroups of
the study population regarding sex, age, body mass index
(BMI), smoking status, school education and under-
reporting. BMI was calculated as weight in kilograms
divided by height in metres squared and was categorised
according to sex and age
36
. Underreporting was defined
according to Goldberg et al.
37
and an energy intake
(EI)/basal metabolic rate (BMR) ratio of 0.92 was the cut-
off point. Besides Spearman correlation coefficients, the
ability of the short version to correctly classify people
according to their total HCA intake as well as their intake
of the single HCA was measured by cross-classification
into quartiles in comparison with the longer version
38
. All
statistical analyses were performed using SAS version 6.12
(SAS Institute Inc., Cary, NC, USA).
Results
The participation rate in this pilot study was 68.8%
(344/500). Among the 344 participants, 45.6% were men
and 54.4% were women. Their age ranged from 36 to 66
years, with a median of 49 years. All questionnaires had a
sufficient quality regarding completeness to be included in
the study.
Using the results of the first longer version of the HCA
questionnaire of 385 participants, we calculated the
contribution to the interpersonal variation of HCA intake
of each combination of food item and cooking method.
The Pearson correlation coefficient between the HCA
intake from broiled roast beef and the total HCA intake
calculated from all foods and cooking methods amounted
to r¼0:76 (Table 2). This item explained most of the
between-person variance ðr2¼0:58Þ:Three combinations
of food and cooking method explained more than 90% of
between-person variance in HCA intake. Twelve combi-
nations reached a Pearson correlation coefficient of r¼
1:0:These combinations of food items and cooking
methods were included in the short questionnaire.
When comparing HCA intake calculated from the long
version with the intake calculated from the short version,
the results did not differ significantly (Table 3). The median
intakes of 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyri-
dine (PhIP), 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline
(MeIQx) and 2-amino-3,4,8-trimethylimidazo[4,5-f]quin-
oxaline (DiMeIQx) calculated from the short version were
Table 2 Ranking of food items and cooking methods according to between-person variance of
HCA intake chosen by Max_r ðn¼385Þ:These food items and cooking methods were used in the
short version of the HCA questionnaire
Number Food item and cooking method
Cumulated Pearson
correlation coefficient, r
1 Roast beef, broiled 0.761
2 Pork roast, broiled 0.896
3 Fried chicken, turkey, grilled 0.980
4 Fish, broiled 0.985
5 Pork chops, cutlet, fillet, loin, steak, pan-fried 0.990
6 Beef steak, fillet, loin, pan-fried 0.993
7 Fried chicken, turkey, broiled 0.995
8 Fried chicken, turkey, pan-fried 0.997
9 Beef steak, fillet, loin, broiled 0.998
10 Fish, grilled 0.999
11 Pork chops, cutlet, fillet, loin, steak, broiled 0.999
12 Ribs of pork, pan-fried 1.000
Table 3 5th, 25th, 50th (median), 75th and 95th percentiles, and maximum of HCA intakes (in ng day
21
) calculated from the long and
the short versions of the questionnaire ðn¼344Þ
Percentile
Questionnaire 5th 25th 50th 75th 95th Maximum P-value*
Total HCA Long 2 21 89 333 801 10 432 0.64
Short 2 29 103 303 804 4194
PhIP Long 0 7 53 253 646 9208 0.25
Short 0 6 63 225 650 4060
MeIQx Long 0 8 25 66 173 1085 0.10
Short 1 12 34 118 160 880
DiMeIQx Long 0 0 2 5 25 140 0.11
Short 0 0 2 4 17 123
IQ Long 0 0 0 0 3 17 0.15
Short 0 0 0 1 3 8
* Wilcoxon rank sum test.
PhIP– 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine; MelQx – 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline; DiMelQx –2-amino-3,4,8-trimethylimidazo
[4,5-f]quinoxaline; IQ–2-amino-3-methylimidazo[4,5-f]quinoline.
Assessment of heterocyclic aromatic amines intake 701
63, 34 and 2 ng day
21
, respectively. The median intake of
total HCA amounted to 102 ng day
21
as assessed with the
short version, with a range from 0 to 4194 ng day
21
. The
range in the longer version was larger, but in both
versions the 95th percentile of total HCA intake was
about 800 ng day
21
. PhIP contributed most to the total
HCA intake, followed by MeIQx.
Spearman rank correlation coefficient between the long
and the short versions was 0.5 for total HCA, 0.46 for PhIP,
0.58 for DiMeIQx and 0.6 for MeIQx. All correlations were
significant at P,0:001 (Table 4). Fig. 1 shows the
relationship between the total HCA intakes calculated
from the long and short versions of the questionnaire in
detail. When looking at the correlation coefficients of
subgroups, some (although not statistically significant)
differences could be observed (Table 5). The correlation
coefficient was higher for men than for women and an
increasing correlation was observed for an increasing
number of years of school education. Among persons with
different BMI, no tendency could be observed. In quartile
cross-classification between the long and the short
versions of the questionnaire, 42.4% (PhIP) and 50.0%
(total HCA) were assigned to the same quartile of intake
(Table 4). Between 32.8% and 40.7% were classified into
an adjacent quartile and gross misclassification amounted
to 5.8% of the participants.
Discussion
The aim of this pilot study was the development of a short
questionnaire for the assessment of HCA intake. Instru-
ments for the assessment of these substances are needed
since several studies that investigated the association
between HCA intake and the cancer risk of different sites
gave conflicting results. A recently published US study
indicates a higher risk of breast cancer with an increasing
intake of PhIP
9
, while an older study was not able to
demonstrate any association
15
. Furthermore, two studies
point at a higher risk of lung cancer and colorectal
adenomas, respectively, and the intake of MeIQx
7,8
.
Unlike several older studies
10 – 13
, these studies used a
specially designed questionnaire
39
. In these older studies
instead, assumptions about the frequency of cooking
Fig. 1 Scatter plot of total HCA intakes calculated from the long
and short versions of the HCA questionnaire (logarithmic scale)
Table 4 Spearman rank correlation coefficients between the HCA intakes computed from the long and short HCA
questionnaires and results of quartile cross-classification between HCA intakes calculated from the long and short
versions of the questionnaire ðn¼344Þ
Correlation coefficient
(95% confidence limit)
Quartile cross-classification
Same quartile Adjacent quartile Opposite quartile
Total HCA 0.51 (0.42–0.60) 172 (50.0%) 113 (32.8%) 14 (4.1%)
PhIP 0.46 (0.37–0.56) 146 (42.4%) 140 (40.7%) 20 (5.8%)
MeIQx 0.60 (0.52–0.68) 163 (47.4%) 124 (36.0%) 5 (1.5%)
DiMeIQx 0.58 (0.50–0.66) 169 (49.1%) 126 (36.6%) 12 (3.5%)
IQ 0.48 (0.38–0.58) NC NC NC
NC – quartile cross-classification was not calculated for IQ.
Table 5 Spearman rank correlation coefficients between total
HCA intakes computed from the long and short HCA question-
naires for different subgroups of the study population ðn¼344Þ
Characteristic Subgroup n
Correlation
coefficient
(95% confidence
limits)
Sex Female 186 0.49 (0.36– 0.62)
Male 158 0.55 (0.42–0.68)
Age group (years) 35– 44 96 0.46 (0.27 –0.64)
45–54 139 0.56 (0.42– 0.69)
55+ 109 0.49 (0.31– 0.66)
School education #8 years 120 0.39 (0.21–0.57)
10 years 70 0.49 (0.29 –0.69)
High school degree 154 0.56 (0.43 –0.68)
Underreporting* Underreporter 41 0.60 (0.37–0.83)
Non-underreporter 303 0.49 (0.39–0.59)
BMI Underweight 60 0.67 (0.48 – 0.85)
Normal 175 0.43 (0.29 –0.57)
Overweight 60 0.56 (0.35 –0.76)
Obese 49 0.48 (0.26–0.70)
Smoking status Non-smoker 146 0.53 (0.39 –0.66)
Ex-smoker 112 0.53 (0.37–0.69)
Smoker 86 0.45 (0.25 – 0.65)
* Underreporting defined as EI/BMR ,0.92 (in accordance with Goldberg
et al.
37
).
S Rohrmann and N Becker702
methods used in the study population and the preferred
degree of browning were made. This probably introduces
a bias in the estimation of HCA intake. Since a
questionnaire used for nutrient assessment should be as
short, but also as accurate as possible, Voskuil and co-
workers
18
evaluated the possibility to reduce the number
of food items used in questionnaires for HCA assessment.
They were able to demonstrate that the resulting loss of
accuracy was negligible when the number of dishes was
reduced from 39 to 15.
Our study was conducted in a similar way. In the first
step of the development of the questionnaire, a longer
version including the most important meat and fish items
was developed. This version was reduced in a second step
and, in addition to what Voskuil et al.
18
did in their study,
validated in a field study.
For calculating HCA intakes no analyses of meat and
fish samples prepared by different cooking methods
were performed. Instead, data from the literature were
used
29 – 33
. We selected those data that were analysed in a
standardised way and were most likely to be appropriate
for the situation in Germany. Nevertheless, this approach
may be a source of bias. Firstly, the HCA concentration in
the foods included in the questionnaire in Heidelberg
might be higher or lower than the concentration in the
foods used in the literature, although the foods look
similar in the pictures used to indicate the degree of
browning. Secondly, HCA concentration in typical German
foods, e.g. liver loaf (‘leberkaese’), had to be assumed
because no corresponding data were available. In these
cases, data for foods that were similar regarding fat, water
and protein contents were chosen, but this does not
automatically imply that HCA content is approximately the
same. In conclusion, using literature data on HCA
concentration in meat and fish to estimate the intake
may introduce an over- or underestimation. This is
important when the aim of a questionnaire is to assess the
intake of a nutrient quantitatively. If the purpose of a
questionnaire, especially of a short questionnaire, is to
assess the variance of the intake and to categorise
individuals according to their intake, an over- or
underestimated intake is negligible if the bias is non-
differential. Non-differential misclassification is often a
problem inherent in data collection methods
40
.
Included in the longer version of the questionnaire were
those food items that are eaten in high amounts according
to the results of the recruitment FFQ. Additionally, food
items that were eaten less frequently in this cohort, e.g.
bacon, were included because they can contribute much
to HCA intake due to the high HCA concentration in the
prepared food
2
. Using the variance-based method
Max_r
34,35
, the list of food items and cooking methods
was reduced. The shorter version of the HCA ques-
tionnaire included those food items that contributed most
to between-person variance of total HCA intake. In our
study, one cooked food item (broiled roast beef)
explained 58% of between-person variance of total HCA
intake, which is comparable to the results of Voskuil
et al.
18
. In their analyses, fried pork chops explained 48%
of between-person variance. Willett
41
suggested that a
short questionnaire should include at least as many food
items as are used to explain 80% of the variation of nutrient
intake. We included all items that explained 99% of the
variance of HCA intake. Since two items explain more than
80% of the variance, this will not give a detailed picture of
important meat items and cooking methods used in
Germany. For that reason 12 combinations of food items
and cooking methods were included in the questionnaire.
The calculated intake of 103 ng HCA per day is
comparable to the intake computed in Swedish
studies
16,28
. They calculated for different groups of
participants a median intake varying from 77 to
160 ng day
21
. PhIP contributed a higher proportion to
HCA intake in the Heidelberg study than in the Swedish
one
28
. The results from Heidelberg concerning PhIP intake
are more comparable with results from US studies
39
.US
data
31
were used to calculate HCA intake from grilled
chicken in our pilot study because no data from Sweden
were available
28
. Therefore, grilled chicken contributes
much to PhIP intake in the present study. In the Swedish
study, HCA concentration of pan-fried chicken was used
for grilled chicken, which might underestimate HCA
intake. However, using US data for grilled chicken in a
German study possibly overestimates HCA concentration,
since Americans may prefer eating meat darker than
Germans.
In the study conducted in Heidelberg, correlation
coefficients between the long and short versions of the
HCA questionnaire are 0.46 or higher, which is an
acceptable result
42
. Although correlation coefficients in
the subgroups do not differ significantly (overlapping
confidence limits), there are some interesting differences.
Increasing correlation coefficients with increasing edu-
cational level might a hint at more accurate filling in of
questionnaires by better-educated subjects. The differing
correlation coefficient between underreporters and non-
underreporters can be due to chance alone, because of the
strata size. However, underreporters might be more
precise in reporting the way in which they prepare their
meals. The results of the quartile cross-classification of the
Heidelberg pilot study are in the range of other studies
comparing a short questionnaire with a longer ver-
sion
43,44
. However, one must keep in mind that the longer
HCA questionnaire used as a reference instrument cannot
be considered as a gold standard. We do not believe that
the long questionnaire reflects the true HCA intake in the
study population, but our assumption behind this kind of
questionnaire development and validation is that the long
version of the questionnaire assesses the HCA intake more
accurately than the short version because it includes more
food items and cooking methods. Therefore, we would
like to know and test whether a version with a reduced
Assessment of heterocyclic aromatic amines intake 703
number of items assesses the intake as well as the long
version without losing too much accuracy. The similarities
in design of the longer and the short versions of the
questionnaire may cause an overestimated validity of the
short version
41
. No gold standard exists to validate an HCA
questionnaire, since measurements of DNA-adducts
45,46
or
excreted HCA metabolites in urine
47,48
are either not
sensitive enough or have other limitations. An appropriate
way could be the double portion technique, where a part
of the food consumed is analysed for its HCA content, but
this method is expensive and only useful in small groups
of individuals.
In conclusion, the developed short version of an HCA
questionnaire demonstrates good validity in comparison
with the longer, more extensive version. The total intake of
HCA as assessed with the short questionnaire is consistent
with the exposure in other studies using a short
questionnaire. According to these results, the short
questionnaire can be used to further examine whether
HCA intake is related to cancer risk. Since HCA exposure
varies depending on cooking methods and the level of
cooked meat consumption, EPIC offers the opportunity to
obtain information on cooking of meat and fish in different
regions of Europe with a variety of consumption habits.
Using the questionnaire at least in some of these countries
and regions provides a good chance to proceed in our
knowledge on the association between HCA intake,
cooking methods and cancer risk.
Acknowledgements
WewouldliketothankallparticipantsofEPIC–
Heidelberg, as well as Dr Gunnar Steineck and Dr
Katarina Augustsson for allowing us to use their photos for
assessment of the degree of browning of different foods.
EPIC–Heidelberg is supported by Deutsche Krebshilfe,
Deutsches Krebsforschungszentrum and the programme
‘Europe against Cancer’ of the European Commission.
References
1 Sugimura T. Overview of carcinogenic heterocyclic amines.
Mut. Res. 1997; 376: 211–9.
2 Skog K, Johansson MAE, Ja
¨gerstad MJ. Carcinogenic
heterocyclic amines in model systems and cooked foods: a
review on formation, occurrence and intake. Food Chem.
Toxicol. 1998; 36: 879–96.
3 Layton DW, Bogen KT, Knize MG, et al. Cancer risk of
heterocyclic amines in cooked foods: an analysis and
implications for research. Carcinogenesis 1995; 16: 39 –52.
4 Gross GA, Gru¨ter A. Quantification of mutagenic/carcino-
genic heterocyclic aromatic amines in food products.
J. Chromatogr. 1992; 592: 271– 8.
5 Adamson RH, Thorgeirsson UP, Snyderwine EG, et al.
Carcinogenicity of 2-amino-3-methylimidazo[4,5-f]quino-
line in nonhuman primates: induction of tumors in three
macaques. Jpn. J. Cancer Res. 1990; 81:10–4.
6 Shirai T, Tamano S, Sano M, et al. Carcinogenicity of
2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) in
rats: dose–respose studies. Princess Takamatsu Symposia
1995; 23: 232–9.
7 Sinha R, Kulldorff M, Chow W-H, et al. Dietary intake of
heterocyclic amines, meat-derived mutagenic activity, and
risk of colorectal adenomas. Cancer Epidemiol. Biomark.
Prev. 2001; 10: 559–62.
8 Sinha R, Kulldorff M, Swanson CA, et al. Dietary heterocyclic
amines and the risk of lung cancer among Missouri women.
Cancer Res. 2000; 60: 3753–7.
9 Sinha R, Gustafson DR, Kulldorf M, et al. 2-Amino-1-methyl-
6-phenylimidazo[4,5-b]pyridine, a carcinogen in high-
temperature-cooked meat, and breast cancer risk. J. Natl.
Cancer Inst. 2000; 92: 1352–4.
10 De Stefani E, Boffetta P, Mendilaharsu M, et al. Dietary
nitrosamines, heterocyclic amines, and risk of gastric cancer:
a case–control study in Uruguay. Nutr. Cancer 1998; 30:
158–62.
11 De Stefani E, Ronco A, Mendilaharsu M, et al. Case–control
study on the role of heterocyclic amines in the atiology of
upper aerodigestive cancers in Uruguay. Nutr. Cancer 1998;
32:43–8.
12 De Stefani E, Ronco A, Mendilaharsu M, et al. Meat intake,
heterocyclic amines, and risk of breast cancer: a case –
control study in Uruguay. Cancer Epidemiol. Biomark. Prev.
1997; 6: 573–81.
13 De Stefani E, Doneo-Pellegrini H, Mendilaharsu M, et al.
Meat intake, heterocyclic amines and risk of colorectal
cancer: a case–control study in Uruguay. Int. J. Oncol. 1997;
10: 573–80.
14 Norrish AE, Ferguson LR, Knize MG, et al. Heterocyclic
amine content of cooked meat and risk of prostate cancer.
J. Natl. Cancer Inst. 1999; 91: 2038–44.
15 Delfino RJ, Sinha R, Smith C, et al. Breast cancer, heterocyclic
amines from meat and N-acetyltransferase 2 genotype.
Carcinogenesis 2000; 21: 607–15.
16 Augustsson K, Skog K, Ja
¨gerstad M, et al.Dietary
heterocyclic amines and cancer of the colon, rectum,
bladder, and kidney: a population-based study. Lancet 1999;
353: 703–7.
17 Lyon JL, Mahoney AW. Fried foods and the risk of colon
cancer. Am. J. Epidemiol. 1988; 128: 1000– 6.
18 Voskuil DW, Augustsson K, Dickman PW, et al. Assessing the
human intake of heterocyclic aromatic amines: limited loss
of information using reduced sets of questions. Cancer
Epidemiol. Biomark. Prev. 1999; 8: 809 –14.
19 Riboli E, Kaaks R. The EPIC project: rationale and study
design. Int. J. Epidemiol. 1997; 26(1): S6–14.
20 Boeing H, Wahrendorf J, Becker N. EPIC – Germany – a
source for studies into diet and risk of chronic diseases. Ann.
Nutr. Metab. 1999; 43: 195– 204.
21 Boeing H, Korfmann A, Bergmann MM. Recruitment
procedures of EPIC–Germany. Ann. Nutr. Metab. 1999;
43: 205–15.
22 Bergmann M, Bussas U, Boeing H. Follow-up procedures in
EPIC–Germany – data quality aspects. Ann. Nutr. Metab.
1999; 43: 225–34.
23 Brandstetter BR, Korfmann A, Bergmann MM. Dietary habits
in the German EPIC cohorts: food group intake estimated
with food frequency questionnaire. Ann. Nutr. Metab. 1999;
43: 246–57.
24 Zhang X-M, Wakabayashi K, Liu Z-C, et al. Mutagenic and
carcinogenic heterocyclic amines in Chinese cooked foods.
Mut. Res. 1988; 201: 181–8.
25 Taylor SL, Berg CM, Shoptaugh NH, et al. Lack of mutagens
in deep-fat fried foods obtained at the retail level. Food
Chem. Toxicol. 1982; 20: 209–12.
26 Nilsson L, O
¨vervik E, Fredholm L, et al. Influence of frying fat
on mutagenic activity in lean pork meat. Mut. Res. 1986; 171:
115–21.
27 Bohlscheid-Thomas S, Hoting I, Boeing H, et al.
S Rohrmann and N Becker704
Reproducibility and relative validity of energy and macro-
nutrient intake of a food frequency questionnaire developed
for the German part of the EPIC project. Int. J. Epidemiol.
1997; 26(Suppl 1): S71–81.
28 Augustsson K, Skog K, Ja
¨gerstad M, et al. Assessment of the
human exposure of heterocyclic amines. Carcinogenesis
1997; 18: 1931–5.
29 Abdulkarim BG, Smith JS. Heterocyclic amines in fresh and
processed meat products. J. Agric. Food Chem. 1998; 46:
4680–7.
30 Sinha R, Knize MG, Salmon CP, et al. Heterocyclic amine
content of pork products cooked by different methods and
by varying degrees of doneness. Food Chem. Toxicol. 1998;
36: 289–97.
31 Sinha R, Rothman N, Brown ED, et al. High concentrations of
the carcinogen 2-amino-1-methyl-6-phenylimidazo[4,5-
b]pyridine [PhIP] occur in chicken but depending on the
cooking temperature. Cancer Res. 1995; 55: 4516–9.
32 Skog K, Augustsson K, Steineck G, et al. Polar and non-polar
heterocyclic amines in cooked fish and meat products and
their corresponding pan residues. Food Chem. Toxicol. 1997;
35: 555–65.
33 Skog K, Steineck G, Augustsson K, et al. Effect of cooking
temperature on the formation of heterocyclic amines in fried
meat products and pan residues. Carcinogenesis 1995; 16:
861–7.
34 Mark SD, Thomas DG, Decarli A. Measurement of the
exposure to nutrients: an approach to the selection of
informative foods. Am. J. Epidemiol. 1996; 143: 514–21.
35 Thomas DG, Mark SD. Max_r: an optimal method for the
selection of subsets of foods for the measurement of specific
nutrient exposures. Comp. Meth. Prog. Biomed. 1997; 54:
151–6.
36 National Research Council. Diet and Health: Implications for
Reducing Chronic Disease Risk. Washington, DC: National
Academy Press, 1989.
37 Goldberg GR, Black AE, Jebb SA, et al. Critical evaluation of
energy intake data using fundamental principles of energy
physiology: 1. Derivation of cut-off limits to identify under-
recording. Eur. J. Clin. Nutr. 1991; 45: 569 – 81.
38 Garrow JS. Validation of methods for estimating habitual
diets: proposed guidelines. Eur. J. Clin. Nutr. 1995; 49:
231–2.
39 Byrne C, Sinha R, Platz EA, et al. Predictors of dietary
heterocyclic amines in three prospective cohorts. Cancer
Epidemiol. Biomark. Prev. 1998; 7: 523 –9.
40 Gordis L. Epidemiology. Philadelphia, PA: WB Saunders
Company, 1996.
41 Willett WC. Nutritional Epidemiology, 2nd ed. New York:
Oxford University Press, 1998.
42 Block G, Hartman AM. Issues in reproducibility and validity
of dietary studies. Am. J. Clin. Nutr. 1989; 50: 1133– 8.
43 Dobson AJ, Blijlevens R, Alexander HM, et al. Short fat
questionnaire: a self-administered measure of fat-intake
behaviour. Aust. J. Public Health 1993; 17: 144– 9.
44 Bingham SA, Gill C, Welch A, et al. Comparison of dietary
assessment methods in nutritional epidemiology: weighed
records v. 24 h recalls, food-frequency questionnaires and
estimated-diet records. Br. J. Nutr. 1994; 72: 619 – 43.
45 Ji H, Yu MC, Stillwell WG, et al. Urinary excretion of
2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline in white,
black, and Asian men in Los Angeles County. Cancer
Epidemiol. Biomark. Prev. 1994; 3: 407 –11.
46 Kidd LCR, Stillwell WG, Yu MC, et al. Urinary excretion of
2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) in
white, African-American, and Asian-American men in Los
Angeles County. Cancer Epidemiol. Biomark. Prev. 1999; 8:
439–45.
47 Mauthe RJ, Dingley KH, Leveson SH, et al. Comparison of
DNA-adduct and tissue-available dose levels of MeIQx in
human and rodent colon following administration of a very
low dose. Int. J. Cancer 1999; 80: 539–45.
48 Wakabayashi K, Totsuka Y, Fukotome K, et al. Human
exposure to mutagenic/carcinogenic heterocyclic amines
and co-mutagenic b-carbolines. Mut. Res. 1997; 376: 253–9.
Assessment of heterocyclic aromatic amines intake 705
