Dietary acrylamide and human cancer
Low levels of acrylamide have been found in several foods cooked at high temperatures. While there is sufficient evidence for the carcinogenicity of acrylamide in experimental animals, the few epidemiologic studies conducted to date on occupational and dietary exposure to acrylamide have found no consistent evidence of association with human cancer risk. Using data from an integrated network of Italian and Swiss hospital-based case-control studies, we analyzed the relation between dietary acrylamide intake and cancers of the oral cavity and pharynx (749 cases, 1,772 controls), esophagus (395 cases, 1,066 controls), large bowel (1,394 cases of colon, 886 cases of rectal cancer, 4,765 controls), larynx (527 cases, 1,297 controls), breast (2,900 cases, 3,122 controls), ovary (1,031 cases, 2,411 controls) and prostate (1,294 cases, 1,451 controls). All the studies included incident, histologically confirmed cancer cases and controls admitted to the same network of hospitals for acute nonneoplastic conditions. We calculated odds ratios (ORs) using multivariate logistic regression models, adjusted for energy intake and other major covariates of interest. The ORs for the highest versus the lowest quintile of acrylamide intake were 1.12 (95% CI = 0.76–1.66) for cancer of the oral cavity/pharynx, 1.10 (95% CI = 0.65–1.86) for esophageal, 0.97 (95% CI = 0.80–1.18) for colorectal, 1.23 (95% CI = 0.80–1.90) for laryngeal, 1.06 (95% CI = 0.88–1.28) for breast, 0.97 (95% CI = 0.73–1.31) for ovarian and 0.92 (95% CI = 0.69–1.23) for prostate cancer. None of the trend in risk was significant. This uniquely large and comprehensive data set does not show any consistent association between intake of acrylamide and the risk of breast and several other common cancers. © 2005 Wiley-Liss, Inc.
Acrylamide is a vinyl monomer that improves the aqueous solubility, adhesion and crosslinking of polymers. The International Agency for Research on Cancer (IARC) classified acrylamide in 1994 as “probably carcinogenic to humans” (Group 2A) on the basis of sufficient evidence for carcinogenicity in experimental animals and mechanistic considerations.1 An experiment in rats, which were given acrylamide in drinking water, found an increased incidence of breast, thyroid and central nervous system tumors.2 Acrylamide exerts its mutagenic effect by the capacity of its epoxide metabolite glycidamide to form DNA adducts.3
In April 2002, a team of Swedish researchers reported the formation of acrylamide in several heated foods.4 This finding caused considerable public concern about cancer risk.5 Since then, a number of studies have confirmed and measured the acrylamide content of many food items6, 7, 8, 9, 10, 11, 12 and have calculated the daily average dietary intakes of acrylamide in various countries. In Europe, these ranged between 0.3 μg/kg body weight (bw) for Switzerland,13 and 0.5 μg/kg bw for Sweden, Norway, The Netherlands and Belgium.7, 8, 14, 15, 16 On the basis of tests in animals, a Joint FAO/WHO Expert Committee on Food Additives (JECFA) concluded that current acrylamide levels in foods may indicate a human health concern and that cancer could be the most important adverse effect of acrylamide.12
To date, only a small number of epidemiologic studies have been published on occupational and dietary exposure to acrylamide and cancer risk. Among occupational studies, a cohort study of 8,508 male workers found no excess risk for mortality from all cancers and from most cancer sites.17 Only pancreatic cancer was significantly associated with acrylamide, with a standardized mortality ratio of 2.26 (95% confidence interval (CI) = 1.03–4.29) for the highest level of cumulative exposure (> 0.30 mg/m3-year). In another cohort study including 371 employees assigned to acrylamide operations, Sobel et al.18 observed no increase in overall mortality or from any specific cancer, except for an excess of digestive tract cancers in a subgroup of workers previously exposed to organic dyes. Occupational exposure to acrylamide, however, mostly involves male subjects. Thus, occupational studies provide no information on breast cancer and other female hormone-related neoplasms. Moreover, occupational exposures differ from dietary ones according to levels of acrylamide and routes of administration (i.e., inhalation or through dermal absorption); therefore, the results from the 2 settings may not be comparable.
Among studies considering dietary intake of acrylamide, a cohort study of 43,404 women and 2 population-based case-control studies from Sweden found no direct association between acrylamide intake and risk of selected cancers. The relative risks for the highest versus the lowest level of intake were 1.2 for breast cancer in the prospective study19 and 0.6 for large bowel, 0.8 for bladder and kidney20 and 1.1 for renal cell cancer in the case-control studies.21 We examined the intake of fried and baked potatoes and coffee, i.e., 2 major sources of acrylamide in foods, in a series of case-control studies from Italy and Switzerland on cancers of the oral cavity/pharynx, esophagus, colorectum, larynx, breast and ovary and found no consistent excess risks. The odds ratios (ORs) for the highest versus the lowest level of intake ranged between 0.8 and 1.1 for consumption of fried/baked potatoes22 and between 0.4 and 1.0 for coffee drinking.23 To provide further information on the issue, we have analyzed the relation between dietary intake of acrylamide and risk of cancer at various sites from a series of large case-control studies conducted in Italy and the Swiss canton of Vaud between 1991 and 2002.
Material and methods
Data were obtained from an integrated series of hospital-based case-control studies with the same design, questionnaire and inclusion criteria.24, 25, 26, 27, 28, 29, 30 Information was collected between 1991 and 2002 in several areas of northern Italy (greater Milan, the provinces of Pordenone, Padua, Udine and Gorizia, the urban area of Genoa and the province of Forlì), central Italy (the provinces of Rome and Latina) and southern Italy (the urban area of Naples). Studies on oral cavity/pharynx, esophagus, colorectum, larynx and breast were also conducted in the Swiss canton of Vaud.
The case-control studies included incident, histologically confirmed cases of cancer. The first study, on cancer of the oral cavity and pharynx, included 749 cases (median age, 57 years) and 1,772 controls (median age, 57 years). The second study, on cancer of the esophagus, included 395 cases (median age, 60 years) and 1,066 controls (median age, 60 years). The third study, on large bowel, included 1,394 cases of colon cancer (median age, 62 years), 886 cases of rectal cancer (median age, 63 years) and 4,765 controls (median age, 58 years). The fourth study, on laryngeal cancer, included 527 cases (median age, 61) and 1,297 controls (median age, 61). The fifth study, on breast cancer, included 2,900 cases (median age, 55 years) and 3,122 controls (median age, 56 years). The sixth study, on cancer of the ovary, was conducted in Italy only and included 1,031 cases (median age, 56 years) and 2,411 controls (median age, 57 years). The seventh study, on prostate cancer, was conducted in Italy only and included 1,294 cases (median age, 66 years) and 1,451 controls (median age, 63 years). Controls were patients hospitalized for a wide spectrum of acute nonneoplastic conditions (overall, 23% had traumas, 28% nontraumatic orthopedic disorders, 23% acute surgical conditions and 26% miscellaneous other illnesses). All cases and controls were < 80 years old and were identified and questioned by trained interviewers during their hospital stay in a network of teaching and general hospitals in the areas under surveillance. The proportion of refusals was less than 5% among cases and controls in all Italian centers and about 15% in Switzerland.
Interviewers used a structured questionnaire, including information on sociodemographic factors, anthropometric variables, smoking, alcohol and other lifestyle habits, a problem-oriented medical history, physical activity and history of cancer in relatives. Information on diet referred to the 2 years preceding diagnosis and was based on a food-frequency questionnaire (FFQ), validated for nutrient intake and tested for reproducibility for specific nutrients and food items. The FFQ included 78 foods, food groups, or recipes and allowed the estimation of energy intake.31, 32, 33 Average weekly frequency of consumption of foods was elicited, and therefore the FFQ looked more like a menu list from a restaurant than a shopping list. Frequencies of less than once per week, but at least once per month, were coded as 0.5 per week, whereas never or less than once per month was coded as 0. The FFQ was divided into 6 sections: (1) bread, cereals, first courses; (2) second courses (i.e., meat, fish and other main dishes); (3) side dishes (i.e., vegetables, fried/baked potatoes); (4) fruits; (5) sweets, desserts and soft drinks; (6) milk, hot beverages and sweeteners. At the end of each section, 1 or 2 open questions were used to include foods that were not in the questionnaire, but were eaten at least once per week. Among the items in the FFQ, 2 questions referred specifically to consumption of fried/baked potatoes, including weekly frequency of consumption and usual portion size. Other items considered for the estimation of the average daily acrylamide intake were consumption of coffee, bread, sweet biscuits, crackers, breadsticks and melba toast and, for Italy, “cappuccino” (white coffee). Also, information on breakfast cereals was asked to Swiss subjects. To obtain estimation of the personal intake of acrylamide, we added up the products of frequency of consumption of these foods for the corresponding average content of acrylamide that was obtained from resources made available on the Internet by the World Health Organization (WHO; http://www.who.int/fsf) and the Swiss Federal Office of Public Health (SFOPH; http://www.bag.admin.ch/).6, 13
Reproducibility of questions on food items was evaluated by Spearman correlation coefficients (r), corrected for study center. Values of r for questions on the main food items related to acrylamide ranged from 0.52 to 0.75.31 Validity of nutrient intakes was estimated comparing the FFQ with two 7-day diaries by means of the Pearson product moment correlation coefficient between intake of nutrients computed from the FFQ and the average value derived from the two 7-day diaries records. Value of the Pearson correlation coefficient for validity of energy intake was 0.63.32
OR and the corresponding 95% CI of selected cancers according to quintiles of acrylamide intake (μg/day) were derived using unconditional multiple logistic regression (in unmatched studies, i.e., breast, ovarian, colorectal and prostate cancers) or conditional one (in studies matched on age, sex and study center, i.e., upper aerodigestive tract cancers).34 All regression models included terms for age, sex, study center, education, body mass index and energy intake. According to the cancer site analyzed, further adjustments were made for alcohol drinking, smoking habit, physical activity, parity and family history of cancer at the same site.
Table I shows the distribution of cases and controls according to sex and age.
Table I. Distribution of Cases and Controls According to Sex and Age in the Studies Investigated, Italy and Switzerland, 1991–2002
| Colon|| ||786:2,403||608:2,362||330:1,892||497:1,528||567:1,345|
| Rectum|| ||532:2,403||354:2,362||201:1,892||319:1,528||366:1,345|
Table II gives information on the intake of acrylamide among controls in the 7 studies considered. In all studies, fried and baked potatoes were the major contributors of acrylamide, accounting for 36–44% of intake. The average daily intake of acrylamide among controls ranged between 23.33 and 29.24 μg/day. The upper cutpoint of the first (lowest) quintile of acrylamide intake ranged between 10.07 μg/day in the ovarian cancer study and 13.19 μg/day in the esophageal cancer study. The lower cutpoint of the fifth (highest) acrylamide quintile ranged between 32.39 μg/day in the ovarian cancer study and 40.38 μg/day in the oral/pharyngeal cancer study. All median values were lower than mean values, suggesting a skewed distribution of acrylamide intake. The prostate cancer study had the lowest values of the average daily acrylamide intake per kg of bw, i.e., 0.33 μg/kg bw, as compared to 0.36 μg/kg bw of esophagus and larynx, 0.37 μg/kg bw of ovary, 0.38 μg/kg bw of breast and 0.40 μg/kg bw of oral cavity/pharynx and large bowel cancer studies.
Table II. Information on the Intake of Acrylamide among Controls in the Studies Considered, Italy and Switzerland, 1991–2002
Table III gives, separately for each case-control study considered, the ORs and 95% CI for subsequent quintiles of daily acrylamide intake (μg/day) as compared to the lowest quintile. The ORs for the highest quintile were 1.12 (95% CI = 0.76–1.66) for cancer of the oral cavity/pharynx, 1.10 (95% CI = 0.65–1.86) for esophageal, 0.97 (95% CI = 0.80–1.18) for colorectal, 1.23 (95% CI = 0.80–1.90) for laryngeal, 1.06 (95% CI = 0.88–1.28) for breast, 0.97 (95% CI = 0.73–1.31) for ovarian and 0.92 (95% CI = 0.69–1.23) for prostate cancer. None of the risk estimates or the p-values for linear trend was significant.
Table III. Odds Ratios and 95% Confidence Intervals of Selected Cancers According to Subsequent Quintiles of Acrylamide Intake (μg/Day), Italy and Switzerland, 1991–2002
|Oral cavity/pharynx2||1||1.10 (0.78–1.57)||1.27 (0.89–1.81)||1.04 (0.72–1.51)||1.12 (0.76–1.66)||0.70|
|Esophagus2||1||1.16 (0.75–1.81)||1.20 (0.75–1.93)||0.74 (0.44–1.24)||1.10 (0.65–1.86)||0.67|
|Large bowel3||1||0.89 (0.75–1.05)||1.06 (0.89–1.26)||1.05 (0.88–1.26)||0.97 (0.80–1.18)||0.56|
| Colon3||1||0.89 (0.73–1.09)||1.11 (0.91–1.36)||1.11 (0.90–1.38)||0.98 (0.78–1.23)||0.45|
| Rectum3||1||0.87 (0.68–1.11)||0.95 (0.75–1.22)||0.96 (0.75–1.24)||0.96 (0.73–1.26)||0.92|
|Larynx2||1||1.04 (0.70–1.57)||0.85 (0.56–1.29)||0.89 (0.59–1.36)||1.23 (0.80–1.90)||0.54|
|Breast4||1||1.01 (0.85–1.20)||1.01 (0.85–1.20)||1.09 (0.92–1.31)||1.06 (0.88–1.28)||0.37|
|Ovary4||1||1.03 (0.79–1.34)||1.09 (0.83–1.44)||1.01 (0.76–1.34)||0.97 (0.73–1.31)||0.80|
|Prostate5||1||1.00 (0.77–1.30)||1.22 (0.94–1.58)||1.01 (0.77–1.33)||0.92 (0.69–1.23)||0.65|
This series of large case-control studies found no association between dietary acrylamide intake and risk of any of the cancers investigated. These included breast cancer, i.e., one of the sites at potentially increased risk on the basis of results from an experimental study.2
A concern of the study was the reliability of the estimate of dietary acrylamide intake,35 since the acrylamide content of foods varies widely according to the modalities of cooking (particularly temperature and duration) or the different product brands.4, 6, 10 This information was not available to us. Moreover, the validity of any such information would be undefined. However, even if there was a variation within samples of the same food, a broad agreement emerged between studies according to the average acrylamide contents of various cooked foods.7, 8, 9, 10
We estimated acrylamide intake from consumption of a limited number of foods, while acrylamide has been found, at very low levels, in several other foodstuffs.7 Consequently, the daily intakes we obtained are probably slightly underestimated. Nevertheless, it was calculated that fried potatoes, coffee and breads together account for 75% and 92% of acrylamide intake in Dutch and Swedish diets, respectively.7, 8 In Switzerland, where diet is more similar to the Italian one and where, in addition, a part of cases and controls of our studies were enrolled, a diet-duplicate study estimated that coffee and fried potatoes together may account for 84–89% of daily acrylamide intake.13 Our data were reassuringly similar to these estimations, since fried/baked potatoes contributed for 36–44% of acrylamide in diet, bread and crackers for 26–35% and coffee for 10–13%. Thus, the foods we considered should cover most of the dietary acrylamide intake. Our estimates of mean acrylamide intake, moreover, were consistent with published ones from other populations from Sweden, Norway and the Netherlands,7, 8, 14 whereas in this population a high intake was rather infrequent (i.e., 95th percentile between 0.71 and 0.91 μg/kg bw per day) as compared to other countries.16
Among the strengths of these studies were the use of a validated and reproducible FFQ,31, 32, 33 the similar interview setting of cases and controls, which provides reassurance against potential information bias,36 and the ability to adjust for major selected potential confounding factors (including energy intake). As proposed by Willett and Stampfer,37 energy intake was simply derived as the sum of proteins, fats, starches and sugars, plus alcohol. The absolute value of this measure of energy intake is known to be imprecise, but it provides a ranking of subjects that can be used as a covariate in multivariate models.38
Since fried potatoes and coffee, i.e., two of the main contributors of dietary acrylamide, could be markers of a less healthy lifestyle, it is likely that other potential confounding factors not considered in the analysis would, if anything, further reduce rather than increase the ORs.
Hospital-based case-control studies have several potential limitations34: information was collected after the diagnosis, and thus it is possible that early symptoms of the disease may have caused changes in the diet, and information on nutritional factors can be subject to selection and recall bias. We tried to minimize bias by excluding all control patients with diagnoses linked to long-term changes in diet or admitted for chronic conditions. Kipnis et al.39 suggested that even multiple-day dietary records used to validate the FFQs may be subject to measurement error. Since this is likely to be correlated with measurement error in the FFQ, coefficients for validity estimated using dietary records could be overestimated.40, 41 This could lead to an attenuation of the true effects of dietary correlates.39, 42
Debates are open on the actual risk levels to humans related to acrylamide intake in consideration of discordant findings of animal and epidemiologic studies.14, 43 Two major lines of reasoning emerged to explain such inconsistency: the first one underlines limitations of extrapolating animal data to humans,44, 45 given the uncertain bioavailability of acrylamide after oral consumption and according to various food matrices in humans as well as differences in the metabolism of acrylamide reported among various animal species,16 and the different level of exposure between laboratory data and human diet.45 The second one criticizes epidemiologic studies on various aspects, particularly for their limited power.43, 46 In fact, due to low levels and limited range of acrylamide exposure from foods, cancer risk may be too small to be detectable by observational studies.43
In any case, studies based on a large number of subjects, such as ours, provide a more detailed risk assessment and additional information on the upper risk limits related to dietary acrylamide intake.45 In this investigation, the breast and the large bowel cancer studies had 80% power to estimate, for the upper quintiles of intake, unadjusted risk increases higher than 20% and 19%, respectively. Corresponding values for other studies were 29% for ovarian cancer, 30% for prostate cancer and 40–50% for upper aerodigestive tract cancers. Thus, our data weigh against important excesses in risk of breast and other selected cancers in relation to dietary intake of acrylamide. This finding is consistent with other epidemiologic studies on acrylamide and cancer risk.17, 18, 19, 20, 21, 44
The authors thank Mrs. C. Pasche and F. Lucchini for Swiss data collection and validation and Mrs. I. Garimoldi for editorial assistance. Supported by a fellowship from the Italian Foundation for Cancer Research (to C.P.).