Dietary Acrylamide Intake and Brain Cancer Risk

Acrylamide has been classified as a probable human carcinogen based on animal studies (1), and is since 2002 known to be present in several carbohydrate-rich heated foods, such as French fries and potato crisps. Recent epidemiologic studies on dietary acrylamide intake and cancer risk reported positive associations with endometrial, ovarian, and renal cell cancer risk (2, 3). An advisory group of the IARC recently gave high priority to acrylamide for assessment in future IARC Monograph series (4).

Because the acrylamide molecule is small and hydrophilic, it passively diffuses throughout the whole body, and for this reason, theoretically all tissues are targets for acrylamide carcinogenesis. The incidence of central nervous system tumors was increased upon acrylamide administration in drinking water in one of the two lifetime carcinogenicity studies with rats (5, 6). In this prospective cohort study, we investigated the association between dietary acrylamide intake and human brain cancer risk.

The study took place within the Netherlands Cohort Study on diet and cancer, which started in September 1986 (7). At baseline, the participants (58,279 men and 62,573 women aged 55-69 years) completed a self-administered questionnaire on diet and other possible risk factors for cancer. The case-cohort approach was used for data processing and analysis; cases were enumerated for the entire cohort, whereas the accumulated person-time for the total cohort was estimated from a subcohort of 5,000 men and women randomly sampled from the full cohort at baseline. Cases and subcohort members were excluded from the analysis if they had been diagnosed with cancer (other than skin cancer) at baseline and if their dietary data were incomplete or inconsistent.

Incident cases in the total cohort were detected by record linkage to the Netherlands Cancer Registry. The completeness of cancer follow-up through linkage with these cancer registries was assessed to be at least 96%, whereas the follow-up of the subcohort at the end of the follow-up period was nearly 100% complete. Further details on the design of the study and methods of follow-up are presented elsewhere (7).

The Netherlands Cohort Study on diet and cancer food frequency questionnaire contained questions on habitual consumption of 150 food items during the year preceding the baseline. We used data on acrylamide levels in foods on the Dutch market in order to obtain an intake estimate representative for the Dutch situation. The acrylamide intake was estimated from the mean acrylamide level of foods, and the frequency of consumption and portion size of the foods. For further details of the intake assessment, see ref. (2).

Hazard ratios (HR) were obtained through Cox proportional hazards regression with STATA software (package 9.2). Smoking does not seem to be a clear risk factor for brain cancer in adults (8), but smokers have, on average, three to four times higher levels of acrylamide-hemoglobin adducts (a marker of internal doses of acrylamide) compared with nonsmokers (9). Therefore, subgroup analyses were done for never-smokers. Other analyses were based on histologic subtype: astrocytic glioma (ICD-O morphology codes: M9384, M9400-9401, M9410-9411, M9420-9421, M9424, M9440-9442, and M9481; other case groups were too small for meaningful analysis), and astrocytic glioma with a high malignancy grade (M9400-9401, differentiation grade 3-4; M9410-9411, differentiation grade 3-4; M9420, differentiation grade 3-4; M9424, differentiation grade 3-4; M9440-9442, and M9481; ref. (10)). The HRs for acrylamide intake were adjusted for sex, age, and the a priori selected covariables smoking (status, quantity, and duration), body mass index, height, education level, and energy intake. The other variables from Table 1 were checked for confounding potential based on causing a change of >10% in the HR of acrylamide and incorporated as covariables if they fulfilled this criterion. We considered 100 cases as the minimum number needed for quintile analyses, 60 for tertile analyses, and 20 for analyses with acrylamide as a continuous variable.

After 16.3 years of follow-up, there were 259 cases of primary brain cancer (ICD-O-3: C71), of which 205 were microscopically verified. One hundred and eighty cases were of the astrocytic glioma type, and there were 158 cases with high-grade astrocytic glioma. Table 1 shows the characteristics of the subcohort and the brain cancer cases. Cases consumed more alcohol, were more physically active, smoked more and longer, had a somewhat higher education level and body mass index, and were slightly taller than the subcohort members. There was no effect modification by sex, and therefore the association between acrylamide intake and brain cancer risk is shown for both sexes combined. Acrylamide intake was not associated with brain cancer risk (Table 2), either adjusted only for sex and age or multivariable-adjusted. The same applied to microscopically verified brain cancer, astrocytic glioma, and high-grade astrocytic glioma. Also, in the never-smokers, no association was found. The most important acrylamide-contributing foods in this cohort (Dutch spiced cake and coffee) were by themselves also not associated with brain cancer risk (results not shown).

This prospective cohort study does not give support for an association between dietary acrylamide intake and human brain cancer risk, which was not studied before. Although the route of exposure differs, these results are in line with studies on occupational acrylamide exposure, which did not show excess brain cancer incidence in association with acrylamide exposure (11, 12). The number of cases in the analyses was modest, especially in the subgroups based on histology. According to the formula of Cai and Zeng (13) for case-cohort studies, the power was 80% to detect a statistically significant (significance level of 0.05) association between acrylamide intake (highest versus lowest quintile) and brain cancer risk of 1.85 or higher. The strengths of this prospective study include the absence of recall bias and selection bias, and the fact that only acrylamide data for foods from the Dutch market were used (2). Also, assigning mean acrylamide concentrations to individual foods that are known to have considerable variation in acrylamide concentrations was shown to result in a reasonable rank ordering of the acrylamide intake of the participants in a 24-hour duplicate diet study (2). Despite the potential of nondifferential misclassification of the acrylamide intake [for a discussion of this issue, see Hogervorst et al. (2)], we observed statistically significant positive associations between acrylamide intake and endometrial, ovarian, and renal cell cancer risk in previous analyses in the Netherlands Cohort Study on diet and cancer (2, 3). In conclusion, this prospective study does not give support for an association between dietary acrylamide intake and brain cancer risk.

No potential conflicts of interest were disclosed.