IJC International Journal of Cancer

Transcription

1 IJC International Journal of Cancer Dietary acrylamide and cancer risk: An updated meta-analysis Claudio Pelucchi 1, Cristina Bosetti 1, Carlotta Galeone 1 and Carlo La Vecchia 2 1 Department of, IRCCS Istituto di Ricerche Farmacologiche Mario Negri, Milan, Italy 2 Department of Clinical Sciences and Community Health, University of Milan, Milan, Italy The debate on the potential carcinogenic effect of dietary acrylamide is open. In consideration of the recent findings from large prospective investigations, we conducted an updated meta-analysis on acrylamide intake and the risk of cancer at several sites. Up to July 2014, we identified 32 publications. We performed meta-analyses to calculate the summary relative risk (RR) of each cancer site for the highest versus lowest level of intake and for an increment of 10 mg/day of dietary acrylamide, through fixed-effects or random-effects models, depending on the heterogeneity test. Fourteen cancer sites could be examined. No meaningful associations were found for most cancers considered. The summary RRs for high versus low acrylamide intake were 0.87 for oral and pharyngeal, 1.14 for esophageal, 1.03 for stomach, 0.94 for colorectal, 0.93 for pancreatic, 1.10 for laryngeal, 0.88 for lung, 0.96 for breast, 1.06 for endometrial, 1.12 for ovarian, 1.00 for prostate, 0.93 for bladder and 1.13 for lymphoid malignancies. The RR was of borderline significance only for kidney cancer (RR ; 95% confidence interval, CI, ). All the corresponding continuous estimates ranged between 0.95 and 1.03, and none of them was significant. Among never-smokers, borderline associations with dietary acrylamide emerged for endometrial (RR ; 95% CI, ) and ovarian (RR ; 95% CI, ) cancers. This systematic review and meta-analysis of epidemiological studies indicates that dietary acrylamide is not related to the risk of most common cancers. A modest association for kidney cancer, and for endometrial and ovarian cancers in never smokers only, cannot be excluded. Epidemiological evidence on the relation between dietary acrylamide and the risk of several cancers has continued to accumulate during the last few years, after the publication of our first systematic review and meta-analysis on acrylamide and human cancer. 1 Recently, several data have been released from large cohort studies, 2 11 including among others the European Prospective Investigation into Cancer and Nutrition (EPIC) and the Nurses Health Study (NHS). In particular, the EPIC study, including over 500,000 participants, reported results for esophageal, 2 pancreatic 3 and endometrial 4 cancer. An increased risk of esophageal cancer (on the basis of 341 cases) emerged in subjects with intermediate levels as Key words: acrylamide, epidemiologic studies, meta-analysis, neoplasms, review Abbreviations: CI: confidence interval; ER: estrogen receptor; EPIC: European Prospective Investigation into Cancer and Nutrition; HR: hazard ratio; NCS: Netherlands Cohort Study; NHS: Nurses Health Study; PR: progesterone receptor; RR: relative risk Disclosure: The authors have declared no conflicts of interest. Grant sponsor: Italian Foundation for Cancer Research (FIRC) DOI: /ijc History: Received 31 July 2014; Accepted 5 Nov 2014; Online 18 Nov 2014 Correspondence to: Dr. Claudio Pelucchi, Department of, IRCCS - Istituto di Ricerche Farmacologiche Mario Negri, Via Giuseppe La Masa 19, Milan 20156, Italy, Tel.: 139/ , Fax: 139/ , claudio.pelucchi@ marionegri.it compared to low acrylamide intake, but no significant association was found for a high intake (hazard ratio, HR ; 95% confidence interval, CI, ). 2 No association emerged for pancreatic cancer, based on 865 incident cases, the HRs being 0.77 (95% CI, ) for high versus low intake and 0.95 (95% CI, ) for an increase in acrylamide intake of 10 mg/day. 3 The analysis of endometrial cancer, including 1,382 cases 627 of which were type-i found no overall association with acrylamide (HR ; 95% CI, ), but a positive one for type-i endometrial cancer in the subgroup of women who never smoked and never used oral contraceptives (HR ; 95% CI, ). 4 The NHS considered the relation between dietary acrylamide and breast, endometrial and ovarian cancer. 5 The study included 6,301 cases of breast, 484 of endometrial, and 416 of ovarian cancer. While no association emerged with breast cancer (relative risk, RR ; 95% CI, , for high vs. low intake), the risk was increased for endometrial (RR ; 95% CI, ) and, to a lower extent, ovarian cancer (RR ; 95% CI, ). 5 A subsequent nested case-control analysis including data from the NHS I and II considered the relation between acrylamide exposure, as measured by hemoglobin adducts, and ovarian cancer risk, providing no evidence of association (RR ; 95% CI, for highest vs. lowest tertile of acrylamide adducts). 6 Our earlier extensive systematic review and meta-analysis on acrylamide and cancer included data of both dietary and occupational exposures published until June That

2 Pelucchi et al What s new? Acrylamide is formed in a variety of foods, and some evidence suggests it may cause cancer. How dangerous is dietary acrylamide? This study collated data from 32 earlier projects evaluating the relationship between acrylamide consumed in foods and cancer risk. Of fourteen cancer sites represented, only kidney cancer showed a possible increase in risk associated with dietary acrylamide. When they narrowed the analysis to people who had never smoked, dietary acrylamide appeared to slightly increase risk of endometrial and ovarian cancer as well. meta-analysis found no increase in risk of most types of cancer in relation to acrylamide exposure. Since for several cancer sites the number of cases available from epidemiological studies on dietary acrylamide is now more than doubled, and given the continuing debate on the issue, we updated the quantitative meta-analysis on dietary acrylamide intake and cancer risk. Material and Methods The methods used are similar to those described in our earlier meta-analysis. 1,12 In the current investigation, however, the interest was focused on the estimate of total dietary acrylamide only, because: (i) there are no new relevant data from occupational studies, and (ii) studies considering specific foods/food groups (e.g., fried potatoes and chips, coffee, etc.) are no longer relevant, given the accumulating number of investigations providing estimates of total dietary acrylamide. In July 2014, we performed a systematic literature search in the Medline database, using PubMed, without restrictions according to language, using the following search string: (acrylamide OR glycidamide) AND (cancer OR neoplasm OR tumor) AND (diet OR dietary OR food OR foods). Since the literature search of our earlier meta-analysis was conducted in June 2009, we limited our search to the period A similar search was conducted in EMBASE, using the same keywords and over the same search period. Potentially relevant articles were retrieved and assessed. Those that were not in the scope for this review were excluded (e.g., animal studies, studies aimed at reducing the quantity of acrylamide in specific foods, toxicity studies, studies not focused on cancer risk, etc.). Abstract and unpublished studies were also excluded. None of the selected, relevant articles was published in a language other than English. No studies were excluded a priori for weakness of design or data quality, and we did not assign quality scores to the studies. A total of 105 publications were identified in the PubMed search and 203 in the EMBASE search. A first selection, mainly based on the title and abstract of the publications, was performed to exclude those studies that were clearly nonrelevant (according to the criteria described above), and 59 publications were retained for further consideration. These were abstracted and reviewed in detail. Among them, 46 did not report original epidemiological results on dietary acrylamide and cancer and were no longer considered, whereas the remaining 13 publications were retained for the review. 2 5,7 11,13 16 These were added to the previously selected 19 publications on dietary intake of acrylamide, for a total of 32 publications. We reviewed all the studies selected and abstracted the following information in a standard format: study design; country; cancer sites; number of cases and person-years (or noncases, or controls), overall and according to each level of acrylamide intake; sex; data categorization (i.e., quartiles or quintiles); mean (or median) acrylamide intake in each category among controls; relative risk estimates (rate ratios, HRs, odds ratios, collectively referred to as RR) and the corresponding 95% CI; and additional results from subgroup analyses. A few publications reported duplicate results from the same populations. A population-based case-control study conducted in Sweden provided early results on dietary acrylamide intake and risk of cancers of the colorectum, bladder and kidney. 30 In a subsequent letter, 17 the authors provided updated results based on a more comprehensive estimate of acrylamide intake (i.e., including additional data on coffee drinking). The latter RRs were thus used for the main meta-analyses. Still, data from the first report 30 were used in subgroup analyses on smoking status, as the corresponding results were not available in the subsequent letter. Two publications 13,20 provided results on acrylamide and colorectal cancer using data from the Netherlands Cohort Study (NCS). We included in the metaanalysis the estimates from the first publication that was based on a larger number of cases and a longer follow-up period. 20 Also, two publications from the NCS were focused on breast cancer. 10,19 We included in the meta-analysis the estimates from the second publication, 10 that had an extended follow-up period and a larger number of cases. Two publications reported the RRs and 95% CIs as figures only. 17,18 The exact RR estimates for colorectal, bladder cancer, 17 and for renal cell cancer 18 could be retrieved (Prof. Lorelei Ann Mucci, personal communication) and were used in the meta-analyses, while the RR estimate for kidney cancer 17 and the corresponding 95% CI were no longer available and were thus derived from Figure 2 of Ref. 17 and from Figure 1 of Ref. 18. When at least two independent risk estimates on dietary acrylamide exposure were available for a specific cancer site, we performed a meta-analysis and calculated pooled RR estimates, and the corresponding 95% CIs, for high versus low level of exposure and for an increment of 10 mg/day of dietary acrylamide. For the latter, when the RR for an increment of 10 mg/day of dietary acrylamide was not available from the

3 2914 Dietary acrylamide and cancer original analysis, we estimated it using a method proposed by Greenland and Longnecker, which relates the natural logarithm of the RR to the corresponding mean value of acrylamide intake across exposure categories, 36,37 taking into account that risk estimates for subsequent levels of intake are correlated. When data on number of subjects in each level were not available, we ignored the correlation between estimates and calculated the dose-risk slopes using variance-weighted least squares regression. To obtain the summary RRs for high versus low exposure and, separately, for an increment of 10 mg/day of dietary acrylamide intake, we pooled the corresponding RR estimates according to a fixed-effects model using the inverse variance method (i.e., computing an average effect by weighting the log RR of each study according to the inverse of their sampling variance), when the test for heterogeneity between estimates was not significant (i.e., p > 0.10), or to a randomeffects model, which considers both within- and betweenstudy variations, using the DerSimonian and Laird method [i.e., using the sum of the inverse of the variance of the log (RR) and the moment estimator of the variance between studies as weights], when the test for heterogeneity between estimates was significant (i.e., p 0.10). 38,39 Heterogeneity between estimates was assessed using the v 2 test. For cancer sites with at least four published studies, we also computed summary estimates in subgroups of study design (cohort and case-control studies) and smoking habit (when sufficient data were available), and according to menopausal status and hormone receptor status for breast cancer. Given the small number of studies providing relevant results across separate strata, we did not perform meta-analyses on subsets of risk estimates defined according to sex, body mass index or other potential effect modifiers. All the statistical analyses were performed using the STATA package (version 11; StataCorp, College Station, TX). Results Table 1 shows the main characteristics and results of the 32 publications that provided data on dietary acrylamide and cancer risk. The publications and their results are ordered by cancer site (according to the International Classification of Diseases) and publication year. Several publications reported risk estimates for more than one cancer site and/or subtype, or separately by sex. Therefore, a total number of 64 estimates for high versus low acrylamide intake and 71 estimates for the continuous measure of intake are given in Table 1. Separate RR estimates for various cancer subtypes (i.e., oral cavity and oro and hypopharynx, 16 or specific lymphoid malignancies 7 ) were pooled, in order to obtain a single RR to be included in the meta-analysis of the corresponding cancer site (i.e., oral cavity/pharynx and lymphoid malignancies, respectively). 40 No meta-analysis was conducted for brain and thyroid cancer (only one estimate available). Therefore, a total number of 55 estimates were included in the metaanalyses of dietary acrylamide and risk of cancer at various sites. Table 2 gives, for each cancer site examined, the number of available estimates (at least two) and the corresponding total number of cases, together with the summary RR and 95% CI for both a categorical (i.e., high vs. low level) and a continuous (i.e., for 10 mg/day increase) measure of acrylamide intake. All the RR estimates for high versus low level of intake were close to unity, six were below, seven above unity and one equal to Except for kidney cancer (1,802 cases), that showed a borderline significant increased RR (1.20; 95% CI, ) for high versus low acrylamide intake using a fixed-effects model (p for heterogeneity ), there was no other significant association. When we repeated the metaanalysis of kidney cancer using a random-effects model, the RR was 1.18 (95% CI, ). With further reference to kidney cancer, the continuous RR was 1.02 (95% CI, ), obtained using a random-effects model (p for heterogeneity ). Using a fixed-effects model, the corresponding RR was 1.02 (95% CI, ). Breast cancer had the largest number of estimates (n 5 7) and cases (n 5 16,773). The corresponding RRs were 0.96 (95% CI, ) for high versus low intake and 1.00 (95% CI, ) for an increase in intake of 10 mg/day, with no evidence of heterogeneity between estimates. Six estimates were available for prostate (n 5 13,559 cases) and colorectal (n 5 6,794 cases) cancers. As concerns prostate cancer, both RRs were equal to 1.00, while for colorectal cancer the RRs were 0.94 (95% CI, ) for high versus low intake and 1.00 (95% CI, ) for an increase of 10 mg/day. For other cancer sites, the number of cases on which the meta-analyses were based was lower. The summary RRs for high versus low acrylamide intake were 0.87 for oral and pharyngeal (933 cases), 1.14 for esophageal (1,546 cases), 1.03 for stomach (787 cases), 0.93 for pancreatic (1,732 cases), 1.10 for laryngeal (707 cases), 0.88 for lung (3,598 cases), 1.06 for endometrial (2,774 cases), 1.12 for ovarian (2,010 cases), 0.93 for bladder (1,838 cases) cancer and 1.13 for lymphoid malignancies (1,208 cases). All the corresponding continuous estimates ranged between 0.95 and 1.03, and none of them was significant. Within most cancer sites, estimates were fairly homogeneous. Significant heterogeneity between estimates was reported only for cancers of the oral cavity and pharynx (p for high vs. low intake, p for the continuous measure), lung (p for high vs. low intake, p for the continuous measure), and ovary (p for high vs. low intake), besides kidney (p for the continuous measure). Table 3 gives the pooled RRs and 95% CIs of selected cancer sites (i.e., those for which at least four studies were available) for high versus low dietary acrylamide intake, according to subgroups of study design, smoking status (categorized as never vs. ever smokers or never/former vs. current smokers, according to which information was mainly reported in the studies of each cancer site) and, for breast cancer,

4 Pelucchi et al Table 1. RR and corresponding 95% CI of specific cancer sites, for high versus low level and for an increase of 10 mg/day of dietary acrylamide intake Cancer site; first author, reference, year Study design No. cases Oral cavity and pharynx High vs. low level of intake RR (95% CI) Pelucchi, Case-control ( ) 1.01 ( ) 1 Schouten, (Oral cavity) 2 Schouten, (Oro and hypopharynx) 2 Esophagus 10 mg/day increase in intake RR (95% CI) Additional results Case-cohort ( ) 0.90 ( ) No heterogeneity by sex. Increased continuous HR in female, but not male, nonsmokers (21 cases) Case-cohort ( ) 0.74 ( ) Borderline decreased continuous HR in men. No association in women Pelucchi, Case-control ( ) 0.99 ( ) 1 Hogervorst, 20 Case-cohort ( ) 0.96 ( ) No heterogeneity by smoking, age, physical activity; higher risk for overweight/obese subjects. Comparable results for squamous cell carcinoma and adenocarcinoma Lin, Case-control ( ) 1.04 ( ) 1 Stronger association for esophageal SCC (particularly in nonsmokers) and in overweight/obese subjects Lujan-Barroso, Stomach Hogervorst, 20 Cohort ( ) 1.09 ( ) 1 Similar results for esophageal SCC and adenocarcinoma, as well as for never smokers/20 years quitters. Lower risk estimates for energy-adjusted acrylamide intake. Case-cohort ( ) 1.02 ( ) No heterogeneity by smoking, age, physical activity, BMI. Comparable results for cardia and noncardia cancer Hirvonen, Cohort ( ) 1.01 ( ) 1 Colorectum Mucci, 17,30 Case-control ( ) 0.92 ( ) 1 No heterogeneity by smoking ,4 Mucci, Cohort ( ) 1.02 ( ) 1 No heterogeneity by age, BMI Pelucchi, Case-control ( ) 1.00 ( ) 1 Hogervorst, 20 Case-cohort ( ) 1.00 ( ) No heterogeneity by smoking, age, physical activity, BMI Larsson, Cohort ( ) 1.00 ( ) 1 No heterogeneity by smoking Hirvonen, Cohort ( ) 0.99 ( ) 1 Hogervorst, , men 5 Case-cohort ( ) 1.03 ( ) No heterogeneity by smoking. Increased risk in tumors with activating KRAS mutation Hogervorst, , women 5 Case-cohort ( ) 0.95 ( ) No heterogeneity by smoking. Decreased risk in tumors with truncating APC mutation Pancreas Hogervorst, 20 Case-cohort ( ) 1.06 ( ) No heterogeneity by smoking, age, physical activity; higher risk for overweight/obese subjects Hirvonen, Cohort ( ) 1.01 ( ) 1 Pelucchi, Case-control ( ) 1.01 ( ) No heterogeneity by sex, age, education, BMI, smoking, alcohol drinking and diabetes

5 2916 Dietary acrylamide and cancer Table 1. RR and corresponding 95% CI of specific cancer sites, for high versus low level and for an increase of 10 mg/day of dietary acrylamide intake (Continued) Cancer site; first author, reference, year Study design No. cases Obon-Santacana, Larynx High vs. low level of intake RR (95% CI) 10 mg/day increase in intake RR (95% CI) Additional results Cohort ( ) 0.95 ( ) No heterogeneity by smoking and sex. Inverse association in obese subjects Pelucchi, Case-control ( ) 1.03 ( ) 1 Schouten, Lung Hogervorst, , men Hogervorst, , women Case-cohort ( ) 1.05 ( ) No association in men nor in nonsmokers Case-cohort ( ) 1.03 ( ) No heterogeneity by age, smoking, BMI, diabetes, nonoccupational physical activity, alcohol. Similar results for different histologies Case-cohort ( ) 0.82 ( ) Lower HRs for intermediate levels of nonoccupational physical activity. No heterogeneity by age, smoking, BMI, diabetes, alcohol. Similar results for different histologies Hirvonen, Cohort ( ) 1.02 ( ) 1 Breast Mucci, Cohort ( ) 1.01 ( ) 1 Pelucchi, Case-control ( ) 1.01 ( ) 1 Hogervorst, 19 Case-cohort ( ) 0.99 ( ) No heterogeneity by smoking Larsson, Cohort ( ) 0.98 ( ) 1 No heterogeneity by hormonal receptor status, smoking Wilson, Cohort ( ) 0.98 ( ) 1 No heterogeneity by hormonal receptor status, smoking, age, BMI, alcohol, glycemic index and glycemic load Burley, Cohort ( ) 1.08 ( ) Positive association in premenopausal women. No association in never smokers Pedersen, Case-cohort ( ) 0.97 ( ) No significant association emerged by ER and PR status, and in never smokers, though risks were somewhat increased in never smoking ER1, PR1 and ER1/PR1 cases Wilson, Cohort ( ) 0.99 ( ) 1 Similar results in never smokers and according to ER/PR status. No heterogeneity by menopausal status and BMI Endometrium Hogervorst, 19 Case-cohort ( ) 1.04 ( ) Higher risk in nonsmokers 2007 Larsson, Cohort ( ) 1.01 ( ) 1 No heterogeneity by smoking, menopausal status Wilson, Cohort ( ) 1.15 ( ) 1 Similar results in never smokers. No heterogeneity by menopausal status and BMI, though risks were somewhat higher in those with BMI <25 kg/m 2 Obon-Santacana, Cohort ( ) 0.98 ( ) No heterogeneity by smoking, alcohol, BMI. Decreased risk of type-i endometrial cancer only in OC users, and in those with BMI <25 kg/m 2.Increased risk of type-i endometrial cancer only in never smokers who never used OC, and decreased risk in ever smokers who ever used OC

6 Pelucchi et al Table 1. RR and corresponding 95% CI of specific cancer sites, for high versus low level and for an increase of 10 mg/day of dietary acrylamide intake (Continued) Cancer site; first author, reference, year Study design No. cases Ovary High vs. low level of intake RR (95% CI) 10 mg/day increase in intake RR (95% CI) Additional results Pelucchi, Case-control ( ) 0.99 ( ) 1 Hogervorst, 19 Case-cohort ( ) 1.11 ( ) Higher risk in nonsmokers 2007 Larsson, Cohort ( ) 0.97 ( ) 1 No heterogeneity by smoking, alcohol, HRT, OC Wilson, Cohort ( ) 1.10 ( ) 1 Similar results in never smokers. No heterogeneity by menopausal status and BMI, though risks were somewhat higher in those with BMI <25 kg/m 2. Higher risk for serous cancers Prostate Pelucchi, Case-control ( ) 0.98 ( ) 1 Hogervorst, 21 Case-cohort ( ) 1.01 ( ) No heterogeneity by smoking, alcohol, diabetes, physical activity Larsson, Cohort ( ) 0.98 ( ) 1 No heterogeneity by smoking Wilson, Case-control ( ) 0.99 ( ) No heterogeneity by smoking Hirvonen, Cohort ( ) 1.01 ( ) 1 Wilson, Cohort ( ) 1.01 ( ) 1 Similar results in never smokers, and in those with lethal, advanced, localized, high grade and low grade disease Bladder Mucci, 17, ,4 Case-control ( ) 1.00 ( ) 1 No heterogeneity by smoking Hogervorst, 21 Case-cohort ( ) 1.00 ( ) No heterogeneity by alcohol, diabetes, physical activity. Lower risk in nonsmokers Hirvonen, Cohort ( ) 0.99 ( ) 1 Kidney Mucci, 17,30 Case-control ( ) 0.91 ( ) 1 No heterogeneity by smoking ,4 Mucci, ,4 Case-control ( ) 1.03 ( ) 1 No heterogeneity by smoking Pelucchi, Case-control ( ) 1.03 ( ) 1 No heterogeneity by sex, age Hogervorst, 21 Case-cohort ( ) 1.10 ( ) No heterogeneity by smoking, alcohol, diabetes, physical activity Hirvonen, Cohort ( ) 1.07 ( ) 1 Brain Hogervorst, Thyroid Schouten, Case-cohort ( ) 1.02 ( ) No heterogeneity by smoking. Similar results for astrocytic glioma Case-cohort ( ) 1.03 ( ) No association in women nor in nonsmokers Lymphoid malignancies Hirvonen, (lymphoma) Cohort ( ) 1.02 ( ) 1 Bongers, Case-cohort ( ) 1.14 ( ) Higher risk in never smokers (MM, men) 7

7 2918 Dietary acrylamide and cancer Table 1. RR and corresponding 95% CI of specific cancer sites, for high versus low level and for an increase of 10 mg/day of dietary acrylamide intake (Continued) Cancer site; first author, reference, year Study design No. cases High vs. low level of intake RR (95% CI) 10 mg/day increase in intake RR (95% CI) Additional results Bongers, Case-cohort ( ) 0.92 ( ) No heterogeneity by smoking (MM, women) 7 Bongers, Case-cohort ( ) 1.04 ( ) No heterogeneity by smoking (DLCL, men) 7 Bongers, Case-cohort ( ) 1.02 ( ) No heterogeneity by smoking (DLCL, women) 7 Bongers, Case-cohort 134 NA 0.88 ( ) No heterogeneity by smoking (CLL, men) 7 Bongers, Case-cohort ( ) 0.83 ( ) No heterogeneity by smoking (CLL, women) 7 Bongers, (follicular lymphoma, men) 7 Case-cohort 42 NA 1.28 ( ) Bongers, (follicular lymphoma, women) 7 Case-cohort 47 NA 1.12 ( ) Bongers, Case-cohort 54 NA 1.18 ( ) (WMI, men) 7 Bongers, Case-cohort 35 NA 1.21 ( ) (WMI, women) 7 Bongers, (mantle cell lymphoma, only men) 7 Case-cohort 38 NA 1.06 ( ) Bongers, (T-cell lymphoma, only men) 7 Case-cohort 35 NA 0.94 ( ) BMI: body mass index; CLL: chronic lymphocytic leukemia; DLCL: diffuse large cell lymphoma; ER: estrogen-receptor; HR: hazard ratio; HRT: hormone replacement therapy; MM: multiple myeloma; NA: not available; OC: oral contraceptives; OR: odds ratio; PR: progesterone receptor; SCC: squamous cell carcinoma; WMI: Waldenstrom macroglobulinemia and immunocytoma. 1 Estimated using the method proposed by Greenland and Longnecker When we pooled the estimates from Schouten et al, 2009, the RRs of all oral cavity and pharynx cancers were 0.66 (95% CI, ) for high versus low intake and 0.85 (95% CI, ) for continuous exposure. Both pooled estimates were obtained using a fixed-effect meta-analysis. These estimates were included in the meta-analysis of Table 2. 3 Mean acrylamide intake of open-ended quartiles was estimated using data from other Swedish studies on dietary acrylamide. 4 The exact RR estimates of Ref. 17 for colorectal and bladder cancer and of Ref. 18 for kidney cancer were provided by Prof. Mucci (personal communication), while the other estimates and the 95% CI were no longer available and were thus derived from Figure 2 of Ref. 17 and from Figure 1 of Ref Estimates were not included in the meta-analysis since the data derived from the same prospective investigation of Hogervorst et al., 20, and the latter was based on a larger number of cases and a longer follow-up period. 6 Estimates were not included in the meta-analysis since the data derived from the same prospective investigation of Pedersen et al., , and the latter had an extended follow-up period. 7 When we pooled the available estimates from Bongers et al, 2012, the RRs of all lymphoid malignancies were 1.30 (95% CI, ) for high versus low intake and 1.07 (95% CI, ) for continuous exposure in men, and 0.98 (95% CI, ) for high versus low intake and 0.98 (95% CI, ) for continuous exposure in women. Pooled estimates were obtained using the Hamling method 40 for high versus low exposure and a fixed-effect meta-analysis for continuous exposure. These estimates were included in the meta-analysis of Table 2. menopausal status and hormone receptor status. The pooled RRs for high versus low acrylamide intake from cohort studies only were 1.01 (95% CI, ) for esophageal, 0.96 (95% CI, ) for colorectal, 0.87 (95% CI, ) for pancreatic, 0.96 (95% CI, ) for breast, 1.06 (95% CI, ) for endometrial, 1.20 (95% CI, ) for ovarian, 1.01 (95% CI, ) for prostate and 1.48 (95% CI, ) for kidney cancer. The risk of endometrial (RR ; 95% CI, ) and ovarian cancer (RR ; 95% CI, ) for high acrylamide intake was of borderline significance in the subgroup of never smoking women. For kidney cancer, the RRs for high versus low acrylamide intake were not materially different in never/ former smokers (RR ; 95% CI, ) and in

8 Pelucchi et al Table 2. Results of meta-analyses of epidemiological studies of dietary acrylamide intake and cancer risk. High vs. low level of intake 10 mg/day increase in intake Cancer site No. of estimates Total no. of cases RR (95% CI) 1 p for heterogeneity RR (95% CI) 1 p for heterogeneity Oral cavity and pharynx ( ) ( ) 0.07 Esophagus ( ) ( ) 0.43 Stomach ( ) ( ) 0.86 Colorectum ( ) ( ) 0.14 Pancreas ( ) ( ) 0.28 Larynx ( ) ( ) 0.82 Lung ( ) ( ) 0.04 Breast 7 16, ( ) ( ) 0.33 Endometrium ( ) ( ) 0.17 Ovary ( ) ( ) 0.12 Prostate 6 13, ( ) ( ) 0.74 Bladder ( ) ( ) 0.97 Kidney ( ) ( ) 0.03 Lymphoid malignancies ( ) ( ) 0.37 RR: relative risk; CI: confidence interval 1 RR from random-effects model when p for heterogeneity 0.10, and from fixed-effects model elsewhere. 2 Based on three estimates from two studies, since for one study the RRs were given separately for men and women. current smokers (RR ; 95% CI, ). No difference in risk of breast cancer with high acrylamide intake emerged according to menopausal status nor for different hormone receptor cancer types. Discussion This updated quantitative review of the available evidence on dietary acrylamide and cancer risk confirms previous indications of a lack of association with most cancer sites, as well as of a potential small increase in the risk of kidney cancer, and of endometrial and ovarian cancer in never smoking women, at high levels of acrylamide intake. The findings reported by different studies were generally not heterogeneous, and results were confirmed when only cohort investigation were retained in the analyses. The total number of cancer cases included in this updated meta-analysis was approximately doubled by including the new data released over the last 5 years, mainly from large prospective cohort studies. 2 5,7 9,11 Thus, we were able to examine a higher number of cancer sites, as well as to achieve an increased statistical power with, therefore, narrower CIs for several estimates. This notwithstanding, for some cancer sites the number of studies is still limited. As concerns kidney cancer, we reported a 20% increase in risk in subjects with high as compared to low dietary acrylamide intake, based on data from three case-control and two cohort studies. Considering only the latter study design, the increase was somewhat higher, that is, about 50%. On the other hand, there was no significant association when the analyses were conducted using a continuous measure of exposure (12% in risk for an increase of 10 mg/day of dietary acrylamide). The latter analysis was however affected by relevant heterogeneity between study estimates, ranging from 0.91 in a Swedish case-control investigation 17,30 to 1.10 in the NCS. 21 This may be at least in part explained by different results according to study design with, again, cohort studies reporting moderately higher risk estimates. 9,21 Smoking is a major risk factor for kidney cancer 41 and a major source of acrylamide exposure, too. 42 Thus, smoking might have a confounding or modifying effect on the relation between dietary acrylamide and kidney cancer. All the five studies included in the meta-analysis of kidney cancer, however, controlled their risk estimates for some measure of tobacco smoking, and when we performed subgroup meta-analyses according to smoking status, we found similar RRs among never/former and current smokers. Limited indications for a potential role of acrylamide exposure on cancer of the kidney were also provided from two occupational cohort studies. 43,44 Given the modest, borderline significant, association and the still limited amount of epidemiological data on kidney cancer (i.e., about 1,800 cases), the issue remains therefore open to discussion. The main pathway to carcinogenesis of acrylamide is through its oxidization to glycidamide, a chemically reactive genotoxic metabolite. 45 Besides damaging and mutagenic effects on DNA, the existence of other modes of action of acrylamide is supported by the observation of a tissuespecific cancerogenicity in both mice and rats. 46 Acrylamide

9 2920 Dietary acrylamide and cancer Table 3. Meta-analysis of high versus low dietary acrylamide intake in relation to the risk of selected cancers, in subgroups of study design and selected covariates. Cancer site/subgroup No. of studies RR (95% CI) 1 Esophagus Cohort studies ( ) Case-control studies ( ) Never/former smokers ( ) Colorectum Cohort studies ( ) Case-control studies ( ) Never/former smokers ( ) Current smokers ( ) Pancreas Cohort studies ( ) Case-control studies ( ) Breast Cohort studies ( ) Case-control studies ( ) Never smokers ( ) Ever smokers ( ) Premenopausal women ( ) Postmenopausal women ( ) ER1PR1 cancer ( ) ER1PR- cancer ( ) ER-PR1 cancer ( ) ER-PR- cancer ( ) Endometrium 3 Never smokers ( ) Ever smokers ( ) Ovary Cohort studies ( ) 4 Case-control studies ( ) Never smokers ( ) Ever smokers ( ) Prostate Cohort studies ( ) Case-control studies ( ) Never smokers ( ) Kidney Cohort studies ( ) Case-control studies ( ) Never/former smokers ( ) Current smokers ( ) CI: confidence interval; ER: estrogen-receptor; PR: progesterone receptor; RR: relative risk 1 RR from random-effects model when p for heterogeneity 0.10, and from fixed-effects model elsewhere. 2 Including case-cohort designs. 3 Analyses by study design are not presented since all the studies were cohorts. 4 p for heterogeneity within subgroup may also exert a carcinogenic role on selected body sites by affecting hormonal balances. 47 This, together with the findings of animal studies that reported an increased occurrence of mammary gland tumors in rats which were given acrylamide through drinking water, 48,49 explains the wide interest in investigating the relation with breast, but also endometrial and ovarian, cancer in epidemiological studies. Our meta-analysis, however, did not report an association with breast cancer risk. In particular, the summary RRs of dietary acrylamide and breast cancer were below unity for high versus low intake, as well as in subgroups of prospective studies, of both never and ever smokers, and of both estrogen receptor positive (ER1)/progesterone receptor positive (PR1) and ER2PR2 breast cancers. Two other publications (both from the Danish, Diet, Cancer and Health study) 42,50 considered acrylamide exposure in relation to breast cancer, but were not included in the present meta-analysis, because: (i) the first investigation considered biomarkers of exposure, that is, acrylamide-hemoglobin (Hb) and glycidamide-hb adducts, but not dietary intake. 42 This study found a positive association between acrylamide-hb level and ER1 (RR 5 2.7; 95% CI, , for a 10-fold increase in adduct concentrations) but not ER- breast cancer, the overall RR being 1.9 (95% CI, ). Results for ER1 breast cancer were lower in nonsmokers (RR 5 1.9) than in smokers (RR 5 4.9), and a significant increase in breast cancer risk emerged only after careful adjustment for smoking habits, including amount, duration and past smoking. Smokers had over threefold higher levels of acrylamide adducts than nonsmokers, in both cases and controls. The lack of matching of cases and controls on smoking status which had an important role in that study makes it, therefore, difficult to interpret the results; (ii) the second investigation analyzed the role of prediagnostic biomarkers of acrylamide exposure on survival after a diagnosis of breast cancer. 50 Breast cancer specific mortality increased with increasing adduct levels, particularly in nonsmoking women with ER1 cancer (HR , 95% CI, , for a 25 pmol/g globin increase in acrylamide- Hb; HR , 95% CI, , for a 25 pmol/g globin increase in glycidamide-hb level). Thus, the results on biomarkers of acrylamide exposure of the Danish, Diet, Cancer and Health study 42,50 indicated significant increases of both ER1 breast cancer risk and mortality, that should be further investigated. This meta-analysis of data on dietary acrylamide, on the other hand, did not find associations with ER1PR1 nor with ER1PR2 breast cancer risk. The role of acrylamide on endometrial and ovarian cancers has been the object of a recent debate In this updated review, we were able to include additional data from the NHS and EPIC (for endometrial cancer only) studies. 4,5 Still, summary results were close to those of our previous report confirming, overall, an absence of association, but a moderate increase in risk in the subgroup of never smoking women. For ovarian cancer, some caution is needed in the interpretation of results, given the unexplained heterogeneity

10 Pelucchi et al between estimates. Another study examined data of the NHS and NHS II on ovarian cancer risk according to Hb adduct levels, and was thus not included in the meta-analysis on dietary acrylamide. 6 This reported no association in the complete dataset (RR ; 95% CI, for highest versus lowest tertile of combined acrylamide-hb and glycidamide- Hb adducts) nor in the subgroup of nonsmoking women (RR ; 95% CI, ). Overall, epidemiological studies published to date do not support an association between dietary acrylamide intake and major female hormone-related cancers. Associations in specific subgroups, particularly among never smokers for endometrial and ovarian cancers, cannot be excluded, 4 and need further investigation. With reference to other cancer sites, a considerable amount of data is now available on the role of dietary acrylamide on colorectal and prostate cancers. Overall results do not indicate any excess risk for these cancer sites. For colorectum, however, a recent study reported for the first time differential results for acrylamide and colorectal cancer with specific somatic mutations, with a more than twofold increased risk of cancers with an activating Kirsten-ras (KRAS) mutation in men and a halved risk of cancers with a truncating adenomatous polyposis coli (APC) mutation in women with high versus low acrylamide intake. 13 To date, specific information on the molecular characteristics of colorectal tumors in relation to acrylamide intake is limited to this investigation. For prostate cancer, one of the studies included in the meta-analysis provided further results for acrylamide-hb adduct levels, besides dietary intake, reporting no association. 34 As concerns pancreatic cancer, a possible link with acrylamide exposure has been suggested by occupational cohort studies. 1,43,44 On the other hand, none of the four epidemiological studies on dietary intake of acrylamide reported any association, and the summary RR of dietary References 1. Pelucchi C, La Vecchia C, Bosetti C, et al. Exposure to acrylamide and human cancer a review and meta-analysis of epidemiologic studies. Ann Oncol 2011;22: Lujan-Barroso L, Gonzalez CA, Slimani N, et al. Dietary intake of acrylamide and esophageal cancer risk in the European Prospective Investigation into Cancer and Nutrition cohort. Cancer Causes Control 2014;25: Obon-Santacana M, Slimani N, Lujan-Barroso L, et al. Dietary intake of acrylamide and pancreatic cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. Ann Oncol 2013;24: Obon-Santacana M, Kaaks R, Slimani N, et al. Dietary intake of acrylamide and endometrial cancer risk in the European Prospective Investigation into Cancer and Nutrition cohort. Br J Cancer 2014;111: Wilson KM, Mucci LA, Rosner BA, Willett WC. A prospective study on dietary acrylamide intake and the risk for breast, endometrial, and ovarian cancers. Cancer Epidemiol Biomarkers Prev 2010; 19: Xie J, Terry KL, Poole EM, et al. Acrylamide hemoglobin adduct levels and ovarian cancer risk: a nested case-control study. Cancer Epidemiol Biomarkers Prev 2013;22: Bongers ML, Hogervorst JG, Schouten LJ, et al. Dietary acrylamide intake and the risk of lymphatic malignancies: the Netherlands Cohort Study on diet and cancer. PLoS One 2012;7: e Burley VJ, Greenwood DC, Hepworth SJ, et al. Dietary acrylamide intake and risk of breast cancer in the UK women s cohort. Br J Cancer 2010; 103: Hirvonen T, Kontto J, Jestoi M, et al. Dietary acrylamide intake and the risk of cancer among Finnish male smokers. Cancer Causes Control 2010;21: Pedersen GS, Hogervorst JG, Schouten LJ, et al. Dietary acrylamide intake and estrogen and progesterone receptor-defined postmenopausal breast studies was below unity. While no relation emerged in the meta-analysis of lymphoid malignancies as a whole, a positive association was found, in the only study that considered different types of lymphoid malignancies separately, 7 with multiple myeloma and with follicular lymphoma in men. For other cancer sites as well, associations were null or weak, although in some cases based on limited data and/or on heterogeneous risk estimates (such as for lung cancer, where heterogeneity was mainly due to different results in men and women in one study 22 ). Epidemiological studies on dietary acrylamide have been criticized for their alleged inadequacy to address its relation with cancer risk Besides general limitations of cohort and case-control studies, we acknowledge several critical aspects related to this specific topic, including the difficulties in estimating dietary acrylamide intake through food frequency questionnaires and databases of acrylamide content in foods, the lack of repeated exposure estimations over an adequately long time-period, and the lack of statistical power to detect small increases in risk. However, epidemiological studies also have several strengths. In particular, they allow to directly address the relation between acrylamide exposure and cancer risk in humans, avoiding the uncertainties deriving from the use of animal data and mathematical models, and to assess the public health relevance of such a relation. 57 During the last decades, epidemiological studies identified several dietary and nutritional factors associated with the risk of various cancers, particularly those of the digestive tract. 58 This weighs in favor of the capability of epidemiological studies to assess the association with dietary acrylamide. Further, the meta-analytic approach in the absence of relevant heterogeneity, as was the case for most of the analyses presented here increases to a great extent the statistical power of the investigation, by joining together the results of several studies. cancer risk. Breast Cancer Res Treat 2010;122: Wilson KM, Giovannucci E, Stampfer MJ, Mucci LA. Dietary acrylamide and risk of prostate cancer. Int J Cancer 2012;131: Stroup DF, Berlin JA, Morton SC, et al. Metaanalysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in (MOOSE) group. JAMA 2000;283: Hogervorst JG, de Bruijn-Geraets D, Schouten LJ, et al. Dietary acrylamide intake and the risk of colorectal cancer with specific mutations in KRAS and APC. Carcinogenesis 2014;35: Lin Y, Lagergren J, Lu Y. Dietary acrylamide intake and risk of esophageal cancer in a population-based case-control study in Sweden. Int J Cancer 2011;128: Pelucchi C, Galeone C, Talamini R, et al. Dietary acrylamide and pancreatic cancer risk in an Italian case-control study. Ann Oncol 2011;22:

11 2922 Dietary acrylamide and cancer 16. Schouten LJ, Hogervorst JG, Konings EJ, et al. Dietary acrylamide intake and the risk of headneck and thyroid cancers: results from the Netherlands Cohort Study. Am J Epidemiol 2009;170: Mucci LA, Dickman PW, Steineck G, et al. Reply: Dietary acrylamide and cancer risk: additional data on coffee. Br J Cancer 2003;89: Mucci LA, Lindblad P, Steineck G, Adami HO. Dietary acrylamide and risk of renal cell cancer. Int J Cancer 2004;109: Hogervorst JG, Schouten LJ, Konings EJ, et al. A prospective study of dietary acrylamide intake and the risk of endometrial, ovarian, and breast cancer. Cancer Epidemiol Biomarkers Prev 2007; 16: Hogervorst JG, Schouten LJ, Konings EJ, et al. Dietary acrylamide intake is not associated with gastrointestinal cancer risk. J Nutr ;138: Hogervorst JG, Schouten LJ, Konings EJ, et al. Dietary acrylamide intake and the risk of renal cell, bladder, and prostate cancer. Am J Clin Nutr ;87: Hogervorst JG, Schouten LJ, Konings EJ, et al. Lung cancer risk in relation to dietary acrylamide intake. J Natl Cancer Inst 2009;101: Hogervorst JG, Schouten LJ, Konings EJ, et al. Dietary acrylamide intake and brain cancer risk. Cancer Epidemiol Biomarkers Prev 2009;18: Larsson SC, Akesson A, Bergkvist L, Wolk A. Dietary acrylamide intake and risk of colorectal cancer in a prospective cohort of men. Eur J Cancer 2009;45: Larsson SC, Akesson A, Wolk A. Long-term dietary acrylamide intake and risk of epithelial ovarian cancer in a prospective cohort of Swedish women. Cancer Epidemiol Biomarkers Prev 2009; 18: Larsson SC, Akesson A, Wolk A. Long-term dietary acrylamide intake and breast cancer risk in a prospective cohort of Swedish women. Am J Epidemiol 2009;169: Larsson SC, Akesson A, Wolk A. Dietary acrylamide intake and prostate cancer risk in a prospective cohort of Swedish men. Cancer Epidemiol Biomarkers Prev 2009;18: Larsson SC, Hakansson N, Akesson A, Wolk A. Long-term dietary acrylamide intake and risk of endometrial cancer in a prospective cohort of Swedish women. Int J Cancer 2009;124: Mucci LA, Adami HO, Wolk A. Prospective study of dietary acrylamide and risk of colorectal cancer among women. Int J Cancer 2006;118: Mucci LA, Dickman PW, Steineck G, et al. Dietary acrylamide and cancer of the large bowel, kidney, and bladder: absence of an association in a population-based study in Sweden. Br J Cancer 2003;88: Mucci LA, Sandin S, Balter K, et al. Acrylamide intake and breast cancer risk in Swedish women. JAMA 2005;293: Pelucchi C, Galeone C, Dal Maso L, et al. Dietary acrylamide and renal cell cancer. Int J Cancer 2007;120: Pelucchi C, Galeone C, Levi F, et al. Dietary acrylamide and human cancer. Int J Cancer 2006; 118: Wilson KM, Balter K, Adami HO, et al. Acrylamide exposure measured by food frequency questionnaire and hemoglobin adduct levels and prostate cancer risk in the Cancer of the Prostate in Sweden Study. Int J Cancer 2009;124: Wilson KM, Mucci LA, Cho E, et al. Dietary acrylamide intake and risk of premenopausal breast cancer. Am J Epidemiol 2009;169: Greenland S, Longnecker MP. Methods for trend estimation from summarized dose-response data, with applications to meta-analysis. Am J Epidemiol 1992;135: Orsini N, Bellocco R, Greenland S. Generalized least squares for trend estimation of summarized dose-response data. Stata J 2006;6: DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7: Greenland S. Quantitative methods in the review of epidemiologic literature. Epidemiol Rev 1987;9: Hamling J, Lee P, Weitkunat R, Ambuhl M. Facilitating meta-analyses by deriving relative effect and precision estimates for alternative comparisons from a set of estimates presented by exposure level or disease category. Stat Med ; 27: McLaughlin JK, Lipworth L, Tarone RE, Blot WJ. Renal cancer. In: Schottenfeld D, Fraumeni JF. Cancer epidemiology and prevention, 3rd ed. New York: Oxford University Press, Olesen PT, Olsen A, Frandsen H, et al. Acrylamide exposure and incidence of breast cancer among postmenopausal women in the Danish Diet, Cancer and Health Study. Int J Cancer ;122: Marsh GM, Youk AO, Buchanich JM, et al. Mortality patterns among workers exposed to acrylamide: updated follow up. J Occup Environ Med 2007;49: Swaen GM, Haidar S, Burns CJ, et al. Mortality study update of acrylamide workers. Occup Environ Med 2007;64: International Agency for Research on Cancer. Some industrial chemicals. Vol. 60. Lyon: International Agency for Research on Cancer, Besaratinia A, Pfeifer GP. A review of mechanisms of acrylamide carcinogenicity. Carcinogenesis 2007;28: Hogervorst JG, Baars BJ, Schouten LJ, et al. The carcinogenicity of dietary acrylamide intake: a comparative discussion of epidemiological and experimental animal research. Crit Rev Toxicol 2010;40: Johnson KA, Gorzinski SJ, Bodner KM, et al. Chronic toxicity and oncogenicity study on acrylamide incorporated in the drinking water of Fischer 344 rats. Toxicol Appl Pharmacol 1986;85: Rice JM. The carcinogenicity of acrylamide. Mutat Res 2005;580: Olsen A, Christensen J, Outzen M, et al. Prediagnostic acrylamide exposure and survival after breast cancer among postmenopausal Danish women. Toxicology 2012;296: Lipworth L, Sonderman JS, Tarone RE, McLaughlin JK. Review of epidemiologic studies of dietary acrylamide intake and the risk of cancer. Eur J Cancer Prev 2012;21: Hogervorst J, Duell E, Schouten L, et al. Reaction on the acrylamide and cancer review by Lipworth and colleagues. Eur J Cancer Prev 2013;22: Lipworth L, Sonderman JS, Tarone RE, McLaughlin JK. Acrylamide: a human cancer risk? Eur J Cancer Prev 2013;22: Granath F, Tornqvist M. Who knows whether acrylamide in food is hazardous to humans? J Natl Cancer Inst 2003;95: Hagmar L, Tornqvist M. Inconclusive results from an epidemiological study on dietary acrylamide and cancer. Br J Cancer 2003;89:774 5; author reply Virk-Baker MK, Nagy TR, Barnes S, Groopman J. Dietary acrylamide and human cancer: a systematic review of literature. Nutr Cancer 2014;66: Mucci LA, Wilson KM. Acrylamide intake through diet and human cancer risk. J Agric Food Chem ;56: American Institute for Cancer Research. Food, nutrition, physical activity, and the prevention of cancer: a global perspective. Washington, DC: American Institute for Cancer Research, 2007.