Acrylamide and Glycidamide Hemoglobin Adduct Levels and Breast Cancer Risk in Japanese Women: A Nested Case–Control Study in the JPHC

The International Agency for Research on Cancer classified acrylamide (AA) as a carcinogen (class 2A) in 1994 (1). Various studies, including those carried out with in vitro and animal models, have suggested the genotoxic and non-genotoxic mechanisms of AA carcinogenicity (2, 3). AA is primarily metabolized by an enzyme called CYP2E1 to the epoxide metabolite, glycidamide (GA), which is mutagenic and genotoxic (2, 3). AA has also the potential to cause hormone-related cancer, including breast, ovarian, and endometrial cancers, via its effects on sex hormones (3–5).

AA intake is markedly influenced by the content of sugar, asparagine, and other amino acids in food, cooking temperature, and time; therefore, AA estimated from food frequency questionnaires (FFQ) may not be accurate (6). The human body is also exposed to AA via cigarette smoking, passive smoking (cigarette smoke present in the environment), and occupational exposure (7, 8). Exposure to GA cannot be estimated from FFQs as it is affected by the difference in metabolism in the body (e.g., differences in enzyme activity). Therefore, the association of AA and GA with cancers should be evaluated using biomarkers that are strongly correlated with AA exposure.

Both AA and GA can interact with hemoglobin to N-terminal hemoglobin adducts (HbAA and HbGA, respectively), which are considered appropriate biomarkers of internal exposure and represent exposure from 4 months prior, to the life span of red blood cells (9–11). The degree of metabolism of AA to GA varies from person to person and with AA intake, so the best way to evaluate AA and GA effect seems to assess the sum or ratio of AA and GA together (11). Carcinogenic chemicals are biotransformed into more toxic metabolites by phase I enzymes, such as CYP450, and nontoxic compounds by phase II enzymes, such as GST (12). A similar mechanism may occur in AA, as individuals with variants in the genes for phase I or phase II enzymes are suggested to be at a higher risk of carcinogenesis when exposed to carcinogens. A high HbGA/HbAA ratio indicates the higher activity of the phase I enzyme (CYP450) that metabolizes AA to GA or the lower activity of the GA detoxification enzyme (phase II enzyme), which may determine individual differences in susceptibility to AA. For instance, the nested case–control studies within the European Prospective Investigation into Cancer and Nutrition cohort have investigated the association of AA and GA with ovarian and endometrial cancers using HbAA, HbGA, their sum (HbAA+HbGA), their ratio (HbGA/HbAA) as exposure variables (13, 14).

Many epidemiologic studies have evaluated the association between the dietary intake of AA and breast cancer and there is contradicting evidence associating AA intake with breast cancer (15, 16). Although there is scarce literature reporting an association between HbAA and HbGA and breast cancer, a study showed that HbAA levels in blood samples were positively associated with estrogen receptor (ER)-positive breast cancer after adjustment for smoking behavior (17).

The Japan Public Health Center-based Prospective Study (JPHC Study) is a prospective population-based cohort study that enrolled approximately 140,000 Japanese individuals in the general population and followed these individuals for more than 20 years (18). Although no association was found between AA estimated by an FFQ and breast cancer in this large cohort, the association may be attenuated by measurement error in the FFQ (19). Therefore, the purpose of this study was to evaluate the association between HbAA and HbGA estimated from blood samples and breast cancer in this cohort.

The JPHC Study was a prospective population-based cohort study that investigated the association between lifestyle, dietary habits, and disease incidence. A total of 140,420 registered residents ages 40 to 69 years in 11 public health centers (PHC; cohort I: Iwate, Akita, Nagano, Tokyo, and Okinawa-Chubu; cohort II: Niigata, Ibaraki, Osaka, Kochi, Nagasaki, and Okinawa-Miyako) were enrolled between 1990 and 1993 and were followed up for over 20 years. The details of the study design are presented elsewhere (18). The 5-year follow-up survey response time was employed as the starting point for this study because more comprehensive participant characteristics and lifestyle information were collected in this survey than in the baseline survey.

The selection of the study population for this case–control study within the JPHC cohort is shown in Fig. 1. As erythrocyte samples were used in this study, participants from the area of erythrocyte sample collection (cohort II) were included (n = 78,825). Participants who met the following criteria were excluded: non-Japanese nationality (n = 22), late report of migration occurring before the starting point (n = 98), incorrect birth date (n = 4), and duplicate registration (n = 4). Participants who died (n = 2,385), moved out of the study area (n = 5,516), or were lost to follow-up (n = 140) before the start of the study were also excluded, resulting in 70,656 eligible participants. To assess only female breast cancer, men (n = 33,809) were excluded, resulting in 36,847 women from cohort II being part of the current study.

Participants completed a self-administered 5-year follow-up survey 5 years after study entry (1995–1998), providing information on their lifestyle and dietary habits. The survey was mainly distributed by hand and partly by mail, and incomplete answers were supplemented by telephone interviews. The following information was collected: smoking and drinking habits, current height (whole centimeters), body weight (kilograms), personal and family medical history, and reproductive factors, including menopausal status, age at menopause, and use of exogenous female hormones. The questionnaire covered the medical history of cancer, cardiovascular disease, diabetes mellitus, and other diseases (20).

A total of 30,475 of the 36,847 women completed the 5-year follow-up survey (response rate: 82.7%). Nonrespondents to the 5-year follow-up survey (n = 6,372) tended to be slightly younger than respondents, with a mean age of 56.1 years (SD, 9.2).

During the health checkups conducted at the time of the 5-year follow-up survey, some participants voluntarily provided 10 mL of venous blood. Blood samples were collected in heparin-filled vacutainer tubes; separated into plasma, buffy layers (in cohort I and II), and erythrocytes (in cohort II only); and stored at −80°C until analysis. Among the respondents to the questionnaire in cohort II (n = 30,475), participants who had no erythrocyte sample (n = 20,174) and participants with no blood data (n = 371) were excluded, resulting in 9,930 women (32.6%) with blood samples. The baseline characteristics of the participants without blood samples (n = 20,174) were shown in the Supplementary Table S1. Participants without blood samples were more likely than subjects included in the study (Table 1) to be current and former smokers, premenopausal, and exogenous hormone users were higher.

The follow-up was conducted from the time of response to the 5-year follow-up survey until the designated end of follow-up (December 31, 2012 for Suita; December 31, 2013 for Kochi and Nagasaki; and December 31, 2015 for the other PHC regions). Cancer incidence was verified by active patient notification from the main regional hospital in the study area and data linkage to population-based cancer registries with permission from the local government responsible for cancer registries. Information from death certificates was used as supplemental information. Breast cancer cases were identified using the International Classification of Disease-Oncology, Third Edition (ICD-O-3) code C50. The first cancer type was used in the analysis when a single participant was diagnosed with more than one cancer.

Among 9,930 women whose blood samples were collected, those with a history of breast cancer identified by the 5-year follow-up survey (n = 81) and those who were diagnosed with breast cancer from the baseline survey (n = 2) were excluded. Up to the end of the study period after blood collection, 136 of the 9,847 women who were eligible for the study developed breast cancer. For each case, two cancer-free controls were randomly selected on the date of case identification using the incidence density sampling protocol. The matching criteria were age (±3 years), region (municipal level), a season of blood collection (4 seasons), time of day of blood collection (±3 hours), and menopausal status (premenopause, spontaneous menopause, or nonspontaneous menopause). As a result, cases with no controls (n = 5) and cases with only one control (n = 6) were excluded. The final analysis included 125 cases and 250 controls (Fig. 1).

Analysis of HbAA and HbGA in erythrocytes was performed using the N-alkyl Edman method, which is used as a standard in other epidemiologic studies. The details are presented elsewhere (8). Briefly, fluorescein isothiocyanate isomer-I (FITC) was added to purified protein fractions containing hemoglobin in red blood cells to release FITC-labeled HbAA or HbGA. They were purified by solid-phase extraction and analyzed by LC/MS-MS. The concentrations were corrected at eight amino acid residues (AA-VHLTPEEK) to correct for differences in Edman degradation rates between samples. HbAA and HbGA levels were adjusted for hemoglobin concentration in erythrocytes and measured using the QuantiChrom Hemoglobin Assay Kit (Bioassay Systems) according to the manufacturer's instructions. For participants (n = 2) with HbAA or HbGA values less than the limits of detection (AA-VHLTPEEK, 102.89 nmol/L; GA-VHLTPEEK, 60.46 nmol/L), a value of half the limit of detection was assigned. The percent coefficient of variation was 8.2.

HbAA, HbGA, their sum (HbAA+HbGA), their ratio (HbGA/HbAA), and potential confounders were compared between the cases and controls. Continuous variables are presented as means and SDs, and categorical variables are presented as percentages.

Conditional logistic regression analysis was used to estimate the association between HbAA, HbGA, HbAA+HBGA, and HbGA/HbAA with breast cancer incidence. HbAA, HbGA, HbAA+HBGA, and HbGA/HbAA were categorized into tertiles (low, medium, and high) using the distribution of the control group. ORs and 95% confidence intervals (CI) were estimated using the lowest tertile category as the reference. P_trend was tested by entering the ordinal tertile into the models as a continuous variable. Potential confounders were defined on the basis of the responses obtained from the 5-year follow-up survey. Missing values for each confounder were added to the models to create a missing category. The basic models were conditioned on the matching factors (age, area, season of blood collection, time of day of blood collection, and menopause status) only. Multivariable-adjusted model adjusted for body mass index (BMI; <25, ≥25, or missing), family history of breast cancer (yes or no), age at menarche (≤13, 14, 15, ≥16, or missing), age at first delivery (<26, ≥26, or missing), use of exogenous female hormones (yes, no, or missing), smoking status (current or past, never, or missing), and alcohol intake (<150 and ≥ 150 g/week). These covariates were known or suspected risk factors for breast cancer in the JPHC Study and were adjusted in our previous study (19). A sensitivity analysis was performed excluding cases within 3 years of the 5-year follow-up survey. Stratified analysis by age group (<60 and ≥ 60 years) and BMI (<25 and ≥25) were also performed. In addition, we conducted 20 rounds of multiple imputations by a fully conditional specification for datasets, using the SAS POC MI procedure.

All reported P values were two sided, with a significance level of P < 0.05. All statistical analyses were performed using SAS software (version9.4, SAS Institute, Inc.) and Stata version 16.0 (Stata Corporation).

This study was launched before the enactment of ethical guidelines in Japan, and thus obtaining written informed consent for the collection of blood samples was not mandatory. Participants were asked to provide blood for this study following a written or verbal explanation, and their informed consent was obtained in writing or verbally. In addition, in accordance with the ethical guidelines enacted subsequent to the launch of this study, a research summary was published on the homepage which guarantees participants the opportunity to refuse participation (https://epi.ncc.go.jp/jphc/764/3701.html). This study was approved by the Institutional Review Board in National Cancer Center, Tokyo, Japan (approved no. 2020-101), and the Ethical Review Board of Osaka University, Osaka, Japan (approved no. 22131).

For information on how to apply for access to JPHC data and/or biospecimens, please follow the instructions provided at https://epi.ncc.go.jp/en/jphc/805/8155.html.

In this nested case–control study within the JPHC cohort, 125 breast cancer cases were matched with 250 cancer-free controls. The baseline characteristics of cases and controls were compared (Table 1). There were no differences in the HbAA, HbGA, HbAA+HbGA, or HbGA/HbAA values. The percentages of current and past smokers in the case and control groups were 0.8% and 0.0%, and 2.4% and 0.8%, respectively, highlighting a low percentage overall. Of note, the case group had a lower percentage of smokers than the control group. The case group contained more participants with a family history of breast cancer, earlier age at menarche, later first birth, more hormone users, and more frequent drinkers than the control group.

The ORs and 95% CIs for breast cancer according to AA exposure biomarkers are shown in Table 2. No association was found between HbAA, HbGA, or HbAA+HbGA and breast cancer incidence in either the unadjusted or multivariable-adjusted model. The ORs (95% CI) at the highest tertile compared with the lowest tertile in the multivariable-adjusted model were 1.34 (0.69–2.59), 1.46 (0.79–2.69), and 1.36 (0.72–2.58) for HbAA, HbGA, and HbAA+HbGA, respectively, with point estimates exceeding 1.0, but no significant associations. In contrast, HbGA/HbAA ratio was positively associated with breast cancer incidence. The OR (95% CI) at the highest tertile compared with the lowest tertile in the multivariable-adjusted model was 2.19 (1.11–4.31) with a statistically significant P_trend (P = 0.027). This association was similar in a sensitivity analysis that excluded breast cancer cases that occurred within 3 years of the 5-year follow-up survey (OR _{Highest vs. Lowest} = 2.03; 95% CI = 0.98–4.16; P_trend = 0.074).

The ORs and 95% CIs for breast cancer according to AA exposure biomarkers by age group and BMI are shown in the Supplementary Table S2. No association was found between HbAA, HbGA, or HbAA+HbGA and breast cancer incidence in either stratum. HbGA/HbAA ratio was positively associated with breast cancer incidence in the age group <60 years (OR _{Highest vs. Lowest} = 3.20; 95% CI = 1.13–9.09; P_trend = 0.022) or BMI <25 (OR _{Highest vs. Lowest} = 2.34; 95% CI = 1.02–5.36; P_trend = 0.060), but not in the age group ≥ 60 years (OR _{Highest vs. Lowest} = 1.60; 95% CI = 0.61–4.20; P_trend = 0.398) or BMI ≥ 25 (OR _{Highest vs. Lowest} = 1.90; 95% CI = 0.63–5.76; P_trend = 0.227).

In this nested case–control study within a large prospective population-based cohort, no association was found between HbAA, HbGA, or HbAA+HbGA and breast cancer; however, a positive association was observed between the ratio of HbGA to HbAA and breast cancer. To our knowledge, this is the first study to report an association between HbAA and HbGA levels estimated from blood samples and breast cancer in Asians.

The association of AA with breast cancer has been evaluated in several epidemiologic studies and there is contradicting evidence associating AA intake with breast cancer. A meta-analysis (15) and our prior study based on the JPHC cohort (19) revealed no association between dietary AA intake and breast cancer. However, another meta-analysis (16) reported positive associations of AA intake levels above 20 μg/day in never-smokers with breast cancer. Because dietary AA and GA exposure estimated using FFQs may be inaccurate as an exposure indicator, it is important to evaluate the association of AA and GA with cancers by using biomarkers that are strongly correlated with AA exposure (9, 10). Only one report (17) revealed a lack of association between HbAA levels from blood samples and breast cancer risk, with an estimated incidence rate ratio (95% CI) of 1.05 (0.66–1.69) per 10-fold increase in HbAA level, despite some evidence of a positive association between HbAA and ER-positive breast cancer after adjusting for smoking status, with an estimated incidence rate ratio (95% CI) of 2.7 (1.1–6.6) per 10-fold increase in HbAA level. The null association between HbAA and HbGA levels and breast cancer (not restricted to subtypes) in the current study is consistent with the results of previous studies (15, 19).

Although most participants in this study were nonsmokers, the median (interquartile range) concentrations of HbAA and HbGA adducts were high [66.9 (51.5–86.1) and 47.9 (34.4–72.0) pmol/gHb, respectively]. Olesen and colleagues (17) found a median (5–95 percentile) of 47 (20–209) and 26 (9–99) pmol/gHb for all participants, including 42% nonsmokers and 35 (18–90) and 21 (8–49) pmol/gHb for nonsmokers. Previous epidemiologic studies based on Western populations reported HbAA levels ranging from 19 to 51 pmol/g Hb in nonsmokers and 80 to 194 pmol/g Hb in smokers (8). In general, the estimated intake of AA is lower for Asian populations, including Japanese people, than that for Western populations (6). In addition, our previous validation study in Japan showed that the estimated amount of AA from the FFQ correlated well with HbAA in nonsmokers (8). Although the reasons for the high HbAA and HbGA in the current study are not clear, the population in this study may have had high exposure to AA from environmental sources, such as passive smoking, or used many foods and cooking methods that are high in AA.

In the current study, the HbGA-to-HbAA ratio was positively associated with breast cancer. According to Duale and colleagues (21), the ratio of HbGA to HbAA is a biomarker for AA-related genotoxic exposure as it can correct for variations in both external AA exposure and changes in AA and GA internal metabolism. A high HbGA/HbAA ratio indicates the higher activity of the phase I enzyme (CYP450) that metabolizes AA to GA or the lower activity of the GA detoxification enzyme (phase II enzyme), which may determine individual differences in susceptibility to AA. Carcinogenic chemicals are biotransformed into more toxic metabolites by phase I enzymes, such as CYP450, and nontoxic compounds by phase II enzymes, such as GST (12). A similar mechanism may occur in AA, as individuals with variants in the genes for phase I or phase II enzymes are suggested to be at a higher risk of carcinogenesis when exposed to carcinogens. Although the biological importance of the ratio of HbGA to HbAA is not fully clear, the results of the current study imply that the association of AA with cancer may need to be evaluated, with a focus not only on the absolute amount of HbAA or HbGA but also on the ratio of HbGA to HbAA and activity of metabolic genes.

The impact of AA on breast cancer may vary depending on the expression status of hormone receptors as hormones play an important role in the etiology of hormone receptor–positive breast cancer (22). AA exposure is known to alter sex hormone levels in the body of rats and humans (4, 5) and AA may affect breast cancer by altering sex hormone levels. AA alkylates the amino and sulfhydryl groups of proteins, which may lead to alterations in ER function (17). Olesen and colleagues reported that HbAA is more strongly associated with ER-positive breast cancer and HbGA has a weaker association with breast cancer than HbAA (17) Such finding suggests that AA may induce cancer through non-genotoxic mechanisms, such as hormone-mediated mechanisms, rather than through GA, a metabolite that is more genotoxic. In contrast, Hogervorst and colleagues (4) reported that AA was not associated with hormones, which could explain the increased risk of breast cancer with biological plausibility. In addition, hormone-related cancers, including ovarian and endometrial cancers, are not associated with HbAA levels in blood samples (13, 14, 23). In the current study, it was not possible to divide breast cancers into subtypes because of the small number of cases in which hormone receptor data were available.

In this study, selection bias was minimized as cases and controls were selected from the same cohort and blood samples were collected before cancer incidence, ultimately serving as a strength of the study. Of note, our study had several limitations. First, the results of this study may not be generalizable to the more general population. The blood samples were voluntarily provided by participants who had undergone health checkups conducted by the local government, mainly for those enrolled in the National Health Insurance Program. Other participants either received or did not receive health checkups at their workplaces. In addition, the period of blood sample collection was limited; 99.2% of the blood samples were collected within 1 year (the other 0.8% within one and half years) of the 5-year follow-up survey. Therefore, blood samples were obtained from only 32.6% of the participants. Thus, participants in this nested case–control study may have been healthier than those in the total JPHC cohort. In fact, the blood sample donors in this study had much smaller proportions of current and past smokers than the participants who did not provide blood samples. Second, as hemoglobin adducts were only measured once, the results may not reflect seasonal intraindividual variations or changes over time. The life span of hemoglobin is approximately 4 months, and hemoglobin adducts represent the total exposure to AA over that period. Therefore, future studies using blood samples collected multiple times are needed. Third, the limited number of cases in this study limits statistical power although we used all breast cancer cases for whom blood samples were available in the JPHC Study. The null results in the current study could be due to low statistical power from both the relatively small sample size as well as potential issues with the one-time biomarker measurement not necessarily reflecting long-term and/or usual exposure.

In conclusion, this study showed no association between HbAA and HbGA and breast cancer, but some evidence of a positive association between HbGA/HbAA and breast cancer.

T. Yamaji reports grants from Ministry of Health, Labor, and Welfare of Japan during the conduct of the study. J. Ishihara reports grants from the Food Safety Commission, Cabinet Office, Government of Japan; National Cancer Center, Japan; and Ministry of Health, Labor, and Welfare of Japan during the conduct of the study. No disclosures were reported by the other authors.

N. Narii: Formal analysis, writing–original draft, writing–review and editing. K. Kito: Conceptualization, formal analysis, writing–review and editing. T. Sobue: Conceptualization, supervision, writing–review and editing. L. Zha: Writing–review and editing. T. Kitamura: Writing–review and editing. Y. Matsui: Data curation, writing–review and editing. T. Matsuda: Data curation, writing–review and editing. A. Kotemori: Conceptualization, writing–review and editing. M. Nakadate: Conceptualization, writing–review and editing. M. Iwasaki: Writing–review and editing. M. Inoue: Writing–review and editing. T. Yamaji: Writing–review and editing. S. Tsugane: Data curation, funding acquisition, project administration, writing–review and editing. J. Ishihara: Conceptualization, data curation, funding acquisition, project administration, writing–review and editing. N. Sawada: Conceptualization, data curation, funding acquisition, project administration, writing–review and editing.

This study was supported by the Food Safety Commission, Cabinet Office, Government of Japan (Research Program for Risk Assessment Study on Food Safety, no 1904), National Cancer Center Research and Development Fund (since 2010), and Grant-in-Aid for Cancer Research from the Ministry of Health, Labor, and Welfare of Japan (from 1989 to 2010).

Members of the Japan Public Health Center-based Prospective Study are listed at the following site (as of April 2021): https://epi.ncc.go.jp/en/jphc/781/8896.html. We would like to thank the Akita, Iwate, Nagano, Niigata, Ibaraki, Osaka, Kochi, Nagasaki, and Okinawa Cancer Registries for providing the incidence data.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.