{"paper_id":"a8e32bbd-2d91-43f3-989e-07514acf409a","body_text":"Differentiated thyroid carcinoma (DTC) accounts for about 90% of all thyroid cancers. The incidence of thyroid cancer is characterized by 4–5 times higher rates in women than in men and considerable ethnic and geographic variation [ 1 , 2 ]. Particularly high incidence rates were observed in Melanesian women of New Caledonia, South Pacific (71.4/100 000 person-years (py) in 1995–1999 [ 1 – 3 ], contrasting with rates ranging from 2 to 8/100 000 py among women in other high income countries [ 2 ]). During the last decades, the incidence of DTC has increased regularly in most high-resource countries. Part of this increase has been attributed mainly to changes in medical screening practices that enhance the detection of small dormant microcarcinomas [ 4 – 6 ], but changes in environmental or lifestyle factors may also contribute to this increase [ 5 ].\nApart from exposure to ionizing radiation during childhood, a well-established risk factor for DTC, increased DTC risk has also been associated with overweight, iodine deficiency, high parity, and late age at menarche. By contrast, several studies have reported that the risk of DTC decreased with alcohol or tobacco consumption [ 7 – 11 ]. Thyroid cancer is one of the cancers with the highest familial risk [ 12 ], suggesting a role of genetic risk factors. However, only a few variants have been identified thus far [ 13 ] and, to the best of our knowledge, no gene-environment interactions have been convincingly reported for DTC.\nGlutathione S-transferase (GST) genes, such as  GSTM1  and  GSTT1 , encode for phase II enzymes, and are involved in the steroid hormone metabolism and in the detoxification of various xenobiotics and products of oxidative stress. The enzyme activity depends on the number of copies of  GSTM1  and  GSTT1  (copy number variation, CNV) in the genome [ 14 , 15 ], with complete gene deletion resulting in a loss of function. It can thus be hypothesized that gene deletion confers higher vulnerability to carcinogen exposure [ 16 ].\nIn a meta-analysis of 12 studies conducted in countries from Europe, Asia and South America that investigated the association between DTC and  GSTM1  or  GSTT1  genotypes, significant heterogeneity between studies was reported [ 17 ]. Heterogeneity may be explained by unmeasured exposures in endogenous or exogenous risk factors that modify the association of DTC with GST genotypes.\nIn the present paper, we used data collected in studies on DTC conducted in populations of European origin in France and in populations of European and Melanesian origins form New Caledonia, to investigate the role of G STM1  and  GSTT1  genotypes in DTC risk, and their interaction with suspected risk factors of DTC that could be modulated by these genes such as hormonal factors, cigarette smoking and alcohol drinking, and obesity.\n\nWe used data from two case-control studies on thyroid cancer conducted in metropolitan France (CATHY study) and in New Caledonia (NC study). All participants provided signed informed consent. The study was approved by the review board of the French institute of health and medical research (INSERM) and authorized by the French data protection authority (CNIL).\nThe CATHY study [ 18 ] is a population-based case-control study performed in three French “ départements ” (Marne, Ardennes, and Calvados) covered by a cancer registry. Cases were patients living in these areas aged 25 years and older diagnosed with DTC between 2002 and 2007. Controls were selected at random using the telephone directory and unlisted phone numbers of all private homes in the study areas, and were frequency-matched to the cases by 5-year age group and study area. To prevent possible selection bias arising from differential participation rates across categories of socioeconomic status (SES), the control group was selected to reflect the distribution by SES categories of the general population, as described in details previously [ 18 ]. From 621 cases and 706 controls recruited for the study, saliva DNA samples (Oragen®) were obtained for 482 cases and 565 controls of self-declared European ancestry.\nThe NC study is a country-wide, population-based, case-control study [ 3 , 7 , 19 ]. The cases included patients with DTC diagnosed between 1993 and 1999 who had been living in NC for at least 5 years at the time of diagnosis. The cases were identified from the two pathology laboratories in NC and from active searches in the medical records of the main hospitals. Age- and sex-matched controls were randomly selected from recently updated electoral rolls. A total of 332 cases and 412 controls were included; 42 cases and 133 controls self-declared Europeans and 206 cases and 156 controls self-declared Melanesians. Saliva DNA samples (Oragen®) were available for 284 Melanesians (164 cases and 120 controls) and 108 Europeans (27 cases and 81 controls).\nIn the two studies, information on ethnicity, personal and familial history of thyroid disease, gynecological and reproductive history, anthropometric factors, diet, alcohol intake, tobacco smoking, and residential and occupational histories was collected during in-person interviews by trained interviewers. BMI was defined as weight (in kilograms) divided by height (in meters) squared. Alcohol drinking was assessed as the lifetime average number of glasses per week using information from the questionnaire for each type of beverage separately (beer, wine, aperitif, and liqueur). Because the ethanol content is approximately the same for an ordinary glass of any alcoholic beverage, the total number of drinks per week was used as an indicator of total alcohol intake. Pack-years of cigarette smoking were calculated from the total number of years of tobacco smoking and from the number of cigarettes smoked per day for each smoking period.\nWomen were considered to be users of oral contraceptives if they had ever taken pills for at least 6 months. Women were considered postmenopausal if they reported no menstruation for at least 1 year or used menopausal hormone therapy (MHT) (natural menopause), or if they had bilateral ovariectomy (artificial menopause). Women with unknown menopausal status, because of hysterectomy before cessation of menstruations or unknown date of last menstruation, were considered postmenopausal if they were 50 years old or more (the median age at menopause in women with natural menopause).\nAssociation between DTC risk and reproductive factors, BMI, tobacco smoking and alcohol drinking were analyzed in detail previously in the NC [ 7 ,  19 ] and CATHY studies [ 18 ]. In the present paper, the analyses were performed in the subset of individuals with genotype data from these two studies.\nDNA extracted from saliva samples was used to determine the copy number variation (CNV) of  GSTM1  and  GSTT1  using TaqMan gene copy number detection designed by Applied Biosystems. The genotyping was processed by Integragen (Evry, France). Real-time PCR was run on an Applied Biosystems 7900HT Fast system with gene-specific primers for  GSTM1  probes (Hs02575461_cn) and  GSTT1  probes (Hs00010004_cn), along with primer for the RNase P gene as a reference. Each sample was run in triplicate using 50 ng of genomic DNA. The gene-specific primers were validated in 90 CEPH individuals and genotyping carried out blindly as to case-control status. CopyCaller software V1 (Applied Biosystems) was used to quantify the number of copies in each sample.\nThirty-one subjects (5 cases and 26 controls) in the CATHY study and 14 subjects (8 cases and 6 controls) in the NC study had missing genotypes for both  GSTM1  and  GSTT1  and were excluded from the analysis. Therefore, a total 1124 Europeans (504 cases and 620 controls) and 270 Melanesians (156 cases and 114 controls) were included in the analyses.\nOdds ratios (ORs) were calculated for  GSTM1  and  GSTT1  genotypes using unconditional logistic regression, comparing categories defined by the gene copy number coded as a categorical variable (0, 1, or ≥2 copies) and adjusting for age (5-year intervals), sex, and area of residence. We also compared non-carriers (null genotype) to carriers of at least one copy of the gene (non-null genotype). We first analyzed Europeans and Melanesians separately and tested the heterogeneity of the ORs using the Cochran’s Q test or likelihood ratio test. As no heterogeneity was detected, the combined dataset was used in further analyses (additionally adjusted for ethnic group).\nWe tested the interaction between  GSTM1  and  GSTT1  genotypes and tobacco smoking, alcohol intake, BMI, and hormonal and reproductive factors using the likelihood ratio test, comparing models with and without the interaction term. Tobacco smoking, alcohol intake, and BMI were analyzed in men and women together, whereas hormonal and reproductive factors were analyzed in women only. To account for the potential confounding effect of hormonal and reproductive factors, we also performed the analysis for tobacco smoking, alcohol drinking, and BMI in women separately. The results were similar to those obtained for men and women combined and are not shown.\n\nThe characteristics of the study sample are presented in  Table 1  for Europeans and Melanesians separately. Cases and controls had similar sex and age distributions in each group. Among cases, the proportion of papillary thyroid cancers was similar in the two groups (~85%).\nObesity (BMI ≥ 30 kg/m2) and high parity were more prevalent in Melanesians than in Europeans. Conversely, alcohol drinking and use of oral contraceptives were less common in Melanesians ( Table 2 ). However odds ratios calculated in Europeans and Melanesians were comparable and none of the p-values for heterogeneity was significant. In the analysis of the two groups combined, thyroid cancer risk was increased in obese subjects and decreased with alcohol consumption. No association with tobacco smoking was observed. Among women, late age at menarche was associated with a non-significant increased risk of DTC.\na  The multivariate regression model includes age, sex, area of residence, ethnic group, BMI, alcohol consumption, and cigarette smoking.\nb  The multivariate regression model includes age, area of residence, ethnic group, BMI, alcohol consumption, cigarette smoking, age at menarche, number of full-term pregnancies, and oral contraception.\nc  The sum does not add up to the total (504 cases/620 controls in European, 156 cases/114 controls in Melanesian) because of missing values for some variables in the multivariate analysis.\nd  Test for heterogeneity of odds ratios across ethnic groups\nOR: odds ratio, CI: confidence interval\nThe proportion of control subjects with a ‘null’ genotype was much higher in Melanesians than in Europeans (82.46% vs 48.06% for GSTM1; 34.21% vs 18.71% for GSTT1; 31.58% vs 8.55% for deletion of both genes, in Melanesians and Europeans, respectively) ( Table 3 ). Odds ratios for DTC associated with null as compared with non-null genotypes did not significantly depart from unity in Europeans or in Melanesians, or in both groups combined. No heterogeneity between groups of subjects was detected ( Table 3 ).\nAll OR are adjusted for age, sex, area of residence, and ethnic group where appropriate.\nOR: Odds Ratio, CI: confidence interval. Het: test for heterogeneity of odds ratios across ethnic groups\nIn  Table 4 , we calculated ORs for BMI, alcohol drinking, and tobacco smoking after stratification of study subjects by  GSTM1  and  GSTT1  genotypes, using a single dataset combining Europeans and Melanesians as we did not detect heterogeneity between the analyses conducted for each group separately (not shown). The association with obesity was greater among individuals with  GSTM1 - and/or  GSTT1 -null genotypes than individuals with non-null genotypes. The OR in obese individuals with null genotype for both genes was 5.75 while it was 1.26 in those with at least one copy of the genes (p-interaction = 0.02). While comparing individuals with null genotype for both genes and BMI≥30 kg/m 2  to individuals with non-null genotype and BMI<25 kg/m 2 , the OR was 2.69 (p = 0.01) ( S1 Table ). The inverse association with alcohol intake was particularly noticeable in subjects with  GSTT1 -null genotypes (OR ≥1 glass/week 0.33; p-interaction 0.01) and in those with deletion of both genes (OR≥1 glass/week 0.21; p interaction 0.12). The OR comparing individuals drinking more than one glass per week with non-null genotypes to never drinkers with null genotypes for both genes was 0.36 (p = 0.01) ( S1 Table ). The association between tobacco smoking and DTC did not differ according to the GSTT1 or GSTM1 genotypes.\na  The multivariate regression model includes age, sex, area of residence, ethnic group, BMI, alcohol consumption, and cigarette smoking.\nb  Null genotype refers to individuals with null genotypes for both genes, while non-null genotype refers to subjects with at least one copy of  GSTM1  or  GSTT1\nc The number of subjects do not add up to the total because of missing values for some variables\nCa/Co: Cases/Controls, OR: Odds ratio, 95% CI: 95% confident Interval,  p-int : p-value of interaction\nThe analyses for hormonal and reproductive factors in women stratified by genotypes are shown in  Table 5 . No significant interaction with GSTM1 was detected for age at menarche, parity, oral contraceptive, or menopausal status. Late age at menarche was positively associated with DTC in individuals with GSTT1 non-null genotype for but not in those with GSTT1-null genotype (p interaction 0.03).\na  The multivariate regression model included age, sex, area of residence, ethnic group, BMI, alcohol consumption, cigarette smoking, age at menarche, number of full-term pregnancies, and oral contraception.\nb  Null genotype refers to individuals with null genotypes for both genes, whereas non-null genotype refers to subjects with at least one copy of  GSTM1  or  GSTT1 .\nc  Test for the interaction between the corresponding variables and the genotypes\nCa/Co: cases/controls, OR: odds ratio, CI: confidence interval\n\nThis study was conducted in two populations with different genetic background and distinct patterns of exposure to environmental factors. The associations we reported between DTC and lifestyle, hormonal and reproductive factors are consistent with those published previously in the total samples of NC and CATHY studies [ 7 ,  18 ,  19 ]. We showed that BMI and alcohol consumption were more strongly associated with DTC in individuals with  GSTM1  and  GSTT1 -null genotypes than in those with non-null genotypes. A significant interaction between  GSTT1 -null and age at menarche was also observed. By considering genes and environmental or lifestyle factors, no heterogeneity between groups was observed.\nThe distribution of  GSTM1  and  GSTT1  CNVs among Europeans in our study was similar to previous reports [ 20 , 21 ]. To the best of our knowledge, this is the first report of CNV distribution in Melanesians. We found that the frequency of null genotypes was notably higher in Melanesians than in Europeans, which is consistent with previous data concerning other Pacific islanders [ 22 ].\nBeing overweight has consistently been associated with DTC risk [ 23 , 24 ]. In the present study, obesity was more strongly associated with DTC in  GST-null  than non-null individuals, particularly when both genes were deleted. This finding was consistent in both ethnic groups. To the best of our knowledge, the only study to investigate the joint effect of BMI and GST genes in DTC was conducted in Korea and found no indication of an interaction [ 25 ]. However, the prevalence of obesity in this population is very low and the study may have been underpowered. Studies in populations with a high prevalence of obesity would be helpful for confirming our finding.\nObesity may increase the risk of DTC through complex mechanisms [ 26 ]. The chronic systemic inflammation induced by obesity may favor cancer development through the formation of reactive oxygen species (ROS) [ 27 ]. Because the enzymes encoded by  GSTM1  and  GSTT1  contribute to reducing oxidative stress by conjugating and eliminating ROS products [ 28 ], obese subjects with deletion of these genes could be at particularly high risk of DTC.\nThe observed inverse association of alcohol drinking with DTC is in line with several studies [ 10 , 11 , 29 – 31 ] and a recent meta-analysis [ 32 ]. This inverse association was greater in  GSTM1 / GSTT1 -null subjects than non-null subjects, particularly in those with a deletion of both genes. An interaction between alcohol intake and  GSTM1  or  GSTT1  was reported previously in cancer of the breast, lung, and stomach [ 33 – 35 ]. Our study is the first report of such an interaction in thyroid cancer.\nThe mechanisms underlying the potentially protective effect of alcohol drinking in thyroid carcinogenesis are not known. Some studies have suggested that free radicals generated by alcohol metabolism may have either a direct toxic effect on the thyroid tissues or disturb the hypothalamus–pituitary–thyroid axis [ 30 , 36 , 37 ]. However, how this can lead to a decreased risk of DTC remains unclear [ 38 ], and it does not explain the synergistic effect of alcohol and GST-null genotypes.\nUnlike alcohol,  GSTM1  and  GSTT1  were not found to modify the association between tobacco smoking and DTC. The interaction between GST genes and smoking was also investigated in a previous study, but no evidence of an interaction was reported [ 22 ].\nGiven the higher incidence of DTC in women than in men, female sex hormones have been suspected to play a major role in thyroid carcinogenesis [ 39 ]. We found that late age at menarche was associated with increased DTC incidence. This association was also reported by several studies [ 7 , 40 , 41 ], but a recent meta-analysis found no conclusive evidence in cohort studies [ 42 ]. We also found that ever users of oral contraceptives were at decreased risk, a finding supported by recent studies [ 43 , 44 ]. Conversely, parity was not associated with DTC in our data. There was also some indication that menopause increases DTC risk. These findings were described in detail previously in the CATHY [ 18 ] and NC [ 7 , 19 ] studies.\nOur study is the first to investigate the interaction between GST genotype and reproductive factors in thyroid cancer. No interaction was observed between  GSTM1  and age at menarche, use of oral contraceptives, parity, or menopausal status, whereas significant interaction was observed between  GSTT1  and age at menarche.\nThe association of DTC with age at menarche has been inconsistent in epidemiological studies, and the potential underlying mechanisms involved are unclear. It has been shown that expression of  GSTM1  differs during menstruation [ 39 ] and polymorphisms in the GSTs genes were associated to various endometrial pathologies such as for instance endometriosis [ 45 ]. However, how  GSTT1  may modulate the association between late age at menarche and DTC risk remains to be elucidated.\nThis study used data collected from two case-control studies in Europeans and Melanesians with a population-based design and exhaustive identification of thyroid cancer cases. Our study was based on a relatively high number of subjects. With the exception of one study in Korea that included 1372 cases and 1669 controls [ 25 ], previous studies of DTC risk in relation to  GSTM1  or  GSTT1  genotype have had limited sample sizes of less than 300 cases. Unlike previous studies contrasting DTC risk in null vs. non-null genotypes, CNVs of  GSTM1  and  GSTT1  were also examined in our study to investigate a potential dose-effect relationship between enzymatic activity and cancer risk. However, such a dose-effect association was not observed in our data. We also examined gene-environment interactions between  GSTM1  and  GSTT1  genotypes and suspected risk factors of DTC. The analyses of risk factors in these two ethnic groups with diverse prevalence of exposure to risk factors and frequency of GST-null genotypes could help explain the difference in DTC incidence between the study populations. In particular, the highest prevalence of both GST-null genotypes and obesity in Melanesians compared to Europeans could partly explain the higher incidence of DTC in Melanesians.\nOne limitation of our study is the limited statistical power in some strata of the stratified analyses. Potential recall bias inherent to case-control studies may have occurred when assessing exposure to several risk factors. But differential misclassification was unlikely as cases and controls were interviewed in the same conditions by trained interviewers. Also, we cannot exclude that some of the associations found are due to the number of tests used. Replication in independent studies are necessary to confirm our results.\nIn conclusion, our results suggest that  GSTM1  and  GSTT1  may modify the associations between DTC risk and obesity, alcohol consumption, and possibly hormonal factors. Disparities in the frequency of  GSTM1  and  GSTT1  deletion and in the prevalence of obesity between populations worldwide, may partly explain differences in thyroid cancer incidence.\n\n(XLSX)\nClick here for additional data file.","source_license":"CC-BY-4.0","license_restricted":false}