Study characteristics
By using keywords, a total of 2422 related reports were found in digital databases. After scanning the titles and abstracts firstly, 2078 articles were excluded, including reviews, meta-analyses, clinical trials, irrelevant reports and duplicate reports. By carefully reading the whole text of remaining studies, we further removed 261 articles for the following reasons: (a) lack of the information for number of samples; (b) lack of research content on the association between genetic polymorphisms and cancers among smoking or drinking population. Finally, 83 articles (19311 cases and 23194 controls) were kept (Fig 1 Table 1, 2), of which 69 articles (15,897 cases and 19,249 controls) were about the relationship between GSTM1 and cancers[27–31, 19, 32–36, 20, 37–81, 3, 82–84, 4, 85–91]; 50 articles (11,800 cases and 14,857 controls) were about the relationship between GSTT1 and cancers[3, 4, 19, 27, 28, 30–34, 36, 37, 39–42, 44, 45, 47–52, 54, 55, 59, 61, 63, 64, 66, 68–71, 73, 77–79, 82, 83, 85–88, 90–95]; and 30 articles (8,417 cases and 9,715 controls) were about GSTP1 and cancers[4, 19, 20, 28, 33, 34, 37, 39, 45, 47, 48, 50, 52, 55, 62, 69, 70, 73, 82, 91, 96–105] among smokers. Among these researches, 8 articles mentioned the classification of smoking levels (20 pack-year) [31, 32, 50, 56, 85, 88, 90, 95]. For alcohol consumption, there were 13 articles (4,308 cases and 5,476 controls) were about the association between GSTM1 and cancers[19, 31, 36, 38, 39, 43, 68, 69, 72, 75, 77, 78, 100], 8 articles (2,949 cases and 4,025controls) were about GSTT1 and cancers[19, 31, 36, 39, 68, 69, 77, 78] and 4 articles (1,797 cases and 2,358 controls) were about GSTP1 and cancers[19, 39, 69, 100] among drinkers.
GSTM1 gene polymorphism studies among smokers and drinkers
Based on the 69 reports on smoking, there were 24 articles for lung cancer, 8 articles for bladder cancer, 6 articles for liver cancer, 5 articles for gastric cancer, 5 articles for cervical cancer, 4 articles for head and neck cancer, 3 articles for esophageal cancer, 2 articles for nasopharyngeal cancer, 2 articles for breast cancer, and 2 articles for prostate cancer; the other types of cancer were not conducted subgroup analysis due to no more than one articles involved. Among subgroup of ethnicities, 34 studies were from Caucasian, 30 studies were from Asian, and 5 studies were mixed.
Of the 13 reports on drinking, 3 articles were for gastric cancer, 2 articles were for head and neck cancer, 2 articles were for liver cancer, and 5 other cancers (breast cancer, thyroid cancer, esophageal cancer, colorectal cancer, and lung cancer). Among subgroup of ethnicities, 6 studies were from Caucasian, 5 studies were from Asian, and 2 studies were mixed (Table 3).
GSTT1 gene polymorphism studies among smokers and drinkers
Of the 50 articles on smoking, 13 articles were for lung cancer, 6 articles were for bladder cancer, 5 articles were for gastric cancer, 5 articles were for cervical cancer, 4 articles were for liver cancer, 2 articles were for head and neck cancer, 2 articles were for breast cancer, 2 articles were for pancreatic cancer, 2 articles were for prostate cancer, as well as other 9 articles that were not performed subgroup analysis for the same reasons as above. Among subgroup of ethnicities, 31 studies were from Caucasian, 14 studies were from Asian, and 5 studies were mixed.
Of the 8 studies on drinking, 2 articles were for gastric cancer and 6 articles were for other cancers (breast cancer, thyroid cancer, esophageal cancer, colorectal cancer, lung cancer, and liver cancer). Among subgroup of ethnicities, 4 studies were from Caucasian, 2 studies were from Asian, and 2 studies were mixed (Table 4).
GSTP1 gene polymorphism studies among smokers and drinkers
Of the 30 reports on smoking, 9 articles were for lung cancer, 4 articles were for bladder cancer, 3 articles were for head and neck cancer, 3 articles were for gastric cancer, 2 articles were for pancreatic cancer, and 2 articles for esophageal cancer as well as other 7 articles without subgroup analysis for the same reasons as above. Among subgroup of ethnicities, 21 studies were from Caucasian, 6 studies were from Asian, and 3 studies were mixed.
Of the 4 studies on drinking, articles were for gastric cancer, head and neck cancer, esophageal cancer, and colorectal cancer; 3 studies were from Caucasian and 1 study was from Asian (Table 5).
GSTM1, GSTT1 and GSTP1 polymorphisms studies in head and neck neoplasm among smokers
For GSTM1, a total of 12 studies were classified as head and neck neoplasm, 6 studies for Caucasian, 4 articles for Asian and 2 articles for mixed. For GSTT1, a total of 7 studies were classified as head and neck neoplasm, 4 studies for Caucasian, 1 article for Asian and 2 articles for mixed. For GSTP1, A total of 6 studies were classified as head and neck neoplasm, 4 studies for Caucasian, 1 article for Asian and 1 article for mixed ethnicities (Table 6).
Quantitative synthesis
The relationship between GSTM1 polymorphism and cancer risks alone and in combination with smoking or drinking
GSTM1 (null/present) might rise the overall cancer risk in both smokers (I2=72.20%, OR (95%CI) =1.305(1.149-1.482), P<0.001) and non-smokers (I2=61.90%, OR (95%CI) =1.362(1.204-1.540), P<0.001) with a similar increase. The subgroup results demonstrated that GSTM1-null might increase the risks of lung and nasopharyngeal cancer in both smokers and non-smokers, but high cancer risks were found in stomach, cervix uteri and prostate among non-smokers, instead of smokers. Moreover, GSTM1-null could increase the ORs in Caucasians, Asians and hospital-based group, regardless of smoking status. (Table 3).
The overall finding showed no statistical association was found between cancer risks and GSTM1-null in either drinkers or non-drinkers. Subgroup results demonstrated that drinking increased the risks of head and neck cancer with a 3.747-fold and liver cancer with a 4.244-fold among GSTM1-null carriers. Moreover, alcohol consumption increased the cancer risk of GSTM1-null carriers by 2.132 times among Caucasians, but no association was found among Asians (Table 3).
The relationship between GSTT1 polymorphism and cancer risks alone and in combination with smoking or drinking
GSTT1 (null/present) and the overall cancer risk were found to have a significant positive correlation (I2=81.00%, OR (95%CI) =1.265(1.032-1.552), P=0.024) among smokers, but no statistical association was found between GSTT1 (null/present) and cancer risks among non-smokers. Subgroup analysis found that smoking might raise the risks of lung and prostate cancer by 1.6-fold and 2.8-fold in GSTT1-null carriers, respectively, whereas this risk was not observed in non-smokers. Moreover, the increased cancer risks in combination with smokingdid not differ by ethnicities among GSTT1-null carriers (Table 4).
GSTT1-null significantly increased the overall cancer risk (I2=61.30%, OR (95%CI) =1.423(1.042-1.942), P=0.026) among drinkers, while no statistical association between GSTT1-null and cancer risk exited among non-drinkers. Subgroup analysis found that GSTT1-null in combination with alcohol consumption might increase the risk of gastric cancer to 1.877 times, but no statistical significance was found among non-drinkers. Moreover, drinking might be more likely to increase the cancer susceptibility of Asian GSTT1-null carriers by 2.246 times compared with 1.543 times in Caucasians (Table 4).
The relationship between GSTP1 polymorphism and cancer risks alone and in combination with smoking or drinking
GSTP1 polymorphism (AG+GG/AA) was not statistically related to the overall cancer risk in either smokers or non-smokers. However, subgroup analysis indicated a negative correlation might exit between the risk of head and neck cancer and GSTP1 polymorphism in combination with smoking (OR (95%CI) = 0.833(0.715-0.970), P=0.019), whereas no statistical correlation was found among non-smokers. Moreover, GSTP1 polymorphism might reduce the risk of cancers to 0.788-fold among Caucasian smokers.
GSTP1 polymorphism might not statistically affect the overall cancer risk in either drinkers or non-drinkers. However, GSTP1 (AG+GG) carriers might have a reduced risk of cancer with 0.724-fold among Caucasian drinkers, whereas such change was not found among non-drinkers (Table 5).
GSTM1, GSTT1 polymorphisms and the risk of head and neck neoplasm alone and in combination with smoking
For GSTM1,no statistical difference was found between the risk of head and neck neoplasm and GSTM1-null, either alone or in combination with smoking. But subgroup results showed that the risk of head and neck neoplasm in GSTM1-null increased to 1.620 (95%CI=1.072-2.449, P=0.022) among Caucasian smokers. For GSTT1, no statistical association exited between the risk of head and neck neoplasmand GSTT1-null, either alone or in combination with tobacco consumption (Table 6).
Publication bias
The results of publication bias for GSTM1, GSTT1, and GSTP1 gene polymorphisms with cancer risks were illustrated in Table 3, 4 and 5, respectively. Some results showed publication bias: Egger testfor GSTM1: 0.007(non-smoking); for GSTT1: 0.016/0.025(smoking/non-smoking) (Table 3).
Sensitivity analysis
No statistically significant variables were found in the sensitivity analysis of GSTM1, GSTT1 and GSTP1 gene polymorphisms. The meta-regression analysis found that the year, the ethnicity, and the source of control population were not associated with experimental heterogeneity.
FPRP and BFDP test
The FPRP and BFDP analysis values of GSTs gene polymorphisms and smoking or drinking were shown in Tables 3, 4, 5 and 6, respectively. According to the results of the FPRP analysis, a part of the results in GSTs polymorphisms models were noteworthy in the FPRP test at the OR of 1.5 with prior probabilities of 0.25 and 0.1, but very few results were noteworthy in the BFDP test at the OR of 1.5 with prior probabilities of 0.01, 0.001 and 0.000001, which suggested that the results of this study should be interpreted with caution and euphemism.