The Epidemiological Study On The Association Of TERT, CLPTM1L, CHRNA3Polymorphisms And Environmental Factors With The Risk Of Lung Cancer In Never-smokers

[Objective]To investigate the main impact factors of lung cancer and the independent and interaction effects of polymorphisms in gene TERT,CLPTM1L,CHRANA3and environmental facts on the development of lung cancer in nonsmokers. To explore the application of the statistical methods in research of interactions of gene‐gene and gene‐environment of complex disease.[Methods]1. A case‐control study was conducted with417patients and453control subjects. Patients were recruited from the First Clinical Medical College and the Affiliated Union Hospital of Fujian Medical University and Fuzhou General Hospital with bronchoscopy and pathologic confirmation. All healthy controls were enrolled from hospital visitors and community population, and were frequency‐matched to the cases on gender and age(±2year). Eligible subjects were personally interviewed using a questionnaire. Univariate and multivariate logistic regression were used to identify the risk factors of lung cancer and to estimate the odd ratios (ORs) and95%confidence intervals (95%CIs). Random forest and classification trees were used to investigate the interaction effects.2.5ml blood sample was collected from each patients. The collection of samples among control subjects was accomplished by using OrageneTM DNA self‐collection kit. The genomic DNA was extracted from all blood and saliva samples. The genotypes of polymorphisms were determined by Sequenom platform. Univariate and multivariate logistic regression, random forest, generalized multifactor dimensionality reduction，crossover analysis were used to identify the independent and interaction effects of polymorphisms in gene TERT,CLPTM1L,CHRANA3and environmental facts to the development of lung cancer in nonsmokers. 3. A comprehensive comparison of parametric statistical methods (logistic regression, crossover analysis) and non‐parametric statistical methods (classification trees, random forest, generalized multifactor dimensionality reduction) was made, including the advantages or the flaws and scope of application, then simulated the use of instance data.[Results]1. The risk factors of lung cancer in never‐smokers included: BMI＜18.5(OR=2.753,95%CI=1.353‐5.602), contamination near residential area (OR=2.609,95%CI=1.567‐4.344), cooking oil fume(light)(OR=2.423,95%CI=1.567‐3.719), cooking oil fume(moderate)(OR=1.742,95%CI=1.032‐2.939), cooking oil fume (heavy)(OR=5.929,95%CI=2.120‐16.584), exposure to ETS(OR=2.582,95%CI=1.852‐3.598), family history of lung cancer(OR=2.608,95%CI=1.098‐6.198), use of pesticide(OR=2.436,95%CI=1.559‐3.806), deep X‐ray exposure(OR=2.753,95%CI=1.353‐5.602), introverted personality (OR=1.791,95%CI=1.149‐2.792). The protective factors of lung cancer in never‐smokers included：intake of fruits (1‐3times per week)(OR=0.496,95%CI=0.284‐0.868), fruits (>3times per week)(OR=0.336,95%CI=0.197‐0.573), kelp (1‐4times per month)(OR=0.564,95%CI=0.351‐0.906), kelp (>4times per month)(OR=0.611，95%CI=0.376‐0.992), drinking tea (OR=0.497,95%CI=0.347‐0.712), work with moderate labor intensity (OR=0.569,95%CI=0.389‐0.832). The result of interaction analysis indicated that cooking oil fume showed main effect in the development of lung cancer in never‐smokers (MeanDecrease Gini, MDG=33.1541).2. CLPTM1L (rs401681) polymorphic variations were associated with lung cancer in never‐smokers. This association was still obvious after stratified by pathological types. No association between the SNP TERT(rs2736100), CHRNA3(rs1051730) and the susceptibility of lung cancer in never‐smokers was observed. We then examined the combined effect of rs2736100and rs401681and observed a significant effect between unfavorable genotypes(rs2736100TG/GG or rs401681GG) and risk of lung cancer. Compared to individuals without unfavorable genotypes, those with1or2unfavorable genotypes, or both, had a significantly higher risk of lung cancer. Cooking oil fume had a major effect in the etiology of lung cancer in nonsmokers (MDG=21.7259). Individuals who carried the TG or GG of TERT (rs2736100), exposure to cooking oil fume and ETS, had higher risk of lung cancer. The positive interaction (P<0.05) on multiplicative scale between rs2736100and cooking oil fume was observed.3. Non‐parametric statistics methods such as classification tree, random forest, and generalized multifactor dimensionality reduction couldn’ t calculate the exact strength of interaction. However, they could be applied to identify the factors with interaction effects, and then the exact value of main effect and interaction effect were calculated by using enter method in logistic regression.[Conclusion]1. BMI＜18.5, contamination near residential area, cooking oil fume and exposure to ETS, family history of lung cancer, use of pesticide, deep X‐ray exposure, introverted personality were the major risk factors of lung cancer in never‐smokers. The frequent intake of fruit, kelp, and drinking tea, moderate labor intensity were identified as protective factors.2. The analysis of TERT‐CLPTM1L polymorphisms with susceptibility to lung cancer suggested that polymorphic variation on5p15.33played a role in determining the risk of developing lung cancer in never‐smokers. The result of interaction analysis indicated that cooking oil fume remained as a major effect in the etiology of lung cancer in nonsmokers and the best model included TERT (rs2736100), exposure to ETS, cooking oil fume.3The interaction between factors couldn’t be ignored. Each statistics methods had advantages and disadvantages. Justifying the most suitable conditions for each of the approaches, combining the use of parametric statistics and non‐parametric statistics may be preferable when evaluating gene‐gene and gene‐environment interacton.