The Meta-analysis Of Risk Factors For Esophageal Cancer In Taihang Mountain Areas And The Modification Of Mothods To Detect Publication Bias

Esophageal cancer（EC）is the eighth commonest type of cancer worldwide,occupying sixth place in terms of mortality.EC incidence rate is reported to be very high in the Taihang Mountain area of north-central China.We performed a meta-analysis and thus evaluated the risk factors for EC in the Taihang Mountain area,so as to provide the basis for drafting preventive measures of esophageal cancer.In this paper,we presented two innovative methods for the detection of publication bias,which was based on arcsine transformation and arctangent transformation.In order to verify the feasibility of the new methods,we compared the performances of two innovative methods developed and the existing methods in the meta-analysis of the risk factors for esophageal cancer by simulation.Firstly,a search was conducted through PubMed,Embase,Medline,CNKI,WANFANG,Google scholar and Cochrane Library.Combining with other retrieval strategy,we collect comprehensively all published case-control studies regarding the esophageal cancer.We formulated the detailed inclusion and exclusion criteria,in order to screen out the appropriate literature.The“Odds Ratio（OR）”was invoked as the key measure of effect size for meta-analysis.The meta-analysis was conducted using STATA 12.0 software for data we extracted.And we calculated the pooled OR and 95%Confidence Interval（95%CI）.Secondly,we developed two new methods able to detect publication bias,which was based on the linear regression methods of effect size and variance.The method 1was based on the arcsine transformation of the effect size,and transformed its approximate variance as the inverse of sample size,which weights is（N_t+N_c）.Let N_t and N_c be,respectively,the sample size of case and control groups.The method 2 was based on the arctangent transformation of the effect size,and transformed its approximate variance as the inverse of sample size,which weights is[1/（N_t）+1/（N_c）]^-1.We compared the type I error rates and power of the new tests we proposed and existing tests based on the meta-analysis of the risk factors for esophageal cancer according to meta-analysis datasets simulated with the R language.And we applied these tests to a specific example for Taihang Mountain areas.In according to meta-analysis,we concluded that hot food and drinks,socio-psychological factor,salted food,family history of esophageal cancer,alcohol intake,pickled vegetables consumption,hard food,and smoking were the risk factors for esophageal cancer.While the intake of meat,eggs and milk,fresh fruit intake,fresh vegetable consumption and high education level appeared to be protective factors.The results of our investigation could contribute to identify the influence factors responsible for esophageal cancer and assisted in the development of preventive measures in order to control the incidence and mortality of esophageal cancer in the Taihang Mountain region.Method 1 of two methods detecting the publication bias we proposed might be appropriate in almost all cases,especially when there was a small effect size or few studies was included.Method 2 had slightly greater properties when the effect size was small.Method 1 and method 2 both had some limitations when the level of heterogeneity was moderate or high,but the method 1 illustrated the acceptable property.We carried out a comprehensive simulation study based on the scenarios that extracted from specific example,and provided two effective analytical methods to test the publication bias in the meta-analysis of risk factor for esophageal cancer,which appeared to be effective and reliable.