Note of the methodological flaws in the paper entitled "Association between the G870A polymorphism of cyclin D1 gene and glioma risk".

Abstract

Dear Editor, Recently, we had the honor of reading a meta-analysis article published in Tumor Biology which was completed by the investigators (Zong et al.) [1]. Because of our interest in the association between the G870A polymorphism of cyclin D1 gene and glioma risk, we carefully read the entire article. In this study, the investigators performed a meta-analysis of four independent studies to examine whether CCND1 G870A polymorphism was associated with glioma risk. A significant association was found between the CCND1 G870A polymorphism and glioma risk in three genetic models (ORA vs. G= 1.178, 95 % CI 1.025–1.354, POR=0.021; ORAA vs. GG= 1.328, 95 % CI 1.007–1.750, POR=0.045; ORAA+AG vs. GG= 1.253, 95 % CI 1.006–1.516, POR=0.044). They finally reached an important conclusion that the CCND1 G870A polymorphism is a risk factor for the development of glioma. It is a valuable study on the association of CCND1 G870A polymorphism with glioma risk. Nevertheless, there are several deficiencies in the methodology we would like to raise related to this article. To begin with, only four electronic databases, namely, PubMed, Embase, Web of Science, and Wanfang databases for articles up to October 31, 2013 were searched for the meta-analysis. The small number of required papers would be an important limitation of the meta-analysis. We recommended that more electronic databases (such as OVID, SCOPUS, Science Citation Index, Cochrane-controlled Trials Register, Chinese Biomedical Literature Database, and China National Knowledge Infrastructure) should be searched. Meanwhile, the investigators just showed us some search terms such as glioma, brain tumor, polymorphism, variant, CCND1, cyclin D1, and rs603965. To make the meta-analysis more credible, the investigators should provide us the flow chart of literature search and the detailed manual search strategy. Furthermore, the investigators did not evaluate the quality of the four included studies. As is well known, articles of high quality are requirements for a convincing meta-analysis. Introduction of much low-quality articles may result in wrong conclusions. To make the meta-analysis more credible, the investigators should perform the quality evaluation by using some authority standards, such as the Newcastle-Ottawa Scale (NOS) [2] and Downs-Black tool [3]. Finally, there was significant heterogeneity in the recessive genetic model in Caucasians (AA vs. AG+GG, I=74.6 %, PH=0.047). However, the investigators did not find out the source of heterogeneity. We recommend that the investigators should perform a one-way sensitivity analysis [4, 5] to find out the source of heterogeneity. In conclusion, thanks go to the investigators for their contribution to supplying us with an assessment of the relationship between CCND1G870A polymorphism and glioma risk. However, considering the above methodological flaws in the meta-analysis, we hope that the investigators would redo the meta-analysis with the correct method and to reconsider the conclusions based on the new results. * Sen-ren Guo fjtcmsenren@163.com