Methodological quality evaluation of systematic reviews or meta-analyses on ERCC1 in non-small cell lung cancer: a
Huan Tao1, Yueyuan Zhang2, Qian Li1
1Department of Evidence-based Medicine and Clinical Epidemiology, West China Medical School of Medicine/West China Hospital, Sichuan University, 37 Guo Xue Xiang, Cheng Du, 610041, China.
Purposes:
To assess the methodological quality of systematic reviews (SRs) or meta-analysis concerning the predictive value of ERCC1 in platinum chemotherapy of non-small cell lung cancer.
Methods:
We searched the PubMed, EMbase, Cochrane library, international prospective register of systematic reviews, Chinese BioMedical Literature Database, China National Knowledge Infrastructure, Wan Fang and VIP database for SRs or meta-analysis. The methodological quality of included literatures was evaluated by risk of bias in systematic review (ROBIS) scale.
Results:
Nineteen eligible SRs/meta-analysis were included. The most frequently searched databases were EMbase (74%), PubMed, Medline and CNKI. Fifteen SRs did additional retrieval manually, but none of them retrieved the registration platform. 47% described the two-reviewers model in the screening for eligible original articles, and seven SRs described the two reviewers to extract data. In methodological quality assessment, inter-rater reliability Kappa was 0.87 between two reviewers. Research question were well related to all SRs in phase 1 and the eligibility criteria was suitable for each SR, and rated as 'low' risk bias. But the 'high' risk bias existed in all the SRs regarding methods used to identify and/or select studies, and data collection and study appraisal. More than two-third of SRs or meta-analysis were finished with high risk of bias in the synthesis, findings and the final phase.
Conclusions:
The study demonstrated poor methodological quality of SRs/meta-analysis assessing the predictive value of ERCC1 in chemotherapy among the NSCLC patients, especially the high performance bias. Registration or publishing the protocol is recommended in future research.
