Strategies to improve the credibility of meta-analyses in spine surgery: a systematic survey.

Background Context: Meta-analyses are powerful tools that can synthesize existing research, inform clinical practice, and support evidence-based care. These studies have become increasingly popular in the spine surgery literature, but the rigor with which they are being conducted has not yet been evaluated.
Purpose: Our primary objectives were to evaluate the methodological quality (credibility) of spine surgery meta-analyses and to propose strategies to improve future research. Our secondary objectives were to evaluate completeness of reporting and identify factors associated with higher credibility and completeness of reporting.
Study Design: This study is based on a systematic survey of meta-analyses.
Outcome Measures: We evaluated credibility according to the Users' Guide to the Medical Literature and completeness of reporting according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist.
Methods: We systematically searched MEDLINE, EMBASE, and The Cochrane Library, and two reviewers independently assessed eligibility, credibility, and completeness of reporting. We used multivariable linear regression to evaluate potential associations. Interrater agreement was quantified using kappa and intraclass correlation (ICC) coefficients.
Results: We identified 132 eligible meta-analyses of spine surgery interventions. The mean credibility score was 3 of 7 (standard deviation [SD], 1.4; ICC, 0.86), with agreement for each item ranging from 0.54 (moderate) to 0.83 (almost perfect). Clinical questions were judged as sensible in 125 (95%), searches were exhaustive in 102 (77%), and risk of bias assessments were undertaken in 91 (69%). Seven (5%) meta-analyses addressed possible explanations for heterogeneity using a priori subgroup hypotheses and 24 (18%) presented results that were immediately clinically applicable. Investigators undertook duplicate assessments of eligibility, risk of bias, and data extraction in 46 (35%) and rated overall confidence in the evidence in 24 (18%). Later publication year, increasing Journal Impact Factor, increasing number of databases, inclusion of Randomized Controlled Trials, and inclusion of non-English studies were significantly associated with higher credibility scores (p<.05). The mean score for reporting was 18 of 27 (SD, 4.4; ICC, 0.94).
Conclusions: The credibility of many current spine surgery meta-analyses is limited. Researchers can improve future meta-analyses by performing exhaustive literature searches, addressing possible explanations of heterogeneity, presenting results in a clinically useful manner, reproducibly selecting and assessing primary studies, addressing confidence in the pooled effect estimates, and adhering to guidelines for complete reporting.
(Copyright © 2015 Elsevier Inc. All rights reserved.)