Assessing the Effects of Publication Bias on Reported Outcomes of Cervical Disc Replacement and Anterior Cervical Discectomy and Fusion: A Meta-Epidemiologic Study

Abstract

There have been several clinical trials as well as observational studies that have compared the outcomes of different cervical disc replacement (CDR) devices with anterior cervical disc replacement and fusion (ACDF). Although the results of these studies have provided sufficient evidence for the safety of CDR, there is still a lack of consensus in terms of longer-term outcomes, with studies providing equivocal results for the 2 procedures. In the current study, we used a novel methodology, a meta-epidemiologic study, to investigate the impact of study characteristics on the observed effects in the literature on CDR and ACDF. Data were abstracted from available meta-analyses regarding author, study author, year, intervention events, control events, and sample size, as well as year and geographic location of each study within the meta-analyses. We grouped the studies based on median year of publication as well as the region of the submitting author(s). Odds ratios, 95% confidence intervals (CIs), and standard errors of individual studies were calculated based on the number of events and sample size for each arm (ACDF or CDR). Further, results of outcomes from individual studies were pooled and a meta-analysis was conducted. Ratio of odds ratio (ROR) was used to assess the impact of each of these factors on estimates of the study for CDR versus ACDF. A total of 13 meta-analyses were analyzed after exclusions. Using the results from 10 meta-analyses, we found that studies published before 2012 reported significantly lower odds of a reoperation after CDR (vs. ACDF), compared with studies published after 2012 (ROR, 0.51; 95% CI, 0.38-0.67; P < 0.001). We did not observe a significant impact of study year on difference in estimates between CDR and ACDF for adjacent segment disease (ROR, 0.99; 95% CI, 0.64-1.55; P= 0.465). The region of submitting author was also found to have no impact on results of published studies. These results indicate that there may be a publication bias regarding the year of publication, with earlier studies reporting lower reoperation rates for CDR compared with ACDF.