Comparing Nonrandomized Observational Studies With Randomized Controlled Trials in Cervical Disc Arthroplasty A Meta-analysis

Study Design. Systematic review and meta-analysis. Objective. To compare the treatment effects of observational studies versus randomized controlled trials (RCTs) in cervical disc arthroplasty. Summary of Background Data. RCTs can be logistically challenging and sometimes insufficiently generalizable; well-designed observational studies have been suggested as an alternative. We hypothesized that treatment effects of observational studies in cervical disc arthroplasty are similar to those of RCTs. Methods. We searched electronic database from 2000 to 2014. The Neck Disability Index (NDI) was the primary outcome from which the standardized pre-and-post mean difference (Hedges's g) was determined. Meta-analysis was performed to compare Hedges's g from observational studies to that of RCTs. Potential moderator variables including study quality, age, gender, industry sponsorship, location by continent, and disc types were also collected and analyzed. Observational studies were further stratified into prospective and retrospective, and they were compared to each other as well as to RCTs. Results. We identified nine RCTs, 28 observational studies, and one hybrid study for meta-analysis. NDI Hedges's g was 2.15 for RCTs and 2.03 for observational studies, which was not significant (P = 0.416). No significant difference was found in secondary outcomes. However, after further stratification, prospective observational studies had less treatment effect in Visual Analog Scale neck compared with that of RCTs (1.60 vs. 2.11, P = 0.006). RCTs recruited younger patients (44.1 vs. 45.6, P = 0.008) with worse NDI at baseline (54.30 vs. 46.92, P < 0.001). Patients treated with ProDisc-C showed less standardized improvement on the NDI compared with the patients treated with Prestige (1.41 vs. 2.48, P = 0.026). Conclusion. Prospective observational studies that utilize the same features of RCTs such as inclusion and exclusion criteria validated clinical outcomes, and statistical methods can provide valuable information about the treatment effects on a generalizable population. Level of Evidence: 4