The Statistical Fragility of Trials Comparing Cervical Disc Arthroplasty and Anterior Cervical Discectomy and Fusion

Study Design. Meta-analysis Objective. Assess the robustness of randomized controlled trials (RCTs) that compared cervical disc arthroplasty (CDA) and anterior cervical discectomy and fusion (ACDF) for the treatment of symptomatic degenerative cervical pathology by using fragility indices. Summary of Background Data. RCTs comparing these surgical approaches have shown that CDA may be equivalent or even superior to ACDF due to better preservation of normal spinal kinematics. Methods. RCTs reporting clinical outcomes after CDA versus ACDF for degenerative cervical disc disease were evaluated. Data for outcome measures were classified as continuous or dichotomous. Continuous outcomes included: Neck Disability Index (NDI), overall pain, neck pain, radicular arm pain, and modified Japanese Orthopaedic Association (mJOA) scores. Dichotomous outcomes included: any adjacent segment disease (ASD), superior-level ASD, and inferior-level ASD. The fragility index (FI) and continuous FI (CFI) were determined for dichotomous and continuous outcomes, respectively. The corresponding fragility quotient (FQ) and continuous FQ (CFQ) were calculated by dividing FI/CFI by sample size. Results. Twenty-five studies (78 outcome events) were included. Thirteen dichotomous events had a median FI of 7 (IQR: 3–10) and the median FQ was 0.043 (IQR: 0.035–0.066). Sixty-five continuous events had a median CFI of 14 (IQR: 9–22) and median CFQ of 0.145 (IQR: 0.074–0.188). This indicates that, on average, altering the outcome of 4.3 patients out of 100 for the dichotomous outcomes, and 14.5 out of 100 for continuous outcomes, would reverse trial significance. Of the 13 dichotomous events that included lost to follow-up data, 8 (61.5%) represented ≥7 patients lost. Of the 65 continuous events reporting lost to follow-up data, 22 (33.8%) represented ≥14 patients lost. Conclusion. RCTs comparing ACDF and CDA have fair to moderate statistical robustness and do not suffer from statistical fragility.