Addressing limitations of previous meta-analyses of uniportal versus biportal endoscopic lumbar surgery: an umbrella review and de-duplicated, pathology-stratified meta-analysis.

PURPOSE: Previous meta-analyses of uniportal versus biportal endoscopic lumbar surgery are affected by overlapping primary studies, discordant study-level extraction, and aggregation of pathologies requiring different operative tasks. We appraised the review-level evidence and generated a de-duplicated, pathology-stratified synthesis of operative time, complications, disability, and leg pain.

METHODS

Five databases were searched for systematic reviews. Primary citations were normalized across reviews, each eligible comparison was included once per outcome, and discordant group allocation, sample size, follow-up, and numerical values were checked against the original primary publication wherever accessible. Random-effects models and meta-regression examined pathology, approach, and country. Review quality, primary-study risk of bias, and certainty were assessed with AMSTAR 2, RoB 2 or the Newcastle-Ottawa Scale, and GRADE, respectively.

RESULTS

Five reviews and 35 analytic comparisons from 34 primary publications were included. Operative time did not differ overall (35 comparisons; 3,338 participants; MD, 0.09 min; 95% CI, - 8.29 to 8.47), but the direction differed by pathology (P < 0.001): the estimate favored shorter uniportal procedures for LDH (13 comparisons; MD, - 17.67 min; 95% CI, - 26.90 to - 8.44; I² = 92.6%) and shorter biportal procedures for LSS (22 comparisons; MD, 10.45 min; 95% CI, 0.73 to 20.17; I² = 98.9%). Complications (OR, 1.44; 95% CI, 0.94 to 2.23), ODI (MD, 1.96; 95% CI, - 0.60 to 4.52), and leg-pain VAS (MD, 0.01; 95% CI, - 0.05 to 0.06) did not differ clinically. Certainty was low to very low.

CONCLUSION

De-duplication and pathology stratification showed that an overall null operative-time estimate may average effects in opposite directions. These findings support possible effect modification by pathology but, given substantial residual heterogeneity and low to very low certainty, do not establish the superiority of either technique for LDH or LSS.