Endoscopic decompression alone for lumbar spinal stenosis with degenerative spondylolisthesis: a systematic review and meta-analysis of paired change scores.

Abstract

To synthesise patient-reported and reoperation outcomes after endoscopic decompression alone for lumbar spinal stenosis with degenerative spondylolisthesis, while accounting for paired measurements, linked reports, clinical importance and certainty. PubMed/MEDLINE, Embase, Scopus, the Cochrane Library, trial registries and supplementary sources were searched to 9 August 2026. Adult cohorts with lumbar stenosis, degenerative spondylolisthesis, endoscopic decompression alone and at least 12 months' follow-up were eligible. Linked reports were clustered to avoid duplicate patients. Leg-pain visual analogue scale (VAS; 0-10) and Oswestry Disability Index (ODI; 0-100) were pooled as paired mean improvements using random-effects restricted maximum likelihood with Hartung-Knapp-Sidik-Jonkman inference. When change-score SDs were unavailable, a pre/post correlation of 0.60 was imputed and varied from 0.30 to 0.80. Reoperations were pooled with a binomial-normal generalized linear mixed model. ROBINS-I and GRADE were applied. Nine independent cohorts (335 patients) informed at least one synthesis. Leg-pain VAS improved by 5.25 points (95% CI 4.57-5.93; I²=93.1%; τ²=0.69; 95% prediction interval 3.16-7.34; nine studies). ODI improved by 38.11 points (30.43-45.79; I²=98.6%; τ²=82.22; prediction interval 14.54-61.68; eight studies, 293 patients). Both prediction intervals exceeded reference MCIDs (1.6 VAS points; 12.8 ODI points), but between-study inconsistency was substantial. Eleven reoperations occurred among 335 patients; the pooled proportion was 3.3% (95% CI 1.5%-5.6%). All quantitative studies had serious overall risk of bias, and certainty was very low. Substantial improvement was observed after endoscopic decompression in selected patients with mostly stable, low-grade degenerative spondylolisthesis. Because evidence consisted predominantly of uncontrolled retrospective cohorts with marked heterogeneity, these data do not establish comparative effectiveness, equivalence to fusion or superiority of uniportal over biportal techniques. CRD420251270278.