Diagnostic Accuracy of Intraoperative Neurophysiological Monitoring in Spine Surgery: An Umbrella Review and Study-Level Reanalysis

Purpose: Overlapping systematic reviews report conflicting accuracy estimates for intraoperative neurophysiological monitoring (IONM). In this umbrella review with study-level re-analysis, we summarized modality accuracy for new postoperative deficit by procedure and tested whether published estimates are reproducible. Methods: PubMed, Scopus and Web of Science were searched to 12 October 2026 for systematic reviews reporting the accuracy of somatosensory evoked potentials (SSEP), motor evoked potentials (MEP), D-wave, free-run or triggered electromyography (EMG), or multimodal IONM in spine surgery. Quality was rated with AMSTAR 2 and overlap with the corrected covered area. One index meta-analysis was selected per modality-procedure cell; where study-level data were available, primary studies were deduplicated and re-pooled with a bivariate model. GRADE was applied. Results: Thirty-seven reviews (31 critically low, 6 low confidence) informed 30 cells; 26 were re-pooled. Specificity was high (88.0%–98.8%) except in intramedullary tumour surgery (80.2%–83.7%). Sensitivity varied: MEP 57.0%–84.0% (mixed spine: 84.0% [80.2–87.1]), multimodal 63.1%–89.1%, SSEP 31.9%–75.6%, free-run EMG 46.8%, intramedullary D-wave 26.6%, and triggered EMG for screw malposition 31.3%–54.9%. The largest review pooled non-independent entries; counting one entry per cohort lowered SSEP sensitivity from 71.4% to 60.7% and MEP from 90.2% to 84.0%. Certainty was moderate in 3 cells, low in 12 and very low in 15. Conclusion: IONM alerts were specific, but a stable tracing did not reliably exclude a deficit, particularly with SSEP alone or triggered EMG. MEP-based multimodal monitoring showed the highest sensitivity. With mostly low or very low certainty, these estimates should inform, not replace, clinical judgement.