MRI-derived necrotic burden and subsequent femoral head collapse in osteonecrosis of the femoral head: a systematic review and meta-analysis

Femoral head collapse substantially influences the management of osteonecrosis of the femoral head (ONFH). Magnetic resonance imaging (MRI) can quantify necrotic lesion extent before radiographic collapse, but the strength, precision, and generalizability of the association between MRI-derived necrotic burden and subsequent collapse remain uncertain. We systematically evaluated MRI-derived necrotic burden in relation to subsequent femoral head collapse. The protocol was registered in PROSPERO (CRD420261456638). PubMed, Embase, the Cochrane Library/CENTRAL, Web of Science, CNKI, Wanfang, and VIP were searched from inception to 8 August 2026. Quantitative MRI-derived measures were synthesized separately as lesion-burden, necrotic-angle, and continuous measurement frameworks. Random-effects meta-analyses used restricted maximum likelihood estimation with Knapp-Hartung-adjusted confidence intervals and 95% prediction intervals. Additional sensitivity analyses examined threshold derivation, minimum follow-up duration, temporal sequencing of MRI before collapse, and individual-study influence. Risk of bias was assessed using QUIPS, and certainty of evidence using GRADE for prognostic-factor evidence. Eighteen reports representing 16 cohort families were included after study-level eligibility and extraction re-audit. In the primary binary analysis of six studies involving 383 hips, higher MRI-quantified necrotic burden was associated with greater odds of subsequent collapse (OR 31.62, 95% CI 11.70–85.50; I² = 23.8%; 95% prediction interval 5.60-178.67). Four studies involving 336 hips evaluating necrotic-angle metrics yielded an OR of 52.55 (95% CI 6.33-436.25; I² = 63.8%), but the prediction interval was extremely wide and crossed the null (0.20–14,090.93). In the continuous analysis of six studies involving 359 hips, subsequently collapsed hips had greater MRI-quantified necrotic burden (Hedges’ g 2.23, 95% CI 1.29–3.16; I² = 86.0%; 95% prediction interval − 0.31 to 4.76). A sensitivity analysis restricted to unequivocally pre-collapse MRI assessment included four studies involving 231 hips and yielded Hedges’ g 1.83 (95% CI 0.51–3.14; 95% prediction interval − 1.82 to 5.48). Sensitivity analyses generally preserved the direction of the average associations, although several prediction intervals crossed the null. Certainty of evidence was low for the primary binary analysis and very low for the necrotic-angle and continuous analyses. Greater MRI-derived necrotic burden is associated with subsequent femoral head collapse in ONFH, but the magnitude of association is imprecisely estimated and varies across measurement approaches and study settings. MRI-derived volume, area, surface, angular, and spatial measures should be regarded as related but non-interchangeable prognostic markers. Current evidence does not support a universal MRI threshold, a stand-alone treatment rule, or a validated individual-risk prediction model. Prospective multicenter validation using clearly pre-collapse MRI assessment, standardized measurements, predefined thresholds, adequate follow-up, and appropriate handling of bilateral hips is needed.