Comparative effects of core decompression combined with autologous cell-based strategies on structural outcomes in early-stage osteonecrosis of the femoral head: a systematic review and Bayesian network meta-analysis of randomized and quasi-randomized controlled trials

Abstract

Core decompression (CD) creates a reparative environment for revascularization and bone repair in early- to mid-stage osteonecrosis of the femoral head (ONFH), but structural outcomes remain heterogeneous and autologous cell-based augmentation strategies are biologically distinct. We compared CD-based strategies for radiographic/structural progression (RSP) and conversion to total hip arthroplasty (THA) using Bayesian network meta-analysis. Six databases and trial registries were searched from inception to August 6, 2026 for randomized and quasi-randomized trials in adults with imaging-confirmed early- to mid-stage ONFH. Five strategies were evaluated: conservative/non-operative control, CD alone, CD plus autologous bone marrow mononuclear cells (CD+BMMNC), CD plus autologous bone marrow aspirate concentrate (CD+BMAC), and CD plus autologous osteoblastic cells. THA conversion and RSP were analyzed separately using Bayesian random-effects consistency models with posterior odds ratios (ORs) and 95% credible intervals (CrIs). Inconsistency, transitivity, SUCRA, sensitivity analyses, comparison-adjusted funnel plots, and GRADE informed robustness and certainty. Nine independent trials (seven randomized and two quasi-randomized) contributed to five-strategy networks; eight contributed to each outcome. For THA conversion, all comparisons involving CD+BMMNC were directionally favorable, but every 95% CrI included 1. For RSP, CD+BMMNC reduced the odds versus CD (OR = 0.02, 95% CrI 0.000–0.31), while CD+BMAC had higher odds than CD+BMMNC (OR = 51.77, 95% CrI 1.95–5795.12); other comparisons remained imprecise. CD+BMMNC ranked first for THA conversion (SUCRA 88.5%) and RSP (98.0%). Sensitivity analyses preserved the favorable direction; for RSP versus CD, 95% CrIs remained below 1 whenever the node was retained. GRADE certainty ranged from moderate to low, with moderate certainty for lower RSP risk with CD+BMMNC versus CD and low certainty for the corresponding THA comparison. Funnel plots showed no clear asymmetry, with only seven direct-comparison estimates per outcome. CD+BMMNC showed the most consistent structural-preservation signal, with the clearest support for reducing RSP. Definitive superiority across both outcomes is not established because THA estimates remain imprecise and cell-based nodes are supported by few trials. Larger multicenter trials with standardized products, techniques, outcomes, follow-up, bilateral-hip handling, and patient stratification are needed to establish durable benefit.