Biologic augmentation with core decompression for early femoral head osteonecrosis: a re-audited evidence map and endpoint-based synthesis

Cellular and platelet-derived products are added to core decompression for osteonecrosis of the femoral head, but comparator structure, overlapping reports and bilateral-hip analyses complicate inference. We re-audited a focused evidence set and examined endpoint-specific estimates and their robustness. The locked 478-record register was classified as direct incremental comparative (A), non-isolating comparative (B), contextual/out-of-scope (C), or no-results (D) evidence. A row-level ledger resolved 601 unmatched PubMed records; citation checking covered 61 seeds and 1,305 unique linked records. Reports were linked to study lineages before analysis. Product, comparator, endpoint, follow-up, population, overlap and analysis-unit gates determined pool membership. Random-effects risk ratios used Paule–Mandel heterogeneity and Hartung–Knapp inference. Approximate bilateral design effects used ICC 0.20 primarily and 0–0.50 in sensitivity analyses. Leave-one-out, all-eligible and explicitly broadened PICO/design-assumption analyses were performed. The locked register comprised 28 A, 15 B, 416 C and 19 D records. External additions yielded 30 Class-A reports across 26 conservatively grouped study lineages (10 randomized; 16 nonrandomized) and 20 Class-B reports; the PBSC–tantalum trial remained Class C. Three exploratory pools were examined. Nontraumatic longer-term PRP–THA yielded RR 0.67 (95% CI 0.17–2.69; k = 3), and 24–30-month marrow-cell structural progression yielded RR 0.41 (0.15–1.13; k = 4). The short-term PRP analysis had particularly unstable precision (k = 2; Hartung–Knapp df = 1). Some membership/design sensitivities excluded the null, whereas primary estimates remained imprecise. Functional and pain findings were heterogeneous and frequently survivor-conditioned; harms ascertainment was sparse. Risk of bias was substantial and certainty was very low. Inference was sensitive to pool membership and analysis assumptions. This fragility, substantial risk of bias and approximate handling of bilateral dependence preclude a robust treatment-effect conclusion. The evidence does not justify routine augmentation, product ranking or a firm treatment recommendation.