Asian Journal of Orthopaedic Research

Osteomyelitis continues to defeat antimicrobial regimens that appear adequate by conventional susceptibility testing, and relapse after apparently successful treatment remains a defining feature of the disease. This critical narrative review examines the biomolecular basis of that discordance and asks which mechanisms are supported by robust evidence, which remain mechanistically plausible but clinically unproven, and which have been over-interpreted. Open scholarly sources were searched from 1998 to 13 July 2026, supplemented by backward and forward citation tracking and by examination of authoritative guideline documents. Sources were appraised for study design, fidelity of the model to human bone infection, analytical transparency and the alignment between each cited claim and the underlying data. Five themes structure the synthesis. Bone behaves as a pharmacologically awkward compartment in which necrosis, thrombosis and matrix binding decouple plasma exposure from the concentration of free drug at the bacterial surface, and in which reported tissue-to-plasma ratios are derived by methods that are not interchangeable. Bacteria occupy niches that confer protection independent of genotype, including the cytoplasm of osteoblasts and osteocytes, the osteocyte lacuno-canalicular network, fibrin-enclosed abscess communities and surface-attached biofilm. Phenotypic tolerance arises through small colony variant formation, adenosine triphosphate depletion, the stringent response and host-derived oxidative stress, and is mechanistically distinct from resistance while predisposing to it. Genetic resistance interacts with these states, most consequentially for rifampicin, whose clinical value rests on a small and ageing randomised evidence base. Susceptibility testing remains calibrated to planktonic growth, and biofilm-directed indices have no validated clinical interpretation. The evidence most securely supports surgical removal of devitalised tissue, niche-aware antimicrobial selection and rifampicin combination therapy in implant-associated staphylococcal infection. Evidence is weakest where it is most often asserted, namely that intracellular and canalicular reservoirs quantitatively determine human relapse. Reconciling this requires human tissue studies, standardised tolerance assays with clinical calibration, and adequately powered trials of tolerance-directed adjuncts.