Immunohistochemistry (IHC) is the most widely used ancillary technique in surgical pathology and the practical foundation of biomarker-directed cancer care, yet adoption remains uneven between and within health systems and the analytical quality of the results produced is more variable than clinicians generally assume. This narrative review synthesises peer-reviewed literature, professional society guidance, external quality assessment (EQA) scheme reports, and global health policy documents published to December 2021, supplemented by a consolidated cross-disciplinary reference set covering laboratory infrastructure, procurement, workforce development, governance, and informatics. Barriers clustered into ten interacting domains spanning capital and recurrent cost, supply chain and equipment sustainability, workforce scarcity and skill mix, pre-analytical variability, analytical validation and standardisation, interpretation and reporting, quality assurance and regulatory constraints, utilisation and tissue stewardship, informatics, and pandemic-related service disruption. Long-running EQA data indicate that a substantial minority of submitted stains have been judged insufficient for diagnostic use, and that approximately nine in ten of those failures represent weak or falsely negative staining, so the characteristic patient-level harm is effective therapy withheld rather than ineffective therapy given. Comparability studies of programmed death ligand 1 assays further show that assay and algorithm selection alone can change patient classification. Reported mitigations converged on tiered antibody menus, hub-and-spoke service networks, written pre-analytical specifications, multi-tissue on-slide controls containing low expressors, formal analytical validation, mandatory EQA with documented corrective action, technologist-focused training, and deliberate procurement. Adoption is therefore constrained more by organisational systems than by technology: availability of IHC is not equivalent to analytical adequacy, and a service reporting predictive biomarkers without demonstrated analytical sensitivity plausibly causes more harm than one that refers.