Poor replicability of recommended exercise interventions for knee osteoarthritis: a descriptive analysis of evidence informing current guidelines and recommendations

Summary Objective To examine the reporting completeness of exercise-based interventions for knee osteoarthritis (OA) in studies that form the basis of current clinical guidelines, and examine if the clinical benefit (pain and disability) from exercise is associated with the intervention reporting completeness. Design Review of clinical OA guidelines Methods We searched MEDLINE and EMBASE for guidelines published between 2006 and 2016 including recommendations about exercise for knee OA. The studies used to inform a recommendation were reviewed for exercise reporting completeness. Reporting completeness was evaluated using a 12-item checklist; a combination of the Template for Intervention Description and Replication (TIDieR) and Consensus on Exercise Reporting Template (CERT). Each item was scored ‘YES' or ‘NO' and summarized as a proportion of interventions with complete descriptions and each intervention's completeness was summarized as the percentage of completely described items. The association between intervention description completeness score and clinical benefits was analyzed with a multilevel meta-regression. Results From 10 clinical guidelines, we identified 103 original studies of which 100 were retrievable (including 133 interventions with 6,926 patients). No interventions were completely described on all 12 items (median 33% of items complete; range 17–75%). The meta-regression analysis indicated that poorer reporting was associated with greater effects on pain and no association with effects on disability. Conclusion The inadequate description of recommended interventions for knee OA is a serious problem that precludes replication of effective interventions in clinical practice. By consequence, the relevance and usability of clinical guideline documents and original study reports are diminished. PROSPERO CRD42016039742.