OBJECTIVE: Discordance between radiographic and pain severity in osteoarthritis (OA) has led researchers to investigate other pain mechanisms, including neuropathic pain. Accurate identification of any neuropathic pain in hip or knee OA is important for appropriate management, but neuropathic pain prevalence is unknown. We aimed to obtain an overall prevalence estimate by systematically reviewing and meta-analysing the prevalence of neuropathic pain in people with hip or knee OA. METHOD: Observational studies which measured neuropathic pain in people aged 18 years and older with hip or knee OA were considered for inclusion. Electronic databases were searched up to February 2016. Two reviewers independently identified eligible studies and assessed methodological quality. Prevalence estimates and 95% confidence intervals were calculated using random effects meta-analytic techniques. RESULTS: Nine studies met the inclusion criteria. Study samples were from general population, hospital and community settings and all used self-report questionnaires to determine neuropathic pain. The overall prevalence estimate was 23% (95% CI: 10-39%), with considerable heterogeneity (I = 97.9%, p < 0.001). This estimate was largely unchanged with subgroup analyses based on index joint, questionnaire type, setting and consideration of other potential causes of neuropathic pain. However, the estimate for two studies that excluded other potential causes of neuropathic pain was substantially higher (32%, 95% CI: 29-35%). CONCLUSIONS: Neuropathic pain prevalence in people with knee or hip OA is considerable at 23%, and may be higher after other potential causes of neuropathic pain are excluded. Concerns regarding the validity of neuropathic pain questionnaires, selection bias, methodological quality and study heterogeneity suggest caution with interpretation of these findings. Prevalence studies using standardised criteria for neuropathic pain are required.