BackgroundAnti-nerve growth factor (NGF) has shown promise for osteoarthritis (OA) pain efficacy, but an unanticipated joint safety signal occurred in trials. To what extent this was related to OA natural history, or was a consequence of anti-NGF agents is unclear. Of the adverse joint safety events identified, osteonecrosis has specific diagnostic codes available in the medical record to enable assessment of its frequency in the general community. We therefore investigated the rates of knee osteonecrosis in three real-world cohorts using two data sources to place these trial data in context. MethodsWe used data from UK-based IQVIA Medical Research Data (IMRD) of adults diagnosed with incident knee OA between 2000 and 2018 to examine the incidence of knee osteonecrosis using different definitions. Additionally, we evaluated the incidence of knee osteonecrosis in the year prior to knee replacement in IMRD and among US Medicare beneficiaries who received a knee replacement in 2011–2014. ResultsIn IMRD, among 122,343 subjects with incident knee OA (mean age 68 years, 58% female), incidence estimates for knee osteonecrosis were 0.006–0.10%, with incidence rates of 0.01–0.17 per 1000 person-years. Among the 81,807 who had a knee replacement, the incidence of knee osteonecrosis in the year prior to knee replacement was 0.004–0.06%. In Medicare, among 316,593 with knee replacement (mean age 74, 68% female), the incidence of knee osteonecrosis was 0.24–0.7%. ConclusionKnee osteonecrosis is rare among people with knee OA, including in the year prior to knee replacement. These data provide context for interpreting osteonecrosis events in NGF trials.