Abstract Objective. The data are conflicting for the association between the index-to-ring finger length ratio (2D:4D) and the risk of OA. The aim of this cohort study was to examine the relationship between 2D:4D and the risk of severe knee and hip OA requiring total joint replacement. Methods. A total of 14 511 participants in the Melbourne Collaborative Cohort Study had 2D:4D assessed from hand photocopies. The incidence of total knee replacement and total hip replacement between 2001 and 2011 was determined by linking the cohort records to the Australian Orthopaedic Association National Joint Replacement Registry. Results. Over an average 10.5 years of follow-up, 580 participants had total knee replacement and 499 had total hip replacement. Greater right 2D:4D [hazard ratio (HR) 0.91 for a S.D. increase in 2D:4D, 95 CI 0.84, 0.99, P = 0.03] and average right and left 2D:4D (HR 0.91 for a S.D. increase in 2D:4D, 95 CI 0.84, 0.99, P = 0.02) were associated with a reduced incidence of total knee replacement. These associations persisted when participants whose fingers had any features that might have affected the validity of 2D:4D measurements were excluded. No significant associations were observed between 2D:4D and the incidence of total hip replacement. Conclusion. A lower 2D:4D is associated with an increased risk of severe knee OA requiring total knee replacement, but not the risk of severe hip OA. The underlying mechanisms for the association warrant further investigation.