Background: This study is a secondary analysis of 12-month follow-ups from two parallel, randomized controlled trials (RCT) in painful knee osteoarthritis patients. RCT1: Total knee replacement (TKR) followed by nonsurgical treatment compared with nonsurgical treatment. RCT2: Nonsurgical treatment compared with usual care. The aims were to investigate (1) possible predictors of treatment outcome after TKR and nonsurgical interventions at 12 months, (2) associations between pain intensity and pressure pain thresholds (PPTs) (pain sensitization) at baseline and after 12 months and (3) possible gender differences. Method: Each RCT included 100 patients. Pain intensities, PPTs and number of painful sites were assessed at baseline and after 12 months. Results: In all groups, pain improved and pain sensitization decreased. In RCT1, the TKR group had the greatest improvements in pain. In RCT2 the nonsurgical group had the greatest improvement, with no between-group differences in PPTs. Lower PPTs at baseline predicted higher pain after TKR. Baseline pain intensity and PPT levels were associated with the number of painful sites. Subjects with the highest pain and lowest PPTs at baseline showed the largest relative improvement in pain and sensitization but were still experiencing highest absolute pain and lowest PPTs after 12 months (combined cohorts). Conclusion: Low PPTs at baseline predicted worse pain outcome after TKR, but did not predict outcome after nonsurgical interventions. The number of painful sites was weakly associated with pain and PPTs, and the higher pain/lower PPTs, the higher pain/lower PPTs at 12 months with females showing the lowest PPT values. Significance: Human experimental pain assessment was used to assess the degree of pain sensitization in patients with painful knee osteoarthritis. High sensitization before TKR predicted worse outcome. Outcome after nonsurgical interventions could not be predicted.