To evaluate the efficacy, safety and carry-over effect of diacerein, in comparison to piroxicam, in the treatment of Thai patients with symptomatic knee osteoarthritis (OA). This was a double-blind, randomised, piroxicam-controlled, parallel-group study. A 7-day non-steroidal anti-inflammatory drug washout period was followed by a 16-week treatment period with either diacerein 100mg/day or piroxicam 20mg/day, and an 8-week treatment-free observation period. The primary efficacy criterion was pain on Western Ontario and McMaster University Osteoarthritis (WOMAC) A. The secondary criteria included WOMAC B, C and total WOMAC, paracetamol intake, Short Form-36 questionnaire and global judgements on efficacy and tolerability by patients and investigators. Of 171 randomised patients, 150 completed the study and 161 were analysed in the intent-to-treat population (diacerein: 82, piroxicam: 79). Pain (WOMAC A) decreased to a similar extent in both groups at Week 16 (diacerein: -69.7%+/-31.5%; piroxicam: -74.1+/-26.2%; P=n.s.). On treatment discontinuation, pain increased in the piroxicam group at Weeks 20 (-47%+/-47.8%) and 24 (-26.8%+/-60.6%) while improvements persisted in the diacerein group at Weeks 20 (-66.9%+/-35.9%) and 24 (-69.5%+/-33.7%), with a significant difference in favour of diacerein at Weeks 20 and 24, demonstrating the carry-over effects of the drug. The incidence of adverse events was similar in both groups but more patients from the piroxicam group dropped out of the study due to these events. Diacerein was as effective as piroxicam in reducing pain and improving function but, unlike piroxicam, displayed a carry-over effect and a better safety profile.