Outcomes were measured at baseline, 13, and 65 weeks at physiotherapy practices not involved in the trial by three trained research assistants

who were blinded to group allocation. Blinding was maintained by instructing participants not to talk about their intervention to the research assistants. Patients were included if they had osteoarthritis of the hip or knee according to the clinical Microbiology inhibitor criteria of the American College of Rheumatology (Altman et al 1986, Altman et al 1991) and were between 50 and 80 years of age. They were excluded if they had other pathology explaining the complaints; complaints in less than 10 out of 30 days; intervention for these complaints with exercise in the preceding six months; indication for hip or knee replacement within one year; contraindication for exercise; inability to understand the Dutch language; and a high level of physical functioning defined as < 2 on the walking ability and physical function sections of the Algofunctional

index (Faucher et al 2003, Lequesne et al 1987). They were recruited directly by the participating physiotherapists or in response to press releases in local newspapers (Veenhof et al 2005). Age, gender, height, weight, location of complaints, duration of complaints, and the presence of other chronic disorders were collected. X-rays of the hip and/or knee were scored by a rheumatologist according to the Kellgren Torin 1 in vivo 4-Aminobutyrate aminotransferase and Lawrence scale; it consists of five levels where 0 = no osteoarthritis, 1 = doubtful osteoarthritis, 2 = minimal osteoarthritis, 3 = moderate osteoarthritis, and

4 = severe osteoarthritis (Kellgren and Lawrence 1957, Ravaud and Dougados 1997). Pain and physical functioning were measured with the WOMAC (Bellamy et al 1988). Physiotherapists working in primary care in the Utrecht region were included in the study. They were recruited using the NIVEL National Database of Primary Care Physiotherapists. A random sample of six hundred physiotherapists from Utrecht region was invited to participate. One hundred physiotherapists responded, of whom 87 (working in 72 practices) were willing and able to participate. The experimental group received a behavioural exercise program (see Appendix 1 on the eAddenda for details). The intervention was directed at a time-effective increase in the level of activities, with the goal of integrating these activities into daily living. The intervention also included individually-tailored exercises aimed at reducing any impairment limiting the performance of these activities. The complete protocol included written materials such as education messages, activity diaries, performance charts. The intervention consisted of a maximum of 18 sessions over a 12-week period, followed by five booster sessions in Week 18, 25, 34, 42, and 55. In Week 18 and 25, participants were allowed to receive 2 sessions.