BACKGROUND: Acupuncture is widely used by patients with chronic pain although there is little evidence of its effectiveness. We investigated the efficacy of acupuncture compared with minimal acupuncture and with no acupuncture in patients with osteoarthritis of the knee. METHODS: Patients with chronic osteoarthritis of the knee (Kellgren grade < or =2) were randomly assigned to acupuncture (n=150), minimal acupuncture (superficial needling at non-acupuncture points; n=76), or a waiting list control (n=74). Specialised physicians, in 28 outpatient centres, administered acupuncture and minimal acupuncture in 12 sessions over 8 weeks. Patients completed standard questionnaires at baseline and after 8 weeks, 26 weeks, and 52 weeks. The primary outcome was the Western Ontario and McMaster Universities Osteoarthritis (WOMAC) index at the end of week 8 (adjusted for baseline score). All main analyses were by intention to treat. RESULTS: 294 patients were enrolled from March 6, 2002, to January 17, 2003; eight patients were lost to follow-up after randomisation, but were included in the final analysis. The mean baseline-adjusted WOMAC index at week 8 was 26.9 (SE 1.4) in the acupuncture group, 35.8 (1.9) in the minimal acupuncture group, and 49.6 (2.0) in the waiting list group (treatment difference acupuncture vs minimal acupuncture -8.8, [95% CI -13.5 to -4.2], p=0.0002; acupuncture vs waiting list -22.7 [-27.5 to -17.9], p<0.0001). After 52 weeks the difference between the acupuncture and minimal acupuncture groups was no longer significant (p=0.08). INTERPRETATION: After 8 weeks of treatment, pain and joint function are improved more with acupuncture than with minimal acupuncture or no acupuncture in patients with osteoarthritis of the knee. However, this benefit decreases over time.