The aim of this paper is to describe the treatment and physician characteristics in a randomised trial of acupuncture for osteoarthritis of the knee.Three-armed, randomised, controlled multicentre trial with 1-year follow-up.Twenty-eight outpatient centres in Germany.A total of 294 patients with osteoarthritis of the knee were randomised to 12 sessions of semi-standardised acupuncture (at least 6 local and 2 distant points needled per affected knee from a selection of predefined points, but individual choice of additional body or ear acupuncture points possible), 12 sessions of minimal acupuncture (superficial needling of at least 8 of 10 predefined, bilateral, distant non-acupuncture points) or a waiting list control (2 months no acupuncture).Participating trial physicians and interventions.Forty-seven physicians specialised in acupuncture (mean age 43±8 years, 26 females) provided study interventions in 28 outpatient centres in Germany. The median duration of acupuncture training completed by participating physicians was 350h (range 140–2508). The total number of needles used was 17.4±4.8 in the acupuncture group compared to 12.9±3.3 in the minimal acupuncture group. In total, 39 physicians (83%) stated that they would have treated patients in either a similar or in exactly the same way outside of the trial, whereas 7 (15%) stated that they would have treated patients differently (1 missing).Our documentation of the trial interventions shows that semi-standardised acupuncture strategy represents an acceptable compromise for efficacy studies. However, a substantial minority of participating trial physicians stated that they would have treated patients differently outside of the trial.