BACKGROUND: Musculoskeletal problems generate high costs. Of these disorders, patients with knee problems are commonly seen by GPs. Magnetic resonance imaging (MRI) of the knee is an accurate diagnostic test, but there is uncertainty as to whether GP access to MRI for these patients is a cost-effective policy. AIM: To investigate the cost-effectiveness of GP referral to early MRI and a provisional orthopaedic appointment, compared with referral to an orthopaedic specialist without prior MRI for patients with continuing knee problems. Design of study: Cost-effectiveness analysis alongside a pragmatic randomised trial. SETTING: Five-hundred and thirty-three patients consulting their GP about a knee problem were recruited from 163 general practices at 11 sites across the UK. METHOD: Two-year costs were estimated from the NHS perspective. Health outcomes were expressed in terms of quality-adjusted life years (QALYs), based on patient responses to the EQ-5D questionnaire administered at baseline, and at 6, 12, and 24 months' follow-up. RESULTS: Early MRI is associated with a higher NHS cost, by pound 294 ($581; euro 435) per patient (95% confidence interval [CI] = pound 31 to pound 573), and a larger number of QALYs, by 0.05 (95% CI = 0.025 to 0.118). Mean differences in cost and QALYs generated an incremental cost per QALY gained of pound 5840 ($11,538; euro 8642). At a cost per QALY threshold of pound 20,000, there is a 0.81 probability that early MRI is a cost-effective use of NHS resources. CONCLUSION: GP access to MRI for patients presenting in primary care with a continuing knee problem represents a cost-effective use of health service resources.