BACKGROUND Health anxiety has been treated by therapists expert in cognitive behaviour therapy with some specific benefit in some patients referred to psychological services. Those in hospital care have been less often investigated. Following a pilot trial suggesting efficacy we carried out a randomised study in hospital medical clinics. METHODS We undertook a multicentre, randomised trial on health anxious patients attending cardiac, endocrine, gastroenterological, neurological, and respiratory medicine clinics in secondary care. We included those aged 16-75 years, who satisfied the criteria for excessive health anxiety, and were resident in the area covered by the hospital, were not under investigation for new pathology or too medically unwell to take part. We used a computer-generated random scheme to allocate eligible medical patients to an active treatment group of five-to-ten sessions of adapted cognitive behaviour therapy (CBT-HA group) delivered by hospital-based therapists or to standard care in the clinics. The primary outcome was change in health anxiety symptoms measured by the Health Anxiety Inventory at 1 year and the main secondary hypothesis was equivalence of total health and social care costs over 2 years, with an equivalence margin of £150. Analysis was by intention to treat. The study is registered with controlled-trials.com, ISRCTN14565822. FINDINGS Of 28 991 patients screened, 444 were randomly assigned to receive either adapted cognitive behaviour therapy (CBT-HA group, 219 participants) or standard care (standard care group, 225), with 205 participants in the CBT-HA group and 212 in the standard care group included in the analyses of the primary endpoints. At 1 year, improvement in health anxiety in the patients in the CBT-HA group was 2·98 points greater than in those in the standard care group (95% CI 1·64-4·33, p<0·0001), and twice as many patients receiving cognitive behaviour therapy achieved normal levels of health anxiety compared with those in the control group (13·9% vs 7·3%; odds ratio 2·15, 95% CI 1·09-4·23, p=0·0273). Similar differences were observed at 6 months and 2 years, and there were concomitant reductions in generalised anxiety and, to a lesser extent, depression. Of nine deaths, six were in the control group; all were due to pre-existing illness. Social functioning or health-related quality of life did not differ significantly between groups. Equivalence in total 2-year costs was not achieved, but the difference was not significant (adjusted mean difference £156, 95% CI -1446 to 1758, p=0·848). INTERPRETATION This form of adapted cognitive behaviour therapy for health anxiety led to sustained symptomatic benefit over 2 years, with no significant effect on total costs. It deserves wider application in medical care. FUNDING National Institute for Health Research Health Technology Assessment Programme.