Generalised anxiety disorder (GAD) is common, causing unpleasant symptoms and impaired functioning. The National Institute for Health and Care Excellence (NICE) guidelines have established good evidence for low-intensity psychological interventions, but a significant number of patients will not respond and require more intensive step 3 interventions, recommended as either high-intensity cognitive behavioural therapy (CBT) or a pharmacological treatment such as sertraline. However, there are no head-to-head comparisons evaluating which is more clinically effective and cost-effective, and current guidelines suggest that treatment choice at step 3 is based mainly on patient preference. To assess clinical effectiveness and cost-effectiveness at 12 months of treatment with the selective serotonin reuptake inhibitor (SSRI) sertraline compared with CBT for patients with persistent GAD not improved with NICE-defined low-intensity psychological interventions. Participant randomised trial comparing treatment with sertraline with high-intensity CBT for patients with GAD who had not responded to low-intensity psychological interventions. Community-based recruitment from local Improving Access to Psychological Therapies (IAPT) services. Four pilot services located in urban, suburban and semirural settings. People considered likely to have GAD and not responding to low-intensity psychological interventions identified at review by IAPT psychological well-being practitioners (PWPs). Those scoring ≥ 10 on the Generalised Anxiety Disorder-7 (GAD-7) anxiety measure were asked to consider involvement in the trial. Aged ≥ 18 years, a score of ≥ 10 on the GAD-7, a primary diagnosis of GAD diagnosed on the Mini International Neuropsychiatric Interview questionnaire and failure to respond to NICE-defined low-intensity interventions. Inability to participate because of insufficient English or cognitive impairment, current major depression, comorbid anxiety disorder(s) causing greater distress than GAD, significant dependence on alcohol or illicit drugs, comorbid psychotic disorder, received antidepressants in past 8 weeks or high-intensity psychological therapy in previous 6 months and any contraindications to treatment with sertraline. Consenting eligible participants randomised via an independent, web-based, computerised system. (1) The SSRI sertraline prescribed in therapeutic doses by the patient's general practitioner for 12 months and (2) 14 (± 2) CBT sessions delivered by high-intensity IAPT psychological therapists in accordance with a standardised manual designed for GAD. The primary outcome was the Hospital Anxiety and Depression Scale - Anxiety component at 12 months. Secondary outcomes included measures of depression, social functioning, comorbid anxiety disorders, patient satisfaction and economic evaluation, collected by postal self-completion questionnaires. Only seven internal pilot participants were recruited against a target of 40 participants at 7 months. Far fewer potential participants were identified than anticipated from IAPT services, probably because PWPs rarely considered GAD the main treatment priority. Of those identified, three-quarters declined participation; the majority (30/45) were reluctant to consider the possibility of randomisation to medication. Poor recruitment was the main limiting factor, and the trial closed prematurely. It is unclear how much of the recruitment difficulty was a result of conducting the trial within a psychological therapy service and how much was possibly a result of difficulty identifying participants with primary GAD. It may be easier to answer this important question by recruiting people from primary care rather than from those already engaged in a psychological treatment service. Current Controlled Trials ISCRTN14845583. This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 21, No. 45. See the NIHR Journals Library website for further project information.