Background. Collaborative care is an effective intervention for depression which includes both organizational and patient-level intervention components. The effect in the UK is unknown, as is whether cluster-or patient-randomization would be the most appropriate design for a Phase III clinical trial.Method. We undertook a Phase II patient-level randomized controlled trial in primary care, nested within a clusterrandomized trial. Depressed participants were randomized to ' collaborative care' -case manager-coordinated medication support and brief psychological treatment, enhanced specialist and GP communication -or a usual care control. The primary outcome was symptoms of depression (PHQ-9).Results. We recruited 114 participants, 41 to the intervention group, 38 to the patient-randomized control group and 35 to the cluster-randomized control group. For the intervention compared to the cluster control the PHQ-9 effect size was 0.63 (95 % CI 0.18-1.07). There was evidence of substantial contamination between intervention and patient-randomized control participants with less difference between the intervention group and patient-randomized control group (x2.99, 95 % CI x7.56 to 1.58, p=0.186) than between the intervention and cluster-randomized control group (x4.64, 95 % CI x7.93 to x1.35, p=0.008). The intra-class correlation coefficient for our primary outcome was 0.06 (95 % CI 0.00-0.32).Conclusions. Collaborative care is a potentially powerful organizational intervention for improving depression treatment in UK primary care, the effect of which is probably partly mediated through the organizational aspects of the intervention. A large Phase III cluster-randomized trial is required to provide the most methodologically accurate test of these initial encouraging findings.