Problems in conducting economic evaluations alongside clinical trials

Background Case management has become the statutory basis of community care in the UK for people with long-term mental disorders, although a randomised controlled trial found no important improvements over standard care. Here we compare the costs and cost consequences of this intervention with standard care. Method Resource-use data were collected over a six-month baseline period and for 14 months after randomisation on all patients in the trial. Results At 14 months the ratio of control group to treatment group weekly costs was 1.09 (95% CI 0.86–1.38) for total costs; 1.12 (0.76–1.65) for state benefits, and 1.21 (0.61–2.42) for health care costs. Costs were thus lower in the treatment group, but these differences were not significant. Conclusions Retrospective power calculations indicated that the trial could have detected differences of 30% in total cost, but would have required 700 patients per arm to detect a 20% difference in health care costs. Hence this study, which had adequate power to detect clinically meaningful differences, was found to be far too small to detect large differences in costs. Funding agencies increasingly request that clinical trials include economic alongside clinical end-points: these findings may have important lessons for that policy.