Exorcising protocol-induced spirits: making the clinical trial relevant for economics

Economic evaluation frequently depends on estimates from clinical trials of both effectiveness of treatment and resource utilization accompanying it. Protocol-driven events in the trial among other influences often imply that both estimates will be inaccurate. This paper indicates how one may supplement a trial with additional data to connect the artificial trial to the real world of clinical practice. It also shows that data required for this model may be estimated from other sources (via Bayesian modeling) if they are not directly available. The required data include (for example) the proportion of patients with disease who would have presented with clinical signs if they had not been part of a trial that allowed early detection and treatment based on subclinical testing mandated by trial protocol. Those presenting with clinical signs would use additional resources for treatment and/or confirmatory diagnostics. Those with subclinical disease either 1) would never use resources in the case where they never developed clinical manifestations or 2) would use resources of a different type or intensity and perhaps have different outcomes by virtue of their disease's being discovered at a later point in time. Basing resource use and ultimate effectiveness on this revised measure of outcome rather than the one in the trial should lead to more accurate predictions for economic purposes.