It is generally agreed that the goal of resuscitation is survival with neurological and physiological status similar to that preceding the cardiac arrest. Previously I have argued that the lack of improvement in outcome from resuscitation over the past 3 to 4 decades, as compared to the substantial progress made in treatment of ischemic heart disease, is a consequence of the absence of randomized clinical trials of new interventions and the use of intermediate endpoints such as return of spontaneous circulation or admittance to hospital. Proponents of these intermediate endpoints have argued that those involved in the resuscitation have no control over what care is undertaken in the hospital and hence hospital mortality only adds noise, at best, thus making survival a less sensitive and less relevant endpoint for evaluation of resuscitation interventions. Recent reports of improvement in hospital survival have caused me to consider that their argument may have more validity than I had supposed. In this note I propose a test that gives weight both to the intermediate endpoint and survival. The test is responsive to the primary goal of testing survival with limited loss of power compared to a test based only on the intermediate endpoint. The test is illustrated with several examples.



Included in

Biostatistics Commons