Surrogate endpoints: a basis for a rational approach.

In clinical trials, the clinical endpoint is often replaced by an intermediate endpoint, known in some instances as a "surrogate" endpoint. The reasons for the substitution are often both practical and financial. At present, no theoretical basis or practical guidelines exist to help in the choice of surrogate endpoints. An approach is proposed here, based on three provisos which can be verified using one of a series of equations, if sufficient data on the pathophysiology and epidemiology of the disease are available. It is shown that even a strong statistical correlation is not a sufficient criterion for the definition of a surrogate endpoint. It is apparent that results obtained with the commonly used "surrogate" endpoints should be cautiously considered, and that the assessment of treatments should, when possible, be based on clinical rather than intermediate endpoints.