Decision-making given surrogate outcomes.

Background Opinions differ about the extent to which intervention research should and can directly assess the main patient-important health outcomes, what role surrogate endpoints can play, and which requirements should then apply to the scientific underpinning of clinical and policy decisions. Method In a commentary we elaborate on this and provide guidance for dealing with related dilemmas. Conclusions Ethical, methodological and practical reasons for decision making based on surrogate endpoints can be that (1) reaching the intended patient-important health outcome would take too long to await direct RCT-based evidence, (2) experimental conditions have limited sustainability over time; and (3) the plausibility of an intervention's clinical efficacy, given the already available evidence regarding surrogate endpoints, goes beyond equipoise. Given an expected increase of interventions with a long term patient-important health outcome perspective, dealing with surrogate endpoints will remain an important challenge. Appropriately dealing with a surrogate endpoint includes (1) the assessment of its predictive value for the intended patient-important outcome, where GRADE guidelines for assessing 'indirectness' and 'causal chain analysis' can be helpful; (2) transparency of (absence of) evidence; (3) adequately updating the 'knowledge mosaic'; (4) weighing different perspectives and values, and (5) monitoring whether adjustments need to be made. The remaining level of uncertainty must be balanced against the urgency of clinical or societal decision making and the disadvantages of postponing this. Criteria for using surrogate endpoints are suggested. Patients, citizens and policy makers can be involved in agreeing upon these criteria