OBJECTIVES: Performance-based payment arrangements for innovative drugs seek to allocate financial risk of interventions between manufacturers and payers. Given operational challenges in monitoring real-world outcomes, effective risk sharing mechanisms may require prediction of incremental survival/quality-adjusted life years (QALYs) conditional on surrogate markers, for example, complete remission (CR) or partial remission (PR). In accepting a risk-sharing arrangement that reimburses only for remitters, payers should minimize the “false-positive” (FP) risk: early remitters who subsequently have limited QALYs. Manufacturers should minimize the “false-negative” (FN) risk: early non-remitters who subsequently have prolonged QALYs. A Bayesian decision framework can be used to choose among multiple likelihood (predictive) priors to optimize the posterior economic risk trade-off for payers and manufacturers. METHODS: A Bayesian decision-analytic, hypothetical data-based, cost-effectiveness model was developed. Prior probabilities and QALYs were assigned for 6-, 3-, and 1-month survival, as were treatment costs. A plausible prior likelihood (predictive) structure represented the (ROC) relationship between the sensitivity and specificity of CR/PR in predicting survival. Expected (posterior) probabilities of survival, conditional on CR/PR, were generated. At a threshold of $50,000/QALY, the cost-effectiveness of the intervention, conditional on achieving CR/PR, and an optimal sensitivity-specificity trade-off was derived. RESULTS: At a hypothetical treatment cost of $5,000/month for a 4-month cycle, a minimal FP of 13% (maximum specificity of 87%) and a minimal FN of 33% (maximum sensitivity of 67%) emerged as necessary to be accepted by payers and manufacturers respectively to ensure viable risk-sharing. At higher sensitivity, payer risk did not meet the reimbursement threshold, while at higher specificity, manufacturers would assume excessive financial risk. Other illustrations will be discussed. CONCLUSIONS: Manufacturers should propose evidence-based payment arrangements that utilize clinical trial data to develop economic implications of being at various points on the ROC curve in order to optimize the trade-offs between payer and manufacturer incentives.