OBJECTIVES: Parameter estimation in cost-effectiveness (CE) analyses often uses published evidence, which can be of varying quality. Parameters with poor quality of evidence (QoE) may lead to biases and/or model imprecision, resulting in potential mis-estimation of economically economically justifiable prices (EJP) and/or incremental cost-effectiveness ratios (ICER). We describe a method for assessing the impact of QoE for clinical parameters (e.g., overall survival, progression-free survival, time to treatment discontinuation), cost measures (e.g., treatment and monitoring-specific, adverse event [AE]-related costs), and health-state utilities on these CE outcomes, illustrating its use for a hypothetical new drug vs endocrine therapy in HR+/HER2- mBC.

METHODS: A rating scheme adapted from the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) system was used to grade QoE for each model parameter for the hypothetical drug. QoE was assessed by 2 reviewers based on 8 attributes, and summary QoE scores were assigned to each parameter based on this review. Every parameter was assigned a ‘credible interval’ based on its associated summary QoE score, ranging from ±10% of base case value for QoE rated as ‘high’ to ±50% for QoE rated as ‘poor’. Subsequently, deterministic sensitivity analyses (DSA) were conducted using upper and lower bounds based on these intervals.

RESULTS: Selected parameters, including AE disutilities and costs of terminal care, had a minimal impact on CE measures despite relatively poor QoE; while other parameters, such as pre-progression health-state utility and time to treatment discontinuation with similar QoE scores, had substantial impacts on CE outcomes.

CONCLUSIONS: The modified quality rating framework we used demonstrates how QoE can significantly affect CE outcomes. Findings may help prioritize evidence generation for CE analyses. Assigning credible intervals for each parameter rather than identical intervals for all variables might provide better guidance on the impact of variable QoE on CE outcomes.