OBJECTIVES: Multi-criteria decision-analysis (MCDA) is an important benefit-risk assessment of medical products by allowing comparisons of competing interventions based multiple criteria via a single utility analysis. However, existing MCDA methods can be limited, as they often require criteria such as time-to-event (e.g. overall survival; OS) and ordinal outcomes (e.g., tumor growth measured via RECIST) to be dichotomized. Here, we present Bayesian multi-criteria augmented decision analysis (MCADA), a new framework that can handle outcomes in their natural forms.

METHODS: Based on a patient-level data of randomized clinical trial conducted for extensive-disease small cell lung cancer (NCT01439568), we performed simulation-guided benefit-risk assessments based on OS, RECIST, and treatment-related grade 3/4 adverse events (AEs). Our simulation compared MCADA with these criteria at their natural forms to probabilistic MCDA and stochastic multi-criteria acceptability analysis (SMAA), two existing MCDA approaches that require these variables to be dichotomized. Treatment performance on OS was assumed to be twice as important as the RECIST, and RECIST was considered as twice as important as AEs (utility weights of 0.57, 0.29, and 0.14, respectively). We calibrated the posterior superiority thresholds to control the expected type I error rate at 10%.

RESULTS: Our simulations showed that MCADA generally had considerably higher statistical power compared to both probabilistic MCDA and SMAA. In the scenario where the treatment was superior to the control in all three parameters (0.90 hazard ratio for OS, 0.85 odds ratio for RECIST, and 0.90 relative risk for AEs), MCADA had the highest power of 95.6% followed by 88.9% for probabilistic MCDA and 70.4% for SMAA.

CONCLUSIONS: MCADA is an important extension of existing MCDA methods. It can handle multiple parameters in their natural forms, including time-to-event and ordinal outcomes. MCADA provides clarity when decision-makers must weigh multiple risks and benefits, facilitating transparent and justifiable decision-making.