OBJECTIVES: A Markov Model assessed the cost-effectiveness of BGF 320/18/9.6 μg versus FF/UMEC/VI 100/62.5/25 μg. Health states were defined based on FEV1 severity (moderate, severe, and very severe). Mortality data for FF/UMEC/VI was only available at the aggregated level from a matching adjusted indirect treatment comparison (MAIC) and required decomposition into mortality associated with FEV1 health states. We aimed to estimate standardized mortality estimates per latent FEV1 health state adjusting for general population mortality from aggregated mortality data to run a Markov model for 5-year time horizon evaluation.

METHODS: All-cause mortality was obtained from UK life tables, accounting for higher mortality rates in males than females. The cumulative mortality curves from MAIC were digitized to extract the cumulative probability of death over-time at 1-month intervals as per model cycle length. A hazard ratio was calibrated using the baseline distribution in different FEV1 states (i.e., 29%, 61%, and 10%) and excess relative risk of dying (i.e., 1.4, 2.6, and 2.6) for FF/UMEC/VI by minimizing the mean square error (MSE). For mortality beyond one year, the 12-month relative hazard was assumed to remain constant or increase linearly to match the FF/UMEC/VI hazard at 5 years.

RESULTS: At one year, the decomposed mortality by moderate, severe, very severe health states for BGF (0.86%, 1.59%, and 1.59%, respectively) was lower compared to FF/UMEC/VI (1.41%, 2.61%, and 2.61%, respectively). BGF was found to increase the life-years by 7.5% over 5 years when assuming a constant rate and by 5.1% when assuming a linear decrease in treatment effect from 1 to 5 years compared to FF/UMEC/VI.

CONCLUSIONS: The methodology to decompose overall mortality into health state-specific mortality, accounting for general population mortality, was successfully implemented. This approach can be applied in health economic models to predict health state specific mortality from overall mortality accounting for general population mortality.