Presentation (CTI)

OBJECTIVES: Our objective is to illustrate how machine learning and microsimulation models can be used by firms to identify the revenue-maximizing indication. Broader indications will dilute the average effect of the technology and thus lower the price. This is the classic monopolist trade-off between volume and price. This is intensified in the context of international reference pricing, where broader indications lower domestic prices (e.g., in the UK) and erode revenue in linked international markets, such as the USA via most-favoured-nation policies.
METHODS: The firm can characterise patient level heterogeneity by integrating a decision analytic model with causal machine learning (Glynn et al 2023). First fit a flexible machine learning model to the Phase II/III trial data to estimate the relative treatment effect on the primary clinical endpoint as a function of baseline covariates. Then feed these tailored effect estimates into a microsimulation model to simulate individual level lifetime incremental quality-adjusted life years (ΔQALYs) (cost savings/increases ignored for simplicity). Ordering patients by ΔQALYs, the average effectiveness declines as the proportion treated increases. Price is a function of the average effectiveness. Revenue is the total size of the indication multiplied by the proportion treated and the price. The firm can then choose the proportion which maximizes revenue.
RESULTS: We illustrate our approach using a hypothetical drug for diabetes type 2 and the UKPDS model. Heterogeneity in ΔQALYs arises due to heterogeneity in 1) treatment effects and 2) baseline risks of clinical outcomes. Revenue was maximized by expanding treatment to all individuals who benefit while strictly excluding those harmed by the therapy. Including international reference pricing restricted domestic indications further.
CONCLUSIONS: Recent advances in machine learning and microsimulation models mean that firms can better understand treatment effect heterogeneity and use this guide indication selection. This is particularly important with international reference pricing.