OBJECTIVES

Biopharmaceutical companies prepare for European Health Technology Assessment (HTA) submission of their products by understanding HTA guidelines and looking at past analogue examples, supplemented with primary market research. This work explores the application of advanced analytics techniques including AIML (artificial intelligence and machine learning) and their potential to improve understanding of submission factors leading to specific HTA outcomes.

METHODS

We utilized a database of all oncology product launches (including indication expansions) over the past 10 years and extracted more than 50 features relating to indication, evidence, competition and HTA. Based on these data, we built several machine learning models to predict HTA outcomes, using a random forest binary classifier: country-specific models for France and Germany, and cross-country models using all data from France, Germany, Italy and Spain. In addition to the AIML models, we conducted data analytics to understand HTA trends and specific analogs.

RESULTS

The binary classifiers to predict HTA outcomes show strong performance with an area under the curve (AUC) of 0.8 to 0.95. Using these models, scenario analyses can be conducted for specific product profiles to obtain a likelihood score of a positive HTA outcome in France, Germany, Italy and Spain. The most important features used by the model to make predictions relate primarily to evidence and indication. Additional analyses of the database allow for insights into the characteristics of products associated with specific HTA outcomes.

CONCLUSIONS

This pilot project shows the potential of AIML to supplement and enhance traditional approaches to HTA submission strategies by leveraging all previous submissions in the therapy area. AIML models can capture non-linear relationships among a large number of features to produce quantitative outputs on the likelihood of favorable HTA outcomes. Such an approach has the potential to radically change how companies successfully prepare for HTA evaluation in the future.