BACKGROUND: Predicting individualized treatment effects is of great importance, so that a treatment will be prescribed to individuals who will benefit from it and not to those who will not benefit or even get harmed. Decision curve analysis (DCA) estimates the net benefit of competing treatment strategies and can be used to evaluate the clinical usefulness of a model that predicts who will benefit from a treatment. DCA is well described for models that compare two treatments using data from a single randomized clinical trial. However, methods related to decisions between multiple treatments using data from multiple independent trials are not yet available.

OBJECTIVES: Hence the main objective of this work is to fill-in this methodological gap by extending DCA into a network meta-analysis framework (NMA), where several treatment options are compared, via synthesizing several trials.

METHODS:We explain the framework by breaking it down into 12 steps DCA method to compare treatment choice based on an NMA-prediction model with “treat-all with a specific drug” strategies. We also exemplify our methodology on an individualized prediction model for deciding between three treatments (Natalizumab, Dimethyl Fumarate, Glatiramer Acetate) and placebo, for patients diagnosed with relapsing-remitting multiple sclerosis.

RESULTS: In many circumstances, the individualized prediction model guides better the clinical decision-making than other “treat-all with a specific drug” strategies, such as the “treat all with Natalizumab” strategy.

CONCLUSIONS: As individualized prediction models comparing different treatments are become increasingly used in personalized medicine, our methodology help assessing the impact of such models in clinical practice.

This project is funded by the European Union’s Horizon 2020 research and innovation program under grant agreement No 825162.