Presentation (CTI)

OBJECTIVES: This analysis assesses the accuracy of parametric model predictions used in Health Technology Assessment (HTA) submissions for larotrectinib, a selective TRK inhibitor approved for patients with TRK fusion cancer. It compares these predictions with observed overall survival (OS) and investigator-assessed progression-free survival (PFS) from integrated trials with more recent data read-outs and longer follow-up. Understanding prediction accuracy is crucial for effective clinical and policy decision-making.
METHODS: Larotrectinib received preliminary regulatory approval in 2018 and full FDA approval in 2025. HTA submissions to NICE and other agencies were based on the 2018 dataset and comprised n=102 patients. The sponsor used a Weibull distribution to extrapolate OS and PFS in a cost-effectiveness model. Prediction accuracy was evaluated by comparing model outputs to 95% confidence intervals (CIs) of observed outcomes from the 2024 data read-out, reflecting a 61.6-month increase in median survival follow-up. Timepoints were set at 6 years for PFS and 8 years for OS, corresponding to 10% patients risk milestones, and used to evaluate the accuracy of model predictions against observed data.
RESULTS: At the 8-year timepoint, the 2018 Weibull model predicted 48% OS, while the 2024 observed OS was 53% - an underestimation of 5 percentage points, though still within the 95% CI. The median OS was not reached, but the model estimated it to be 90 months. At the 6-year timepoint, the model projected 15% PFS, compared to 30% observed - an underestimation of 15 percentage points, outside the 95% CI.
CONCLUSIONS: While the Weibull model provided reasonably projected OS, it underestimated PFS. The model's median OS estimate is considered conservative, given the underestimation of OS prediction. These results highlight the need for improved modelling approaches in HTA submissions to better inform clinical and policy decisions.