Assessing the Impact of Survival Data Maturity in Cost-Effectiveness Estimates in Advanced Non-Small Cell Lung Cancer

OBJECTIVES: A requirement in health technology assessment (HTA) concerns extrapolation of outcomes beyond randomized clinical trial (RCT) duration which is a source of uncertainty in cost-effectiveness results. We investigate the impact of data maturity on the estimation of survival and cost-effectiveness outcomes by examining RCT data from patients with non-small cell lung cancer (NSCLC) undergoing second or subsequent lines of treatment (2L+).

METHODS: A targeted review was conducted in IQVIA’s HTA accelerator to identify two-arm, phase 3 clinical trials in 2L+ NSCLC that reported results for at least two data cuts. The overall survival (OS) and progression-free survival (PFS) data were used in a partitioned survival model. The comparison of data cuts was conducted on two distinct levels. Initially, we examined how data maturity affected long-term extrapolations and which survival models were more appropriate for model selection. We then assessed how the cost-effectiveness estimates were influenced by differences in survival outcomes between data cuts, and different extrapolation models.

RESULTS: Four RCTs were selected. Mean OS and PFS were mostly increased using data from the second data cut but there were exceptions, depending on the model selection, suggesting the absence of a definitive trend. The between data cut difference in extrapolation results was higher in immunotherapies compared to chemotherapies and targeted therapies. Across all RCTs, the ICER increased in the second data cut under different survival models. These findings warrant further investigation to uncover an underlying relationship between data maturity and ICER, as the increase was not consistent between scenarios.

CONCLUSIONS: More mature data resulted in increased ICERs in the identified RCTs for 2L+ NSCLC. Limitations such as the impact of treatment switching, adverse events, time to treatment discontinuation data and indirect comparisons with relevant real-world comparators were not considered. Therefore, the generalizability of these findings warrants further investigation.