OBJECTIVES

To assess the feasibility and usefulness of value of information analyses at early-stage development by investigating parameters influencing uncertainty in decision-making when data are limited, using an internally-developed, R/Shiny indication-agnostic oncology modeling platform.

METHODS

: A case study in pre-treated small cell lung cancer was conducted, assessing the cost-effectiveness of nivolumab against IV topotecan (most prescribed treatment), using only publicly-available data. PFS and OS outcomes were digitized from recently published single-arm CheckMate-032 and , respectively, and extrapolated over a 20-year time horizon using parametric survival models. Time on treatment (ToT) was assumed equal to PFS. The expected value of perfect and partial perfect information (EVPI, EVPPI) were calculated using the BCEA package in the model platform.

RESULTS

: The value of information (VOI) analysis was performed and was replicated outside the platform environment.the per-patient EVPI was $4,330. Based on the info-rank statistic, measuring the proportion of individual-parameter EVPPI over total EVPI, nivolumab PFS/ToT distributional parameters contributed the most to decision uncertainty (log-logistic shape: 56%, scale: 47%), followed by body weight (44%) and nivolumab OS (log-normal shape: 40%, scale: 33%). The 5-parameter info-rank was 96%, while single-parameter EVPPIs of other model inputs were less than 16%. At a conventional acceptability threshold, the 5-parameter EVPPI was $94.

CONCLUSIONS

The VOI analysis identified nivolumab /ToT and body weight, determinants of treatment duration and cost, as the greatest contributors to decision uncertainty, therefore suggesting prioritizing reducing uncertainty associated with those parameters. In the case study, this might entail evaluating flat-dose nivolumab administration and modeling ToT separately to improve predictive accuracy. The model platform allows running EVPPI analyses efficiently for any oncology indication, even when based only on a naïve comparison of publicly-available data, to support strategy development and prioritize evidence generation.