Presentation (CTI)

OBJECTIVES: When the proportional hazards assumption is violated for indirect treatment comparisons (ITCs) due to time-varying hazard ratios, the EU Joint Clinical Assessment recommends submissions use parametric survival curves or fractional-polynomial (fp) methods. This methodology’s relevance extends beyond this landscape. This review aims to understand how fractional-polynomial network meta-analysis (fpNMA) methods have been conducted and reported to support regulatory submissions to Canada’s Drug Agency (CDA-AMC).
METHODS: A pragmatic review of CDA-AMC submissions was conducted by three researchers using fp related search terms. Data extraction was performed using guiding questions to capture reported fpNMA methods and their impact on the reimbursement recommendation. Submissions were reviewed to identify model parameters including polynomial order, follow-up timepoints, and model effects type, alongside criterion to determine best polynomial model fit. Findings were collated to inform standard practices for clinical guidance report submissions.
RESULTS: Twenty reviews were identified, 10 of which reported using fpNMA to support evidence generation. All fpNMA applications were identified in oncology submissions with five submissions identified in non-small-cell lung cancer or renal-cell carcinoma. There was a large variability in how fpNMA results were leveraged to support submissions. Some submissions also included time-constant network meta-analyses to support the results of ITCs. Many submissions reported second order fp as the best fitting model, however little justification was provided to support these choices. The reimbursement decision did not always cite fp methods as a key limitation.
CONCLUSIONS: This research highlights the increasing use of fpNMAs across oncology CDA-AMC submissions and the variation in detail provided in dossiers. In the submissions, ITC criticisms related to the reimbursement decision, for example heterogeneous evidence bases, network structures and limited outcome comparability, were generally not about fpNMA methods.