VALUE ASSESSMENT OF PROPORTIONAL HAZARDS TESTINGS IN ONCOLOGY HTA SUBMISSIONS FOR PD-1/PD-L1 INHIBITORS - A NICE PERSPECTIVE
Author(s)
Sharma S1, Singh B1, Mangat G2, Siddiqui MK1
1Parexel International, Mohali, PB, India, 2Parexel International, Mohali, CH, India
OBJECTIVES : Clinical trials have demonstrated that PD-1/PDL-1 inhibitors are associated with delayed treatment effects and long-term survivors. This results in a survival curve, which is challenging to model due to the complex hazard function and non-proportional treatment effect. We reviewed the recent National Institute for Health and Care Excellence (NICE) submissions for PD-1/PD-L1 inhibitors in oncology to assess the extent of PH assumption testing and the statistical approaches used in the absence of PH assumption. METHODS : We searched the NICE website for the available appraisals of PD-1/PDL-1 agents for different oncology indications in the last five years. The methods for PH assessment and statistical approaches applied post-non-PH identification were extracted. We also reviewed published committee reports to determine the approach used by the agency in its assessment process. RESULTS : We identified 27 NICE appraisals for avelumab, atezolizumab, cemiplimab, durvalumab, nivolumab, and pembrolizumab across eight oncology indications. All except one submission received a positive recommendation from NICE. PH assumption was reported/assessed in 23 of the 27 appraisals, though the method of assessment varied across these submissions. A total of 10 appraisals utilized cumulative hazard plots/the log cumulative hazard plots, and five submissions reported testing using the Schoenfeld residual plots. In-line with the NICE DSU guidelines, the survival curve fitting using standard parametric models (i.e., exponential, Weibull, Gompertz, log-logistic, and log-normal) was carried out in almost all submissions with PH violation. The most common methods used for network meta-analyses (NMA) were parametric NMA using Gompertz, Weibull, and a set of 2nd order fractional polynomials. CONCLUSIONS : The current practice indicates that the PH assumption was tested consistently across PD1/PD-L1 NICE submissions. The parametric and spline-based models were the most common methods for curve fitting, and Gompertz, Weibull, and 2nd order FPs were the most common methods for conducting NMA in case of PH violation.
Conference/Value in Health Info
2020-05, ISPOR 2020, Orlando, FL, USA
Value in Health, Volume 23, Issue 5, S1 (May 2020)
Code
PCN269
Topic
Health Policy & Regulatory, Health Technology Assessment
Topic Subcategory
Approval & Labeling, Decision & Deliberative Processes, Reimbursement & Access Policy
Disease
Oncology