Moving Beyond Partitioned Survival Models in Oncology Submissions: Incorporation of Canada's Drug Agency Guidance on Model Structure
Author(s)
Katherine Park, B.A., M.P.H. (Candidate), Aashna Shah, B.Sc., M.Sc., Akshaya Jayakarunakaran, B.Sc., Molly Atkinson, B.A., M.Sc;
Costello Medical, Boston, MA, USA
Costello Medical, Boston, MA, USA
Presentation Documents
OBJECTIVES: To assess how 2023 Canada's Drug Agency (CDA) guidance, which highlights limitations of partitioned survival models (PSMs), has influenced model structure choice in oncology health technology assessment submissions.
METHODS: An up-to-date list of Reimbursement Review Reports from the CDA website was extracted as of November 26 2024. “Complete” reports for oncology drugs submitted after May 1 2023 (when the guidance was released), were downloaded. An equal number of reports submitted prior to this date were used for comparison of reports pre- and post-published guidance. Assessment details (including disease area, intervention and comparators, and CDA recommendation), the sponsor’s model structure, whether it was considered adequate to address the decision problem (used as a proxy for acceptance of model structure), and any additional comments relating to model structure by the CDA were extracted.
RESULTS: Fifty CDA submissions were reviewed (25 before and 25 after the threshold date). Two non-sponsored submissions (pre-May 2023) without manufacturer models were excluded. Of the remaining 48 submissions, 29/48 (60%), 4/48 (8%), and 15/48 (31%) were PSMs, (semi-)Markov’s, and other model structures, respectively. Pre-2023 guidance, 16/23 (70%) and 3/23 (13%) were PSMs and (semi-)Markov’s, respectively, of which 6/16 (38%) and 0/3 (0%) were accepted. Post-guidance, 13/25 (52%) and 1/25 (4%) were PSMs and (semi-)Markov’s, respectively, with 7/13 (54%) and 1/1 (100%) accepted. In their comments, the CDA has consistently highlighted limitations of PSMs, including independence of progression-free and overall survival, and challenges in modeling subsequent therapies, suggesting that alternative modeling approaches may be more appropriate in some circumstances.
CONCLUSIONS: From the reviewed submissions, there was not an increase in the use of (semi-)Markov structures following the 2023 guidance; however, sample size may limit interpretation. Comments from the CDA align with their guidance, indicating that model structure should be carefully considered based on the disease area being modeled.
METHODS: An up-to-date list of Reimbursement Review Reports from the CDA website was extracted as of November 26 2024. “Complete” reports for oncology drugs submitted after May 1 2023 (when the guidance was released), were downloaded. An equal number of reports submitted prior to this date were used for comparison of reports pre- and post-published guidance. Assessment details (including disease area, intervention and comparators, and CDA recommendation), the sponsor’s model structure, whether it was considered adequate to address the decision problem (used as a proxy for acceptance of model structure), and any additional comments relating to model structure by the CDA were extracted.
RESULTS: Fifty CDA submissions were reviewed (25 before and 25 after the threshold date). Two non-sponsored submissions (pre-May 2023) without manufacturer models were excluded. Of the remaining 48 submissions, 29/48 (60%), 4/48 (8%), and 15/48 (31%) were PSMs, (semi-)Markov’s, and other model structures, respectively. Pre-2023 guidance, 16/23 (70%) and 3/23 (13%) were PSMs and (semi-)Markov’s, respectively, of which 6/16 (38%) and 0/3 (0%) were accepted. Post-guidance, 13/25 (52%) and 1/25 (4%) were PSMs and (semi-)Markov’s, respectively, with 7/13 (54%) and 1/1 (100%) accepted. In their comments, the CDA has consistently highlighted limitations of PSMs, including independence of progression-free and overall survival, and challenges in modeling subsequent therapies, suggesting that alternative modeling approaches may be more appropriate in some circumstances.
CONCLUSIONS: From the reviewed submissions, there was not an increase in the use of (semi-)Markov structures following the 2023 guidance; however, sample size may limit interpretation. Comments from the CDA align with their guidance, indicating that model structure should be carefully considered based on the disease area being modeled.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
HTA93
Topic
Health Technology Assessment
Topic Subcategory
Decision & Deliberative Processes
Disease
SDC: Oncology