Adjusting to Nonoverlapping Variables in the Unanchored Matching-Adjusted Indirect Comparisons of Survival Outcomes: Application of Monte Carlo Simulations
Author(s)
Olga Mironenko, PhD1, Andrei Lazarev, MS2, Kirill Sapozhnikov, PhD2, Daria Tolkacheva, MS2, Natalia Sableva, MS1, Taras Khimich, MS1.
1BIOCAD JSC, Moscow, Russian Federation, 2BIOCAD JSC, Saint Petersburg, Russian Federation.
1BIOCAD JSC, Moscow, Russian Federation, 2BIOCAD JSC, Saint Petersburg, Russian Federation.
OBJECTIVES: This research aims to compare overall (OS) and progression-free (PFS) survival between two anti-PD1+anti-CTLA4 combinations (prologolimab+nurulimab and nivolumab+ipilimumab) in the first line therapy of patients with advanced melanoma.
METHODS: Systematic literature review revealed three randomized clinical trials: OCTAVA (NCT05732805, 135 patients on prolgolimab+nurulimab (PROLGO+NURU)), CheckMate-067 (NCT01844505, 314 patients on nivolumab 1 mg/kg + ipilimumab 3 mg/kg (NIVO1+IPI3)), CheckMate-511 (NCT02714218, 178 patients on NIVO1+IPI3, 180 patients on nivolumab 3 mg/kg + ipilimumab 1 mg/kg (NIVO3+IPI1)). For OCTAVA individual patient data (IPD) were available, for comparators IPD on OS and PFS were restored from Kaplan-Meier curves and aggregated data on baseline characteristics were used. Due to the absence of common comparators unanchored matching-adjusted indirect comparisons (MAICs) in target populations of each CheckMate trial were performed. Patients with mucosal melanoma and CNS metastases (unfavorable survival predictors) were not included into OCTAVA, while their proportion in CheckMate-067 and CheckMate-511 was 12.4% and 10.6%, respectively. Survival distributions for these patients on PROLGO+NURU in both target populations were assumed similar to ones in patients with mucosal melanoma on NIVO1+IPI3 in CheckMate-067. To adjust MAIC estimates to these non-overlapping variables, we simulated 10 thousand direct comparisons in target populations including the respective share of patients with mentioned unfavorable characteristics.
RESULTS: Hazard ratios (95% confidence intervals) for OS on PROLGO+NURU compared to NIVO1+IPI3 (CheckMate-067), NIVO1+IPI3 (CheckMate-511), NIVO3+IPI1 (CheckMate-511) were 0.40 (0.20-0.66), 0.44 (0.21-0.73), 0.43 (0.21-0.72) in MAICs without adjustment to non-overlapping groups, and 0.53 (0.29-0.82), 0.57 (0.31-0.89), 0.56 (0.30-0.88) after adjustment, respectively. For PFS: 1.09 (0.79-1.45), 1.14 (0.85-1.51), 1.04 (0.77-1.37) before and 1.12 (0.82-1.49), 1.16 (0.86-1.55), 1.06 (0.78-1.41) after adjustment, respectively.
CONCLUSIONS: Monte-Carlo simulations are applicable for adjustment to non-overlapping variables in MAICs, although they require additional assumptions about survival distribution on the study treatment in unstudied populations, which are subject to further validation.
METHODS: Systematic literature review revealed three randomized clinical trials: OCTAVA (NCT05732805, 135 patients on prolgolimab+nurulimab (PROLGO+NURU)), CheckMate-067 (NCT01844505, 314 patients on nivolumab 1 mg/kg + ipilimumab 3 mg/kg (NIVO1+IPI3)), CheckMate-511 (NCT02714218, 178 patients on NIVO1+IPI3, 180 patients on nivolumab 3 mg/kg + ipilimumab 1 mg/kg (NIVO3+IPI1)). For OCTAVA individual patient data (IPD) were available, for comparators IPD on OS and PFS were restored from Kaplan-Meier curves and aggregated data on baseline characteristics were used. Due to the absence of common comparators unanchored matching-adjusted indirect comparisons (MAICs) in target populations of each CheckMate trial were performed. Patients with mucosal melanoma and CNS metastases (unfavorable survival predictors) were not included into OCTAVA, while their proportion in CheckMate-067 and CheckMate-511 was 12.4% and 10.6%, respectively. Survival distributions for these patients on PROLGO+NURU in both target populations were assumed similar to ones in patients with mucosal melanoma on NIVO1+IPI3 in CheckMate-067. To adjust MAIC estimates to these non-overlapping variables, we simulated 10 thousand direct comparisons in target populations including the respective share of patients with mentioned unfavorable characteristics.
RESULTS: Hazard ratios (95% confidence intervals) for OS on PROLGO+NURU compared to NIVO1+IPI3 (CheckMate-067), NIVO1+IPI3 (CheckMate-511), NIVO3+IPI1 (CheckMate-511) were 0.40 (0.20-0.66), 0.44 (0.21-0.73), 0.43 (0.21-0.72) in MAICs without adjustment to non-overlapping groups, and 0.53 (0.29-0.82), 0.57 (0.31-0.89), 0.56 (0.30-0.88) after adjustment, respectively. For PFS: 1.09 (0.79-1.45), 1.14 (0.85-1.51), 1.04 (0.77-1.37) before and 1.12 (0.82-1.49), 1.16 (0.86-1.55), 1.06 (0.78-1.41) after adjustment, respectively.
CONCLUSIONS: Monte-Carlo simulations are applicable for adjustment to non-overlapping variables in MAICs, although they require additional assumptions about survival distribution on the study treatment in unstudied populations, which are subject to further validation.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
SA7
Topic
Clinical Outcomes, Methodological & Statistical Research, Study Approaches
Topic Subcategory
Decision Modeling & Simulation
Disease
Biologics & Biosimilars, Oncology