The Impact of Incorporating Survival Heterogeneity into Modeling of Survival Outcomes for Patients with Muscle-Invasive Bladder Cancer after Radical Resection: Analyses from a Retrospective Real-World Study in France
Author(s)
Chepynoga K1, De T2, Kurt M3, Colrat F4, Branchoux S5, Patel M6, Teitsson S7
1Access Consulting, Parexel, Hackensack, NJ, USA, 2Access Consulting, Parexel, Billerica, MA, USA, 3Bristol Myers Squibb, Princeton, NJ, USA, 4Bristol Myers Squibb, Rueil Malmaison, 92, France, 5Bristol Myers Squibb, Rueil-Malmaison, France, 6Bristol Myers Squibb, Summit, NJ, USA, 7Bristol Myers Squibb, Uxbridge, UK
OBJECTIVES: To examine the impact of accounting for potential long-term survivors (LTS) on the analysis and projection of survival for patients with muscle-invasive bladder cancer at high risk of recurrence after radical resection from the COBLAnCE registry cohort in France. METHODS: Publicly available disease-free survival (DFS) and overall survival (OS) data on 256 patients were digitized. Median data follow-up was 5.03 years. Mixture cure models (MCMs), standard parametric models (SPMs), and cubic-spline models (CSMs) were fitted separately to the DFS and OS data. MCMs classified patients into 2 latent subpopulations, as LTS and non-LTS, where LTS were subject only to non-disease-related mortality estimated from general population mortality rates in France, adjusted with baseline age, sex, and smoking prevalence in the cohort. Time-to-event outcomes for non-LTS were modeled via SPMs. Statistical goodness-of-fit metrics, and visual inspection of survival distributions and hazards were used to select the base-case models and their alternatives for sensitivity analyses. Restricted mean survival times (RMSTs) were estimated and compared over 6, 10, and 20 years. RESULTS: Estimated percent of LTS was 19.9% (95% CI, 11.3%-32.9%) for DFS and 25.8% (95% CI, 16.3%-38.3%) for OS. Estimated RMST by the MCM showed only marginal differences versus those by SPM and CSM over 6 and 10 years (DFS, <1.1 months; OS, <1.4 months). Twenty-year RMST projections from the MCMs were 61 months for DFS and 71.4 months for OS. These projections exceeded those from the SPM and CSM by 10.7 and 6.7 months, respectively, for DFS, and by 8.1 and 10 months, respectively, for OS. Results were not sensitive to the selected alternative models in each family. CONCLUSIONS: A considerable proportion of the cohort can be projected as LTS. Accounting for the latent survival heterogeneity in the cohort borne by these patients via MCMs can yield higher long-term survival projections.
Conference/Value in Health Info
2023-05, ISPOR 2023, Boston, MA, USA
Value in Health, Volume 26, Issue 6, S2 (June 2023)
Acceptance Code
P27
Topic
Methodological & Statistical Research
Disease
no-additional-disease-conditions-specialized-treatment-areas