244 - Cure Estimation using Reconstructed Kaplan-Meier Analyses in Randomized Phase III Oncology Trials (CUREMA): A Meta-Epidemiological Study
Presenter(s)
K. Chauhan1, P. Msaouel2, N. Meimoun3, A. M. Miller3, J. Liu3, T. A. Lin4, D. M. Routman5, W. Breen5, Z. R. McCaw6, E. B. Ludmir4, and A. D. Sherry5; 1Mayo Clinic Alix School of Medicine, Rochester, MN, 2Department of Genitourinary Medical Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 3The University of Texas MD Anderson Cancer Center, Houston, TX, 4Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 5Department of Radiation Oncology, Mayo Clinic, Rochester, MN, 6Insitro, South San Francisco, CA
Purpose/Objective(s): The most common primary endpoints of phase III oncology trials include disease control measures, such as progression-free survival, or overall survival. However, aside from time to progression or death, the probability of improving cancer cure with new therapies is seldom evaluated. Motivated by wholesale improvements in oncology over the last two decades, we performed the largest investigation of cure probabilities in modern phase III oncology trials using reconstructed individual patient-level data (IPD).
Materials/Methods: CUREMA was a prospectively registered meta-epidemiological study. Published two-arm phase III superiority oncology trials with matured composite primary endpoints assessing both survival and disease control were screened for eligibility from ClinicalTrials.gov. Kaplan–Meier curves were manually reconstructed using established methods including quality assessment to obtain IPD (Guyot et al). IPD were then fit for each trial with semiparametric proportional hazards mixture cure models to distinguish susceptible versus cured/non-susceptible individuals, defined as being alive with no evidence of disease recurrence at long term follow up, using the approach of Cai et al. Cure fractions were compared between arms within each trial using bootstrap resampling for variance estimation. Multivariable logistic regressions using confounders identified from a directed acyclic graph evaluated trial-level features and odds of improving cure probabilities.
Results: After screening 1184 trials, a total of 365 trials published between 2002 and 2024 and enrolling 313,000 patients met inclusion criteria. 330 trials (90%) were sponsored by industry, 362 tested systemic therapy (including with combinations of standard-of-care local therapy) (99%), and 227 (62%) met the primary endpoint. Cure models met convergence criteria for 359 trials (98%). Median cure probabilities for experimental and control arms were 24% (IQR, 8% to 52%) and 14% (IQR, 4% to 38%), respectively, including 61% and 59% in non-metastatic trials and 13% and 8% in metastatic trials. The experimental arm increased the odds of cure in 98 trials, including 40 metastatic. More recent trials were associated with greater odds of demonstrating improved cure probability (OR 1.07; 95% CI: 1.02 to 1.13; p = 0.01).
Conclusion: CUREMA demonstrates that cure probability, a highly patient-centric outcome, can be readily inferred from standard oncology trial data. Moreover, cure probabilities appear to be improving over time and are observed even in the metastatic setting, suggesting the benefits of modern cancer therapies extend beyond simply delaying time to progression or death. Incorporating cure probability as an endpoint during phase III trial design, especially among trials evaluating local therapies such as radiotherapy, may better identify and prioritize treatments promising clinically relevant improvements in achieving disease eradication. (PROSPERO: CRD420261299690)