Citations 2
Interpreting Treatment Effects Using Posterior Probabilities: A Bayesian Reanalysis of 230 Phase III Oncology Trials
Sherry, Alexander D. , Msaouel, Pavlos, Kupferman, Gabrielle S. , Lin, Timothy A. , Jaoude, J. Abi, Kouzy, Ramez, El-Alam, Molly B , Patel, Roshal , Koong, Alex , Lin, Christine , Beck, Esther J., Fuller, Clifton David, Ludmir, Ethan B, Msaouel, Pavlos , Kouzy, Ramez , Jaoude, Joseph Abi , Msaouel, Pavlos , Abi Jaoude, Joseph , Kouzy, Ramez , Passy, Adina H , Miller, Avital M , Beck, Esther J , Fuller, Clifton David , Meirson, Tomer , McCaw, Zachary , Ludmir, Ethan B , Sherry, Alexander D , Lin, Timothy A , Miller, Avital M. , Meirson, Tomer , McCaw, Zachary R. , Ludmir, Ethan B.
JCO Clinical Cancer Informatics | May 1, 2026 | Volume: 10 | Issue: 2 | Pages: No page available
PURPOSE Most oncology trials define superiority according to dichotomized P value thresholds, which are frequently misinterpreted. Posterior probability, however, directly estimates the probability of the hypothesis at hand. Here, we reanalyze a large collection of modern phase III trials and benchmark posterior probability versus the standard trial interpretation based on statistical significance. METHODS Outcomes from 194,129 patients were manually reconstructed from the primary end points of 230 phase III, superiority-design oncology trials. Posterior probabilities of treatment effect were then calculated across multiple priors and several effect sizes of clinical relevance, including minimum clinically important difference (MCID) defined as hazard ratio (HR) < 0.8 per ASCO criteria or HR < 0.64 per European Society of Medical Oncology (ESMO) criteria. RESULTS All trials interpreted as superior using P value thresholds had probabilities >90% for achieving at least marginal benefits (HR < 1). However, only 62% of positive trials (74/120) had >90% probabilities of achieving the ASCO MCID (HR < 0.8), even under an enthusiastic prior, including 70% of trials (57/82) leading to regulatory approval. Only 30% of positive trials (36/120) had >90% probability of achieving the ESMO MCID (HR < 0.64). Conversely, 24% of trials (26/110) interpreted as not superior had >90% probability of achieving marginal benefits (HR < 1), even under a skeptical prior. CONCLUSION Bayesian models, although often in agreement with statistical significance thresholds, add considerable unique interpretative value for a subset of phase III oncology trials. Posterior probability may provide a solution for overcoming the discrepancies between refuting the null hypothesis and detecting clinically relevant effects.
Differential censoring in overall survival in phase 3 oncology clinical trials
Sherry, Alexander D. , Lin, Timothy A. , Jaoude, Joseph Abi , Kouzy, Ramez , McCaw, Zachary R. , Meirson, Tomer , Msaouel, Pavlos , Ludmir, Ethan B. , Hsu, Eric J , Sherry, Alexander D , Lin, Timothy A , Abi Jaoude, Joseph , Kouzy, Ramez , McCaw, Zachary , Meirson, Tomer , Msaouel, Pavlos , Ludmir, Ethan B
JNCI: Journal of the National Cancer Institute | December 22, 2025 | Volume: 118 | Issue: 8 | Pages: 1527-1531
Abstract In phase 3 oncology clinical trials, identifying and understanding potential sources of differential censoring (DC), which describes censoring imbalance between treatment arms, in overall survival (OS) endpoints is vital to reduce bias and misinterpretation of OS benefit. Among 322 trials, 34 (11%) trials exhibited DC overrepresented in the control arm (DCC). Compared to trials without DC or those with DC overrepresented in the experimental arm (DCE), trials with DCC showed higher rates of OS benefit (56% vs. 32%; odds ratio [OR] = 2.26, 95% CI = 1.04 to 4.90; P = .04) and were associated with patient withdrawal (median 8.0% vs. 3.8%; P = .0002), lack of double blinding (15% vs. 48%, OR = 0.19, 95% CI = 0.06 to 0.56; P = .003), and better global quality of life in the experimental arm (32% vs. 13%, OR = 3.01, 95% CI = 1.26 to 7.19; P = .01). OS DCC is a prominent challenge that requires rigorous trial design and comprehensive patient and mortality follow-up to reduce bias in interpreting OS benefit.