Interpreting a positive PFS trial with immature overall survival
A phase III trial randomizes 600 patients with advanced biliary tract cancer to FOLFOX alone or FOLFOX plus durvalumab. At the first interim analysis, median progression-free survival is 7.1 months with durvalumab versus 5.3 months with chemotherapy alone (HR 0.74, 95% CI 0.60–0.91; p=0.004). Overall survival data are immature: median 18.2 versus 16.9 months (HR 0.92, 95% CI 0.75–1.12). Objective response is 21% versus 12%. Grade 3 or higher adverse events are 58% versus 49%. Twenty-eight percent of control patients received subsequent immunotherapy off protocol. The trial's primary endpoint is PFS, with OS a key secondary endpoint. A colleague proposes adopting the regimen for all fit patients at your tumor board.
1. How do you interpret the PFS result, including statistical versus clinical significance and the interim analysis?
Read the worked answer
The PFS difference is statistically positive at the interim analysis, but that alone does not establish practice-changing benefit. The absolute gain is 1.8 months (7.1 vs 5.3), and the HR of 0.74 with an upper confidence limit of 0.91 indicates a modest effect. Whether p=0.004 crosses the prespecified interim boundary depends on the alpha-spending plan; a p-value without that context is insufficient. I would also check whether PFS was centrally assessed, whether the curves separate early or late, and whether the trial was powered for OS. In biliary cancer, PFS is not a fully validated surrogate for OS or quality of life. The higher grade 3+ toxicity (58% vs 49%) matters. My provisional view is that this is a promising but not yet standard option for all fit patients.
2. What are the implications of immature OS and 28% crossover for your recommendation?
Read the worked answer
Immature OS means the confidence interval still includes no benefit (HR 0.92, 95% CI 0.75–1.12), so I cannot tell a patient that survival is improved. The 28% crossover from control to immunotherapy is a major confounder: it can dilute an OS effect and makes a simple intention-to-treat OS comparison hard to interpret. I would look at maturity, follow-up duration, subsequent therapies, and a prespecified crossover-adjusted or RPSFT analysis, but these are supportive, not definitive. For a patient with good performance status and limited alternatives, I might discuss the regimen as an option with uncertain survival benefit and more toxicity, preferably in a trial or registry. I would not make it the default for all fit patients until OS or a validated quality-of-life benefit is mature.
3. What additional information would you need before adopting this regimen, and what would you tell the patient?
Read the worked answer
I would want the full protocol: alpha-spending plan, primary PFS definition, scan schedule, follow-up maturity, subgroup results by PD-L1, biliary obstruction, prior therapy and geography, and quality-of-life data. I would also want the final OS analysis plan and the toxicity details, including immune-related events and treatment-related deaths. For the patient, I would say that the trial shows a small average delay in progression but no proven survival gain yet, that the control arm often received immunotherapy later, and that the new regimen adds immune toxicity and more grade 3+ events. We would decide together based on his goals, performance status, liver function, access and willingness to accept toxicity. I would prefer a clinical trial or a shared decision rather than a blanket tumor-board policy.