Conference Program

RL22: Impact of Missing or Alternative Tumor Assessments and Intercurrent Events on the Validity of PFS Results (303577)

Cong Chen, Merck & Co., Inc. 
Joyce Cheng, FDA/CDER 
Susan Halabi, Duke University 
*Ray Lin, Genentech/Roche 
Yuan-Li Shen, US Food and Drug Administration 

Keywords: PFS, missing assessment, alternative assessment schedules, intercurrent event, interval censoring, survival analysis

To evaluate the impact of missing, alternative tumor assessments, or the occurrences of intercurrent events (e.g., patient dropout or the initiation of non-protocol therapies) on Progression-free survival (PFS), several analysis methods have been suggested, including (1) using the observed event time as is regardless of missing assessments or intercurrent events, (2) censoring the observation at the last event-free assessment prior to the missing assessments or intercurrent events (FDA guidance 2015, 2018), and (3) methods for interval-censoring analysis (PhRMA working group 2011 and EMA guidance 2008).

In this roundtable, discussants from industry, FDA, and academia will discuss the statistical characteristics of various analysis methods. The FDA will share their current practice for evaluating the impact of missing assessments or intercurrent events. The values and challenges of these methods, the statistical and practical considerations, and their potential use in the regulatory setting will be discussed.

[Introduction] Interval-censoring and right-censoring methods in PFS analysis with missing tumor assessments. (Ray Lin, Genentech/Roche).

[Panel Discussants] Ray Lin, Genentech/Roche; Cong Chen, Merck; Yuan-Li Shen, FDA; Joyce Cheng, FDA; Prof. Susan Halabi, Duke University.

[Discussion points]

1. Historical perspective: PhRMA cross-industry working group on PFS assessment (2011)

2. PFS assessment process at the FDA: What types of sensitivity analysis are conducted; How to drawn the final conclusion based on multiple analysis results

3. Considerations from the academia: The statistical justification and operating characteristics of each method

4. Considerations from the industry: Concerns about interval censoring results, feasibility of more frequent assessments and efforts of reducing missing data

5. Considerations from the regulators: Considerations for regulatory approval, interpretability of the results, comparability across studies