Use of duration of response as an interpretable endpoint in non-small cell lung cancer drug approvals.

Abstract

e23020 Background: Overall response rate (ORR) is used as an early endpoint indicative of efficacy in oncology clinical trials, especially in accelerated approvals. Trials reporting ORR often include duration of response (DoR) as a secondary endpoint contextualizing durability. However, as a time-to-event endpoint, interpreting DoR, including in single-arm trials (SATs), is challenging given the inability to separate treatment effect from disease natural history and dependent censoring. In randomized controlled trials (RCTs), as DoR is limited only to responders, randomization is lost and potential subgroup effects arise. We analyzed DoR use in trials supporting original approvals in non-small cell lung cancer (NSCLC). Methods: The study included original approvals indicated for the treatment of NSCLC from August 26, 2011 through July 2, 2025. Data from the labels and multi-disciplinary review were found using the Drugs@FDA database. Descriptive analyses assessed the use of DoR as an endpoint, its reporting on the label, definitions, and associations with approval pathway and trial design. Results: Thirty drugs were included, representing 20 accelerated and 11 traditional approvals based on 39 trials (9 RCTs and 30 SATs). 74.2% of approvals used ORR as the primary endpoint (20 accelerated, 3 traditional) and DoR was a secondary endpoint in 87.1% (27/31) of approvals, including all accelerated approvals. The majority of trials (80.7%) supporting approvals with DoR as a secondary endpoint were SATs for 20 accelerated and 1 traditional approval. All approved drugs using DoR as an endpoint reported DoR in the label, with heterogenous metrics reported: 1 (3.8%) only DoR range reported, 1 (3.8%) median (mDoR) and range, 7 (26.9%) mDoR and 95% confidence interval (CI), 15 (57.7%) mDoR, 95% CI, and landmark(s), and 2 (7.7%) DoR range and landmark(s). For the 17 labels reporting landmark DoRs, 15 reported at 6 months, 3 reported at 9 months, 9 reported at 12 months, and 1 reported at 18 months. Only 4/26 (15.4%) labels where DoR was reported also reported follow-up duration and 6/26 (23.1%) did not report follow-up time anywhere in the review documents, despite its direct impact on DoR interpretation. Lastly, 10/26 (38.5%) of drugs with a DoR endpoint did not define or had incomplete definitions of DoR; within those with detailed definitions, there were discrepancies in how the start and end date for DoR were defined. Conclusions: DoR is a secondary endpoint in the majority of NSCLC approvals, especially in accelerated approvals. Most trials assessing DoR are SATs, in which time-to-event endpoints are known to be less interpretable. There was also significant variation in the metrics reported and definitions used, which can greatly impede endpoint interpretation. While DoR provides valuable context for ORR, these irregularities impede clinical interpretation. Further alignment of its appropriate use is needed.