Line of therapy (LOT) definition impacts second-line (2L) survival duration in a metastatic pancreatic cancer (mPCa) electronic health record database.

Abstract

300 Background: Establishing an accurate LOT definition in real-world data (RWD) oncology studies is critical for measuring the effects of treatments on outcomes, which informs decisions made by clinicians, regulators, and payers. However, measuring LOT in RWD analyses is challenging and varies across disease states. LOT is generally defined by treatment (tx) start and stop dates. When available, the date of disease progression (PD) can be used to enhance the LOT definition. We explored the analytical impact of varying LOT definitions on 2L survival outcomes. Methods: A retrospective analysis of patients (pts) with mPC treated in US community practices was conducted using the Flatiron Health EHR–derived database. Two algorithms were applied to define LOT: tx change or ≥ 60-day gap in tx (LOTtx) and tx change after first PD (LOTpd). The impact on 1L and 2L sample sizes as well as 2L overall survival (OS) and progression-free survival (PFS) were assessed. Results: 251 pts with an mPC diagnosis between 01 March and 31 October 2015, who initiated chemotherapy ≤ 60 days after mPC diagnosis, were analyzed. Applying the LOTtx definition, 121 pts received 1L tx only and 130 received a 2L tx. Using the LOTpd definition, 145 pts received 1L tx only and 105 received a 2L tx (12 pts were excluded due to unclear time of PD relative to tx). Applying these definitions to time-to-event analyses (from start of 2L therapy) revealed longer 2L OS with LOTtx vs LOTpd (n = 130 vs 105; median 2L OS [95% CI], 5.4 [4.1-7.4] versus 4.5 [3.7-7.3] mo) but similar 2L PFS (n = 122 vs 105; median 2L PFS [95% CI], 2.8 [2.3-3.5] versus 2.8 [2.3-3.6] mo). Conclusions: In this analysis, inclusion of PD in the LOT definition impacted 1L and 2L sample sizes and 2L survival results. In aggressive diseases like mPC, a 1-mo difference in OS estimates is clinically meaningful and may influence treatment decisions. The choice of LOT definition should be based on the research question at hand, available data, and the effort to decrease misclassification bias. Additional LOT definitions, with associated strengths and limitations, will also be presented.