Augmenting Real-World Data Through Modeling Key Clinical Trial Eligibility Criteria: An Example of Patients With...

To compare and characterize baseline characteristics and overall survival (OS) differences by key oncology eligibility criteria for real-world patients from the Flatiron Health database with advanced non-small-cell lung cancer (NSCLC) who received pembrolizumab monotherapy.

Real world data (RWD) were from the Flatiron Health advanced NSCLC database and include patients who initiated pembrolizumab monotherapy (first, second, or third line of therapy) by November 30, 2019. At the data cutoff (May 31, 2020), the median survival follow-up time was 8.4 months. Eligible patients satisfy the criteria of Eastern Cooperative Oncology Group performance status of 0/1 and laboratory values indicative of adequate organ function. RWD were analyzed for all patients and patients with a programmed cell death ligand-1 tumor proportion score ≥ 1%. Patients were divided into three categories: ineligible, eligible, and unknown (who satisfy all observed criteria, with at least one missing). An augmented population was also formed, which combines the latter two groups through a propensity-based adjustment.

At the data cutoff, N = 3,877 patients with NSCLC received pembrolizumab monotherapy (1L = 2,682, 2L = 946, and 3L = 249). OS was consistently lower for the ineligible with similar survival for the eligible and augmented. Among all patients, the median OS in months (95% CI) was 8.2 (7.5 to 9.6), 16.3 (14.5 to 18.4), 16.4 (15.1 to 19.3), and 16.8 (15.6 to 18.5) for the ineligible (47%, n = 1,827), unknown (27%, n = 1,045), eligible (26%, n = 1,005), and augmented, respectively. The results were similar for patients with a programmed cell death ligand-1 tumor proportion score ≥ 1%.

Real-world patients who received pembrolizumab monotherapy and meet key clinical eligibility criteria exhibited similar baseline characteristics and OS profiles as the unknown and augmented patient groups. Population augmentation is a feasible approach for improving the power of RWD analysis.