Propensity score analysis of overall survival between first- and second-generation EGFR-TKIs using real-world data

Abstract

We constructed a data set of EGFR-mutant non-small-cell lung carcinoma (NSCLC) patients, and compared the overall survival of first-generation (1G), and second-generation (2G) EGFR-tyrosine kinase inhibitors (TKIs) in clinical practice using a propensity score. We reviewed the clinical data of consecutive EGFR-mutated NSCLC patients who received EGFR-TKI therapy between January 2008 and August 2017 at 11 institutions in Japan. The primary endpoint was overall survival (OS). When comparing OS between 1G and 2G EGFR-TKIs, propensity score analyses were performed using 2 methods: matching and inverse probability of treatment weighting (IPTW). (Clinical Trial information: UMIN000030121) In total, 1400 patients from 11 institutions were enrolled in this study, and the data from the 1366 patients who received only EGFR-TKI therapy were analyzed (gefitinib [GEF], N = 732; erlotinib [ERL], N = 416; afatinib, N = 218). Median OS times (months [95%CI]) were 29.7 [27.5-33.5] in the 1G group (gefitinib, 32.0 [28.1-35.8]; erlotinib, 27.5 [23.9-31.7]), and 38.6 [32.2-NR] in the 2G group (afatinib), respectively. The trend of longer OS for afatinib against 1G EGFR-TKIs remained, even after adjusted by propensity score. (unadjusted, hazard ratio [HR] 0.676, P = .0023; adjusted by IPTW, HR 0.685 P < .0001; adjusted by matching, HR 0.725, P = .0418). Exploratory analysis showed that OS using the 2G EGFR-TKI was superior to that of the 1G EGFR-TKIs, suggesting the potential of sequential therapy of 2G EGFR-TKI followed by osimertinib. (HR 0.419, P = .0519) Real-world data analysis using 1354 data records, using propensity scoring, indicated that 2G EGFR-TKI had a trend of longer OS compared with 1G EGFR-TKIs.

Keywords: EGFR-TKI; non-small-cell lung cancer; propensity scoring analysis; real-world data.