Background: Novel therapies have improved the survival for patients with advanced non-small cell lung cancer (NSCLC). Clinical scenarios with high treatment variation indicate a lack of consensus about the standard of care, and can identify instances there is less clinical equipoise, and comparative effectiveness research is urgently needed. We quantified the variation in the first-line therapy, stratified by patient’s biomarker and Eastern Cooperative Oncology Group (ECOG) performance score. Methods: Using Flatiron database, we categorized patients with advanced NSCLC diagnosed between 1/1/2016 and 6/30/2020 into groups by biomarker findings and performance status. Patients with unknown status of biomarkers or negative actionable gene alterations, or missing ECOG score were excluded. The degree of variation in first-line treatment regimens was measured using the Herfindahl-Hirschman Index (HHI), which calculates the sum of squared proportion of patients receiving different regimen categories. The HHI ranges from close to 0 to 10,000, with low values indicate heterogeneous treatment patterns whereas high values indicate more uniformity in treatment patterns. We used bootstrapping to calculate 95% confidence intervals (CIs) for the HHI scores. Results: Our cohort consisted of 12,708 patients. Among them, 327 were ALK-positive, 1,816 were EGFR-positive, and 10,565 were ALK/EGFR negative or unknown (including 3,616 patients who were PD-L1 <1%, 3,139 PD-L1 1-49%, and 3,810 PD-L1 ≥50%, respectively). The HHI score for first-line therapy in our overall cohort was 2171 (95% CI: 2167, 2175). The HHI differed by biomarker status: 4403 for the EGFR+ group, 3809 for the ALK+ group, 3065 for the PD-L1 <1% group, 2925 for the PD-L1 1-49% group, and 3370 for the PD- L1 ≥50% group. When we stratified patients by both biomarker and ECOG status, the HHI ranged from 2791 (95% CI: 2740, 2840) for patients with PD-L1 1-49% and ECOG 2-4 to 5364 (95% CI: 5161, 5649) for patients with EGFR+ and ECOG 2-4 (details in Table). Conclusions: Using HHI, we were able to quantify and compare variation in treatment patterns across clinically distinct subgroups of patients. Large variation in the first-line therapy for patients with PD-L1 1-49% suggests lack of strong consensus regarding the optimal treatment regimens for these patients.

ECOG: Eastern Cooperative Oncology Group; HHI: Herfindahl–Hirschman Index; CI: Confidence interval