Value of the lung immune prognostic index in patients with non‐small cell lung cancer initiating first‐line atezolizumab combination therapy: Subgroup analysis of the impower150 trial

Abstract

The lung immune prognostic index (LIPI) is proposed to differentiate prognosis and treatment benefit from immune checkpoint inhibitors (ICIs) in advanced non‐small cell lung cancer (NSCLC). There is minimal information on the predictive importance with first‐line, combination ICI approaches. In post‐hoc analysis of IMpower150, Cox‐proportional hazard analysis assessed the association between LIPI groups and overall survival (OS)/progression free survival (PFS). IMpower150 involved chemotherapy‐naïve, metastatic non‐squamous NSCLC participants randomized atezolizumab‐carboplatin‐paclitaxel (ACP), bevacizumab‐carboplatin‐paclitaxel (BCP), or atezolizumab‐BCP (ABCP). Good (0 factors), intermediate (1 factor), and poor LIPI (2 factors) were defined via derived neutrophil‐to‐lymphocyte ratio >3, and lactate dehydrogenase > upper limit of normal. Of 1148 participants, 548 had good, 479 intermediate, and 121 poor LIPI. In 385 participants randomised ABCP, a significant association between LIPI and OS (HR (95%CI): intermediate LIPI = 2.16 (1.47–3.18), poor LIPI = 5.28 (3.20–8.69), p < 0.001) and PFS (HR (95%CI): intermediate LIPI = 1.47 (1.11–1.95), poor LIPI = 3.02 (2.03–4.50), p < 0.001) was identified. Median OS was 24, 16, and 7 months for good, intermediate, and poor LIPI, respectively. ACP associations were similar. Relative OS treatment effect (HR 95%CI) of ABCP vs. BCP was 0.78 (0.53–1.15), 0.67 (0.49–0.91), and 0.87 (0.51–1.47) for the good, intermediate, and poor LIPI groups, respectively (P(interaction) = 0.66), with no benefit in median OS observed in the poor LIPI group. LIPI identified subgroups with significantly different survival following ABCP and ACP initiation for chemotherapy‐naïve, metastatic non‐squamous NSCLC. There was insufficient evidence that LIPI identifies patients unlikely to benefit from ABCP treatment.