Retrospective analysis of digital biomarker predictability of immune checkpoint inhibitor therapy outcomes in patients with NSCLC.

Abstract

e14713 Background: Immune checkpoint inhibitor (ICI) therapy (pembrolizumab monotherapy and pembrolizumab combined with platinum-based chemotherapy) is recommended as first line for patients with PD-L1 positive advanced or metastatic non-small cell lung cancer (NSCLC) without actionable biomarkers due to incredible improvements in overall survival (OS). Despite this, 15-20% of patients do not respond to ICI therapy in NSCLC. This study examines the effect of body composition features, via analytic morphomics, in conjunction with current clinical and laboratory data to predict ICI response. Methods: This retrospective analysis included 35 patients from the University of Michigan (UM) with newly diagnosed advanced or metastatic NSCLC (adenocarcinoma) receiving single-agent pembrolizumab or in combination with chemotherapy between January 2018 – December 2020. Clinical/ laboratory data included patient gender, age, BMI, neutrophil to lymphocyte ratio (NLR), PDL1 %, and albumin. Analytic Morphomics was performed using proprietary, semi-automated high-throughput CT image processing (mBIOHEALTH, Inc.). Multivariable Cox regression analysis identified clinical and morphomic characteristics predictive of OS at 6 months. Risk models were developed on the UM cohort and tested on an external validation dataset of 24 adenocarcinoma patients from Emory Healthcare who received similar treatment between January 2018 to December 2021. Results: Among the evaluable patients at both sites, 15 (43%) at UM and 19 (79%) at Emory were female, mean age was 66 years at UM and 70 years at Emory, and 6-month OS (95% CI) was 89% (80%, 100%) at UM and 71% (55%, 92%) at Emory. Five regression models with risk scores were assessed with the UM data and then tested with the Emory dataset. Of the five models, 2 matched in their predictability of 6-month OS at both sites. The first, a model of clinical and laboratory features only (age, BMI, PDL1 > = 50%, NLR, and albumin) which had an AUC of 0.79 at UM and 0.66 at Emory. The other utilized an L3 skeletal muscle z-score threshold ( < = -1) alone (previously studied as a measurement of sarcopenia) with an AUC of 0.741 at UM and 0.57 at Emory. Conclusions: Our analysis indicates that the measurement of muscle z-score prior to initiating treatment with ICIs in patients with NSCLC could serve as an indicator for those patients who are unlikely to benefit from therapy. This may be especially important for ICI therapy as skeletal muscle effects the immune function through myokines. Sarcopenia has been associated with immune senescence and thus may play a significant role in decreasing response to ICI therapy. More analysis is warranted with a larger patient population.