P1.15-04 Dynamic Profiling of Blood Immunophenotypes and Radiomic Features to Predict Immunotherapy Response in Advanced Non-small Cell Lung Cancer

Introduction Up to date, no predictive biomarkers can robustly identify patients with non-small cell lung cancer (NSCLC) who will benefit from immune checkpoint inhibitors (ICIs). Thus, we sought to non-invasively decode tumor-immune interactions implicated in ICI response by exploring the dynamic of blood immune-inflammatory markers and radiomic features in a cohort of advanced NSCLC treated with ICIs. Methods On 58 stage IV NSCLC patients undergoing ICI-based therapy, blood immune-phenotyping data and CT-derived radiomic features (RFs) were acquired at baseline (T0) and at first disease assessment (T1). In detail, we performed a flow-cytometric analysis of circulating CD3+, CD8+, CD4+, NK, NKT and Tregs as their expression of functional molecules (PD-1, Granzyme B [GnzB], Perforin [Perf]) and proliferative index (Ki67). Overall, 851 RFs were extracted from T0 and T1 CT scans through a dedicated software (SlicerRadiomics). Time/treatment-dependent changes in blood parameters were expressed as percentage delta variation (Δ% = [T1 value - T0 value/T0 value]*100), while delta-RFs were computed as follows: (T1-T0)/T0. Primary endpoint was disease response per RECIST. CR/PR or SD ≥ 6 months defined clinical benefit (CB), while SD < 6 months or PD non-responders (NR). Results From October 2020 to August 2021, 58 advanced NSCLC patients candidate to receive ICI-based therapy were enrolled. Median age was 69 years (range: 41-88) and 80% underwent first-line ICIs (mostly consisting of pembrolizumab + platinum-based chemotherapy). According to disease response, 32 patients (55%) belonged to CB, while the remaining 26 (45%) were NR. Focusing on blood immune descriptors, we observed a significant increase in the overall number (n/μL) of CD3+, CD8+ and NK cells in CB, with a marked proliferative burst (Ki67+) of CD8+ lymphocytes carrying cytotoxic molecules (GnzB+, Perf+). Specifically, mean delta variations of NK, CD8+Ki67+ and CD8+GnZ+Perf+ phenotypes were, respectively, +22%, +170% and +65% in CB compared to -20%, -0.4% and -41% in NR (p<0.05, U-Mann Whitney test). Furthermore, the kinetic and extent of Treg (CD4+CD25+FOXP3high) counteraction, likely triggered by the expanding (Ki67+) and activated (GnZ+/Perf+) pool of effector T lymphocytes, appeared more pronounced in CB patients, reaching a mean Δ% variation of +375%. Delta-RFs were subjected to feature pre-processing, including redundant features elimination (Spearman correlation, cut-off=0.99) and Z-score standardization, and the remaining 657 features were correlated to ICI efficacy. We interestingly identified 11 delta-RFs differentially regulated in CB vs NR (p<0.05, U-Mann Whitney test). Subsequently, Principal Component Analysis (PCA) was computed to extract the main sources of variation from radiomic data, and principal components (PCs) were correlated with circulating delta-immune parameters (Pearson test). A noteworthy trend towards positive correlation was observed between PCs 21-22 (mostly contributing RFs: wavelet-LLH_firstorder_Mean and wavelet-HLL_firstorder_Maximum, respectively) and proliferating cytotoxic T phenotypes. Conversely, PC12 (mostly contributing RF: wavelet-HLL_firstorder_Skewness) negatively correlated with the same subsets of immune cells. Conclusions Our results suggest that tracking the evolution of blood immune-inflammatory and radiomic profiles may provide a more faithful portray of tumor-host interactions following ICIs, potentially representing a step toward the achievement of individualized decision support in advanced NSCLC patients.