LONGITUDINAL MONITORING OF RADIOMIC AND BLOOD IMMUNE-INFLAMMATORY DESCRIPTORS AS A NON-INVASIVE APPROACH TO PREDICT ICI EFFICACY IN ADVANCED NSCLC PATIENTS

Background: Clinically suitable biomarkers to foresee the response to immune checkpoint inhibitors (ICIs) can be achieved by tracking tumor heterogeneity and its evolution during treatment. Thus, we explored whether the longitudinal monitoring of radiomic features (RFs) and blood immunophenotypes may predict ICI benefit in advanced NSCLC. Methods: We prospectively enrolled NSCLC patients undergoing ICI-based regimens in our 5-year AIRC granted project. Peripheral blood (PB) and CT-scans were acquired at baseline (T0) and at first disease assessment (T1). On PB samples we computed the Lung Immune Prognostic Index (LIPI) and analysed by flow-cytometry circulating CD3+, CD8+, CD4+, NK and Tregs as the expression of PD-1, Granzyme B [GnzB], Perforin [Perf] and Ki67. From CT-images, 851 RFs were extracted through a dedicated software (SlicerRadiomics). Time-dependent changes in PB parameters were expressed as percentage delta immune variation (ΔI= [T1 value – T0 value/T0 value]*100), while ΔRFs as (T1–T0)/T0. Primary endpoints were PFS and tumor response per RECIST (v.1.1). CR/PR or SD ⩾ 6 months defined clinical benefit (CB), while SD < 6 months or PD non-responders (NR). Results: Our preliminary data on 58 NSCLC underlined a significant increase in the overall number of PB CD3+, CD8+ and NK cells in CB group, with a remarkable proliferation (Ki67+) and cytotoxic activity (GnzB+, Perf+) of CD8+ lymphocytes. Mean Δ variations of NK, CD8+Ki67+ and CD8+GnzB+Perf+ phenotypes were, respectively, -20%, -0.4% and -41% in NR vs +22%, +170% and +65% in CB (P<0.05, Mann Whitney), which also exhibited a more pronounced Tregs counterbalance. Strikingly, cases displaying concomitant rise of all these meaningful PB immune feature had a significantly prolonged PFS compared to the counterpart (HR=0.49, P<0.01). In addition, we found that following ICIs most patients preserving good LIPI or experiencing LIPI improvement belonged to CB (P<0.05, Fisher exact) and presented a longer PFS. Pre-processing and Z-score standardization narrowed ΔRFs to 657, among which 11 were differentially regulated in CB vs NR (P<0.05, Mann Whitney). Distinct ΔRFs principal components, largely embodied by wavelet-HLL_firstorder_Maximum and -HLL_firstorder_Skewness, showed, respectively, direct and inverse correlation with ΔI CD8+Ki67+. Conclusions: Dynamic monitoring of immunophenotypic and radiomic cues may implement our current prognostic and predictive models in advanced NSCLC treated with ICIs.