Blood-based kinome profiling predicts immune checkpoint inhibitor response in advanced NSCLC patients, outperforming PD-L1 TPS

For patients with non-small cell lung cancer (NSCLC), immune checkpoint blockade (ICB) targeting programmed death-1 (PD-1) or PD-L1 offers significant benefit, but response rates vary. The PD-L1 tumor proportion score (TPS) is a standard biomarker, yet its predictive power is limited, leaving a critical need for more accurate biomarkers. Kinase activity profiling of peripheral blood mononuclear cells (PBMCs) has shown promise in discovery studies, suggesting a potential new avenue for predicting ICB efficacy.

This multicenter, prospective study included N=210 patients with advanced stage NSCLC treated with anti-PD-1 immune checkpoint blockade (±chemotherapy). Prior to ICB treatment, peripheral blood mononuclear cells (PBMCs) were collected from all patients. These PBMCs were then profiled using a PamChip peptide microarray, which assesses multiple kinase substrates to determine kinase activity. Classification analysis was performed to differentiate between patients who experienced progressive disease within 24 weeks of treatment initiation and those who did not, based on RECIST V.1.1 criteria. A predictive model was initially developed using both PD-L1 TPS and kinase activity profiles, followed by rigorous testing in a separate validation cohort.

In the validation cohort, the kinome-based prediction model demonstrated a significantly higher progression-free survival (PFS) rate for patients predicted to benefit compared to those not predicted to benefit, with a hazard ratio (HR) of 0.56 (p=0.01). This performance tended to be superior when compared to PD-L1 TPS alone, which yielded an HR of 0.66 (p=0.07). The most robust predictive power was observed when the kinome-based model was combined with PD-L1 TPS. This combined approach further amplified the difference in PFS between predicted benefit and non-benefit groups.

When the kinome-based model was combined with TPS, the difference in PFS between patients with and without predicted benefit was further increased (HR=0.38, p<0.001). Notably, the kinome-based model also improved prediction in subgroups with lower PD-L1 expression: for TPS <1%, the HR was 0.46 (p=0.027), and for TPS 1-50%, the HR was 0.20 (p=0.005), indicating its utility across varying PD-L1 levels.