Multimodal profiling for prediction of primary resistance to anti-PD-(L)1 therapy in advanced non-small-cell lung cancer: the prospective PIONeeR biomarkers study
Journal:
medRxiv
Published Date:
Jul 7, 2026
Abstract
Background Pretreatment prediction of primary resistance to anti-PD-(L)1 therapy in advanced non-small-cell lung cancer (NSCLC) remains an unmet clinical need. Existing biomarkers - including PD-L1 expression and tumour mutational burden (TMB) - are insufficiently discriminatory, and multimodal predictive models targeting primary resistance are still lacking to precisely drive patients treatment strategy. Methods PIONeeR was a prospective, multicentre biomarker study conducted across 17 centres (NCT03493581). Adults with advanced NSCLC initiating standard-of-care first-line platinum-based chemotherapy plus anti-PD-(L)1 therapy, or second-or-later-line anti-PD-(L)1 monotherapy, were enrolled. Pretreatment multimodal profiling spanned six biological layers: clinical, routine medical biology, high-dimensional circulating immune phenotyping, soluble vascular markers, tumour immune contexture with digital pathology analysis of multiplex immunohistochemistry and immunofluorescence, transcriptomics and genomics. The primary endpoint was prediction of primary resistance (PrR). Thirty-six feature-selection methods and ten machine learning models were benchmarked within a .632 optimism-correction framework. Results Between March 2018 and January 2023, 439 patients were enrolled (269 first-line, 170 subsequent-line); 435 were evaluable for PrR, which occurred in 39.5%. Individual biomarkers showed broad but modest associations with PrR (maximum AUROC 0.64). The multimodal signature achieved an optimism-corrected AUROC of 0.73 (s.d. 0.05) and a Positive Predictive Value (PPV) of 0.63 (s.d. 0.07) overall, and a PPV of 0.51 (s.d. 0.11) in the first-line setting, outperforming PD-L1 (PPV 0.36, s.d. 0.02) and TMB (PPV 0.35, s.d. 0.03). The 18-feature signature included stromal regulatory T-cell infiltration, circulating transitional B cells, and multiple routine laboratory variables, with treatment-setting-dependent contributions. Interpretation Despite comprehensive multimodal pretreatment profiling, primary resistance to anti-PD-(L)1 therapy in advanced NSCLC cannot be captured by isolated biomarkers alone. Complete results are publicly accessible through an interactive dashboard (https://compo.inria.fr/pioneer-dashboard/), a reference resource for future meta-analyses. Multimodal integration improved risk stratification while highlighting the strong contribution of routinely available baseline variables. These findings support the development of clinically pragmatic models for identifying patients who need to accede to next generation immunotherapy ovecoming primary resistances to PD-(L)1 inhibition.
