Molecular progression defined by longitudinal ctDNA dynamics for prediction of outcomes in immune checkpoint inhibitor–treated solid tumors.

2572 Background: Reliable early biomarkers of immune checkpoint inhibitor (ICI) resistance remain an unmet need due to delayed imaging and non-classical response kinetics. While ctDNA has shown promise, current methods often require tumor tissue or rely on static comparisons. We investigated whether longitudinal ctDNA dynamics measured using a tumor-naive, methylation-based assay could define molecular progression and predict outcomes in patients receiving ICIs. Methods: Patients with advanced solid tumors treated with ICI regimens were analyzed in a primary cohort (n=65) with serial plasma sampling. Tumor Methylation Scores (TMS) were generated using Northstar Response. Molecular progression based on ctDNA trajectory was assessed for associations with PFS and OS and compared with conventional landmark ctDNA approaches and RECIST v1.1. Independent validation was performed in a second cohort (n=72). Results: Molecular progression strongly stratified survival in the primary cohort (OS HR=4.9, p<0.001; PFS HR=5.3, p=0.0002) and demonstrated superior prognostic performance relative to RECIST. These results were independently validated (PFS HR=4.6, p=0.00004; OS HR=4.3, p=0.005). Notably, among RECIST-defined non-progressors at 6 months, molecular progression identified patients with markedly worse outcomes (PFS HR=6.7, p=0.002; OS HR=4.5, p=0.011). Molecular progression preceded radiographic and clinical progression by median lead times of 63.5 and 77 days, respectively. Conclusions: Longitudinal ctDNA dynamics measured using a tumor-naive methylation assay define molecular progression that robustly predicts survival and identifies early ICI resistance, including in RECIST-stable patients. This approach offers clinically meaningful lead time and supports integration of ctDNA dynamics into response assessment frameworks for immunotherapy.