Tumor-Informed ddPCR for Personalized Longitudinal ctDNA Monitoring in Advanced Solid Tumors: Molecular-Radiological...

Serial circulating tumor DNA (ctDNA) monitoring may complement episodic radiological response assessment during systemic therapy. We evaluated personalized tumor-informed droplet digital PCR (ddPCR) for longitudinal ctDNA monitoring in advanced solid tumors, focusing on RECIST-associated ctDNA levels, anatomical tumor burden, molecular-radiological concordance, molecular lead time (MLT), and routine implementation. The clinical cohort comprised 61 patients with breast, colorectal, gynecological, head and neck, lung, pancreaticobiliary, or other advanced solid tumors and 385 longitudinal patient-specific ddPCR assessments; 357 assessments were imaging-concurrent. Repeated RECIST-stratified measurements were evaluated descriptively and using patient-clustered generalized estimating equations (GEEs). Longitudinal trajectories were classified using study-specific Longitudinal Concordance Level (LCL) and Molecular Evidence Level (MEL) frameworks. A separate real-world implementation cohort comprised 270 patients and 650 personalized ddPCR assays; 88 patients underwent two or more serial assessments. Among the 357 imaging-concurrent assessments, 90 were partial response (PR), 117 stable disease (SD), and 150 progressive disease (PD). Median VAF was 0.055% in PR, 0.190% in SD, and 0.205% in PD. Patient-clustered GEEs confirmed an overall association between VAF and RECIST category (Wald χ² = 12.60, p = 0.00183), with lower VAF in PR than SD (Holm-adjusted p = 0.00119) or PD (p = 0.00247), but no separation between SD and PD (p = 0.539). In 210 imaging-ctDNA pairs from 53 patients, correlations with estimated tumor volume (ρ = 0.169) and RECIST SLD (ρ = 0.177) were weak. Of 50 evaluable longitudinal trajectories, 21 (42.0%) were LCL-1. Among 14 patients who developed radiological PD after an initial PR or SD, molecular progression preceded radiological progression in 9 (64.3%), was concurrent in 3 (21.4%), and no qualifying pre-radiological molecular-progression event was identified in 2 (14.3%); positive MLTs had a median of 55 days (IQR 46-83; range 28-209). The real-world cohort included five genomic alteration classes and a median assay-specific LoD of 0.020% VAF (IQR 0.020-0.030%). Personalized tumor-informed ddPCR showed population-level association with RECIST response and structured longitudinal molecular-radiological relationships, while substantial VAF overlap and weak anatomical-burden correlations limit interpretation of isolated measurements. Molecular progression preceded imaging-defined progression in a subset of patients, supporting prospective evaluation of ctDNA-triggered reassessment rather than establishing benefit from ctDNA-guided intervention. The separate real-world cohort supports operational feasibility across diverse assay designs.