Work

Tumor-Informed ddPCR for Personalized Longitudinal ctDNA Monitoring in Advanced Solid Tumors: Molecular–Radiological Concordance and Molecular Lead Time

Year: 2026

Type: article

Abstract: Background: Radiological assessment of treatment response in advanced solid tumors is episodic and may delay recognition of biological progression. We evaluated personalized tumor-informed circulating tumor DNA (ctDNA) monitoring using droplet digital PCR (ddPCR) to determine its association with radiological response and tumor burden, characterize longitudinal molecular–radiological concordance, quantify molecular lead time (MLT), and assess real-world implementation. Methods: Serial plasma ctDNA monitoring was performed in 61 patients between 2017-2019 with advanced solid tumors using personalized ddPCR assays targeting patient-specific somatic driver variants. ctDNA variant allele fraction (VAF) was compared across Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1) categories and evaluated in relation to estimated tumor volume and RECIST sum of longest diameters (SLD). Longitudinal trajectories were assessed using predefined Longitudinal Concordance Level (LCL), molecular progression, and Molecular Evidence Level (MEL) frameworks. A follow-up independent real-world cohort of 270 patients tested between 2021 and 2026 was analyzed to assess operational feasibility and scalability under routine clinical laboratory conditions. Results: A total of 385 longitudinal ctDNA assessments were performed, of which 357 were imaging-concurrent and comprised 90 partial response (PR), 117 stable disease (SD), and 150 progressive disease (PD) observations. ctDNA VAF differed significantly across RECIST categories (Kruskal–Wallis H = 19.09, p = 0.000072), with median VAF increasing from 0.055% in PR to 0.190% in SD and 0.210% in PD. In the primary tumor-burden analysis of 210 imaging–ctDNA pairs from 53 patients, ctDNA VAF showed weak positive correlations with estimated tumor volume (Spearman ρ = 0.169, p = 0.0144) and RECIST SLD (ρ = 0.177, p = 0.0101); both associations remained significant in mixed-effects models accounting for repeated measurements. Across 50 evaluable longitudinal trajectories, 21 (42.0%) demonstrated strong molecular–radiological concordance. Among 14 patients who developed radiological progression after an initial PR or SD, molecular progression preceded radiological progression in 9 (64.3%), with a median MLT of 55 days (IQR 46–83; range 28–209; Wilcoxon signed-rank p = 0.0039). In the real-world cohort, 650 personalized ddPCR assays were performed across five genomic alteration classes - 494 single-nucleotide variant, 93 insertion/deletion, 53 copy-number amplification, 7 fusion, and 3 promoter mutation assays - with a median assay-specific limit of detection of 0.020% VAF (IQR 0.020–0.030%). Conclusions: Personalized tumor-informed ddPCR complements radiological assessment by distinguishing response states at the population level, characterizing longitudinal molecular–radiological concordance, and identifying molecular progression before conventional imaging in selected patients. The substantial overlap in absolute VAF across response categories and the weak association with anatomical tumor burden support interpretation of serial within-patient molecular trajectories rather than universal VAF thresholds. Real-world implementation across 270 patients and five genomic alteration classes further supports the operational feasibility and scalability of this approach and provides a framework for prospective ctDNA-guided interventional studies.

Source: International Journal of Molecular Sciences

DOI: 10.3390/ijms27188340