METHYLSCAPE-COÉvaluation du test Methylscape pour la détection et le suivi universels du cancer

Background. Most cancers lack reliable screening methods, often leading to advanced-stage diagnosis. Multi-cancer early detection (MCED) tests based on body fluids can screen for several cancer types from a single sample. DNA methylation, which occurs early in carcinogenesis and is tissue-specific, is the most commonly used marker in MCED assays. Methylscape leverages global differences in the genomic distribution of methylation between cancerous and normal tissues, detected electrochemically through differential adsorption of DNA on gold electrodes. Objective. To determine whether Methylscape can detect the methylation landscape across multiple cancer types with sufficiently high specificity to predict the cancer signal origin (CSO), and to assess its potential application as a population-scale MCED test among Colombians. Design. Single-center prospective observational study conducted at CTIC (Bogotá, Colombia), with sample processing at CTIC and Columbia University. Phase 1 (N=250 patients with cancer) provides initial validation across solid tumor types selected by local incidence. Sub-study 2a (1,500 patients with cancer and 1,500 matched cancer-free volunteers) provides large-scale clinical validation. Sub-study 2b (N=300 early-stage patients) evaluates serial blood/urine Methylscape testing for relapse detection during follow-up every 6 months. Phase 1 participants may enter Sub-studies 2a/2b only if they meet the corresponding criteria and provide new informed consent; cross-phase participation is flagged in the database to adjust statistical estimates. Biospecimens \& reference standards. Blood/plasma (K2EDTA), urine, saliva, and FFPE tumor tissue are collected and stored at -80 °C in the CTIC biobank. Cancers are coded with WHO ICD-O-3; stage per AJCC 8th edition. Reference standards: histopathology (MCED), RECIST 1.1 or biopsy (relapse), and absence of cancer by record review plus 12-month follow-up (cancer-free). Statistical analysis. Sensitivity and specificity are estimated overall and by stage/type (expected sensitivity 85-90%, specificity 98%; 95% confidence, 80% power). PPV/NPV are derived via Bayes' theorem with prevalence from GLOBOCAN 2022; 95% CIs by stratified bootstrap (1,000 resamples). Budget impact and cost-effectiveness (QALYs) are projected via microsimulation and Markov models.