Analytical validation of EPISEEK, an epigenomic blood-based assay for multicancer detection.

3144 Background: Early cancer detection significantly improves treatment outcomes and survival rates. However, cancer screening faces key challenges: (1) current tests detect only 14% of new cases and cover a limited range of cancers, (2) each cancer requires its own costly and complex screening process, (3) limited patient awareness of suitable screenings, and (4) poor adoption among marginalized and underinsured groups. EPISEEK was developed for multicancer detection using minimal cell-free DNA from plasma. Here, we present validation study results assessing its robustness and accuracy across 20+ cancer types and stages. Methods: EPISEEK is a cfDNA-based methylation assay optimized for 20 ng of cfDNA input from plasma. Plasma cfDNA underwent bisulfite conversion followed by methylation-specific quantitative PCR targeting 10 cancer biomarkers and 3 internal control markers. 251 plasma samples from four cancer stages across 25 cancer types, and 57 samples from individuals over 40 with no known cancer history, were used to establish reference ranges and assess assay specificity and sensitivity. Additional contrived and clinical samples were used to determine the assay's analytical LOD, reproducibility and stability. Results: 57 non-cancer samples were used to train the classifier, achieving 99% specificity at a 95% confidence level. Accuracy testing included 251 cancer samples representing > 20 primary cancer sites, including lung, colon, cervix, esophagus, head and neck, kidney, liver, breast, bladder, skin, testis, thyroid, ovary, pancreas, prostate, stomach, brain, bone marrow, and others. The cancer samples spanned all stages: I (29%), II (13%), III (29%), IV (22%), and cases with missing stage data (6.7%). Sensitivity increased with advancing stage, with observed sensitivity rate of 52%. By stage, observed sensitivity was stage I: 42%, stage II: 46%, stage III: 57%, and stage IV: 64%. Due to the limited and non-representative sample distribution for a typical multicancer screening population, SEER data were utilized to estimate EPISEEK’s real-world performance by adjusting tumor incidence and stage when estimating positive predictive value and negative predictive value. At 99% specificity and based on adjusted performance, EPISEEK achieved a positive predictive value (PPV) of 40% and a negative predictive value (NPV) of 99%. Seven markers were detectable with < 0.1 ng DNA, while the remaining three markers had an LOD95 of 0.1–0.37 ng. Comparing Ct values of each cancer target both intra and inter runs, EPISEEK demonstrated high reproducibility with standard deviation of 0.383 (high positive), 0.232 (low positive) and 1.063 (negative samples). Conclusions: EPISEEK is a sensitive, specific, accurate, and reproducible multicancer detection test. Compared to comprehensive genomic profiling techniques, it offers an affordable testing option for broader populations with a fast turnaround time.