22 Background: Blood testing for CRC screening is a non-invasive option that could improve suboptimal adherence. One such test, SimpleScreen CRC, detects base-level methylation patterns in cell-free DNA (cfDNA) isolated from plasma, using an artificial intelligence/machine learning-enabled classification model. Its clinical performance was recently validated in a pivotal study of 27,010 individuals at average risk for CRC (PREEMPT CRC). Since then, we improved the assay’s signal-to-noise ratio by optimizing key aspects of our methylated cfDNA detection platform, more than doubling molecular signal. We compared performance between the improved and previous assay versions in a case-control validation study, applying the locked SimpleScreen CRC classification model to paired data from independent clinical samples. We also projected the change in lifetime CRC incidence, CRC mortality, and life-years gained (LYG) in a validated microsimulation model. Methods: Clinical test performance was assessed using both versions in an average-risk cohort (N = 965; 210 CRC, 407 advanced precancerous lesions APL; 329 were collected pre-colonoscopy, 144 non-APL, and 204 negative). Specificity was set to 90% for each version to enable direct comparison. Sensitivity was calculated nominally and, to reflect the intended-use population (IUP), adjusted for age and sex using distributions from the US Census, and for CRC stage, APL subtype, and lesion size using distributions from PREEMPT CRC. Two-sided confidence intervals (CIs) were calculated using Wilson’s method. Clinical blends and contrived samples, respectively, were used to determine limit of detection (LoD; n = 90) and limit of blank (LoB; n = 92). Modeled outcomes were assessed per 1000 people for a cohort screened from ages 45 to 75 y (COSMOS CRC). Results: The improved test showed an IUP-adjusted CRC and APL sensitivity of 85.2% (95% CI: 78.5 - 90.1%) and 21.7% (17.1 - 27.2%), respectively, at 90% IUP-adjusted specificity, 1.7 points (-2.8 to +6.2) and 5.4 points (-0.2 to +10.9) higher than the previous version (CRC: 83.5% 76.6 - 88.7%; APL: 16.3% 12.3 - 21.4%). Nominal CRC sensitivity was 88.6% for the improved test vs 86.5%, and APL sensitivity was 27.0% vs 22.3%, respectively, at 90% specificity. Stage I, II and III CRC sensitivities were higher for the improved test, while both versions achieved 100% sensitivity in stage IV. The improved test had a 2.6-fold lower LoD and 2.5-fold lower LoB. Improvements were predicted to reduce lifetime CRC cases by 9% (39.0 vs 42.9) and CRC deaths by 10% (12.5 vs 13.9), and improve LYG by 7% (246 vs 229) compared with the previous version. Conclusions: Assay improvements to a CRC screening blood test resulted in higher CRC and APL sensitivity, and better predicted patient outcomes. The improved test will continue to be evaluated, and associated data will be submitted to the Food and Drug Administration after the approval of SimpleScreen CRC.
Shaukat et al. (Sat,) studied this question.