Clinical performance of an automated stool DNA assay for detection of colorectal neoplasia.

ABSTRACT

BACKGROUND & AIMS: Colorectal cancer (CRC) and advanced precancers can be detected noninvasively by analyses of exfoliated DNA markers and hemoglobin in stool. Practical and cost-effective application of a stool DNA-based (sDNA) test for general CRC screening requires high levels of accuracy and high-capacity throughput. We optimized an automated sDNA assay and evaluated its clinical performance. METHODS: In a blinded, multicenter, case-control study, we collected stools from 459 asymptomatic patients before screening or surveillance colonoscopies and from 544 referred patients. Cases included CRC (n = 93), advanced adenoma (AA) (n = 84), or sessile serrated adenoma ≥1 cm (SSA) (n = 30); controls included nonadvanced polyps (n = 155) or no colonic lesions (n = 641). Samples were analyzed by using an automated multi-target sDNA assay to measure ß-actin (a marker of total human DNA), mutant KRAS, aberrantly methylated BMP3 and NDRG4, and fecal hemoglobin. Data were analyzed by a logistic algorithm to categorize patients as positive or negative for advanced colorectal neoplasia (CRC, advanced adenoma, and/or SSA ≥1 cm). RESULTS: At 90% specificity, sDNA analysis identified individuals with CRC with 98% sensitivity. Its sensitivity for stage I cancer was 95%, for stage II cancer it was 100%, for stage III cancer it was 96%, for stage IV cancer it was 100%, and for stages I-III cancers it was 97% (nonsignificant P value). Its sensitivity for advanced precancers (AA and SSA) ≥1 cm was 57%, for >2 cm it was 73%, and for >3 cm it was 83%. The assay detected AA with high-grade dysplasia with 83% sensitivity. CONCLUSIONS: We developed an automated, multi-target sDNA assay that detects CRC and premalignant lesions with levels of accuracy previously demonstrated with a manual process. This automated high-throughput system could be a widely accessible noninvasive approach to general CRC screening.