Predicting CRC in Unscreened Populations by CBC
Douglas K. Rex, MD, FASGE, reviewing Schneider J, et al. Clin Gastroenterol Hepatol 2020 Apr 29.
About one-third of the U.S. population is unscreened for colorectal cancer (CRC). One approach to early detection in this population is to screen health care databases for laboratory, demographic, and/or other clinical features associated with a risk of CRC.
In the current nested case-control study, an algorithm based largely on levels and trends in complete blood count (CBC) as well as demographic factors was tested. Its sensitivity for predicting CRC within 6 months was 35.4%, including 51.8% in the proximal colon and 27.3% in the distal colon. The sensitivity for CRC stage II or less was 28.8%, and stage III and IV combined was 43.4%. The model’s performance did not differ by gender, race, or body mass index. Using the cutoff in the algorithm, the predictive odds ratio for proximal cancer was 34.7 and for distal cancer, it was 12.1.
![d-rex](/wp-content/uploads/d-rex.jpg)
COMMENTAn important question is how to identify persons at high risk for CRC in the large unscreened U.S. population without active participation by those persons. These data indicate that application of a machine learning algorithm based on CBC data can identify a cohort of persons within that population who have a relatively high risk of current CRC. Patients could then be actively notified of their need for colonoscopy, or if they decline, fecal immunochemical testing.
Note to readers: At the time we reviewed this paper, its publisher noted that it was not in final form and that subsequent changes might be made.
CITATION(S)
Schneider J, Layefsky E, Udaltsova N, Levin TR, Corley DA. Validation of an algorithm to identify patients at risk for colorectal cancer based on laboratory test and demographic data in diverse, community-based population. Clin Gastroenterol Hepatol 2020 Apr 29. (Epub ahead of print) (https://doi.org/10.1016/j.cgh.2020.04.054)