Risk Prediction Model for Advanced Colorectal Neoplasia May Aid in Selection of Screening Options

colon cancer
colon cancer, CRC, colorectal neoplasia
Knowing risk for advanced colorectal neoplasia (AN) could help patients and providers choose among screening tests, researchers created a risk prediction model for AN to help decide which test might be selected.

A risk prediction model based on sociodemographic, physical, and lifestyle features may accurately estimate the risk for advanced colorectal neoplasia (AN) with high discrimination in persons at average risk, researchers reported in Gut.

The study authors recruited persons with an average risk of AN (aged 50 to 80 years) who were undergoing their first colonoscopy screening. All participants completed a mailed survey regarding sociodemographic features, family history, personal medical history, lifestyle habits, and medication use and were asked to measure and record their height, weight, and waist and hip circumferences. Prior to the colonoscopy, nursing personnel recorded the 4 physical measures, which were used preferentially in the analysis.

The investigators created a risk equation for two-thirds of participants (the derivation set) and assigned points to each variable to create a risk score; the remaining one-third of participants were the validation set. Demographic and clinical variables of the derivation (n = 3025; mean age 57.3 ± 6.5 years; 52% women; 94% white) and validation (n = 1475; mean age 57.2 ± 7.0 years; 52% women; 94% white) subgroups were comparable.

In the derivation set, the AN prevalence was 9.4%. The multivariable model yielded 3 risk groups, with AN risks of 1.5% (95% CI, 0.72-2.74) in the low-risk group, 7.06% (CI, 5.89-8.38) in the intermediate-risk group, and 27.26% (CI, 23.47-31.30) in the high-risk group (P < .001), including 23%, 59%, and 18% of participants, respectively. A total of 10 low-risk participants had AN, none had colorectal cancer, and 6 of 10 low-risk participants with AN had a distal location. Based on finding a distal polyp, a screening sigmoidoscopy followed by colonoscopy to examine the proximal colon would have detected 7 (70%) of these advanced polyps.

In the validation set, the AN prevalence was 8.4%. The model demonstrated comparable results, with AN risks of 2.73% (CI, 1.25-5.11) in the low-risk group, 5.57% (CI, 4.12-7.34) in the intermediate-risk group, and 25.79% (CI, 20.51-31.66) in the high-risk groups (P < .001), including 23%, 59%, and 18% of participants, respectively. A total of 9 low-risk participants had AN, none had colorectal cancer, and 5 of this group had distal lesions. Screening sigmoidoscopy followed by colonoscopy for a distal polyp would have detected 6 of 9 persons with AN.

“Among average-risk persons, this model estimates AN risk with high discrimination, identifying a lower risk subgroup that may be screened non-invasively and a higher risk subgroup for which colonoscopy may be preferred,” stated the investigators.

One of several study limitations is the predominantly white cohort undergoing a first screening colonoscopy. The study also included 13 variables, some of which may be difficult for users to understand and respond accurately, noted the study authors.

“Use of a risk prediction model, that personalizes one’s current risk for advanced neoplasia, could guide doctor–patient discussions about screening options and may increase screening uptake and efficiency, especially under conditions of constrained colonoscopy resources,” the researchers commented.


Imperiale TF, Monahan PO, Stump TE, Ransohoff DF. Derivation and validation of a predictive model for advanced colorectal neoplasia in asymptomatic adults. Published online September 29, 2020. Gut. doi: 10.1136/ gutjnl-2020-321698