A simple scoring model for advanced colorectal neoplasm in asymptomatic subjects aged 40–49 years

A cross-sectional study was conducted with 2781 asymptomatic adults aged 40–49 years who underwent screening colonoscopy for the first time from January 2008 to January 2012 at the Health Promotion Center of the Gangnam Severance Hospital in Seoul, South Korea. We excluded the following criteria: 1) previous colorectal examinations, including colonoscopy, sigmoidoscopy, or barium enema; 2) colonoscopies that had inadequate bowel preparation and did not reach the cecum; 3) individuals who had a personal history of colorectal neoplasm and inflammatory bowel disease; 4) a history of colorectal surgery, and 5) any symptoms, weight loss, anemia, or bleeding. Subjects were randomly allocated to a development or validation set in a 2:1 ratio.

Bowel preparation was performed using 4 L of polyethylene glycol solution, and subjects underwent a 3-day dietary restriction. The quality of bowel preparation was determined by the physician who used the following descriptors: excellent, good, fair/adequate, inadequate, and poor. Those that were inadequate and poor were labeled as poor preparation and were not included in the final analysis. Colonoscopies were performed by four endoscopists who had performed a minimum of 1000 colonoscopies, and all endoscopists were gastroenterology fellowship-trained and board-certified in their respective field. All examinations were performed using a standard video colonoscope (CF-H260AI; Olympus, Tokyo, Japan). All detected polyps were biopsied or removed. All polyp characteristics such as the size, number, shape, and location were documented. Polyp size was grossly estimated using open-biopsy forceps (Olympus FB-28U-1; Aomori Olympus Co., Ltd., Aomori, Japan). The polyp shape was classified as sessile (Is), semipedunculated (Isp), or pedunculated (Ip) type [12].

Subjects’ information, including demographics, laboratory tests, colonoscopic findings, and pathology reports, by review of electronic medical records was collected. Age was categorized into two groups: 40–44 years and 45–49 years. Body mass index (BMI) was categorized by the Western Pacific Regional Office of the World Health Organization criteria: normal (<23 kg/m²), overweight (23–24.9 kg/m²), or obese (≥25 kg/m²) [13]. According to the American Joint Committee’s (Joint National Committee) seventh report, blood pressure was classified as normal (<120/80 mmHg), prehypertension (120–139/80–89 mmHg), or hypertension (≥140–90 mmHg or taking antihypertensive drugs). Laboratory tests included immunoglobulin G specific for Helicobacter pylori, which was screened by an enzyme-linked fluorescence assay in each serum (ELFA, enzyme-linked via Vidas; bioMerieux Vitek, Inc., USA); elevated total cholesterol (≥240 mg/dL); elevated triglycerides (≥200 mg/dL); elevated low-density cholesterol (≥100 mg/dL); and low high-density lipoprotein cholesterol (HDL-c, <40 mg/dL). All lipid and lipoprotein levels were measured using a Hitachi 7600 Modular Dp-110 auto-analyzer, which included enzymatic colorimetric tests. Based on the normal reference range, serum lipid profiles were classified as normal or abnormal results.

We collected polyp data from colonoscopy reports. Advanced colorectal neoplasms were defined as follows: 1) CRC; 2) adenoma with a diameter of ≥10 mm; 3) tubular adenoma with high-grade dysplasia; and 4) tubulovillous or villous adenoma. When two or more adenoma were detected, the most advanced lesion was analyzed based on the largest diameter or advanced histology. All polyps were evaluated histologically by gastroenterology pathologists.

Continuous variables were presented as mean ± standard deviation, and categorical variables were expressed as percentages. Continuous variables were compared using Student t-test, whereas categorical variables were compared using chi-square or Fisher exact tests. For prediction model development, differences between variables were compared in the training set. Before univariate analysis, continuous variables were converted to categorical variables. Variables with a P-value <0.1 in univariate analysis were included in subsequent multivariate regression analysis in order to select variables to be implemented in the final model. Backward elimination (i.e., removing the covariate with the largest P value, one at a time) was performed until we developed a final model with statistically significant covariates. We established a risk score model by excluding less significant variables in a risk assessment. We intentionally only used categorized variables that captured easy but relevant and validated information in the prediction model to develop an easy to use screening score. We used a weighted scoring system by rounding down odds ratios (ORs) to the nearest integer in the final model. For example, an OR of 1.96 was rounded to 1 and an OR of 2.26 was rounded to 2. Based on this statistical analysis, we developed a formula for predicting asymptomatic subjects aged 40–49 years with advanced neoplasm, and these subjects may be primary candidates for screening colonoscopy. To validate the model, we assessed the diagnostic accuracy of the model in the validation set. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated using the computed area under receiver operating characteristic curve (AUROC). Thereafter, we validated the diagnostic value of the scoring model. P < 0.05 was considered to indicate statistical significance. All analyses were performed using SPSS software (version 20.0; SPSS Inc.) and MedCalc software (version 11.1; Mariakerke) for the receiver operating characteristic analysis.

Among 2781 subjects, mean age was 44.8 ± 2.8 years, and 58.7% (1633/2781) were men. Helicobacter serology positivity was observed in 1623 (58.4%) patients. Regarding the prevalence of colorectal neoplasm, advanced neoplasm was detected in 70 (2.5%) of patients, whereas any type of colorectal neoplasm was found in 561 (20.2) of patients. Baseline characteristics of 2781 patients in the training and validation sets are summarized in Table 1. There was no statistically significant difference in patient characteristics between the training and validation sets.

Details of the clinicopathological findings of neoplasm are summarized in Table 2. Among 561 neoplasms, 64 (11.4%) were larger than 10 mm, 537 (95.7%) had a low-grade tubular adenoma, and 24 (4.3%) had advanced neoplasm, including 2 patients (0.4%) with cancer.

Univariate analysis showed that male sex (P ≤ 0.001), the obese group (P = 0.016), positive serology of H. pylori (P = 0.009), low HDL level (P ≤ 0.001), and high triglyceride level (P ≤ 0.001) were significantly associated with the presence of advanced neoplasm. Moreover, the older age group (45–49 years) was marginally associated with the presence of advanced neoplasm (P = 0.053). Details of the statistical analysis are summarized in Table 3. In subsequent multivariate analysis, the older age group (OR 1.967, 95% CI 1.191–3.749, P = 0.040), male sex (OR 2.763, 95% CI 1.032–6.409, P = 0.018), positive serology of H. pylori (OR 2.262, 95% CI 1.108–4.621, P = 0.025), low HDL level (OR 2.219, 95% CI 1.073–4.308, P = 0.031), and high triglyceride level (OR 1.967, 95% CI 1.143–4.252, P = 0.018) were identified as independent factors for advanced neoplasm. BMI (body mass index) was not significant in multivariable analysis. Details of these values are depicted in Table 4.

A simple scoring model was constructed based on five independently significant variables in multivariate analysis by using rounded down ORs of each variable (Table 4). The final model is as follows:

A simple scoring model for advanced colorectal neoplasm = Age [0: 40–44, 1: 45–49 years] × 1 + Sex [0: female, 1: male] × 2 + Serology of H. pylori [0: negative, 1: positive] × 2 + High triglyceride level [0: normal range, 1: high] × 2 + Low HDL level [0: normal range, 1: low] × 2

The range of the total score for this risk model was 0–9. This model yielded an AUROC of 0.74 for predicting advanced neoplasm in the development set (Fig. 1a).

We investigated the predicted value of different total score cut points in the validation sets (Table 5). A cutoff point of 4 was selected because it results in the highest value for the AUROC to indicate an individual at high risk for advanced neoplasm. In the validation set, this cutoff point designated 43.1% of subjects at high risk for advanced colorectal neoplasm, and yielded a sensitivity of 79.2%, specificity of 57.8%, PPV of 4.7%, and NPV of 99.1% with an AUROC of 0.72 (Fig. 1b). The prevalence of advanced neoplasm gradually increased as the total risk score increased, and these findings are depicted in Fig. 2.