Objective To construct and validate a predictive model for gastric precancerous lesions based on urea breath test, serum pepsinogen (PG) and gastrin-17 (G-17).
Methods Participants who underwent endoscopic screening for upper gastrointestinal tumors were retrospectively enrolled as study subjects. Using random function, all participants were divided into training cohort of 2, 788 cases (comprising 1, 290 cases in precancerous lesion group and 1, 498 cases in control group) and validation cohort of 1, 194 cases (comprising 581 cases in precancerous lesion group and 613 cases in control group) at a ratio of 7 to 3. A simple model was established based on urea breath test, PG and G-17. Clinical data between the precancerous lesion group and the control group in the training cohort were compared. A predictive model for gastric precancerous lesions was constructed using multifactorial Logistic regression analysis, and a scoring model for gastric precancerous lesions (the complete model) was developed based on this predictive model. The complete model, the simple model, the new ABC method, and the Li's score were all included in the validation cohort to compare the predictive performance of the four models.
Results Multifactorial Logistic regression analysis indicated that male, smoking, positive Helicobacter pylori (Hp) infection, PGⅡ≥10.19 μg/L, the ratio of PGⅠ to PGⅡ(PGR) ≤11.87, and G-17≥3.82 pmol/L were independent risk factors for gastric precancerous lesions (P < 0.05). A predictive model for gastric precancerous lesions was constructed based on these risk factors, and the complete model was established based on the predictive model. The total score ranged from 0 to 12 (with 6 to 12 indicating a high-risk population for gastric precancerous lesions and 0 to 5 indicating a low-risk population). When the complete model, the simple model, the new ABC method, and the Li's score were included in the validation cohort for comparison, the predictive values of the complete model and the simple model were similar. Both models demonstrated higher sensitivity, specificity, positive predictive value, negative predictive value, and accuracy compared to the new ABC method and the Li's score. Furthermore, the diagnostic value of the simple model in the high-sensitivity region was slightly superior to that of the complete model.
Conclusion The simple model constructed based on the urea breath test, PG and G-17 exhibits favorable predictive efficacy, calibration, and clinical utility, and is of positive significance for the early identification of patients with gastric precancerous lesions.
DownLoad: