ABSTRACT
BACKGROUND: The guidelines for performing endoscopy in dyspeptic patients based on clinical parameters alone have shown variable performance, and there is a need for better prediction tools. AIM: We aimed to prospectively develop and validate a simple clinical-cum-laboratory test-based scoring model to identify dyspeptic patients with high risk of upper gastrointestinal malignancy (UGIM). METHODS: Adult patients with dyspeptic symptoms were prospectively recruited over 5 years. Clinical details including alarm features were recorded, and blood tests for hemoglobin and albumin were done before endoscopy. The presence of UGIM was the primary outcome. Risk factors for UGIM were assessed, and based on the OR of significant factors, a predictive scoring model was constructed. ROC curve was plotted to identify optimal cutoff score. The model was validated using bootstrapping technique. RESULTS: The study included 2324 patients (41.9 ± 12.8 years; 33.4% females). UGIM was noted in 6.8% patients. The final model had following five positive predictors for UGIM-age > 40 years (OR 3.3, score 1); albumin ≤ 3.5 g% (OR 3.4, score 1); Hb ≤ 11 g% (OR 3.3, score 1); alarm features (OR 5.98, score 2); recent onset of symptoms (OR 8.7, score 3). ROC curve had an impressive AUC of 0.9 (0.88-0.93), and a score of 2 had 92.5% sensitivity in predicting UGIM. Validation by bootstrapping showed zero bias, which further strengthened our model. CONCLUSION: This simple clinical-cum-laboratory test-based model performed very well in identifying dyspeptic patients at risk of UGIM. This can serve as a useful decision-making tool for referral for endoscopy.
