Prediction of celiac disease at endoscopy.

BACKGROUND AND STUDY AIMS: Celiac disease is increasingly recognized worldwide, but guidelines on how to detect the condition and diagnose patients are unclear. In this study the prevalence and predictors of celiac disease were prospectively determined in a cross-sectional sample of Lebanese patients undergoing esophagogastroduodenoscopy (EGD). PATIENTS AND METHODS: Consecutive consenting patients (n = 999) undergoing EGD answered a questionnaire and had blood taken for serologic testing. Endoscopic markers for celiac disease were documented and duodenal biopsies were obtained. The diagnosis of celiac disease was based on abnormal duodenal histology and positive serology. Risk factors were used to classify patients to either high or low risk for celiac disease. Independent predictors of celiac disease were derived via multivariate logistic regression. RESULTS: Villous atrophy (Marsh 3) and celiac disease were present in 1.8 % and 1.5 % of patients, respectively. Most were missed on clinical and endoscopic grounds. The sensitivity of tissue transglutaminase (tTG) testing for the diagnosis of villous atrophy and celiac disease was 72.2 % and 86.7 %, respectively. The positive predictive value of the deamidated gliadin peptide (DGP) test was 34.2 % and that of a strongly positive tTG was 80 %. While the strongest predictor of celiac disease was a positive tTG (odds ratio [OR] 131.7, 95 % confidence interval [CI] 29.0 - 598.6), endoscopic features of villous atrophy (OR 64.8, 95 %CI 10.7 - 391.3), history of eczema (OR 4.6, 95 %CI 0.8 - 28.8), anemia (OR 6.7, 95 %CI 1.2 - 38.4), and being Shiite (OR 5.4, 95 %CI 1.1 - 26.6) significantly predicted celiac disease. A strategy of biopsying the duodenum based on independent predictors had a sensitivity of 93 % - 100 % for the diagnosis of celiac disease, with an acceptable (22 % - 26 %) rate of performing unnecessary biopsies. A strategy that excluded pre-EGD serology produced a sensitivity of 93 % - 94 % and an unnecessary biopsy rate of 52 %. CONCLUSION: An approach based solely on standard clinical suspicion and endoscopic findings is associated with a significant miss rate for celiac disease. A strategy to biopsy based on the derived celiac disease prediction models using easily obtained information prior to or during endoscopy, maximized the diagnosis while minimizing unnecessary biopsies.