A Predictive Model of Duodenal Atrophy in Adults With Celiac Disease: Could Diagnostic Endoscopy Be Avoided?

In adults, the diagnosis of celiac disease (CeD) relies on the presence of disease-specific antibodies and the confirmation of duodenal atrophy. This study aimed to develop a diagnostic model to predict duodenal atrophy in adults, with the goal of assessing the reliability of a non-biopsy diagnostic approach based on serology.

We conducted a retrospective observational study including consecutive adult patients who attended the Gastroenterology Department of Hospital de Clínicas between 2004 and 2023. Group 1 comprised patients with positive tTG and Marsh 3 lesions, whereas group 2 included patients with normal duodenal biopsies, irrespective of tTG results. Logistic regression models were applied using different tTG cutoff values (dependent variable) to discriminate between "CeD" and "non-CeD".

A total of 211 patients met criteria for group 1 and 267 for group 2. Specificity exceeded 98% at tTG levels ≥2.5 upper the normal level (UNL), with the highest specificity observed at 109 U (5.5 UNL). Model development was restricted to individuals with complete data on all significant variables (n=290; 134 cases, 156 controls). Female sex and folic acid deficiency emerged as significant predictors, increasing sensitivity but not specificity at the 3 and 4 UNL.

In this cohort, a TTG value >5.5 UNL achieved 100% specificity and could potentially obviate the need for diagnostic endoscopy in adults with suspected CeD. Although female sex and folic acid deficiency were independent predictors of CeD, their inclusion did not enhance the predictive performance at a lower tTG cutoffs.