Optimization of serologic diagnosis of celiac disease in the pediatric setting

Background: The clinical presentation of celiac disease (CD) varies between children. The objective of this study was to document the pre-test probability for CD based on symptoms and routine laboratory test and to evaluate the performance of two IgA anti-tissue transglutaminase (tTG) assays. We critically reviewed the concept of using multiples of the manufacturer's upper limit of normal (ULN), as proposed in the ESPGHAN guidelines (if IgA tTG is > 10 times ULN, no biopsy is needed). Methods: The retrospective study included 91 children with newly diagnosed CD and 605 controls ( < 16 years). All underwent upper endoscopy with small bowel biopsies. Four laboratory parameters and 16 symptoms were registered. All patients were tested for IgA anti-tTG antibodies with assays from Inova Diagnostics and Thermo Fisher Scientific. Results: Some combinations of clinical symptoms and laboratory parameters had a high pre-test probability for CD, such as (combinations of) anorexia, failure to thrive, low ferritin level and elevated AST. The diagnostic performance of both IgA anti-tTG assays was excellent and comparable (no difference in ROC curve area under the curve). At a threshold that corresponds to a specificity of 100% (5 times ULN for Inova Diagnostics and 2 times ULN for Thermo Fisher), the sensitivity was 82% for both assays. At the 10 times ULN threshold, the sensitivity differed between the assays (77% vs. 57%), indicating that such threshold does not completely align interpretation across companies. Conclusions: Our study showed that some combinations of symptoms and aberrant laboratory parameters had a high pre-test probability. The use of the ESPGHAN non-biopsy approach could reduce small bowel biopsies, but thresholds for IgA-tTG levels are not aligned across assays and should be based on predefined likelihood ratios or specificity.