One threshold does not fit all: assay-specific validation of anti-tTG IgA tests for biopsy-free diagnosis of pediatric celiac disease.

Abstract

Background: The ESPGHAN biopsy-free diagnostic algorithm for pediatric celiac disease (CeD) requires serum IgA anti-tissue transglutaminase antibody (anti-tTG IgA) concentrations ≥10 times the upper limit of normal (ULN) in the first blood sample, together with positive IgA endomysial antibodies (EMA) in a second blood sample. However, the transferability of the 10×ULN anti-tTG IgA threshold across different analytical platforms remains uncertain owing to the lack of assay harmonization.

Methods: Eighty-four children (6-10 years) who underwent duodenal biopsy during the diagnostic work-up for suspected CeD were retrospectively evaluated. Histology was classified according to the Marsh-Oberhuber criteria (65 Marsh-Oberhuber 3, 12 Marsh-Oberhuber 2, and 7 Marsh-Oberhuber 1). Anti-tTG IgA antibodies were measured using one FEIA assay (EliA Celikey IgA) and two CLIA platforms (QUANTA Flash® and Maglumi®). EMA were assessed by indirect immunofluorescence on distal monkey esophagus. ROC analysis evaluated discrimination of mucosal atrophy (Marsh-Oberhuber 3) from non-atrophic lesions (Marsh-Oberhuber 1-2).

Results: Anti-tTG IgA concentrations correlated significantly with histological severity for all assays (Spearman's ρ=0.670-0.736; p<0.0001). AUCs ranged from 0.961 to 0.992. At 10×ULN, QUANTA Flash® showed 100% sensitivity and 47.4% specificity, whereas Maglumi® and EliA Celikey IgA showed 100% specificity with sensitivities of 70.8% and 86.2%, respectively. For QUANTA Flash®, 42×ULN was required to achieve 100% specificity. In analyses targeting PPV ≥95%, the corresponding thresholds were 7.54×ULN for Maglumi®, 28.55×ULN for QUANTA Flash®, and 6.10×ULN for EliA Celikey IgA.

Conclusions: In this selected anti-tTG IgA-positive cohort, the quantitative relationship between anti-tTG IgA and mucosal atrophy differed across automated assays. Expression of results as multiples of the ULN did not fully compensate for inter-assay analytical variability. These findings support assay-specific verification of decision limits, but the retrospective design and selection based on QUANTA Flash positivity limit generalization of the diagnostic performance estimates to unselected children undergoing evaluation for CeD.