P-13: Toward population-based type 1 diabetes screening: agreement between capillary and venous blood for islet autoantibody detection using lips in a pediatric cohort
R Andrade, R Coelho, R Ribeiro, M A Silva, V Lampasona, I Marzinotto, C Brigatti, C Mathieu, A -G Ziegler, J RaposoAbstract
Introduction
Islet autoantibodies are established biomarkers of type 1 diabetes (T1D), traditionally measured in venous blood. Capillary sampling is a minimally invasive alternative for large-scale pediatric screening if adequate sample volume and analytical equivalence to venous blood are demonstrated.
Objectives
To assess agreement between capillary and venous blood for detection and quantification of islet autoantibodies associated with early-stage T1D.
Methods
Within the European EDENT1FI project, a Portuguese cohort of 12,312 children was screened (early-stage T1D prevalence 0.32%). Capillary blood (200 µL) was first tested with a combined GAD/IA-2/ZnT8 ELISA (3ScreenICA; 25 µL); positives were re-tested in single-antigen LIPS (1 µL) for GAD, IA-2, ZnT8 and IAA, with venous confirmation at a separate visit. A subset of 99 paired capillary and venous samples was analysed, using Cohen’s κ and McNemar’s test (qualitative), and Spearman correlation, Wilcoxon test, linear regression and Bland–Altman analysis (quantitative).
Results
High qualitative agreement (≥89%) was observed across all autoantibodies, with good-to-excellent κ (GAD 0.664; IA-2 1.000, no discordant pairs; ZnT8 0.846; IAA 0.768). No significant asymmetry was found for IA-2, ZnT8 or IAA (p>0.05); GAD showed asymmetric discordance (p=0.021), with capillary detecting additional positives not confirmed on venous blood. Quantitatively, no significant differences were observed (Wilcoxon p>0.05), with moderate-to-strong correlations (Spearman r=0.684–0.941, p<0.0001). Bland–Altman showed minimal bias for IA-2, IAA and GAD, higher variability for ZnT8, and proportional bias for GAD and ZnT8.
Conclusions
Capillary blood is highly reliable for qualitative autoantibody detection by LIPS. Quantitative agreement varies by analyte, strongest for IA-2 and IAA and lower for GAD and ZnT8. These findings support a two-step strategy combining capillary screening with venous confirmation to balance feasibility, adherence and diagnostic accuracy.