Low agreement between radio binding assays in analyzing glutamic acid decarboxylase (GAD65Ab) autoantibodies in patients classified with type 2 diabetes

Abstract

Autoantibodies against glutamic acid decarboxylase (GAD65Ab) are used in the classification of diabetes in adults. We assessed the concordance in GAD65 autoantibody levels within subjects between three different GAD65Ab radio binding assays (RBA). Plasma samples from 112 diabetes patients (median age 50 years) initially classified with type 2 diabetes was randomly selected from a local diabetes registry. Coded samples were analyzed with two RBA employing (35)S-labeled GAD65. The first used the pEx9 plasmid (pEx9 RBA), the second employed the pThGAD65 plasmid (pThGAD65 RBA) to label GAD65 by in vitro transcription translation. We also used a commercial kit employing plasmid pGAD17 labelled with (125)I (pGAD17 RBA). Subsequent analyses followed standard procedures. Two different cut-offs for GAD65Ab positivity were used in all three assays. We calculated the correlation, concordance, and agreement between the assays. The proportion of GAD65Ab positivity differed between assays when low cut-offs were used (pEx9 RBA 25%, pThGAD65 RBA 17.9%, and pGAD17 RBA 12.5%, respectively). When high cut-offs were applied, the concordance between the pEx9 RBA and the pThGAD65 RBA was 97.3 while their concordance to the pGAD17 RBA was lower (88.4 and 87.4, respectively). There was a low agreement between both pEx9 RBA and pGAD17 RBA (0.45, 95% CI 0.20-0.70) and between pThGAD65 RBA and pGAD17 RBA (0.43, 95% CI 0.18-0.68). We found discrepancies in determining the GAD65Ab positivity, which constitutes a problem when GAD65Ab are used clinically. Further methodological GAD65Ab assays studies are warranted.