2.6. Measurements of beta‐cell autoantibodies

GAD65Ab was analyzed in all 5898 samples by radioligand binding assay (RBA) as previously described. 15 Antibody levels were expressed as a relative index to correct for inter‐assay variation using the WHO standard for GAD65Ab. 16 To determine the relative index, positive and negative control samples were included in all assays. The assay showed 70% sensitivity and 98% specificity for GAD65Ab in the International Combined Autoantibody Workshop. 17 For statistical analyses and comparability, the index was translated into U/mL (Index 1 = 1000 U/mL).

To establish specific cut‐off values for GAD65Ab positivity in this population, we refrained from using arbitrary cut‐offs that are highly laboratory specific. Rather, a competition assay was performed in 59 participants in Ghana without diabetes and in 58 participants in Europe without diabetes, employing recombinant human (rh) GAD65 (Diamyd Medical, Sweden). 18 The samples were incubated with [³⁵S]‐GAD65 in the absence or presence of rhGAD65 (200 ng/mL). Samples whose binding to [³⁵S]‐GAD65 were reduced by 40% in the presence of rhGAD65 were considered as having specific GAD65Ab (true positive) (Figure S1). This binding strength was achieved at higher GAD65Ab concentrations in samples from Ghana than in samples from Europe, and was set as the gold standard in subsequent receiver operating characteristic analysis (Figure S2). At 78% sensitivity and 70% specificity, the GAD65Ab cut‐offs were 121 and 97 U/mL for the Ghanaian and European study sites, respectively.

In subsamples, we assessed the robustness of GAD65Ab detection using alternative laboratory methods (Table S1). These methods used for GAD65Ab detection include (a) enzyme‐linked immunosorbent assay (ELISA) using a kit by Kronus (Boise) (n = 124), (b) ELISA using a kit by DRG International Inc (n = 84), and (c) luciferase immunoprecipitation system (LIPS) (n = 140) performed in an independent laboratory at Helmholtz Centre Munich, Germany. 19 Also, we measured zinc transporter‐8 autoantibody (ZnT8Ab) as a novel marker for autoimmune diabetes. ZnT8Ab was measured under similar conditions as described for GAD65Ab with constructs containing the cytosolic segments (aa268‐369) encoding the aa325 codon variants, CGG (T) and TGG (W). Results for ZnT8Ab were converted into arbitrary units by extrapolation using a pan‐reactive positive serum from a patient with T1D with designated 1000 arbitrary units. Cut‐off was set at 10 U/mL for autoantibodies to ZnT8R and 18 U/mL for ZnT8W based on the 98th percentile observed in 162 individuals without diabetes.

Lastly, isotype‐specific identification of immunoglobulins (Ig) was performed using an ELISA (Bio‐Rad) in 316 samples (rural Ghana: n = 124; Europe: n = 192) to account for isotype shifts during different phases of the humoral immune response. Specifically, we measured total IgG 1, 2, 3, 4, IgE, IgA, and IgM. This analysis was performed because the RBA for GAD65Ab measurement does not discriminate GAD65Ab of different Ig isotypes.