Recoveries at GADA concentrations near the diagnostic cutoff were 88-99% for the RBA, 106% for the ELISA, and 89% for the TR-IFMA.
The ELISA and RBA GADA showed diagnostic specificities of 99% (95% CI, 93-100%) and 96 (89-99)%, respectively.
The IA-2A TR-IFMA had a diagnostic specificity similar to that of the RBA [99 (93-110)% and 97 (91-99)%, respectively].
Areas under the ROC curves showed a significant difference (P <0.001) between GADA assays: 0.84 in the RBA (95% CI, 0.77-0.90) and 0.99 in the ELISA (95% CI, 0.95-1.00).
Passing-Bablok regression analysis for GADA yielded the following equation: ELISA = 0.6 RBA - 3.9 kIU/L (95% CI for intercept, -4.5 to -3.0 kIU/L; 95% CI for slope, 0.5-0.7), which shows a high correlation between methods (r = 0.94; P <0.001).
Regarding IA-2As, Passing-Bablok regression analysis yielded the following equation: TR-IFMA = 2.9 RBA - 1.0 kIU/L (95% CI for intercept, -6.0 to 2.2 kIU/L; 95% CI for slope, 1.9-4.2; r = 0.83; P <0.0001).
The diagnostic sensitivities and specificities obtained in the Diabetes Antibody Standardization Program 2003 for the GADA RBA and ELISA and the IA-2A RBA and TR-IFMA are listed in Table 1.
The recovery, [C.sub.VWR], and [CV.sub.T] near the cutoff value were similar in the RBA and ELISA and were comparable to the intra- and interassay CVs obtained previously (10).
The [CV.sub.WR] for IA-2A concentrations near the cutoff was better for the TR-IFMA than for the RBA, although the CVT behaved in an opposite manner at the same threshold.
The diagnostic sensitivities of IA-2A measurements obtained with the TR-IFMA and RBA in our patients with new-onset DM1 and in the Diabetes Antibody Standardization Program 2003 for patients with new-onset DM1 were consistent with the previously reported prevalence of IA-2A (40-70%) (1), although lower than the prevalences reported by others using the same methods (3).