Comparison of three methods for determining anti-thyrotropin receptor antibodies (TRAb) for diagnosis of Graves’ disease: a clinical validation

Graves' disease is secondary to the presence of anti-thyrotropin receptor antibodies (TRAb), which stimulate thyroid hormones. TRab determination is crucial for etiological diagnosis. The objectives of this study were (i) to compare two methods for determining TRab by chemoluminiscence vs. standard TRACE-immunofluorescence; (ii) to determine the diagnostic validity of the three methods. A retrospective study in 194 patients with a TRAb determination request. TRAb were determined by immunofluorescence (Kryptor, ThermoFisher) and chemiluminescence (Immulite, Siemens and Maglumi, Snibe). Clinical validation: medical records were reviewed and categorized according to thyroid function. Statistical analysis: Differences in were assessed by intraclass correlation coefficient, Bland-Altman plot, and mean differences (mD). Qualitative variables were dichotomized by cut-off points; Kappa coefficient was calculated. Correlations were evaluated by Pearson's coefficient and Passing-Bablok regression analysis. The diagnostic validity of the three methods was investigated. Kryptor-Immulite: mD: 1.2 (95%CI: -16 to >18). Passing-Bablok: Constant error (95%CI: -0.8349 to -0.5987). Proportional error (95%CI: 0.7862-1.0387). ICC: 0.86 (95%CI: 0.82-0.89). Kappa coefficient: 0.68 (95%CI 0.59-0.78). Kryptor-Maglumi: mD: -0.3 (95%CI: -12 to >12). Passing-Bablok: Constant error (95%CI: -0.7701 to >0.1621. Proportional error (95%CI: 0.8571 to 1.3179. ICC: 0.93 (95%CI: 0.89-0.97). Kappa coefficient: 0.53 (95%CI: 0.32-0.74). Diagnosis of Graves' disease was confirmed in 113 patients (Kryptorf showed better specificity and positive predictive value, whereas Immulite demonstrated better sensitivity and negative predictive value). The three methods have a good diagnostic performance for Graves' disease, with superimposable results on Bland-Altman plot. Interchangeability was not confirmed on the regression and agreement analysis, with the presence of biases.