Abstract
The existence of an imperfect reference standard presents complications when evaluating the unbiased performance of novel diagnostic techniques. This is especially true in the absence of a gold standard, as is the case in chronic Chagas disease (CD) diagnosis. To circumvent this constraint, we elected to use latent class analysis (LCA). Previously, our group demonstrated the high performance of four Trypanosoma cruzi-chimeric proteins (Molecular Biology Institute of Paraná [IBMP]-8.1, -8.2, -8.3, and -8.4) for CD diagnosis using several distinct immunoassays. Although commercial tests had previously been established as a reference standard, the diagnostic performance of these chimeric antigens could present bias because these tests fail to produce 100% accurate results. Thus, we used LCA to assess the performance of these IBMP chimeric antigens in chronic CD diagnosis. Using the LCA model as a gold standard, sensitivity and specificity values ranged from 93.5% to 99.4% and 99.6% to 100%, respectively. The accuracy values were 96.2% for IBMP-8.2, approximately 98% for IBMP-8.1 and IBMP-8.3, and nearly 100% for IBMP-8.4. For IBMP-8.1 and IBMP-8.2, higher positive predictive values were associated with increases in hypothetical prevalence. Similarly, higher hypothetical prevalence resulted in lower negative predictive values for IBMP-8.1, IBMP-8.2, and IBMP-8.3. In addition, samples with serodiscordant results from commercial serological tests were analyzed using LCA. Molecular Biology Institute of Paraná -8.1 demonstrated potential for use in confirmatory testing with regard to samples with inconsistent results. Moreover, our findings further confirmed the remarkable performance of the IBMP-8.4 antigen to diagnose chronic CD in both endemic and non-endemic areas.