RESUMEN
BACKGROUND: The 2017 revision of the McDonald criteria highlights the usefulness of cerebrospinal fluid (CSF) immunoglobulin G (IgG) analysis to diagnose multiple sclerosis (MS). The objective of this study was to assess the diagnostic performances of CSF IgG analysis in the absence of a gold standard. METHODS: All patients who underwent CSF IgG analysis for events suggestive of MS in Nancy University Hospital (France) from 2008 to 2011 were retrospectively included. A latent class analysis with Bayesian approach was used to infer MS prevalence (latent variable) as well as the diagnostic properties of the 2005 and 2010 McDonald criteria and CSF IgG analysis (observed variables). RESULTS: Data from 673 patients were analysed. For CSF IgG analysis, the Bayesian latent class analysis estimated sensitivity of 0.93 (95% CrI 0.89-0.96) and specificity of 0.81 (95% CrI 0.77-0.85). The true prevalence estimate was 36% (95% CrI 0.33-0.40). Sensitivity and specificity estimates for patients with events suggestive of remitting-onset MS were similar to those for the whole sample-0.92 (95% CrI 0.85-0.96) and 0.80 (95% CrI 0.76-0.84), respectively-but higher for patients with signs of progressive-onset MS-0.95 (95% CrI 0.84-0.99) and 0.88 (95% CrI 0.78-0.94), respectively. CONCLUSIONS: In the absence of a gold standard, latent class analysis indicates good diagnostic properties of CSF IgG analysis for MS. This test could thus be useful, especially for patients who tested negative for the 2005 and 2010 McDonald criteria. These findings deserve to be confirmed prospectively.