Retrospective Evaluation of Various Serological Assays and Multiple Parameters for Optimal Diagnosis of Lyme Neuroborreliosis in a Routine Clinical Setting

Abstract

Laboratory diagnosis of Lyme neuroborreliosis (LNB) is challenging, and validated diagnostic algorithms are lacking. Therefore, this retrospective cross-sectional study aimed to compare the diagnostic performance of seven commercial antibody assays for LNB diagnosis. Random forest (RF) modeling was conducted to investigate whether the diagnostic performance using the antibody assays could be improved by including several routine cerebrospinal fluid (CSF) parameters (i.e., leukocyte count, total protein, blood-CSF barrier functionality, and intrathecal total antibody synthesis), two-tier serology on serum, the CSF level of the B-cell chemokine (C-X-C motif) ligand 13 (CXCL13), and a Borrelia species PCR on CSF. In total, 156 patients were included who were classified as definite LNB (n = 10), possible LNB (n = 7), or non-LNB patient (n = 139) according to the criteria of the European Federation of Neurological Societies using a consensus strategy for intrathecal Borrelia-specific antibody synthesis. The seven antibody assays showed sensitivities ranging from 47.1% to 100% and specificities ranging from 95.7% to 100%. RF modeling demonstrated that the sensitivities of most antibody assays could be improved by including other parameters to the diagnostic repertoire for diagnosing LNB (range: 94.1% to 100%), although with slightly lower specificities (range: 92.8% to 96.4%). The most important parameters for LNB diagnosis are the detection of intrathecally produced Borrelia-specific antibodies, two-tier serology on serum, CSF-CXCL13, Reibergram classification, and pleocytosis. In conclusion, this study shows that LNB diagnosis is best supported using multiparameter analysis. Furthermore, a collaborative prospective study is proposed to investigate if a standardized diagnostic algorithm can be developed for improved LNB diagnosis. IMPORTANCE The diagnosis of LNB is established by clinical symptoms, pleocytosis, and proof of intrathecal synthesis of Borrelia-specific antibodies. Laboratory diagnosis of LNB is challenging, and validated diagnostic algorithms are lacking. Therefore, this retrospective cross-sectional study aimed to compare the diagnostic performance of seven commercial antibody assays for LNB diagnosis. Multiparameter analysis was conducted to investigate whether the diagnostic performance using the antibody assays could be improved by including several routine (CSF) parameters. The results of this study show that LNB diagnosis is best supported using the detection of intrathecally produced Borrelia-specific antibodies, two-tier serology on serum, CSF-CXCL13, Reibergram classification, and pleocytosis. Furthermore, we propose a collaborative prospective study to investigate the potential role of constructing a diagnostic algorithm using multiparameter analysis for improved LNB diagnosis.

Keywords: Borrelia; Lyme neuroborreliosis; Reibergram; antibody index; cerebrospinal fluid; intrathecal antibody synthesis; multiparameter analysis; random forest; two-tier serology.

FIG 1

Overview of the sensitivity and specificity (A) and the positive (PPV) and negative predictive value (NPV) (B) and 95% confidence intervals (CIs) of the five antibody assays tested on cerebrospinal fluid (CSF)-serum pairs and the two antibody assays tested on CSF only for IgM (M), IgG (G), or IgM and IgG combined (M+G). Cases consisted of definite and possible LNB patients, and controls consisted of non-LNB patients. The positives per total (A) are based on the number of pathological AI values (CSF-serum assays) or positive test results (CSF-only assays) among all the cases and are used to calculate the sensitivity. The negatives per total (A) are based on the number of normal AI values (CSF-serum assays) or negative test results (CSF-only assays) among all the controls and are used to calculate the specificity. The true positives (B) are cases that have either a pathological AI value (CSF-serum assays) or a positive test result (CSF-only assays) per total positives (i.e., all patients that have a pathological AI value [CSF-serum assays] or a positive test result [CSF-only assays]). The true negatives (B) are controls that have either a normal AI value (CSF-serum assays) or a negative test result (CSF-only assays) per total negatives (i.e., all patients that have a normal AI value [CSF-serum assays] or a negative test result [CSF-only assays]). The manufacturer of the recomLine immunoblot (IB) revised the interpretation of the recomLine IgG IB in January 2019 by increasing the point value of the VlsE band (Table S3), which had an effect on the test result (Table 3 and S2). Consequently, results are shown that include both the revised (rev) and old interpretation criteria. For the recomLine IgM IB, the PPV could not be calculated as this assay yielded no positive test results.