gms | German Medical Science

62. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

17.09. - 21.09.2017, Oldenburg

Bivariate random effects meta-analysis assessing test accuracy when using the cerebrospinal fluid CXCL13 for diagnosis of acute Lyme neuroborreliosis

Meeting Abstract

  • Markus Pfirrmann - IBE - Institut für Medizinische Informationsverarbeitung, Biometrie und Epidemiologie, Ludwig-Maximilians-Universität München, München, Deutschland
  • Kirsi M. Manz - IBE - Institut für Medizinische Informationsverarbeitung, Biometrie und Epidemiologie, Ludwig-Maximilians-Universität München, München, Deutschland
  • Tobias A. Rupprecht - Abteilung für Neurologie, HELIOS-Klinik München West, München, Deutschland
  • Volker Fingerle - Bayerisches Landesamt für Gesundheit und Lebensmittelsicherheit (LGL), Oberschleißheim, Deutschland
  • Christian Lechner - Abteilung für Neurologie, HELIOS-Klinik München West, München, Deutschland
  • Uwe Koedel - Neurologische Klinik, Ludwig-Maximilians-Universität München, München, Deutschland

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. 62. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Oldenburg, 17.-21.09.2017. Düsseldorf: German Medical Science GMS Publishing House; 2017. DocAbstr. 265

doi: 10.3205/17gmds123, urn:nbn:de:0183-17gmds1236

Veröffentlicht: 29. August 2017

© 2017 Pfirrmann et al.
Dieser Artikel ist ein Open-Access-Artikel und steht unter den Lizenzbedingungen der Creative Commons Attribution 4.0 License (Namensnennung). Lizenz-Angaben siehe http://creativecommons.org/licenses/by/4.0/.


Gliederung

Text

Introduction: The utility of cerebrospinal fluid (CSF) CXCL13 for diagnosis of acute Lyme neuroborreliosis (LNB) has been debated [1]. Our study’s aim was to evaluate its overall diagnostic accuracy through bivariate random effects meta-analysis.

Methods: Electronic searches in PubMed MEDLINE and Web of Science were performed to identify relevant articles published before January 2016. All articles were evaluated using the Quality Assessment of Diagnostic Accuracy Studies method as suggested by the Cochrane Collaboration [2]. In accordance with prior publications, for CXCL13, the cut-off levels 142, 250, and 500 pg/ml were investigated. Since the sensitivity and specificity of a diagnostic test are correlated and dependent on each other, for their common analysis, the bivariate model approach of Reitsma et al. [3] was used. The resulting bivariate summary statistic of sensitivity and 1 – specificity (false positive rate) was given together with its confidence region. Diagnostic odds ratios (DOR) were calculated under consideration of a continuity correction of 0.5 for empty cell frequencies. For assessing the presence of publication bias, the method proposed by Deeks et al. [4] was applied. All statistical analyses were performed using the function “mada”, part of the statistical software R, Version 3.2.2.

Results: Of 136 studies identified by electronic searches, 12 qualified for the meta-analysis. They comprised 1766 patients, 376 with LNB and 1,390 with competing diseases. With 250 pg/ml as the best cut-off value, the pooled sensitivity for CSF CXCL13 was 0.86 (95% confidence interval (CI) 0.78-0.92) and the pooled specificity 0.94 (95% CI 0.87-0.98). Of note, the overall specificity depended on the proportion of the competing diseases in the control samples. To reduce variation between studies, investigators chose inclusion criteria for their control samples which provided the typical constitution of a patient sample with regard to diseases alternative to LNB. While, in differentiating from LNB, specificities for CSF CXCL13 were between 0.90 and 0.98 for viral meningitis, Bell’s palsy, inflammatory autoimmune polyneuritis, multiple sclerosis, tick-borne encephalitis, or non-inflammatory neurological diseases, modest results were observed for bacterial meningitis (0.68), neurosyphilis (0.57), and central nervous system lymphoma (0.13). The funnel plots with regression lines showing non-significant slope coefficients did not hint at publication bias.

Discussion: Bivariate random effects meta-analysis as presented by Reitsma et al. [3] provides an appropriate tool to consider the correlation between sensitivity and specificity in diagnostic accuracy studies. The actual performance of the analysis is decisively supported by the R function “mada”, maintained by Philipp Doebler. At the cut-off value 250 pg/ml, our meta-analysis showed that CSF CXCL13 has the potential to become a useful adjunct in the diagnosis of acute LNB. However, varying specificities of potential alternative diseases have to be taken into account.



Die Autoren geben an, dass kein Interessenkonflikt besteht.

Die Autoren geben an, dass kein Ethikvotum erforderlich ist.


References

1.
Rupprecht TA, Pfister HW, Angele B, Kastenbauer S, Wilske B, Koedel U. The chemokine CXCL13 (BLC): A putative diagnostic marker for neuroborreliosis. Neurology. 2005;65(3):448-50.
2.
Reitsma JB, Rutjes AWS, Whiting P, Vlassov VV, Leeflang MM G, Deeks JJ. Chapter 9: Assessing methodological quality. In: Deeks JJ, Bossuyt PM, Gatsonis C, editors. Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy Version 1.0.0. The Cochrane Collaboration; 2009. Available from: http://srdta.cochrane.org/ Externer Link
3.
Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol. 2005;58(10):982-90.
4.
Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol. 2005;58(9):882-93.