Downloaded from www.microbiologyresearch.org by IP: 54.162.190.106 On: Fri, 18 Mar 2016 12:01:17 Diagnostic accuracy and comparison of two assays for Borrelia-specific IgG and IgM antibodies: proposals for statistical evaluation methods, cut-off values and standardization Ram Benny Dessau Correspondence Ram Benny Dessau ramd@regionsjaelland.dk Received 8 March 2013 Accepted 20 September 2013 Department of Clinical Microbiology, Slagelse Hospital, Region Sjælland 18 Ingemannsvej, DK4200 Slagelse, Denmark Two assays (Liaison, Diasorin; IDEIA, Oxoid) for detection of Borrelia-specific antibodies were compared. A case–control design using patients with neuroborreliosis (n548), laboratory defined by a positive Borrelia-specific antibody index in the spinal fluid, was available and was intended to represent the serological response of disseminated early Lyme borreliosis in general. Serum samples were obtained from 216 Danish blood donors as controls. By comparing sensitivity and specificity using pre-specified cut-off values, significant differences were found. However, using receiver operating characteristic (ROC) curves to optimize and standardize test interpretation, it was shown that testing with both IDEIA IgG and IgM was comparable to testing with Liaison IgG alone by comparing the area under the curve of the diagnostically relevant 25 % partial ROC curve (P50.1). When using the Liaison OspC/VlsE IgM assay, the specificity was decreased without a gain in sensitivity. This study proposes standardizing of reporting by using a control population as the reference and choosing decision thresholds guided by the risk of false-positive results at 2 and 8 %. The sensitivities for IDEIA (IgG and IgM combined) were 85 and 95 % and for the Liaison (VlsE IgG) method were 67 and 96 %, respectively. Methods for test evaluation, test interpretation and statistical testing are presented and discussed. In conclusion, Liaison VlsE IgG alone and IDEIA IgG/IgM combined showed a high and comparable discriminatory ability to distinguish serum samples from patients with neuroborreliosis from blood donor controls. However, cut-off values should be adjusted for a proper comparison. INTRODUCTION Lyme borreliosis (LB) is caused by infection with the tick- borne bacterium Borrelia burgdorferi sensu lato. The most common clinical manifestation is a rash called erythema migrans. The disseminated forms of LB are rarer and include Lyme neuroborreliosis (NB), arthritis, multiple erythema migrans, lymphocytoma and carditis (Stanek et al., 2011, 2012). To support clinical diagnosis of dis- seminated LB, the standard laboratory tests use detection of the antibody response in blood or spinal fluid. Borrelia- specific antibody detection is an excellent tool, as sensi- tivities are high in disseminated disease. The analytical specificity will be high, as B. burgdorferi possesses diagnos- tically relevant antigens that are distinct from other related bacteria. The clinical specificity will also be high, except in certain subpopulations with continuous exposure to ticks. As with the use of serology for diagnostic purposes for any infectious disease, interpretation requires knowledge of the risk of background immunity, and the fact that sensitivity may be low in early clinical disease but should be detectable in all patients by 6–8 weeks after onset of symptoms. Laboratory support is not necessary or useful for diagnosis of localized erythema migrans, as some patients may not develop a detectable antibody response at all. Studies on diagnostic accuracy for Borrelia-specific IgG and IgM antibodies are most often presented in the published literature by counting positive and negative results, without analysis of the quantitative results. The cut-off recom- mendation provided by the developer of the assay has been accepted uncritically in many studies without analysis or discussion (Busson et al., 2012; Cerar et al., 2010; Marangoni et al., 2006, 2008; Petersen et al., 2008; Riesbeck & Hammas, 2007; Tjernberg et al., 2007). It has, however, been generally recommended that an explanation of the definition of and rationale for the units and cut-off values is given when reporting studies of diagnostic accuracy (Bossuyt et al., 2003a, b). The purpose of the Abbreviations: AUC, area under the curve; LB, Lyme borreliosis; NB, neuroborreliosis; NPV, negative predictive value; PPV, positive predictive value; ROC, receiver operating characteristic. Supplementary material is available with the online version of this paper. Journal of Medical Microbiology (2013), 62, 1835–1844 DOI 10.1099/jmm.0.061275-0 061275 G 2013 SGM Printed in Great Britain 1835