Abstract
Tick-borne encephalitis (TBE) is the most important arboviral disease in many parts of Europe and Asia. Both the diagnosis of TBE as well as the conduction of surveillance studies are based on the demonstration of specific antibodies. For reasons of simplicity, automatization, and quick availability of test results, enzyme-linked immunosorbent assays (ELISAs) are the method of choice for anti-TBE virus antibody detection. In this study, we evaluated four commercial IgG-ELISAs using 876 epidemiological plasma samples: the Enzygnost Anti-TBE/FSME Virus IgG assay (Siemens; assay 1), the Anti-FSME/TBE Virus ELISA (IgG) assay (Euroimmun; assay 2), the Anti-FSME/TBE Virus ELISA “Vienna” (IgG) assay (Euroimmun; assay 3), and the RIDASCREEN® FSME/TBE IgG EIA assay (R-Biopharm; assay 4). In total, discrepant results were observed for 37.2% of all samples. The evaluated assays significantly differed in qualitative data (p < 0.0001, Cochran–Mantel–Haenszel test) and showed Spearman's rank correlation coefficients ranging between 0.88 and 0.97 for quantitative data. The degree of disagreement between the different assays was exceptionally high for samples originating from blood donors with vaccination against TBE virus. For this sample group, the proportion of positive results was considerably higher for assay 3 (52.7%) and assay 4 (57%) than for assay 1 (7.5%) and assay 2 (6.4%), respectively, indicating that assays 1 and 2 are less suitable for the detection of vaccination antibodies than assays 3 and 4. Indirect immunofluorescence testing data available for a subset of samples (n = 238) mostly originating from nonflavivirus-vaccinated blood donors (n = 234) revealed problems in both sensitivity and specificity of the evaluated assays; whereas sensitivity issues were most prominent for the Euroimmun assay, specificity concerns were most pronounced for the Euroimmun Vienna and the RIDASCREEN assays.
Introduction
Tick-borne encephalitis virus (TBEV) is the causative agent of tick-borne encephalitis (TBE), one of the most important arboviral diseases in Europe and Asia. The TBEV species (family Flaviviridae, genus Flavivirus) includes three subtypes: a European, a Siberian, and a Far Eastern subtype (Ecker et al. 1999). The virus is typically transmitted by tick bites, wherefore the occurrence of TBEV correlates with the distribution of the vector ticks Ixodes ricinus and Ixodes persulcatus (Lindquist 2014).
TBE may manifest as a disease of variable severity, ranging from subclinical infections to severe courses with neurological involvement and potentially fatal outcome. While the infection is asymptomatic in 70–95% of cases, symptomatic disease may occur as meningeal, encephalitic, poliomyelitic, and myeloradiculitic forms (Lindquist 2014). TBE is typically biphasic when caused by European subtype viruses, whereas Far Eastern and Siberian viruses most often induce a monophasic disease. Chronic forms are observed in association with the Siberian subtype (Gritsun et al. 2003, Lindquist 2014).
Since clinical symptoms of TBE are often unspecific and similar to other central nervous system diseases, the diagnosis of TBEV infections has to be established in the laboratory (Holzmann 2003). The method of choice for the diagnosis of TBE is serological assays. Generally, specific IgM and usually IgG antibodies are present in the first serum sample taken when neurological symptoms are present. In addition to the diagnosis of clinical TBE cases, serological assays may be used in surveillance studies or for the detection of vaccination antibodies (Holzmann 2003, Lindquist 2014). For both purposes, most laboratories use enzyme-linked immunosorbent assays (ELISAs) due to their simplicity, automatization, and quick availability of test results.
Indirect immunofluorescence testing (IIFT) is used only in a few settings. The neutralization test is the most specific assay but is only performed in special laboratories (Litzba et al. 2014). Because of the high cross-reactivity among all flaviviruses, interference of either cross-neutralizing or antigen-binding but non-neutralizing antibodies has to be considered especially when using ELISA tests. False-positive results in these assays may be obtained at a higher frequency in areas where different flaviviruses co-circulate, for persons having traveled to areas endemic for other flaviviruses, or for persons recently vaccinated against other flaviviruses (Lindquist, 2014).
In this study, we compared four different commercial IgG-ELISAs using 876 plasma samples derived from a healthy blood donor population in Switzerland. Two hundred and thirty-eight of these samples were additionally analyzed for other flavivirus reactivity using IIFT.
Materials and Methods
Study population
From July 2014 to January 2015, plasma samples from 9328 healthy blood donors were collected in endemic as well as nonendemic regions of Switzerland. Participants gave written informed consent to the protocol approved by the cantonal ethics commission Bern, Nordwest Schweiz, St. Gallen, Thurgau, Ticino, and Vaud. From each participant, ethylenediaminetetraacetic acid (EDTA)-anticoagulated blood was collected at the time of regular blood donation. In addition, a questionnaire on personal details, vaccination status, and risk factors for TBE was filled in. Plasma samples were stored at −20°C; during the testing period, they were thawed and temporarily stored at 4°C.
Sample selection and summary of performed tests
All 9328 samples were screened for the presence of anti-TBEV-IgG antibodies using the Enzygnost Anti-TBE/FSME Virus IgG assay (Siemens, Marburg, Germany), in the following called “assay 1.” Based on the results of this screening assay, a total of 876 plasma samples were chosen to be included in the ELISA evaluation study, including 348 samples yielding positive, 74 samples yielding equivocal, and 454 samples yielding negative results in the screening test; quantitative test results of these samples covered the entire range of possible results ranging from very low to very high antibody concentrations (data not shown).
Based on questionnaire data on vaccination status, the samples were divided into five categories: (a) blood donors with vaccination against TBEV (n = 93), (b) blood donors without vaccination against TBEV but with vaccination against other flaviviruses (n = 103), (c) blood donors without any vaccinations against any flaviviruses (n = 601), (d) blood donors with vaccinations against TBEV plus other flaviviruses (n = 24), and (e) blood donors giving no information on vaccination status (n = 55).
All selected samples (n = 876) were further subjected to testing with three additional ELISAs, namely the Anti-FSME/TBE Virus ELISA (IgG) assay (Euroimmun, Lübeck, Germany), in the following called “assay 2”; the Anti-FSME/TBE Virus ELISA “Vienna” (IgG) assay (Euroimmun, Lübeck Germany), in the following called “assay 3”; and the RIDASCREEN® FSME/TBE IgG EIA assay (R-Biopharm, Darmstadt, Germany), in the following called “assay 4” (Table 1). Moreover, selected samples (n = 238) were analyzed using IIFT (see Indirect immunofluorescence testing section).
Summary of the Tested Commercial IgG-Enzyme-Linked Immunosorbent Assays
ELISA, enzyme-linked immunosorbent assay; RT, room temperature; RU, relative units; TBE, tick-borne encephalitis; TBEV, tick-borne encephalitis virus; U, units; VIEU, Vienna units.
Testing using commercial IgG-ELISAs
Assays 1, 2, and 3 were processed automatically according to the manufacturer's instructions, whereas assay 4 was processed manually. All analyses were performed according to the manufacturer's instructions (Table 1).
Indirect immunofluorescence testing
Two hundred and thirty-eight selected plasma samples, including 190 samples with discordant qualitative results for at least one ELISA and 48 samples with concordant positive results for all ELISAs, were tested for other flavivirus reactivity by IIFT (Euroimmun Flavivirus Profile 2; reaction sites for TBEV, West Nile virus [WNV], Japanese encephalitis virus [JEV], yellow fever virus [YFV], and dengue virus [DENV] 1–4).
Among the samples tested by IIFT, 4 originated from blood donors having received a vaccination against TBEV (2 with vaccination against TBEV, 1 with vaccination against TBEV and YFV, and 1 with vaccination against TBEV, YFV, and JEV), and 59 originated from blood donors having received a vaccination against one or several other flaviviruses (51 with vaccination against YFV, 1 with vaccination against JEV, and 7 with vaccinations against YFV and JEV). Plasma samples were tested in dilutions of 1:10, 1:100, 1:1000, and 1:10,000 to derive the most likely specific response from the highest titer.
Statistical analyses
The stats and coin packages of the R software were used for all statistical analyses. To assess whether there was an overall difference between qualitative (negative; equivocal; positive) results of the different assays, the asymptotic generalized Cochran–Mantel–Haenszel (CMH) test, stratified for samples, was used; a p-value of <0.05 was regarded as significant. Qualitative test results were then pair-wise compared in contingency tables, and an asymptotic linear-by-linear association test was applied to detect whether the assay test results significantly differ from each other. To account for multiple testing, the Bonferroni correction was applied, and a p-value of <0.0083 was regarded as significant.
Following the overall analyses, samples were separated into sample groups (see Sample selection and summary of performed tests section), and all tests (CMH test, linear-by-linear association test) were repeated for the individual groups (a), (b), and (c). While the Bonferroni correction was applied again to account for the multiple pair-wise comparisons, no further correction was used for the analysis of the separate groups, that is, the significance levels used were the same as in the analysis of all samples together. Sample groups (d) and (e) were not included in these analyses due to their small sample. Finally, correlation of quantitative results (U/mL) was assessed using Spearman's rank correlation coefficient. Approximate 95% confidence intervals for Spearman's rank correlation coefficient were obtained via confidence intervals for the Pearson product-moment correlation applied to the ranks; no further correction was used for the level of confidence.
Results
Qualitative test results
Overall, qualitative test results (negative; equivocal; positive) significantly differed between the different assays (p < 0.0001, CMH test). In total, discrepant results were observed for 326/876 samples (37.2%). The differences were significant in all pair-wise comparisons (p < 0.0001, linear-by-linear association test). Also, the differences remained significant when restricting the analyses to distinct sample groups: p-values of the CMH test were <0.0001 for the individual sample groups (a), (b), and (c), and pair-wise comparison was nonsignificant in only three cases (Tables 2 and 3). The degree of disagreement was particularly high for sample group (a), with 61/93 samples (65.6%) yielding discrepant results for the different assays. For sample groups (b) and (c), 40/103 (37.9%) and 190/601 (31.6%), respectively, of the samples yielded discrepant results.
Pair-Wise Comparison of Anti-Tick-Borne Encephalitis Virus IgG Qualitative Test Results for All Samples (n = 876)
The number of samples with the respective results is given; total number of samples tested = 876.
Assay 1, Enzygnost Anti-TBE/FSME Virus IgG assay (Siemens); assay 2, Anti-FSME/TBE Virus ELISA (IgG) assay (Euroimmun); assay 3, Anti-FSME/TBE Virus ELISA “Vienna” (IgG) assay (Euroimmun); assay 4, RIDASCREEN® FSME/TBE IgG EIA assay (R-Biopharm); equ, equivocal; neg, negative; pos, positive.
Pair-Wise Degree of Agreement Between the Evaluated Anti-Tick-Borne Encephalitis Virus IgG Enzyme-Linked Immunosorbent Assays, Overall and for Different Sample Categories
Assay 1, Enzygnost Anti-TBE/FSME Virus IgG assay (Siemens); assay 2, Anti-FSME/TBE Virus ELISA (IgG) assay (Euroimmun); assay 3, Anti-FSME/TBE Virus ELISA “Vienna” (IgG) assay (Euroimmun); assay 4, RIDASCREEN® FSME/TBE IgG EIA assay (R-Biopharm).
Independent of the sample groups, assay 4 overall yielded the highest proportion of positive results (53.3% of all samples tested positive), followed by assay 3 (47.3%), assay 1 (39.7%), and assay 2 (23.7%). When focusing on sample group (a), the proportion of positive results was considerably lower for assay 1 (6.0%) and assay 2 (4.3%) than for assay 3 (30.8%) and assay 4 (40.2%) (Table 2; Supplementary Table S1).
IIFT results
From all samples tested using IIFT (n = 238), 197 were reactive with at least one flavivirus antigen present on the IIFT slide, whereas 41 tested negative for all antigens. Detailed IIFT results are shown in Table 4.
Indirect Immunofluorescence Testing Results for n = 238 Samples and Respective Proportions of Positive, Equivocal, or Negative Enzyme-Linked Immunosorbent Assay Results for the Evaluated Anti-Tick-Borne Encephalitis Virus IgG Enzyme-Linked Immunosorbent Assays
Assay 1, Enzygnost Anti-TBE/FSME Virus IgG assay (Siemens); assay 2, Anti-FSME/TBE Virus ELISA (IgG) assay (Euroimmun); assay 3, Anti-FSME/TBE Virus ELISA “Vienna” (IgG) assay (Euroimmun); assay 4, RIDASCREEN® FSME/TBE IgG EIA assay (R-Biopharm); DENV, dengue virus; IIFT, indirect immunofluorescence testing; YFV, yellow fever virus.
Quantitative test results
The Spearman rank correlation coefficients (95% confidence intervals) estimated for the correlation of quantitative test results (U/mL) are listed in Table 5. Whereas correlation coefficients were reasonably high for analyses including all samples (0.88–0.97), correlation was partially low (0.37–0.97) when focusing on sample group (a).
Spearman Rank Correlation Coefficients and Approximate Confidence Intervals for Quantitative Test Results (U/mL) Obtained with the Four Evaluated Anti-Tick-Borne Encephalitis Virus IgG Enzyme-Linked Immunosorbent Assays
Approximate 95% confidence intervals were obtained via confidence intervals for the Pearson product-moment correlation applied to the ranks.
Assay 1, Enzygnost Anti-TBE/FSME Virus IgG assay (Siemens); assay 2, Anti-FSME/TBE Virus ELISA (IgG) assay (Euroimmun); assay 3, Anti-FSME/TBE Virus ELISA “Vienna” (IgG) assay (Euroimmun); assay 4, RIDASCREEN® FSME/TBE IgG EIA assay (R-Biopharm).
Discussion
The method of choice for both the diagnosis of infection as well as for epidemiological studies on TBE is serological testing. Although ELISAs produce false-positive results with antibodies directed against other flaviviruses, they remain the method of choice for serological testing due to their simplicity and quick availability of test results (Litzba et al. 2014). In this work, we compared 4 commercially available anti-TBEV IgG ELISAs using a total of 876 epidemiological plasma samples. For a subset of 238 samples, IIFT was performed to assess the influence of cross-reactive antibodies directed against other flaviviruses.
Overall, the qualitative test results (negative; equivocal; positive) significantly differed between the different assays (p < 0.0001, CMH test over all samples and p < 0.0001 for all pair-wise comparisons using linear-by-linear association test). The differences remained highly significant when restricting analyses to distinct sample groups (a), (b), and (c) (p < 0.0001, CMH test for all individual groups). Also, the majority of pair-wise assay comparisons (linear-by-linear association test) showed significant differences (Tables 2 and 3). In total, 37.2% of all samples yielded discrepant results. Qualitative test results are thus influenced by the test system; both the results of seroepidemiological surveys and clinical diagnostic testing are significantly affected by the choice of the diagnostic assay.
Interestingly, the degree of agreement between the different tests can be considerably improved by counting equivocal results of the distinct tests as positive or negative: in pair-wise comparison of assays 1 and 2, when counting equivocal results of assay 1 as negative and equivocal results of assay 2 as positive, the proportion of consistent results rises from 74.8% (655/876) to 88.1% (772/876). Likewise, in pair-wise comparison of assays 3 and 4, when counting equivocal results of assay 3 as positive and equivocal results of assay 4 as negative, the proportion of consistent results rises from 84.6% (741/876) to 91.9% (805/876) (Table 2). Thus, adjusting the cutoff values of the different assays (i.e., raise the lower cutoff value of assays 1 and 4 and lower the upper cutoff value of assays 2 and 3) would improve the agreement of test results of the different assays.
With a proportion of 65.6% of samples yielding discrepant results, the degree of disagreement between the four assays was particularly high for sample group (a), that is, blood donors with vaccination against TBEV. For this sample group, we found that assays 1 and 2 yielded a much higher proportion of negative results (92.5% and 93.6%, respectively) than assay 3 (47.3%) and assay 4 (43.0%). In contrast to individuals naturally infected with TBEV, seroprotection following vaccination is known to remarkably decline over time (Baldovin et al. 2012). Unfortunately, our questionnaire data on the vaccination status do not include information on the interval between the last vaccination and the collection of study samples. However, it seems likely that assays 1 and 2 are less suitable for the detection of vaccination antibodies than assays 3 and 4. Also, it must be noted that Euroimmun recommends using assay 3 but not assay 2 for the detection of vaccination antibodies.
A study by Jilkova et al. (2009) found that results significantly differed between two evaluated ELISAs as an effect of homologous or heterogeneous TBEV strain in the vaccine and ELISA. In Switzerland, the FSME-IMMUN vaccine using TBEV strain Neudörfl and the Encepur vaccine based on TBEV strain K23 are in use. Since our questionnaire data do not include complete information on which type of vaccine blood donors received, we cannot evaluate this effect for our sample population.
For this study, we used EDTA plasma samples. Since EDTA is toxic to cell cultures required for serum neutralization testing, we were not able to perform serum neutralization testing with our samples. We therefore decided to perform IIFT, which may provide at least some more precise information with respect to antibody specificity compared with ELISA testing. We tested a total of 238 samples using the Euroimmun Flavivirus Profile 2 assay. In this test, the antigen yielding highest antibody titers is supposed to represent the specificity of the antibodies (Litzba et al. 2014).
Forty-one of 238 samples yielded a negative IIFT test result for antibodies against all viruses included in the panel. The IIFT manufacturer indicates a sensitivity of 93–100%, depending on the antigen (TBEV, WNV, JEV, YFV, DENV 1–4) (package insert Euroimmun Flavivirus Profile 2). For 116/238 (48.7%) of the samples tested by IIFT, highest titers were identified against TBEV. For these, positive ELISA results were obtained for only 23.3% using assay 2, indicating major sensitivity problems for this assay. For samples yielding highest titers against YFV (36/238), DENV (18/238), or JEV (1/238) in IIFT testing, variable degrees of cross-reactivity for the evaluated ELISAs were observed (Table 4).
Interestingly, samples originating from blood donors having received vaccinations against TBEV (n = 4) or at least one flaviviruses other than TBEV (n = 59), only 15 gave highest titers against one or both antigens stated in the vaccination questionnaire (detailed results not shown), indicating that the IIFT may have some problems in detecting vaccination antibodies. These findings correspond with the information given by the manufacturer, stating that IIFT is not as sensitive as ELISA and is therefore not suitable as an assay for detecting vaccination antibodies.
Although cross-reactive antibodies directed against other flaviviruses will generate false-positive results in all ELISAs evaluated in this study, the problem seems more pronounced for assays 3 and 4 than for assays 1 and 2. Spearman's rank correlation coefficients were high for overall analyses (0.88–0.97). As it would be expected, rank correlation coefficients were particularly high for assays 2 and 3 originating from the same manufacturer. However, quantitative test results showed a pronounced variation between the different assays when focusing on sample group (a), that is, blood donors with vaccination against TBEV (Table 5). Consequently, direct comparison of quantitative data from different assays is not appropriate.
Titers of IgG antibodies determined by ELISA are described to serve as a marker for predicting the presence of neutralizing antibodies against TBEV, given that cross-reactive antibodies directed against other flaviviruses are absent. For assays using the so-called Vienna standard (e.g., anti-FSME/TBE virus ELISA Vienna IgG; Euroimmun), cutoffs valid for vaccinated persons shortly after basic immunization have been described (Holzmann et al. 1996, Kunz 2003). Based on our results for TBEV-vaccinated blood donors, detection of antibodies following vaccination against TBEV strongly depends on the test assay used. Determination of antibody titers following vaccination, including but not limited to the definition of cutoffs defining protective immunity, is challenging and must be implemented with care.
Conclusion
In our study comparing four commercially available anti-TBEV IgG ELISAs, we have observed prominent differences for both qualitative and quantitative data, with 37.2% of all samples yielding discrepant results, indicating the need of standardization of the different ELISAs.
When focusing on sample group (a), blood donors with vaccination against TBEV, assays 1 and 2 gave a very high proportion of negative results (92.5% and 93.6%, respectively) compared with assays 3 and 4 (47.3% and 43.0%, respectively), indicating that assays 3 and 4 are more suitable for the detection of vaccination antibodies than assays 1 and 2.
IIFT data available for a subset of samples mostly originating from nonflavivirus-vaccinated blood donors revealed problems in both sensitivity and specificity of the evaluated assays; whereas sensitivity issues were most prominent for assay 2, specificity concerns were most prominent for assays 3 and 4.
Footnotes
Acknowledgments
We thank Euroimmun, R-Biopharm, and Siemens for their interest in our evaluation work and for offering some of the assays used in this evaluation study. Special thanks go to Michael Vock, Institute of Mathematical Statistics and Actuarial Science, University of Berne, for his very competent support in statistical analyses. Finally, we thank Olivier Engler, Marc Strasser, and Roland Züst for their support and helpful discussions.
Author Disclosure Statement
No competing financial interests exist.
Supplementary Material
Supplementary Table S1
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
