- Email: [email protected]
Performance of 14 rubella IgG immunoassays on samples with low positive or negative haemagglutination inhibition results
Journal of Clinical Virology 74 (2016) 13–18
Contents lists available at ScienceDirect
Journal of Clinical Virology journal homepage: www.elsevier.com/locate/jcv
Performance of 14 rubella IgG immunoassays on samples with low positive or negative haemagglutination inhibition results Daniela Huzly ∗ , Ingeborg Hanselmann, Dieter Neumann-Haefelin, Marcus Panning Institute of Virology, University Medical Center Freiburg, Germany
a r t i c l e
i n f o
Article history: Received 13 August 2015 Received in revised form 14 October 2015 Accepted 16 November 2015 Keywords: Rubella CRS Assay comparison Seroepidemiologic studies
a b s t r a c t Background: Rubella IgG testing is routinely done in prenatal care and seroepidemiological studies. Recently concern was raised that seropositivity rates were decreasing questioning vaccination policies. Manufacturers of rubella IgG assays and authors of seroepidemiological studies use different cut-offs for the definition of seropositivity. As rubella virus circulation is reduced since many years, seronegativity rates might be overestimated using an inappropriate cut-off. Objectives: Using different cut-off definitions we compared fourteen current rubella IgG immunoassays for sensitivity and qualitative result concordance in samples with low positive or negative haemagglutination inhibition (HI) titre. Study design: 150 clinical samples from patients and health care workers were included in the study. All samples were measured in 14 different rubella IgG immunoassays. Seropositivity was defined using recombinant rubella IgG immunoblot as reference standard. Results: The concordance of qualitative results using the manufacturers cut-off definitions was 56.4% if grey-zone results were analysed separately and 69.8% if grey-zone results were defined as positive. Using universal cut-offs of 10 IU/ml or 15 IU/ml the concordance was 70% and 61.4% respectively. Using the different cut-off definitions up to 71 out of the 124 immunoblot-positive samples tested negative in the immunoassays. The mean coefficient of variation (CV) of quantitative results in positive samples was 51% (range 19–113%). Conclusions: Determination of rubella immunity by measurement of rubella-IgG in a population with high vaccination coverage with current assays leads to a high number of false negative results. The value of routine rubella antibody testing in countries with high vaccination coverage should be discussed. © 2015 Elsevier B.V. All rights reserved.
1. Background Rubella is an acute infection usually occurring during childhood with a mild clinical course. Primary rubella virus infection during early pregnancy, however, carries a high risk for fetal death and the development of congenital rubella syndrome (CRS) with malformations of eyes, ears and heart. Vaccination programs have reduced the incidence of CRS in most industrialized countries dramatically but in some countries, cases still occur. As systematic data on vaccination history are rarely available, seroepidemiological studies are conducted to evaluate the efficacy of vaccination policy. According to recent studies, seropositivity rates seem to decrease in countries with high vaccination coverage [1–3]. Based on this finding, concern
∗ Corresponding author at: Institute of Virology, University Medical Center, Hermann-Herder−Str. 11, 79104 Freiburg, Germany. Fax: +49 761 2036603. E-mail address: [email protected] (D. Huzly). http://dx.doi.org/10.1016/j.jcv.2015.11.022 1386-6532/© 2015 Elsevier B.V. All rights reserved.
was raised that an increasing number of susceptible women in child bearing age might question the effectiveness of present vaccination strategies. Numerous commercial assays are available to detect and quantify anti-rubella IgG. These assays are calibrated with an international reference preparation and results are reported in international units per ml, which in theory should lead to a standardization of test results and interpretation. However, serological assays for rubella IgG use different cut-offs for the definition of seropositivity. In addition, authors of seroepidemiological studies frequently use a universal cut-off of 10 or 15 IU/ml based on published recommendations and not the cut-off recommended by the manufacturer. With reduced virus circulation, antibody levels decline over time and therefore seronegativity rates might be overestimated due to a wrong cut-off definition.
14
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
2. Objectives The aim of this study was to compare the sensitivity and concordance of qualitative results of current rubella-IgG immunoassays defining seropositivity by IgG recombinant immunoblot.
3. Study design Study samples: A total of 150 clinical samples were included in the study. The samples obtained from health care workers, pregnant women, and children had been submitted to the Institute of Virology for routine rubella antibody testing and had been stored at −20 ◦ C thereafter. A total of 121/150 samples were collected prospectively based on a haemagglutination inhibition (HI) titre below 1:32. Another 29/150 samples with a HI titre of 1:32 or 1:64 were chosen retrospectively. To avoid multiple freeze–thaw cycles, samples were stored at 4 ◦ C during the main study period. The first International Standard for anti-rubella immunoglobulin RUBI-1-94 containing 1600 IU per vial (NIBSC) was measured in different dilutions to check the calibration of each immunoassay. The standard preparation was diluted in a human serum pool which tested negative for rubella-IgG using the reference assay. Recombinant immunoblot (reference assay): All samples were measured in a commercially available recombinant immunoblot (recomBlot Rubella IgG, Mikrogen, Martinsried, Germany) on a Tecan Profiblot. All kits were of the same lot, and a negative and an E2 weakly positive control as well as the 12.5 IU/ml NIBSC standard dilution were run with each kit. The E2 band intensity of the E2 weakly positive control was used as reference for the definition of band positivity. All sera showing at least E1-reactivity on the immunoblot were defined as rubella IgG positive. Haemagglutination-inhibition test (HI): All sera had previously been tested by a commercially available CE labelled rubella HI assay (Labor Dr. Merck & Kollegen, Ochsenhausen, Germany) as recommended by the manufacturer. A titre of 1:8 is considered as positive but a confirmation of results below 1:32 is recommended. The distribution of HI results is shown in Supplementary file 1. Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.jcv.2015.11.022. Enzyme immunoassays: Seven different enzyme immunoassays were used: Enzygnost Anti-Rubella-Virus/IgG (Siemens Healthcare, Fernwald, Germany), Serion ELISA classic Rubella Virus IgG quant. (Institut Virion\Serion GmbH, Würzburg, Germany) Elimmun–Rubella G (Labor Dr. Merck & Kollegen, Ochsenhausen, Germany), Rubella-IgG-ELISA PKS medac (medac Diagnostica, Hamburg, Germany), Vircell Rubella ELISA IgG (Vircell, Granada, Spain), Euroimmun Anti-Rubella virus ELISA (IgG) (Euroimmun, Lübeck, Germany), Dia Sorin ETI-RUBEK-G PLUS (DiaSorin, Dietenbach, Germany). All sera were diluted manually and automatically processed in a Siemens Elisa Processor BEP III. As the Vircell Rubella ELISA IgG generates semiquantitative results (Signal/Cutoff*10) we used the international standard preparation to obtain the results in IU/ml, as recommended by the manufacturer (personal communication). (see Supplementary file 2 for technical specifications). Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.jcv.2015.11.022. Automated immunoassays: Seven automated immunoassays were used: Architect Rubella IgG on the Abbott Architect system (Abbott, Chicago, USA), Abbott Axsym Rubella IgG on the Abbott Axsym system (Abbott, Chicago USA), ADVIA Centaur Rubella G on the Siemens ADVIA Centaur system (Siemens Healthcare, Fernwald, Germany), DiaSorin Liaison Rubella IgG on the LIAISON system (DiaSorin, Torino, Italy), Cobas Rubella IgG on the Roche ELECSYS system (Roche, Mannheim, Germany), Vitros Rubella IgG on the Vitros ECi System (Ortho Clinical Diagnostics, Raritan, USA), and VIDAS RUB
IgG II on the mini VIDAS System (Biomerieux, Lyon, France) (see Supplementary file 3 for technical specifications). Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.jcv.2015.11.022. The tests by Abbott Architect and by Roche Cobas were performed at the end of the study. The leftover volume of the samples was small, thus, the number of sera measured in these systems was limited to 130 in the Abbott Architect System and 96 sera in the Roche system. Statistical analysis: The recombinant immunoblot was used as reference standard for the definition of seropositivity. For assessment of immunoassay results we first used the cut-offs given by the manufacturers and then reassessed the results using 10 IU/ml and 15 IU/ml as universal cut-off for all assays. The universal cut-offs were also in accordance with supplementary threshold recommendations of several manufacturers. Statistical analyses were done with SPSS Version 20 and MEDCALC. The serum with questionable reactivity in the immunoblot was excluded from specificity, sensitivity as well as ROC analyses. Negative predictive values were calculated using published vaccination data of Germany and assuming a 95% efficacy of vaccination. For all positive samples the coefficient of variation of quantitative values between the different assays was determined. For comparison of the different assays with HI results, mean IU concentrations and range as well as 95% confidence interval at different HI titres were calculated. 4. Results 4.1. Baseline demographic data Median age of patients was 29 years (range 2–67 years), and 122/150 (81%) were female. Indication for rubella serology were in-vitro fertilisation in 79/150, routine health care in 27/150, serostatus before immunosuppressive therapy in 26/150, suspicion of acute rubella virus infection in 13, and pregnancy in 5/150 patients. 4.2. Assessment of seropositivity using recombinant immunoblot A total of 124/150 (83%) samples were positive and 25 samples were negative for rubella-IgG, based on immunoblot results. One sample could not be definitely defined as there was only a weak E1band with lower intensity than the weakly positive control band. In all immunoblot-positive samples, at least two bands (E1 and E1–E2) were reactive. 4.3. Assessment of test results using manufacturers recommendations Using the cut-offs given by the manufacturers, five of the 25 immunoblot-negative samples were reactive and two were equivocal by Abbott Axsym. Four of these samples also showed positive or borderline reactivity in the DiaSorin ELISA, but were measured below 2 IU in all other assay systems and negative by HI. Three different samples showed borderline results in the LIAISON, Centaur and Architect. 16 samples were concordantly negative. A summary of the results are shown in Table 1. Between one and 46 out of the 124 immunoblot-positive samples are negative in the immunoassays (Table 2). Between 4 and 21 samples had grey zone results in the different assays, of which most tested positive by immunoblot. A summary of the analysis of greyzone results are shown in Table 3. The concordance of qualitative results using the manufacturers cut-off definitions was 56.4% if grey-zone results were analysed separately and 69.8% if grey-zone results were defined as positive.
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
15
4.4. Assessment of test results using universal cut-offs
4.6. Comparison of quantitative results of immunoassays
Using 10 IU/ml as a universal cut-off, 7/25 immunoblot-negative samples were reactive in one (n = 3), two (n = 3) or three (n = 1) immunoassays, whereas 18/25 samples were concordantly negative (Table 1). Between 5 and 41 out of the 124 immunoblot-positive samples were negative in the immunoassays (Table 2). The overall concordance of results was 70%. Next, using 15 IU/ml as universal cut-off, 4/25 immunoblot-negative samples were reactive in one (n = 3) or two immunoassays (Table 1) and between 12 and 71 out of the 124 immunoblot-positive samples are negative in the immunoassays (Table 2). The concordance of results using 15 IU/ml as cut-off was 61.4%.
The mean coefficient of variation (CV) of positive samples was 51% (range 19%–113%). There was no significant difference between CVs of sera with low values (mean < 35 IU/ml) and sera with high values.
4.5. Correlation of HI results with immunoblot and with IU/ml measured by different immunoassays
4.7. Measurement of standard serum using immunoassays Next, twofold dilutions of the NIBSC standard were analysed using the immunoassays. The 100 IU/ml standard was measured too low or too high in some systems (Fig. 1). The assay from Vircell was not able to detect the 12.5 and 25 IU/ml standard dilution (Fig. 2).
4.8. Specificity, sensitivity and ROC analysis of immunoassays In comparison to the immunoblot results, four out of the 25 immunoblot-negative samples had a titre of 1:8 by HI and 11/124 immunoblot-positive samples tested negative by HI yielding an overall concordance of 88%. There was no clear correlation between HI results and results in IU/ml (data not shown). Immunoassay results in samples with a titre of 1:8 by HI are shown in Fig. 1.
Finally, specificity and sensitivity analyses using the different cut-off definitions were determined in our cohort with low HI result. Specificity ranged from 80 to 100% depending on the cutoff used (Table 1). On the other hand, sensitivity ranged from 67 to 100% (Table 2).
Table 1 Specificity of the different assay systems in 25 immunoblot negative samples using 10 IU/ml or 15 IU/ml in comparison to the manufacturers cut-off. Assay system
Abbott Architecta Abbott Axsym Siemens Centaur Siemens enzygnost EIA Biomérieux Vidas DiaSorin EIA DiaSorin LIAISON Euroimmun EIA Medac EIA Merck EIA Ortho ECI Roche COBASb Serion EIA Vircell EIA a b
Number of false positive results (% specificity) Manufacturers’ cut-off, equivocals excluded
Manufacturers’ cut-off, equivocals = positive
Cut-off at 10 IU/ml
Cut-off at 15 IU/ml
0 (100) 5 (80) 0 (100) 0 (100) 0 (100) 2 (92) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100)
1 (95) 7 (72) 1 (96) 0 (100) 0 (100) 2 (92) 1 (96) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100)
0 (100) 5 (80) 0 (100) 0 (100) 0 (100) 2 (92) 0 (100) 0 (100) 1 (96) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100)
0 (100) 3 (88) 0 (100) 0 (100) 0 (100) 2 (92) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100)
21 negative samples. 16 negative samples.
Table 2 Number and percentage of false negative samples in the different immunoassays using manufacturers cut-offs, or a universal cut-off at 10 IU/ml or 15 IU/ml, respectively. b 80 positive samples. c Manufacturer’s cut-off in Arbitrary Units (AU). Assay system
Abbott Axsym Abbott Architecta Bayer Centaur Siemens enzygnost EIA Biomérieux Vidas DiaSorin EIA DiaSorin LIAISON Medac EIA Merck EIA Ortho ECI Roche Cobasb Serion EIA Vircell EIA a
109 positive samples. 80 positive samples. c Manufacturer’s cut-off in Arbitrary Units (AU). b
Number of false negative results (% sensitivity) Manufacturers’ cut-off
Cut-off at 10 IU/ml
Cut-off at 15 IU/ml
1 (99.2) 5 (95.4) 2 (98.4) 0 (100) 10 (91.9) 12 (90.3) 5 (96.0) 20 (83.9) 29 (76.6) 6 (95.2) 10 (87.5) 7 (94.4) 46 (62.9)c
8 (93.5) 20 (81.7) 12 (90.3) 20 (83.9) 10 (91.9) 12 (90.3) 5 (96.0) 13 (89.5) 20 (83.9) 6 (95.2) 11 (86.3) 7 (94.4) 41 (67.0)
23 (81.5) 34 (68.8) 19 (84.7) 36 (71.0) 21 (83.1) 19 (84.7) 13 (89.5) 33 (73.4) 31 (75.0) 12 (90.3) 17 (78.8) 19 (84.7) 71 (42.7)
16
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
Fig. 1. Boxplot of measured values by immunoassays in samples with a titre of 1:8 by HI (minimum, 1. quartile, median, 3. quartile, maximum, outliers with sample ID). Automated assays on the left.
Table 3 Immunoblot results of samples with equivocal results using manufacturers cut-offs. Assay system
Number of equivocal results
Positive by immunoblot (n)
Abbott Architect Abbott Axsym Biomérieux Vidas DiaSorin EIA DiaSorin Liaison Euroimmum EIA Medac EIA Merck EIA Ortho ECI Roche (no greyzone) Serion EIA Siemens Centaur Siemens Enzygnost EIA Vircell EIA
17 9 11 7 6 13 21 4 6 0 24 11 11 8
16 7 11 7 5 13 21 4 6 0 24 10 11 8
ROC analysis yielded cut-offs below 10 IU/ml and below the manufacturer’s cut-off in most assays (Table 4).
Fig. 2. Measurement of the first international standard for anti-rubella immunoglobulin RUBI-1-94, NIBSC.
Table 4 ROC analysis PPV = positive predictive value, NPV = negative predictive value. Assay
Highest Youden Index at IU/ml
Sensitivity (%)
Specificity (%)
PPV (%)
Abbott Architect Abbott Axsym Biomérieux Vidas DIaSorin EIA DiaSorin LAISON Euroimmun EIA Medac EIA Merck EIA Ortho ECI Roche cobas Serion EIA Siemens Centaur Siemens Enzygnost Vircell
5.5 7.1 4 6.4 5 4.1 5.2 6.5 1 2 1 5.3 4 3
96.3 97.6 97.6 91.1 100 97.6 99.2 93.5 95 98.7 98.4 98.4 100 80.6
100 80 96 92 100 100 96 96 100 100 100 100 100 100
100 96 99.2 95.5 100 100 99.2 99.1 100 100 100 100 100 100
NPV (%) 84.5 92.6 88.9 68 100 89.3 96 75 80.6 94.1 92.6 92.6 100 51
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
5. Discussion The presence of specific rubella antibodies correlates with immunity, but a specific type or level of antibodies correlating with absolute protection has not been identified [4]. To avoid unnecessary vaccination of individuals and irritation during pregnancy, an appropriate definition of seropositivity is essential. According to recent seroepidemiological studies from countries with high vaccination coverage, the number of women not protected against rubella is growing [1,2,5,6]. However, despite an apparently increasing number of seronegative individuals, rubella cases did not increase and practically no CRS cases occur in countries with high vaccination coverage. Studies have shown, that in the absence of widespread virus circulation, individual antibody concentrations will get lower over time and the increasing number of seronegative individuals could just reflect a low negative predictive value of rubella IgG assays [2,7,8]. Similar effects have been shown for measles and varicella-zoster virus, where routine antibody assays have a low sensitivity in highly vaccinated populations [9,10]. Technically, to determine the serostatus for rubella, different test systems are on the market and an international standard preparation is available, which should render assay systems comparable. This has, however, not been successfull, as shown by our study. In our cohort of individuals with low antibody titres by HI, up to 37% of the immunoblot-positive samples (our reference assay) were negative in various immunoassays. Importantly, the concordance of commercial immunoassay results was only 56% using manufacturers cut-off definitions including the greyzone results. Concordance of results with manufacturers cut-offs was better when grey zone results were interpreted as seropositive, indicating that cut-off definition could be the main problem. This is also reflected by the fact that almost all samples with equivocal results in one or more immunoassays were positive in our reference assay. Using universal cut-offs of 10 IU/ml as recommended by Skendzel [11] or 15 IU/ml, as recommended by several authors [12–15], the concordance of immunoassay results did not improve. It is important to note, that the first IgG anti-rubella standard prepared in 1970 was based on human normal immunoglobulin. The International Unit (IU) for anti-rubella serum was then defined on the basis of neutralization test results from a collaborative study. This standard was termed Second International Reference Preparation of Anti-Rubella Serum. All subsequent standards, including RUBI-1-94 introduced in 1995, were calibrated against this reference preparation. The RUBI-1-94 standard consists of a pool of human normal immunoglobulin and was estimated to contain 1600 IU per ampoule. It has been used to calibrate recent commercial anti-rubella IgG assays. However, the dilutions of the reference standard were not measured accurately by the assays in our study. The use of a universal cut-off implies comparability of quantitative assay results. However, numerous studies comparing quantitative results of different assay systems come to the conclusion, that measured values are highly discrepant and comparability of results is not accomplished by the use of standard preparations [16–23]. We have shown very high discrepancies even in our small cohort of individuals with low HI titres. We were also not able to show a correspondence of HI titres with certain mean IU values [11,24]. Although, these high discrepancies between assay results have been described from the beginning, most assay manufacturers used the proposed universal cut-offs of 10 IU/ml or 15 IU/ml for their assays and relied on these fixed values instead of defining the limit of detection for the newer assay techniques. We used the recombinant immunoblot as reference standard to define true seropositivity in our set of low positive samples. Immunoblot testing was chosen as reference assay in our study as in contrast to neutralization assays it is highly standardized [4,25]. E1 antibodies on the immunoblot can be shown in all individuals after rubella vac-
17
cination or wildtype virus infection [26], indicating high sensitivity. It has been shown that E1 reactivity on westernblot correlates with neutralising antibodies [27]. Specificity of E1 antibody positivity in the immunoblot has been shown by the fact that individuals with E1 antibodies showed booster responses upon challenge [28]. Defining immunoblot-positive samples as true seropositive, most assays, like in previous studies, showed very high specificity. The comparably low specificity of the Abbott Axsym had been shown before [29]. False positive HI results at a dilution of 1:8 are probably due to the presence of non-specific inhibitors of haemagglutination, as has been shown in previous studies [30,31]. We performed ROC analyses for the different assays and in all cases the calculated cut-offs were below 10 IU/ml, without losing specificity in the majority of assays. One seroepidemiological study used the lower limit of quantification of the Enzygnost assay (4 IU/ml), which was also the cut-off result of our ROC analysis, and showed a persistently high rate of seropositivity over the years in the US [32]. This reflects rubella epidemiology much better. Recently, the PanAmerican Health Organization of the World Health Organization declared that the Americas region had eliminated endemic transmission of rubella and CRS [33]. In light of our and previously published results, the value of reference standards for the quantification of antibodies should be critically addressed. We have shown that even anti-HBs results are highly discrepant although most assays use the same antigen [34]. Rubella IgG assays use different antigens and different assay constructions, so differences are to be expected. Reporting results in IU/ml with the high variation between different assays can lead to wrong interpretations by clinicians as results from different labs may imitate a significant titre rise. A recent publication even postulates high rubella exposition rates in pregnant women, based on increasing titres from the beginning to the end of pregnancy. However, prenatal rubella testing had been done by different laboratories using various assay systems [35]. No CRS case had occurred in that study. The value of routine rubella antibody testing in countries with high vaccination coverage has to be discussed. Several countries now recommend testing only in individuals without documented vaccination history [36,37].
6. Conclusion Definition of rubella immunity by measurement of rubella-IgG in a population with high vaccination coverage with current assays does lead to a relatively high number of false negative results. Assay manufacturers should reevaluate the cut-off of their assays. To avoid the diagnosis of rubella reinfection on the basis of significant titre rise between measurements by different laboratories, quantification of IgG assays using International Units should be abandoned. After 40 years trying with different standards and different assay concepts we should accept that standardization of IgG quantitation is not reliably feasible. Assay developers should concentrate on true test accuracy parameters when developing new assays.
Conflict of interest The authors have no conflict of interest to declare.
Sources of funding None.
18
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
Ethical approval Not required. Acknowledgement
[18]
[19]
All test kits were contributed by the manufacturers free of charge.
[20]
References
[21]
[1] R. Barlinn, K. Vainio, H.H. Samdal, S.A. Nordbo, H. Nokleby, S.G. Dudman, Susceptibility to cytomegalovirus, parvovirus B19 and age-dependent differences in levels of rubella antibodies among pregnant women, J. Med. Virol. 86 (2014) 820–826. [2] G. Smits, L. Mollema, S. Hahne, H. de Melker, I. Tcherniaeva, F. van der Klis, G. Berbers, Seroprevalence of rubella antibodies in The Netherlands after 32 years of high vaccination coverage, Vaccine 32 (2014) 1890–1895. [3] Coyle P. Viral Congenital and Neonate Infections ESCV Book of Abstracts 2014; Keynote lecture at the ESCV congress in Prague 2014. [4] W.H.O. The Immunological Basis for Immunization Series. Immunization, Vaccines and Biologicals 2008, Module 11: Rubella: 1–25. [5] S.A. Alsibiani, Rubella immunity among pregnant women in Jeddah, Western Region of Saudi Arabia, Obstet. Gynecol. Int. 2014 (2014), 659838. [6] S. Mortlock, S. Farthing, Rubella antibody status of patients attending a south-west London antenatal clinic, 2007–2012, Br. J. Biomed. Sci. 71 (2014) 111–114. [7] J.C. Watson, S.C. Hadler, C.A. Dykewicz, S. Reef, L. Phillips, Measles, mumps, and rubella-vaccine use and strategies for elimination of measles, rubella, and congenital rubella syndrome and control of mumps: recommendations of the Advisory Committee on Immunization Practices (ACIP), MMWR Recomm. Rep. 47 (1998) 1–57. [8] C.C. Lin, C.Y. Yang, Y.L. Shih, Y.Y. Huang, T.H. Yang, J.Y. Liang, C.F. Chang, H.S. Hsieh, Y.L. Huang, Persistence and titer changes of rubella virus antibodies in primiparous women who had been vaccinated with strain RA 27/3 in Junior High School, Clin. Vaccine Immunol. 19 (2012) 1–4. [9] A. Sauerbrei, P. Wutzler, Serological detection of varicella-zoster virus-specific immunoglobulin G by an enzyme-linked immunosorbent assay using glycoprotein antigen, J. Clin. Microbiol. 44 (2006) 3094–3097. [10] B.J. Cohen, D. Doblas, N. Andrews, Comparison of plaque reduction neutralisation test (PRNT) and measles virus-specific IgG ELISA for assessing immunogenicity of measles vaccination, Vaccine 26 (2008) 6392–6397. [11] L.P. Skendzel, Rubella immunity defining the level of protective antibody, Am. J. Clin. Pathol. 106 (1996) 170–174. [12] P.P. Mortimer, J.M.B. Edwards, A.D. Porter, R.S. Tedder, J.E. Mace, A. Hutchinson, Are many women immunized against rubella unnecessarily? J. Hyg. 87 (1981) 131–138. [13] J.B. Kurtz, P.P. Mortimer, P.R. Mortimer, P. Morgancapner, M.S. Shafi, G.B.B. White, Rubella antibody measured by radial hemolysis—characteristics and performance of a simle screening method for use in diagnostic laboratories, J. Hyg. 84 (1980) 213. [14] P.W. Neumann, J.M. Weber, Single radial hemolysis test for rubella immunity and recent infection, J. Clin. Microbiol. 17 (1983) 28–34. [15] B.T.H. Pustowoit, J. Hofmann, Aspects of Antibody Quantification in Rubellaand CMV-serology, European Group for Rapid Viral Diagnosis, 1994. [16] W. Dimech, N. Arachchi, J. Cai, T. Sahin, K. Wilson, Investigation into low-level anti-rubella virus IgG results reported by commercial immunoassays, Clin. Vaccine Immunol. 20 (2013) 255–261. [17] W. Dimech, A. Bettoli, D. Eckert, B. Francis, J. Hamblin, T. Kerr, C. Ryan, I. Skurrie, Multicenter evaluation of five commercial rubella virus
[22]
[23]
[24]
[25]
[26]
[27] [28]
[29]
[30]
[31]
[32]
[33] [34]
[35]
[36]
immunoglobulin G kits which report in international units per milliliter, J. Clin. Microbiol. 30 (1992) 633–641. W. Dimech, L. Panagiotopoulos, B. Francis, N. Laven, J. Marler, D. Dickeson, T. Panayotou, K. Wilson, R. Wootten, E.M. Dax, Evaluation of eight anti-rubella virus immunoglobulin g immunoassays that report results in international units per milliliter, J. Clin. Microbiol. 46 (2008) 1955–1960. N.P. Greenwood, I.G. Ovsyannikova, R.A. Vierkant, M.M. O’Byrne, G.A. Poland, A qualitative and quantitative comparison of two rubella virus-specific IgG antibody immunoassays, Viral Immunol. 23 (2010) 353–357. A.S. Weissfeld, W.D. Gehle, A.C. Sonnenwirth, Comparison of several test systems used for determination of rubella immune status, J. Clin. Microbiol. 16 (1982) 82–85. I.C. Shekarchi, J.L. Sever, N. Tzan, A. Ley, L.C. Ward, D. Madden, Comparison of hemagglutination inhibition test and enzyme-linked immunosorbent assay for determining antibody to rubella virus, J. Clin. Microbiol. 13 (1981) 850–854. R.A. Wittenburg, M.A. Roberts, L.B. Elliott, L.M. Little, Comparative evaluation of commercial rubella virus antibody kits, J. Clin. Microbiol. 21 (1985) 161–163. M. Enders, U. Bartelt, F. Knotek, K. Bunn, S. Strobel, K. Dietz, G. Enders, Performance of the elecsys rubella IgG assay in the diagnostic laboratory setting for assessment of immune status, Clin. Vaccine Immunol. 20 (2013) 420–426. C.M. Bradstreet, B. Kirkwood, J.R. Pattison, J.O. Tobin, The derivation of a minimum immune titre of rubella haemagglutination-inhibition (HI) antibody. A Public Health Laboratory Service Collaborative survey, J. Hyg. (Lond.) 81 (1978) 383–388. S. O’Shea, K.M. Corbett, S.M. Barrow, J.E. Banatvala, J.M. Best, Rubella reinfection; role of neutralising antibodies and cell-mediated immunity, Clin. Diagn. Virol. 2 (1994) 349–358. J. Nedeljkovic, T. Jovanovic, S. Mladjenovic, K. Hedman, N. Peitsaro, C. Oker-Blom, Immunoblot analysis of natural and vaccine-induced IgG responses to rubella virus proteins expressed in insect cells, J. Clin. Virol. 14 (1999) 119–131. M.G. Cusi, R. Metelli, P.E. Valensin, Immune responses to wild and vaccine rubella viruses after rubella vaccination, Arch. Virol. 106 (1989) 63–72. T. Zhang, C.A. Mauracher, L.A. Mitchell, A.J. Tingle, Detection of rubella virus-specific immunoglobulin G (IgG), IgM, and IgA antibodies by immunoblot assays, J. Clin. Microbiol. 30 (1992) 824–830. R.J. Diepersloot, H. Dunnewold-Hoekstra, J. Kruit-Den Hollander, F. Vlaspolder, Antenatal screening for hepatitis B and antibodies to Toxoplasma gondii and rubella virus: evaluation of two commercial immunoassay systems, Clin. Diagn. Lab. Immunol. 8 (2001) 785–787. D.B. Budzko, D.F. Jelinek, B.W. Wilcke Jr., Nonspecific reactions in the hemagglutination inhibition test for detection of rubella antibodies, J. Clin. Microbiol. 13 (1981) 818–823. H. Blom, G. Haukenes, Identification of non-specific serum inhibitors of rubella virus haemagglutination, Med. Microbiol. Immunol. 159 (1974) 271–277. I. Davidkin, S. Jokinen, M. Broman, P. Leinikki, H. Peltola, Persistence of measles, mumps, and rubella antibodies in an MMR-vaccinated cohort: a 20-year follow-up, J. Infect. Dis. 197 (2008) 950–956. PAHO/WHO, Elimination of rubella and congenital rubella syndrome in the Americas, PAHO Fact sheet 2015, April. D. Huzly, T. Schenk, W. Jilg, D. Neumann-Haefelin, Comparison of nine commercially available assays for quantification of antibody response to hepatitis B virus surface antigen, J. Clin. Microbiol. 46 (2008) 1298–1306. J. Hutton, P. Rowan, A. Greisinger, M. Mouzoon, Rubella monitoring in pregnancy as a means for evaluating a possible reemergence of rubella, Am. J. Obstet. Gynecol. 211 (2014) 534, e1-4. STIKO am RKI Empfehlungen der Ständigen Impfkommission (STIKO) am Robert Koch-Institut/Stand, August 2015, Epidemiologisches Bulletin 2015, Nr 34.
Contents lists available at ScienceDirect
Journal of Clinical Virology journal homepage: www.elsevier.com/locate/jcv
Performance of 14 rubella IgG immunoassays on samples with low positive or negative haemagglutination inhibition results Daniela Huzly ∗ , Ingeborg Hanselmann, Dieter Neumann-Haefelin, Marcus Panning Institute of Virology, University Medical Center Freiburg, Germany
a r t i c l e
i n f o
Article history: Received 13 August 2015 Received in revised form 14 October 2015 Accepted 16 November 2015 Keywords: Rubella CRS Assay comparison Seroepidemiologic studies
a b s t r a c t Background: Rubella IgG testing is routinely done in prenatal care and seroepidemiological studies. Recently concern was raised that seropositivity rates were decreasing questioning vaccination policies. Manufacturers of rubella IgG assays and authors of seroepidemiological studies use different cut-offs for the definition of seropositivity. As rubella virus circulation is reduced since many years, seronegativity rates might be overestimated using an inappropriate cut-off. Objectives: Using different cut-off definitions we compared fourteen current rubella IgG immunoassays for sensitivity and qualitative result concordance in samples with low positive or negative haemagglutination inhibition (HI) titre. Study design: 150 clinical samples from patients and health care workers were included in the study. All samples were measured in 14 different rubella IgG immunoassays. Seropositivity was defined using recombinant rubella IgG immunoblot as reference standard. Results: The concordance of qualitative results using the manufacturers cut-off definitions was 56.4% if grey-zone results were analysed separately and 69.8% if grey-zone results were defined as positive. Using universal cut-offs of 10 IU/ml or 15 IU/ml the concordance was 70% and 61.4% respectively. Using the different cut-off definitions up to 71 out of the 124 immunoblot-positive samples tested negative in the immunoassays. The mean coefficient of variation (CV) of quantitative results in positive samples was 51% (range 19–113%). Conclusions: Determination of rubella immunity by measurement of rubella-IgG in a population with high vaccination coverage with current assays leads to a high number of false negative results. The value of routine rubella antibody testing in countries with high vaccination coverage should be discussed. © 2015 Elsevier B.V. All rights reserved.
1. Background Rubella is an acute infection usually occurring during childhood with a mild clinical course. Primary rubella virus infection during early pregnancy, however, carries a high risk for fetal death and the development of congenital rubella syndrome (CRS) with malformations of eyes, ears and heart. Vaccination programs have reduced the incidence of CRS in most industrialized countries dramatically but in some countries, cases still occur. As systematic data on vaccination history are rarely available, seroepidemiological studies are conducted to evaluate the efficacy of vaccination policy. According to recent studies, seropositivity rates seem to decrease in countries with high vaccination coverage [1–3]. Based on this finding, concern
∗ Corresponding author at: Institute of Virology, University Medical Center, Hermann-Herder−Str. 11, 79104 Freiburg, Germany. Fax: +49 761 2036603. E-mail address: [email protected] (D. Huzly). http://dx.doi.org/10.1016/j.jcv.2015.11.022 1386-6532/© 2015 Elsevier B.V. All rights reserved.
was raised that an increasing number of susceptible women in child bearing age might question the effectiveness of present vaccination strategies. Numerous commercial assays are available to detect and quantify anti-rubella IgG. These assays are calibrated with an international reference preparation and results are reported in international units per ml, which in theory should lead to a standardization of test results and interpretation. However, serological assays for rubella IgG use different cut-offs for the definition of seropositivity. In addition, authors of seroepidemiological studies frequently use a universal cut-off of 10 or 15 IU/ml based on published recommendations and not the cut-off recommended by the manufacturer. With reduced virus circulation, antibody levels decline over time and therefore seronegativity rates might be overestimated due to a wrong cut-off definition.
14
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
2. Objectives The aim of this study was to compare the sensitivity and concordance of qualitative results of current rubella-IgG immunoassays defining seropositivity by IgG recombinant immunoblot.
3. Study design Study samples: A total of 150 clinical samples were included in the study. The samples obtained from health care workers, pregnant women, and children had been submitted to the Institute of Virology for routine rubella antibody testing and had been stored at −20 ◦ C thereafter. A total of 121/150 samples were collected prospectively based on a haemagglutination inhibition (HI) titre below 1:32. Another 29/150 samples with a HI titre of 1:32 or 1:64 were chosen retrospectively. To avoid multiple freeze–thaw cycles, samples were stored at 4 ◦ C during the main study period. The first International Standard for anti-rubella immunoglobulin RUBI-1-94 containing 1600 IU per vial (NIBSC) was measured in different dilutions to check the calibration of each immunoassay. The standard preparation was diluted in a human serum pool which tested negative for rubella-IgG using the reference assay. Recombinant immunoblot (reference assay): All samples were measured in a commercially available recombinant immunoblot (recomBlot Rubella IgG, Mikrogen, Martinsried, Germany) on a Tecan Profiblot. All kits were of the same lot, and a negative and an E2 weakly positive control as well as the 12.5 IU/ml NIBSC standard dilution were run with each kit. The E2 band intensity of the E2 weakly positive control was used as reference for the definition of band positivity. All sera showing at least E1-reactivity on the immunoblot were defined as rubella IgG positive. Haemagglutination-inhibition test (HI): All sera had previously been tested by a commercially available CE labelled rubella HI assay (Labor Dr. Merck & Kollegen, Ochsenhausen, Germany) as recommended by the manufacturer. A titre of 1:8 is considered as positive but a confirmation of results below 1:32 is recommended. The distribution of HI results is shown in Supplementary file 1. Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.jcv.2015.11.022. Enzyme immunoassays: Seven different enzyme immunoassays were used: Enzygnost Anti-Rubella-Virus/IgG (Siemens Healthcare, Fernwald, Germany), Serion ELISA classic Rubella Virus IgG quant. (Institut Virion\Serion GmbH, Würzburg, Germany) Elimmun–Rubella G (Labor Dr. Merck & Kollegen, Ochsenhausen, Germany), Rubella-IgG-ELISA PKS medac (medac Diagnostica, Hamburg, Germany), Vircell Rubella ELISA IgG (Vircell, Granada, Spain), Euroimmun Anti-Rubella virus ELISA (IgG) (Euroimmun, Lübeck, Germany), Dia Sorin ETI-RUBEK-G PLUS (DiaSorin, Dietenbach, Germany). All sera were diluted manually and automatically processed in a Siemens Elisa Processor BEP III. As the Vircell Rubella ELISA IgG generates semiquantitative results (Signal/Cutoff*10) we used the international standard preparation to obtain the results in IU/ml, as recommended by the manufacturer (personal communication). (see Supplementary file 2 for technical specifications). Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.jcv.2015.11.022. Automated immunoassays: Seven automated immunoassays were used: Architect Rubella IgG on the Abbott Architect system (Abbott, Chicago, USA), Abbott Axsym Rubella IgG on the Abbott Axsym system (Abbott, Chicago USA), ADVIA Centaur Rubella G on the Siemens ADVIA Centaur system (Siemens Healthcare, Fernwald, Germany), DiaSorin Liaison Rubella IgG on the LIAISON system (DiaSorin, Torino, Italy), Cobas Rubella IgG on the Roche ELECSYS system (Roche, Mannheim, Germany), Vitros Rubella IgG on the Vitros ECi System (Ortho Clinical Diagnostics, Raritan, USA), and VIDAS RUB
IgG II on the mini VIDAS System (Biomerieux, Lyon, France) (see Supplementary file 3 for technical specifications). Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.jcv.2015.11.022. The tests by Abbott Architect and by Roche Cobas were performed at the end of the study. The leftover volume of the samples was small, thus, the number of sera measured in these systems was limited to 130 in the Abbott Architect System and 96 sera in the Roche system. Statistical analysis: The recombinant immunoblot was used as reference standard for the definition of seropositivity. For assessment of immunoassay results we first used the cut-offs given by the manufacturers and then reassessed the results using 10 IU/ml and 15 IU/ml as universal cut-off for all assays. The universal cut-offs were also in accordance with supplementary threshold recommendations of several manufacturers. Statistical analyses were done with SPSS Version 20 and MEDCALC. The serum with questionable reactivity in the immunoblot was excluded from specificity, sensitivity as well as ROC analyses. Negative predictive values were calculated using published vaccination data of Germany and assuming a 95% efficacy of vaccination. For all positive samples the coefficient of variation of quantitative values between the different assays was determined. For comparison of the different assays with HI results, mean IU concentrations and range as well as 95% confidence interval at different HI titres were calculated. 4. Results 4.1. Baseline demographic data Median age of patients was 29 years (range 2–67 years), and 122/150 (81%) were female. Indication for rubella serology were in-vitro fertilisation in 79/150, routine health care in 27/150, serostatus before immunosuppressive therapy in 26/150, suspicion of acute rubella virus infection in 13, and pregnancy in 5/150 patients. 4.2. Assessment of seropositivity using recombinant immunoblot A total of 124/150 (83%) samples were positive and 25 samples were negative for rubella-IgG, based on immunoblot results. One sample could not be definitely defined as there was only a weak E1band with lower intensity than the weakly positive control band. In all immunoblot-positive samples, at least two bands (E1 and E1–E2) were reactive. 4.3. Assessment of test results using manufacturers recommendations Using the cut-offs given by the manufacturers, five of the 25 immunoblot-negative samples were reactive and two were equivocal by Abbott Axsym. Four of these samples also showed positive or borderline reactivity in the DiaSorin ELISA, but were measured below 2 IU in all other assay systems and negative by HI. Three different samples showed borderline results in the LIAISON, Centaur and Architect. 16 samples were concordantly negative. A summary of the results are shown in Table 1. Between one and 46 out of the 124 immunoblot-positive samples are negative in the immunoassays (Table 2). Between 4 and 21 samples had grey zone results in the different assays, of which most tested positive by immunoblot. A summary of the analysis of greyzone results are shown in Table 3. The concordance of qualitative results using the manufacturers cut-off definitions was 56.4% if grey-zone results were analysed separately and 69.8% if grey-zone results were defined as positive.
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
15
4.4. Assessment of test results using universal cut-offs
4.6. Comparison of quantitative results of immunoassays
Using 10 IU/ml as a universal cut-off, 7/25 immunoblot-negative samples were reactive in one (n = 3), two (n = 3) or three (n = 1) immunoassays, whereas 18/25 samples were concordantly negative (Table 1). Between 5 and 41 out of the 124 immunoblot-positive samples were negative in the immunoassays (Table 2). The overall concordance of results was 70%. Next, using 15 IU/ml as universal cut-off, 4/25 immunoblot-negative samples were reactive in one (n = 3) or two immunoassays (Table 1) and between 12 and 71 out of the 124 immunoblot-positive samples are negative in the immunoassays (Table 2). The concordance of results using 15 IU/ml as cut-off was 61.4%.
The mean coefficient of variation (CV) of positive samples was 51% (range 19%–113%). There was no significant difference between CVs of sera with low values (mean < 35 IU/ml) and sera with high values.
4.5. Correlation of HI results with immunoblot and with IU/ml measured by different immunoassays
4.7. Measurement of standard serum using immunoassays Next, twofold dilutions of the NIBSC standard were analysed using the immunoassays. The 100 IU/ml standard was measured too low or too high in some systems (Fig. 1). The assay from Vircell was not able to detect the 12.5 and 25 IU/ml standard dilution (Fig. 2).
4.8. Specificity, sensitivity and ROC analysis of immunoassays In comparison to the immunoblot results, four out of the 25 immunoblot-negative samples had a titre of 1:8 by HI and 11/124 immunoblot-positive samples tested negative by HI yielding an overall concordance of 88%. There was no clear correlation between HI results and results in IU/ml (data not shown). Immunoassay results in samples with a titre of 1:8 by HI are shown in Fig. 1.
Finally, specificity and sensitivity analyses using the different cut-off definitions were determined in our cohort with low HI result. Specificity ranged from 80 to 100% depending on the cutoff used (Table 1). On the other hand, sensitivity ranged from 67 to 100% (Table 2).
Table 1 Specificity of the different assay systems in 25 immunoblot negative samples using 10 IU/ml or 15 IU/ml in comparison to the manufacturers cut-off. Assay system
Abbott Architecta Abbott Axsym Siemens Centaur Siemens enzygnost EIA Biomérieux Vidas DiaSorin EIA DiaSorin LIAISON Euroimmun EIA Medac EIA Merck EIA Ortho ECI Roche COBASb Serion EIA Vircell EIA a b
Number of false positive results (% specificity) Manufacturers’ cut-off, equivocals excluded
Manufacturers’ cut-off, equivocals = positive
Cut-off at 10 IU/ml
Cut-off at 15 IU/ml
0 (100) 5 (80) 0 (100) 0 (100) 0 (100) 2 (92) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100)
1 (95) 7 (72) 1 (96) 0 (100) 0 (100) 2 (92) 1 (96) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100)
0 (100) 5 (80) 0 (100) 0 (100) 0 (100) 2 (92) 0 (100) 0 (100) 1 (96) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100)
0 (100) 3 (88) 0 (100) 0 (100) 0 (100) 2 (92) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 0 (100)
21 negative samples. 16 negative samples.
Table 2 Number and percentage of false negative samples in the different immunoassays using manufacturers cut-offs, or a universal cut-off at 10 IU/ml or 15 IU/ml, respectively. b 80 positive samples. c Manufacturer’s cut-off in Arbitrary Units (AU). Assay system
Abbott Axsym Abbott Architecta Bayer Centaur Siemens enzygnost EIA Biomérieux Vidas DiaSorin EIA DiaSorin LIAISON Medac EIA Merck EIA Ortho ECI Roche Cobasb Serion EIA Vircell EIA a
109 positive samples. 80 positive samples. c Manufacturer’s cut-off in Arbitrary Units (AU). b
Number of false negative results (% sensitivity) Manufacturers’ cut-off
Cut-off at 10 IU/ml
Cut-off at 15 IU/ml
1 (99.2) 5 (95.4) 2 (98.4) 0 (100) 10 (91.9) 12 (90.3) 5 (96.0) 20 (83.9) 29 (76.6) 6 (95.2) 10 (87.5) 7 (94.4) 46 (62.9)c
8 (93.5) 20 (81.7) 12 (90.3) 20 (83.9) 10 (91.9) 12 (90.3) 5 (96.0) 13 (89.5) 20 (83.9) 6 (95.2) 11 (86.3) 7 (94.4) 41 (67.0)
23 (81.5) 34 (68.8) 19 (84.7) 36 (71.0) 21 (83.1) 19 (84.7) 13 (89.5) 33 (73.4) 31 (75.0) 12 (90.3) 17 (78.8) 19 (84.7) 71 (42.7)
16
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
Fig. 1. Boxplot of measured values by immunoassays in samples with a titre of 1:8 by HI (minimum, 1. quartile, median, 3. quartile, maximum, outliers with sample ID). Automated assays on the left.
Table 3 Immunoblot results of samples with equivocal results using manufacturers cut-offs. Assay system
Number of equivocal results
Positive by immunoblot (n)
Abbott Architect Abbott Axsym Biomérieux Vidas DiaSorin EIA DiaSorin Liaison Euroimmum EIA Medac EIA Merck EIA Ortho ECI Roche (no greyzone) Serion EIA Siemens Centaur Siemens Enzygnost EIA Vircell EIA
17 9 11 7 6 13 21 4 6 0 24 11 11 8
16 7 11 7 5 13 21 4 6 0 24 10 11 8
ROC analysis yielded cut-offs below 10 IU/ml and below the manufacturer’s cut-off in most assays (Table 4).
Fig. 2. Measurement of the first international standard for anti-rubella immunoglobulin RUBI-1-94, NIBSC.
Table 4 ROC analysis PPV = positive predictive value, NPV = negative predictive value. Assay
Highest Youden Index at IU/ml
Sensitivity (%)
Specificity (%)
PPV (%)
Abbott Architect Abbott Axsym Biomérieux Vidas DIaSorin EIA DiaSorin LAISON Euroimmun EIA Medac EIA Merck EIA Ortho ECI Roche cobas Serion EIA Siemens Centaur Siemens Enzygnost Vircell
5.5 7.1 4 6.4 5 4.1 5.2 6.5 1 2 1 5.3 4 3
96.3 97.6 97.6 91.1 100 97.6 99.2 93.5 95 98.7 98.4 98.4 100 80.6
100 80 96 92 100 100 96 96 100 100 100 100 100 100
100 96 99.2 95.5 100 100 99.2 99.1 100 100 100 100 100 100
NPV (%) 84.5 92.6 88.9 68 100 89.3 96 75 80.6 94.1 92.6 92.6 100 51
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
5. Discussion The presence of specific rubella antibodies correlates with immunity, but a specific type or level of antibodies correlating with absolute protection has not been identified [4]. To avoid unnecessary vaccination of individuals and irritation during pregnancy, an appropriate definition of seropositivity is essential. According to recent seroepidemiological studies from countries with high vaccination coverage, the number of women not protected against rubella is growing [1,2,5,6]. However, despite an apparently increasing number of seronegative individuals, rubella cases did not increase and practically no CRS cases occur in countries with high vaccination coverage. Studies have shown, that in the absence of widespread virus circulation, individual antibody concentrations will get lower over time and the increasing number of seronegative individuals could just reflect a low negative predictive value of rubella IgG assays [2,7,8]. Similar effects have been shown for measles and varicella-zoster virus, where routine antibody assays have a low sensitivity in highly vaccinated populations [9,10]. Technically, to determine the serostatus for rubella, different test systems are on the market and an international standard preparation is available, which should render assay systems comparable. This has, however, not been successfull, as shown by our study. In our cohort of individuals with low antibody titres by HI, up to 37% of the immunoblot-positive samples (our reference assay) were negative in various immunoassays. Importantly, the concordance of commercial immunoassay results was only 56% using manufacturers cut-off definitions including the greyzone results. Concordance of results with manufacturers cut-offs was better when grey zone results were interpreted as seropositive, indicating that cut-off definition could be the main problem. This is also reflected by the fact that almost all samples with equivocal results in one or more immunoassays were positive in our reference assay. Using universal cut-offs of 10 IU/ml as recommended by Skendzel [11] or 15 IU/ml, as recommended by several authors [12–15], the concordance of immunoassay results did not improve. It is important to note, that the first IgG anti-rubella standard prepared in 1970 was based on human normal immunoglobulin. The International Unit (IU) for anti-rubella serum was then defined on the basis of neutralization test results from a collaborative study. This standard was termed Second International Reference Preparation of Anti-Rubella Serum. All subsequent standards, including RUBI-1-94 introduced in 1995, were calibrated against this reference preparation. The RUBI-1-94 standard consists of a pool of human normal immunoglobulin and was estimated to contain 1600 IU per ampoule. It has been used to calibrate recent commercial anti-rubella IgG assays. However, the dilutions of the reference standard were not measured accurately by the assays in our study. The use of a universal cut-off implies comparability of quantitative assay results. However, numerous studies comparing quantitative results of different assay systems come to the conclusion, that measured values are highly discrepant and comparability of results is not accomplished by the use of standard preparations [16–23]. We have shown very high discrepancies even in our small cohort of individuals with low HI titres. We were also not able to show a correspondence of HI titres with certain mean IU values [11,24]. Although, these high discrepancies between assay results have been described from the beginning, most assay manufacturers used the proposed universal cut-offs of 10 IU/ml or 15 IU/ml for their assays and relied on these fixed values instead of defining the limit of detection for the newer assay techniques. We used the recombinant immunoblot as reference standard to define true seropositivity in our set of low positive samples. Immunoblot testing was chosen as reference assay in our study as in contrast to neutralization assays it is highly standardized [4,25]. E1 antibodies on the immunoblot can be shown in all individuals after rubella vac-
17
cination or wildtype virus infection [26], indicating high sensitivity. It has been shown that E1 reactivity on westernblot correlates with neutralising antibodies [27]. Specificity of E1 antibody positivity in the immunoblot has been shown by the fact that individuals with E1 antibodies showed booster responses upon challenge [28]. Defining immunoblot-positive samples as true seropositive, most assays, like in previous studies, showed very high specificity. The comparably low specificity of the Abbott Axsym had been shown before [29]. False positive HI results at a dilution of 1:8 are probably due to the presence of non-specific inhibitors of haemagglutination, as has been shown in previous studies [30,31]. We performed ROC analyses for the different assays and in all cases the calculated cut-offs were below 10 IU/ml, without losing specificity in the majority of assays. One seroepidemiological study used the lower limit of quantification of the Enzygnost assay (4 IU/ml), which was also the cut-off result of our ROC analysis, and showed a persistently high rate of seropositivity over the years in the US [32]. This reflects rubella epidemiology much better. Recently, the PanAmerican Health Organization of the World Health Organization declared that the Americas region had eliminated endemic transmission of rubella and CRS [33]. In light of our and previously published results, the value of reference standards for the quantification of antibodies should be critically addressed. We have shown that even anti-HBs results are highly discrepant although most assays use the same antigen [34]. Rubella IgG assays use different antigens and different assay constructions, so differences are to be expected. Reporting results in IU/ml with the high variation between different assays can lead to wrong interpretations by clinicians as results from different labs may imitate a significant titre rise. A recent publication even postulates high rubella exposition rates in pregnant women, based on increasing titres from the beginning to the end of pregnancy. However, prenatal rubella testing had been done by different laboratories using various assay systems [35]. No CRS case had occurred in that study. The value of routine rubella antibody testing in countries with high vaccination coverage has to be discussed. Several countries now recommend testing only in individuals without documented vaccination history [36,37].
6. Conclusion Definition of rubella immunity by measurement of rubella-IgG in a population with high vaccination coverage with current assays does lead to a relatively high number of false negative results. Assay manufacturers should reevaluate the cut-off of their assays. To avoid the diagnosis of rubella reinfection on the basis of significant titre rise between measurements by different laboratories, quantification of IgG assays using International Units should be abandoned. After 40 years trying with different standards and different assay concepts we should accept that standardization of IgG quantitation is not reliably feasible. Assay developers should concentrate on true test accuracy parameters when developing new assays.
Conflict of interest The authors have no conflict of interest to declare.
Sources of funding None.
18
D. Huzly et al. / Journal of Clinical Virology 74 (2016) 13–18
Ethical approval Not required. Acknowledgement
[18]
[19]
All test kits were contributed by the manufacturers free of charge.
[20]
References
[21]
[1] R. Barlinn, K. Vainio, H.H. Samdal, S.A. Nordbo, H. Nokleby, S.G. Dudman, Susceptibility to cytomegalovirus, parvovirus B19 and age-dependent differences in levels of rubella antibodies among pregnant women, J. Med. Virol. 86 (2014) 820–826. [2] G. Smits, L. Mollema, S. Hahne, H. de Melker, I. Tcherniaeva, F. van der Klis, G. Berbers, Seroprevalence of rubella antibodies in The Netherlands after 32 years of high vaccination coverage, Vaccine 32 (2014) 1890–1895. [3] Coyle P. Viral Congenital and Neonate Infections ESCV Book of Abstracts 2014; Keynote lecture at the ESCV congress in Prague 2014. [4] W.H.O. The Immunological Basis for Immunization Series. Immunization, Vaccines and Biologicals 2008, Module 11: Rubella: 1–25. [5] S.A. Alsibiani, Rubella immunity among pregnant women in Jeddah, Western Region of Saudi Arabia, Obstet. Gynecol. Int. 2014 (2014), 659838. [6] S. Mortlock, S. Farthing, Rubella antibody status of patients attending a south-west London antenatal clinic, 2007–2012, Br. J. Biomed. Sci. 71 (2014) 111–114. [7] J.C. Watson, S.C. Hadler, C.A. Dykewicz, S. Reef, L. Phillips, Measles, mumps, and rubella-vaccine use and strategies for elimination of measles, rubella, and congenital rubella syndrome and control of mumps: recommendations of the Advisory Committee on Immunization Practices (ACIP), MMWR Recomm. Rep. 47 (1998) 1–57. [8] C.C. Lin, C.Y. Yang, Y.L. Shih, Y.Y. Huang, T.H. Yang, J.Y. Liang, C.F. Chang, H.S. Hsieh, Y.L. Huang, Persistence and titer changes of rubella virus antibodies in primiparous women who had been vaccinated with strain RA 27/3 in Junior High School, Clin. Vaccine Immunol. 19 (2012) 1–4. [9] A. Sauerbrei, P. Wutzler, Serological detection of varicella-zoster virus-specific immunoglobulin G by an enzyme-linked immunosorbent assay using glycoprotein antigen, J. Clin. Microbiol. 44 (2006) 3094–3097. [10] B.J. Cohen, D. Doblas, N. Andrews, Comparison of plaque reduction neutralisation test (PRNT) and measles virus-specific IgG ELISA for assessing immunogenicity of measles vaccination, Vaccine 26 (2008) 6392–6397. [11] L.P. Skendzel, Rubella immunity defining the level of protective antibody, Am. J. Clin. Pathol. 106 (1996) 170–174. [12] P.P. Mortimer, J.M.B. Edwards, A.D. Porter, R.S. Tedder, J.E. Mace, A. Hutchinson, Are many women immunized against rubella unnecessarily? J. Hyg. 87 (1981) 131–138. [13] J.B. Kurtz, P.P. Mortimer, P.R. Mortimer, P. Morgancapner, M.S. Shafi, G.B.B. White, Rubella antibody measured by radial hemolysis—characteristics and performance of a simle screening method for use in diagnostic laboratories, J. Hyg. 84 (1980) 213. [14] P.W. Neumann, J.M. Weber, Single radial hemolysis test for rubella immunity and recent infection, J. Clin. Microbiol. 17 (1983) 28–34. [15] B.T.H. Pustowoit, J. Hofmann, Aspects of Antibody Quantification in Rubellaand CMV-serology, European Group for Rapid Viral Diagnosis, 1994. [16] W. Dimech, N. Arachchi, J. Cai, T. Sahin, K. Wilson, Investigation into low-level anti-rubella virus IgG results reported by commercial immunoassays, Clin. Vaccine Immunol. 20 (2013) 255–261. [17] W. Dimech, A. Bettoli, D. Eckert, B. Francis, J. Hamblin, T. Kerr, C. Ryan, I. Skurrie, Multicenter evaluation of five commercial rubella virus
[22]
[23]
[24]
[25]
[26]
[27] [28]
[29]
[30]
[31]
[32]
[33] [34]
[35]
[36]
immunoglobulin G kits which report in international units per milliliter, J. Clin. Microbiol. 30 (1992) 633–641. W. Dimech, L. Panagiotopoulos, B. Francis, N. Laven, J. Marler, D. Dickeson, T. Panayotou, K. Wilson, R. Wootten, E.M. Dax, Evaluation of eight anti-rubella virus immunoglobulin g immunoassays that report results in international units per milliliter, J. Clin. Microbiol. 46 (2008) 1955–1960. N.P. Greenwood, I.G. Ovsyannikova, R.A. Vierkant, M.M. O’Byrne, G.A. Poland, A qualitative and quantitative comparison of two rubella virus-specific IgG antibody immunoassays, Viral Immunol. 23 (2010) 353–357. A.S. Weissfeld, W.D. Gehle, A.C. Sonnenwirth, Comparison of several test systems used for determination of rubella immune status, J. Clin. Microbiol. 16 (1982) 82–85. I.C. Shekarchi, J.L. Sever, N. Tzan, A. Ley, L.C. Ward, D. Madden, Comparison of hemagglutination inhibition test and enzyme-linked immunosorbent assay for determining antibody to rubella virus, J. Clin. Microbiol. 13 (1981) 850–854. R.A. Wittenburg, M.A. Roberts, L.B. Elliott, L.M. Little, Comparative evaluation of commercial rubella virus antibody kits, J. Clin. Microbiol. 21 (1985) 161–163. M. Enders, U. Bartelt, F. Knotek, K. Bunn, S. Strobel, K. Dietz, G. Enders, Performance of the elecsys rubella IgG assay in the diagnostic laboratory setting for assessment of immune status, Clin. Vaccine Immunol. 20 (2013) 420–426. C.M. Bradstreet, B. Kirkwood, J.R. Pattison, J.O. Tobin, The derivation of a minimum immune titre of rubella haemagglutination-inhibition (HI) antibody. A Public Health Laboratory Service Collaborative survey, J. Hyg. (Lond.) 81 (1978) 383–388. S. O’Shea, K.M. Corbett, S.M. Barrow, J.E. Banatvala, J.M. Best, Rubella reinfection; role of neutralising antibodies and cell-mediated immunity, Clin. Diagn. Virol. 2 (1994) 349–358. J. Nedeljkovic, T. Jovanovic, S. Mladjenovic, K. Hedman, N. Peitsaro, C. Oker-Blom, Immunoblot analysis of natural and vaccine-induced IgG responses to rubella virus proteins expressed in insect cells, J. Clin. Virol. 14 (1999) 119–131. M.G. Cusi, R. Metelli, P.E. Valensin, Immune responses to wild and vaccine rubella viruses after rubella vaccination, Arch. Virol. 106 (1989) 63–72. T. Zhang, C.A. Mauracher, L.A. Mitchell, A.J. Tingle, Detection of rubella virus-specific immunoglobulin G (IgG), IgM, and IgA antibodies by immunoblot assays, J. Clin. Microbiol. 30 (1992) 824–830. R.J. Diepersloot, H. Dunnewold-Hoekstra, J. Kruit-Den Hollander, F. Vlaspolder, Antenatal screening for hepatitis B and antibodies to Toxoplasma gondii and rubella virus: evaluation of two commercial immunoassay systems, Clin. Diagn. Lab. Immunol. 8 (2001) 785–787. D.B. Budzko, D.F. Jelinek, B.W. Wilcke Jr., Nonspecific reactions in the hemagglutination inhibition test for detection of rubella antibodies, J. Clin. Microbiol. 13 (1981) 818–823. H. Blom, G. Haukenes, Identification of non-specific serum inhibitors of rubella virus haemagglutination, Med. Microbiol. Immunol. 159 (1974) 271–277. I. Davidkin, S. Jokinen, M. Broman, P. Leinikki, H. Peltola, Persistence of measles, mumps, and rubella antibodies in an MMR-vaccinated cohort: a 20-year follow-up, J. Infect. Dis. 197 (2008) 950–956. PAHO/WHO, Elimination of rubella and congenital rubella syndrome in the Americas, PAHO Fact sheet 2015, April. D. Huzly, T. Schenk, W. Jilg, D. Neumann-Haefelin, Comparison of nine commercially available assays for quantification of antibody response to hepatitis B virus surface antigen, J. Clin. Microbiol. 46 (2008) 1298–1306. J. Hutton, P. Rowan, A. Greisinger, M. Mouzoon, Rubella monitoring in pregnancy as a means for evaluating a possible reemergence of rubella, Am. J. Obstet. Gynecol. 211 (2014) 534, e1-4. STIKO am RKI Empfehlungen der Ständigen Impfkommission (STIKO) am Robert Koch-Institut/Stand, August 2015, Epidemiologisches Bulletin 2015, Nr 34.