- Email: [email protected]
Comparison of four enzyme immunoassays for the detection of cytomegalovirus IgG antibodies
Clinical and Diagnostic Virology, 1 (1993) 215-223
© 1993 ElsevierSciencePublishers B.V. All rights reserved0928-0197/93/$06.00
Clinical and
DIAVIR00027
Diagnostic
Virology Comparison of four enzyme immunoassays for the detection of cytomegalovirus IgG antibodies Louise Pedneault a.b, Louise Robillard b, Pierre Harvey a and Jean Joncas a.b aDepartment of Microbiology and Immunology, Universit~de Montreal, Montreal, Quebec, Canada and bDepartment of Microbiology and Infectious Diseases, H~pital Sainte-Justine, Montreal, Quebec, Canada
(Received8 March 1993;revisionreceived25 May 1993;accepted31 May 1993)
Abstract Four commercial enzyme immunoassays (EIA), namely the Behring Enzygnost EIA (BE-EIA), Abbott IMx, Whittaker CMV STAT Test Kit and Diamedix assay, were evaluated for the detection of CMV IgG. The methods were compared as to sensitivity, specificity, positive and negative predictive values, global agreement, ease of performance and, for a small number of specimens, reproducibility. Discordant results were resolved by using the Gull CMV indirect fluorescent antibody (IFA) method. Our data suggest that all four assays were valuable screening tools for the detection of CMV IgG based on their high sensitivity and high negative predictive value. However, differences were noted in the reproducibility level and in the incidence of false-positive, equivocal and nonspccific results regarding certain tests in particular. In our hands, the Abbott IMx and the BE-EIA ranked high in the performance characteristics for a good screening test, yet the Abbott IMx offers the added advantages of being the easiest to perform and having the most rapid turnaround time. Key words: Cytomegalovirus;IgG antibody;Enzymeimmunoassay.
Introduction Cytomegalovirus (CMV) infects approximately 50% of adults in developed countries and nearly 100% of the adult population in developing countries (Gold and Nankervis, 1991). In the normal host, CMV infection is usually mild and asymptomatic or, occasionnally, manifests itself as a mononucleosis syndrome. In the immunosuppressed host, primary or reactivated CMV infection may be asymptomatic, but may also give rise to a mononucleosis-like syndrome more often than in a healthy individual, and to complications such as fever, pneumonitis, chorioretinitis, hepatitis Correspondence to: L. Pedneault,D6partement de Microbiologie,H6pital Sainte-Justine,3175 Chemin C6te Sainte-Catherine,Montr6al,Qu6bec,Canada H3T 1CS.Fax:(514) 345-4818;Tel:(514) 345-4643. Note: This work was presentedin part at the 8th AnnualClinicalVirologySymposium,Clearwater,FL, USA, May 1992. Abbreviations: CMV, cytomegalovirus,IgG, immunoglobulinG, EIA, enzyme immunoassay,IFA, indirect fluorescenceassay.
216 or gastrointestinal lesions (Ho, 1990). Furthermore, in utero infection may result in various sequelae in the newborn such as mental retardation, chorioretinitis, hearing loss and neurologic problems (Pass et al., 1980). CMV can be transmitted in several ways; an infant may acquire CMV transplacentally, during birth (Reynolds et al., 1973) or through ingestion of infected breast milk (Stagno et al., 1980). Seronegative individuals of all ages may acquire CMV through infected blood products (Preiksaitis et al., 1988; Yeager et al., 1981) or donated organs (Ho et al., 1975; Preiksaitis et al., 1983), or through direct contact with the secretions of an infected individual (Gold and Nankervis, 1991; Ho, 1990). The selection and use of seronegative blood products or of donated organs from seronegative individuals to immunocompromised patients remains a vital consideration in patient management. The decision to use such products is usually based on the CMV serologic status of the donor. For that matter, the choice of an adequate serologic test for the detection of CMV antibodies becomes crucial. In this study, we present the results from 519 serum specimens analyzed for the presence of CMV IgG antibodies by four EIA procedures. Discordant results were resolved by using the Gull CMV indirect fluorescent antibody (IFA) method.
Materials and Methods
Specimens A total of 519 sera were evaluated for the presence of CMV IgG antibodies by the various EIA tests described below. Sera were randomly selected from individuals for whom a CMV serology had been requested. Specimens were stored at -20°C and underwent on average two freeze-thaw cycles.
Comparison testing procedures BE-EIA (Behringwerke, Marburg, Germany) is based on the use of two wells per serum sample at a time, one of them being coated with viral antigen and the other with control antigen. Virus antigen and control antigen are obtained from human cell cultures infected with CMV and uninfected, respectively, and inactivated before coating. The test was performed as part of the routine workload by a laboratory technician at Sainte-Justine Hospital according to the manufacturer's instructions. BE-EIA requires an initial serum volume of 20/al for a final dilution of the sample of 1:44. A specimen is considered positive if the A viral antigen minus A control antigen (AA) is ~>0.2, negative if the AA is <0.2, and doubtful if the AA value lies between 0.2 and 0.3. The latter needs to be repeated. If the result remains doubtful after the retest, it should be confirmed using another approved method. In cases where the background seemed particularly high (AA<0.2 when the control antigen value is >0.2), we decided to consider the result 'nonspecific' (NS) (this is assessed as negative by Behring). Comparison testing with the IMx CMV IgG microparticle based-EIA (Abbott Laboratories, Abbott Park, IL), the CMV STAT Test Kit (Whittaker Bioproducts, Walkersville, MD) and Diamedix (Diamedix Corporation, Miami, FL) EIAs for detection of CMV IgG antibodies was performed blindly by a laboratory technician at Sainte-Justine Hospital according to the manufacturer's instructions. The STAT Test and Diamedix assay are standard EIAs using wells coated with inactivated CMV as a source of antigen. However, to perform the IMx microparticle based-EIA
217
(MEIA), the patient's serum is added to a microparticle coated with inactivated CMV and, after incubation, an aliquot of the antigen-antibody complex formed is transferred to a glass fiber matrix. The microparticles will bind irreversibly to this matrix. An anti-human IgG conjugated to alkaline phosphatase is then added, followed by the appropriate substrate to allow detection of CMV IgG antibodies if present. No control antigen wells or microparticles are used in any of the three comparison tests, the source of CMV antigen being partially purified in all cases. Initial serum volumes of 150 ~1, 10 ~1 and 5 #1 are necessary to perform the IMx, STAT Test, and Diamedix EIAs, respectively. These test samples are used undiluted in the IMx assay, but must be brought to a final dilution of 1:21 (STAT Test) or 1:41 (Diamedix) for the two other assays. The interpretation of the results for the various EIAs was as follows: (1) IMx: a sample is considered positive if it contains at least 15 AU (for antibody units)/ml of serum ( >~ 15 AU/ml), and negative if the reading is < 15 AU/ml. (2) STAT Test: a sample is considered positive if its predicted index value (PIV) is ~> 1.0, equivocal if the PIV lies between 0.80 and 0.99, and negative if the PIV is ~<0.79. (3) Diamedix: a sample is considered positive if it contains more than 23 EU (for EIA units)/ml of serum (> 23 EU/ml), negative if it contains less than 18 EU/ml, and equivocal if the reading is between 18 and 23 EU/ml.
Resolution of discordant results We chose IFA as a reference method because it allows to visualize and discriminate among various reactivities by evaluating the cellular and subcellular patterns of fluorescence, and because we considered that an alternate method might be useful to resolve some of the discrepant EIA results. Thirty-eight sera that gave a discordant result in only one EIA, the 8 sera that were found NS with BE-EIA and positive with the three other EIAs, 15 randomly chosen sera that were NS when tested by BE-EIA and negative with the three other EIAs, and four other randomly selected sera were evaluated by IFA (Gull Laboratories, Salt Lake City, Utah) according to the manufacturer's instructions. Furthermore, 12 serum specimens positive by all 4 EIAs and 12 serum specimens negative by all 4 EIAs were used as controls. At the screening dilution (1:10), a greenish-yellow fluorescence of about 10% or more of the infected cells in each field, with a whole cell or nuclear pattern, is considered positive. Absence of fluorescence is read as negative.
Analysis of results All testing was carried out without knowledge of the results obtained with the other EIAs. For a given serum specimen to be considered positive for the presence of CMV IgG antibody, concordance of at least three of the four EIA methods or, in cases where IFA was performed, confirmation of positivity by IFA was required. Conversely, for a sample to be considered negative, concordance of three or more EIAs or, when relevant, confirmation of negativity by IFA must be obtained. Equivocal, non-specific and doubtful results for each method, as well as the five serum specimens which gave positive results by two EIAs and negative results with the two other assays, were excluded from the preliminary analysis table (Table IVA., unresolved data). Therefore, only the serum samples considered positive or negative by concordance of at least three EIAs were included in this table (Table IVA., unresolved data). The IFA results obtained on the 89 serum samples tested by this
218
method (Table III) were included in the final analysis of the data (Table IVB., resolved data). Sensitivity, specificity, positive and negative predictive values, incidence of false-positive and -negative results, and global agreement were calculated using the formulas recommended by Griner et al. (1981).
Results
A total of 519 sera were evaluated by four different EIAs for the detection of CMV IgG antibodies (Table I). According to our seropositivity and seronegativity criteria, 159 sera (31%) were CMV IgG positive and 226 (44%) were found negative, for a complete agreement of 75%. However, 104 (20%) sera gave NS results when tested with the BE-EIA. Excluding these 104 sera, the complete agreement for the four assays becomes 385/415 (93%). The Behring Enzygnost enzyme immunoassay (BE-EIA) for the detection of CMV IgG antibodies frequently gives NS results, thereby rendering the interpretation of the test difficult. The results for the 104 samples yielding NS results by BE-EIA are detailed in Table II. Thirty-seven (36%) of these specimens were found negative with the three other EIAs. If we presume that a NS result is, in fact, negative, as suggested by the manufacturer, 31 more serum samples would therefore be considered negative, for a total of 68/104 (65%), based on the criteria previously established for seronegativity where concordance of at least three EIAs is necessary to categorize a specimen as positive or negative. Furthermore, eight (8%) of the NS specimens were positive by all three other EIAs. We decided that, to resolve some of the discrepant results obtained when comparing EIA methods among themselves, we should use a different assay, in this case, IFA. Eighty-nine serum samples were thus submitted to IFA testing (Table III). After resolution, there were still 2, 11, and 18 specimens previously considered false-positive by IMx, STAT Test and Diamedix, respectively, that remained classified as such. TABLE I C M V s e r o r e a c t i v i t y r e s u l t s o b t a i n e d w i t h e a c h a s s a y o n 519 s e r u m s a m p l e s BE-EIA . + NS
IMx
S T A T test
Diamedix
Number
+
+
+
193 (37.2) 157 (30.3)
+ /-
+ / -/equi
+ / -/equi
.
.
-
-
-
--
DO
+
.
+ +
equi +
-
equi +
104 (20.0) a 20 11 10 9 4
--
+
-
+
3
-
÷
-
-
2
-
-
+
+
2
+
--
+
+
1
--
+
+
+
1
+
--
+
I 1
DO DO
NS, nonspecific; equi, equivocal; DO, doubtful. aSee T a b l e I I f o r m o r e d e t a i l s o n t h e s e s a m p l e s .
(%)
219
TABLE II C o r r e l a t i o n o f t h e 104 B E - E I A n o n - s p e c i f i c r e s u l t s w i t h t h e t h r e e o t h e r a s s a y s BE-EIA
IMx
S T A T test
Diamedix
Number
NS NS
-
-
+
37 15
NS NS NS NS NS
+ + + _ +
+ + equi
-
12
NS
-
equi
-
3
NS
-
-
equi
3
NS NS NS
+
+
+
-
-
+
equi equi
2 2 2
NS NS
+ +
equi +
equi
I 1
NS NS
-
equi +
+ +
S T A T test
Diamedix
+ + +
9 8 4 3
1 1
N S , n o n - s p e c i f i c ; equi, e q u i v o c a l . TABLE III R e s u l t s o f I F A t e s t i n g o n 89 s e r a BE-EIA
IMx
-
--
NS .
-
.
+
.
.
.
.
.
IFA -
.
Number (%) 18
15 12
.
-
-
+
-
-
11
+
+
+
+
+
11
NS NS --
+ + -
+ + -
+ -
4 4 3
--
2
÷
2 2
-
.
-¢-
.
--
.
+ + equi -
.
-
-
-
+
+
-
-
+
+
-
I
+ +
+
+ +
+ +
-
1 1
--
+
+
+
+
1
-
+
+
1
NS
equi
NS, nonspecific; equi, equivocal
Three serum samples found equivocal by Diamedix were negative by all other means. The 15 BE-EIA NS specimens that were negative when tested with the three other EIAs were also found negative by IFA. On the contrary, among the eight serum samples found NS by BE-EIA and positive by the other EIA methods, four were found positive and four were negative b~ IFA. The performance characteristics of each EIA are detailed in Table IV. Before and after resolution, the percent sensitivities obtained with each assay remain fairly equivalent when compared to each other. On the other hand, BE-EIA and IMx seem more specific than the two other tests. The positive predictive values obtained with
220 TABLE IV Analysis of test performance for the four EIAs Test (supplier)
Sensitivity (%)
Specificity (%)
False result rate
Predictive value
Positive
Negative
Positive
Negative
(%)
(%)
(%t
I%)
I00 98.8 93.5 88.8
99.6 99.6 100 100
98.7 98.1 92.4 88.8
97.8 98.2 98.l 99.0
(A) Before resolution with the Gull IFA procedure (Unresolved Data) BE-EIA (Behringwerke) 99.4 IMx (Abbott) 99.4 STAT test (Whittaker) 100 Diamedix (Diamedix) 100
100 99.1 95.1 91.2
0 0.9 4.9 8.8
0.6 0.6 0 0
(B) After resolution of 89 sera with the Gull IFA procedure (Resolved Data) BE-EIA (Behringwerke) IMx (Abbott) STAT test (Whittaker) Diamedix (Diamedix)
96.9 97.5 97.5 98.8
99.1 98.7 94.2 91.1
0.9 1.3 5.8 8.9
3.1 2.5 2.5 1.2
the BE-EIA and IMx assay are higher than those of the STAT Test and the Diamedix assay, whereas the negative predictive values are very similar for all tests. The IMx test was found to be the easiest to perform, followed by the STAT Test, Diamedix assay and BE-EIA in that order. Diamedix EIA seemed to be the least reproducible method when a few sera were tested repeatedly (Table V). Nine of the 13 serum samples found discrepant when repeatedly tested by Diamedix gave a positive and a negative result, four other sera gave an equivocal result combined with a positive or a negative result on repeat. However, the two BE-EIA discrepant serum specimens were doubtful on one run and, on repeat, one was weakly positive and the other one was negative. Finally, no grey zone has been defined by the manufacturer for the Abbott IMx, and the discrepant serum gave a negative result (A value of 9.2 AU/ml) when first tested, and a weakly positive result (15.7 AU/ml) on repeat. Therefore, a minimal concentration of antibody present in the serum is likely to explain the discrepant results obtained with the BE-EIA and the IMx assay, but not most of those obtained with Diamedix.
Discussion Documentation of CMV serologic status of a transfusion or transplant candidate and that of the blood or tissue donor may have a major impact in the future management of such individuals. Serological testing must therefore be optimal, and TABLE V Evaluation of reproducibility on sera tested repeatedly using the same EIA Test BE-EIA IMx STAT test Diamedix
Number of sera repeatedly tested 11 12 14 29
Discrepancy
Agreement
2 1 0 13
9 11 14 16
I
221
technically easy to perform. Even though numerous reports based on laboratory comparison of various assays have been published, the choice of the most reliable method for donor and recipient screening remains a controverted one (Phipps et al., 1983; Cheeseman et al., 1984; Beckwith et al., 1985; McHugh et al., 1985; Chou et al., 1988; Booth et al., 1989; Leland et al., 1989; Miller et al., 1989; Van Enk et al., 1991; Kraat et al., 1992). Ideally, the chosen assay should be highly sensitive; have a high negative predictive value to avoid CMV transmission to recipients; exhibit minimal equivocal, doubtful or NS results, which would require retesting or exclusion of the serum sample; and be reproducible and easy to perform. According to these criteria, the four EIAs were found very satisfactory insofar as sensitivity and negative predictive values are concerned. However, equivocal results as well as a higher number of false-positive results were observed with the Diamedix assay and the STAT Test. Equivocal results, if not retested, could lead to erroneous donor-recipient serologic matching. Furthermore, the Diamedix EIA showed poor reproducibility, in contrast to the results obtained by Van Enk et al. (1991) on a panel of 76 sera tested. The discrepancies observed among the tests for the positive and negative results cannot be explained by the screening dilution and the initial volume of serum used for each assay. On the contrary, as already mentioned, more false-positive results were obtained with the Diamedix and, to a lesser extent, with the STAT Test assays where the smallest initial volumes of serum were used at the highest dilutions. Technically, even though the IMx assay allows for only 24 specimens to be processed at once, it was found to be the easiest to perform and had the most rapid turnaround time. Regarding the BE-EIA NS results, after resolution by IFA, four of these sera would still have to be considered false-negative. Several of the serum samples that gave BE-EIA hiS results were obtained from immunocompromised patients. It is possible that antibodies directed against cellular components of the antigen mixture contribute for a majority of these so-called NS results with high nonspecific absorbance. Another possible explanation is that much NS activity in EIA could be attributed to reactions between anti-bovine antibodies commonly present in human serum samples and residual calf serum in viral antigens (Disc and Brunell, 1987). The use of an uninfected cell well to control for nonspecific reactivity in the BE-EIA could be seen as an advantage of this method over the other EIAs which do not include such a control antigen. This could explain, at least in part, the false-positive results obtained with the other assays. Therefore, it no longer appears justified to recommend a modification of the BE-EIA criterion of negativity to include a 'NS' category. Interestingly enough, the Abbott IMx performed as well as the BE-EIA without such a cell control antigen. Furthermore, the possibility of low serum dilutions increasing the chances of specific binding inhibition did not appear to be a problem with the Abbott IMx even though the latter used undiluted serum. Last year, Kraat et al. (1992) published a study where the Abbott IMx compared favorably with three other serological methods for the detection of CMV IgG, despite a small number of sera tested. Our work, performed on a larger number of serum samples, tends to support these data. In our hands, the Abbott IMx ranked high in the performance characteristics for a good screening test; it was highly sensitive, had a high negative predictive value, showed good reproducibility and was easy to perform.
222 Incorrect classification of a n o r g a n recipient a n d of its d o n o r as being C M V seropositive or seronegative, based o n a n e r r o n e o u s assay interpretation, can lead to i n a p p r o p r i a t e m a n a g e m e n t of such patients with potentially d r a m a t i c clinical consequences. Therefore, o u r d a t a suggest that all four assays were valuable screening tools for the detection of C M V I g G based o n their high sensitivity a n d high negative predictive value. However, differences were n o t e d in the reproducibility level a n d in the incidence of false-positive, equivocal a n d nonspecific results regarding certain tests in particular.
Acknowledgements L.P. is a clinical research scholar s u p p o r t e d in part by the F o n d s de la Recherche en Sant6 d u Qu6bec.
References Beckwith, D.G., Halstead, D.C., Alpaugh, K., Schweder, A., Blount-Fronefield, D.A. and Toth, K. (1985) Comparison of a latex agglutination test with five other methods for determining the presence of antibody against cytomegalovirus. J. Clin. Microbiol., 21,328-331. Booth, J.C., Kangro H.O., Liu, K.M., El Mohandes, L. and Tryhorn, Y.S. (1989) Discordant results obtained on testing sera from immunocompromised patients for cytomegalovirus IgG by enzymelinked immunosorbent assay and radioimmunoassay. J. Virol. Methods, 26, 77-89. Cheeseman, S.H., Doern, G.V., Ferriani, R.A., Keville, M.W., McGraw, B.R. and Stewart, J.A. (1984) Detection of cytomegalovirus antibody with two commerciallyavailable assays, an indirect hemagglutination test and an enzyme immunosorbent assay. J. Clin. Microbiol., 20, 9-11. Chou, S. and Scott, K.M. (1988) Latex agglutination and enzyme-linked immunosorbent assays for cytomegalovirus serologic screening of transplant donors and recipients. J. Clin. Microbiol., 26, 2116-2119. Dise, T. and Brunell, P.A. (1987) Anti-bovine antibody in human sera as a cause of nonspecificity in enzyme immunoassay. J. Clin. Microbiol., 25, 987-990. Gold, E. and Nankervis, G.A. (1991) Cytomegalovirus. In: A.S. Evans (Ed), Viral Infections of Humans: Epidemiology and Control, 3rd edn., Plenum Medical, New York, pp. 169-189. Griner, P.F., Mayewski, R.J., Mushlin, A.I. and Greenland, P. (1981) Selection and interpretation of diagnositc tests and procedures. Principles and applications. Ann. Intern. Med., 94, 553-600. Ho, M., Suwansirikul, S., Dowling, J.N., Youngblood, L.A. and Armstrong, J.A. (1975) The transplanted kidney as a source of cytomegalovirus infection. N. Engl. J. Med., 293, 1109-I112. Ho, M. (1990) Cytomegalovirus. In: G.L. Mandell, R.G. Douglas Jr. and J.E. Bennett (Eds), Principles and Practice of Infectious Diseases, 3rd Edn, Churchill Livingstone, New York, pp. 1159-1172. Kraat, Y.J., Hendrix, R.M.G., Landini, M.P. and Bruggeman, C.A. (1992) Comparison of four techniques for detection of antibodies to cytomegalovirus. J. Clin. Microbiol., 30, 522-524. Leland, D.S., Barth, K.A., Cunningham, E.B., Jansen, J., Tricot, G.J. and French, M.L.V. (1989) Evaluation of four methods for cytomegalovirus antibody detection for use by a bone marrow transplantation service. J. Clin. Microbiol., 27, 176-178. McHugh, T.M., Casavant, C.H., Wilber, J.C. and Stites, D.P. (1985) Comparison of six methods for the detection of antibody to cytomegalovirus. J. Clin. Microbiol., 22, 1014-1019. Miller, H., McCulloch, B., Landini, M.P. and Rossier, E. (1989) Comparison of immunoblotting with other serological methods and virus isolation for the early detection of primary cytomegalovirus infection in allograft recipients. J. Clin. Microbiol., 27, 2672-2677. Pass, R.F., Stagno, S., Myers, G.J. and Alford, C.A. (1980) Outcome of symptomatic congenital cytomegalovirus infection: results of long-term longitudinal follow-up. Pediatrics, 66, 758-762. Phipps, P.H., Gregoire, L., Rossier, E. and Perry, E. (1983) Comparison of five methods of cytomegalovirus antibody screening of blood donors. J. Clin. Microbiol., 18, 1296-1300. Preiksaitis, J.K., Rosno, S., Grumet, C. and Merigan, T.C. (1983) Infection due to herpesvirusesin cardiac
223 transplant recipients: role of the donor heart and immunosuppressive therapy. J. Infect. Dis., 147, 974-981. Preiksaitis, J.K., Brown, L. and McKenzie, M. (1988) The risk of cytomegalovirus infection in seronegative transfusion recipients not receiving exogenous immunosuppression. J. Infect. Dis., 157, 523-529. Reynolds, D.W., Stagno, S., Hosty, T.S., Tiller, M. and Alford, C.A. Jr (1973) Maternal cytomegalovirus excretion and perinatal infection. N. Engl. J. Med., 289, 1-5. Stagno, S., Reynolds, D.W., Pass, R.F. and Alford, C.A. (1980) Breast milk and the risk of cytomegalovirus infection. N. Engl. J. Med., 302, 1073-1076. Van Enk, R.A., James, K.K. and Thompson, K.D. (1991) Evaluation of three commercial enzyme immunoassays for toxoplasma and cytomegalovirus antibodies. Am. J. Clin. Pathol., 95, 428-434. Yeager, A.S., Grumet, F.C., Hafleigh, E.B., Arvin, A.M., Bradley, J.S. and Prober, C.G. (1981) Prevention of transfusion-acquired cytomegalovirus infections in newborn infants. J. Pediatr., 98, 281-287.
© 1993 ElsevierSciencePublishers B.V. All rights reserved0928-0197/93/$06.00
Clinical and
DIAVIR00027
Diagnostic
Virology Comparison of four enzyme immunoassays for the detection of cytomegalovirus IgG antibodies Louise Pedneault a.b, Louise Robillard b, Pierre Harvey a and Jean Joncas a.b aDepartment of Microbiology and Immunology, Universit~de Montreal, Montreal, Quebec, Canada and bDepartment of Microbiology and Infectious Diseases, H~pital Sainte-Justine, Montreal, Quebec, Canada
(Received8 March 1993;revisionreceived25 May 1993;accepted31 May 1993)
Abstract Four commercial enzyme immunoassays (EIA), namely the Behring Enzygnost EIA (BE-EIA), Abbott IMx, Whittaker CMV STAT Test Kit and Diamedix assay, were evaluated for the detection of CMV IgG. The methods were compared as to sensitivity, specificity, positive and negative predictive values, global agreement, ease of performance and, for a small number of specimens, reproducibility. Discordant results were resolved by using the Gull CMV indirect fluorescent antibody (IFA) method. Our data suggest that all four assays were valuable screening tools for the detection of CMV IgG based on their high sensitivity and high negative predictive value. However, differences were noted in the reproducibility level and in the incidence of false-positive, equivocal and nonspccific results regarding certain tests in particular. In our hands, the Abbott IMx and the BE-EIA ranked high in the performance characteristics for a good screening test, yet the Abbott IMx offers the added advantages of being the easiest to perform and having the most rapid turnaround time. Key words: Cytomegalovirus;IgG antibody;Enzymeimmunoassay.
Introduction Cytomegalovirus (CMV) infects approximately 50% of adults in developed countries and nearly 100% of the adult population in developing countries (Gold and Nankervis, 1991). In the normal host, CMV infection is usually mild and asymptomatic or, occasionnally, manifests itself as a mononucleosis syndrome. In the immunosuppressed host, primary or reactivated CMV infection may be asymptomatic, but may also give rise to a mononucleosis-like syndrome more often than in a healthy individual, and to complications such as fever, pneumonitis, chorioretinitis, hepatitis Correspondence to: L. Pedneault,D6partement de Microbiologie,H6pital Sainte-Justine,3175 Chemin C6te Sainte-Catherine,Montr6al,Qu6bec,Canada H3T 1CS.Fax:(514) 345-4818;Tel:(514) 345-4643. Note: This work was presentedin part at the 8th AnnualClinicalVirologySymposium,Clearwater,FL, USA, May 1992. Abbreviations: CMV, cytomegalovirus,IgG, immunoglobulinG, EIA, enzyme immunoassay,IFA, indirect fluorescenceassay.
216 or gastrointestinal lesions (Ho, 1990). Furthermore, in utero infection may result in various sequelae in the newborn such as mental retardation, chorioretinitis, hearing loss and neurologic problems (Pass et al., 1980). CMV can be transmitted in several ways; an infant may acquire CMV transplacentally, during birth (Reynolds et al., 1973) or through ingestion of infected breast milk (Stagno et al., 1980). Seronegative individuals of all ages may acquire CMV through infected blood products (Preiksaitis et al., 1988; Yeager et al., 1981) or donated organs (Ho et al., 1975; Preiksaitis et al., 1983), or through direct contact with the secretions of an infected individual (Gold and Nankervis, 1991; Ho, 1990). The selection and use of seronegative blood products or of donated organs from seronegative individuals to immunocompromised patients remains a vital consideration in patient management. The decision to use such products is usually based on the CMV serologic status of the donor. For that matter, the choice of an adequate serologic test for the detection of CMV antibodies becomes crucial. In this study, we present the results from 519 serum specimens analyzed for the presence of CMV IgG antibodies by four EIA procedures. Discordant results were resolved by using the Gull CMV indirect fluorescent antibody (IFA) method.
Materials and Methods
Specimens A total of 519 sera were evaluated for the presence of CMV IgG antibodies by the various EIA tests described below. Sera were randomly selected from individuals for whom a CMV serology had been requested. Specimens were stored at -20°C and underwent on average two freeze-thaw cycles.
Comparison testing procedures BE-EIA (Behringwerke, Marburg, Germany) is based on the use of two wells per serum sample at a time, one of them being coated with viral antigen and the other with control antigen. Virus antigen and control antigen are obtained from human cell cultures infected with CMV and uninfected, respectively, and inactivated before coating. The test was performed as part of the routine workload by a laboratory technician at Sainte-Justine Hospital according to the manufacturer's instructions. BE-EIA requires an initial serum volume of 20/al for a final dilution of the sample of 1:44. A specimen is considered positive if the A viral antigen minus A control antigen (AA) is ~>0.2, negative if the AA is <0.2, and doubtful if the AA value lies between 0.2 and 0.3. The latter needs to be repeated. If the result remains doubtful after the retest, it should be confirmed using another approved method. In cases where the background seemed particularly high (AA<0.2 when the control antigen value is >0.2), we decided to consider the result 'nonspecific' (NS) (this is assessed as negative by Behring). Comparison testing with the IMx CMV IgG microparticle based-EIA (Abbott Laboratories, Abbott Park, IL), the CMV STAT Test Kit (Whittaker Bioproducts, Walkersville, MD) and Diamedix (Diamedix Corporation, Miami, FL) EIAs for detection of CMV IgG antibodies was performed blindly by a laboratory technician at Sainte-Justine Hospital according to the manufacturer's instructions. The STAT Test and Diamedix assay are standard EIAs using wells coated with inactivated CMV as a source of antigen. However, to perform the IMx microparticle based-EIA
217
(MEIA), the patient's serum is added to a microparticle coated with inactivated CMV and, after incubation, an aliquot of the antigen-antibody complex formed is transferred to a glass fiber matrix. The microparticles will bind irreversibly to this matrix. An anti-human IgG conjugated to alkaline phosphatase is then added, followed by the appropriate substrate to allow detection of CMV IgG antibodies if present. No control antigen wells or microparticles are used in any of the three comparison tests, the source of CMV antigen being partially purified in all cases. Initial serum volumes of 150 ~1, 10 ~1 and 5 #1 are necessary to perform the IMx, STAT Test, and Diamedix EIAs, respectively. These test samples are used undiluted in the IMx assay, but must be brought to a final dilution of 1:21 (STAT Test) or 1:41 (Diamedix) for the two other assays. The interpretation of the results for the various EIAs was as follows: (1) IMx: a sample is considered positive if it contains at least 15 AU (for antibody units)/ml of serum ( >~ 15 AU/ml), and negative if the reading is < 15 AU/ml. (2) STAT Test: a sample is considered positive if its predicted index value (PIV) is ~> 1.0, equivocal if the PIV lies between 0.80 and 0.99, and negative if the PIV is ~<0.79. (3) Diamedix: a sample is considered positive if it contains more than 23 EU (for EIA units)/ml of serum (> 23 EU/ml), negative if it contains less than 18 EU/ml, and equivocal if the reading is between 18 and 23 EU/ml.
Resolution of discordant results We chose IFA as a reference method because it allows to visualize and discriminate among various reactivities by evaluating the cellular and subcellular patterns of fluorescence, and because we considered that an alternate method might be useful to resolve some of the discrepant EIA results. Thirty-eight sera that gave a discordant result in only one EIA, the 8 sera that were found NS with BE-EIA and positive with the three other EIAs, 15 randomly chosen sera that were NS when tested by BE-EIA and negative with the three other EIAs, and four other randomly selected sera were evaluated by IFA (Gull Laboratories, Salt Lake City, Utah) according to the manufacturer's instructions. Furthermore, 12 serum specimens positive by all 4 EIAs and 12 serum specimens negative by all 4 EIAs were used as controls. At the screening dilution (1:10), a greenish-yellow fluorescence of about 10% or more of the infected cells in each field, with a whole cell or nuclear pattern, is considered positive. Absence of fluorescence is read as negative.
Analysis of results All testing was carried out without knowledge of the results obtained with the other EIAs. For a given serum specimen to be considered positive for the presence of CMV IgG antibody, concordance of at least three of the four EIA methods or, in cases where IFA was performed, confirmation of positivity by IFA was required. Conversely, for a sample to be considered negative, concordance of three or more EIAs or, when relevant, confirmation of negativity by IFA must be obtained. Equivocal, non-specific and doubtful results for each method, as well as the five serum specimens which gave positive results by two EIAs and negative results with the two other assays, were excluded from the preliminary analysis table (Table IVA., unresolved data). Therefore, only the serum samples considered positive or negative by concordance of at least three EIAs were included in this table (Table IVA., unresolved data). The IFA results obtained on the 89 serum samples tested by this
218
method (Table III) were included in the final analysis of the data (Table IVB., resolved data). Sensitivity, specificity, positive and negative predictive values, incidence of false-positive and -negative results, and global agreement were calculated using the formulas recommended by Griner et al. (1981).
Results
A total of 519 sera were evaluated by four different EIAs for the detection of CMV IgG antibodies (Table I). According to our seropositivity and seronegativity criteria, 159 sera (31%) were CMV IgG positive and 226 (44%) were found negative, for a complete agreement of 75%. However, 104 (20%) sera gave NS results when tested with the BE-EIA. Excluding these 104 sera, the complete agreement for the four assays becomes 385/415 (93%). The Behring Enzygnost enzyme immunoassay (BE-EIA) for the detection of CMV IgG antibodies frequently gives NS results, thereby rendering the interpretation of the test difficult. The results for the 104 samples yielding NS results by BE-EIA are detailed in Table II. Thirty-seven (36%) of these specimens were found negative with the three other EIAs. If we presume that a NS result is, in fact, negative, as suggested by the manufacturer, 31 more serum samples would therefore be considered negative, for a total of 68/104 (65%), based on the criteria previously established for seronegativity where concordance of at least three EIAs is necessary to categorize a specimen as positive or negative. Furthermore, eight (8%) of the NS specimens were positive by all three other EIAs. We decided that, to resolve some of the discrepant results obtained when comparing EIA methods among themselves, we should use a different assay, in this case, IFA. Eighty-nine serum samples were thus submitted to IFA testing (Table III). After resolution, there were still 2, 11, and 18 specimens previously considered false-positive by IMx, STAT Test and Diamedix, respectively, that remained classified as such. TABLE I C M V s e r o r e a c t i v i t y r e s u l t s o b t a i n e d w i t h e a c h a s s a y o n 519 s e r u m s a m p l e s BE-EIA . + NS
IMx
S T A T test
Diamedix
Number
+
+
+
193 (37.2) 157 (30.3)
+ /-
+ / -/equi
+ / -/equi
.
.
-
-
-
--
DO
+
.
+ +
equi +
-
equi +
104 (20.0) a 20 11 10 9 4
--
+
-
+
3
-
÷
-
-
2
-
-
+
+
2
+
--
+
+
1
--
+
+
+
1
+
--
+
I 1
DO DO
NS, nonspecific; equi, equivocal; DO, doubtful. aSee T a b l e I I f o r m o r e d e t a i l s o n t h e s e s a m p l e s .
(%)
219
TABLE II C o r r e l a t i o n o f t h e 104 B E - E I A n o n - s p e c i f i c r e s u l t s w i t h t h e t h r e e o t h e r a s s a y s BE-EIA
IMx
S T A T test
Diamedix
Number
NS NS
-
-
+
37 15
NS NS NS NS NS
+ + + _ +
+ + equi
-
12
NS
-
equi
-
3
NS
-
-
equi
3
NS NS NS
+
+
+
-
-
+
equi equi
2 2 2
NS NS
+ +
equi +
equi
I 1
NS NS
-
equi +
+ +
S T A T test
Diamedix
+ + +
9 8 4 3
1 1
N S , n o n - s p e c i f i c ; equi, e q u i v o c a l . TABLE III R e s u l t s o f I F A t e s t i n g o n 89 s e r a BE-EIA
IMx
-
--
NS .
-
.
+
.
.
.
.
.
IFA -
.
Number (%) 18
15 12
.
-
-
+
-
-
11
+
+
+
+
+
11
NS NS --
+ + -
+ + -
+ -
4 4 3
--
2
÷
2 2
-
.
-¢-
.
--
.
+ + equi -
.
-
-
-
+
+
-
-
+
+
-
I
+ +
+
+ +
+ +
-
1 1
--
+
+
+
+
1
-
+
+
1
NS
equi
NS, nonspecific; equi, equivocal
Three serum samples found equivocal by Diamedix were negative by all other means. The 15 BE-EIA NS specimens that were negative when tested with the three other EIAs were also found negative by IFA. On the contrary, among the eight serum samples found NS by BE-EIA and positive by the other EIA methods, four were found positive and four were negative b~ IFA. The performance characteristics of each EIA are detailed in Table IV. Before and after resolution, the percent sensitivities obtained with each assay remain fairly equivalent when compared to each other. On the other hand, BE-EIA and IMx seem more specific than the two other tests. The positive predictive values obtained with
220 TABLE IV Analysis of test performance for the four EIAs Test (supplier)
Sensitivity (%)
Specificity (%)
False result rate
Predictive value
Positive
Negative
Positive
Negative
(%)
(%)
(%t
I%)
I00 98.8 93.5 88.8
99.6 99.6 100 100
98.7 98.1 92.4 88.8
97.8 98.2 98.l 99.0
(A) Before resolution with the Gull IFA procedure (Unresolved Data) BE-EIA (Behringwerke) 99.4 IMx (Abbott) 99.4 STAT test (Whittaker) 100 Diamedix (Diamedix) 100
100 99.1 95.1 91.2
0 0.9 4.9 8.8
0.6 0.6 0 0
(B) After resolution of 89 sera with the Gull IFA procedure (Resolved Data) BE-EIA (Behringwerke) IMx (Abbott) STAT test (Whittaker) Diamedix (Diamedix)
96.9 97.5 97.5 98.8
99.1 98.7 94.2 91.1
0.9 1.3 5.8 8.9
3.1 2.5 2.5 1.2
the BE-EIA and IMx assay are higher than those of the STAT Test and the Diamedix assay, whereas the negative predictive values are very similar for all tests. The IMx test was found to be the easiest to perform, followed by the STAT Test, Diamedix assay and BE-EIA in that order. Diamedix EIA seemed to be the least reproducible method when a few sera were tested repeatedly (Table V). Nine of the 13 serum samples found discrepant when repeatedly tested by Diamedix gave a positive and a negative result, four other sera gave an equivocal result combined with a positive or a negative result on repeat. However, the two BE-EIA discrepant serum specimens were doubtful on one run and, on repeat, one was weakly positive and the other one was negative. Finally, no grey zone has been defined by the manufacturer for the Abbott IMx, and the discrepant serum gave a negative result (A value of 9.2 AU/ml) when first tested, and a weakly positive result (15.7 AU/ml) on repeat. Therefore, a minimal concentration of antibody present in the serum is likely to explain the discrepant results obtained with the BE-EIA and the IMx assay, but not most of those obtained with Diamedix.
Discussion Documentation of CMV serologic status of a transfusion or transplant candidate and that of the blood or tissue donor may have a major impact in the future management of such individuals. Serological testing must therefore be optimal, and TABLE V Evaluation of reproducibility on sera tested repeatedly using the same EIA Test BE-EIA IMx STAT test Diamedix
Number of sera repeatedly tested 11 12 14 29
Discrepancy
Agreement
2 1 0 13
9 11 14 16
I
221
technically easy to perform. Even though numerous reports based on laboratory comparison of various assays have been published, the choice of the most reliable method for donor and recipient screening remains a controverted one (Phipps et al., 1983; Cheeseman et al., 1984; Beckwith et al., 1985; McHugh et al., 1985; Chou et al., 1988; Booth et al., 1989; Leland et al., 1989; Miller et al., 1989; Van Enk et al., 1991; Kraat et al., 1992). Ideally, the chosen assay should be highly sensitive; have a high negative predictive value to avoid CMV transmission to recipients; exhibit minimal equivocal, doubtful or NS results, which would require retesting or exclusion of the serum sample; and be reproducible and easy to perform. According to these criteria, the four EIAs were found very satisfactory insofar as sensitivity and negative predictive values are concerned. However, equivocal results as well as a higher number of false-positive results were observed with the Diamedix assay and the STAT Test. Equivocal results, if not retested, could lead to erroneous donor-recipient serologic matching. Furthermore, the Diamedix EIA showed poor reproducibility, in contrast to the results obtained by Van Enk et al. (1991) on a panel of 76 sera tested. The discrepancies observed among the tests for the positive and negative results cannot be explained by the screening dilution and the initial volume of serum used for each assay. On the contrary, as already mentioned, more false-positive results were obtained with the Diamedix and, to a lesser extent, with the STAT Test assays where the smallest initial volumes of serum were used at the highest dilutions. Technically, even though the IMx assay allows for only 24 specimens to be processed at once, it was found to be the easiest to perform and had the most rapid turnaround time. Regarding the BE-EIA NS results, after resolution by IFA, four of these sera would still have to be considered false-negative. Several of the serum samples that gave BE-EIA hiS results were obtained from immunocompromised patients. It is possible that antibodies directed against cellular components of the antigen mixture contribute for a majority of these so-called NS results with high nonspecific absorbance. Another possible explanation is that much NS activity in EIA could be attributed to reactions between anti-bovine antibodies commonly present in human serum samples and residual calf serum in viral antigens (Disc and Brunell, 1987). The use of an uninfected cell well to control for nonspecific reactivity in the BE-EIA could be seen as an advantage of this method over the other EIAs which do not include such a control antigen. This could explain, at least in part, the false-positive results obtained with the other assays. Therefore, it no longer appears justified to recommend a modification of the BE-EIA criterion of negativity to include a 'NS' category. Interestingly enough, the Abbott IMx performed as well as the BE-EIA without such a cell control antigen. Furthermore, the possibility of low serum dilutions increasing the chances of specific binding inhibition did not appear to be a problem with the Abbott IMx even though the latter used undiluted serum. Last year, Kraat et al. (1992) published a study where the Abbott IMx compared favorably with three other serological methods for the detection of CMV IgG, despite a small number of sera tested. Our work, performed on a larger number of serum samples, tends to support these data. In our hands, the Abbott IMx ranked high in the performance characteristics for a good screening test; it was highly sensitive, had a high negative predictive value, showed good reproducibility and was easy to perform.
222 Incorrect classification of a n o r g a n recipient a n d of its d o n o r as being C M V seropositive or seronegative, based o n a n e r r o n e o u s assay interpretation, can lead to i n a p p r o p r i a t e m a n a g e m e n t of such patients with potentially d r a m a t i c clinical consequences. Therefore, o u r d a t a suggest that all four assays were valuable screening tools for the detection of C M V I g G based o n their high sensitivity a n d high negative predictive value. However, differences were n o t e d in the reproducibility level a n d in the incidence of false-positive, equivocal a n d nonspecific results regarding certain tests in particular.
Acknowledgements L.P. is a clinical research scholar s u p p o r t e d in part by the F o n d s de la Recherche en Sant6 d u Qu6bec.
References Beckwith, D.G., Halstead, D.C., Alpaugh, K., Schweder, A., Blount-Fronefield, D.A. and Toth, K. (1985) Comparison of a latex agglutination test with five other methods for determining the presence of antibody against cytomegalovirus. J. Clin. Microbiol., 21,328-331. Booth, J.C., Kangro H.O., Liu, K.M., El Mohandes, L. and Tryhorn, Y.S. (1989) Discordant results obtained on testing sera from immunocompromised patients for cytomegalovirus IgG by enzymelinked immunosorbent assay and radioimmunoassay. J. Virol. Methods, 26, 77-89. Cheeseman, S.H., Doern, G.V., Ferriani, R.A., Keville, M.W., McGraw, B.R. and Stewart, J.A. (1984) Detection of cytomegalovirus antibody with two commerciallyavailable assays, an indirect hemagglutination test and an enzyme immunosorbent assay. J. Clin. Microbiol., 20, 9-11. Chou, S. and Scott, K.M. (1988) Latex agglutination and enzyme-linked immunosorbent assays for cytomegalovirus serologic screening of transplant donors and recipients. J. Clin. Microbiol., 26, 2116-2119. Dise, T. and Brunell, P.A. (1987) Anti-bovine antibody in human sera as a cause of nonspecificity in enzyme immunoassay. J. Clin. Microbiol., 25, 987-990. Gold, E. and Nankervis, G.A. (1991) Cytomegalovirus. In: A.S. Evans (Ed), Viral Infections of Humans: Epidemiology and Control, 3rd edn., Plenum Medical, New York, pp. 169-189. Griner, P.F., Mayewski, R.J., Mushlin, A.I. and Greenland, P. (1981) Selection and interpretation of diagnositc tests and procedures. Principles and applications. Ann. Intern. Med., 94, 553-600. Ho, M., Suwansirikul, S., Dowling, J.N., Youngblood, L.A. and Armstrong, J.A. (1975) The transplanted kidney as a source of cytomegalovirus infection. N. Engl. J. Med., 293, 1109-I112. Ho, M. (1990) Cytomegalovirus. In: G.L. Mandell, R.G. Douglas Jr. and J.E. Bennett (Eds), Principles and Practice of Infectious Diseases, 3rd Edn, Churchill Livingstone, New York, pp. 1159-1172. Kraat, Y.J., Hendrix, R.M.G., Landini, M.P. and Bruggeman, C.A. (1992) Comparison of four techniques for detection of antibodies to cytomegalovirus. J. Clin. Microbiol., 30, 522-524. Leland, D.S., Barth, K.A., Cunningham, E.B., Jansen, J., Tricot, G.J. and French, M.L.V. (1989) Evaluation of four methods for cytomegalovirus antibody detection for use by a bone marrow transplantation service. J. Clin. Microbiol., 27, 176-178. McHugh, T.M., Casavant, C.H., Wilber, J.C. and Stites, D.P. (1985) Comparison of six methods for the detection of antibody to cytomegalovirus. J. Clin. Microbiol., 22, 1014-1019. Miller, H., McCulloch, B., Landini, M.P. and Rossier, E. (1989) Comparison of immunoblotting with other serological methods and virus isolation for the early detection of primary cytomegalovirus infection in allograft recipients. J. Clin. Microbiol., 27, 2672-2677. Pass, R.F., Stagno, S., Myers, G.J. and Alford, C.A. (1980) Outcome of symptomatic congenital cytomegalovirus infection: results of long-term longitudinal follow-up. Pediatrics, 66, 758-762. Phipps, P.H., Gregoire, L., Rossier, E. and Perry, E. (1983) Comparison of five methods of cytomegalovirus antibody screening of blood donors. J. Clin. Microbiol., 18, 1296-1300. Preiksaitis, J.K., Rosno, S., Grumet, C. and Merigan, T.C. (1983) Infection due to herpesvirusesin cardiac
223 transplant recipients: role of the donor heart and immunosuppressive therapy. J. Infect. Dis., 147, 974-981. Preiksaitis, J.K., Brown, L. and McKenzie, M. (1988) The risk of cytomegalovirus infection in seronegative transfusion recipients not receiving exogenous immunosuppression. J. Infect. Dis., 157, 523-529. Reynolds, D.W., Stagno, S., Hosty, T.S., Tiller, M. and Alford, C.A. Jr (1973) Maternal cytomegalovirus excretion and perinatal infection. N. Engl. J. Med., 289, 1-5. Stagno, S., Reynolds, D.W., Pass, R.F. and Alford, C.A. (1980) Breast milk and the risk of cytomegalovirus infection. N. Engl. J. Med., 302, 1073-1076. Van Enk, R.A., James, K.K. and Thompson, K.D. (1991) Evaluation of three commercial enzyme immunoassays for toxoplasma and cytomegalovirus antibodies. Am. J. Clin. Pathol., 95, 428-434. Yeager, A.S., Grumet, F.C., Hafleigh, E.B., Arvin, A.M., Bradley, J.S. and Prober, C.G. (1981) Prevention of transfusion-acquired cytomegalovirus infections in newborn infants. J. Pediatr., 98, 281-287.