Unsatisfactory Specificities and Sensitivities of Six Enzyme ...

JOURNAL OF CLINICAL MICROBIOLOGY, Sept. 1989, p. 2067-2072

Vol. 27, No. 9

0095-1137/89/092067-06$02.00/0 Copyright C 1989, American Society for Microbiology

Unsatisfactory Specificities and Sensitivities of Six Enzyme Immunoassays for Antibodies to Hepatitis B Core Antigen GREGOR CASPARI,' HANS-JOACHIM BEYER,' GABI ELBERT,2 KLAUS KOERNER,3 PETER MUSS,4 FRIEDRICH WILHELM SCHUNTERS ANGELA UY,6 WOLFRAM GERLICH,6* REINER THOMSSEN,6 AND HEINZ SCHMITT' Blood Transfusion Service of German Red Cross, Institute Springe, Niedersachsen,' Blood Transfusion Service of German Red Cross, Institute Bad Kreuznach, Rheinland-Pfalz/Saarland,2 Blood Transfusion Service of German Red Cross, Institute Ulm, Baden-Wurttemberg? Blood Transfusion Service of German Red Cross, Institute Breitscheid, Nordrhein-Westfalen,4 Blood Transfusion Service of German Red Cross, Institute Rotenburg, Niedersachsens and Department of Medical Microbiology, National Reference Center for Viral Hepatitis, University of Gottingen, D-3400 Gottingen,6 Federal Republic of Germany Received 13 March 1989/Accepted 7 June 1989

To examine the consistency and comparability of anti-hepatitis B core antigen (anti-HBcAg) assays, four blood donation centers of the Red Cross in the Federal Republic of Germany tested 4,080 unselected blood donors with six different tests in parallel. Confirmation testing of reactive samples was done in the National Reference Center for Viral Hepatitis. Depending on the test kit used, 4.1 to 9.9% of serum samples were initially positive and 2.9 to 7.5% were repeatedly positive. Sixteen percent of serum samples were positive in at least one test but only three percent were positive in ail six tests. Statistical analysis of frequency distribution of optical densities for each test suggested that there should be a correction of the cutoff values. This reduced the number of false-positive results by half, but a significant proportion of discrepant results could not be resolved. The lack of specificity and consistency requires cautious interpretation of isolated anti-HBcAg results in clinical specimens. Screening of predominantly anti-HBcAg-negative populations (e.g., blood donors) by the current anti-HBcAg test kits will almost necessarily give unsatisfactory results.

The core particle of the hepatitis B virus (HBV) is an extremely potent B-cell antigen (4). Persons with acute or chronic HBV infection have very high titers of antibody to HBV core antigen (anti-HBcAg). After resolution of transient HBV infection, anti-HBcAg usually persists longer than other HBV antibodies, such as those against the HBV surface or e antigen (anti-HBsAg or anti-HBeAg). Thus, it is the most universal HBV marker for diagnostic and epidemiologic purposes. Several studies suggest that anti-HBcAg may be an indirect serological marker for an elevated risk to transmit an unknown hepatitis virus(es) by blood donation (3, 5). As a consequence, compulsory screening of blood donations has been introduced in the United States and France and is under consideration in several other European countries. In the current standard technique to assay antiHBcAg, the sample antibody inhibits an enzyme-labeled anti-HBcAg from binding to an immobilized HBcAg. Preliminary observations of the authors suggested that this type of assay generates a significant rate of irreproducible or nonspecific results. To identify the magnitude of this problem, four blood donation centers of the Red Cross in the Federal Republic of Germany tested a large number of unselected blood donors for anti-HBcAg by using six different commercially produced test kits in parallel. All reactive samples were subjected to confirmation testing in the National Reference Center for Viral Hepatitis. Criteria were developed to identify both false-positive and false-negative results. The distribution of test results was analyzed, and the abilities of the test kits to differentiate between positive, borderline, and negative results were evaluated.

*

MATERIALS AND METHODS Samples. The Blood Transfusion Services of the Red Cross in Niedersachsen, Rheinland-Pfalz/Saarland, Nordrhein-Westfalen, and Baden-Wurttemberg collected an additional vial of about 5 to 10 ml of blood per donor (for a total of 4,080 samples) during routine blood collections in March and April 1987. No selection was applied other than the normal admission criteria for blood donation. The serum was either kept at 4°C and tested within 3 days or portioned and kept frozen until testing. Test kits. Anti-HBcAg enzyme immunoassays were supplied by the following producers: Abbott Laboratories (North Chicago, Ill.), Behringwerke AG (Marburg, Federal Republic of Germany), Du Pont/Medical Products (Wilmington, Del.), Organon Teknika (Oss, The Netherlands), Viramed GmbH (Martinsried, Federal Republic of Germany), and Wellcome Diagnostics (Dartford, England). One of the producers participated with two tests; another did not provide test kits for the initial screening but did provide kits for additional confirmation in the National Reference Center for Viral Hepatitis in Gottingen. Since not all producers agreed to public identification, results in this report are given in code: A to F for the six tests used for initial screening and G and H for the tests used for additional confirmation of positive samples. Six of the commercial tests were inhibition assays using peroxidase-labeled anti-HBcAg; the seventh was a direct immunoglobulin G-binding assay using peroxidase-labeled anti-immunoglobulin G. Two of the inhibition assays (B and C) were one-step competitive assays using monoclonal antibodies. All tests used recombinant HBcAg from Escherichia coli at the solid phase. Cutoif and results. As high absorbance values in competitive inhibition enzyme immunoassays correspond to nega-

Corresponding author. 2067

2068

CASPARI ET AL.

J. CLIN. MICROBIOL.

TABLE 1. Rate of initially and repeatedly anti-HBcAg-positive samples from 4,080 blood donors as detected by enzyme immunoassays A through F' % of samples with indicated result (reproducibility) for test:

Test result

Initially positive (reproducibility)

Repeatedly positive

A

B

C

D

E

F

7.7 (86) 6.6

9.4 (80) 7.5

9.9 (67) 6.6

5.4 (68) 3.7

4.1 (71) 2.9

5.6 (85) 4.8

a Only valid results were considered (see text). b See Materials and Methods.

tive samples, the results could be standardized as a percentage of the absorbance values of the negative controls. Cutoff values were calculated as suggested by the producers. In three tests (A, B, and C), they were defined as the percentage of absorbance of the negative control serum (50% in two tests and 25% in the other). In test F, the cutoff was defined as the arithmetic mean of absorbance of the positive (P) and negative (N) controls: (N + P)/2 (corresponds to 51 to 60% of N), and in test E, the cutoff was defined as N x 0.33 + P (corresponds to 34 to 43%). Only for test E (the immunoglobulin G-binding assay) did results have to be expressed as a multiple or fraction of the cutoff. Controls. Tests required between two and four negative controls and two positive controls. One test required a blank. User instructions. User instructions were checked for answers to the following questions. (i) Quality control. Do the test instructions indicate for which positive and negative control results the tests are no longer to be evaluated? Are these criteria unequivocal, sufficient, and understandable? (ài) Grey zone. Do the user instructions state for which test result a single serum sample must be retested? (iii) Definitive test results. In case initial testing and retesting are contradictory or retesting gives (again) a grey-zone result, how is a final decision to be reached? Testing procedure. Test producers were invited to instruct and train laboratory staff if they considered it necessary. All tests were performed between March and June 1987. User instructions were followed carefully. Results were evaluated only if quality-control criteria of user instructions were met. All serum samples which had a positive or borderline result according to the user instructions were retested once. Because some serum samples were unavailable for retesting, a reproducibility rate was calculated as the number of samples reactive in repeated testing divided by the number of samples actually retested. The repeatedly reactive rate was then calculated by multiplying the initially reactive rate with the reproducibility rate. Reference testing. All serum samples positive in at least one test and an additional number of serum samples negative in all tests were sent to the National Reference Center for Viral Hepatitis, Gottingen, for confirmation. Two tests were used: a blocking assay (test H) developed in this laboratory (in duplicate; test described in reference 1) and an additional commercial assay (test G). In case of contradictory results, both tests were repeated once. If both confirmatory tests were positive, serum samples were considered to be positive. When the National Reference Center test was positive and the commercial test was negative serum samples were designated +D, and when the opposite was the case serum samples were designated -D. Serum samples negative in both tests were considered negative.

RESULTS User instructions, quality control, and evaluation of results. Besides explaining to the user how to perform the enzyme immunoassay, the user instructions should also state which results, depending on the control sera, may be accepted and which results must not. This internal quality control was missing in test A. It allowed considerable variations between different microplates in tests B, C, D, and F. Even then, some test results had to be eliminated from the evaluation because quality-control criteria of the user instructions were not met. For test B, 2,984 results were valid; for test C, 3,850 results were valid; for test D, 4,080 results were valid; for test E, 3,172 results were valid; and for test F, 2,982 results were valid. Possible reasons for elimination of test results besides mistakes in performing the tests are insufficiently stable control sera, incorrect ranges of desired values for the controls stated by the producers, insufficient test procedures or descriptions of these, and within-lot variances of test kits. For test D, no results had to be excluded, in spite of substantial variations between microplates, because of its lenient quality-control criteria. Only one user instruction required retesting of all initially positive results, three instructions defined arbitrary and partly insufficient grey zones, and two instructions did not require retesting of initial results at all. In this study, all positive and grey-zone samples were retested in order to evaluate the reproducibility of these results. In case of contradictions between the initial result and the retest result, only one user instruction provided a criterion for deciding whether the sample should be defined positive or negative, another instruction was confusing on this point, and the other four instructions left the choice up to the user. Therefore, for most tables in this paper only the initial results were taken into account. Positive results per test. Table 1 compiles the largely varying rate of initially positive results and the limited reproducibility of the tests. Test C, with the largest number of positive results, found more than twice as many serum samples anti-HBcAg positive as did test E, with the lowest number of positive results. Tests A and F had the best reproducibility, test B had a slightly lower reproducibility, and tests C, D, and E had the worst reproducibilities. Positive results per serum sample. Regarding these differences, the question of how comparable the tests are for one individual serum sample arose. The left side of Table 2 gives the number of serum samples for which none or up to six positive test results could be obtained by the six different tests used. To avoid any possible bias the table includes only those 1,838 serum samples for which all six tests gave valid results (according to user instructions). Most of the other sera had five valid test results.

VOL. 27, 1989

ASSESSMENT OF SIX ASSAYS FOR ANTIBODIES TO HBcAg

TABLE 2. Number of serum samples with concordant negative, concordant positive, and discrepant anti-HBcAg results (1 c xa c 5) depending on the cutoff value used Original cutoff X

0 1 2 3 4 5 6

% of %o No of %of No total %of No. of total %of tot samples positive samples samples positive samples

1,548 141 38 25 14 10 62

No. positive in at least one test a

Readjusted cutoff

290

splssamples 84.2 48.6 7.7 13.1 2.1 8.6 1.3 4.8 0.8 3.5 0.5 21.4 3.4 100.0

15.8

samples

1,724 35 7 3 3 10 56

30.7 6.2 2.6 2.6 8.8 49.1

93.8 1.9 0.4 0.2 0.2 0.5 3.0

114

100.0

6.2

x, Number of positive test results per serum sample.

Frequency distribution of optical densities. The next step of our evaluation took into account the degree of positivity (or negativity) of the test results. Figure 1 shows the frequency distribution of the absorbance values in the competitive inhibition assays standardized as percentages of the absorbance values of the negative control sera. Strongly positive samples had absorbance values of