Utility of quantitative enzyme immunoassay reactivity for predicting human immunodeficiency virus seropositivity in low and high prevalence populations

JOURNAL OFCLINICALMICROBIOLOGY,Jan. 1994,p.220-223 0095-1137/94/$04.00+0

Copyright X 1994, American SocietyforMicrobiology

Utility of

Quantitative

Enzyme Immunoassay

Reactivity for

Predicting Human

Immunodeficiency Virus

Seropositivity

in Low-

and

High-Prevalence

Populations

XINYUEHOU,1 PAULA L. BREESE,1 ANDJOHN M. DOUGLAS,

JR.l12*

Denver DiseaseControl

Service'

and Division of Infectious Diseases, Department ofMedicine,

University

of

ColoradoHealth Sciences

Center,2

Denver, Colorado80204

Received 3June 1993/Returned formodification 19 August 1993/Accepted 19 October 1993

To assesstheutility of quantitativeenzymeimmunoassay (EIA)

reactivity

forpredicting human immuno-deficiency virusseropositivity,weevaluated22,823serumsamples from homo- and bisexualmen,heterosexual

intravenous drugusers,and otherheterosexuals with initial screening byEIA, retesting of reactive samples in

duplicate, and confirmatory Western blot (immunoblot) testing. Quantitative EIA

reactivity

wasdetermined by

a meanof theopticaldensityratio of the threeassaysperformed for each reactive specimen. A total of1,773 samples (7.8%) were

repeatedly

reactive, and 1,747 (7.7%) were confirmed Western blot positive. All 26

ETA-reactive-Westernblot-negative samples had low-level EIAreactivity(ratio <2.2), whilemost(86%) of the Westernblot-positive samples hadhigh-level

reactivity

(ratio, >3.0). The positivepredictive value for samples withmoderate-to-high-level ETA

reactivity

(ratio, >2.2)was100%o for all riskgroups. These resultssupport

the value ofquantitative ETA

reactivity

inpredicting human immunodeficiency virus seropositivit andsuggest

thatconfirmatorytesting ofspecimens with high-levelreactivityis notnecessaryin all situations.

Enzyme immunoassay (EIA) kits for detection of human immunodeficiency virus type 1 (HIV-1) antibody, first li-censedby the U.S. Food and Drug Administration in 1985,

arethemostwidely used HIV-1 tests, and eightassays are

nowFood andDrug Administration approved forusein the United States (11). Because the EIA format is designedto provide optimal sensitivity, use of a supplemental test to confirmHIVseropositivity has been highly recommendedto enhancespecificity and is widely practiced (4, 5, 10, 11). The algorithm generally used is an initial EIA screening test, repeated in duplicate if initially reactive, followed by a confirmatory test, usually a Western blot (immunoblot). Whilethisapproach has been extremely specific and has had excellent predictive value in clinical practice (2, 17), it has severaldrawbacks, includingavariableturnaround time(1),

and problems with the Western blot procedure, which in-clude inconsistent interpretive criteria (6), subjectivity in interpretation of results, and highcosts(19).

Although EIA results are generally interpreted

qualita-tivelyaseitherreactive (aboveanassay-defined threshold) ornonreactive(12, 25), itwasrecognized within the first few

years of the development of commercial assays that the

quantitative degree of EIAreactivity influenced the predic-tivevalue, withmorehighly reactive results havingagreater

positive predictive value (3, 14, 15, 21, 23, 25-27). With recent improvements in commercial EIA kit sensitivity and specificity, we reviewed this approach and conducted an

evaluation of the utility of quantitative EIA reactivity in

predictingHIV-1 seropositivity.

From1 October 1988to31December1991, 22,823 serum

specimens were tested for HIV-1 antibody. Samples were

initially screened once by EIA, and reactive samples were

then retested by EIA in duplicate. Samples repeatedly reactiveby EIAwereconfirmedaspositive by Western blot.

*Corresponding author. Mailing address: DenverDisease

Con-trolService, 605 Bannock Street, Denver, CO80204.Phone: (303)

436-7200. Fax:(303) 436-7211.

Allsampleswereobtainedfrompatientswho were tested for

clinicalpurposes or as part ofresearchprotocols orblinded seroprevalence studies. For analyses, patientswere

hierar-chically

stratifiedon thebasis of available riskfactor infor-mationas homosexual orbisexualmen(n = 5,067),

hetero-sexual intravenous (i.v.) drug users (n =

2,254),

or other heterosexuals (n = 15,502). HIV-1 antibody testing was

performedwithcommercially available Bio-Enzabead(prior

toAugust 1991) and VironostikaHIV-1 (since August 1991)

EIA kits (Organon Teknika Corporation, Durham, N.C.). Samples with optical density (OD) ratios greater than 1.0

wereconsidered reactive, asdetermined bythesample OD dividedby the assaycutoff

(calculated

in accordance with the manufacturer's instructions)for each run. Quantitative

EIA reactivity of repeatedly reactive samples was deter-mined as a mean of the sampleODratio ofthe threeassays

(initial assay and duplicate follow-up assay). Western blot

assays wereperformedwith thecommercially available HIV

Western Blot Kit from

Biotech-Dupont, Wilmington,

Del.

(prior to November 1991) or the HIV-1 Western Blot Kit from Epitope, Beaverton, Oreg.

(since

November

1991).

Westernblot resultsweredefined

by

theAssociation ofState andTerritorialPublic HealthLaboratory Directors

consen-suscriteria(5).

TABLE 1. SequentialEIAandWestern blottesting ofserum

samples byrisk group

No.%)

ni-No.

(%)re- No.(%)

con-Risk group No.of No.ll(%) m eaedl re- firmedHIV

samples reactive activeby positiveby

EIA Westernblot

Homo-or 5,067 1,686(33.3) 1,647(32.5) 1,645(32.4)

bisexualmen

i.v.drugusers 2,254 74(3.3) 59(2.6) 53(2.4)

Otherhetero- 15,502 126(0.8) 67(0.4) 49(0.3)

sexuals

Total 22,823 1,886(8.3) 1,773(7.8) 1,747(7.7) 220

Vol. 32,No. 1

on May 15, 2020 by guest

http://jcm.asm.org/

Downloaded from

on May 15, 2020 by guest

http://jcm.asm.org/

Downloaded from

on May 15, 2020 by guest

http://jcm.asm.org/

Downloaded from

on May 15, 2020 by guest

http://jcm.asm.org/

NOTES 221 TABLE 2. CorrelationofquantitativeEIAreactivityand

Western blotresults among1,773serumsamples repeatedly reactive by ELA testing

No. of samples with following level Western ofEIA reactivity: Riskgroup blot

result' High Moderate Low (>3.Op (>2.2-3.0)p (>1.0-2.2)

Homo-or bisexual + 1,436 196 13

men i 0 0 1

- 0 0 1

I.v. drugusers + 50 0 3

i 0 0 2

- 0 0 4

Otherheterosexuals + 37 12 0

i 0 0 6

- 0 0 12

Total + 1,523 208 16

i 0 0 9

- 0 0 17

a +,positive; i, indeterminate; -,negative.

bOD ratio.

Asshown in Table 1, of the total of 22,823 samples, 1,886 (8.3%) were initially reactive by EIA, 1,773 (7.8%) were repeatedly reactive by ETA, and 1,747 (7.7%) were con-firmedasHIV-1 positive by Western blot.

By using the meanEIA OD ratio to stratify samples by quantitative ETA reactivity, repeatedly reactive samples weredivided into threegroups: high level (OD ratio, >3.0), moderate level (OD ratio, >2.2to3.0), and low level (OD ratio,>1.0to2.2). Of the 1,773 repeatedly reactive samples, 1,523 (85.9%) had high-level reactivity, 208 (11.7%) had moderate-levelreactivity, and 42 (2.4%) had low-level reac-tivity. As outlined in Table 2, all samples with high and

moderate levels of ETAreactivitywere confirmed as HIV positive by Western blot (1,731 of 1,731, i.e., 100%). All repeatedly reactive samples found to be false positive by Western blot had low-level reactivity; of 42 low-level-reac-tive samples, only 16 (38.1%) were confirmed positive by

Western blot. All 16 samples occurred among the higher-prevalence populations of homo- orbisexual men and het-erosexual i.v.drugusers.

The positive predictive value of a sample repeatedly

reactiveby ETAwas99.9% amonghomo-orbisexualmen, but it was only 89.8% among i.v. drug users and 73.1% among other heterosexuals (Table 3). This difference was

directly related to theproportion of low-level ETA-reactive samplesamong repeatedly reactive samples in eachgroup: 0.9%amonghomo-orbisexualmen,15.3% amongi.v.drug

users, and 26.9% among other heterosexuals. While the positive predictive value ofalow-level-reactivesamplewas only86.7%amonghomo-orbisexualmen,33.3%amongi.v. drugusers,and 0%amongotherheterosexuals, the positive predictive of moderate- and high-level-reactive samples did not varywith HIV prevalence butwas100% for allgroups. With the development ofcommercially available HIV-1 EIAtesting for screening of blood donors, test kit cutoffs were chosentoprovide maximal sensitivity, with a

conse-quent reduction in specificity. While this approach was effective inprotecting the blood supply, it created problems with theuse of the EIA for diagnosis of HIV-1 infection,

especially in low-prevalence populations, for which the positive predictive value waslow. Consequently, the U.S. Public Health Service (4) and others (11, 14) have recom-mended routine use of confirmatory tests, such as the Western blot or indirect immunofluorescence assay, for verification ofsamplesrepeatedly reactive by EIAas HIV positive, a practice which has been widely accepted.

Fol-lowing initial concernsabout unacceptably high false posi-tivityratesinthegeneralpopulation (10, 18), this sequential testingstrategyhas beendemonstrated in severallarge-scale evaluations amonggeneralpopulationstohaveaspecificity of>99.999%,eveninlow-prevalencegroups (2, 17).

Despite these excellent results, there have been draw-backs tothis approach. All of the confirmatory assays are moretedioustoperform than theEIA, resulting in increased turnaroundtimes, rangingupto2weeks inonerecentsurvey (1). Additionally, the most common confirmatorytest, the Westernblot, has had severalproblems. Because the proce-dure relies on visible detection of characteristic bands, interpretation of results is inherently less objective than with the EIA. Interpretation has been further confounded by disagreement on the criteria that distinguish positive from indeterminate reactions, withatleast four differentsystems inuse (6). Finally, because production ofreagents is com-plex,assaycosts arerelatively high (19).

These issues have given rise to several attempts to de-velop simplified alternative algorithms for HIV testing (12, 13, 16, 19, 24). While these approaches have the greatest

value forsettings with limitedresources, suchasAfrica(13, 24), they are also relevant for developed countries. For example, in the United States, overtwo million HIV tests

were performed in public clinics in 1991 (8) and approxi-mately twice that numberwereperformed in private settings (9); an estimated additional annual one million tests have been performed recentlyas partof blinded seroprevalence surveys(7). The alternativealgorithms have generally relied upon aseries ofassayswhichemploy different methods and whichusequalitative (reactiveversusnonreactive) interpre-tations. Little attention has beenpaid totheuseof quanti-tative dataprovided by procedures suchastheEIA, despite substantial evidence that specificity andpositive predictive

TABLE 3. Positivepredictive value of repeatedly reactive EIA results for HIV-1 infection among various risk groupsby quantitativeETAreactivity

No.ofsamples confirmed positive/no. repeatedly reactive in EIA

(PPV0

[%])

Group HIVprevalence atfollowinglevel:

Any Low Moderate High

Homo- orbisexualmen 32.4 1,645/1,647 (99.9) 13/15 (86.7) 196/196 (100) 1,436/1,436(100)

I.v.drugusers 2.4 53/59 (89.8) 3/9 (33.3) 50/50 (100)

Other heterosexuals 0.3 49/67 (73.1) 0/18 (0) 12/12 (100) 37/37 (100)

a PPV,positive predictivevalue.

VOL.32, 1994

on May 15, 2020 by guest

http://jcm.asm.org/

J. CLIN.MICROBIOL.

Initial EIA

/\

Report Second EIA (duplicate) asnegative

(-I-)

(-I+and

+/+)

Report O.D. Ratio O.D. Ratio

as negative >3.0 >1.0-3.0

Report Confirmatory

as positive Test

(W.B.

or l.I.F.

Non-reactive Reactive

Report Report

as negative as positive

FIG. 1. Proposed HIV-1 testing algorithm usingquantitative EIA

reactivity. Sampleswouldbe screenedinitially byEIA and reactive

samples wouldbe retested in duplicate. Aminus sign indicates a sampleOD ratio below the assaycutoff,andaplus signindicatesan OD ratio above the cutoff.SampleswithmeanOD ratios of>3.0 for the three EIAs would be considered HIV-1 positive; those with

meanODratios of>1.0to3.0would be consideredpossibly positive

and evaluatedby a confirmatory assay such asthe Western blot

(W.B.)orindirectimmunofluorescence assay(I.I.F.).

value increase with more

highly

reactive results

(3,

15,

21-23, 25-27).

Our

study,

done witha

current-generation

HIV-1 EIAkit

and evaluation of over

20,000 samples

from risk group

populations

with infectionrates

ranging

from0.3%to

32.4%,

indicates that

high-level-reactive

EIA results are strongly

predictive

for HIV-1

infection, especially

for

high-preva-lencegroups, and insome

settings

mayreduceoreliminate

the need for

other,

confirmatory

tests. An

algorithm

which

would utilize such an

approach

is shown in

Fig.

1. To minimize

performance

errorsandassay-to-assay

variability,

repeat

testing

in

duplicate

of

initially

reactive

samples

is

proposed.

For

optimal specificity,

low-level-reactive

sam-ples

would

undergo confirmatory testing by

another

method,

although

most

specimens

(86%

of our

sample)

would not

require

this.

By

using

a

quantitative interpretation

of ETA data to

stratify qualitative

EIA

results,

the

sensitivity

and

specificity

of the conventional

algorithm

would be

main-tained,

without the timeandcostof

confirmatory testing

for most

samples.

Ifourresultsarecorroborated

by others,

thereareseveral situations in which this

approach might

beuseful.

Examples

include blinded

seroprevalence

surveys,

rapid diagnosis

in

patients

with

suspected

infection,

and resource-poor

set-tings,

suchas

developing

countries. For the

former,

in which results are used for

epidemiologic

purposes and are not

communicated to

patients

(8),

this

approach might

be of

immediate

value,

with

significant

cost

savings. Regarding

rapid diagnosis,

therearecircumstances in which

obtaining

highly predictive diagnostic

information within

days

rather than a week or more may be useful in redirecting the

managementof ill patients. Finally, limited resources have made the performance of traditional HIV-1 confirmatory

testing problematic in many settings in Africa (13, 24) and with the increasing difficulty of funding HIV-relatedservices there (20), the ability to provide highly predictive results with asingle assay would be of widespread value. Given the potential for its use in these and other settings, further studies to clarify the value of quantitative EIA testing are

indicated.

WethankMelisa Marquez for expert assistance with preparation of themanuscript.

REFERENCES

1. Association of State and Territorial Public Health Laboratory

Directors. 1992. Survey results. 7th Annual Conference on Human Retrovirus Testing: report. Association of State and Territorial Public Health Laboratory Directors, Chicago, Ill. 2. Burke,D.S., J. F.Brundage,R.R.Redfield, J. J. Damato, C. A.

Schable,P.Putman,and R. Visintine.1988. Measurement of the falsepositiveratein ascreening program for human

immuno-deficiencyvirusinfections. N.Engl. J. Med. 319:961-964. 3. Carlson,J. R.,M.L.Bryant,S.H.Hinrichs, J.K.Yamamoto,

N. B.Levy, J. Yee, J. Higgins,A. M.LeAvine,P. Holland,M. B. Gardner,and N. C.Pedersen. 1985. AIDS serology testing in low- andhigh-risk groups. JAMA 23:3405-3408.

4. Centersfor Disease Control. 1987. Publichealth service guide-lines forcounseling and antibody testingtopreventHIV infec-tionand AIDS. Morbid. Mortal.Weekly Rep. 36:509-515. 5. Centers forDiseaseControl. 1989.Interpretation anduseof the

Western blot assay for serodiagnosis of human immunodefi-ciency virus type-1 infections. Morbid. Mortal. Weekly Rep. 38:1-6.

6. Centers forDisease Control. 1991. Interpretive criteria used to reportWestern blot results for HIV-1-antibody testing-United States. Morbid. Mortal.Weekly Rep. 40:692-695.

7. Centers for DiseaseControl. 1991. National HIV serosurveil-lance summary, results through 1990, vol. 2, p. 1-29. U.S. Department of Health andHumanServices, Atlanta, Ga. 8. Centers forDiseaseControl.1992. Publicly funded HIV

coun-seling and testing-United States, 1991. Morbid. Mortal. Weekly Rep. 41:613-616.

9. Centers for Disease Control.1992. HIVcounseling and testing services from public and private providers-United States, 1990. Morbid. Mortal.Weekly Rep. 41:743-752.

10. Cleary,P. D., M.J. Barry, K. H. Mayor, A. M. Brandt, L. Gostin, and H. V. Fineberg. 1987. Compulsory premarital screening for the human immunodeficiency virus. JAMA 258: 1757-1762.

11. Consortium for RetrovirusSerology Standardization. 1988.

Se-rological diagnosisof humanimmunodeficiencyvirusinfection by Western blot testing. JAMA 260:674-679.

12. Davey, R. T.,and H.C.Lane.1990.Laboratory methods in the diagnosis and prognostic staging of infection with human immu-nodeficiency virus type-1.Rev.Infect. Dis. 12:912-930. 13. Fleming, A. F. 1988. Simplified confirmatory HIV testing.

Lancet ii:848.(Letter.)

14. Goedert, J. J.1986.Testingfor humanimmunodeficiencyvirus. Ann.Intern. Med.105:609-610.

15. Handsfield,H.H.,M.Wandell,L.Goldstein,K.Shriver,and the Cooperative Study Group. 1987. Screeningand diagnostic per-formance of enzymeimmunoassay for antibody to lymphade-nopathy-associatedvirus. J. Clin.Microbiol. 25:879-884. 16. Lepine, D. G., P. W. Neumann, S. L. Frenette, and M. V.

O'Shaughnessy.1990.Evaluation ofahumanimmunodeficiency

virus test algorithm utilizing a recombinant protein enzyme immunoassay.J. Clin. Microbiol. 28:1169-1171.

17. MacDonald,K.L., J.B.Brooks,R.J. Bowman,H.F.Polesky, F. S. Rhame, H. H. Balfour, and M. T. Osterholm. 1989. Performance characteristics ofserologictestsforhuman immu-nodeficiency virus type 1 (HIV-1) antibody amongMinnesota blooddonors. Ann. Intern.Med. 110:617-621.

222 NOTES

on May 15, 2020 by guest

http://jcm.asm.org/

VOL. 32, 1994 NOTES 223 18. Meyer, K. B., and S. G. Pauker. 1987.Screening for HIV:can

we affordthe falsepositive rate?N.Engl. J. Med. 317:238-241. 19. Mortimer, P. P. 1991. The fallibility of HIV Western blot.

Lancet 337:286-287.

20. N'Galy, B., S. Bertozzi, andR.W.Ryder. 1990.Obstaclestothe optimal management ofHIVinfection/AIDS in Africa. J. Ac-quiredImmuneDefic. Syndr. 3:430-437.

21. Nishanian, P., J.M.Taylor,E.Korns,R.Detels,A. Saah,and J. L. Fakey. 1987. Significance of quantitative enzyme-linked immunosorbent assay (ELISA) results in evaluation of three ELISAs andWestern blot tests for detection of antibodies to human immunodeficiency virus in a high-risk population. J. Clin. Microbiol. 25:395-400.

22. Schwartz, J. S., P.E.Dans, andB. P.Kinosian. 1988.Human immunodeficiency virus test evaluation, performance, and use. JAMA259:2574-2579.

23. Sivak,S. L., and G.P.Wormser. 1986.Predictive value ofa

screeningtestforantibodiestoHTLV-III.Am. J. Clin. Pathol. 85:700-703.

24. Spielberg,F.,C.M. Kabeya, T. C. Quinn, R. W. Ryder, N. K. Kifuani, J. Harris,T. R. Bender, and W. L. Heyward. 1990. Performance andcost-effectiveness ofadualrapid assay system for screening and confirmation of human immunodeficiency

virus type1seropositivity. J. Clin. Microbiol. 28:303-306. 25. Steckelberg, J. M.,and F. R.Cockerill. 1988. Serologictesting

for humanimmunodeficiency virus antibodies. Mayo Clin. Proc. 63:373-380.

26. Ward, J. W., A. J. Grindon,P. M. Feorino, C. Schable, M. Parvin, and J. R. Allen. 1986. Laboratory and epidemiologic evaluation ofanenzymeimmunoassay for antibodiesto HTLV-III. JAMA 256:357-361.

27. Weiss, S. H., J. J. Goedert,M.G.Sarngadharan,A.J.Bodner, R. C. Gallo, and W. A. Blattner. 1985. Screening test for HTLV-III (AIDS agent) antibodies. JAMA 253:221-225.

on May 15, 2020 by guest

http://jcm.asm.org/

2340 LETTERS TO THE EDITOR

accuratediagnosis is of vital importance. Antigen detection is very important in this respect. Over the years several methods for the detection of Aspergillus antigens have been published

(1,

3,

4).

Haynes' method uses concentrated urine specimens which areelectrophoresed on sodium dodecyl sulfate-polyacrylamide gels and blotted to polyvinyldifluoride paper. The blotted antigens are detected by a detector serum, rabbit serum directed against cell wall antigens from Aspergillus fumigatus. Haynes also publishedthe molecular weights of antigens which he claims to be specific for aspergillosis. Considering that this method seemed very promising for the rapid and accurate detection of Aspergillus antigens and that it is able to detect defined antigens, we introduced it in our laboratory.

Experiments withurine specimens from patients with proven invasive aspergillosis obtained from Haynes showed that our detector serum was able to detect the same antigenic bands on the blots as did Haynes' detector serum. Our detector serum

wasimmuneserumdirected against cell wall antigens fromA.

fuimigatus from a conventionally raised New Zealand White rabbit(Haynesdid notspecify the type of rabbit to be used for

immunization). Because wewanted to have negative controls in the test, weinvestigated 10 urine specimens from healthy individuals and 10 urine specimens from patients without aspergillosis butwith urinary tract infections.

In the urine specimens from patients with urinary tract infections, we found strong reactions on the blots in places specific for Aspergillus antigens. From two patients, the caus-ative organism could be isolated; both isolates were

Esche-richia coli. We preincubated the detector serum from the immunoblot withE. coli that had been treated for 1 h at 100°C. This treatment was intended to remove specific

lipopolysac-charide epitopes on the E. coli surface and to improve the exposure of common gram-negative surface antigens (5).

When this "absorbed" detectorserum wasused in the immu-noblot test, the previously found reactions in the patients' urinespecimensdisappearedalmost completely.

When investigating urine specimens from patients with

urinarytractinfectionsby usingunabsorbedserafrom

conven-tionallyraisedrabbits, oneshould beawareof this pitfall.

E. coli is not the only organism able to cause urinary tract

infections, we investigated three Aspergillus detector serum specimens for the presence of antibodiesagainstseveral other

organisms.Inallthreewefound,alongwithantibodiesagainst

E. coli, antibodies against (non-protein A-producing)

Staphy-lococcussaprophyticus, Klebsiella pneumoniae, Enterobacter clo-acae, and Proteusmirabilis invarying amounts.

Two of the Aspergillus detector serum specimens were

absorbed with E. coli, S. saprophyticus, and K pneumoniae

organisms. For thispurposeboth merthiolate-killed and heat-treated(1 h, 100°C) organismswereused. The reactions of the "absorbed" detector sera with water-soluble antigen from A. fumigatus NCPF 2109 on the immunoblot were not affected. On thecontrary,theintensity and the number of bands found on blots when "absorbed" detectorsera wereused in combi-nation with urine specimens from patients with urinary tract infections were lower than when "unabsorbed" detectorsera wereused.

Theuseof S.saprophyticusas anabsorbing organism hadno effect in the reaction between the urine antigens and the treatedsera.

After the problemsweexperienced,werecommend thatone

check theAspergillus detectorsera before using them in the immunoblot test for thepresence ofantibodies against com-mon organisms causing urinarytractinfections and, if neces-sary, oneincubate the serabefore usingthemwiththe

organ-isms against which antibodies have been found.

REFERENCES

1. de Repentigny, L. ,M. Boushira, L. Ste.-Marie, and G. Bosisio. 1987.Detection of galactomannanantigenemiaandantigenuria in aspergillosis: studies in patients andexperimentally infected rabbits. J.Clin. Microbiol25:863-867.

2. Haynes, K. A., J. P. Latge, and T. R. Rogers. 1990. Detection of Aspergillus antigens associated with invasive infection. J. Clin.

Microbiol. 28:2040-2044.

3. Reiss, E., and P. F. Lehmann. 1979. Galactomannanantigenemiain invasiveaspergillosis. Infect.Immun.25:357-365.

4. Sabetta,J. R., P.Minter,andV. T.Andriole.1985. Thediagnosis of invasive aspergillosis by an enzyme-linked immunosorbent assay for circulatingantigen. J. Infect. Dis. 152:946-953.

5. Vreede, R. 1988. Ph.D. thesis. Rijks UniversiteitUtrecht, Utrecht, The Netherlands.

LucM.WiJnands Frans M. vanLeusden RobJ. T. Puyk Marcel P. M.Hofstee H. W.Boudewijn Engel

Laboratoryfor Parasitology andMycology National Instituteof Public Health and

Environmental Protection P.O. Box 1

3720 BABilthoven, The Netherlands Ed. Note: The author of the published article declined to

respond.

Utility

of

Quantitative

Enzyme

Immunoassay Reactivity

for

Predicting

Human

Immunodeficiency

Virus

Seropositivity

in

Low-

and

High-Prevalence Populations

Hou et al. (2) reported the practical value ofquantitative

enzyme immunoassay (EIA) reactivity in predicting human

immunodeficiency virus type 1 (HIV-1) seropositivity. For HIV-1 EIAs, the relationship between antibody titer and absorbance is not linear, and therefore this test was never

intended to be quantitative (1). Because of the serious

conse-quences ofapositivediagnosisofHIV-1 infection,areport of

areactive HIV-1 EIAshould neverbe made without supple-mentary, confirmatorytesting.

REFERENCES

1. George, J. R. 1992.Qualitycontrol forserologic testing,p. 79-89. In G. Schocheteman and J. R. George(ed.),AIDStesting, method-ologyandmanagement issues.Springer-Verlag,New York. 2. Hou, X., P. L. Breese, and J. M. Douglas, Jr. 1994. Utility of

LETTERS TO THE EDITOR 2341 quantitative enzymeimmunoassay reactivity for predicting human

immunodeficiency virusseropositivity in low- and high-prevalence populations. J.Clin. Microbiol.32:220-223.

Frank J.Michalski, Ph.D. Diagnostic Virology Laboratory MetPath

One Malcolm Avenue Teterboro, New Jersey 07608 Author's Reply

We appreciate Dr. Michalski's comments on our study describingthe use ofquantitative EIAreactivity inpredicting HIV seropositivity by Western blot. We agree that the rela-tionship between HIV antibody titer and absorbance is not

directly linear, as demonstrated by George (1). Furthermore, one cannot define a true quantitative relationship between a

screening test which generates continuous numerical data (HIV EIA) and a confirmatory test with a categorical inter-pretation (HIV Western blot). However, there is a generally directrelationship between EIA absorbance andHIVantibody titer asmeasured by serial dilutions (1), andwe have demon-strated that one can make semiquantitative use of these

quantitative data topredictHIVWestern blotpositivity with a high degree ofreliability.

Given the significant clinical and psychological impact of a

positive diagnosis ofHIV-1 infection, we do not suggest that

ouralgorithm is currently appropriateforwidespreaduse. We

do feel that it is an approach which warrants furtherevaluation

and which may findutility at present insituations such asthe selected circumstances we suggested (e.g., epidemiologic sur-veys, rapid diagnosis, and resource-poor settings).

REFERENCE

1. George, J. R. 1992.Quality control for serologic testing, p. 79-89.In

G. Schocheteman and J. R. George (ed.), AIDStesting, method-ologyand management issues. Springer-Verlag, New York.

Xinyue Hou PaulaL. Breese

John M.Douglas, Jr. DenverDiseaseControlService 605Bannock Street

Denver, Colorado 80204

E

Test

as

Susceptibility Test for Evaluation of Neisseria meningitidis Isolates

We read withinterest the article by Hughesetal. (4) about theirexperience with ETest(AB Biodisk, Solna, Sweden)as a

susceptibilitytestforevaluating Neisseria meningitidis isolates. We agree thatthe test is useful and think that itcould be the method of choice for separating penicillin-sensitive strains from penicillin-resistant strains in laboratories without facili-ties for agar dilutiontechniques.

In recent years, N. meningitidis strains with low levels of penicillin resistance have been reported in Great Britain (5), Canada(8), Spain (7, 9, 10), and elsewhere (1, 11). In these strains, the MIC ofpenicillin is 5-to50-foldhigher than it is in susceptible strains (6). Tentative criteria have been proposed fordiscriminating betweenstrainswith moderatesusceptibility and those with full susceptibility to penicillin by the disk diffusion method (2,3), but we have found noantibiotic disk sufficiently sensitive and specific to separate clearly the two

bacterial populations.

In our opinion, the oxacillin 1-pLg disk is not suitable for differentiating these populations. Thirty of 65 strains with

penicillinMICs of 0.03to0.06 pLg/mltestedin anearlierstudy producedno inhibition zone around the oxacillin disk.

We studied 187N.meningitidis isolates. Noneof the strains produced ,B-lactamase. Nonduplicated organisms from

re-cently obtained clinical isolates (cerebrospinal fluid, blood cultures, orpharyngealswabs from carriers)were maintained

as stock cultures at -70°C until just before testing. Stock cultureswerethawed andsamples were inoculated onto plates containing 5% chocolate horse blood agar. After incubation for20 to 24h,isolate colonies were subcultured onto a second

chocolate agarplate whichwasincubated foranother 20 to 24 h.Growth from thisplatewasused to prepareinocula.

GC agar (BBL, Cockeysville, Md.) supplemented with 5% chocholate horse bloodwasusedfor classicaldiskdiffusion, the

ETest, andagardilution.

Plates were inoculated with 5.107 CFU for classical disk

diffusion andthe E Test and with 104 CFU for dilution agar (the reference MIC method).Allplateswere incubated in5% CO2at 35°C for24h.

Of the 187 strainsstudied,61 strains hadpenicillin MICs of -0.25 ,ug/ml by the reference agar dilution method. By the diffusionmethod, the penicillin 2U disk (P2) and the amdino-cillin 10-,ug disk (AMD10) were useful for discriminating betweenpenicillin-resistant and penicillin-sensitive strains, but theirspecificityandsensitivitywere notoptimal.No

penicillin-sensitive strain had a P2zoneof less than 22 mm indiameter

or an AMD10 zone of less than 16 mm, and no penicillin-resistant strain had a P2 zone of more than 27 mm or an AMD1O zone of more than 21 mm. However, many strains produced intermediate results that were between these limits:

72 (38.5%) strains screened with the P2 disk and 36(21.2%) strains screenedwith the AMD10 disk.

Comparison of the MICs obtained by the agar dilution method and the E Test showed agreement of the results to within 1 log2dilution in 97.3% (182 of 187) of strains. For the remainingfive strains, the results of these two methods agreed to within 2 log2 dilutions, and only one of these strains was

penicillin resistant (MIC = 0.25 p.g/ml).

REFERENCES

1. Botha,P. 1988. Penicillin-resistantNeisseriameningitidis in South Africa. Lancet i:54.

2. Campos, J.,P. M.Mendelman, M. U.Sako, D.0. Chaffin,A. L.

Smith,andJ.A. Saenz-Nieto. 1987. Detection ofrelatively peni-cillin G-resistantNeisseriameningitidis bydisk susceptibility

test-ing. Antimicrob. Agents Chemother. 31:1478-1482.

3. Campos, J., G. Trujillo, T. Seuba, and A. Rodriguez. 1992. Discriminative criteria forNeisseria meningitidis isolates that are

moderately susceptible to penicillin and ampicillin. Antimicrob.

AgentsChemother. 36:1028-1031.

4. Hughes, J. H., D.J. Biedenbach,M. E.Erwin, and R.N. Jones. 1993. E test assusceptibility test and the epidemiologic tool for evaluation of Neisseria meningitidis isolates. J. Clin. Microbiol. 31:3255-3259.

5. Jones,D.M.,and E. M.Sutclife.1990.Meningococci with reduced susceptibility topenicillin. Lancet 335:863-864.