17.zeh (2011) commercial elisa testing in hiv

8/13/2019 17.Zeh (2011) Commercial ELISA Testing in HIV

1/8

Journal of Virological Methods 176 (2011) 2431

Contents lists available at ScienceDirect

Journal ofVirological Methods

journal homepage: www.elsevier .com/ locate / jv i romet

Performance ofsix commercial enzyme immunoassays and two alternative

HIV-testing algorithms for the diagnosis ofHIV-1 infection in Kisumu, Western

Kenya

C. Zeh a,, B. Oyaro c, H. Vandenhoudtf, P. Amornkula, A. Kasembelic, P. Bondo c, D. Mwaengo c,T.K. Thomas a, C. Hart b, K.F. Laserson a,e, P. Ondoag,J.N. Nkengasongd

a US-Centers forDisease Control andPrevention (CDC), Kisumu, Kenyab Division of HIV/AIDS Prevention, National Center forHIV, Viral Hepatitis, STD, andTB Prevention,CDC, Atlanta, GA,USAc KenyaMedical Research Institute, Center for GlobalHealthResearch, Kisumu, Kenyad Global AIDS Program, National Center for HIV, Viral Hepatitis, STD, andTB Prevention, CDC, Atlanta, GA, USAe

US-Centers forDisease Control and Prevention, Center for GlobalHealth, Atlanta, GA,USAf Institute of Tropical Medicine, Antwerp, Belgiumg Center for Poverty-related CommunicableDiseases (CPCD), Department of Internal Medicine, Center for Infection and Immunity (CINIMA), Amsterdam Institute for GlobalHealth

and Development (AIGHD), AcademicMedical Center, University of Amsterdam, The Netherlands

Article history:

Received 30 December 2010

Receivedin revised form 11 May 2011

Accepted 12 May 2011

Available online 26 May 2011

Keywords:

HIV-1

SensitivitySpecificity

Diagnosis

Immunoassay

Alternative

Algorithm

a b s t r a c t

Performances ofserological parallel and serial testing algorithms were analyzed using a combination of

three ELISA and three rapid tests for the confirmation ofHIV infection.

Each was assessed individually for their sensitivity and specificity on a blinded panel of 769 retro-

spective sera ofknown HIV status. Western blot was used as a confirmatory assay for discordant results.

Subsequently, one parallel and one serial testing algorithm were assessed on a new panel of 912 HIV-

positive and negative samples. Individual evaluation ofthe ELISAs and rapid tests indicated a sensitivity

of100% for all assays except Uni-Gold with 99.7%. The specificities ranged from 99.1% to 99.4% for rapid

assays and from 97.5% to 99.1% for ELISAs. A parallel and serial testing algorithms using Enzygnost and

Vironostika, and Determine followed by Uni-Gold respectively, showed 100% sensitivity and specificity.

The cost for testing 912 samples was US$4.74 and US$ 1.9 per sample in parallel and serial testing respec-

tively. Parallel or serial testing algorithm yielded a sensitivity and specificity of 100%. This alternative

algorithm is reliable and reduces the occurrence ofboth false negatives and positives. The serial testing

algorithm was more cost effective for diagnosing HIV infections in this population.

Published by Elsevier B.V.

1. Introduction

Accurate serological HIV testing is key to the early diagnosis

and timely counseling of HIV-infected people, to effective screen-

ing of blood andblood products used in transfusions, to prevention

of mother-to-child transmission, and to monitor HIV prevalence

in the population (Nkengasong et al., 1999). The development ofrapid, reliable and less expensive serological testing assays and

Disclaimer: The findings and conclusions in this article are those of the authors

and do not necessarily represent the views of the U.S. Centers for Disease Control

and Prevention.Use of trade namesis foridentificationpurposes only anddoes not

constitute endorsement by the U.S. Centers for Disease Control and Prevention or

theDepartment of Health andHuman Services. Correspondingauthor at:Centersfor DiseaseControland Prevention, Kenya, Off

Kisumu-Busia Road, P.O Box 1578-40100, Kisumu, Kenya. Tel.: +254 724 255 639;

fax: +254 572 022981.

E-mail address: [email protected] (C. Zeh).

algorithms continue to be a challenge in resource-limited settings

(Butler et al., 2007; Foglia et al. , 2004; Owen et al., 2008; WHO,

1997). HIV-1 subtype diversity remains one of the main challenges

of accurate screening strategies due to the emergence of new

viruses and inter- and intra-subtype to HIV-1 as well as recom-

bination to HIV-1 group M viruses (Butler et al., 2007). Another

challenge is the affordability of HIV diagnostic assays that need tobe inexpensive, require minimal infrastructure and be easily trans-

ported. Choosing the appropriate serological screening algorithms

in geographical areas with multiple HIV subtypes and viral recom-

binants combined with limited resources require identification of

assays that have optimal sensitivity and specificity for that partic-

ular setting (Owen et al., 2008). In addition, the complexity and

feasibility of these protocols need to be addressed.

Most sub-Saharan countries follow the World Health Organi-

zation (WHO) guidelines on use of HIV antibody detection tests

(WHO,1997). The recommended testalgorithmincludesa sensitive

enzyme-linked immunosorbent assay (ELISA) for initial detection

0166-0934/$ see front matter. Published by Elsevier B.V.

doi:10.1016/j.jviromet.2011.05.021
http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.jviromet.2011.05.021http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.jviromet.2011.05.021http://www.sciencedirect.com/science/journal/01660934http://www.elsevier.com/locate/jvirometmailto:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.jviromet.2011.05.021http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.jviromet.2011.05.021mailto:[email protected]://www.elsevier.com/locate/jviromethttp://www.sciencedirect.com/science/journal/01660934http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.jviromet.2011.05.021


2/8

C. Zeh et al. / Journal of Virological Methods 176 (2011) 2431 25

Table 1A

Characteristics of thethreeELISAs evaluated on 769 samples of Blood Bank samples in Kisumuand Nairobi,Kenyain 2004.

ELISA (manufacturer) Antigen (Ag) Group O Ag Test principle Cost/test ($) Comments

HIV-1 HIV-2

Murex HIV-1.2.0 (Abbott) r-p24, r-gp41, sp-gp41 sp-gp36 + Sandwich 2.1 Ag and Ab

Vironostika HIV Uni-Form II Plus O (Organon Teknika) r-p24, r-gp41 sp-gp36 + Sandwich 2.12 Ag and Ab

Enzygnost HIV-1/2 sp-gp41 sp-gp36 + Indirect 2.42 Ag and Ab

of HIV serum antibodies, followed by a confirmation of all positive

samples by Western blot (WB) (Beelaert et al., 2002; Nkengasong

et al., 1999; Respess et al., 2001; WHO, 1998). This algorithm is

impractical in many developing countries due to the prohibitive

cost of WB, its difficult interpretation, and need for specific instru-

mentation. The format of an ELISA is the most appropriate for

screening large numbers of specimens with limited infrastructure.

The improvements in HIV ELISAs from first to fourth generation,

including better HIV target antigens plus antibodies have greatly

enhanced their sensitivities and specificities. Detection of diver-

gent HIV-1 strains like group O as well as reactivity against both

IgMandIgG arenow acceptedstandards forthese assays(Aghokeng

et al., 2009). New generation ELISAs have also reduced the window

period between infection and appearance of detectable antibodiesin the blood to between 15 and 20 days (Busch and Hecht, 2005;

Koch et al., 2001; Mylonakis et al., 2000; Nguyen and Busch, 2000;

Parry et al., 2003; Stramer et al., 2004). In addition to ELISAs, a vari-

etyof rapidtestsare also availableand suitable forlaboratories with

limited space and equipment to perform tests on a low number of

samples daily.

Diagnostic algorithms which combine two or more serological

tests are reliable and greatly reduce the frequency of false pos-

itivity hence reduce misdiagnosis as compared to using a single

ELISA (Owen et al., 2008) and most importantly, limits the neces-

sity for WB testing. There has been an upsurge in the use of testing

algorithms that combine both ELISA and rapid tests in African

and Asian countries (Respess et al., 2001; WHO, 1998). These

algorithms have invalidated the use of the expensive and time-consuming WB for confirmation (Bebell et al., 2010; Sommerfelt

et al., 2004; Tenenbaum et al., 2005). In addition, these combi-

nation algorithms have very attractive operational characteristics

in terms of simplicity of sample, reagent preparation and high

throughput in addition to allowing prompt reporting and patient

counseling often on the same day (Owen et al., 2008). Western

Kenya is known to harbor a variety of HIV-1 subtypes including

A and D, circulating recombinant forms (CRFA/D) and other, unty-

peable HIV-1 viruses (Yang et al., 2004). Effective HIV diagnosis in

this resource-limited region could be undermined by emergence

of new subtypes and recombinant forms. Hence, it is important to

evaluate periodically and select optimally performing serological

assays before their use in large scale. The HIV screening algorithm

in use during the time of this study was serial testing using a com-bination of rapid tests; Determine followed by Bioline (NASCOP,

2001).

The WHO provides a template (WHO/UNAIDS I, II and III)

which sets a framework for developing region-specific parallel

and serial algorithms for the diagnosis of HIV infection (WHO,

1997). Developing such an algorithm involves an initial evalua-

tion of the sensitivity and specificity of ELISAs and rapid tests for

a particular region followed by development of a suitable algo-

rithm, which combines the most sensitive and specific assays

(Razafindratsimandresy et al., 2006).

Three ELISAs and three rapid tests were evaluated for capac-

ity to diagnose HIV infection in western Kenya. Based on these

results, parallel and serial algorithms, based on strategy II of WHO

but including WB and/or HIV-1 DNA PCR in cases of discordant or

indeterminate results, were designed and evaluated. Operational

characteristics and costs related to the algorithms were assessed.

2. Materials and methods

2.1. Study design and specimen source

Between 2003 and 2004, 769 discarded human sera were

obtained from the bloodtransfusion centers in Nairobi and Kisumu,

Kenya. The panel of serum samples was designed to obtain

approximately 50% HIV prevalence. These sera were analyzed ret-

rospectively in order to determine the sensitivity and specificity

of three ELISAs (Vironostika, Enzygnost, Murex) and three rapidtests (Determine, Uni-Gold and Capillus). Subsequently, the two

best ELISAs and rapid tests were selected for the prospective anal-

ysis on the basis of their sensitivity, specificity, positive predictive

values (PPVs) negative predictive values (NPVs), delta () values,

cost and availability of the assays locally.

The prospective algorithms were tested using 912 sera drawn

from a cross-sectional survey aimed at estimating the prevalence

of HIV and sexually transmitted infections (STIs) and associated

risk factors in preparation for a future biomedical HIV prevention

and intervention trial in rural western Kenya conducted between

January and May 2005. Peripheralbloodmononuclear cells(PBMCs)

from each participant were cryopreserved for future analyses.

The study was approved by the Kenyan Medical Research Insti-

tute (KEMRI) Ethical Review Committee (protocol #886) and U.S.

Centers for Disease Control and Prevention (CDC) Institutional

Review Board (protocol #4366). All participants were counseled

and those who agreed to participate in the study were consented.

Participantsaged below 18 years were considered minorsand were

accompanied by a guardian to consent/assent.

2.2. Characteristics of ELISA and rapid tests

Theessentialcharacteristics of theplated HIV-1 and-2 antigens,

Group O compatibility, testing principles, and acceptable speci-

mens used in three ELISAs: Enzygnost anti HIV-1/HIV-2 Plus (Dade

Behring Diagnostics, Marburg, Germany); Vironostika HIV Uni-

form II Plus O (Biomerieux, Boxtel, Netherlands); Murex HIV-1.0.2

(Murex Biotech Limited, Dartford, UK) and three rapid tests: Capil-

lus HIV-1/2 (Trinity Biotech, Wicklow, Ireland); Uni-Gold HIV-1/2

(Trinity Biotech, Wicklow, Ireland) and Determine HIV-1/2 (Abbott

Labs, Tokyo, Japan) are summarized in Tables 1A and 1B. These

assays were selected for this evaluation based on at least one of the

following criteria: availability of published performance informa-

tion, capacity of the assay to detect HIV-1 group M and O infections

and ability to detect IgG and IgM antibodies.

2.3. ELISA and rapid test procedures

ThethreeELISAswereperformedonallseraandreadonanELISA

reader (Sunrise Remote/Touch Screen A-5002 Austria) according to

the manufacturers guidelines and JointUnited Nations Programme

on HIV AIDS.


3/8

26 C. Zeh et al. / Journal of VirologicalMethods 176 (2011) 2431

Table 1B

Characteristics of thethreerapid assays evaluated on 769 samples of Blood Bank samples in Kisumu andNairobi Kenya in 2004.

Rapid test (manufacturer) Antigen (Ag) Test principle Cost/test ($) Specimen type

HIV-1 HIV-2

Determine HIV-1/2 (Abbott Lab, Tokyo, Japan r and sp g p41, gp-120 r and sp-gp36 Immunochromatographic 1.23 Serum, plasma o r w hole blood

Uni-Gold HIV-1/2 r-gp41,gp-120 r-gp36 Immunochromatographic 2.15 Serum, plasma or whole blood

Capillus HIV-1/2 r-gp41, gp-120 r-gp36 Agglutination test 3.69 Serum, plasma or whole blood

gp, glycoprotein; r, recombinant; sp, synthetic protein; Env, envelope.

The three rapid HIV tests were also performed on all sam-

ples and HIV results were reviewed and recorded according to

manufacturers criteria for interpretation of positive, negative or

inconclusive results. Repeat testing was done on any sample that

gave invalid/equivocal result as per manufacturers recommenda-

tions.

2.4. Western blots (WB) and HIV-1 DNA PCR

HIV-1 WB (HIV Blot v2.2, Gene Labs Diagnostic, Singapore)

was performed on any sample discordant by the six assays for

the retrospective analysis and for any specimen discordant by the

two selected assays in the prospective analysis. The WB results

were interpreted using the manufacturers instructions and theCDC guideline (CDC, 1989). HIV results of individual samples were

recorded after reading by two different laboratory technologists

and according to manufacturers criteria for interpretation of pos-

itive, negative or inconclusive results. If the assay criteria were

not fulfilled, the test was repeated and results reviewed before

beingrecorded. For the prospectiveanalysis, HIV-1DNA PCR (Roche

Amplicor v1.5, Roche Molecular Systems, Inc Branchburg, NJ USA)

was used to test all samples with WB indeterminate results as pre-

viously described (Nkengasong et al., 1999; Silverstein et al., 2004).

2.5. HIV parallel and serial algorithms

For the prospective analysis,sera weretested in a parallel testing

algorithm using two ELISAs (Vironostika HIV Uniform Integral andEnzygnost HIV-1/2 plus) chosen based on their test performance

in the retrospective study previously described. In the algorithm

(Fig. 1A), concordantly reactive or non-reactive sera by these two

ELISAs were considered to be true positives or true negatives. Sera

discordant by the two ELISAs were tested by WB. Sera negative by

WB were considered true negatives. For sera with indeterminate

WB results, a stored aliquot of cryopreserved PBMCs was tested

by HIV-1 DNA PCR (Roche DNA PCR version 1.5 Roche Diagnostic

System Branchburg, NJ, USA) and results from nucleic acid ampli-

fication were considered definitive.

In addition, the World Health Organizations proposed serial

testing algorithm (WHOII), which is recommended for testing of

asymptomatic seropositive persons in areas with HIV prevalence

greater than 10% (Razafindratsimandresy et al., 2007; WHO, 1997)was evaluated. In this algorithm (Fig. 1B), sera that were non-

reactive by Determine HIV 1/2 were considered as true negatives.

Sera that were reactive were confirmed using Uni-Gold HIV 1/2.

Reactive results were considered true positives. Sera discordant by

thetwo assayswereretestedby WB.Positive or negative WB results

were considered definitive. In case WB results were indeterminate,

a stored aliquot of cryopreserved PBMCs was tested by HIV-1 DNA

PCR and the results were considered definitive.

2.6. Data analysis

2.6.1. Analysis of individual ELISAs and rapid tests

In the assessment of performance of the various assays, true

positives were defined as those samples which were concordantly

reactive by all six assays and/or positive by WB if the assays

were discordant. True negatives were defined as concordantly non-

reactive by all six assays and/or negative by WB if the assays were

discordant. A false-positive result wasdefinedasareactiveresultby

the assay under question whereas other assays were non-reactive

and the WB was negative. A false-negative was defined as a non-

reactive result by the assayunder question whereother assays were

reactive and WB was also positive. All samples with indeterminate

WB results (n= 13) were excluded from the analyses of sensitivity

and specificity.

2.6.2. Sensitivity, specificity, positive predictive value (PPV),

negative predictive value (NPV), likelihood ratio and delta value()

Sensitivity and specificity: Sensitivity of an assay was calculated

as the sum of concordant reactive sera by all the six assays and

WB positive of the discordant divided by the sum of concordant

reactive sera by all the six assays and WB positive of the discor-

dant plus number of discordant by the assay. Specificity of an assay

was calculated as the total number of concordant non-reactive sera

by all the six assays and WB negative of the discordant divided

by the total number of concordant non-reactive sera in all the

six assays and WB negative of the discordant plus number of dis-

cordant by the assay. PPV and NPV were calculated as previously

described (Nkengasong et al., 1999). Optical density ratio for the

prospective analysis samples were categorized into three quanti-

tative groups: highly reactive (OD ratio > 5.0), moderately reactive(OD ratio > 3.05.0), and weakly reactive (OD ratio> 1.03.0) by the

ELISAs and WB for discordants. These OD ratios have previously

been used in different settings for the validation of ELISAs in Cote

dIvoire (Nkengasong et al., 1999), Tanzania (Urassa et al., 1999)

and China (Hei et al., 2007). The usefulness of this quantitative

ELISA reactivity in predicting HIV seropositivity was assessed for

all three ELISA methods. The positive predictive value of the quan-

titative ELISA results was defined as the number of reactive sera

with high OD ratio divided by the total number of sera with high

OD ratio.

Likelihood ratios are preferred for expressing generalizable test

performance since they are not affected by disease prevalence

(Grimes and Schulz, 2005). The positive likelihoodratio was defined

as the ratio of the test sensitivityto the false positive rate while thenegative likelihood ratio was defined as the ratio of false negative

rate to the specificity (Eller et al., 2007).

Delta values of the anti-HIV positive and negative samples were

calculated by dividing the mean OD ratio (log10) by the standard

deviation of each sample group/population as described previ-

ously in other publications (Bassett et al., 2011; Eller et al., 2007;

Lyamuya et al., 2009; Respess et al., 2001). In the event of an over-

flow, usually denoted as Over (above the upper limit of detection

of the ELISA plate reader), or **** in the data print out, an OD

of 3.500 or 4.000 was assigned to the specimen depending on

the ELISA used. The higher the positive (+) and negative ()

delta values, the higher the probability that the test will distin-

guish antibody positive and negative specimens (Lyamuya et al.,

2009).


4/8


Determine

n = 912

Determine (+)

n = 138

Uni-Gold

Discordant

n = 3

Western Blot

Uni-Gold (+)

n = 135

Indeterminate

n = 2

HIVTotal positive

n = 136

HIV-1 DNA PCR

HIV

Total negativen = 776

2Neg

Determine (-)

n = 774

1 Pos

FIG. 1A.

Enzygnost and Vironostikan = 912

Discordant

n = 4

Enzygnost (+)/Vironostika (+)

n = 136

Enzygnost (-)/

Vironostika (-) n = 772

Western Blot

Indeterminate

n = 3

HIV-1 DNA PCR3 Neg

HIV

Total negativen = 776

1 Neg

HIVTotal Positive

n = 136

A

B

Fig. 1. (A) Parallel testing algorithm: Results of the parallel serologic algorithm for the diagnosis of HIV infection (negative, ; positive, +; discordant or indeterminate).

A true positive serum was concordantly positive by both ELISA (enzyme-linked immunosorbent assay; i.e. Enzygnost and Vironostika, or positive by Western blot. A true

negativeserumwas definedas concordantlynegative reactivity by thetwo initial ELISA (Enzygnost and Vironostika) or discordant by the ELISAs, indeterminate by Western

blot, and negative by HIV-1 DNA PCR. B. Serial testing algorithm: Results of serial serologic algorithm for the diagnosis of HIV infection (Negative, Positive, Discordant or

Indeterminate). A truepositive serum wasdefinedas positiveby bothDetermineand Uni-Goldor positive by Westernblot. A truenegative serum wasdefinedas non-reactive

by Determine or discordant by Uni-Gold, indeterminate by Western blot andnegative by HIVDNA PCR.

2.6.3. Quality assurance

To assess the possible occurrence of false-positive or false-

negative results from concordantly reactive or non-reactive sera

using ELISAor rapidtests, a randomlyselectedsubsetof serumsam-

ples (10% of concordantly reactive and non-reactive) were tested

by WB. In addition, all the six assays used in this evaluation are

enrolled in the College of American Pathologists external quality

assurance program.

2.6.4. Cost estimate

The costassociated withusing the parallel testing algorithmwas

compared to that associated with serial testing algorithm. The cost


5/8

28 C. Zeh et al. / Journal of VirologicalMethods 176 (2011) 2431

estimate was determined by the reagents cost per algorithm only

(Fig. 1A and B).

3. Results

3.1. Evaluation of individual assay performance: evaluation of

assay sensitivity

Of the 769 sera, 408 (53.1%) were concordantly HIV reactive

across all six assays, 330 (42.9%) were concordantly HIV non-

reactive across all six assays and 31 (4.0%) were discordant. Of

the discordant samples tested by WB, one (3.2%) was reactive, 17

(54.9%) were non-reactive and 13 (41.9%) remained indeterminate.

The 13 samples with indeterminate WB results were excluded in

the analysis of sensitivity and specificity (Table 2). All the ELISAs

andrapid tests correctlyidentified the HIV reactive sera (100% sen-

sitivity)except Uni-Gold which failedto identify onesample (99.7%

sensitivity). All assays had specificities of98.9% with the excep-

tion of Murex which had a specificityof 97.5% (Table 2). Most of the

HIV reactive sera reacted with high OD ratios (OD ratio > 5.0). Sera

that reacted with moderate to low OD ratios (OD ratio97.8%) and negative predictive values (NPV) (>99.7%).

Positive likelihood ratios for the ELISAs ranged from 38.6 for

murex and 115.7 for vironostika while for the rapid tests it

ranged from 115.7 to 173.5 for Determine and Capillus respec-

tively. The negative predictive value was zero for all the six

assays.

Delta () values for positive sera ranged from 4.50 for Murex

to 5.25 for Vironostika. Enzygnost had the highest negative

value (7.67) followed by Murex (6.27) and Vironostika (5.14)

(Table 2). The results among these tests were comparable to a

sensitivity of >99.7% and a specificity of >98.9%, with the excep-

tion of the Murex ELISA which had a lower specificity of 97.5%(Table 2).

To evaluate the performance of the parallel testing algorithm

in the prospective analysis, Enzygnost and Vironostika were cho-

sen based on their performance in the retrospective analysis. Both

ELISAs had 100% sensitivity while Vironostika recorded the highest

specificity (99.1%), PPV (99.3), positive likelihood ratio (115.7) and

+ value (5.25) of all the ELISAs. For the serial testing algorithm,

Determine and Uni-Gold were similarly chosen, with Determine

more sensitive (100%) and Uni-Gold more specific (99.4%) with a

high positive likelihood ratio (173.1). In both algorithms, WB was

used as a confirmatory assay for discordant results and HIV DNA

PCR for indeterminate WB results.

3.2. Evaluation of the parallel testing algorithm

Of 912 sera samples from the cross-sectional study, 136 (14.9%)

were reactive and 772 (84.6%) were non-reactive by both ELISAs

while 4 (0.4%) were discordantly reactive. Of these, one was WB-

negative and three were WB-indeterminate but were subsequently

negative by HIV-1 DNA PCR (Fig. 1A).

All 136 sera concordantly reactive by both ELISAs had high OD

ratios (>5.0) resulting in 100% sensitivity and negative predictive

value. Of the four discordant sera, one (WB-indeterminate) had a

moderate OD ratio and three (WB-negative) had low OD ratios. Of

the 776 non-reactive sera, 772 were reported as negative in the

parallel testing algorithm resulting in a specificity of 99.5% (95% CI:

98.699.7%) and 99.3% positive predictive value (Table 3). Table

2

Sensitivity,specificity,predictive,likelihoodratio

and

valuesofthesixEIAonseraunequivocally

positiveornegativebyWesternblotoftheretrospectiveanalysisofBloodBanksamplesinKisum

uandNairobi,Kenyain2004.

Tests

No.o

fsamplesa

Sensitivity(95%CI)

Specificity(95%CI)

Predictivevaluesb(%)

Positivelikelihoodratio

c

Positive

Negative

Positive

N

egative

(95%CI)

+

1.V

ironostika

409

344

100(99.1100)

99.1

(97.599.7

)

99.3

(97.999.9

)

100(98.9100)

115.7

(37.5356.9

)

5.2

5

5.1

4

2.Enzygnost

409

343

100(99.1100)

98.9

(97.199.6

)

99.0

(97.599.7

)

100(98.9100)

86.8

(32.7229.8

)

4.9

8

7.6

7

3.Murex

409

338

100(99.1100)

97.4

(95.198.6

)

97.8

(96.099.0

)

100(98.9100)

38.6

(20.273.5

)

4.5

6.2

7

4.Determine

409

344

100(99.1100)

99.1

(97.599.7

)

99.3

(97.999.9

)

100(98.9100)

115.7

(37.5356.9

)

n/a

n/a

5.Uni-Gold

408

45

99.8

(98.699.8

)

99.4

(97.999.8

)

99.5

(98.399.9

)

99.7

(98.4100)

173.1

(43.5689.3

)

n/a

n/a

6.Capillus

409

345

100(99.0

7100)

99.4

(97.999.8

)

99.5

(98.399.9

)

100(98.9100)

173.5

(43.6691.0

)

n/a

n/a

aAllserareactive/non-reactivebyeachtests.

b

ThepositiveandnegativepredictivevalueofanELISAandrapidtest.

cPositiveandnegative

values.CI,confidence

interval.


6/8


Table 3

Positive predictive values of quantitative reactivity of all reactive sera by the two ELISA(n= 140) prospectively on Gem Baseline Cross-sectional Survey samples in Kisumu,

Western Kenya in 2005.

ELISA Western Blot results No of sera with OD/cut-off value ELISA reactivitya

High (>5.0) Moderate (>3.05.0) Low (1.03.0) PPVb (%)of ELISA ODratio> 5.0 (95%CI)

Enzygnost Positive 136 0 0 100(97.3100)

Negative 0 0 0

Indeterminate 0 0 0

Vironostika Positive 136 0 0 99.3(97.8100)

Negative 0 1 0

Indeterminate 0 0 3

a All serareactive byat least one ofthe ELISA.b Thepositivepredictive value (PPV) of an ELISA. Optical density (OD)ratio wasdefined as the sumof Western blot-confirmedpositivesera andELISA reactivesera, divided

by thetotalnumber of sera reactive by theELISA.

3.3. Evaluation of the serial testing algorithm

Ofthe 912seratestedusingDetermine HIV1/2,138(15.1%) were

reactive, and 774 (84.9%) were non-reactive. Of these 138 reactive

sera, 135 were also reactive using Uni-Gold HIV-1/2. Upon further

testing by WB of the three samples with discordant results, one

was confirmed positive by WB and two were indeterminate but

subsequently tested negative by HIV-1 DNA PCR (Fig. 1B) resulting

in sensitivity of 99.7% (95% CI: 98.899.8%) and 99.9% negative pre-dictive value. Of the 776 non-reactive sera, 774 were concordantly

non-reactive in the serial algorithm,resultingin specificityof 98.9%

(95% CI, 97.9199.84%); 98.9% positive predictive value.

3.4. Quality assurance

As part of a quality control assessment, 10% (77) of the 776 non-

reactive sera were randomlyselectedfor testing byWB andall were

confirmed negative. In addition, 10% (14) of all reactive sera were

similarly randomly selected for testing by WB and were similarly

confirmed positive.

3.5. Estimated cost associated with parallel and serial testing

algorithms

The cost associated with using the parallel testing algorithm

was compared to that associated with serial testing algorithm. In

Kenya, at the time of this analysis, the cost per sample for Enzyg-

nost and Vironostika was US$2.42 and US$2.12, respectively. The

cost per sample for Determine and Uni-Gold was US$ 1.3 and US$

2.1, respectively. Additional costs per sample are US$35 for West-

ern blot and US$45 for DNA PCR (Tables 1A and 1B). Overall, the

kit costs for testing 912 samples was US$ 4,327.32 (US $ 4.74 per

sample) in the parallel testing algorithm and US$1,746.20 (US$ 1.9

per sample) in the serial testing algorithm. Thus, a substantial cost

saving of 61% was realized using the serial versus the parallel algo-

rithm. The results were recorded each day within 1 h for the serial

testing algorithm. However, an ELISA-based parallel testing algo-rithm required 6 h to produce the final results and it typically takes

10 working days to transcribe and report results to the patients.

4. Discussion

The cost of HIV serologic screening is challenging, especially in

resource-limited countries with high HIV prevalence and for labo-

ratories that still useexpensive Western blot as a confirmatory test.

With interest in widespread testing, particularly in countries with

high HIV prevalence, the cost of testing has significantly increased.

In these resource-limited settings, the use of WB has limited the

expansion of testing programs. Use of alternative HIV testing algo-

rithms that do not rely on WB confirmation could substantially

reduce the cost of HIV diagnostics. In this study, six commercially

availableELISAs andrapid tests were evaluatedand those with high

sensitivities andspecificities identifiedallowingtheir use in a paral-

lelor serialtesting algorithm. This analysis shows that both parallel

and serial HIV testing algorithms have a high level of accuracy

for diagnosing HIV infection, and this reduces the need for sup-

plemental confirmatory assays. In the retrospective analysis, any

combination of ELISA or rapid tests, excluding Murex and Capillus,

qualified for both parallel and serial testing algorithms. Murex was

excludeddue toits lowspecificityandhigh falsepositiveresultsandhas previously been reported with high misclassification among

East African subjects when evaluating its performance (Everett

et al., 2007) andhenceexcludedfromthe prospective phase of anal-

ysis. Despite having high sensitivity and specificity, Capillus was

excluded from the prospective analysis due to erratic local supply

and high cost in conducting the assay.

Both ELISAs included in the parallel algorithm showed excel-

lent performance including evaluated specificity, sensitivity and

delta values. High values indicate the capacity of an assay to

consistently produce high OD ratios for reactive sera and low OD

ratios for non-reactive sera thereby enhancing the confidence put

in the sensitivity and specificity for the ELISA (Andersson et al.,

1997; Nkengasong et al., 1999). Results of the retrospective eval-

uation, showed a strong correlation between high positive andnegative values for quantitative level of OD reactivity and speci-

ficity (Table 2). Excellent ELISA sensitivities, specificities, high

values and positive predictive values for moderate or high OD

reactivity suggest that a combination of ELISAs with different char-

acteristics in a parallel testing algorithm provide an efficient and

cost effective alternative to WB as a reliable algorithmfor serodiag-

nosis of HIV infection. In this case, Enzygnost utilizes the synthetic

protein gp41 while Vironostika utilizes the recombinant proteins

gp41 and p24 antigens. Furthermore, the two tests differ in their

test principles with indirect andsandwich principles forEnzygnost

and Vironostika respectively.

The 13 indeterminate specimens by WB in the retrospective

evaluation were excluded from the calculation of sensitivity and

specificity since WB has traditionally been the gold standardagainst which HIV assays are compared. It should be noted, how-

ever, that false-positive WB have been reported (Kleinman et al.,

1998; Willman et al.,2001). Thesefindings showthat a combination

of paired ELISA or rapid tests algorithm (Tables 1A and 1B) based

on different principles can provide results as accurate as those for

standard testing algorithm for serodiagnosis of HIV infection and is

in accordance with other studies (Philiand Vardas, 2002; Wilkinson

et al., 1997).

These results are comparable to earlier findings in various set-

tings that have unequivocally shown that a combination of ELISAs

and rapid tests in a parallel or serial testing algorithm can produce

accurate results for diagnosing HIV infections (Nkengasong et al.,

1999; Phili and Vardas, 2002; Wilkinson et al., 1997). The choice of

assay depends entirely on the sensitivity and specificity while the


7/8


8/8


Koch, W.H., Sullivan, P.S., Roberts, C., Francis, K., Downing, R., Mastro, T.D., Nken-gasong, J., Hu, D., Masciotra, S., Schable, C., Lal, R.B., 2001. Evaluation of UnitedStates-licensed human immunodeficiency virus immunoassays for detection ofgroup M viral variants. J. Clin. Microbiol. 39, 10171020.

Louie, B., Pandori,M.W., Wong,E., Klausner, J.D.,Liska, S.,2006. Useof anacute sero-conversion panel to evaluate a third-generation enzyme-linked immunoassayfor detection of human immunodeficiency virus-specific antibodies relative tomultiple other assays. J. Clin. Microbiol. 44, 18561858.

Ly,T.D., Ebel, A.,Faucher, V.,Fihman,V., Laperche, S.,2007.Could thenew HIVcom-bined p24 antigen and antibody assays replace p24 antigen specific assays? J.Virol. Methods 143, 8694.

Lyamuya, E.F., Aboud, S., Urassa, W.K., Sufi, J., Mbwana, J., Ndugulile, F., Massambu,C., 2009. Evaluation of simple rapid HIV assays and development of nationalrapid HIVtest algorithms in Dar es Salaam. Tanzan. BMC Infect. Dis. 9, 19.

Meda, N., Gautier-Charpentier, L., Soudre, R.B., Dahourou, H., Ouedraogo-Traore,R., Ouangre, A., Bambara, A., Kpozehouen, A., Sanou, H., Valea, D., Ky, F., Car-toux, M., Barin, F., Van de Perre, P., 1999. Serological diagnosis of humanimmuno-deficiency virus in BurkinaFaso:reliable, practicalstrategiesusinglessexpensive commercial test kits. Bull. World Health Organ. 77, 731739.

Mylonakis, E., Paliou, M., Lally, M., Flanigan, T.P., Rich, J.D., 2000. Laboratory testingfor infection with the human immunodeficiency virus: established and novelapproaches. Am.J. Med. 109, 568576.

NASCOP, 2001. National AIDS and STD Control Programme. NationalGuidelines for Voluntary Counseling and Testing. Nairobi.www.drh.go.ke/documents/National%20VCT%20guidelines.pdf. (accessed22.03.2011).

Nguyen, K.A., Busch, M.P., 2000. Evolving strategies for diagnosing human immun-odeficiency virus infection. Am.J. Med. 109, 595597.

Nkengasong, J.N., Maurice, C., Koblavi, S., Kalou, M., Yavo, D., Maran, M., Bile, C.,Nguessan, K., Kouadio,J., Bony, S.,Wiktor, S.Z., Greenberg, A.E., 1999. Evaluation

ofHIV serial and parallel serologic testing algorithms in Abidjan, Cote dIvoire.AIDS 13, 109117.

Nunn, A.J., Biryahwaho, B., Downing, R.G., Van der, G.G., Ojwiya, A., Mulder, D.W.,1993.Algorithmsfor detectingantibodiesto HIV-1:resultsfrom a rural Ugandancohort. AIDS 7, 10571061.

Owen, S.M., Yang, C., Spira, T., Ou, C.Y., Pau, C.P., Parekh, B.S., Candal, D., Kuehl, D.,Kennedy, M.S., Rudolph, D., Luo, W., Delatorre, N., Masciotra, S., Kalish, M.L.,Cowart, F., Barnett, T., Lal, R., McDougal, J.S., 2008. Alternative algorithms forhuman immunodeficiency virus infectiondiagnosis using tests thatare licensedin the United States. J. Clin. Microbiol. 46, 15881595.

Pandori, M.W., Hackett Jr., J.,Louie, B.,Vallari,A., Dowling, T., Liska, S.,Klausner, J.D.,2009. Assessment of the ability of a fourth-generation immunoassay for humanimmunodeficiency virus (HIV) antibody and p24 antigen to detect both acuteandrecentHIV infectionsin a high-risksetting. J. Clin.Microbiol.47, 26392642.

Parry, J.V., Mortimer, P.P., Perry, K.R., Pillay, D., Zuckerman, M., 2003. Towardserror-free HIVdiagnosis:guidelines on laboratory practice.Commun. Dis.PublicHealth 6, 334350.

Phili, R.,Vardas, E.,2002.Evaluation of a rapid humanimmunodeficiency virus testat two community clinics in Kwazulu-Natal. S. Afr. Med. J. 92, 818821.

Razafindratsimandresy, R., Hollanda, J., Soares, J.L., Rousset, D., Reynes, J.M., 2007.HIVtype 1 diversityin the Seychelles. AIDSRes. Hum.Retroviruses 23,761763.

Razafindratsimandresy, R., Rajaonatahina, D., Soares, J.L., Rousset, D., Reynes, J.M.,2006. High HIV type 1 subtype diversity and few drug resistance mutationsamong seropositive people detected during the 2005 second generation HIVsurveillance in Madagascar. AIDS Res. Hum. Retroviruses 22, 595597.

Respess, R.A., Rayfield, M.A., Dondero, T.J., 2001. Laboratory testing and rapid HIVassays: applicationsfor HIV surveillance in hard-to-reachpopulations. AIDS 15,S49S59.

Silverstein, D.M., Aviles, D.H., Vehaskari, V.M., 2004. False-positive human immun-odeficiency virus antibody test in a dialysis patient. Pediatr. Nephrol. 19,547549.

Sommerfelt, M.A., Ohlsson, I., Flolid, I., Thorstensson, R., Sorensen, B., 2004. A sim-ple semi-rapid HIV-1&2confirmatory immunoassay using magneticparticles. J.Virol. Methods 115, 191198.

Stramer,S.L., Glynn, S.A., Kleinman,S.H., Strong, D.M., Caglioti,S., Wright, D.J.,Dodd,R.Y., Busch, M.P., 2004. Detectionof HIV-1 and HCV infections among antibody-negativeblood donors by nucleic acid-amplificationtesting.N. Engl. J. Med.351,760768.

Tenenbaum, S.A., Morris, C.A., Alexander, S.S., McFerrin, H.E., Garry, R.F., Leissinger,C.A., 2005. Evidence of HIV exposure and transient seroreactivity in archivedHIV-negative severe hemophiliac sera. Virol. J. 2, 65.

Urassa, W., Godoy, K., Killewo, J., Kwesigabo, G., Mbakileki, A., Mhalu, F., Biberfeld,G., 1999. The accuracy of an alternative confirmatory strategy for detection ofantibodies to HIV-1: experience from a regional laboratory in Kagera. Tanzan.J.

Clin. Virol. 14, 2529.Wilkinson,D.,Wilkinson, N.,Lombard,C., etal., 1997.On-siteHIVtestingin resource-

poor settings: is onerapid test enough? AIDS 11, 377381.Willman, J.H., Hill, H.R., Martins, T.B., Jaskowski, T.D., Ashwood, E.R., Litwin, C.M.,

2001. Multiplexanalysisof heterophil antibodiesin patientswithindeterminateHIV immunoassay results. Am. J. Clin. Pathol. 115, 764769.

World Health Organization, 1997. Revised Recommendation for the Selection anduse of HIVantibody tests. Wkly. Epidemiol. Rec. 20, 145152.

World Health Organization, 1998. The importance of simple/rapid assays in HIVtesting. Wkly. Epidemiol. Rec. 73, 321328.

Yang, C.,Li, M.,Cowart,F., Rudolph,D., Parekh,B.S., McDougal, J.S., Lal,R., 2004. Char-acterization of human immunodeficiency virus type-1 from HIV-1 seropositivecases with undetectable viremia. J. Clin. Virol. 30, 224228.
http://www.drh.go.ke/documents/National%2520VCT%2520guidelines.pdfhttp://www.drh.go.ke/documents/National%2520VCT%2520guidelines.pdfhttp://www.drh.go.ke/documents/National%2520VCT%2520guidelines.pdf

17.zeh (2011) commercial elisa testing in hiv

Documents