T R A N S F U S I O N C O M P L I C A T I O N S

Impact of changes in viral marker screening assays

Usha K. Sharma, Susan L. Stramer, David J. Wright, Simone A. Glynn, Sigurd Hermansen,

George B. Schreiber, Steven H. Kleinman, and Michael P. Busch

for the Retrovirus Epidemiology Donor Study

BACKGROUND: Monitoring the performance of rou-tinely used infectious disease serologic tests is neces-sary to evaluate their effectiveness in identifying true-positive units and erroneously disqualifying safe blooddonors.METHODS: With two large screening test data sets col-lected between 1991 and 1998 and between 1997 and2000, the impact of changes in screening assays forHIV, HCV, and HBsAg was analyzed with regard to theprevalence of confirmed-positive, indeterminate, andconfirmed-negative results and the deferral of donorswith an indeterminate or negative results (donor loss).RESULTS: The prevalence of indeterminate results anddonors loss increased significantly in the 6 months afterintroduction of an HIV-1/2 EIA. A second-generationHCV EIA increased the detection of confirmed-positivedonations in repeat donors (p < 0.001) and increasedthe prevalence of indeterminate donations. Implementa-tion of a third-generation HCV EIA resulted in a signifi-cant decrease in indeterminate results in first-time do-nors. Nonspecific test results increased when HBsAgtest kits from a different manufacturer were introducedor different lots of HIV antibody screening test kits fromthe same manufacturer were used.CONCLUSION: Introduction of newly licensed versionsof assays, switching kit manufacturers, and lot-to-lotvariations have an impact on rates of deferrals of safedonors as well as sensitivity of routine screening. Be-fore considering changes in screening tests, blood cen-ters should be aware of, and evaluate, the potential im-pact on donor loss.

Since 1985, the safety of the blood supply hasbeen greatly improved with the introduction ofscreening tests for HIV-1, HTLV-I and -II, andHCV.1 First-generation screening tests for these

agents have been replaced by assays with increased sen-sitivity, resulting in earlier detection of infected donorsand improved detection of donors with immunologicallydivergent strains of a particular virus.2 However, in-creased sensitivity has often been associated with de-creased specificity and consequently an increased rate offalse-positive test results. During the 1990s, several re-ports have raised concerns over the positive-predictivevalue and the false-positive rates of some screeningtests.3-5

False-positive screening test results in otherwisehealthy blood donors have been attributed to the pres-ence of cross-reactive circulating antigens and antibod-ies. Many of the early screening assays used viral lysatepreparations contaminated with other cellular proteinsresulting in false-positive test results.6 In a case-controlstudy conducted at nine blood centers, a history of influ-

ABBREVIATIONS: ARC = American Red Cross; FT = first-time;

neotest = 1 to 6 months after the new screening test was intro-

duced; posttest = 7 to 12 months after the new screening test

was introduced; pretest = the 6-month period before introduc-

tion of the new screening test; REDS = Retrovirus Epidemiol-

ogy Donor Study; RPT = repeat; RR = repeat-reactive; WB =

Western blot.

From Westat, Rockville, Maryland; the National Confirmatory

Testing Laboratory, American Red Cross, Gaithersburg, Mary-

land; and the Blood Centers of the Pacific, San Francisco, Cali-

fornia.

Address reprint requests to: Usha K. Sharma, Westat,

1650 Research Boulevard, Rockville, MD 20850; e-mail:

ushasharma@westat.com.

Supported by NHLBI Contracts N01-HB-97077 (super-

seded by N01-HB-47114), N01-HB-97079, N01-HB-97080, N01-

HB-97081, and N01-HB-97082.

Received for publication June 21, 2002; revision received

September 16, 2002, and accepted September 23, 2002.

TRANSFUSION 2003;43:202-214.

202 TRANSFUSION Volume 43, February 2003

enza vaccination, acute illness, and allergies within 6months before blood donation was reported to be asso-ciated with false-positive screening test results.7 Further,test specificities have been reported to vary with thepopulations being tested, the assay chosen, the testmanufacturers, and minor batch-to-batch variations ofthe same test kit. For example, when comparing HCV testresults, an approximate twofold difference in specificitywas observed when two test systems were compared(PRISM, Abbott Laboratories, Abbott Park, IL; and Ortho3.0, Ortho Clinical Diagnostics, Raritan, NJ).8

While increases in test sensitivity are highly desirablefor ensuring recipient safety, an undesirable conse-quence may be increased donation loss because of false-positive test results. In situations where test kits arechanged based on considerations of cost rather than in-creases in sensitivity, an increased loss of safe donorsmay totally negate the benefit of switching assays. Nu-merous studies have found that donors who have repeat-reactive (RR) test results by screening assays but arenegative or indeterminate upon further testing are usu-ally not infected with the viral agent in question.4,5,9-11

However, donations obtained from these donors are dis-carded and the donors are deferred from donating blood.In the US, each year, a large number of eligible donors arelost because of apparently false-positive screening testresults.2,12 False-positive screening test results have alsobeen reported to cause psychological distress and confu-sion because the clinical significance of these results issometimes not known or, if known, may not be under-stood by the donor.13-15

With data collected between 1991 and 1998 by theRetrovirus Epidemiology Donor Study (REDS), a mul-tidisciplinary research study sponsored by the NationalHeart, Lung, and Blood Institute and data collected be-tween 1997 and 2000 by the American Red Cross (ARC),we studied the impact on confirmed-positive and false-positive rates associated with the introduction of newand more sensitive screening tests, the use of test kitsfrom different manufacturers, and lot-to-lot variation inscreening test kits from the same manufacturer.

MATERIALS AND METHODSImpact of introducing new versions ofscreening testsData on community whole-blood donations have beencollected since 1991 at five REDS blood centers: the ARCBiomedical Services in the Greater Chesapeake and Po-tomac regions (Baltimore, MD; and Washington, DC),Southeastern Michigan (Detroit, MI), and Southern Cali-fornia (Los Angeles, CA); the Blood Centers of the Pacific(San Francisco, CA); and the Oklahoma Blood Institute(Oklahoma City, OK). These centers provide informationon donation type, donation dates, donor demographics

(age, sex, race and ethnicity, education level, country ofbirth), and serologic screening and confirmed test resultsto the study coordinating center (Westat, Rockville, MD)for compilation of a central database. The REDS protocolwas approved by the Institutional Review Board at eachcenter.

To study the impact of introducing new versions ofscreening tests, HIV and HCV antibody screening testdata collected between 1991 and 1998 from the five REDSblood centers were evaluated. All five blood centers usedthe same screening tests for HIV antibodies and madechanges in the types of kits used either simultaneously oraround the same period. For screening, HIV-1 EIA kits(Abbott) were used until March 1992, when the bloodcenters changed to the newly licensed and more sensitiveHIV-1 and HIV-2 EIA (Abbott HIV-1/2 EIA) kits. At four ofthe five blood centers, HCV antibody screening with afirst-generation HCV EIA (Ortho) was replaced by an-other first-generation HCV EIA (Abbott) in the spring of1991. In early spring 1992, the centers changed to thenewly licensed second-generation HCV EIA (Ortho). Thistest was replaced by a third-generation HCV EIA (Ortho)in June 1996. For HCV, screening data from only fourblood centers were included. Data from the OklahomaBlood Institute were excluded from the HCV analysis be-cause this center followed a slightly different HCV testingalgorithm during this period.

For HIV-1, a Western blot (WB) test (Cambridge Bio-tech, Rockville, MD) was used as confirmatory assaythroughout the study period. In addition, after an HIV-1/2 EIA was implemented, HIV-2 EIA and WB (unli-censed) were used as needed. For HCV, the unlicensedsecond-generation RIBA (Chiron Corporation, Emery-ville, CA) was used as a confirmatory test from May 1991through June 1993. A licensed version of this assay wasused from July 1993 to June 1996. Coincident with theimplementation of a third-generation HCV EIA in June1996, confirmatory testing was subsequently conductedwith the unlicensed third-generation HCV RIBA (Chiron).

To examine the impact of HIV and HCV antibodyscreening test changes, data were analyzed from donorswho had given allogeneic whole-blood donations from1991 to 1998. More specifically, we compared the preva-lence of RR screening test results that were confirmatory-test-positive, confirmatory-test-negative, and indetermi-nate during the 6-month period before introduction ofthe new screening test (pretest) to similar data obtained1 to 6 months (neotest) and 7 to 12 months (posttest)after the new screening test was introduced.

Monthly and 6-month prevalence estimates of sero-positive, indeterminate, and false-positive (screening testreactive but confirmatory negative) results were calcu-lated per 105 screened donations. Safe donor loss wasdefined as the number of donors with a RR screening testresult and either a negative or an indeterminate confir-

IMPACT OF NEW SCREENING TESTS

Volume 43, February 2003 TRANSFUSION 203

matory test result. These measures were calculated for alldonors and for first-time (FT) and repeat (RPT) donorsseparately.

Impact of introducing test kits fromdifferent manufacturersHBsAg screening data from the ARC collected betweenJanuary 1997 and September 1998 were evaluated to ana-lyze the impact of switching manufacturer of the test kits.

During this period, approximately 10.8 � 106 dona-tions were screened for HBsAg. Screening test kits fromAbbott, in use from January 1997 through June 1997, werereplaced by kits from Ortho in July 1997. A HBsAg neu-tralization assay (Abbott) was used to confirm sampleswith RR screening test results from either manufacturer.During the Ortho test period, RR samples were retestedby the Abbott screening test; concordant RR sampleswere tested by an Abbott neutralization assay. During thisperiod, all donors with RR test (Ortho) results were de-ferred. The prevalence of RR screening test results andconfirmatory-test-positive (neutralizable) and -negativeresults during the period each test kit was in use (6 and 15months, respectively) was compared.

Impact of lot-to-lot changes in screening test fromthe same manufacturerBetween May 2000 and August 2001, approximately 9.02� 106 donations were screened by an HIV-1/2 EIA (Ab-bott) and an HIV-1 p24 antigen EIA (Coulter Corporation,Miami, FL) at the ARC. A new lot of the test kits was usedat approximate monthly intervals. RR test results ob-tained with the HIV-1/2 EIA screening test were con-firmed by HIV-1 WB (Cambridge Biotech). In addition, allHIV-1 WB-negative and -indeterminate samples weretested by HIV-2 EIA and WB (unlicensed). Samples thatwere RR for the HIV-1 p24 antigen were tested by anHIV-1 p24 antigen neutralization assay (Coulter).Monthly prevalence estimates were calculated.

Statistical analysisWith ANOVA,16 we compared the monthly prevalence ofconfirmed-positive, indeterminate, and confirmed-negative results and evaluated whether these prevalenceestimates changed before and after implementation of anew screening test or before and after the introduction ofa test kit from a different manufacturer. An F test wasused to compare prevalence in the three periods (pretest,neotest, and posttest), assuming that the marker preva-lence during the period of observation was constant.

It should be noted that confirmed-positive preva-lence rates were not expected to change significantly overthe three periods for two reasons: First, the sensitivity ofscreening tests was already very high, and a new test, newkit, or new lot was unlikely to affect the sensitivity enough

to be detectable in our observational study. Second, thepopulation prevalence and incidence rates were expectedto be relatively constant over an 18-month period, andany change would be masked by variation in the monthly“sample” of donors.

RESULTSImpact of introducing new versions ofscreening test

HIV. The impact of replacing the HIV-1 EIA with theHIV-1/2 EIA as a screening test in March 1992 was anincreased overall rate of RR and confirmed-positivescreening test results in FT donors when compared toRPT donors (Table 1 and Fig. 1). As expected, the preva-lence of HIV-antibody-seropositive donations did not dif-fer significantly (p > 0.05); that is, there was no demon-strable change in the yield of HIV-infected donors, whenresults of the HIV-1 EIA during the pretest period werecompared to the HIV-1/2 EIA in the neotest and posttestperiods (Table 1). However, the prevalence of indetermi-nate results (in all donors and for FT and RPT donorsseparately) increased significantly in the 12-month pe-riod after implementation of the HIV-1/2 EIA. The preva-lence of confirmed-negative results also varied signifi-cantly during the 18-month period of evaluation (in alldonors, p < 0.01; Fig. 1), being significantly higher in FTdonors during the first 1 to 6 months (neotest) after theintroduction of the HIV-1/2 EIA (p = 0.03; Table 1) anddecreasing back to pretest levels in the posttest period(p = 0.6).

Safe donor loss increased significantly during thefirst 6 months after implementation of the HIV-1/2 EIA(p � 0.05; Table 1) but returned to baseline 7 to 12months posttest implementation.

HCV. The proportion of donations that were con-firmed positive by the second-generation HCV EIA in theneotest period was significantly higher than the propor-tion of donations that were confirmed positive by thefirst-generation HCV EIA (228.5 vs. 172.1 per 105 dona-tions, respectively; p = 0.01; Table 2 and Fig. 2). Stratifiedby FT and RPT donors, the number of confirmed-positiveresults in FT donors during the neotest and posttestperiods increased substantially, compared to the pretestperiod (Table 2). However, the difference was not signifi-cant (p = 0.4). The confirmed-positive detection of HCVantibodies in donations from RPT donors increased sig-nificantly (p < 0.001) during the neotest period of obser-vation. However, the prevalence of donations that wereconfirmed positive for HCV antibodies (all donors) dur-ing the posttest period decreased to a level similar to thatobserved during pretest period (175.6 vs. 172.1 per 105

donations, respectively; Table 2).While a significant decrease was observed in the

number of donations that were confirmed negative for

SHARMA ET AL.

204 TRANSFUSION Volume 43, February 2003

HCV antibodies after implementation of the second-generation HCV EIA (neotest and posttest periods), therewas a significant increase in the number of indeterminatedonations (in both FT and RPT donors, p < 0.001; Table2). Therefore, no significant change was observed in safedonor loss in the 6 months (neotest period) after the sec-ond-generation HCV EIA was introduced. However, thesafe donor loss because of nonspecific HCV antibodyscreening test results decreased significantly (p = 0.001;Table 2) in the posttest period for both FT and RPT do-nors.

A third-generation HCV EIA was implemented as ascreening test in June 1996. At this time, a switch alsooccurred in confirmatory tests, with the third-generationRIBA replacing the second-generation RIBAs. Althoughthe prevalence of confirmed-positive donations in-creased with the implementation of the third-generationHCV EIA, the difference was not significant (Table 3).While the proportion of indeterminate donations amongFT donors decreased significantly (p < 0.05) following thechange in tests, the frequency of indeterminate or nega-tive results in RPT donors did not show a significant dif-ference when the performance of the second-generationHCV EIA was compared to that of the the third-generation HCV EIA (p > 0.05; Table 3). The prevalence ofdonations that were confirmed positive or confirmednegative for HCV antibodies did not change significantly

in the 12 months (neotest and posttest) after the imple-mentation of the third-generation HCV EIA (Table 3 andFig. 3).

Impact of introducing test kits fromdifferent manufacturersOverall, 10.8 � 106 donations were screened by the ARCfor HBsAg between January 1997 and September 1998.During this period, 7239 donations were RR for HBsAg; ofthese approximately 34 percent (22.9 per 105 screened)were confirmed positive for antibodies against HBsAg.The rate of RR screening test results increased from 33.7per 105 to 80.2 per 105 when the test kits from one manu-facturer were replaced by kits obtained from anothermanufacturer (p < 0.001). However, the increased pro-portion of donations that showed RR screening test re-sults was not associated with an increase in proportion ofHBsAg confirmed-positive donations, which remainedrelatively constant from January 1997 to September 1998(approximately 23.8 per 105 donations tested, p = 0.3; Fig.4). While 75 percent of donations that had RR HBsAgscreening test results (Abbott) from January 1997 to June1997 were confirmed positive for antibodies against HB-sAg, only 27.4 percent of donations that had RR screeningtest (Ortho) results from July 1997 to September 1998were confirmed positive for antibodies against HBsAg.

TABLE 1. Impact of a change from HIV-1 EIA to HIV-1/2 EIA as a screening test in whole-blood donors

Donations

Donors

All FT RPT

Pretest Neotest Posttest Pretest Neotest Posttest Pretest Neotest Posttest

Total number ofdonationsscreened(×105)

4.96 4.99 4.88 1.03 0.99 0.94 3.93 4.00 3.94

Total number ofRR testresults

474 722 396 122 195 125 352 527 271

Prevalence (per105) of RRtest results

95.5 144.4 81.2 118.6 196.5 132.4 89.4 131.5 68.8

RR screening testresults andconfirmed*

Positive 47 (9.9) 43 (5.9) 31 (7.8) 31 (25.4) 32 (16.4) 23 (18.4) 16 (4.5) 11 (2.1) 8 (2.9)Negative 346 (72.9) 513 (71.1) 208 (52.5) 71 (58.2) 122 (62.6) 52 (41.6) 275 (78.1) 391 (74.2) 156 (57.6)Indeterminate 76 (16.0) 164 (22.7) 154 (38.9) 20 (16.4) 41 (21) 48 (38.4) 56 (15.9) 123 (23.3) 106 (39.1)Missing 5 (1.1) 2 (0.3) 3 (0.8) 0 (0.0) 0 (0.0) 2 (1.6) 5 (1.4) 2 (0.4) 1 (0.4)

Prevalence per105† p value p value p value p value p value p valueSeropositive 9.5 8.6 0.8† 6.3 0.2 30.1 32.2 0.7 24.3 0.7 4.1 2.7 0.3 2.0 0.1Negative 69.7 102.6 0.1 42.6 0.1 69.0 122.9 0.03 55.1 0.6 69.9 97.6 0.1 39.6 0.1Indeterminate 15.3 32.8 0.005 31.5 0.01 19.4 41.3 0.04 50.9 0.01 14.2 30.7 0.01 26.9 0.03

Number of donorslost per 105

85.0 135.4 0.03 74.1 0.6 88.5 164.2 0.02 106.0 0.5 84.1 128.3 0.05 66.5 0.4

* Number in parentheses represents the percentage of total RR donations.† p value for F test compares prevalence in the neotest and posttest periods to prevalence in the pretest period (September 1992 through February

1992).

IMPACT OF NEW SCREENING TESTS

Volume 43, February 2003 TRANSFUSION 205

Thus, the increased screening test reactivity only repre-sented increased false reactivity, resulting in increaseddonor loss estimates from 8.4 per 105 to 58.2 per 105

during the later period (Fig. 4). The use of the Abbott

screening test on samples that were RR by the Ortho testbefore neutralization testing eliminated greater than 95percent of the number of donations requiring neutraliza-tion testing (data not shown).

Fig. 1. Prevalence of HIV antibody RR (�), confirmed-positive (�), indeterminate (�), and confirmed-negative (�) test (HIV-1

and HIV-1/2 EIA) results.

SHARMA ET AL.

206 TRANSFUSION Volume 43, February 2003

Impact of lot-to-lot changes in screening test fromthe same manufacturerTo evaluate the variation in results when different test kitlots from the same manufacturer were used, the results ofHIV antibody screening were analyzed with different lotsof anti-HIV-1/2 EIA kits and HIV-1 p24 antigen test kitsfrom May 2000 through August 2001 (Figs. 5 and 6). Lotchanges occurred approximately monthly during this pe-riod.

No trend was apparent when the results of the HIV-1/2 EIA over the first 8 months (May-December 2000)were evaluated (Fig. 5). However, from January 2001through May 2001, a steady increase in screening testreactivity was observed with a corresponding increase inthe rate of WB-negative and -indeterminate donations.This nonspecific reactivity decreased with the change inlots after June 2001 (Fig. 5). While the prevalence of HIVantibody confirmed-positive donations did not show anysignificant change (as to be expected) during this period,the safe donor loss varied between 28.4 and 110.8 per 105

donations screened per month.From May 2000 through November 2000, the

monthly HIV-1 p24 antigen RR donation rates were foundto show minor variation (between 22 and 41 per 105). Anabrupt increase, however, in the number of HIV-1 p24antigen RR donations was observed during the 4-monthperiod from December 2000 to March 2001 (Fig. 6). Witha change in test kit lot after March 2001, the monthlyrates of HIV-1 p24 antigen RR donations decreased (Fig.6), but did not return to the levels observed before De-

cember 2000. The vast majority (97.9%) of HIV-1 p24 an-tigen RR samples did not show evidence of HIV infectionupon further testing with p24 neutralization, RNA assays,or HIV antibody assays. The rates of HIV-1 antibody con-firmed-positive donations remained constant over thisperiod of time (data not shown). Because of the high rateof indeterminate test results, the safe donor loss variedbetween 22.5 and 85.9 per 105 during the 16-month pe-riod of observation.

DISCUSSION

A significant number of safe donations are removed fromthe blood supply because of reactive screening test re-sults, but subsequent confirmatory test results are inde-terminate or negative.3,5,17 In the US, such results are thebasis for indefinite deferral of a large number of volunteerdonors.18 Data from different blood centers in the USindicate that each year approximately 500,000 individualsare informed of a reactive infectious disease screeningtest results, of which approximately 15 percent haveHIV-1 antibodies, 1.5 percent have antibodies againstHBsAg, and 18 percent show HCV antibodies.2 Differentstudies have reported that less than 1 percent of HIV-1p24 antigen-reactive, antibody-negative donors and lessthan 10 percent of the donors with RR results for HIVantibody testing are truly infected,3,4,19,20 while 56 and 69percent of the donors with RR test results for HCV anti-body and antibodies against HBsAg, respectively, arefound positive on confirmatory testing.11 Indeterminate

TABLE 2. Impact of a change from a first-generation HCV EIA to a second-generation HCV EIA as a screening testin whole-blood donors

Donations

Donors

All FT RPT

Pretest Neotest Posttest Pretest Neotest Posttest Pretest Neotest Posttest

Total number of

donations

screened (×105)

4.25 4.26 4.15 0.92 0.88 0.81 3.33 3.39 3.34

Total number of RR

test results

2129 2074 1386 899 890 704 1230 1184 682

Prevalence (per 105)

of RR test results

501.2 486.5 334.3 979.6 1016 870.5 369.3 349.6 204.4

RR screening test

results and

confirmed*

Positive 731 (34.3) 974 (46.9) 728 (52.5) 451 (50.2) 541 (60.8) 448 (63.6) 280 (22.8) 433 (36.5) 280 (41.1)

Negative 1307 (61.4) 486 (23.4) 248 (17.9) 403 (44.8) 152 (17.1) 100 (14.2) 904 (73.5) 334 (28.2) 148 (21.7)

Indeterminate 76 (3.6) 588 (28.4) 394 (28.4) 40 (4.5) 185 (20.8) 152 (21.5) 36 (2.9) 403 (34) 242 (35.5)

Missing 15 (0.7) 26 (1.3) 16 (1.2) 5 (0.6) 12 (1.3) 4 (0.6) 10 (0.8) 14 (1.2) 12 (1.8)

Prevalence per 105† p value p value p value p value p value p value

Seropositive 172.1 228.5 <0.01† 175.6 0.8 491 617.6 0.2 553.9 0.6 84.1 127.9 <0.001 83.9 1.0

Negative 307.7 114.0 <0.001 59.8 <0.001 439 173.5 <0.001 123.6 <0.001 271.4 98.6 <0.001 44.3 <0.001

Indeterminate 17.9 137.9 <0.001 95 <0.001 43.6 211.1 <0.001 187.9 <0.001 10.8 119.0 <0.001 72.5 <0.001

Number of donors lost

per 105325.6 252.0 0.1 154.8 0.001 482.7 384.7 0.1 311.6 0.01 282.3 217.6 0.1 116.9 <0.001

* Number in parentheses represents percentage of total RR donations.† p value for F test compares prevalence in the neotest and posttest periods to prevalence in the pretest period (September 1992 through February 1992).

IMPACT OF NEW SCREENING TESTS

Volume 43, February 2003 TRANSFUSION 207

Fig. 2. Prevalence of HCV antibody screening test (first- and second-generation EIA) RR (�), confirmed-positive (�), indetermi-

nate (�), and negative (�) results.

SHARMA ET AL.

208 TRANSFUSION Volume 43, February 2003

HIV and HCV antibody test results have only rarely beenassociated with infection, and in the case of HCV, thisoccurred primarily with the older versions of RIBA.4,5,9-11

Notification of indeterminate results has been reportedto cause psychological distress because the cause, andtherefore the clinical significance, of these results is oftennot known; consequences are negative social effects in-cluding disruptions to family and work, the costs of medi-cal follow-up, and the effect of this information on em-ployment or insurability.13-15 Other potentially damagingoutcomes of indeterminate and false-positive infectiousdisease screening results are the loss of otherwise quali-fied blood donors and all their subsequent donationsgiven the lack of an acceptable reentry algorithm.

Several factors may be associated with screening-test-reactive but confirmatory-test-negative and -inde-terminate results. A history of allergy, acute illness, oralloimmunization; the presence of autoimmune antibod-ies; or vaccination may result in false-reactive test re-sults.7,21,22 Increased prevalence of HIV antibody screen-ing-test-reactive but unconfirmed results have beenreported among multiparous women, younger agegroups, black people, and FT donors; increased preva-lence of HCV antibody screening-test-reactive but un-confirmed results have been found in older donors, indonors with a history of transfusion, and in FT donors.23

Unresolved screening test specificity is responsiblefor the increased frequency of lost donors. While themore sensitive HIV EIAs are helpful in detecting earlyseroconverting donors, an increase in nonspecific reac-

tivity results in increasing rates of WB-indeterminate do-nations.19,24 In contrast, some of the more specific HIVantibody assays based on synthetic peptides or recombi-nant proteins have proven to be less sensitive than theolder assays based on whole viral lysates.25,26 Our resultsshow that after the introduction of HIV-1/2 EIA, in-creased reactive rates were associated with an increase inconfirmatory-test-negative and -indeterminate results.Thus, enhanced assay sensitivity did not result in a com-parable improvement in specificity.

An increase in confirmatory-test-negative and -inde-terminate test results, when the less sensitive screeningtests were replaced with more sensitive assays, is alsoevident from our evaluation of HCV antibody screeningtests. To reduce the risk of posttransfusion hepatitis C inrecipients of blood components, blood donor screeningwith a first-generation HCV antibody EIA introduced in1990 was replaced by the highly sensitive second-generation HCV antibody EIA in 1992, which further de-creased posttransfusion hepatitis C.17 Our results showthat the detection of HCV antibody confirmed-positivedonations from RPT donors increased significantly afterthe introduction of the second-generation HCV EIA.These data are best explained by the detection and sub-sequent culling of additional infections among RPT do-nors that had been missed by the first-generation HCVEIA. In contrast, there was a significant increase in thenumber of HCV antibody screening-test-reactive but -in-determinate donations that persisted through the neotestand posttest periods after the introduction of the second-

TABLE 3. Impact of a change from a second-generation HCV EIA to a third-generation HCV EIA as a screening testin whole-blood donors

Donations

Donors

All FT RPT

Pretest Neotest Posttest Pretest Neotest Posttest Pretest Neotest Posttest

Total number ofdonations screened(×105)

4.25 4.01 3.86 0.84 0.70 0.76 3.41 3.31 3.10

Total number of RRtest results

775 735 734 520 447 466 255 288 268

Prevalence (per 105)of RR test results

182.4 183.1 190 618.2 634.2 609.3 74.8 87 86.5

RR screening testresults andconfirmed*

Positive 441 (56.9) 481 (65.4) 422 (57.5) 335 (64.4) 352 (78.7) 335 (71.9) 106 (41.6) 129 (44.8) 87 (32.5)Negative 157 (20.3) 114 (15.5) 159 (21.7) 77 (14.8) 42 (9.4) 65 (13.9) 80 (31.1) 72 (25.0) 94 (35.1)Indeterminate 170 (21.9) 133 (18.1) 149 (20.3) 106 (20.4) 51 (11.4) 65 (13.9) 64 (25.1) 82 (28.5) 84 (31.3)Missing 7 (0.9) 7 (0.9) 4 (0.5) 2 (0.4) 2 (0.4) 1 (0.2) 5 (1.9) 5 (1.7) 3 (1.1)

Prevalence per 105 p value p value p value p value p value p valueSeropositive 103.8 119.8 0.1† 109.2 0.7 398.3 499.4 0.1 438.0 0.6 31.1 38.9 0.1 28.1 0.4Negative 37.0 28.4 0.3 41.1 0.6 91.5 59.6 0.2 85.0 0.8 23.5 21.8 0.8 30.3 0.2Indeterminate 40.0 33.1 0.2 38.6 0.7 126.0 72.3 0.01 85.0 0.03 18.8 24.8 0.2 27.1 0.1

Number of donors lostper 105

77.0 61.5 0.2 79.7 0.8 217.6 131.9 0.02 170.0 0.1 42.3 46.5 0.6 57.4 0.1

* Number in parentheses represents percentage of total RR donations.† p value for F test compares prevalence in the neotest and posttest periods to prevalence in the pretest period (December 1995 through May 1996).

IMPACT OF NEW SCREENING TESTS

Volume 43, February 2003 TRANSFUSION 209

generation HCV EIA. Introduced in 1996, the third-generation HCV EIA has been reported to be either com-parable to or more sensitive than the second-generation

HCV EIA, and it has been reported to detect HCV sero-conversion earlier, as well as detect infections not iden-tified by the earlier assays.27-32 In our study, the sensitiv-

Fig. 3. Prevalence of HCV antibody screening test (second- and third-generation EIA) RR (�), confirmed-positive (�), indetermi-

nate (�), and negative (�) results.

SHARMA ET AL.

210 TRANSFUSION Volume 43, February 2003

ity of the second-generation HCV EIA comparedfavorably to the third-generation HCV EIA as indicated bythe similar rate of RR screening test and confirmed-positive test results during the period of observation. Fur-ther, the incorporation of the third-generation HCV EIAand third-generation RIBA has been demonstrated to re-solve the status of a large proportion of second-generation HCV EIA-RR, second-generation RIBA-indeterminate specimens.28 Our results in FT donors alsoindicate a significant decrease in indeterminate resultsafter the implementation of the third-generation HCVEIA and third-generation RIBA. However, approximately35 percent to 43 percent of the screening-test-RR dona-tions continue to be either confirmatory-test-negative or-indeterminate. Although it has been reported that per-sons with HCV genotype 3 infection and hepatitis C vi-remia may have EIA-reactive but third-generation RIBA-

indeterminate test results, these findings are not likely toapply to the majority of RIBA-indeterminate results in theUS where genotype 3 infection is uncommon.33

Introduction of screening test kits from differentmanufacturers for operational reasons (e.g., cost or con-tractual) may also have an impact on confirmed-positiveand indeterminate results. Cross-reactions to the murineMoAbs being used in most current HBsAg screening testsmay be responsible for a false-positive screening test re-sult.8 The use of murine MoAbs may vary with the manu-facturer, along with the manufacturer’s strategies forblocking murine-nonspecific reactions. After the screen-ing test kits from one manufacturer were replaced bythose from an alternative manufacturer, the rate of de-tection of HBsAg confirmed-positive blood donations re-mained stable; however, the rate of HBsAg screening-test-reactive but confirmatory-test-negative donations

increased and fluctuated markedly.These results further mandate a carefulevaluation of screening test assay per-formance before change.

Lot-to-lot variation in test kits re-agents from the same manufacturermay account for variability in preva-lence of screening-test-reactive butconfirmatory-test-negative and -inde-terminate results over time. Althoughlot changes should not lead to changesin test kits performance, an evaluationof the results of HIV-1 and -2 antibodyand HIV-1 p24 antigen testing showthat quite wide variation of test resultswere likely attributed to changes in re-

Fig. 5. HIV-1/2 EIA screening test results from May 2000 through August 2001.

(�) Confirmed-positive, (�) confirmed-negative, (�) indeterminate, (�) screen-

ing-test-positive.

Fig. 4. Impact of change of HBsAg screening test kits from two manufacturers. (�) Confirmed-positive, (�) screening-test-

positive.

IMPACT OF NEW SCREENING TESTS

Volume 43, February 2003 TRANSFUSION 211

agent lots. Suspected causes for the observed increase inreactive rates included lab-to-lab variation, variation inperformance of equipment, the influence of technolo-gists performing the assay, and the master lots used.While instrumentation and technologists were ruled outas possible causes of the variation observed, the numberof screening-test-reactive WB-indeterminate and -nega-tive or p24-antigen-nonneutralizable donations in-creased and then decreased following switches in reagentlots. Because the increase was system-wide and not con-fined to a specific location, the test kits were implicatedas the cause. Manufacturers were unable to identify anyparticular component in the test kit responsible for thesechanges in specificity. Because many of the critical rarereagents used in these components extend over multiplemaster lots, lot-related poor specificity can extend forseveral months.

It should be noted that as a part of the FDA licensureprocess of screening test kits used in the US, rigorousclinical trials are performed. These trials are usually lim-ited to 10,000 donations and only three master lots. Partof the licensure process also includes validation of allmanufacturing steps such that all processes occur withinwell-specified conditions to prevent changes in test kitperformance. However, these safeguards may not pre-vent the large variability in specificity observed whentesting millions of donation samples by test kits fromdifferent manufacturers or different lots of the same re-agent.

Our results clearly indicate that while manufacturersof new screening test kits have worked hard to developassays with enhanced sensitivity, further efforts are re-quired to enhance and maintain the specificity of theseassays. In evaluating test specificity (both in clinical trialsbefore licensure and in ongoing lot validation analyses),it is important to analyze results both in RPT (screened

negative) and in FT (previously un-screened) donors to accurately quantifythe impact of assay specificity on donorloss.8 The assessment of sensitivity andspecificity of screening tests also de-pends on the confirmatory tests beingused. It is important that developmentof highly sensitive donor screening testsbe accompanied by a parallel develop-ment of supplemental assays with im-proved sensitivity and specificity. Also,reentry algorithms should also be prac-tical and appropriate, plus include theresults of NAT, where available. Fur-ther, appropriate counseling guidelinesshould be available for unconfirmeddonors.

Longitudinal surveys to determineadequacy of the blood supply have

shown that despite the increases in blood collection be-tween 1997 and 1999, the margin between supply anddemand has declined.12 Each year a large number of do-nors are deferred from donating blood because of RRscreening test results, a high proportion of which arefalse-positive.12 The introduction of highly sensitivescreening tests in developed countries has significantlyreduced the risk of transmission of clinically significantviral infections by blood transfusion. Changing tests maybe justified when a new version of the assay has demon-strably enhanced sensitivity, irrespective of the impact onspecificity and donor loss. However, if there is only mar-ginal increase in sensitivity and a large increase in loss ofdonors because of poor specificity of serologic test, thenthe tradeoff may not be appropriate. Furthermore, therecent introduction of NAT screening negates the impactof window-period sensitivity differences between sero-logic tests. In the absence of evidence of enhanced sen-sitivity, blood centers should carefully weigh the adverseeffect of decreased test specificity and increased safe do-nor loss when considering changes in serologic assays.Because donations from individuals who are NAT-negative and serologically positive may remain infec-tious,34,35 it is unlikely that the implementation of geno-mic testing will replace serologic screening as a tool fordetection of transfusion-transmitted infections in blooddonors. Hence, serology and NAT screening will operatein parallel. Therefore, manufacturers, the FDA, and bloodcenters should be encouraged to introduce systems withhigh sensitivity and specificity and also to validate thespecificity of new reagent lots before their implementa-tion in routine screening. Together these changes wouldreduce the loss of safe donors from the declining donorpool and, more importantly, minimize the adverse im-pact of notification of false-positive test results on peoplewho give blood.

Fig. 6. HIV p24 antigen screening test results from May 2000 through August 2001.

(�) Confirmed-positive, (�) indeterminate, (�) screening-test-positive.

SHARMA ET AL.

212 TRANSFUSION Volume 43, February 2003

ACKNOWLEDGMENTS

The Retrovirus Epidemiology Donor Study (REDS Study Group)

is presently the responsibility of the following persons:

Blood Centers:

American Red Cross Blood Services Greater Chesapeake and

Potomac Region: C.C. Nass

American Red Cross Blood Services South-eastern Michigan

Region: M. Higgins and J. Campbell

American Red Cross Blood Services Southern California Re-

gion: G. Garratty and S. Hutching

Blood Centers of the Pacific: E.L. Murphy (UCSF) and M.P.

Busch

Oklahoma Blood Institute: R.O. Gilcher and J.W. Smith

Medical Coordinating Center:

Westat, Inc.: G.B. Schreiber and M.R. King

National Heart, Lung, and Blood Institute, NIH:

G.J. Nemo

Steering Committee Chairman:

T.F. Zuck

REFERENCES

1. Busch MP. To thy (reactive) donors be true! Transfusion

1997;37:117-20.

2. McCullough J. The nation’s changing blood supply sys-

tem. JAMA 1993;269:2239-45.

3. Busch MP, Laycock M, Kleinman SH, et al. Accuracy of

supplementary testing for human T-lymphotropic virus

types I and II in US blood donors. Retrovirus Epidemiol-

ogy Donor Study. Blood 1994;83:1143-8.

4. Busch MP, Kleinman SH, Williams AE, et al. Frequency of

human immunodeficiency virus (HIV) infection among

contemporary anti-HIV-1 and anti-HIV-1/2 supplemental

test-indeterminate blood donors. The Retrovirus Epidemi-

ology Donor Study. Transfusion 1996;36:37-44.

5. Kleinman S, Busch MP, Hall L, et al. False-positive HIV-1

test results in a low-risk screening setting of voluntary

blood donation. Retrovirus Epidemiology Donor Study.

JAMA 1998;280:1080-5.

6. Kuhnl P, Seidl S, Holzberger G. HLA DR4 antibodies

cause positive HTLV-III antibody ELISA results. Lancet

1985;1:1222-3.

7. Simonsen L, Buffington J, Shapiro CN, et al. Multiple false

reactions in viral antibody screening assays after influ-

enza vaccination. Am J Epidemiol 1995;141:1089-96.

8. Dow BC. ‘Noise’ in microbiological screening assays.

Transfus Med 2000;10:97-106.

9. Dodd RY, Houlihan K, Lamberson HV. The specificity of

enzyme immunoassays for antibodies to HIV: impact on

record and donor management. Transfusion 1993;33:

693.

10. Jackson JB, Hanson MR, Johnson GM, et al. Long term

follow-up of blood donors with indeterminate HIV-1 re-

sults on Western blot. Transfusion 1995;35:98-102.

11. Dodd RY. Current viral risks of blood and blood products.

Ann Med 2000;32:469-74.

12. Brittenham GM, Klein HG, Kushner JP, Ajioka RS. Pre-

serving the national blood supply. Hematology (Am Soc

Hematol Educ Program) 2001;Jan:422-32.

13. Guiltinan AM, Murphy EL, Horton JA, et al. Psychological

distress in blood donors notified of HTLV-I/II infection.

Retrovirus Epidemiology Donor Study. Transfusion 1998;

38:1056-62.

14. Wu Y, Schreiber GB, Busch MP, et al. Donor behavior af-

ter notification of abnormal laboratory test results (ab-

stract). Transfusion 1999;39(Suppl):113S.

15. Kleinman SH, Williams AE, Busch MP, Wu Y. Donor reac-

tions to notification of abnormal test results (abstract).

Transfusion 1999;39(Suppl):113S.

16. Snedecor GW, Cochran WG. Statistical methods. 7th ed.

Ames (IA): Iowa State University Press; 1980:215-20.

17. Kleinman S, Alter H, Busch M, et al. Increased detection

of hepatitis C virus (HCV)-infected blood donors by a

multiple-antigen HCV enzyme immunoassay. Transfusion

1992;32:805-13.

18. Comprehensive report on blood collection and transfu-

sion in the United States in 1999. Bethesda: National

Blood Data Resource Center; 2001.

19. Sayre KR, Dodd RY, Tegtmeier G, et al. False-positive hu-

man immunodeficiency virus type 1 Western blot tests in

noninfected blood donors. Transfusion 1996;36:45-52.

20. Dodd RY, Stramer SL. Indeterminate results in blood do-

nor testing: what you don’t know can hurt you. Transfus

Med Rev 2000;14:151-60.

21. Celum CL, Coombs RW, Jones M, et al. Risk factors for

repeatedly reactive HIV-1 EIA and indeterminate western

blots: a population-based case-control study. Arch Intern

Med 1994;154:1129-37.

22. Dow BC, Munro H, Frame D, Yates P. Several assays show

hepatitis B positivity soon after vaccination. BMJ 1998;

316:311.

23. Ownby HE, Korelitz JJ, Busch MP, et al. Loss of volunteer

blood donors because of unconfirmed enzyme immuno-

assay screening results. Retrovirus Epidemiology Donor

Study. Transfusion 1997;37:199-205.

24. Busch MP, Lee LL, Satten GA, et al. Time course of detec-

tion of viral and serologic markers preceding human im-

munodeficiency virus type 1 seroconversion: implications

for screening of blood and tissue donors. Transfusion

1995;35:91-7.

25. Schable C, Zekeng L, Pau CP, et al. Sensitivity of United

States HIV antibody tests for detection of HIV-1 group O

infections. Lancet 1994;344:1333-4.

26. Hu DJ, Dondero TJ, Rayfield MA, et al. The emerging ge-

netic diversity of HIV: the importance of global surveil-

lance for diagnosis, research and prevention. JAMA 1996;

275:210-6.

27. Dow BC, Follett EA, Davidson F, Calmann M. Comparison

IMPACT OF NEW SCREENING TESTS

Volume 43, February 2003 TRANSFUSION 213

of second- and third-generation hepatitis C virus enzyme-

linked immunosorbent assays in a low-risk blood donor

population. Transfusion 1996;36:477-8.

28. Busch MP, Tobler LH, Tegtmeier G, et al. Use of third-

generation hepatitis C virus (HCV) enzyme immunoassay

(EIA) to resolve second-generation HCV EIA-reactive and

second-generation recombinant immunoblot assay-

indeterminate blood samples: data to support current

Food and Drug Administration guidance on HCV look-

back. Transfusion 2000;40:10-4.

29. Courouce AM, Le Marrec N, Girault A, et al. Anti-hepatitis

C virus (anti-HCV) seroconversion in patients under-

going hemodialysis: comparison of second- and third-

generation anti-HCV assays. Transfusion 1994;34:790-5.

30. Barrera JM, Francis B, Ercilla G, et al. Improved detection

of anti-HCV in post-transfusion hepatitis by a third-

generation ELISA. Vox Sang 1995;68:15-8.

31. Vrielink H, Zaaijer HL, Reesink HW, et al. Sensitivity and

specificity of three third-generation anti-HCV ELISAs. Vox

Sang 1995;69:14-7.

32. Vrielink H, Zaaijer HL, Reesink HW, et al. Comparison of

two anti-hepatitis C virus enzyme-linked immunosorbent

assays. Transfusion 1995;35:601-4.

33. Dow BC, Buchanan I, Munro H, et al. Relevance of

RIBA-3 supplementary test to HCV PCR positivity and

genotypes for HCV confirmation of blood donors. J Med

Virol 1996;49:132-6.

34. Allain JP. Will genome detection replace serology in blood

screening for microbial agents? Baillieres Best Pract Res

Clin Hematol 2000;13:615-29.

35. Sato S, Ohhashi W, Ihara H, et al. Comparison of the sen-

sitivity of NAT using pooled donor samples for HBV

and that of a serologic HBsAg assay. Transfusion

2001;41:1107-13.

SHARMA ET AL.

214 TRANSFUSION Volume 43, February 2003