JOURNAL OF CLINICAL MICROBIOLOGY, June 1991, p. 1137-1142 0095-1137/91/061137-06$02.00/0 Copyright C 1991, American Society for Microbiology Evaluation of a Commercially Available Enzyme-Linked Immunosorbent Assay for Giardia lamblia Antigen in Stool DAVID G. ADDISS,1* HENRY M. MATHEWS,1 JAMES M. STEWART,1 SUSANNE P. WAHLQUIST,1 RUBY M. WILLIAMS,1 ROBERT J. FINTON,2 HARRISON C. SPENCER,' AND DENNIS D. JURANEK1 Parasitic Diseases Branch, Division of Parasitic Diseases, Center for Infectious Diseases, Centers for Disease Control, Atlanta, Georgia 30333,1 and Office of Epidemiology, Fulton County Health Department, Atlanta, Georgia 303032 Received 9 October 1990/Accepted 26 February 1991 The lack of a quick, simple, and inexpensive diagnostic test has limited the ability of public health officials to rapidly assess and control outbreaks of Giardia lamblia in child day-care centers. We evaluated the performance of a commercially available enzyme-linked immunosorbent assay (ELISA) for the detection of a G. lamblia-associated antigen in stool. Stool specimens were collected from the diapers of 426 children attending 20 day-care centers, fixed in 10% Formalin and polyvinyl alcohol, and examined by microscopy by Formalin concentration and trichrome staining techniques. Specimens were also tested visually and spectrophotometri- cally by ELISA. Of 99 tests positive by microscopy, 93 were visually positive by ELISA (sensitivity, 93.9%). Of 534 tests negative for G. lamblia by microscopy, 32 (6.0%) were ELISA positive. However, on the basis of examination of multiple specimens from the same child, none of these could be considered false-positive ELISAs; the specfficity of the ELISA was therefore 100%. The sensitivity of both microscopy and ELISA improved as the number of specimens per child increased. An optical density value of >0.040 was 98.0% sensitive and 100% specific for G. lamblia. This ELISA, which appeared to be more sensitive for G. lamblia than did microscopic examination of stool, should be useful as an epidemiologic tool, particularly in day-care settings, and may also have a role in confirming clinical diagnoses of giardiasis. Giardia lamblia is endemic in child day-care centers (DCCs) in the United States (2, 12), with estimates of 5 to 15% of diapered children attending DCCs being infected. G. lamblia is also a cause of epidemic diarrhea in DCCs; conventional measures to control these outbreaks, including the diagnosis, treatment, and follow-up of infected children and their household contacts, may be costly for local public health agencies to implement and of questionable effective- ness (13). Diagnosis of Giardia infection by microscopic examina- tion of stool for ova and parasites can be time-consuming and expensive; furthermore, examination of a single stool specimen may miss 10 to 50% of Giardia infections because of intermittent or low-level shedding of cysts in stool (3). The need for rapid, inexpensive methods to diagnose Giardia infection has led to the recent development of enzyme-linked immunosorbent assays (ELISAs) for the detection of G. lamblia-associated antigens in stool. As part of an epidemiologic study of endemic giardiasis in DCCs in Fulton County, Ga., we evaluated the first commer- cially available ELISA kit for G. lamblia in stool (ProSpecT/ Giardia, Alexon, Inc., Mountain View, Calif.). This ELISA detects a G. lamblia-specific antigen (GSA-65) associated with the Giardia cyst wall and, to a lesser extent, with the trophozoite (11). In a previous study, this ELISA appeared to be both sensitive and specific for Giardia infection in hospi- talized adults and refugees attending a screening clinic (10), but neither its use in children nor its correlation with standard microscopic examination of several stools from the same person had been evaluated. Similarly, most studies of other ELISAs for Giardia antigen in stool (4, 6, 7, 14, 15) have * Corresponding author. evaluated only single specimens from each study participant. Because microscopic examination of a single stool specimen for G. lamblia lacks sensitivity, the use of a single specimen as a "gold standard" in these studies may have resulted in artificially low estimates of test specificity. MATERIALS AND METHODS After informed consent was obtained, stool specimens were collected from the soiled diapers of children attending 17 randomly selected DCCs and three DCCs with outbreaks of diarrheal illness in Fulton County, Ga., from February 1989 to March 1990. When diapers were changed, stool specimens were collected by the investigators or by caregiv- ers at the centers and placed in separate vials of 10% Formalin and polyvinyl alcohol preservative (Para-Paks, Meridian Diagnostics). At the Centers for Disease Control, specimen containers were labeled numerically so that labo- ratory investigators were unaware of the source of the specimen. Whenever possible, three or more specimens were collected on different days from each child. Each specimen was examined microscopically after being processed by three different methods. Stools fixed in 10% Formalin were concentrated by the Formalin-ethyl acetate sedimentation technique (8). The concentrate was placed on a slide and covered with a 22-mm coverslip; the entire slide was then examined microscopically for the presence of ova and parasites. Using the Kinyoun carbol-fuchsin (modified acid-fast) procedure (1), a stained slide was also prepared from the concentrate and examined microscopically for the presence of Cryptosporidium species. A permanent slide of stool fixed in polyvinyl alcohol was stained with trichrome (8) and examined microscopically for the presence of G. lamblia and other protozoa. All slides were read for at least 1137 Vol. 29, No. 6 on May 2, 2019 by guest http://jcm.asm.org/ Downloaded from
Transcript
JOURNAL OF CLINICAL MICROBIOLOGY, June 1991, p. 1137-11420095-1137/91/061137-06$02.00/0Copyright C 1991, American Society for Microbiology
Evaluation of a Commercially Available Enzyme-LinkedImmunosorbent Assay for Giardia lamblia Antigen in Stool
DAVID G. ADDISS,1* HENRY M. MATHEWS,1 JAMES M. STEWART,1 SUSANNE P. WAHLQUIST,1RUBY M. WILLIAMS,1 ROBERT J. FINTON,2 HARRISON C. SPENCER,'
AND DENNIS D. JURANEK1
Parasitic Diseases Branch, Division of Parasitic Diseases, Center for Infectious Diseases, Centers forDisease Control, Atlanta, Georgia 30333,1 and Office of Epidemiology, Fulton County Health Department,
Atlanta, Georgia 303032
Received 9 October 1990/Accepted 26 February 1991
The lack of a quick, simple, and inexpensive diagnostic test has limited the ability of public health officialsto rapidly assess and control outbreaks of Giardia lamblia in child day-care centers. We evaluated theperformance of a commercially available enzyme-linked immunosorbent assay (ELISA) for the detection of a
G. lamblia-associated antigen in stool. Stool specimens were collected from the diapers of 426 children attending20 day-care centers, fixed in 10% Formalin and polyvinyl alcohol, and examined by microscopy by Formalinconcentration and trichrome staining techniques. Specimens were also tested visually and spectrophotometri-cally by ELISA. Of 99 tests positive by microscopy, 93 were visually positive by ELISA (sensitivity, 93.9%).Of 534 tests negative for G. lamblia by microscopy, 32 (6.0%) were ELISA positive. However, on the basis ofexamination of multiple specimens from the same child, none of these could be considered false-positiveELISAs; the specfficity of the ELISA was therefore 100%. The sensitivity of both microscopy and ELISAimproved as the number of specimens per child increased. An optical density value of >0.040 was 98.0%sensitive and 100% specific for G. lamblia. This ELISA, which appeared to be more sensitive for G. lambliathan did microscopic examination of stool, should be useful as an epidemiologic tool, particularly in day-caresettings, and may also have a role in confirming clinical diagnoses of giardiasis.
Giardia lamblia is endemic in child day-care centers(DCCs) in the United States (2, 12), with estimates of 5 to15% of diapered children attending DCCs being infected. G.lamblia is also a cause of epidemic diarrhea in DCCs;conventional measures to control these outbreaks, includingthe diagnosis, treatment, and follow-up of infected childrenand their household contacts, may be costly for local publichealth agencies to implement and of questionable effective-ness (13).
Diagnosis of Giardia infection by microscopic examina-tion of stool for ova and parasites can be time-consumingand expensive; furthermore, examination of a single stoolspecimen may miss 10 to 50% of Giardia infections becauseof intermittent or low-level shedding of cysts in stool (3). Theneed for rapid, inexpensive methods to diagnose Giardiainfection has led to the recent development of enzyme-linkedimmunosorbent assays (ELISAs) for the detection of G.
lamblia-associated antigens in stool.As part of an epidemiologic study of endemic giardiasis in
DCCs in Fulton County, Ga., we evaluated the first commer-cially available ELISA kit for G. lamblia in stool (ProSpecT/Giardia, Alexon, Inc., Mountain View, Calif.). This ELISAdetects a G. lamblia-specific antigen (GSA-65) associatedwith the Giardia cyst wall and, to a lesser extent, with thetrophozoite (11). In a previous study, this ELISA appeared tobe both sensitive and specific for Giardia infection in hospi-talized adults and refugees attending a screening clinic (10),but neither its use in children nor its correlation with standardmicroscopic examination of several stools from the same
person had been evaluated. Similarly, most studies of otherELISAs for Giardia antigen in stool (4, 6, 7, 14, 15) have
* Corresponding author.
evaluated only single specimens from each study participant.Because microscopic examination of a single stool specimenfor G. lamblia lacks sensitivity, the use of a single specimenas a "gold standard" in these studies may have resulted inartificially low estimates of test specificity.
MATERIALS AND METHODS
After informed consent was obtained, stool specimenswere collected from the soiled diapers of children attending17 randomly selected DCCs and three DCCs with outbreaksof diarrheal illness in Fulton County, Ga., from February1989 to March 1990. When diapers were changed, stoolspecimens were collected by the investigators or by caregiv-ers at the centers and placed in separate vials of 10%Formalin and polyvinyl alcohol preservative (Para-Paks,Meridian Diagnostics). At the Centers for Disease Control,specimen containers were labeled numerically so that labo-ratory investigators were unaware of the source of thespecimen. Whenever possible, three or more specimenswere collected on different days from each child.Each specimen was examined microscopically after being
processed by three different methods. Stools fixed in 10%Formalin were concentrated by the Formalin-ethyl acetatesedimentation technique (8). The concentrate was placed on
a slide and covered with a 22-mm coverslip; the entire slidewas then examined microscopically for the presence of ovaand parasites. Using the Kinyoun carbol-fuchsin (modifiedacid-fast) procedure (1), a stained slide was also preparedfrom the concentrate and examined microscopically for thepresence of Cryptosporidium species. A permanent slide ofstool fixed in polyvinyl alcohol was stained with trichrome(8) and examined microscopically for the presence of G.lamblia and other protozoa. All slides were read for at least
TABLE 1. Results of 633 tests for G. lamblia by both ELISA and microscopic examination for ova and parasites,with classification of discordant specimens
Test result category No.
Total ...................................................................................................... 633Positive by microscopic examination ...................................................................................................... 99
Positive by ELISA....................................................................................................... 93Negative by ELISA ...................................................................................................... 6a
Negative by microscopic examination ...................................................................................................... 534Negative by ELISA ...................................................................................................... 502Positive by ELISA...................................................................................................... 32
False-positive ELISA ...................................................................................................... 0False-negative microscopy ...................................................................................................... 20Repeat microscopic examination of same specimen was positive ..................................................................................... 8Children with intermittent cyst shedding; other stools from these children were positive on microscopic examination .............. 7Posttreatment: specimens were obtained a few days after children were treated for giardiasis; specimens obtained before
treatment were positive for G. lamblia on microscopic examination ..................... ........................................................ 3Probable prepatent or recent infection; specimens obtained a few days later were positive for G. lamblia on microscopic
Children were treated for giardiasis after these initial specimens were obtained; all subsequent stools werenegative for G. lamblia by microscopic examination and ELISA.................................................................................. 4
Other specimens from these children were positive for G. lamblia on microscopic examination, but were obtained morethan 1 week later; therefore, the children could have become infected after these specimens were obtained ........................ 8
" See Table 2.
10 min by one of two microscopists (S.P.W. or R.M.W.),who noted the presence and number of organisms per 10fields at x40, as follows: none; few, <3 organisms; moder-ate, 3 to 9 organisms; or many, .10 organisms. An indepen-dent reviewer confirmed all positive or questionable speci-mens and a 10% subset of negative specimens that wererandomly selected by computer.The first Formalin-preserved stool specimen from each
child and all additional specimens from children with evi-dence of Giardia infection were tested by ELISA, accordingto the instructions of the manufacturer, by one of the authors(H.M.M.), who was blinded to the source of the stool andthe results of microscopic examination. Two readers(H.M.M. and another reader) independently scored theirvisual interpretation of the colormetric reactions using theindicator card provided by the manufacturer (from 0 to 4+;a faint color lighter than the 1+ standard on the indicatorcard was coded as 0.5). If two readers disagreed regardingpositive or negative results, a third reader scored all the testsin that run. Aliquots of 200 ,ul each were then transferred toa microtitration tray for determination of the optical density(OD) at 490 nm with a microplate spectrophotometer (Mo-lecular Devices, Palo Alto, Calif.).
Lyophilized antigen (GSA-65) from a single lot was recon-stituted and stored at -20°C for use as a positive control;immediately before each run, the antigen was warmed toroom temperature. In addition, aliquots of stools from twopeople with giardiasis and from an uninfected asymptomaticperson were fixed in 10% Formalin in February 1989, storedat room temperature, and tested with each run to evaluateantigen stability after fixation in Formalin.
If microscopic examination and ELISA results were dis-cordant, the specimen was relabeled with a new number andevaluated again for G. lamblia by both methods. Somespecimens were therefore tested more than once; however,laboratory investigators did not know whether a specimenhad previously been tested.To evaluate the specificity of the ELISA, 131 stool spec-
imens from a Bolivian village where G. lamblia and otherintestinal parasites are endemic were also examined bymicroscopy and tested by ELISA as described above.
Definitions. A child was defined as infected if Giardia cystsor trophozoites were observed by microscopic examinationof one or more stool specimens; the specificity of micro-scopic examination for G. lamblia was therefore consideredto be 100%. A child was defined as uninfected if no Giardiacysts or trophozoites were visualized on microscopic exam-ination of three or more stool specimens. Except whereotherwise noted, a specimen was considered to have apositive visual ELISA result if two readers indicated a visualcolorimetric result of 0.5 or more. The Spearman rank testwas used to test for correlations between variables.
RESULTS
A total of 1,148 stool specimens were obtained from 426diapered children; 609 specimens were examined by bothmicroscopy and ELISA. Twenty-four specimens with dis-cordant or questionable results were repeated by micros-copy and ELISA, for a total of 633 tests examined by bothmethods.Of the 633 tests, 99 (15.6%) were positive for G. lamblia
by microscopy; 93 (93.9%) of these were visually positive byELISA (Table 1). Of the six tests that were visually negativeby ELISA, five tests had fewer than five cysts per slide onmicroscopic examination and ELISA OD readings of 0.030to 0.093; the sixth had only a few trophozoites on micro-scopic examination (Table 2).The remaining 534 tests were negative for G. lamblia by
microscopy; of these, 32 (6.0%) were visually positive byELISA (Table 1). These discordant tests were classified asfalsely positive by ELISA, falsely negative by microscopy,or indeterminate on the basis of subsequent stool specimensfrom the same child or repeat examinations of the samespecimen. Twenty discordant tests were classified as falselynegative by microscopy, and the remaining 12 tests wereconsidered to be indeterminate for the reasons listed in Table1; the specificity of the ELISA for G. lamblia was therefore100%. Of 119 true-positive tests (99 specimens positive bymicroscopy and an additional 20 positive only by ELISA), 99(83.2%) were identified as positive by microscopy and 113(95.0%) were positive by ELISA (Table 3).
EVALUATION OF AN ELISA FOR GIARDIA ANTIGEN IN STOOL 1139
TABLE 2. Results for six discordant tests that were positive for G. lamblia on microscopic examination for ova and parasites andnegative by visual ELISA reading
Results of repeat ELISA testingChild no. Specimen no. OD Microscopic finding
OD Visual reading
1 759 0.053 Four cysts seen 0.192 Positive; two of two readers
2 821 0.018 Few trophozoites 0.027 Negative; two of two readers
3 1103 0.030 Two cysts seen 0.048 Positive; one of three readers8893 (1103) 0.048 Five cysts seen
4 8595 0.093 Four cysts seen 0.163 Positive; three of three readers8888 (8595) 0.089 One cyst seen 0.046 Positive; two of three readers
Initial specimens. In the diagnostic laboratory, only one test on an individual stool specimen. Because the averagestool specimen might be received from each patient, and prepatent or incubation period for giardiasis is approxi-reading of the ELISA by multiple technicians may not be mately 9 days (16), we reasoned that children were alreadypractical. To estimate the sensitivity and specificity of the infected during the 4 days before they produced a stool thatELISA in a routine diagnostic laboratory, we compared was microscopically positive for G. lamblia. One wouldvisual ELISA results for a single reader (H.M.M.) with therefore expect a test with 100% sensitivity for Giardiamicroscopic results of initial stool specimens from all 426 infection to be positive on any specimen obtained duringchildren. Twenty-nine (6.8%) children were positive for G. these preceding 4 days. To compare the ability of micros-lamblia by microscopy on the first stool (Table 3). Two of copy and ELISA to detect G. lamblia in individuals knownthese specimens (specimens 1103 and 8595, Table 2) were to be infected, we identified 31 specimens obtained from 15visually negative by ELISA; no follow-up specimens were G. lamblia-positive children during the 4 days before aavailable from either child. Eight children had microscopi- microscopically positive specimen was obtained. The sensi-cally negative specimens that were ELISA positive. Seven tivities of microscopy and ELISA were 70.0 and 83.9%,of these specimens were from children shown to have G. respectively (Table 3).lamblia on subsequent specimens or on repeat microscopic One or more specimens. Collection of more than one stoolexamination of the same stool. The eighth child, who was specimen from the same child permitted us to assess theexcluded from the calculations of test sensitivity presented degree to which multiple specimens may improve the sensi-in Table 3, was treated for symptomatic giardiasis after tivity of both microscopic examination and ELISA. Wesubmission of the first stool specimen; the subsequent spec- reviewed data for all 38 children who had a stool specimenimen was negative. Of 36 children with initial true-positive positive by microscopy within 10 days after submitting theirspecimens (29 by microscopy and 7 more positive only by first specimen. Specimens submitted more than 10 days afterELISA), 29 (80.6%) were positive by microscopy and 34 the initial specimen were excluded because some children,(94.4%) were positive by ELISA. particularly in outbreak centers, were initially negative forKnown infected children. Clinicians may be more inter- G. lamblia and were found to be positive 2 to 3 weeks later.
ested in the degree to which the ELISA correctly identifies These children were considered likely to have had newlypeople infected with G. lamblia than in the sensitivity of the acquired infections rather than false-negative initial test
TABLE 3. Results and sensitivity of ELISA and microscopic (ova and parasite) examination for G. lamblia in stool
Result (no.) of microscopic examination Sensitivity of both methods vsand ELISA Sensitivity of ELISA all true positivesb
Specimens submitted within 10 daysafter first specimen from each child
First specimen 27 2 4 5 31/38 (81.6) 29/38 (76.3)Two specimens 30 3 3 2 33/38 (86.8) 33/38 (86.8)Three specimens 34 2 2 0 36/38 (94.7) 36/38 (94.7)da Values are number of ELISA-positive, microscopically positive tests/number of microscopically positive tests (%).b Values are number of tests positive by ELISA or microscopy/number of "true-positive" tests (t). "True positives" include specimens positive for G. lamblia
by microscopic examination as well as ELISA-positive/microscopically negative specimens from children who were microscopically positive for G. lamblia ona subsequent stool specimen or repeat examination of the same specimen. Specimens considered "indeterminate" (Table 1) are excluded.
I See Table 2.d The sensitivity of microscopy did not reach 100%o because two children had specimens that were ELISA positive but microscopically negative on initial
examination; on repeat examination of the same stool, a few Giardia cysts were found.
FIG. 1. ELISA OD results for microscopically negative speci-mens from uninfected children (those who had three or morespecimens negative for G. lamblia by microscopic examination) andmicroscopically positive specimens from infected children.
results. The sensitivities of both microscopy and ELISAimproved as the number of specimens increased; for a singlespecimen, the sensitivities of microscopy and ELISA were76.3 and 81.6%, respectively. For three specimens, bothmethods were 94.7% sensitive (Table 3). These values prob-ably underestimate test sensitivity since the four childrenwho remained negative for G. lamblia by either method weretreated for giardiasis after the first specimen was obtained.
Correlation of ELISA OD with visual readings and micros-copy. To determine the OD that would best distinguishbetween G. lamblia-positive and G. Iamblia-negative speci-mens, OD values of 99 microscopically positive specimenswere compared with OD values of specimens from 261uninfected people with three or more microscopically nega-tive specimens (Fig. 1). All negative specimens had ODs ofc.0.040; two positive specimens (specimens 821 and 1103,Table 2) had ODs of c0.040. A positive OD value of >0.040would therefore be 98.0% sensitive and 100% specific.The OD was highly correlated with the mean visual score
(Table 4) (r2 = 0.92, P = 0.0001). To a lesser extent, both the
OD and the mean visual ELISA reading were also correlatedwith the number of organisms seen on microscopy (r2 = 0.66and 0.70, respectively; P = 0.001). The lower degree ofcorrelation appeared to be due primarily to the fact thatmany of the discordant (microscopically negative/visualELISA positive) specimens considered to be true positiveshad a relatively high OD.
Correlation between readers. ELISA readers agreed onpositive or negative visual results for all but 10 (1.6%) tests,which had ODs of between 0.046 and 0.153. None of the ninepeople who read ELISAs during the study were significantlymore likely than the others to score these tests as negative;therefore, the discrepancies did not appear to be due todifferences in the ability to perceive slight colorimetricchanges. Actual visual scores (0 to 4+) were highly corre-lated (r2 = 0.90 to 1.00, P 0.0001) for eight of the ninereaders. Scores for the remaining reader were less highlycorrelated with those of one other reader (r2 = 0.58, P =
0.001).Lack of evidence for ELISA cross-reactivity. No evidence
of cross-reactivity with Cryptosporidium species or otherorganisms was found. Sixty-nine DCC specimens were mi-croscopically positive for Cryptosporidium species; theELISA was positive only for the seven DCC specimens thatwere also microscopically positive for G. lamblia. Of 131single stool specimens from Bolivia, 115 were negative forG. lamblia by both microscopy and ELISA; the specificity ofthe ELISA was therefore 100% in this group. Forty-four ofthese specimens were positive for other organisms, including20 (17%) for Ascaris lumbricoides, 20 (17%) for Trichuristrichiura, 11 (10%) for Entamoeba histolytica, and 5 (5%) forTaenia species.Decay in antigen reactivity with time. The OD of the
lyophilized positive controls provided by the manufacturerranged from 0.557 to 1.174 (mean + standard deviation,0.868 + 0.152) and did not significantly change with time(Fig. 2). In contrast, the ELISA reactivity of the Formalin-preserved G. lamblia-positive stool prepared in our labora-tory dropped sharply in the first week, followed by a gradualdecay over a period of 9 months. Overall, the OD of theFormalin-preserved stools decreased by 78.2% (an averageof 8.7% per month), from a mean of 1.100 during the firstmonth to a mean of 0.240 during the ninth month.
Intermittent detection of antigen and cysts. Two children inthe same DCC who were considered infected with G. lam-blia had stool specimens that were negative by both micro-scopic examination and ELISA. Both children had stoolsthat were microscopically positive for G. lamblia within 26days before and 25 days after these negative specimens;thus, the negative stools were "bracketed" by G. lamblia-positive stools. Neither child was treated for giardiasis;
TABLE 4. Correlation of ELISA OD with ELISA visual readings and number of organisms seen on microscopya
OD rangeVisual reading (04+)b No. of organisms seen on microscopyc No. of
a There were 633 tests performed on stools from 426 children.b A trace of color (less than 1+ on indicator card) was scored as 0.5.c Few, c2 organisms per 10 (x40) fields; moderate, 3 to 9 organisms per 10 fields; many, .10 organisms per 10 fields.
Days After Preparation of Formalin SpecimenFIG. 2. ELISA OD over time for Formalin-preserved, G. lamblia-positive stools and the lyophilized positive test kit control (GSA-65)
provided by the manufacturer.
however, because of the length of time between specimens,self-cure and reinfection cannot be entirely excluded as apossibility. OD values for these specimens were 0.000 and0.013.
DISCUSSION
The lack of a quick, simple, and inexpensive diagnostictest has limited the ability of public health officials to rapidlyassess and control outbreaks of G. lamblia in DCCs. Al-though ELISAs for Giardia antigen in stool have recentlybeen developed in several laboratories and appear to bepromising (4-7, 9, 14, 15), these tests have not been rigor-ously assessed in the DCC setting.We evaluated the performance of a commercially available
ELISA using stool specimens from diapered children attend-ing DCCs in Fulton County, Ga. Compared with micro-scopic examination of stool specimens for ova and parasites,the ELISA was more sensitive and virtually 100% specificfor G. lamblia. Because many specimens can be read by asingle technician in a short period of time, the ELISA is alsopotentially less expensive than microscopic examination ofstools. The ELISA is particularly useful for epidemiologicinvestigations when rapid results are required for largenumbers of specimens and the etiologic agent is known to beG. lamblia.An OD cutoff of >0.040 was 98% sensitive and 100%
specific for Giardia infection. Rosoff et al. (10) recom-mended the same positive cutoff value for this ELISA in astudy of hospitalized patients and people attending a refugeescreening clinic. However, the assay is designed to be readvisually, and for routine use in the diagnostic laboratory, ODdetermination does not appear to be necessary. In ourlaboratory, visual ELISA results and OD values were highlycorrelated, and reader-to-reader variation was minimal. Ourdata suggest that if even a faint colorimetric change is
observed, the test should be interpreted as positive for G.lamblia.The sensitivities of this and other ELISAs have compared
favorably with microscopic examination in previous studies(4-7, 9, 14, 15). Nash et al. (9) examined serial stoolspecimens from 18 infected adult volunteers; antigen-posi-tive but cyst-negative stools occurred primarily during treat-ment and in the prepatent period. We also observed ELISA-positive but cyst-negative stools in children who hadrecently begun antimicrobial therapy and in those fromoutbreak centers who were apparently infected after theinitial specimens had been obtained. Investigators haveexpressed concern about the potential lack of specificity ofother Giardia antigen assays (17), but we were unable todocument a single false-positive ELISA result.Although the ELISA appeared to be more sensitive for G.
lamblia than microscopic examination was, on a single stoolspecimen the sensitivity was <100% in children known to beinfected. As with microscopy, the sensitivity of ELISAimproved as the number of specimens increased. Two dif-ferent types of false-negative ELISA results (discordant andconcordant) were observed. Discordant results (those thatwere microscopically positive on the same test) had border-line OD readings and were likely to be ELISA positive onrepeat testing of the same specimen. These false-negativeresults may have been caused by test-to-test variation inthe laboratory. By contrast, two specimens with negativeELISAs and negative microscopic results were obtainedfrom children who were microscopically positive both onprevious and subsequent specimens. These specimens hadlower OD readings and did not become ELISA positive on
repeat testing. Although self-cure and reinfection could notbe excluded as a possibility, these findings suggest thatshedding of both cysts and Giardia antigen into stool mayoccur intermittently. The presence of GSA-65 in the stool
could therefore be expected to be irregular in infected peoplewho shed cysts intermittently.The functional nature of the antigen detected by the
ELISA remains obscure, but the antigen has been shown tobe associated primarily with the Giardia cyst wall (11). Anideal stool test for G. lamblia would be 100% sensitive forthe presence of trophozoites in the small intestine. We foundonly one stool specimen that contained trophozoites (in lownumbers) and no Giardia cysts; this specimen was negativeby ELISA.
It is unknown whether GSA-65 is associated with allhuman Giardia strains. However, our results in childrenattending DCCs in Georgia and people in a Bolivian villagewere comparable to those of Rosoff et al. (10), who studiedthis ELISA with specimens from refugees and symptomatichospital patients in California. OD readings of Formalin-fixed positive control specimens decreased gradually duringthe study, suggesting that the antigen is not stable over longperiods of time in 10% Formalin at room temperature. Formaximum ELISA sensitivity, preserved specimens shouldbe tested within a few months after being submitted.To limit costs, only the first stool specimen from each
child was tested by ELISA, unless that specimen wasELISA positive or any of the child's specimens were posi-tive for G. lamblia by microscopy. A few more ELISA-positive, cyst-negative discordant specimens may have beenidentified if all specimens had been tested by both ELISAand microscopy. However, because the prevalence of Gia-rdia infection in this population was low, virtually all of theuntested specimens would have been negative by bothmethods, i.e., concordantly negative. Thus, routine testingof only the first stool specimen by ELISA may have led us tooverestimate the proportion of discordant test results andunderestimate the performance of the ELISA.
In summary, when compared with microscopic examina-tion of stool, the ELISA was more sensitive and 100%specific for Giardia infection. The assay appears to be usefulas an epidemiologic tool, especially in day-care settings, andin the investigation and control of outbreaks of giardiasis.The ELISA may also have a role in confirming the clinicaldiagnosis of giardiasis by physicians. It is unlikely to replacemicroscopic examination of stool for ova and parasites as aroutine diagnostic test, however, because other potentialpathogens would escape detection. Although more sensitivethan microscopic examination, intermittent antigen sheddingand laboratory variation may result in an occasional false-negative ELISA, particularly on a single stool specimen.
Additional studies are needed to further characterize thenature of the antigen and evaluate the diagnostic potential ofthe ELISA, especially in people with chronic giardiasis whoare negative for G. lamblia on microscopic examination ofstool.
ACKNOWLEDGMENTS
We thank the directors, staff, children, and parents of the DCCsparticipating in this study; Barry Whitset and William Elsea, FultonCounty Health Department; Chandra Jones, Morehouse UniversitySchool of Medicine; and Henry Bishop and James Sullivan at the
Centers for Disease Control. We also thank Phyllis Moir for edito-rial assistance, Ralph Bryan for providing stool specimens fromBolivia, Meridian Diagnostics for providing specimen containers,and Alexon, Inc., for providing reagents for the ELISA.
REFERENCES
1. Ash, L. R., and T. C. Orihel. 1987. A guide to laboratoryprocedures and identification, p. 51-52. Society of ClinicalPathologists Press, Chicago.
2. Black, R. E., A. C. Dykes, S. P. Sinclair, and J. G. Wells. 1977.Giardiasis in day-care centers: evidence of person-to-persontransmission. Pediatrics 60:486-491.
3. Burke, J. A. 1977. The clinical and laboratory diagnosis ofgiardiasis. Crit. Rev. Lab. Sci. 7:373-391.
4. Carlson, J. R., P. S. Sullivan, D. J. Harry, M. A. Stork, S. A.Thorton, and H. L. DuPont. 1988. Enzyme immunoassay for thedetection of Giardia lamblia. Eur. J. Clin. Microbiol. Infect.Dis. 7:538-540.
5. Goldin, A. J., W. Apt, X. Aguilera, I. Zulantay, D. C. Warhurst,and M. A. Miles. 1990. Efficient diagnosis of giardiasis amongnursery and primary school children in Santiago, Chile bycapture ELISA for the detection of fecal Giardia antigens. Am.J. Trop. Med. Hyg. 42:538-545.
6. Green, E. L., M. A. Miles, and D. C. Warhurst. 1985. Immu-nodiagnostic detection of Giardia antigen in faeces by a rapidvisual enzyme-linked immunosorbent assay. Lancet ii:691-693.
7. Janoff, E. N., J. C. Craft, L. K. Pickering, T. Novotny, M. J.Blaser, C. V. Knisley, and L. B. Reller. 1989. Diagnosis ofGiardia lamblia infection by detection of parasite-specific anti-gens. J. Clin. Microbiol. 27:431-435.
8. Melvin, D. M., and M. M. Brooke. 1982. Laboratory proceduresfor the diagnosis of intestinal parasites, 3rd ed., p. 91-131.Centers for Disease Control laboratory manual. U.S. Depart-ment of Health and Human Services publication no. (CDC)82-8282. Centers for Disease Control, Atlanta.
9. Nash, T. E., D. A. Herrington, and M. M. Levine. 1987.Usefulness of an enzyme-linked immunosorbent assay for de-tection of Giardia antigen in feces. J. Clin. Microbiol. 25:1169-1171.
10. Rosoff, J. D., C. A. Sanders, S. S. Sonnad, P. R. DeLay, W. K.Hadley, F. F. Vincenzi, D. M. Yajko, and P. D. O'Hanley. 1989.Stool diagnosis of giardiasis using a commercially availableenzyme immunoassay to detect Giardia-specific antigen 65(GSA 65). J. Clin. Microbiol. 27:1997-2002.
11. Rosoff, J. D., and H. H. Stibbs. 1986. Isolation and identificationof a Giardia lamblia-specific stool antigen (GSA 65) useful incoprodiagnosis of giardiasis. J. Clin. Microbiol. 23:905-910.
12. Sealy, D. P., and S. H. Schuman. 1983. Endemic giardiasis andday care. Pediatrics 72:154-158.
13. Steketee, R. W., S. Reid, T. Cheng, J. S. Stoebig, R. G.Harrington, and J. P. Davis. 1989. Recurrent outbreaks ofgiardiasis in a child day care center, Wisconsin. Am. J. PublicHealth 79:485-490.
14. Ungar, B. L. P., R. H. Yolken, T. E. Nash, and T. C. Quinn.1984. Enzyme-linked immunosorbent assay for the detection ofGiardia lamblia in fecal specimens. J. Infect. Dis. 149:90-96.
15. Wienecka, J., E. Olding-Stenkvist, H. Schroder, and G. Huldt.1989. Detection of Giardia antigen in stool samples by a semi-quantitative enzyme immunoassay (EIA) test. Scand. J. Infect.Dis. 21:443-448.
16. Wolfe, M. S. 1979. Giardiasis. Pediatr. Clin. North Am. 26:295-303.
17. Yolken, R. H., and B. Ungar. 1985. ELISA detection of Giardialamblia. Lancet ii:1120. (Letter.)
