JOURNAL OF CLINICAL MICROBIOLOGY, Apr. 1989, p. 752-7570095-1137/89/040752-06$02.00/0Copyright 1989, American Society for Microbiology

Evaluation of Culture, Immunofluorescence, and Serology for theDiagnosis of Pertussis

S. A. HALPERINl2* R. BORTOLUSSI,"2 AND A. J. WORT2

Departments of Pediatrics' and Microbiology,2 Dalhousie University and the Izaak Walton Killam Hospital for Children,Halifax, Nova Scotia, Canada B3J 3G9

Received 1 September 1988/Accepted 3 January 1989

Nasopharyngeal culture, direct immunofluorescence, and serology of acute-phase and paired serum

specimens were compared for the laboratory diagnosis of infections due to Bordetella pertussis in a

community-based pediatric population with both high vaccine usage and high pertussis incidence. In 77 (37%)of 210 patients evaluated, one or more tests were positive for pertussis. A clinical illness compatible withpertussis was present in 52 (71%) of 73 pertussis test-positive and 42 (35%) of 119 test-negative patients (P <0.001). Nasopharyngeal culture was of low sensitivity (20 [26%] of 77 positive tests) but was most commonlyconfirmed by another positive pertussis test (85%). Direct immunofluorescence was both insensitive andnonspecific; only 6 (30%) of 20 cases positive by culture were positive by immunofluorescence, and only 4 (33%)of 12 of the culture-negative, immunofluorescence-positive cases could be confirmed by another positivepertussis test. Although serology by enzyme immunoassay proved to be the most sensitive of the laboratory tests(87%), this sensitivity could be achieved only by assaying both acute-phase and paired serum specimens andmeasuring immunoglobulin G (IgG), IgA, and IgM antibodies to two pertussis antigens (pertussis toxin andfilamentous hemagglutinin). Loss of sensitivity occurred with any reduction in the number of these serologicassays performed. Optimal laboratory diagnosis of endemic pertussis in a pediatric population requires bothnasopharyngeal culture and serology by enzyme immunoassay.

The unavailability of sensitive laboratory tests for thediagnosis of pertussis hampers investigators who attempt toevaluate the epidemiology of the disease or the efficacy ofthe vaccine in preventing illness. Techniques routinely avail-able in clinical laboratories include culture and direct immu-nofluorescence on nasopharyngeal secretions (2, 30); how-ever, both tests lack sensitivity, and the latter also lacksspecificity (3, 31). Some investigators have recently sug-gested that serologic methods may increase the diagnosticsensitivity. Such tests have usually measured specific anti-body to the whole organism, Bordetella pertussis, in na-sopharyngeal secretions (10) or in serum (21, 27, 39). Morerecently, these assays have been modified to detect antibod-ies to purported B. pertussis virulence factors (filamentoushemagglutinin [FHA] and pertussis toxin [PT]) in serum (1,13, 15, 38) and in secretions (11). Unfortunately, the lack ofstandardized methods has prevented these tests from be-coming universally available.

Difficulties in a prompt laboratory diagnosis of pertussisderive from the nature of the pathophysiology of the infec-tion caused by B. pertussis. Typical signs and symptoms ofpertussis do not occur until week 3 of infection, during theparoxysmal phase of illness (33). However, isolation of B.pertussis is most consistent in the catarrhal or early parox-ysmal stage, often before the diagnosis of pertussis is sus-

pected (24). In addition, infants, older children, adults, andpreviously immunized individuals may never develop theclassical clinical manifestations of whooping cough. Al-though diagnostic methods relying on paired serum speci-mens (PS) may be more sensitive, they do not permitconfirmation of infection during the acute phase, sinceproper timing of the follow-up serum is critical (31). Sero-logic methods using a single acute-phase serum specimen(AS) have also been proposed; these rely on combinations of

* Corresponding author.

titers of various antibody classes to differentiate antibodydue to acute infection from antibody due to prior immuniza-tion (39).Improved laboratory methods to aid in the diagnosis of

pertussis are crucial in obtaining accurate epidemiologic dataon pertussis incidence and vaccine efficacy. Recent studieshave evaluated pertussis diagnostic tests which use crudebacterial antigen preparations (21), for diagnosis in unimmu-nized populations (39), or for diagnosis in adults during anoutbreak (38). Although the results of these studies aregenerally favorable toward the tests, it is not clear whetherthese methods would be suitable for laboratory diagnosis ofpertussis in an immunized population or during nonoutbreaksituations. In this study we assessed a number of assays,including nasopharyngeal culture, direct immunofluores-cence, and antibody to PT and FHA in AS and PS, in orderto determine their suitability as diagnostic tests. The studywas done in a community-based pediatric population withboth high vaccine usage and high pertussis incidence.

MATERIALS AND METHODS

Population, specimen collection, and processing. The IzaakWalton Killam Hospital for Children (IWK) is the only pedi-atric hospital and diagnostic center for the city of Halifax(population, 250,000) and serves as a referral center for themaritime provinces of Canada (population, 1,500,000). Theincidence of reported pertussis is approximately 13 cases per100,000 people. Patients in whom pertussis was suspected wereevaluated by nasopharyngeal culture and smear and by serol-ogy. At the time of presentation, secretions were collected bysyringe aspiration through a fine flexible plastic catheter passedthrough the nose into the nasopharynx as previously described(10). In addition, a blood sample was obtained by finger stick,and a questionnaire detailing symptomatology, immunizationstatus, and recent use of antimicrobial agents was completed.

752

Vol. 27, No. 4

on June 29, 2018 by guesthttp://jcm

.asm.org/

Dow

nloaded from

EVALUATION OF METHODS FOR THE DIAGNOSIS OF PERTUSSIS

Arrangements were made for collection of a 3-week follow-upserum specimen; patients were reminded by telephone prior totheir appointment. Nasopharyngeal secretions were immedi-ately transported in the catheter from the Outpatient Diagnos-tic Centre to the microbiology laboratory of the IWK. There,nasopharyngeal secretions were expelled onto agar plates forculture and onto two clean glass slides for immunofluores-cence. Smears were processed by treatment with aprotinin (4U/ml) (18) and were stored at 4°C until examination. Serumwas separated and stored at -20°C. Most specimens were

completely processed within 1 h of collection, thereby elim-inating the need for transport media; the few specimens col-lected overnight were held at 4°C prior to processing.

Control sera. A single serum specimen was collected fromeach of 210 children with no symptoms of respiratory illnesswho were admitted to the hospital for elective minor sur-

gery. The control group included 88 children (42%) under-going either insertion or removal of tympanostomy tubes, 53(25%) having dental extractions or restorations, 14 (7%)undergoing circumcision or repair of circumcision, 13 (6%)undergoing herniorrhaphy, 13 (6%) undergoing superficialcyst removal or other minor plastic surgery procedures, and23 (11%) undergoing other miscellaneous procedures (orchi-opexy, cystoscopy, strabismus repair, etc.). The mean agewas 6.5 years, with a range of 2 months to 15.9 years; 44%were less than 5 years of age. The immunization status ofthese children was not determined; however, immunizationrates for the community were 94.8% in children under 5years and increased to 98.3% after school entry (unpublishedstatistics, Nova Scotia Department of Health and Fitness).

Nasopharyngeal culture. Nasopharyngeal secretions werecultured on charcoal agar CM 119 (Oxoid Ltd., London,England) supplemented with 5% defibrinated horse bloodand 40 ,ug of cephalexin per ml (35). Secretions were

expelled directly onto two charcoal plates, one with and onewithout the cephalexin supplement. Plates were incubated at36°C in a humid environment and were examined daily for 7days for colonies typical of B. pertussis. Colonies withtypicàI morphology were examined by Gram stain, andorganisms with characteristic gram-negative morphologywere confirmed by agglutination with B. pertussis antiserumand by absence of agglutination with B. parapertussis anti-serum (Wellcome Research Laboratories, Beckenham,England).

Immunofluorescence. Direct smears of nasopharyngeal se-cretions were examined by standard methods (32). Slideswere heat fixed and stained with either fluorescein-isothio-cyanate-conjugated B. pertussis or B. parapertussis antise-rum (Difco Laboratories, Detroit, Mich.). Positive and neg-ative control slides were included in all assays; a smear wasconsidered positive if three or more organisms with brightfluorescence.and typical morphology were observed on theslide stained with B. pertussis antiserum in the absence offluorescence on the B. parapertussis-stained slides. Thesingle observer who read all the slides was unaware ofwhether the smear was from a patient or was a positivecontrol.Enzyme immunoassay. Antibodies to PT and FHA in AS

and PS were measured by enzyme immunoassay. PT, puri-fied by the method of Sekura et al. (37), and FHA, purifiedby the method of Sato et al. (36), were adsorbed underalkaline conditions to duplicate wells of polystyrene micro-dilution plates (Nunc, Roskilde, Denmark) at concentrationsof 5 ,ug/ml for 2 h at 37°C and then at 4°C overnight.Following a series of washes in phosphate-buffered salinewith 0.05% Tween 20 (Sigma Chemical Co., St. Louis, Mo.)

TABLE 1. Comparison of IWK positive control sera with U.S.reference pertussis antiserum hutnan lot 3

U.S. referenceIWK control pertussis antiserumAntigen Antibody serum antibody lot 3

specificity class titer'eAntibody ELISAtiter unitageb

PT IgG 2,400 3,977 200IgA 600 162 15IgM 201 505 15

FHA IgG 1,200 1,400 200IgA 1,000 1,182 100IgM 71 420 15

"As calculated in Materials and Methods.b Enzyme-linked immunosorbent assay (ELISA) as calculated by Manclark

et al. (25).

(PBS-T), remaining binding sites were blocked by a 1-hincubation at 37°C with PBS-T-10% normal goat serum(Woodlyn Laboratories, Guelph, Ontario, Canada). Follow-ing another series of washes, 100 pi of a 1:106 dilution ofserum to be tested was added to duplicate wells. After ànincubation at 37°C of 1 h (for immunoglobulin G [IgG] andIgA assays) or 24 h (for IgM assays) and another series ofwashes, 100 ,ul of peroxidase-conjugated goat F(ab')l frag-ment antibody to either human IgG (Fc specific; dilution,1:20,000), IgA (a. specific; dilution, 1:4,000) or IgM (puspecific; dilution, 1:1,000) (all from Organon Teknika, Mal-vern, Pa.) was added to the wells, and they were incubatedat 37°C for 1 h. Following a final series of washes, substrateconsisting of o-phenylenediamine and hydrogen peroxide ina citrate-phosphate buffer was added to the plates. Afterincubation for 30 min at room temperature (-22°C), thereaction was stopped with 100 pul of 4 N sulfuric acid and theA492 was determined in a Biotek 309 microelisa reader(Biotek Corp., Winooski, Vt.). Sera with high IgG, IgA, orIgM titers to PT dr FHA, obtained from children with naturalpertussis infection (IWK positive control sera), were seriallydiluted and run with each assay.The enzyme-linked immunosorbent assay titer was de-

fined as the reciprocal of the serum dilution that resulted inan A492 at 30 min of 0.200. The titer was derived from theequation for the straight line described by the IWK positivecontrol serum and the absorbance of the test serum at adilution of 1:100 (16, 17). Standard sera, kindly provided byCharles Manclark at the Center for Drugs and Biologics,Food and Drug Administration, Bethesda, Md. (U.S. refer-ence pertussis antiserum, human, lot 3), were assayed andcompared with the IWK positive control sera used for thisstudy (Table 1).

In PS, a fourfold rise in IgG or IgA or a fourfold rise or fallin IgM to PT or FHA was interpreted as a positive test. OnlyIgA and IgM were evaluated in single serum specimens;samples exceeding the mean of the control group by 2standard deviations or more were interpreted as positive(Table 2). A patient was considered to have a serologicdiagnosis of pertussis when one or more of the followingconditions were met: (i) fourfold rise in IgG titer to PT orFHA, (ii) fourfold rise in IgA titer to PT or FHA, (iii)fourfold rise or fall in IgM titer to PT or FHA, or (iv) positiveIgM and IgA titer in AS.Data analysis and ethical considerations. Sensitivity, spec-

ificity, and positive and negative predictive values weredetermined by standard methods (14, 34). False-positive and

753VOL. 27, 1989

on June 29, 2018 by guesthttp://jcm

.asm.org/

Dow

nloaded from

754 HALPERIN ET AL.

TABLE 2. PT and FHA antibody measured by enzymeimmunoassay in sera from 210 control children

Immunoglobulin Antibody titerand antigen Oa Mean Highest Mean ± 2 SD

IgAPT 173 12 945 28FHA 183 104 2,000 607

IgMPT 208 14 65 38FHA 208 27 192 140

" Number of sera assayed.

false-negative rates can be calculated by subtracting speci-ficity and sensitivity values, respectively, from 100 (14);these values are not included in the text. Differences inproportions were tested by chi square; Yates' correction wasapplied for contingency tables with 1 df. Student's t test wasused for statistical analysis of the differences in antibodyresponse between groups of patients. All statistical calcula-tio'ns were performed with a computer-programmed statis-tics package (Stats Plus; Human Systems Dynamics, NorthRidge, Calif.) on an Apple IIC personal computer (AppleComputers, Cupertino, Calif.); P c 0.05 was consideredstatistically significant.The protocol for this study was approved by the ethical

review committee of the IWK.

RESULTS

Between November 1986 and December 1987, specimensfrom -502 patients were submitted for the diagnosis of per-tussis. In 210 cases (42%o), specimens included a nasopha-ryngeal aspirate for culture (NPAC) and smear direct fluo-rescent-antibody assay (DFA) and at least one serumsample. This subgroup constituted the study population. In150 (71%) of 210 cases, PS were available. In 77 (37%) of the210 patients, one or more tests were positive for pertussis.Culture was positive in 20 patients (9.5%), and immunoflu-orescence was positive in 18 (8.6%). A serologic diagnosiscould be made in 67 patients, i.e., in 25 (12%) of 210 AS and48 (32%) of 150 cases in which PS were available (thisincluded 6 specimens also meeting the diagnostic criteria forAS).

Culture and immunofluorescence. Of the 77 pertussis test-positive patients, 20 (26%) were positive by NPAC and 18(23%) were positive by DFA. Six patients were positive byboth NPAC and DFA. A total of 17 (85%) of 20 NPAC-positive patients and 10 (56%) of 18 DFA-positive patientswere also positive by serologic criteria. A total of 13 (93%) of14 NPAC-positive DFA-negative cases and 4 (33%) of 12NPAC-negative DFA-positive cases were also positive byserologic criteria. Assuming that culture positivity and afourfold antibody rise are indicative of true-positives, NPAChad a sensitivity of 37%, a specificity of 100%, a positivepredictive value of 100%, and a negative predictive value of76%. DFA had values of 18, 93, 56, and 69%, respectively.By the less stringent criteria for true positivity of positiveNPAC, DFA, PS, or AS, culture had a sensitivity of 26%, aspecificity of 100%, a positive predictive value of 100%, anda negative predictive value of 60%; DFA had values of 22,100, 100, and 58%, respectively.

Serology. Mean IgG, IgA, and IgM antibody titers to PTand FHA were higher in the test-positive group than in thetest-negative group (Table 3; P < 0.05 for all comparisons).Mean antibody titers in the test-negative group did not differstatistically from those that had been assessed in the controlgroup (IgA and IgM to PT and FHA; Table 2).

AS. AS was available from all 210 patients; for 60 patients,this was the-only serum specimen. A total of 25 (12%) of thepatients had high IgM and IgA titers to PT or FHA in theirfirst serum specimen; this accounted for 32% (25 of 77) of thepositive patients. Only four patients positive by AS had apositive NPAC, suggesting that the organism may disappearcoincidentally with antibody appearance. Sensitivity was14% for single-specimen serology when stringent true-posi-tive criteria were used and 31% by the less stringent defini-tion. Specificity and positive and negative predictive valueswere 85 and 100%, 32 and 100%, and 66 and 61%, respec-tively, for stringent and less stringent criteria. The use ofacute-phase serology detected 65% of positive patients whencombined with NPAC and 76% when combined with NPACand DFA.

PS. Two serum samples were available from each of 150patients; 48 (32%) demonstrated a fourfold rise in IgG, IgA,or IgM titer to PT or FHA or a fourfold decrease in IgM. PSserology was the laboratory test which most often contrib-uted to a positive result (67% of those positive); 17 (35%) of48 patients positive by PS serology also had a positive

TABLE 3. PT and FHA antibody titers on 210 patients whose specimens were submitted for diagnosis of pertussis

Mean titer (SEM)Laboratory test Condition (no.) of PT FHA

status patientsIgG IgA IgM IgG IgA IgM

Negative Acute (133) 582 (99) 28 (7.2) 14 (0.9) 292 (41) 213 (58) 23 (2.6)Convalescent (78) 602 (126) 14 (3.9) 16 (1.5) 246 (40) 293 (107) 25 (4.0)

PositiveTotal Acute (77) 1,267 (194) 99 (23) 54 (9.2) 490 (67) 466 (101) 63 (11)

Convalescent (72) 1,823 (196) 112 (28) 49 (5.9) 700 (80) 755 (158) 63 (9.5)NPAC Acute (20) 999 (331) 110 (46) 58 (24) 442 (102) 468 (174) 51 (15)

Convalescent (19) 2,342 (276) 144 (38) 46 (8.8) 746 (142) 1,301 (467) 67 (26)FA Acute (18) 437 (121) 30 (11) 22 (6.2) 348 (140) 295 (144) 22 (5.1)

Convalescent (16) 1,507 (414) 207 (112) 40 (11) 481 (201) 637 (362) 27 (5.4)PS Acute (48) 681 (190) 47 (18) 36 (11) 417 (87) 259 (68) 62 (15)

Convalescent (48) 1,678 (235) 128 (41) 49 (8.0) 726 (108) 640 (177) 62 (12)AS Acute (25) 2,866 (335) 220 (54) 133 (21) 932 (127) 1,045 (254) 106 (17)

Convalescent (22) 3,017 (305) 84 (14) 60 (8.8) 1,027 (121) 1,330 (328) 81 (15)

J. CLIN. MICROIBIOL.

on June 29, 2018 by guesthttp://jcm

.asm.org/

Dow

nloaded from

EVALUATION OF METHODS FOR THE DIAGNOSIS OF PERTUSSIS

TABLE 4. Sensitivity of antibody-class measurements in paired sera for the diagnosis of pertussis

Sensitivity (<X) with:Antigen(s)

IgG IgA IgM IgG + IgA IgG + IgM IgA + IgM IgG + IgA + IgM

PT 42 44 33 52 63 67 73FHA 29 21 27 38 48 46 56PT and FHA 58 54 48 73 85 88 100

NPAC. When either definition of true positivity (see above)is used, sensitivity (94 or 67%), specificity (100%), positivepredictive value (100%), and negative predictive value (97 or76%) all compared favorably with values of the other diag-nostic tests evaluated.Of the 48 patients with serologic evidence in PS of

infection, 34 had a rise in antibody to PT and 27 had a rise inantibody to FHA. An IgG rise was demonstrated in 28, anIgA rise was demonstrated in 26, and an IgM rise or fall wasdemonstrated in 23 patients. No antibody class or pertussisantigen alone was adequate to assess infection by PS serol-ogy. Measurement of a single antibody class to a singleantigen detected 21 to 44% of the total seropositive cases(Table 4); the use of two antigens increased this to as high as58%. Measuring two antibody classes to a single antigendetected 38 to 67% of positive cases, while measuring twoantibody classes to two antigens made possible a detectionrate of 88%. An additional six cases (12%) were detectedonly by measurement of IgG, IgA, and IgM to both PT andFHA.The data were also assessed excluding the fluorescent

antibody results. Of the 66 specimens positive by NPAC, orby PS or AS serology, 19 (29%) were positive by NPAC, 48(73%) were positive by PS serology, and 22 (33%) were

positive by AS serology. NPAC and PS serology togetheridentified 51 (77%); NPAC and AS serology detected 37(56%). Use of serologic PS or AS criteria without cultureyielded 63 (95%) positives.

Clinical data. Clinical information was available for anal-ysis on 192 patients; 119 patients were negative by alldiagnostic tests, and 73 patients were positive by one ormore assays.

(i) Age. The mean age of patients with a positive pertussistest was 4.9 years; the mean for patients who tested negativewas 3.6 years (P = 0.01). There was no statistically signifi-cant difference in age of patients positive by different tests(Table 5).

(ài) Immunization. Both pertussis test-positive and test-negative patients had received a mean of 3.6 doses ofpertussis vaccine prior to onset of illness (P > 0.5). The

number of pertussis immunizations received by childrenpositive by the various diagnostic tests ranged from 3.2 to4.0 (Table 5); the differences were not statistically signifi-cant. A total of 17 (13Y%) of 133 test-negative patients and 14(19%) of 73 test-positive patients had received fewer thanthree immunizations (P = 0.3); half of these patients wereunder 6 months of age.Immunity to B. pertussis following vaccination decreases

with the interval since immunization, with no demonstrableprotection by 12 years (20). A total of 78% of the patients inthis study had received a pertussis vaccination in the previ-ous 2 years. Patients with negative pertussis tests had amean interval since the most recent pertussis vaccine of 11.8months. Patients with positive tests tended to have a longerinterval since the last immunization (18.1 months) (Table 5);however, this reached statistical significance only in thegroup positive by AS serology criteria (22.1 months; P =

0.04).(iii) Clinical illness. A total of 52 (71%) of 73 of pertussis

test-positive patients and 42 (35%) of 119 of pertussis test-negative patients had an illness characteristic of pertussis(cough for more than 2 weeks, with vomiting, cyanosis,paroxysms, or apnea) (P < 0.001). A clinically compatibleillness was present in significantly more patients positive byeach test than in test-negative patients (Table 5) (P < 0.006).There was no difference demonstrated among the varioustests. Duration of symptoms tended to be longer in test-positive patients but reached statistical significance only inpatients positive by NPAC (34.1 versus 21.5 days; P = 0.02)and AS serology criteria (32.3 versus 21.5 days; P = 0.04).

(iv) Antibiotic use. A total of 63 patients (33%) werereceiving antibiotics at the time of specimen collection; 37(19%) of 192 were receiving an erythromycin-containingpreparation. Three of these patients had a positive NPAC.No significant difference in antibiotic use was found betweenpertussis test-positive and test-negative patients (Table 5).

DISCUSSIONEvaluation of pertussis diagnostic tests is made difficult

because there is no sensitive and consistently positive test (a

TABLE 5. Clinical data on 192 patients whose specimens were submitted for diagnosis of pertussis

Patients Immunizations Signs and symptomsLaboratory test status Mean age Mean no. Mean interval since No. (%) meeting case Mean interval since No. (%) receiving

(yr) received last dose (mo) definition" onset (days) erythromycin

Negative 119 3.6 3.6 11.8 42 (35) 21.5 21 (18)

PositiveTotal 73 4.9 3.6 18.1 52 (71) 26.6 16 (22)NPAC 20 4.4 3.5 19.2 16 (80) 34.1 3 (15)DFA 18 4.8 3.8 20.6 13 (72) 23.0 6 (33)PS 47 4.3 3.2 16.4 32 (68) 26.0 8 (17)AS 23 5.9 4.0 22.1 20 (87) 32.3 4 (17)

Cough longer than 1 week, with paroxysms. vomiting, apnea, or cyanosis.

VOL. 27? 1989 755

on June 29, 2018 by guesthttp://jcm

.asm.org/

Dow

nloaded from

756 HALPERIN ET AL.

so-called "gold standard") against which other tests can beassessed. The use of multiple diagnostic tests in combinationmay result in improved sensitivity, but often with a concur-rent loss of specificity (14). Additionally, evaluation ofdiagnostic tests in an outbreak situation can mask tests oflow specificity, since the true positive rate is so uncommonlyhigh. In this study we assessed this issue by using acombination of tests as potential "gold standards," andthese were then compared with single or multiple combina-tions of tests. The study was performed during a 14-monthperiod of moderate endemic pertussis activity rather thanduring an outbreak. During this period, 20 culture-confirmedcases occurred; by contrast, 88 culture-confirmed cases hadoccurred during an epidemic year encompassed in the pre-ceding 14-month period (unpublished data).Nasopharyngeal culture proved to be a test of low sensi-

tivity but high specificity, while direct immunofluorescencelacked both sensitivity and specificity. These results of ourassessment of methods for detecting bacteria in nasopharyn-geal secretions are similar to those in previous reports. By avariety of criteria, culture sensitivity has been estimated tobe 30 to 60% (6, 12, 13, 15, 27, 28, 38, 39; S. A. Young, G. L.Anderson, and P. D. Mitchell, Clin. Microbiol. Newsl. 9:176-179, 1987), while direct immunofluorescence has beenreported to be positive in 20 to 100% of culture-positivecases (5, 7-9, 19, 26; Young et al., Clin. Microbiol. Newsl.).Our apparent false-positivity rate of 44% (those positive byDFA only) is higher than that in other reports (0 to 40%) (5,8, 9, 19; Young et al., Clin. Microbiol. Newsl.) and may beattributable to a number of factors. Most important is theactual definition of false positivity, since our study requiredanother positive laboratory test to confirm the DFA results,whereas other researchers often used a compatible clinicalillness as sufficient confirmation. Observer inexperience andvariability have also been linked to the difficulties with theDFA assay (3); however, the use of a single experiencedobserver in this study should have decreased this type oferror. Prior use of antimicrobial agents has often beenadvanced to explain DFA-positive culture-negative speci-mens (31; Young et al., Clin. Microbiol. Newsl.). Indeed, ahigher proportion of DFA-positive culture-negative patients(5 [56%] of 9) than of DFA-positive culture-positive patients(1 [17%] of 6) were receiving erythromycin. These results didnot reach statistical significance.Antibody determinations in PS proved to be the most

sensitive ofthe laboratory tests; however, no single antibodyor antigen test was sufficient to consistently confirm thediagnosis. Although optimal results were achieved by usingtwo antigens and measuring three antibody classes, it ispossible that additional cases could have been detected byusing additional antigens or by increasing the interval be-tween the two serum specimens. The ideal serologic testwould be available for a single specimen at the time ofpatient presentation and would differentiate between naturalinfection and immunity derived through immunization. Fol-lowing infection or vaccination, both IgG and IgM antibodiesare formed; IgA is elicited only following infection (4, 22, 23,27, 29, 39). Since IgG antibodies are long lived, a rising titerin PS is required for the diagnosis of pertussis. In contrast,IgM antibodies have a short half-life, and their presenceindicates recent exposure to pertussis antigens. IgA mayremain elevated for 6 months to 2 years (23, 29). Thus, thepresence in a single serum of both IgA and IgM suggests arecent natural infection.The single-serum test results suffered from a lack of

sensitivity and specificity, since positive samples frequently

were not confirmed by other tests. It is possible that theserologic cutoff titers used for IgM and IgA were too low andthat children with high titers from previous immunization orillness were included. This is unlikely, since only titersgreater than 2 standard deviations above those of a compa-rably aged population were considered positive and sincepositivity was limited to individuals exceeding those valuesfor both IgM and IgA. None of the 210 healthy controlchildren met these criteria for infection. Similarly, the erro-neous inclusion of children with persistent high IgA from aprevious natural infection and high IgM from a recentimmunization is unlikely, since children with previous infec-tion would more likely have an anamnestic IgG rather thanIgM rise (12). Interestingly, the interval since a previouspertussis vaccine dose was longest in children positive byAS criteria. It is more likely that our criteria for seroposi-tivity on single specimens were too restrictive, particularlysince the control children came from the community inwhich there is a high rate of endemic pertussis infection.However, these criteria have previously correlated well withclinical pertussis (38, 39); in our study, over 91% of patientspositive by AS serology criteria had a compatible illness.

It is not surprising that there was poor correlation betweenpatients testing positive by culture and PS serology andthose testing positive by AS serology. Nasopharyngeal cul-tures for pertussis are uniformly positive early in the illnessand become negative during the paroxysmal stage (24, 31,33). Similarly, patients demonstrating a rise in antibody topertussis antigens would likely have been seen early in theirillnesses, at the time the first serum specimen was taken. Incontrast, patients in whom high titers are already present atthe time of first sampling would tend to be later in theircourse, when cultures are consistently negative. Indeed, thepatients positive by AS serology criteria had the longestinterval since onset of symptoms.The results of this study corroborate and extend the work

of previous investigators with certain exceptions. In our

study, DFA was difficult to interpret and of low specificityand may be best suited for areas where there is a highpertussis prevalence or for an outbreak situation. Like otherinvestigators (12, 15, 39), we found that serologic methodswere the most sensitive and specific but that a combinationof assays was required for optimal results. Our data supportthe findings of Steketee et al. (38), which demonstrated theutility of the enzyme immunoassay with purified pertussisantigens for the diagnosis of pertussis during an outbreakamong institutionalized adults, and we extend these obser-vations to include endemic community-based illness in a

pediatric population. In contrast to their study, ours was

unable in most cases to correlate the type of pertussistest-positive case with the duration of illness. However,more-accurate clinical data from a closed adult residentialfacility were probably available in their population and mayexplain these differences.

Pertussis continues to be an elusive infection in terms ofits immunity, diagnosis, and control through immunization.Our report documents the continued need for improvementsin diagnostic tests for pertussis and demonstrates that opti-mal laboratory diagnosis currently requires both nasopha-ryngeal culture and serology by enzyme immunoassay.Epidemiologic data based only on typical clinical presenta-tion and culture are likely to underestimate the true inci-dence of pertussis in a community.

J. CLIN. MICROBIOL.

on June 29, 2018 by guesthttp://jcm

.asm.org/

Dow

nloaded from

EVALUATION OF METHODS FOR THE DIAGNOSIS OF PERTUSSIS

ACKNOWLEDGMENTS

This research was supported by National Health Research andDevelopment Program grants 6603-1244-54 and 6603-1245-54 fromHealth and Welfare Canada.We thank Edith Foy for technical assistance and Nancy Steel for

preparation of the manuscript.

LITERATURE CITED1. Askelof, P., M. Granstrom, P. Gillenius, and A. A. Lindberg.

1982. Purification and characterisation of a fimbrial haemagglu-tinin from Bordetella pertussis for use in an enzyme-linkedimmunosorbent assay. J. Med. Microbiol. 15:73-83.

2. Bradstreet, C. M. P., A. J. Tannahill, and J. M. B. Edwards.1972. Detection of Bordetella pertessis antibodies in human sera

by complement-fixation and immunofluorescence. J. Hyg. 70:75-83.

3. Broome, C. V., D. W. Fraser, and W. J. English. 1978. Pertussis-diagnostic methods and surveillance, p. 19-22. In C. R. Man-clark and J. C. Hill (ed.), International Symposium on Pertus-sis. National Institutes of Health, Bethesda, Md.

4. Burstyn, D. G., L. J. Baraff, M. S. Peppler, R. D. Leake, J. St.Geme, Jr., and C. R. Manclark. 1983. Serological response tofilamentous hemagglutinin and lymphocytosis-promoting toxinof Bordetella pertussis. Infect. Immun. 41:1150-1156.

5. Chalvardjian, N. 1966. The laboratory diagnosis of whoopingcough by fluorescent antibody and by culture methods. Can.Med. Assoc. J. 95:263-266.

6. Combined Scottish Study. 1970. Diagnosis of whooping cough:comparison of serological tests with isolation of Bordetellapertussis. Br. Med. J. 4:637-639.

7. Donaldson, P., and J. A. Whitaker. 1960. Diagnosis of pertussisby fluorescent antibody staining of nasopharyngeal smears. Am.J. Dis. Child. 99:423-427.

8. Field, L. H., and C. D. Parker. 1977. Pertussis outbreak inAustin and Travis County, Texas, 1975. J. Clin. Microbiol.6:154-160.

9. Gilligan, P. H., and M. C. Fisher. 1984. Importance of culture inlaboratory diagnosis of Bordetella pertussis infections. J. Clin.Microbiol. 20:891-893.

10. Goodman, Y. E., A. J. Wort, and F. L. Jackson. 1981. Enzyme-linked immunosorbent assay for detection of pertussis immuno-globulin A in nasopharyngeal secretions as an indicator ofrecent infection. J. Clin. Microbiol. 13:286-292.

11. Granstrom, G., P. Askelof, and M. Granstrom. 1988. Specificimmunoglobulin A to Bordetella pertussis antigens in mucosalsecretion for rapid diagnosis of whooping cough. J. Clin. Micro-biol. 26:869-874.

12. Granstrom, M., G. Granstrom, A. Lindfors, and P. Askelof.1982. Serologic diagnosis of whooping cough by an enzyme-linked immunosorbent assay using fimbrial hemagglutinin as

antigen. J. Infect. Dis. 146:741-745.13. Granstrom, M., A. A. Lindberg, P. Askelof, and B. Hederstedt.

1982. Detection of antibodies in human serum against thefimbrial haemagglutinin of Bordetella pertuessis by enzyme-linked immunosorbent assay. J. Med. Microbiol. 15:85-96.

14. Griner, P. F., R. J. Mayewski, A. I. Mushlin, and P. Greenland.1981. Selection and interpretation of diagnostic tests and proce-dures. Ann. Intern. Med. 94:553-600.

15. Hakansson, S., C. G. Sundin, M. Granstrom, and B. Gastrin.1984. Diagnosis of whooping cough-a comparison of culture,immunofluorescence and serology with ELISA. Scand. J. In-fect. Dis. 16:281-284.

16. Halperin, S. A., P. Ferrieri, E. D. Gray, E. L. Kaplan, and L. W.Wannamaker. 1987. Antibody response to bacteriophage hyalu-ronidase in acute glomerulonephritis after group A streptococ-cal infection. J. Infect. Dis. 155:253-261.

17. Halperin, S. A., E. D. Gray, P. Ferrieri, and L. W. Wannama-ker. 1985. Enzyme-linked immunosorbent assay for identifica-tion and measurement of antibodies to group A streptococcalbacteriophage. J. Lab. Clin. Med. 106:505-511.

18. Harris, P. P., B. Thomason, and R. M. McKinney. 1980.Preservation of nasopharyngeal smears for fluorescent antibodydetection of Bordetella pertussis. J. Clin. Microbiol. 12:799-801.

19. Holwerda, J., and G. Eldering. 1963. Culture and fluorescent-antibody methods in diagnosis of whooping cough. J. Bacteriol.86:449-451.

20. Lambert, H. J. 1965. Epidemiology of a small pertussis outbreakin Kent County, Michigan. Public Health Rep. 80:365-369.

21. Lawrence, A. J., and J. C. Paton. 1987. Efficacy of enzyme-linked immunosorbent assay for rapid diagnosis of Bordetellapertussis infection. J. Clin. Microbiol. 25:2102-2104.

22. MacAulay, M. E. 1979. The serological diagnosis of whoopingcough. J. Hyg. 83:95-102.

23. MacAulay, M. E. 1981. The IgM and IgG response to Bordetellapertussis vaccination and infection. J. Med. Microbiol. 14:1-7.

24. Manclark, C. R., and J. L. Cowell. 1984. Pertussis, p. 69-106. liR. Germanier (ed.), Bacterial vaccines. Academic Press, Inc..New York.

25. Manclark, C. R., B. D. Meade, and D. G. Burstyn. 1986.Serological response to Bordetella pertissis, p. 388-394. hIN. R. Rose, H. Friedman. and J. L. Fahey (ed.), Manual ofclinical laboratory immunology, 3rd ed. American Society forMicrobiology, Washington, D.C.

26. Marcon, M. J., A. C. Hamoudi, H. J. Cannon, and M. M.Hribar. 1987. Comparison of throat and nasopharyngeal swabspecimens for culture diagnosis of Bordetella pertissis infec-tion. J. Clin. Microbiol. 25:1109-1110.

27. Mertsola, J., T. Kuronen, A. Turunen, M. K. Viljanen, and O.Ruuskanen. 1984. Diagnosis of pertussis. J. Infect. Dis. 8:149-156.

28. Mertsola, J., O. Ruuskanen, T. Kuronen, and M. K. Viljanen.1983. Serologic diagnosis of pertussis: comparison of enzyme-linked immunosorbent assay and bacterial agglutination. J.Infect. Dis. 147:252-257.

29. Nagel, J., and E. J. Poot-Scholtens. 1983. Serum IgA antibody toBordetella periussis as an indicator of infection. J. Med. Micro-biol. 16:417-426.

30. Nelson, J. D., B. Hempstead, R. Tanaka, and F. P. Pauls. 1964.Fluorescent antibody diagnosis of infections. J. Am. Med.Assoc. 188:1121-1124.

31. Onorato, I. M., and S. G. F. Wassilak. 1987. Laboratorydiagnosis of pertussis: the state of the art. Pediatr. Infect. Dis.J. 6:145-151.

32. Parker, C. D., and B. J. Payne. 1985. Bordetella, p. 394-399. InE. H. Lennette, A. Balows, W. J. Hausler, Jr., and H. J.Shadomy (ed.), Manual of clinical microbiology, 4th ed. Amer-ican Society for Microbiology, Washington, D.C.

33. Pittman, M. 1984. The concept of pertussis as a toxin-mediateddisease. Pediatr. Infect. Dis. J. 3:467-486.

34. Radetsky, M., and J. K. Todd. 1984. Criteria for the evaluationof new diagnostic tests. Pediatr. Infect. Dis. J. 3:461-466.

35. Regan, J., and F. Lowe. 1977. Enrichment medium for theisolation of Bordetella. J. Clin. Microbiol. 6:303-309.

36. Sato, Y., J. L. Cowell, H. Sato, D. G. Burstyn, and C. R.Manclark. 1983. Separation and purification of the hemaggluti-nins for Bordetella peruissis. Infect. Immun. 41:313-320.

37. Sekura, R. D., F. Fish, C. R. Manclark, B. Meade, and Y. L.Zhang. 1983. Pertussis toxin: affinity purification of a newADP-ribosyltransferase. J. Biol. Chem. 258:14647-14651.

38. Steketee, R. W., D. G. Burstyn, S. G. F. Wassilak, W. N. Adkins,M. B. Polyak, J. P. Davis, and C. R. Manclark. 1988. A com-parison of laboratory and clinical methods for diagnosing per-tussis in an outbreak in a facility for the developmentallydisabled. J. Infect. Dis. 157:441-449.

39. Viljanen, M. K., O. Ruuskanen, C. Granberg, and T. T. Salmi.1982. Serological diagnosis of pertussis: IgM, IgA, and IgGantibodies against Bordetella periussis measured by enzyme-linked immunosorbent assay (ELISA). Scand. J. Infect. Dis.14:117-122.

VOL. 27, 1989 757

on June 29, 2018 by guesthttp://jcm

.asm.org/

Dow

nloaded from