A Comparison of Gross Pathology, Histopathology, andMycobacterial Culture for the Diagnosis of Tuberculosis in Elk

(Cervus elaphus)

Elizabeth B. Rohonczy, Arumuga V. Balachandran, Thomas W. Dukes, Janet B. Payeur, Jack C. Rhyan,Dennis A. Saari, Terry L. Whiting, Susan H. Wilson, Jerald L. Jarnagin

ABSTRACT

Using the isolation of Mycobac-terium bovis as the reference stan-dard, this study evaluated the sensi-tivity, specificity and kappa statisticof gross pathology (abattoir post-mortem inspection), histopathology,and parallel or series combinationsof the two for the diagnosis oftuberculosis in 430 elk and reddeer. Two histopathology interpre-tations were evaluated: histopathol-ogy I, where the presence of lesionscompatible with tuberculosis wasconsidered positive, and histo-pathology II, where lesions compat-ible with tuberculosis or a selectgroup of additional possible diag-noses were considered positive. Inthe 73 animals from which M. boviswas isolated, gross lesions of tuber-culosis were most often in the lung(48), the retropharyngeal lymphnodes (36), the mesenteric lymphnodes (35), and the mediastinallymph nodes (16). Other mycobac-terial isolates included: 11 M. para-tuberculosis, 11 M. avium, and 28rapidly growing species or M. terraecomplex. The sensitivity estimatesof gross pathology and histopathol-ogy I were 93% (95% confidencelimits [CL] 84,97%) and 88%[CL 77,94%], respectively, and thespecificity of both was 89%[CL 85,92%]). The sensitivity andspecificity of histopathology II were89% (CL 79,95%) and 77%(CL 72,81%), respectively. Thehighest sensitivity estimates (93-95% [CL 84,98%]) were obtainedby interpreting gross pathology andhistopathology in parallel (where ananimal had to be positive on at leastone of the two, to be classified as

combination positive). The highestspecificity estimates (94-95%[CL 91-97%]) were generated whenthe two tests were interpreted inseries (an animal had to be positiveon both tests to be classified as com-bination positive). The presence ofgross or microscopic lesions showedmoderate to good agreement withthe isolation of M. bovis (Kappa =

65-69%). The results show thatpost-mortem inspection, histo-pathology and culture do not neces-sarily recognize the same infectedanimals and that the spectra of ani-mals identified by the tests overlaps.

RESUME

En se servant de l'isolement deMycobacterium bovis comme pointde reference, nous avons evalue lasensibilite', la spe'cificite' et la statis-tique kappa des pathologies macro-scopiques (examen post-mortem al'abattoir), de l'histopathologie, etdes combinaisons en parallele ou enserie de ces deux examens pourdiagnostiquer la tuberculose chez430 e'lans et cerfs rouges. Deuxinterpretations histopathologiquesont ete evaluees; histopathologie I,

ou la presence de lesions compati-bles avec la tuberculose etait con-sideree positive, et histopatholo-gie II, ou des lesions compatiblesavec la tuberculose ou un ensemblerestreint de diagnostics additionnelspossibles etaient consideres positifs.Chez les 73 animaux a partir des-quels M. bovis a ete isole, des lesionsmacroscopiques de tuberculoseetaient le plus souvent observees auniveau du poumon (48), des gan-glions lymphatiques retropharynges

(36), des ganglions lymphatiquesmesenteriques (35), et des ganglionslymphatiques mediastinaux (16).D'autres especes de mycobacte'riesont 'egalement 'ete' isole'es; 11M. paratuberculosis, 11 M. avium, et28 autres especes 'a croissancerapide ou appartenant au complexeM. terrae. La sensibilite de lapathologie macroscopique et del'histopathologie I etaient respec-tivement de 93 % (intervalle de con-fiance [IC] 94,97 % avec un seuil deconfiance de 95 %) et de 88 %(IC 77,94 %), alors que leur speci-ficite etait de 89 % (IC 85,92 %). Lasensibilite' et la spe'cificite' de l'his-topathologie II etaient respective-ment, de 89 % (IC 79,95 %) et 77 %(IC 72,81 %). La sensibilite la pluselevee (93-95 % [IC 84,98 %]) etaitobtenue en interpretant en parallelela pathologie macroscopique etl'histopathologie, un animal etantconsidere positif lorsqu'il etait posi-tif a l'un ou l'autre des tests de lacombinaison. La plus grande speci-ficite (94-95 % [IC 91,97]) etaitobtenue quand les deux tests etaientinterpretes en serie, un animal etantconsidere positif lorsque trouvepositif aux deux tests de la combi-naison. La pre'sence de le'sionsmacroscopiques ou microscopiquesmontrait une relation moderee abonne avec l'isolement de M. bovis(Kappa = 65-69 %). Les resultatsdemontrent que l'examen post-mortem, 1'examen histopatho-logique et la culture bacterienne nereconnaissent pas necessairement lememe animal infecte comme positifet qu'un eventail superpose d'ani-maux infectes sont identifies par lesdiff6rentes methodes de diagnostic.

(Traduit par docteur Serge Messier)

Can J Vet Res 1996; 60: 108-114

Animal Diseases Research Institute, Agriculture and Agri-Food Canada Box 11300, Station "H", Nepean, Ontario K2H 8P9 (Rohonczy, Balachandran,Dukes, Wilson); National Veterinary Services Laboratories, USDA, Box 844, Ames, Iowa 50010 USA (Payeur, Rhyan, Saari, Jarnagin); Agricultureand Agri-Food Canada 613-269 Main Street, Winnipeg, Manitoba R3C 1B2 (Whiting).Submitted January 3, 1995.

108

INTRODUCTION

It is often difficult to compare theresults of one trial that evaluates testsfor the diagnosis of tuberculosis tothose of another, or to directly applythe conclusions to control and eradica-tion programs. One reason for this isbecause researchers have not consis-tently used the same reference stan-dard to estimate the sensitivity andspecificity of diagnostic tests forMycobacterium bovis infection. Thepresence of gross lesions consistentwith M. bovis infection has been usedby some investigators to judge the per-formance of diagnostic tests (1,2,3),while others chose the isolation ofM. bovis as their reference standard(4,5). As well, various combinationsof gross pathology, histopathologyand M. bovis isolation have been usedto establish the disease status of ani-mals in tuberculosis studies (6,7,8,9).The Canadian Captive Ungulate

Program (CCUP) for the eradicationof brucellosis and tuberculosis requiresthe mandatory testing of captiveungulates on premises involved in thecommercial production of Cervidae(10). Cervidae on these premises aretested with the mid-cervical intrader-mal tuberculin test at 3-year intervalsand ungulate carcasses are monitoredin abattoirs for lesions compatiblewith mycobacterial infection. In 1989,a British Columbian fallow deer(Dama dama) herd became the firstherd infected with M. bovis to bedetected under the CCUP. During thefollowing 3 years, two separate out-breaks of tuberculosis occurred inOntario and Alberta which eventuallyled to the detection and depopulationof 23 cervid and mixed species herds.Because of the large number of ani-mals scheduled for slaughter, theopportunity was taken to compareabattoir post-mortem inspection andhistopathology with the isolation ofM. bovis for the diagnosis of tubercu-losis in elk and red deer. Our objec-tive was to investigate the relationshipbetween these tests and determinehow that relationship might affect theuse of the tests as reference standards.

MATERIALS AND METHODS

ANIMALS

A total of 502 elk and red deer(Cervus elaphus) were selected from

5 herds from which Mycobacteriumbovis had previously been isolatedfrom at least one animal as a result of aCCUP investigation. The animals wereindividually identified on the quaran-tined farm of origin with tamper-proof,metal eartags. Most of the subjectswere female since most of the testingoccurred during the rut and the maleswere too difficult to handle. The elkwere tuberculin tested and bled fortests of cell-mediated and humoralimmunity. The results of the ante-mortem diagnostic tests are reportedseparately (1 1).

SAMPLE COLLECTION ANDDISTRIBUTION

All of the animals in the study weresubject to a post-mortem examinationconforming to Canadian guidelinesfor the inspection of cattle (12). Inaddition to the lymph node examina-tion specified in the guidelines, thecarcass lymph nodes of each animalwere incised and examined. The fol-lowing tissues were collected fromevery elk in the study: tonsils and thesubmandibular, parotid, and retropha-ryngeal lymph nodes (Group 1); lungand the bronchial and mediastinallymph nodes (Group 2); the mesen-teric, ileocecal, and hepatic lymphnodes (Group 3); and the inguinal,iliac, prefemoral, and prescapularlymph nodes (Group 4). If multiplelesions were found, representativesamples of each were collected andpooled (Group 5). Each tissue andlesion was divided into 2 parts: 1 wasplaced in 10% buffered formalin forhistopathology and the other frozen at-70°C for subsequent mycobacterialculture. All tissues were excised afterthe hide was removed and, betweenanimals, table surfaces were disin-fected with 1-Stroke Environ (Sanofi,Victoriaville) and/or with hot waterrinse (90°C), and knives wereimmersed in thermal sterilizers at85-900C.When the National Veterinary

Services Laboratories, Ames, Iowa(NVSL) joined the study, cultures of104 elk were in progress at the Ani-mal Diseases Research Institute,Nepean, Ontario (ADRI). All the remai-ning tissues from one farm (29 elk)were sent to NVSL and specimensfrom 300 elk, from 4 premises, wererandomly allocated to the 2 laborato-ries. Similarly, ADRI had completed

the histopathology on 83 animals, andthe remaining specimens (406) wererandomly divided between ADRI andNVSL. In total, NVSL processed tis-sues of 179 elk for culture and 203 elkfor histopathology. ADRI processedthe tissues of 254 elk for culture and286 for histopathology. Specimens forhistopathology and culture, from anygiven animal, did not necessarily go tothe same laboratory.

GROSS PATHOLOGY

Elk with lesions varying from firmor hard white, grey, or yellow nodulewith a yellow, caseous, necrotic cen-tre that was dry and solid to thin-walled suppurative abscesses wereclassified as post-mortem positive (12).Animals without these lesions wereclassified as post-mortem negative.

HISTOPATHOLOGY

The pathologists trimmed all sub-mitted tissues, and therefore, wereaware of the gross pathological statusof the animal, but were not aware ofother diagnostic test results. Micro-scopic criteria for diagnoses weredefined and agreed upon by both labo-ratories. Each piece of submitted tis-sue from every elk was transectedevery 5 mm and examined carefullyfor gross lesions. A section from eachpiece (with or without gross lesions)was embedded in paraffin, sectionedat 5 ,um, and stained with haema-toxylin and eosin (H&E) using con-ventional methods (13). Mineralizedtissues were treated with Decalcifer II(Surgipath Medical Industries, Rich-mond, Illinois) or 5% trichloroaceticacid prior to the processing. Sectionswith lesions were stained by a newfuchsin-methylene blue (NF) (14), amodified Ziehl-Neelsen stain (13),and/or an acridine orange phenyl-auramine 0 (AOAO) technique (15).These sections were examined for thepresence of acid-fast bacilli or AOAOfluorescent-stained bacilli. If thelesions were not positive for myco-bacteriosis, as defined below, addi-tional sections were stained withGiemsa, Gram, Glynn, or Grocott'smethenamine silver, as required, toexclude other bacterial, fungal orparasitic agents (13).

If caseous necrosis, mineralization,epithelioid cells, multinucleated giantcells, macrophages, and acid-fast orfluorescent bacilli were present, the

109

lesion was considered positive formycobacteriosis (16,17). Lesions con-sistent with mycobacterial infection inCervidae, but without demonstrableorganisms, were classified as myco-bacteriosis suggestive. All otherlesions were classified morphologi-cally, or etiologically when an agentcould be demonstrated using the histo-chemical techniques mentioned above.Changes in lymph nodes such as fol-licular or paracortical hyperplasia orregression, and nonspecific alterationsassociated with drainage from areas ofinflammation were also noted.Two schemes were used to classify

the animals as positive or negative forthe statistical analysis. In the first(histopathology I), animals were testpositive if lesions compatible withtuberculosis (mycobacteriosis ormycobacteriosis suggestive) were seen.In the second scheme (histopathologyII), animals were test positive if theyhad diagnoses of mycobacteriosis,mycobacteriosis suggestive, or one of:granulomatous lymphadenitis, lym-phadenitis with mineralization, focalgranulomatous tonsillitis, or tonsillitiswith mineralization. These diagnosesinclude lesions that could have beencaused either by mycobacteria or byother agents such as Actinobacillussp., fungi or parasites.

MYCOBACTERIAL CULTURE

At ADRI, all tissue from each elkwere thinly sliced and examined forlesions. A pool of tissue with lesionsor without lesions, (3-10 g) washomogenized in 5 mL 0.67 M phos-phate buffer (pH 7.0) using Ten Broektissue grinders. Ten mL of 2% NaOHsolution with 0.2% phenol red wasadded to 4 mL of the resulting tissuehomogenate and neutralized with 2 NHCl after 12 min at room temperature.Sterile distilled water was added tobring the total sample volume to35 mL. The sample was then cen-trifuged at 2200 X g for 30 min, thesupernatant decanted, and the pelletinoculated onto the isolation mediausing sterile cotton swabs.At NVSL, the groups of tissues

(4-5 bags per elk) were separatelytrimmed of excess fat and homoge-nized for two min in a 480 mL blenderjar containing 50 mL of tryptose brothwith 0.04% phenol red. Using a widebore pipette, 10.0 mL of the blendedtissue suspension was transferred to a

TABLE I. Comparison of gross pathology and histopathology I to mycobacterial culture

Culture status

Post-mortema/ Interpretationc M. bovis M. bovishistopathology lb In parallel In series Isolated Not isolated Total

Negative/Negative - 5 292 297Negative/Positive + - 0 24 24Positive/Negative + - 4 23 27Positive/Positive + + 64 16 80

Total 73 355 428a Positive: gross lesions compatible with tuberculosisbPositive: histopathological lesions of tuberculosis consistent with or suggestive of mycobacterialinfection

c+ = positive, - = negative

TABLE II. Distribution of lesions at post-mortem

Number (%) of animals found at post-mortem to have gross lesionscompatible with tuberculosis in the tissue or lymph node listedTotal M. bovis not isolated M. bovis isolated

Tissue (430 animals) (357 animals) (73 animals)Lung 48/430 (11) 12/357 (3) 36/73 (49)Retropharyngeal 36 (8) 7 (2) 29 (40)Mesenteric 35 (8) 12 (3) 23 (32)Mediastinal 16 (4) 7 (2) 9 (12)Bronchial 8 (2) 3 (1) 5 (7)Parotid 4 (1) 1 (0) 3 (4)Tonsil 14(3) 11 (3) 3 (4)Prescapular 4 (1) 2 (1) 2 (3)Ileocecal 3 (1) 2 (1) 1(1)Prefemoral 1 (<1) 0 (0) 1 (1)Submandibular 1 (<1) 1 (<1) 0 (0)Inguinal 0 (0) 0 (0) 0 (0)Iliac 0 (0) 0 (0) 0 (0)

conical centrifuge tube. An equal vol-ume of a 1:1 (VIV) solution of 0.3%Zephiran (Winthrop, New York, NewYork) and 1:500 sodium hypochloritewas added, and the mixture was cen-trifuged for 30 min at 2640 X g. Thesupernatant was discarded, and 2 mLof 10% egg yolk solution was added.The sediment was resuspended andinoculated onto the isolation media.The isolation media used by both

laboratories included: LowensteinJensen medium with and withoutglycerol (18), Herrold egg yolkmedium with mycobactin J (18),Herrold egg yolk medium withoutglycerol, mycobactin J, and malachitegreen (18), Stonebrink medium (18),and Modified Middlebrook 7H 1Imedium (19). In addition NVSLused BACTEC 12B medium (BectonDickinson, Towson, Maryland) andADRI used Middlebrook's 7H9broth + OADC (Difco Laboratories,Detroit, Missouri) with and without4.1 mg/L Na pyruvate. The mediawere incubated at 37°C for 12 wk andexamined every 2 wk for colony for-mation. All acid-fast isolates were

classified using standard growth andbiochemical characteristics to deter-mine mycobacterial species (18,20).

STATISTICAL ANALYSES

Sixty-nine elk were excluded fromthe final statistical analysis since theirtissues were either delayed in transitor not clearly identified, thus wereunfit for mycobacterial culture. Therates of M. bovis isolation from theanimals divided between the two lab-oratories were compared using aFisher's exact 2-sided test (21). Forgross pathology and the twohistopathology interpretations, 4-foldtables were generated comparing theirdiagnostic test results with those ofmycobacterial culture. Elk were clas-sified as disease positive if M. boviswas isolated, and as disease negativeif mycobacteria other than M. boviswere isolated or if no mycobacteriawere isolated. Using the 4-fold tables,the sensitivity, specificity, positiveand negative predictive values andkappa statistic were calculated(21,22). The 95% confidence limitswere calculated for the estimates of

110

TABLE III. Distribution of diagnoses by group of tissues collected at post-mortem(One animal can have more than one diagnosis for each group)

DiagnosisMycobacteriosisAbscess(G+Cocci/Nocardia)Actinobacillosis/ActinomycosisAmyloid like depositBALT6- ProminentEmphysemaFat necrosisGranuloma eosinophilicHepatitisHepatitis abscessHepatitis parasiticHyperplasia follicularHyperplasia paracorticalLipidosis focalLymphadenitis granulomatousLymphadenitis mineralizationLymphadenitis mycoticLymphadenitis parasiticLymphadenitis suppurativeLymphangiectasiaMastitisPleural fibrosisPleuritisPneumoniaSialoadenitisTonsillitis inflamatoryTonsillitis focal granulomatousTonsillitis mineralizationTonsillitis noninflamatoryNo significant findings

Group 1'

371

541

000000

3440

20709

42000002

96912

24203

Group 2234001

12100000

1100

11301011

011

23700000

352

Group 3318010000800713109

1 11

3251

00000000

388

Group 44

30036002000010

04

08

160000000000

418

Group 55143

010000012

121420

330340

1198

01310005

17

Tonsils, submandibular, parotid and retropharyngeal lymph nodes2 Lung, bronchial and mediastinal lymph nodesIMesenteric, ileocecal and hepatic lymph nodes4Inguinal, prefemoral, and prescapular lymph nodesIGross lesions compatible with mycobacterial infection6 Bronchiolar associated lymphoid tissueIIncludes 22 interstitial, 1 mycotic8 Includes 5 interstitial, 2 abscess, 1 parasitic, I mycotic9Includes 43 crypt abscess, 1 eosinophilic, 2 foreign body, 2 mycotic, 48 non-specific inflamatory

lesions'° Includes 9 crypt abscess, 4 non-specific inflamatory lesions

sensitivity, specificity, and positiveand negative predictive values using aformula based on a normal approxima-tion to the binomial distribution (21).Gross pathology plus histopathology Ior II was also evaluated in parallel (ananimal had to be positive on at leastone of the two tests to be classified ascombination positive) and in series (ananimal had to be positive on both teststo be classified as combination posi-tive). The total number of animalsused in each analysis varied slightly asanimals were excluded from the calcu-lations if valid data were missing forone or more tests.

RESULTS

In 356 (83%) of the 428 animals,for which there were complete grosspathology, histopathology, and myco-

bacterial culture results, the 3 testsagreed on the tuberculosis status of ananimal: 64 were disease positive and292 were disease negative by all3 tests (Table I). Gross pathology didnot agree with histopathology andculture results in 5.4% (23) of the ani-mals studied, and histopathology dis-agreed with the post-mortem and cul-ture results in 5.6% (24) of cases.Post-mortem and histopathologyfindings did not agree with the myco-bacterial culture results in 4.9%(16+5/428) of the animals.Complete post-mortem and myco-

bacteriology data were obtained for430 animals. Of these, 107 had grosslesions compatible with tuberculosis.The distribution of gross lesions con-sistent with tuberculosis in M. bovispositive and M. bovis negative elk islisted in Table II; the lung, theretropharyngeal lymph node, and the

mesenteric lymph node were the mostcommonly affected tissues in culturepositive animals.Complete histopathological and

mycobacterial culture results wereobtained for 429 animals. Table IIIlists the histopathological diagnosesmade from each group of tissues.Note that Group 5 contains all tissuesfound to have lesions consistent withtuberculosis at the post-mortem exam-ination. Fluke-induced lesions andchronic bacterial (non-mycobacterial)lesions were present in some elk.Lesions containing small, corporaamylacea-like, mineralized foci, afeature frequently seen in tuberculosisin cattle, were observed in somelymph nodes and tonsils. Other miner-alized lesions in these organs weresuggestive of nonspecific, chronicinflammatory foci.There was no difference between

ADRI and NVSL rates of identifica-tion of M. bovis positive animals inthe randomly divided group (P = 0.7for a 2-sided Fisher's exact test). Onehundred and twenty-three mycobac-terial isolates were recovered byculture: 73 (59%) M. bovis, 11 (9%)M. paratuberculosis, 11 (9%)M. avium, and 28 (23%) rapidly grow-ing species or M. terrae complex(Table IV). Mixed mycobacterialinfections were identified in twocases; one M. avium plus an undeter-mined rapid grower and one M. aviumplus M. bovis.The 95% confidence limits for the

sensitivity estimates of all testsand test combinations overlapped(Table V). The sensitivity figures forgross pathology and histopathology Iwere 93% (CL 84,97%) and 88%(CL 77,94%). The highest sensitivitywas obtained by interpreting tests inparallel (93-95% [CL 84,98%]). Thehighest specificity estimates weregenerated when the tests wereinterpreted in series (94-95%[CL 91-97%]). The 95% confidencelimits of specificity estimates did notoverlap those of the other tests andtest combinations (Table V).At a 17% prevalence of tuberculo-

sis infection, based on the isolation ofM. bovis, the combination of post-mortem and histopathology I in seriesgave the highest positive predictivevalue (80%) (Table VI). All the testsand test combinations had negativepredictive values of 97-98%.

111

There was moderate to excellent(44-80%) agreement (kappa) betweenthe tests evaluated and the isolation ofM. bovis (Table VII). The highestagreement beyond chance withM. bovis isolation was obtained withthe series interpretation of post-mortem and histopathology 1 (80%).

DISCUSSION

Studies using post-mortem exami-nation to determine disease status orto select tissues for further analysismay miss M. bovis infection in ani-mals without lesions. As reflected bythe high negative predictive value(NPV) (98% Table VI), there werefew culture positive, lesion negativeanimals in this sample (5/323 = 1.5%,Table VII). Of animals from which M.bovis was isolated, 7% (5/73) had nogross lesions of tuberculosis at abat-toir inspection (Table I). Other studieshave found similar rates (3.7-6%) ofno gross lesion/culture positive animalsin heavily infected herds (23,24).

If post-mortem findings had beenused as the reference standard for thisanalysis, more animals would be clas-sified as positive (107/430 = 25%)than by culture (17%). Of the 107 ani-mals positive for tuberculosis on post-mortem examination, M. bovis wasisolated from 68 (positive predictivevalue (PPV) = 64%, Table VI).Twenty-three of the 39 culture nega-tive animals were also negative byhistopathology (Table I). The mostcommon histopathology diagnoses forthese animals included: actinobacillo-sis, parasitic hepatitis or lymphadeni-tis, tonsillar cysts, and pneumonia.

In the Canadian abattoir surveil-lance program for bovine tuberculo-sis, tissues with gross lesions typicalof tuberculosis are examined byhistopathology, and those diagnosedas mycobacteriosis, or possibly beingcaused by mycobacteriosis, are cul-tured. The histopathology II interpre-tation included lesions, that couldhave been caused by mycobacteria,such as granulomatous lymphadenitis,lymphadenitis with mineralization,and granulomatous tonsillitis, in thepositive definition. The sensitivityincreased minimally using thisscheme, but the specificity decreasedby 12% (Table V) and the PPVdecreased by 18% (Table IV).

TABLE IV. Mycobacteria isolated compared with the histopathology I diagnosis

Histopathology diagnosisMycobacteriosis

or Other than No significantIsolate suggestive of Mycobacteriosis findings Totala

M. bovis 64 8 1 73M. avium 1 9 1 11M. paratuberculosis 4 5 2 11M. terrae complex 4 7 4 15Rapid growerb 2 10 1 13No isolate 29 211 68 308

Total 104 250 77 431

aAn animal may have more than one isolatebRapidly growing Mycobacterium spp. other than M. fortuitum or M. chelonae

TABLE V. Sensitivity and specificity estimates of post-mortem, histopathology I, andhistopathology II compared to the isolation of M. boviss

Number 95% 95%of Sensitivity Confidence Specificity Confidence

Diagnostic test animals estimate (%) limits (%) estimate (%) limits (%)Post-mortemb 430 93 84, 97 89 85, 92Histopathology Ic 429 88 77, 94 89 85, 92Histopathology IId 429 89 79, 95 77 72, 81Parallel interpretationePMb/HistoIc 428 93 84, 97 82 78, 86PM/Histo Ild 428 95 86, 98 72 67, 77Series interpretationfPMb/Histo Ic 428 88 77, 94 95 93, 97PM/Histo IId 428 88 77, 94 94 91, 96a Reference standard; positive: M. bovis isolated; negative: M. bovis not isolatedEnhanced abattoir inspection; positive: gross lesions compatible with tuberculosis seen at post-mortem inspection

c Positive: histopathological lesions consistent with, or suggestive of, mycobacterial infectiondPositive: histopathological lesions consistent with or suggestive of mycobacterial infection,granulomatous lymphadenitis, lymphadenitis with mineralization, focal granulomatous tonsillitis,or tonsillitis with mineralization

e Combined screening tests, with test results interpreted in parallel (i.e. an animal is consideredpositive" if found positive by at least one test)fCombined screening tests, with test results interpreted in series (i.e. an animal is consideredpositive" only if "positive" on both tests)

The positive predictive value mea-sures the likelihood of an animal,found positive by post-mortem and/orhistopathology, being M. bovis posi-tive at culture. The likelihood that ananimal found negative on post-mortem examination and/or histo-pathology, being M. bovis negative atculture, is the negative predictivevalue. As the PPV and NPV of a testvary with changes in the diseaseprevalence, it is not appropriate to usepredictive values for "best" test selec-tion. The predictive values were cal-culated here as another method ofexamining the effect of various inter-pretations on test performance in thefield.

Tests are interpreted in parallelwith the general aim of increasingsensitivity. But, the specificity andthe PPV of such a combination

decreases. Post-mortem/histopathol-ogy I, interpreted in parallel, had1.6 times the number (63) of false posi-tives than either test alone (39 & 40;Table VII). This indicates that post-mortem and histopathology did notidentify the same animals as positive(otherwise the number of false posi-tives with respect to culture wouldremain the same).Agreement between biologically

independent tests can provide a rea-sonable summary of overall test per-formance at determining infection(25). As opposed to sensitivity andspecificity, kappa measures the agree-ment in the positive as well as thenegative categories. It has been sug-gested that a kappa of 40-75% indi-cates a fair to good level of agreementbetween tests and over 75% indicatesexcellent agreement between tests

112

TABLE VI. Positive and negative predictive values of post-mortem, histopathology I, andhistopathology II at a prevalence of 17%, based on M. bovis isolationa

Positive 95% Negative 95%Number of predictive Confidence predictive Confidence

Diagnostic test elk value (%) limits value (%) limits

Post-mortem 430 64 54, 72 98 96, 99Histopathology I 429 62 51, 71 97 95, 99Histopathology II 429 44 36, 53 97 94, 99Parallel interpretationPM/Histopathology I 428 52 43, 61 98 96, 99PM/Histopathology II 428 41 34,49 98 96, 100Series interpretationPM/Histopathology I 428 80 69, 88 97 95, 99PM/Histopathology II 428 74 64, 83 97 95, 99a Definitions and abbreviations as in Table V

TABLE VII. Kappa statistic of post-mortem, histopathology I, and histopathology II againstmycobacterial culturea

Test #1/Test #2bTest #1 +1+ +- -1+ -I- Kappa%c

Post-mortem 68 39 5 318 69Histopathology I 64 40 9 316 65Histopathology II 65 82 8 274 47Parallel interpretationPM/Histopathology I 68 63 5 292 57PM/Histopathology II 69 98 4 257 44Series interpretationPM/Histopathology I 64 16 9 339 80PM/Histopathology II 64 22 9 333 76a Definitions and abbreviations as in Table VTest #1 as listed; Test #2 = mycobacterial culture, where + = M. bovis isolated, - = M. bovis notisolated

c Kappa statistic (percent) of agreement beyond chance beteen test #1 and test #2

(21). The presence of gross lesionstypical of mycobacteriosis at post-mortem inspection showed goodagreement beyond chance with theisolation of M. bovis (kappa = 69%;Table VII). Reasons for lack of ahigher kappa may include: lesionscaused by agents other than M. bovis(e.g. actinobacillosis), lesions causedby mycobacteria other than M. bovis,early M. bovis infections wherelesions are not yet grossly visible,resolving lesions of tuberculosiswhere M. bovis is no longer viable, orfailure of mycobacterial culture toisolate the organism.

Histopathology I had a moderateagreement beyond chance with cul-ture (kappa = 65%; Table VII) and amoderate PPV (62%; Table VI).Infections with mycobacteria otherthan M. bovis and the fact that the twoprocedures cannot physically test thesame portion of tissues may accountfor the lack of stronger agreement.Also, 29 of 104 tissues classified aspositive by histopathology did notyield a mycobacterial isolate(Table IV). This may have been due

to false positive histopathologyresults and/or false negatives by thereference standard (culture). Whenfalse negatives occur with a referencestandard, any test evaluated againstthat standard may have an overesti-mated sensitivity and an underesti-mated specificity.The impact of missing true posi-

tives by the reference standard wouldnot have had a large impact on thesensitivity and specificity figures inthis trial. If, for example, there was a15% increase in the number of culturepositive animals (that is, 10 histo-pathology positive/culture negativewere reclassified as histopathologypositive/culture positive and 1 histo-pathology negative/culture negativeanimal was reclassified as histo-pathology negative/culture positive),the sensitivity of histopathology Iwould remain at 88% and the speci-ficity would increase from 89% to91%.

Interpreting the tests in series,where only tissues with gross lesionswere considered for histopathologyand only those which were histo-

pathology positive were compared toculture, detected 87.7% (64/73) ofthe M. bovis infected animals in thissample (Table I). The kappa statisticand the PPV, for this interpretation,were the highest of all analyses. Aseach animal was required to meet twopositive criteria, only the most likelyinfected animals were considered pos-itive for comparison to culture; there-fore, the false positive rate (thegreater source of error in this popula-tion) was minimized (Table VII).The need for mycobacterial isola-

tion and identification, or more spe-cific histopathology techniques, incases where a definitive diagnosis isrequired is illustrated in Table IV.Fourteen percent of histopathologypositive elk yielded mycobacteriaother than M. bovis (Table IV).Norton et al (26) report the effect ofthe sample collection techniques onthe incidence of environmentalmycobacterial contamination; in theirstudy, 63 isolates of mycobacteriaother than M. bovis were obtainedfrom tissues collected by routine fieldnecropsy, as opposed to none fromtissues collected using aseptic tech-nique. The abattoir post-mortem tech-nique we used should have minimizedenvironmental contamination of thetissues. The M. terrae and rapidlygrowing mycobacteria we isolatedwere probably present within the cul-tured tissues. Note that 61% (17/28)of saprophytic isolates were from thetissues of animals having histopathol-ogy diagnoses other than mycobacte-riosis or mycobacteriosis suggestive,whereas only 18% (5/28) were iso-lated from the tissues of animals hav-ing no significant findings at histo-pathology (Table IV).The prevalence of M. paratubercu-

losis infection in this population mayhave been higher than indicated bythe 11 isolates recovered. The sam-pling protocol, decontamination meth-ods, and incubation time used weredesigned for M. bovis isolation andnot for the recovery of M. paratuber-culosis. Histopathology changes dueto M. paratuberculosis infection indeer may be very similar to thosecaused by M. bovis infection (27). Aportion of the histopathology posi-tive/culture negative cases may havebeen M. paratuberculosis infectionmissed by the culture protocol used.Mycobacterium paratuberculosis

113

infection in farmed Canadian elk willrequire further study before its preva-lence and implications are understood.Our study supports the concept that

gross pathology, histopathology, andculture do not consistently identifythe same animals as tuberculosis posi-tive. It follows then, that a parallelinterpretation of all three methodstogether would maximize the detec-tion of M. bovis infected animals (rec-ognizing that specificity woulddecrease). Our results also show thatusing gross pathology and histo-pathology alone, or in combination,resulted in only small changes in sen-sitivity. Specificity, however, canvary greatly depending on the inter-pretation or combination of the two.As animal populations, production

techniques, tuberculosis prevalence,diagnostic resources, and controlefforts are unique to each country oroutbreak, it is not in the scope of thisreport to discuss the application ofour findings to tuberculosis surveil-lance or eradication schemes. Thedecision to choose gross pathology,histopathology, mycobacterial cul-ture, or combinations thereof, as a ref-erence standard for test evaluationwill depend on the needs of the pro-gram in which the test(s) will be used.

ACKNOWLEDGMENTS

The authors acknowledge the coop-eration and tremendous effort of theAgriculture and Agri-Food Canadapersonnel in Alberta, Saskatchewan,and Ontario in collecting the requiredtissues. The work by staff of theNational Veterinary Services Labora-tories and the Animal DiseasesResearch Institute, in receiving andprocessing the specimens, is greatlyappreciated. We thank the producersof cervids who allowed the use oftheir animals and facilities.

REFERENCES

1. LESSLIE IW, HEBERT CN, BURN KJ,MacCLANCY BN, DONNELLLY WJC.Comparison of the specificity of human

114

and bovine tuberculin PPD for testing cat-tle. Vet Rec 1975; 96: 332-334.

2. FRANCIS J, SEILER RJ, WILKIE IW,O'BOYLE D, LUMSDEN MJ, FROSTAJ. The sensitivity and specificity of vari-ous tuberculin tests using bovine PPD andother tuberculins. Vet Rec 1978; 103:420-425.

3. BUCHAN GS, GRIFFIN JFT. Tubercu-losis in domesticated deer (Cervuselaphus): a large animal model for humantuberculosis. J Comp Path 1990; 103:11-22.

4. WOOD PR, CORNER LA, ROTHELJS, BALDOCK C, JONES SL, COUSINSDB, McCORMICK BS, FRANCIS BR,CREEPER J, TWEDDLE NE. Fieldcomparison of the interferon-gamma assayand the intradermal tuberculin test for thediagnosis of bovine tuberculosis. Aust VetJ 1991; 68: 286-290.

5. STUART FA, MANSER PA, McIN-TOSH FG. Tuberculosis in imported reddeer (Cervus elaphus). Vet Rec 1988; 122:508-511.

6. de JONG H, EKDAHL MO. Evaluationof a number of ancillary tuberculin tests incattle. NZ Vet J 1969; 17: 215-226.

7. LIVINGSTONE PG. The evaluation oftuberculin tests in a tuberculous farmed,red deer Cervus elaphus herd in NewZealand (Master of Preventative Veteri-nary Medicine thesis). Davis: Universityof California, 1980.

8. ROSWURM JD, KONYHA LD. Thecomparative-cervical tuberculin test as anaid to diagnosing bovine tuberculosis. ProcUS Anim Health Assoc 1973; 77:368-389.

9. PLACKETT P, RIPPER J, CORNERLA, SMALL K, de WITTE K,MELVILLE L, HIDES S, WOOD PR.An ELISA for the detection of anergictuberculous cattle. Aust Vet J 1989; 66:15-19.

10. BULMER WS. Canada's captive wildungulate program. Proc US Anim HealthAssoc 1989; 93: 595-598.

11. HUTCHINGS DL, WILSON SH. Evalua-tion of lymphocyte stimulation tests for thediagnosis of bovine tuberculosis in elk(Cervus elaphus). Am J Vet Res 1995; 56:27-33.

12. AGRICULTURE CANADA. MeatHygiene Manual. Ottawa: Canadian Gov-ernment Publishing Centre, CatalogueNo. A62-38/1983-1, Supply and ServicesCanada, 1983.

13. LUNA LG, ed. Manual of histologic stain-ing methods of the Armed Forces Instituteof Pathology. 3rd ed. New York: McGraw-Hill, 1968.

14. HENRY EJ, CASSIDY DR. Rapid decal-cification followed by a new fuchsin-methylene blue stain for acid-fast bacilli infrozen sections. Stain Technol 1966; 41:345-346.

15. MOTE RF, MUHM RI, GIGSTAD DC.A staining method using acridine orangeand auramine for fungi and mycobacteriain bovine tissue. Stain Technol 1974; 50:5-9.

16. DUNGWORTH DL. The RespiratorySystem: Tuberculosis. In: Jubb KVF,Kennedy PC, Palmer N, eds. Pathology ofDomestic Animals. 3rd ed. New York:Academic Press, 1985: 493-504.

17. JONES TC, HUNT R, eds. Tuberculosis.In: Veterinary Pathology. 5th ed. Philadel-phia: Lea and Febiger, 1983: 648-661.

18. PAYEUR JB, JARNAGIN JL,MARQUARDT JG, SCHAPER LA,MARTIN BM. Laboratory Methods inVeterinary Mycobacteriology for the Isola-tion and Identification of Mycobacteria.Ames: National Veterinary Services Labo-ratories, United States Department ofAgriculture, 1993.

19. CORNER LA. Australian Standard Diag-nostic Techniques for Animal Diseases;No. 52, Bovine Tuberculosis. East Mel-bourne: CSIRO Australia, 1989.

20. KENT PT, KUBICA GP. Public HealthMycobacteriology: A Guide for the LevelIII Laboratory. Atlanta: US Department ofHealth and Human Services, 1985.

21. FLEISS JL. Statistical methods for ratesand proportions, 2nd ed. In: Wiley Seriesin Probability and Mathematical Statistics.Toronto: John Wiley & Sons, 1981.

22. MARTIN SW, MEEK AH, WILLEBERGP. Veterinary Epidemiology, Principlesand Methods. Ames, Iowa: Iowa State Uni-versity Press, 1987.

23. de LISLE GW, HAVILL PF. Mycobacte-ria isolated from deer in New Zealandfrom 1970-1983. NZ Vet J 1985; 33:138-140.

24. MILLER MW, WILLIAMS JM,SCHIEFER TJ, SEIDEL JW. Bovinetuberculosis in a captive elk herd inColorado: epizootiology, diagnosis, andmanagement. Proc US Anim Health Assoc1991; 95: 533-542.

25. McNAB WB, MEEK AH, DUNCAN JR,BROOKS BW, VAN DREUMEL AA,MARTIN SW, NIELSEN KH, SUGDENEA, TURCOTTE C. An evaluation ofselected screening tests for bovine paratu-berculosis. Can J Vet Res 1991; 55:252-259.

26. NORTON JH, DUFFIELD BJ,COWARD AJ, HIELSCHER RW,NICHOLLS RF. A necropsy techniquefor cattle to eliminate contamination oflymph nodes by mycobacteria. Aust Vet J1984; 61: 75-76.

27. de LISLE GW, YATES GF, COLLINSDM. Paratuberculosis in farmed deer: casereports and DNA characterization of iso-lates of M. paratuberculosis. J Vet DiagnInvest 1993; 5: 567-571.