Evaluation of pulmonary tuberculosis diagnostic tests in children and adolescents at a pediatric reference center

(1)

www.journalpulmonology.org

ORIGINAL ARTICLE

Evaluation of pulmonary tuberculosis diagnostic tests in children and adolescents at a pediatric reference center

Andrea M.O. Rossoni

^a

, Kathryn L. Lovero

^b

, Tonny T. Tahan

^a

, Antônio R. Netto

^c

, Marssoni D. Rossoni

^d

, Isabela N. Almeida

^e^,∗

, Elisangela A.S. Lizzi

^f

, Afranio Kritski

^g

, Cristina O. Rodrigues

^a

aUniversidadeFederaldoParaná(UFPR)theGraduatePrograminChildandAdolescentHealth,Curitiba,PR,Brasil

bUniversityofCaliforniaBerkeley,SchoolofPublicHealth,DivisionofInfectiousDiseasesandVaccinologyandDivisionof Epidemiology,Berkeley,CA,UnitedStates

cUniversidadedeSãoPaulo(USP),EscoladeMedicinadeRibeirãoPreto,DepartamentodeMedicinaSocial,RibeirãoPreto,SP, Brazil

dUniversidadeFederaldoParaná(UFPR),DepartamentodeMedicina,HospitaldasClínicas,Curitiba,PR,Brasil

eUniversidadeFederaldeMinasGerais(UFMG),FaculdadedeMedicina,LaboratóriodePesquisaemMicobactérias,Belo Horizonte,MG,Brazil

fUniversidadeTecnológicaFederaldoParaná(UTFPR),DepartamentodeMatemática,Curitiba,PR,Brazil

gUniversidadeFederaldoRiodeJaneiro(UFRJ),FaculdadedeMedicina,ProgramaAcadêmicodeTuberculose,RiodeJaneiro, RJ,Brazil

Received18September2019;accepted12January2020 Availableonline31January2020

KEYWORDS Tuberculosis;

Pediatrichospital;

Diagnosis;

Clinicalstudy

Abstract

Introduction:Thisstudyevaluatestheperformanceofindividualandcombinationstestsused forpediatrictuberculosisdiagnosisatareferencecenter.

MaterialsandMethods:Diagnostic test outcomes from children with presumed pulmonary tuberculosisevaluatedfromJanuary2005-July2010werecomparedtoastandarddiagnosis madebyanexpertpanelofphysicians.

Results:Presenceofatleastonesign/symptom,historyofcontact,orabnormalchestX-ray (aCXR)individuallyshowedthehighestsensitivity(85.7%).Whilethecombinationofhistoryof contact,atleastonesign/symptom,positivetuberculinskintest,andaCXRhadlowsensitivityof 20%,butthespeciﬁcityandapositivepredictivevaluewere100%,respectively.Thecombination oftestsusedintheInternationalUnionAgainstTuberculosisandLungDiseaseandtheBrazilian MinistryofHealthsystemsshowedsensitivityof28.6%and71.4%andspeciﬁcityof95.8%and 97.0%,respectively.

∗Correspondingauthorat:FaculdadedeMedicina---LaboratóriodePesquisaemMicobactérias,UniversidadeFederaldeMinasGerais, BeloHorizonte,MinasGerais,AvenidaAlfredoBalena,190,Brazil.

E-mailaddress:[email protected](I.N.Almeida).

https://doi.org/10.1016/j.pulmoe.2020.01.001

(2)

Conclusions: Intheabsenceofagoldstandard,thecombinationofclinicalhistory,tuberculin skintest,andaCXR,aswellastheBrazilianscoringsystemserveassimple,low-costapproach thatcanbeusedforpediatricTBdiagnosisbyﬁrst-contactcareproviders.

Introduction

Tuberculosis (TB) remains a major global public health problem and according to the World Health Organization (WHO) in 2017, 10.0 million people developed TB disease,1.0millionofwhichwerechildren(aged<15years).

In cases of deaths caused by TB, (aged <15 years) 15%

of total deaths and 10% of total deaths in HIV positive cases these values are higher than their share of esti- mated cases, suggesting poorer access to diagnosis and treatment.¹

Many clinical and laboratory tests have been pro- posed for the diagnosis of pediatric TB, but a gold standard is lacking.² Currently, pediatric TB diagnosis is based on history of contact, clinical signs, chest radiog- raphy, tuberculin skin testing (TST), and microbiological examination. However, children with TB can present clinical signs and abnormalities on chest x-rays (aCXR) that are nonspecific. Moreover,respiratory specimens are difficult tocollect and bacteriologicyield is low in pediatric patients, greatly reducing rates of bacteriological confirmation.^3,4

Thereisagreatneedtoidentifydiagnosticteststhatare moresensitive andspeciﬁcforthe diagnosisofpulmonary TBinpediatricpatients.Anumberofscoringsystemshave beenproposedforthediagnosisofpediatricTB,thoughno singlesystemhasbeenadequatelyvalidated.^5---8TheBrazil- ianMinistryofHealth(BMoH)systemhasbeenevaluatedfor its use in HIV-infectedand uninfected children.⁹ In 1998, theInternational UnionAgainstTuberculosis andLungDis- ease(IUATLD)proposedasystem¹⁰basedondifferentscores accordingtothelocalTBepidemiology,whichhasyettobe validatedinothersettings.⁷

Inthiscontexttheaimofthisstudywastoevaluatethe performance of singleand combinationstests usedin the diagnosisofpediatricTB,aswellastwoscoringsystemsin aReferenceCenterinBrazil.

Materials and methods

Studypopulation

Thestudymodelwasadescriptive,surveyofacohortina lowHIVprevalencesetting.Studycohortincludedchildren of14 yearsofageandunder,evaluatedfor pulmonaryTB betweenJanuary2005andJuly2010attheClinicalHospital oftheFederalUniversityofParaná,thereferencecenterfor pediatrictuberculosisinCuritiba,Brazil.

Datacollect

Data on epidemiological, clinical, laboratory, radiological and treatment outcomes were extracted from medical recordsusingastandardizedquestionnaire.Medicalrecords forallchildrenfulﬁllinginclusioncriteriainthistimeperiod wereabletoberecovered.However,patientswereexcluded iftherecorddidn’tcontainedkeydataforexpertpaneldiag- nosis,iftheyweretransferredtoanotherservice orwere losttofollow-up beforetheattendingphysicianmadethe diagnosis.

Evaluationofdiagnostictests

Fordiagnostic test evaluation, a presumedpulmonary TB wasconsideredifthesubjectspresentedoneofmoreofthe followingsignsorsymptoms:coughfortwoweeksormore, fever,sweating, pneumoniaor wheezingwithnoimprove- ment aftertreatment withantibiotics or bronchodilators, loss of appetite, adynamia, and loss or stabilization of weight.Inchildren2yearsandunderwhohadreceivedBCG vaccination,TSTgreaterthanorequalto10mmwasconsid- eredpositive;inchildrenover2yearsofage,independent ofvaccinationstate,TSTgreaterthanorequalto5mmwas consideredpositive.⁹

Evaluationofscoringsystems

Datafromsubjectmedicalrecordswerealsousedtoapply twocommonscoringsystemsusedfordiagnosisofpediatric tuberculosis,theIUATLD[10]andBMoH[9]systems.Forthe BMoH system, both a cutoff of 30 (including ‘‘possible’’

and ‘‘very likely’’ TB classiﬁcations) and a cutoff of 40 (only‘‘verylikelyTB’’subjects)wereevaluated.AsCuritiba CityisclassiﬁedashavinglowTBprevalence,¹¹theIUATLD scoring system for areasof low prevalence wasemployed here. Additionally, because the IUATLD system does not includefollow-up,weevaluatedsubjectsatinitialappoint- ment(IUATLD I)andat thesubsequent consultationwhen physiciandiagnosiswasmade(IUATLDS),assomesubjects presented change in evaluation parameters after admin- istrationof non-tuberculosis treatments. Forboth IUALTD evaluations,ascoreof9ormorewasconsideredaTBcase.

As awayofcreatingdiagnosticgroups forthis study,a panel of experts (consisting of an infectious disease spe- cialist and a pulmonologist, both specializing in pediatric TB) evaluated subjectdata and diagnosed them as a TB, latentTB,ornoTBcase.Incaseswheretherewasdisagree- 84

(3)

mentinthediagnosisprovidedbythetwo-personpanel,a third expertissued aﬁnal decision.Expertshadaccessto allpatients’data,includingTBtreatmentoutcomethrough astandardizedformandclassiﬁedcasesaccordingtotheir clinicalexperience.

AllanalyseswerecarriedoutinSASv9.2.Thestatistical relationshipofsociodemographiccharacteristicsofthedif- ferentchildren´sgroupssuspectedofpulmonarytuberculosis wereevaluatedbyPearson´schi-squared andPearson´schi- squared withYates correctionand Mann-Whitney.Positive predictivevalue(PPV)andnegativepredictivevalue(NPV) ofdiagnostictestswerecalculatedusingtheprevalenceof pediatricTBatthestudysite(11%),aswellasratesfoundin otherhealthcaresettings(1%and5%).⁹Toassessthedegree ofagreementbetweenexpert diagnoses,aKappastatistic wascalculatedandinterpretedaccordingtothecriteriaof LandisandKoch.¹²

Results

Fromatotalof236childrenwithpresumedpulmonaryTB, 21.2%(50/236)wereexcludedfromthestudycohortforthe followingreasons:24losttofollow-upbeforediagnosis,20 incompletemedicalrecords, and6 inconclusivediagnoses byexpertpanel.

Oftheremaining186children,34(18.3%)wereclassiﬁed asnot TB, 131 (70.4%) aslatent TB and 21(11.3%) asTB cases.Diagnosticagreement betweenexpertsvariedfrom substantialtoalmostperfect(Kappa=0.94,0.75,0.69).

Thesubjectswerethendividedintotwogroupsforanal- ysis: active TB group, includingthe 21 TB cases, and No TB (NTB) group includingin this group the 131 latent TB infectionand34notTBcases.

The Sociodemographic characteristics did not differ betweenthetwogroups andisimportanttohighlight that theproportionofboysintheTBgroupwas52%andintheNTB group50%(p=0.95);whiteskincolorratiowas91.7%inthe TBgroupand79.8%intheNTBgroup(p=0.38);themedian ageinyearofTBgroupwas5.7(0.7---13.9)andinNTBgroup was5.8 (0.4---14.9) (p=0.94)and ﬁnallythe median num- berpeoplelivinginahousewas5(4---13)inTBgroupand5 (3---10)inNTBgroup(p=0.13).

Analysis of epidemiological historyshowed that 85% of theTBand91.6%ofNTBgroupshadhistoryofcontactwith atleastoneindexcase.Inbothgroups,ahouseholdcontact wasmostcommon(70%ofTBgroup,79.3%ofNTBgroup), withthegreatestpercentageofindexcasesbeingparents forbothgroups (27.7%TBgroup,43.5%NTBgroup).Inthe TBgroup,38.9%ofchildrenhadcontactwithmorethanone adultTBcase,whichwassigniﬁcantlyhigher(p=0.02)than the percentageof NTB caseswithmore than onecontact (18.4%).

AsigniﬁcantlyhigherpercentageofTBgroupversusNTB group was positive for all signs and symptomsevaluated, exceptfordrycough(Table1)(p<0.001).

To evaluateother potentialdiagnoses, 46% of subjects thatpresentedat leastonesign orsymptomwereinitially treatedforotherconditions(e.gpneumonia,asthma)prior toﬁnalTBdiagnosis(datanotshown).TBgroupsubjectshad signiﬁcantlylessimprovementafterthisinitialnon-TBtreat- ment(p<0.01).However,itisworthnotingthat35%percent

ofTBgroupsubjectsimprovedclinically,andforthesesub- jectsthereturnofsymptomsorthemaintenanceofaltered radiologicalexamswascriticaltothelaterTBdiagnosis.No signiﬁcantdifferencewasfoundbetweengroupswhenana- lyzingco-morbiditiesthatcouldinterferewiththediagnosis ofTB(28.6%TBand24.4%NTB;p=0.78),andonly2subjects wereHIV-positive,bothintheTBgroup.

All subjects in the TB group had previously received the BCG vaccination, while 95.1% of the NTB group had been vaccinated (p=0.6). The TST was positive in 82.4%

of the TB group and 57.1% of the NTB group (p<0.001), andtheaverage diameterofpositive responseswassigni- ﬁcantlylargerintheTBgroup(TBgroup=19.5±5mm.NTB group=15.4mm±5mm; p<0.001). aCXR was observed in 85.7%of the TB group and7.9% of NTB group(p<0.001).

IntheTB group, smearandculture positivewasobserved in 23.5% (n=4/17) and 26.7% (n=4/15), respectively. In theNTBgroup,allsubjectsevaluatedweresputumsmear (n=0/44)andculturenegative(n=0/25).

WhenanalyzingtheaccuracyofsingletestsforTBdiag- nosis,thepresenceofatleastonesignorsymptom(85.7%), historyof adultcontact(85.7%) andaCXR (85.7%)showed thehighestsensitivity.Evaluatingcombinationsoftests,we found that the BMoH system witha cut-off of 30showed a higher sensitivity (95.2%) than any single or combined test.Whilealone,theaCXRshowedhighaccuracy(91.4%), thecombination ofat leastone sign/symptom,history of contactandaCXRincreasedaccuracy(95.1%).

The sensitivity, speciﬁcity, and accuracy for individual and combinations of tests are described in Table 2. The sensitivityandaccuracyprovideusefulinformationtocom- pareperformanceofdiagnostictests,butnotthepositive predictivevalue(PPV)ornegativepredictivevalue(NPV).¹³ Therefore, we assessed these accuracy tests usingthe prevalenceofourstudycohortaswellassimulatingpreva- lence rates found in other healthcare settings (Table 3 and Table 4). Overall, single-test PPV was low. However, looking at test combinations, presence of at least one sign/symptom,historyof contact,positive tuberculinskin testandaCXRhadaPPVof100%.BothBMoHandIUTLDsys- temsshowedhigherNPVvalues,butlowerPPVvalues,than thiscombinationinallprevalencescenarios.

Discussion

Inthisstudy,anexpertpanelofphysicianswasemployedto directlycomparetheperformanceofdifferenttestsusedin thediagnosisofpediatricTB.Wefoundthatwhilethepres- enceofoneormoresignorsymptom,historyofTBcontact, andaCXRhadthehighestsingletestsensitivities,thecombination of thesetests withTST showedhighest accuracy andresultedinaPPVof100%inTBprevalenceratesvarying from1%to11%.

Evaluatingthetwodiagnosticscoringsystems,ourfind- ingsreaffirmedhighsensitivityoftheBMoHsystemwithcut offof30 andhighspecificity witha cutoff40, aswell as anNPVabove95%.¹⁴ Ourstudyis thefirsttoevaluatethe IUATLDsystem inaBrazilian population, whichwasprevi- ously shown a range of sensitivity and specificity of this system in different populations.¹⁰ In our cohort, this sys- temshowedlow sensitivityand highspecificity. aCXR was

(4)

Table1 SignsandsymptomspresentinTBandNotTBgroups.

Sign/Symptom TBGroup

n=21

%

NotTBGroup n=165

%

p-value

Cough 16

7

64 38.8%

0.001^a

Lengthofcough(days) 21.0^d

1---330^e

60.0^d 30---730^e

<0.001^b

Productivecough 14

66.7%

45 27.3%

<0.001^a

Lengthproductivecough(days) 21.0^d 3--- 330^e

60.0^d 30--- 730^e

<0.001^b

Drycough 2

9.5%

20 12.1%

1.00^a

Lengthdrycough(days) 15.0^d

1---180^e

30.0^d 30---30^e

0.52^b

Sweating 10

47.6%

26 15.8%

0.001^a

Weightloss 11

52.4%

25 15.1%

<0.001^a

Fever 9

42.9%

22 13.3%

0.002^a

Anorexia 7

33.3%

14 8.5%

0.002^a

Adynamia 5

23.8%

7 4.2%

0.003^a

Alteredauscultation^c 7

33.3%

7 4.2%

<0.001^a

a Fisher’sexacttest.

b Mann-Whitney.

c allreportedpresenceofcough.

d Averageofdays.

e Variationofdays;TB=tuberculosis.

the single diagnostic test that showed highest sensitivity, accuracy andPPV. While thisstrongly supportsthe use of CXR in diagnosis, it is importantto note that theclinical implementationofthisexaminationcanbecumbersome,as goodimagequalityandtrainedreadersarerequiredforreli- ableinterpretation.¹⁵TheIUATLDsystemwasdevelopedfor low-resourcesettingsanddoesnotincludeCXR.¹⁰Compared toouranalysisofcombinationsoftestsexcludingCXR,the IUATLDsystemhadthehighestPPV,conﬁrmingthatthissys- tem canbe useful asa TB diagnostic approachin regions whereCXRisnotavailable.

Tofurtherevaluatethetestsasapointofcarediagnosis approach,wealsoassessedperformanceintheabsenceof CXRandTSTresults.Wefoundthat----comparedtothecom- bination ofsigns/symptoms,historyofcontact,aCXR, and TST----signs/symptomsandhistoryofcontactaloneshoweda doublinginsensitivityandareductioninspeciﬁcity.More- over,whilePPVwasgreatlyreduced,NPVincreasedinthe absenceofCXRandTSTresults.Together,theseﬁndingsindi- catethatpresenceofsigns/symptomsandhistoryofcontact areusefultestsinpointofcarediagnosis forrulingoutTB suspects.

Thepresenceofatleastonesignorsymptomalonealso showed high singletest sensitivity, though each signs and symptoms evaluatedindividually had a muchlowersensi-

tivity.In previous studies, the individual sign or symptom withthe best performance has varied.^16---20 Together with ourdata,thissuggestsitisimportanttoconsiderallsigns andsymptomsratherthanfocusonaparticularonewhen diagnosingpediatricpatients.Further,manycasesofpedi- atricTB maybeasymptomatic.^21,22 Alongwithourﬁndings showingtheincreasedsensitivity,accuracy,andPPVofsigns andsymptomsin combination withother diagnostictests, thisindicatesthatsignsandsymptomsarebestinterpreted alongwithotherdiagnostictests.

PediatricTBmayhavehighmortalityifnotdetectedand treated,thoughwithpropertreatment,outcomesaregener- allygoodandfewadverseeffectsareobserved.²³Therefore, diagnostic tests should prioritize the avoidance of false negativesoverfalsepositives.⁹Thereforein thisstudywe focusedonidentifyingtestswithhighsensitivityandaccu- racy,ratherthanspeciﬁcity.

Inourcohort,aslightlylargerpercentageof NTBcases hadhistoryofTBcontact,yetTBcontactwaspartofthetest combinationthatshowedthehighestaccuracy.Ithasprevi- ouslybeenshownthathistoryofTBcontactisanimportant riskfactorforchildhoodTBinlow-incidencesettings,though itislessinformativeinhigh-incidencesettings.^19,22,24Atour teachinghospital,themajority ofpediatric patientseval- uatedforTB arecontactsofadult TBcases,andthusitis

(5)

Table2 Sensitivity,speciﬁcity,andaccuracyofsingleandcombinationsoftests.

SingleTests

Test(s) Sensitivity Speciﬁcity Accuracy

Cough 76.2 61.2 62.9

Sweating 47.6 84.2 80.1

Weightloss 52.4 84.9 81.2

Fever 42.9 86.7 81.7

Anorexia 33.3 91.5 85.0

Adynamia 23.8 95.8 87.6

Alteredauscultation 33.3 95.8 88.7

Oneormoresignorsymptom(SS) 85.7 56.4 59.7

Historyofcontact(HC) 85.7 7.9 16.7

Alteredchestx-ray(aCXR) 85.7 91.1 91.4

Positivetuberculinskintest(TST) 60.0 17.6 22.2

Smearpositive 23.5 100.0 69.1

Culturepositive 26.7 100.0 67.7

Combinationsoftests Sensitivity Speciﬁcity Accuracy

Cough+weightloss+anorexia 14.3 98.8 89.2

Cough+sweating+anorexia 14.3 95.8 86.6

SS+HC 71.4 61.8 62.9

SS+HC+TST 35.0 72.1 68.1

SS+HC+aCXR 60.0 99.4 95.1

HC+TST+aCXR 35.0 97.0 90.3

SS+HC+TST+aCXR 20.0 100.0 91.4

BMoH:cutoff30 95.2 91.5 91.9

BMoH:cutoff40 71.4 97.0 94.1

IUATLDI 28.6 95.8 88.2

IUATLDS 23.8 98.8 90.3

Legend:BMoH:BrazilianMinistryofHealth;IUATLD:InternationalUnionAgainstTuberculosisandLungDisease.

Table3 Positivepredictivevalueandnegativepredictivevalueforindividualtestsatvariousprevalencerates.

Test(s) Prevalenceoftuberculosisin children14andunder

1% 5% 11%

PPV NPV PPV NPV PPV NPV

Cough 2.0 99.6 9.4 98.0 19.5 95.4

Sweating 3.0 99.4 13.7 96.8 27.2 92.9

Weightloss 3.4 99.4 15.4 97.1 29.9 93.5

Fever 3.2 99.3 14.5 96.7 28.4 92.5

Anorexia 3.8 99.3 17.1 96.3 32.7 91.7

Adynamia 5.4 99.2 22.8 96.0 41.0 91.1

Alteredauscultation 7.4 99.3 4.7 91.3 10.3 81.7

Oneormoresignorsymptom(SS) 2.0 99.8 9.8 96.7 19.5 97.0

Historyofcontact(HC) 0.9 98.2 4.7 91.3 10.3 81.7

AlteredchestX-ray(aCXR) 8.9 99.8 33.7 99.2 54.4 98.1

Positivetuberculinskintest(TST) 0.7 97.8 3.7 89.3 8.3 78.1

Smearpositive 100.0 99.2 100.0 96.1 100.0 91.4

Culturepositive 100.0 99.3 100.0 96.3 100.0 91.7

Legend:VPP:PositivePredictiveValue;VPN:NegativePredictiveValue.

notsurprisingthathistoryofcontactaloneisinsufﬁcientfor diagnosis.TheTBcaseshadasigniﬁcantlyhigherrateexpo- suretomultiplecontacts(p=0.02),whichmayhaveledto a greater burden of exposure and illness. Thus, while TB

contactisanimportantdiagnostictest,ourresultsindicate thatitisbestappliedincombinationwithothertests.Thus, theuseofdiagnosticscoresasauxiliarytoolsinthediagno- sisofTBisfavored,sinceinisolationbothTSTandcontact

(6)

Table4 Positivepredictivevalueandnegativepredictivevalueforcombinationsoftestsatvariousprevalencerates.

Test(s) Prevalenceoftuberculosisin children14andunder

1% 5% 11%

PPV NPV PPV NPV PPV NPV

SS+HC 1.9 99.5 9.0 97.6 18.8 94.6

SS+HC+TST 1.3 99.1 6.2 95.5 13.4 90.0

SS+HC+aCXR 49.8 99.6 83.8 97.9 92.4 95.3

HC+TST+aCXR 10.5 99.3 37.8 96.6 58.8 92.4

SS+HC+TST+aCXR 100.0 99.2 100.0 95.7 100.0 91.0

BMoH:cutoff30 10.2 100.0 37.2 99.7 58.1 99.4

BMoH:cutoff40 19.2 99.7 55.4 98.5 74.5 96.5

IUATLDI 6.4 99.3 26.3 96.2 45.6 91.6

IUATLDS 16.6 99.2 50.9 96.1 70.9 91.3

Legend:VPP:PositivePredictiveValue; VPN:NegativePredictiveValue; SS:Oneormoresignorsymptom;HC:Historyofcontact;

TST:Positivetuberculinskintest;aCXR:AlteredchestX-ray;BMoH:BrazilianMinistryofHealth;IUATLD:InternationalUnionAgainst TuberculosisandLungDisease.

historyarelittlehelpinthediagnosis,especiallyinplaces ofhighincidenceofTB.

OtherspotentiallimitationsofourstudyisthatTSTwas conductedandreadat differentlocationspriortopatient arrival at the hospital.However,in Brazil thesetests can onlybeconducted,inthepublichealthsystem,byatrained health ofﬁcial using standardized tool. Thus, we believe variability in interpretation to be low.Additionally, chest radiographs wereread bytheattending physicianofeach subject, and therefore also may vary in their interpretation.BMoH,sinceitsﬁrstpublicationhasundergone some changes,especiallyrelatedtotheTST,thelastchangewas in2019.Thisscoreshouldstillbevalidatedwiththecurrent data.

Conclusion

In conclusion, using an expert panel to deﬁne the gold standard for TB diagnosis, we were able tocompare the performance of individual and combinations of pediatric diagnostictests.Thesearesimple,low-costtriageteststhat can beused asa rule-outdiagnostic byﬁrst-contact care providers,includingphysiciansandcommunityhealthwork- ers,asrecommendedbyWHO.Moreover,theBMoHsystem alsoperformedwellinourcohort.Thus,itshouldcontinue tobewidely usedinBrazilinsettingswithlowHIVpreva- lence.

Funding

ThisworkwassupportedbyProjectICOHRTA(International Clinical, Operational, and Health Services Research and Training Award) and Fogarty International Center grants

# U2RTW006885 ICOHRTA and #R25TW009338. Dr Isabela Almeida,receivedfounds for thiswork underthepos-doc schollarship.Process---CNPq/INCT465318/2014-2.

Ethical approval

This study was approved by the Ethics Committee of the ClinicalHospitaloftheFederalUniversityofParaná(CAAEE 0126.0208.000-08.26/06/2008).

Conﬂicts of interest

Theauthorsdeclarethatthereisnoconﬂictofinterest.

Acknowledgments

TheauthorsthankthesupportofProjectICOHRTAandFog- artyInternational Centergrants. The National Council for ScientiﬁcandTechnological Development--- CnpQandThe BrazilianTuberculosisResearchNetwork---REDE-TB.

References

1.Globaltuberculosisreport.Geneva:WorldHealthOrganization;

2018.

2.GrahamSM,AhmedT,Amanullah F,BrowningR, CardenasV, CasenghiM,etal.Evaluationoftuberculosisdiagnosticsinchil- dren:1.Proposedclinicalcasedeﬁnitionsforclassiﬁcationof intrathoracic tuberculosisdisease. Consensusfrom anexpert panel.JInfectDis.2012;205:S199---208.

3.Swaminathan S, Rekha B. Pediatric Tuberculosis: Global OverviewandChallenges.ClinInfectDis.2010;50:S184---94.

4.CarvalhoACC,CardosoCAA,MartireTM,MiglioriGB,Sant’Anna CC.Epidemiologicalaspects,clinicalmanifestations,andpre- ventionofpediatrictuberculosisfromtheperspectiveofthe EndTBStrategy.JBrasPneumol.2018;44(2):134---44.

5.EdwardsDJ,KiteteleF,VanRieA.Agreementbetweenclinical scoringsystemsusedforthediagnosisofpediatrictuberculosis intheHIVera.IntJTubercLungDis.2007;11(3):263---9.

6.HesselingAC,SchaafHS,GieRP,StarkeJR,BeyersN.Acritical reviewofdiagnosticapproachesusedinthediagnosisofchild- hoodtuberculosis.IntJTubercLungDis.2002;6(12):1038---45.

7.Perez-Velez CM,MaraisBJ.Tuberculosis inchildren. NEngl J Med.2012;367:348---61.

(7)

8.RossoniAMO,RossoniMD,RodriguesCO.CritériosdePontuac¸ão para Diagnóstico de Tuberculose em Crianc¸as. Pulmão RJ.

2013;22(3):65---9.

9.MS.Ministério daSaúdedoBrasil.Manualde recomendac¸ões paraocontroledatuberculosenoBrasil.SecretariadeVigilân- ciaemSaúde;2011.

10.Fourie PB,Becker PJ, Festenstein F, Migliori GB, Alcaide J, AntunesM,etal. Proceduresfor developingasimplescoring methodbasedonunsophisticatedcriteriaforscreeningchildren fortuberculosis.IntJTubercLungDis.1998;2(2):116---23.

11.MS.MinistériodaSaúdedoBrasil.Panoramadatuberculoseno Brasil:AmortalidadeemnúmerosSecretariadeVigilânciaem Saúde;2016.

12.LandisJR,KochGG.TheMeasurementofObserverAgreement forCategoricalData.Biometrics.1977;33(1):159---74.

13.DeeksJJ.Usingevaluationsofdiagnostictests:Understanding theirlimitationsand makingthemostofavailableevidence.

AnnOncol.1999;10(7):761---8.

14.PearceEC, WoodwardJF, Nyandiko WM,Vreeman RC, Ayaya SO. A systematic review of clinical diagnostic systems used in thediagnosis of tuberculosisin children. AIDS Res Treat.

2012:401896.

15.DuToitG,SwinglerG,IloniK.Observervariationindetecting lymphadenopathyonchestradiography.IntJTubercLungDis.

2002;6(9):814---7.

16.FrancoR, Santana MA,Matos E,Sousa V,Lemos ACM.Clini- calandradiologicalanalysisofchildrenandadolescentswith

tuberculosisinBahia,Brazil.BrazilianJInfectDis.2003;7(1):

73---81.

17.Marais BJ, Gie RP, Hesseling AC, Schaaf HS, Lombard C, Enarson DA, et al. A Reﬁned Symptom-Based Approach to Diagnose Pulmonary Tuberculosis in Children. Pediatrics.

2006;118(5):e1350---9.

18.MaraisBJ,ObiharaCC,GieRP,SchaafHS,HesselingAC,Lombard C,etal.Theprevalenceofsymptomsassociatedwithpulmonary tuberculosisinrandomlyselectedchildrenfromahighburden community.ArchDisChild.2005;90(11):1166---70.

19.RigoutsL.Clinicalpractice:Diagnosisofchildhoodtuberculosis.

EurJPediatr.2009;168(11):1285---90.

20.Rahman N, Pedersen KK, Rosenfeldt V, Johansen IS. Chal- lenges in diagnosing tuberculosis in children. Dan Med J.

2012;59(7):A4463.

21.Marais BJ, Gie RP, Schaaf HS, Hesseling AC, Obihara CC, StarkeJJ,etal.Thenaturalhistoryofchildhoodintra-thoracic tuberculosis: A critical review of literature from the pre- chemotherapyera.IntJTubercLungDis.2004;8(4):392---402.

22.NewtonSM,BrentAJ,AndersonS,WhittakerE,KampmannB.

Paediatrictuberculosis.LancetInfectDis.2008;8(8):498---510.

23.Guidancefornationaltuberculosisprogrammesonthemana- gement of tuberculosis in children. Geneva: World Health Organization;2014.

24.Hsu KHK. Contact Investigation: A Practical Approach to Tuberculosis Eradication. Am J Public Heal Nations Heal.

2008;53:1761---9.