DOI: 10.1542/peds.2006-0519 2006;118;e1350Pediatrics

Donald A. Enarson and Nulda BeyersBen J. Marais, Robert P. Gie, Anneke C. Hesseling, H. Simon Schaaf, Carl Lombard,

ChildrenA Refined Symptom-Based Approach to Diagnose Pulmonary Tuberculosis in

  

  http://pediatrics.aappublications.org/content/118/5/e1350.full.html

located on the World Wide Web at: The online version of this article, along with updated information and services, is

 

of Pediatrics. All rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.Boulevard, Elk Grove Village, Illinois, 60007. Copyright © 2006 by the American Academy published, and trademarked by the American Academy of Pediatrics, 141 Northwest Pointpublication, it has been published continuously since 1948. PEDIATRICS is owned, PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly

by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

ARTICLE

A Refined Symptom-Based Approach to DiagnosePulmonary Tuberculosis in ChildrenBen J. Marais, MRCP, FCP, MMed, PhDa, Robert P. Gie, FCPa, Anneke C. Hesseling, MBChB, MSca, H. Simon Schaaf, MMed, MDa,

Carl Lombard, PhDb, Donald A. Enarson, MD, PhDc, Nulda Beyers, FCP, MSc, PhDa

aDesmond Tutu Tuberculosis Centre and Department of Paediatrics and Child Health, Tygerberg Children’s Hospital, Stellenbosch University, Cape Town, South Africa;bMedical Research Council of South Africa, Cape Town, South Africa; cInternational Union Against Tuberculosis and Lung Disease, Paris, France

The authors have indicated they have no financial relationships relevant to this article to disclose.

ABSTRACT

BACKGROUND. Tuberculosis control programs place an almost exclusive emphasis onadults with sputum smear-positive tuberculosis, because they are most infectious.However, children contribute a significant proportion of the global tuberculosiscaseload and experience considerable tuberculosis-related morbidity and mortal-ity, but few children in endemic areas have access to antituberculosis treatment.The diagnostic difficulty experienced in endemic areas with limited resources hasbeen identified as a major factor contributing to poor treatment access. In general,there is a sense of scepticism regarding the potential value of symptom-baseddiagnostic approaches, because current clinical diagnostic approaches are oftenpoorly validated. The natural history of childhood tuberculosis demonstrates thatsymptoms may offer good diagnostic value if they are well defined and if appro-priate risk stratification is applied. This study aimed to determine the value ofwell-defined symptoms to diagnose childhood pulmonary tuberculosis in a tuber-culosis-endemic area.

METHODS.A prospective, community-based study was conducted in Cape Town,South Africa. Specific well-defined symptoms were documented in all children�13 years of age reporting a persistent, nonremitting cough of �2 weeks’ dura-tion; study participants were thoroughly evaluated for tuberculosis. In addition, allof the children who received antituberculosis treatment during the study periodwere reviewed by the investigator, irrespective of study inclusion. This concurrentdisease surveillance provided a comprehensive overview of all of the childhoodtuberculosis cases, allowing accurate assessment of the possible disadvantagesassociated with this symptom-based diagnostic approach. In the absence of anacceptable gold standard test, optimal case definition is an important consider-ation. Children were categorized as “bacteriologically confirmed tuberculosis,”“radiologically certain tuberculosis,” “probable tuberculosis,” or “not tuberculosis.”Bacteriologically confirmed tuberculosis was defined as the presence of acid-fastbacilli on sputum microscopy and/or Mycobacterium tuberculosis cultured from arespiratory specimen. Radiologically certain tuberculosis was defined as agreement

www.pediatrics.org/cgi/doi/10.1542/peds.2006-0519

doi:10.1542/peds.2006-0519

Dr Marais had full access to all the datareported in this study and had finalresponsibility for the decision to submit forpublication.

KeyWordssymptom-based, diagnosis, tuberculosis,children

AbbreviationsTST—tuberculin skin testCXR—chest radiographPPV—positive predictive valueOR—odds ratioCI—confidence interval

Accepted for publication Jun 12, 2006

Address correspondence to Ben J. Marais,MRCP, FCP, MMed, PhD, Department ofPaediatrics and Child Health, Desmond TutuTuberculosis Centre, Faculty of HealthSciences, Stellenbosch University, PO Box19063, Tygerberg, 7505, South Africa. E-mail:bjmarais E-mail: bjmarais@sun.ac.za

PEDIATRICS (ISSN Numbers: Print, 0031-4005;Online, 1098-4275). Copyright © 2006 by theAmerican Academy of Pediatrics

e1350 MARAIS et al by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

between both independent experts that the chest radio-graph indicated certain tuberculosis in the absence ofbacteriologic confirmation. Probable tuberculosis wasdefined as the presence of suggestive radiologic signs andgood clinical response to antituberculosis treatment inthe absence of bacteriologic confirmation or radiologiccertainty. Good clinical response was defined as com-plete symptom resolution and weight gain of �10% ofbody weight at diagnosis, within 3 months of startingantituberculosis treatment. Not tuberculosis was definedas spontaneous symptom resolution or no response toantituberculosis therapy in the absence of bacteriologicconfirmation or radiologic signs suggestive of tuberculo-sis. Pulmonary tuberculosis was defined as a symptom-atic child with: (1) bacteriologically confirmed tubercu-losis, (2) radiologically confirmed tuberculosis, or (3)probable tuberculosis (as defined), excluding isolatedpleural effusion.

RESULTS. In total, 1024 children were referred for evalua-tion. Resolving symptoms were reported in 596 children(58.2%); 428 (41.8%) children with persistent, nonre-mitting symptoms at evaluation were investigated fortuberculosis. Pulmonary tuberculosis was diagnosed in197 children; 96 were categorized as bacteriologicallyconfirmed tuberculosis, 75 as radiologically certain tu-berculosis, and 26 as probable tuberculosis. Combining apersistent nonremitting cough of �2 weeks’ duration,documented failure to thrive (in the preceding 3months), and fatigue provided reasonable diagnostic ac-curacy in HIV-uninfected children (sensitivity: 62.6%;specificity: 89.8%; positive predictive value: 83.6%); theperformance was better in the low-risk group (�3 years;sensitivity: 82.3%; specificity: 90.2%; positive predictivevalue: 82.3%) than in the high-risk group (�3 years;sensitivity: 51.8%; specificity: 92.5%; positive predictivevalue: 90.1%). In children with an uncertain diagnosisat presentation, clinical follow-up was a valuable diag-nostic tool that further improved diagnostic accuracy,particularly in the low-risk group. Symptom-based ap-proaches offered little diagnostic value in HIV-infectedchildren. Three (15%) of the 20 HIV-infected childrendiagnosed with pulmonary tuberculosis failed to reportsymptoms of sufficient duration to warrant study inclu-sion, whereas 25% reported persistent, nonremittingsymptoms in the absence of tuberculosis. In addition, thetuberculin skin test was positive in �20% of HIV-in-fected children diagnosed with pulmonary tuberculosis.

DISCUSSION. The combined presence of 3 well-definedsymptoms at presentation (persistent, nonremittingcough of �2 weeks’ duration; objective weight loss [doc-umented failure to thrive] during the preceding 3months; and reported fatigue) provided good diagnosticaccuracy in HIV-uninfected children �3 years of age,with clinical follow-up providing additional value. Theapproach performed less well in children �3 years.

However, the presence of a persistent, nonremittingcough together with documented failure to thrive stillprovided a fairly accurate diagnosis (sensitivity: 68.3%;specificity: 80.1%; positive predictive value: 82.1%), il-lustrating the importance of regular weight monitoringin young children. Clinical follow-up also offered addi-tional diagnostic value, but caution is required, becausevery young children have an increased risk of rapiddisease progression. The approach performed poorly inHIV-infected children. Recent household contact withan adult index case seemed to provide more diagnosticvalue than a positive tuberculin skin test, but novelT-cell-based assays may offer the only real improvementin sensitivity to diagnose M tuberculosis infection in HIV-infected children. The variable diagnostic value offeredby this symptom-based diagnostic approach illustratesthe importance of risk stratification, as demonstrated bythe fact that 11 (91.7%) of 12 children with severedisease manifestations who failed to meet the entry cri-teria were �3 years of age or HIV infected. Particularemphasis should be placed on the provision of preven-tive chemotherapy after documented exposure and/orinfection in these high-risk children. Study limitationsinclude the small number of HIV-infected children, buton the positive side, the large number of HIV-uninfectedchildren permitted adequate evaluation in this impor-tant group. It is often forgotten that HIV-uninfectedchildren constitute the majority of child tuberculosiscases, even in settings where HIV is endemic. This studydemonstrates the importance of ascertaining a child’sHIV status before symptom-based diagnosis is attempted.Because children were recruited at both the clinic andhospital level, some selection bias may have been intro-duced; however, the only significant difference betweenthe 2 groups was the proportion of HIV-infected chil-dren. Pulmonary tuberculosis was diagnosed with differ-ent levels of certainty, but no significant differenceswere recorded between these groups.

CONCLUSIONS. Pulmonary tuberculosis can be diagnosedwith a reasonable degree of accuracy in HIV-uninfectedchildren using a simple symptom-based approach. Thisoffers the exciting prospect of improving treatment ac-cess for children, particularly in resource-limited settingswhere current access to antituberculosis treatment ispoor.

TUBERCULOSIS CONTROL PROGRAMS place an almostexclusive emphasis on adults with sputum smear-

positive disease, because they are most infectious. How-ever, children contribute a significant proportion of theglobal tuberculosis caseload1,2 and experience consider-able tuberculosis-related morbidity and mortality.3 Fewchildren in endemic areas have access to antituberculosis

PEDIATRICS Volume 118, Number 5, November 2006 e1351 by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

treatment,4 and diagnostic difficulty is a major contrib-uting factor.5,6

The diagnosis of tuberculosis in children is compli-cated by the absence of a practical reference test, becauserespiratory specimens are difficult to collect, and thereported bacteriologic yield is low5,6; sputum smearsare positive in �10%–15% of children diagnosed withtuberculosis, and confirmation by culture is achievedin only 30%–40%.4,7 The value of the classic diagnostictriad: (1) exposure to an adult index case, (2) a posi-tive tuberculin skin test (TST), and (3) the presence ofsuggestive signs on the chest radiograph (CXR), isgreatly reduced in endemic areas where exposure toand/or infection with Mycobacterium tuberculosis are com-mon. Consequently, in endemic areas, the diagnosisof childhood tuberculosis rests predominantly on thesubjective interpretation of the CXR,8,9 which is sub-ject to pronounced interobserver and intraobserver vari-ability10 and is rarely available in resource-limited set-tings.

In general, there is a sense of scepticism regardingthe potential diagnostic value of symptom-based ap-proaches. A recent comprehensive review of symptom-based diagnostic approaches emphasized the absence ofstandard symptom definitions and the inadequate val-idation of current approaches.11 Previous descriptivestudies were mostly hospital based,12,13 complicating theextrapolation of results to the community level. A recentcommunity-based survey reported that poorly definedsymptoms traditionally associated with tuberculosis oc-cur commonly in randomly selected healthy children;they occur too commonly to offer any real diagnosticvalue.14

However, the natural history of childhood tubercu-losis demonstrates that symptoms may have diagnosticvalue if appropriate risk stratification is applied.15 Inlow-risk children (immune competent, �3 years of age),disease progression after primary M tuberculosis infec-tion is rare and is associated with the presence of persis-tent, nonremitting symptoms.15 In high-risk children(�3 years of age and/or immune compromised), diseaseprogression occurs more frequently and may be as-sociated with acute symptom onset,15 which reducesthe opportunity for and safety of symptom-based diag-nosis.

In a small pilot study, the use of well-defined symp-toms with a persistent, nonremitting character providedgood diagnostic value in low-risk children.16 However,the true value of these well-defined symptoms re-quired further evaluation in a large prospective study,with community-based recruitment to limit selectionbias and concurrent disease surveillance to documentpotential disadvantages. The current study aimed to de-termine the value of well-defined symptoms to diagnosechildhood pulmonary tuberculosis in a tuberculosis-en-demic area.

METHODS

Study SettingA prospective, community-based study was conductedfrom February 2003 through January 2005 in CapeTown, South Africa. In the study area, the incidence ofall adult tuberculosis was 845 in 100 000, and the tu-berculosis incidence in children �13 years of age was407 in 100 000 in 2004.17 The prevalence of HIV infec-tion among children treated for tuberculosis was 8.8%.17

Five primary health care clinics were selected that allrefer to Tygerberg Children’s Hospital. The study popu-lation rarely access private medical services, and childrendiagnosed with tuberculosis are routinely referred to theprimary health care clinic for supervised antituberculosistreatment. Pediatric services are accessible to children�13 years of age.

Study PopulationStudy entry criteria were children �13 years of age whoreported a persistent, nonremitting cough of �2 weeks’duration, without response to a course of oral antibiot-ics. Nurses within the selected clinics were requested torefer all of the eligible children to the investigator. Theinvestigator visited each clinic twice weekly to evaluatereferred children. The following measures were put inplace to ensure that all of the children who met the entrycriteria were included in the study: (1) training of all ofthe pediatric nurses within the study clinics, (2) postingof prominent reminders, and (3) giving weekly feedbackand motivation to the nurses. In addition, a study nursewas employed to screen all of the children admitted tothe local referral hospital (Tygerberg Children’s Hospi-tal). Hospitalized children who met the entry criteriawere referred to the investigator by the study nurse andwere evaluated in the same fashion as those referred bythe clinic nurses.

Data CollectionA screening questionnaire documented the presence anduninterrupted duration of symptoms. Children whowere referred but whose symptoms resolved or showedmarked improvement before evaluation were not inves-tigated for tuberculosis. However, all of the referredchildren completed the screening questionnaire andwere followed until symptom resolution. Children wereinstructed to return if symptoms recurred. Referral wasnot restricted to a single episode; children were evalu-ated in a similar fashion in the event of a second referral.

Initial Assessment andManagementInvestigation for tuberculosis included completion of acomprehensive questionnaire, weight measurement andassessment of weight gain on the Road-to-Health growthchart during the preceding 3 months, performance of aTST and CXR, and attempts to collect �1 gastric aspirate

e1352 MARAIS et al by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

and/or sputum specimen. HIV testing was offered to-gether with standard pretest and posttest counseling toall of the children.

Children diagnosed with tuberculosis were started onstandard antituberculosis therapy: 2 months of 3 drugs(isoniazid, rifampin, and pyrazinamide) followed by 4months of 2 drugs (isoniazid and rifampin), unless theywere exposed to an index case with known drug resis-tance in which case treatment was individualized. Chil-dren �2 years of age with known exposure and/or apositive TST result received isoniazid preventive chemo-therapy once active tuberculosis was excluded. All of theother children were treated according to the most likelyalternative diagnosis and reviewed after 2 to 4 weeks. Ifsymptoms persisted beyond an additional 4 weeks, fullassessment, including TST (if previously negative), CXR,and mycobacterial cultures, were repeated. At this time,every effort was made to establish a final diagnosis andall children �5 years of age with documented tubercu-losis exposure or infection, but not diagnosed with activetuberculosis, were put on isoniazid preventive chemo-therapy as prescribed by the National Tuberculosis Pro-gramme. The duration of uninterrupted symptoms untilspontaneous symptom resolution or the onset of antitu-berculosis chemotherapy was recorded. Children whoreceived antituberculosis treatment were reviewed after3 months to document response to therapy.

Concurrent Disease SurveillanceAll of the children who received antituberculosis treat-ment during the period of study enrollment and for anadditional surveillance period of 3 months were docu-mented and reviewed by the investigator. This concur-rent disease surveillance provided a comprehensiveoverview of all the children treated for tuberculosis,irrespective of study inclusion, allowing an accurate as-sessment of the possible disadvantages associated withthis symptom-based diagnostic approach.

TSTA TST, using intradermal injection of 2 tuberculin unitsof purified protein derivative (PPD RT 23; Statens SerumInstitut, Copenhagen, Denmark), was performed on thevolar aspect of the left forearm. The transverse diameterof induration was measured in millimeters after 48 to 72hours. A positive TST was regarded as a measurement of�10 mm in HIV-uninfected children, and �5 mm inHIV-infected children.

CXRStandard anteroposterior and lateral views were doneand read by 2 independent experts, blinded to all of theclinical information and to each other’s interpretation.Findings were documented on a standard report formand categorized as “certain tuberculosis,” “uncertain tu-berculosis,” or “certain not tuberculosis.” Disease mani-

festations were classified according to a recently pro-posed radiologic classification of childhood intrathoracictuberculosis.18

HIV TestingA rapid test was used to screen for HIV infection (Deter-mine HIV 1/2 rapid test; Abbott, Tokyo, Japan). All ofthe children with a positive rapid test result were re-ferred to the HIV family clinic at Tygerberg Children’sHospital for confirmatory tests (by polymerase chainreaction in children �18 months and ELISA in olderchildren) and clinical management.

Specimen Collection andMycobacterial CultureWe attempted to collect �1 culture specimen from eachchild; multiple specimens were collected from childrensent to the referral hospital. Samples were inoculatedinto liquid media using either BACTEC or MGIT systems(Becton Dickinson, Sparks, MD). Positive cultures wereconfirmed to be M tuberculosis by routine polymerasechain reaction speciation.19

DefinitionsChildren were categorized as “bacteriologically con-firmed tuberculosis,” “radiologically certain tuberculo-sis,” “probable tuberculosis,” or “not tuberculosis.” Bac-teriologically confirmed tuberculosis was defined asthe presence of acid-fast bacilli on sputum microscopyand/or M tuberculosis cultured from a respiratory speci-men. Radiologically certain tuberculosis was defined asagreement between both independent experts that theCXR indicated certain tuberculosis in the absence ofbacteriologic confirmation. Probable tuberculosis wasdefined as the presence of suggestive radiologic signs andgood clinical response to antituberculosis treatment inthe absence of bacteriologic confirmation or radiologiccertainty. Good clinical response was defined as com-plete symptom resolution and weight gain of �10% ofbody weight at diagnosis, within 3 months of startingantituberculosis treatment. Not tuberculosis was definedas spontaneous symptom resolution or no response toantituberculosis therapy in the absence of bacteriologicconfirmation or radiologic signs suggestive of tuberculo-sis. Pulmonary tuberculosis was defined as a symptom-atic child with: (1) bacteriologically confirmed tubercu-losis, (2) radiologically confirmed tuberculosis, or (3)probable tuberculosis (as defined), excluding isolatedpleural effusion.

EthicsParents gave written informed consent for study partic-ipation and gave separate written consent for HIV test-ing. Ethics approval was obtained from the InstitutionalReview Board of Stellenbosch University, the City ofCape Town Health Department, and from local commu-nity health advisory boards.

PEDIATRICS Volume 118, Number 5, November 2006 e1353 by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

Statistical AnalysisData were dually entered into an Access (Microsoft,Redmond, WA) relational database and validated. De-scriptive analyses were done using SPSS 13.0 (SSPS Inc,Chicago, IL). The sensitivity, specificity, and positive pre-dictive value (PPV) of individual variables were calcu-lated. Multivariate tree regression analysis was used toidentify the individual variables that statistically contrib-uted most diagnostic value. For analysis purposes, 2high-risk groups were identified: (1) HIV-uninfectedchildren �3 years of age and (2) HIV-infected children.HIV-uninfected children �3 years were regarded as low-risk. The sensitivity, specificity, and PPV were calculatedfor each of these relevant risk groups when cumulativelycombining the 3 variables that statistically contributedthe most diagnostic value on multivariate tree regressionanalysis.

RESULTSIn total, 1024 children were referred; 596 (58.2%) re-ported marked improvement before evaluation by theinvestigator, and 428 (41.8%) were enrolled with a per-sistent, nonremitting cough, of whom 197 were diag-nosed with pulmonary tuberculosis (Fig 1). There wereno significant differences in age, gender, or HIV statusbetween children with bacteriologically confirmed tu-berculosis and those with radiologically certain tubercu-losis or probable tuberculosis.

Table 1 reflects the demographics and clinical charac-teristics of all of the children referred for study partici-pation. Demographic variables (gender and age group)were comparable between the hospital- and clinic-re-cruited cohorts. The only significant difference was theproportion of HIV-infected children (proportion of thosetested: hospital: 26 of 74 [35.1%]; clinic: 11 of 223

FIGURE 1Flow diagram of all children evaluated. TB indicates tuberculosis; PTB, pulmonary tuberculosis defined as a symptomatic child with: (1) bacteriologically confirmed tuberculosis (froma respiratory or gastric aspirate specimen), (2) radiologically confirmed tuberculosis, or (3) probable tuberculosis (as defined), excluding isolated pleural effusion.

e1354 MARAIS et al by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

[4.9%]; odds ratio [OR]: 7.12; 95% confidence interval[CI]: 3.18–16.22).

Table 2 reflects concurrent disease surveillance data,

including all 425 children treated for tuberculosis dur-ing the study period, irrespective of study inclusion. Ofthe 214 children diagnosed with pulmonary tubercu-losis (as defined), 197 (92.1%) met the study entrycriteria. Of the 17 children who did not meet the studyentry criteria, 5 (29.4%) had isolated hilar adenopathyand were actively traced household contacts of an adultindex case; 3 (17.6%) had parenchymal consolidationand/or airway compression and reported a cough of �2weeks’ duration (1 HIV infected and 2 HIV uninfected�3 years of age); and 9 (53.0%) had disseminated (mil-iary) disease, of whom 3 reported a cough of �2 weeks’duration, and 6 reported no cough at all (2 HIV infected,6 HIV uninfected �3 years, and 1 HIV uninfected �3years). The only mortality was a 3-month-old HIV-un-infected child who presented with acute symptom onsetand died in hospital with disseminated (miliary) disease.

None of the children with an uncertain diagnosis atthe initial assessment progressed to severe disease, suchas disseminated (miliary) disease, tuberculous meningi-tis, or cavitation during the surveillance period. Four-teen children were referred twice, of whom 2 werediagnosed with pulmonary tuberculosis during the sec-ond episode. None of the children discharged without atuberculosis diagnosis were subsequently treated for tu-berculosis during the surveillance period, apart from the2 who were referred a second time.

Table 3 indicates the value of individual variables todiagnose pulmonary tuberculosis at presentation withinthe relevant risk groups. The value of a positive TSTdiffered significantly between relevant risk groups. Com-paring the value of a positive TST in HIV-uninfected andHIV-infected children demonstrated that the sensitivityof the TST was significantly better in HIV-uninfectedchildren (156 of 179 [87.2%] vs 3 of 17 [17.6%]; OR:31.6; 95% CI: 7.6–151.5). In HIV-infected children, ahistory of household contact with an adult tuberculosis

TABLE 1 Demographics and Clinical Characteristics of All ChildrenReferredWith a Cough of More Than 2Weeks’ Duration(n � 1024)

Characteristics N (%)

Not investigated for tuberculosis, resolvingcough at evaluation (n � 596)

Age group�3 y 287 (48.2)�3 y 309 (51.8)

GenderMale 297 (49.8)Female 299 (50.2)

HIV tested 0PTB 0

Investigated for tuberculosis, persistent, nonremittingcough at evaluation (n � 428)

Age group�3 y 185 (43.2)�3 y 243 (56.8)

GenderMale 225 (52.6)Female 203 (47.4)

HIV tested 297 (69.4)Ratio infectedOverall 37/297 (12.4)Recruited in the hospital 26/74 (35.1)Recruited in the clinic 11/223 (4.9)

TST done 421 (98.4)�10 mm (HIV-uninfected) 226/384 (58.8)�5 mm (HIV-infected) 6/37 (16.2)

PTB 197 (46.0)Relevant risk groupsHIV-infected 17/197 (8.6)�3 y (HIV-uninfected) 94/197 (47.7)�3 y (HIV-uninfected) 86/197 (43.6)

PTB indicates pulmonary tuberculosis defined as a symptomatic childwith: (1) bacteriologicallyconfirmed tuberculosis (from a respiratory or gastric aspirate specimen), (2) radiologically con-firmed tuberculosis, or (3) probable tuberculosis (as defined), excluding isolated pleuraleffusion.

TABLE 2 Concurrent Disease Surveillance Data, Reflecting All Children Treated for Tuberculosis During the Study Period, Irrespective of StudyInclusion

Certainty of Diagnosis Total No. (%) PTB Diagnosis Study Inclusiona

CXR signs absentExposure/infection only 84 (19.8) 0/84 4/84 (4.8)Extrathoracic tuberculosis only 45 (10.6) 0/45 0/45CXR signs presentUncertain HA 69 (16.2) 3/69 (4.3) 27/69 (39.1)Certain HA 76 (1.2) 76/76 (100) 71/76 (93.4)Parenchymal consolidation and/or airway compression 107 (25.2) 107/107 (100) 104/107 (97.2)Disseminated (miliary) disease 14 (3.0) 14/14 (100) 5/14 (35.7)Adult-type disease 14 (3.3) 14/14 (100) 14/14 (100)Pleural effusion only (excluded from PTB outcome definition) 17 (4.0) 0/17 4/17 (23.5)Total 425 (100)Total treated for tuberculosis with PTB 214/425 (50.4)Total with PTB included in the study 197/214 (92.1)

CXR signs indicates signs suggestive of PTB on the CXR; HA, hilar adenopathy; PTB, pulmonary tuberculosis defined as a symptomatic child with: (1) bacteriologically confirmed tuberculosis (froma respiratory or gastric aspirate specimen), (2) radiologically confirmed tuberculosis, or (3) probable tuberculosis (as defined), excluding isolated pleural effusion.a In total, 229 of the 425 children treated for TB met the study entry criteria, of whom 197 (86%) were diagnosed with pulmonary TB (Fig 1).

PEDIATRICS Volume 118, Number 5, November 2006 e1355 by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

index case seemed more sensitive (9 of 17 [52.9%]) thana positive TST (3 of 17 [17.6%]), but this failed to reachstatistical significance (OR: 5.2; 95% CI: 0.89–34.7).

Table 4 reflects the sensitivity, specificity, and PPVachieved within the relevant risk groups by combiningthe 3 variables that statistically contributed the mostdiagnostic value on multivariate tree regression analysis.Figure 2 graphically illustrates the improved diagnosticaccuracy achieved by cumulatively combining thesevariables in HIV-uninfected children. In children �3years, the use of a positive TST as the third cumulativevariable provided better diagnostic accuracy (sensitivity:67.3%; specificity: 93.8%; PPV: 93.2%) than fatigue(sensitivity: 51.8%; specificity: 92.5%; PPV: 90.1%).

In case of an uncertain diagnosis at presentation (all 3criteria not met), persistence of a nonremitting cough at

clinical follow-up provided additional diagnostic value.Children diagnosed with tuberculosis during the initialassessment received treatment without delay; for thepurposes of this analysis, we accepted that their coughwould have persisted in the absence of treatment (implyingcomplete sensitivity). Persistence of a nonremitting coughafter �2 weeks of clinical follow-up considerably im-proved the diagnostic accuracy in HIV-uninfected chil-dren (�3 years: specificity: 98.9%; PPV: 85.1%; �3years: specificity: 82.6%; PPV: 88.6%), with further im-provement after �4 weeks in children �3 years of age(specificity: 98.6%; PPV: 88.6%). However, persistenceof a nonremitting cough added less diagnostic value inHIV-infected children, in whom optimal accuracy wasachieved when objective weight loss and reported fa-tigue at presentation were combined with a persistentcough at follow-up (sensitivity: 56.2%; specificity:75.0%; PPV: 77.2%).

DISCUSSIONIn this study, a simple symptom-based approach diag-nosed childhood pulmonary tuberculosis with a highdegree of accuracy in HIV-uninfected children �3 yearsof age. It is often assumed that these low-risk childrencontribute little to the tuberculosis disease burden, butthey contributed �50% of the pulmonary tuberculosiscaseload in this study. The vast majority of individuals inhighly endemic areas become infected with M tuberculosisduring childhood, but after 2 years of age. The largenumber of infections probably explains the high diseaseburden in these children, despite their relatively low riskto progress to disease.

TABLE 3 The Value of Individual Variables, Documented at Presentation in Relevant Risk Groups, to Diagnose Pulmonary Tuberculosis inChildren

Individual Variablesat Presentationa

Low RiskHIV Uninfected �3 y

High Risk

HIV Uninfected �3 y HIV Infected

Sensitivity Specificity PPV Sensitivity Specificity PPV Sensitivity Specificity PPV

Cough �3 wk 90.7 65.7 33.8 80.9 58.1 33.5 100 45.0 60.7Cough �4 wk 73.3 90.1 58.9 59.6 89.9 60.2 88.2 65.0 68.2Chest pain 9.3 96.9 36.4 3.2 99.7 75.0 0 95.0 0Hemoptysis 5.8 98.1 38.4 1.1 100 100 0 95.0 0Respiratory distress 4.7 99.1 50.0 8.5 94.9 30.8 35.3 60.6 42.8Fever 37.2 92.4 48.5 43.6 89.9 52.6 52.9 30.0 39.1Night sweats 57.0 93.9 64.5 40.4 95.2 70.3 41.2 60.0 46.7Fatigue 94.2 87.0 58.3 68.1 95.5 78.0 94.1 30.0 53.3Weight loss subjective 70.9 84.8 47.3 60.6 91.9 65.5 88.2 30.0 51.7Weight loss objective 82.4 84.4 75.0 75.3 79.1 82.1 82.3 35.0 51.9TST positive 89.3 62.9 59.1 81.9 70.1 77.7 17.6 85.0 50.0Household contact 48.8 72.3 52.5 69.1 53.6 67.0 52.9 60.0 52.9

Cough indicates persistent, nonremitting cough not responding to a course of oral antibiotics; Chest pain, chest pain reported by the child; Hemoptysis, blood in the sputum (not hematemesis ora nose bleed); Respiratory distress, difficult breathing at rest, reported by the parent/caregiver; Fever, fever reported by the parent/caregiver; Night sweats, regular sweating that requires a dry setof nightclothes; Fatigue, perceived decrease in playfulness/activity since the onset of coughing, reported by the parent/caregiver;Weight loss subjective, perceivedweight loss during the preceding3months, reported by the parent/caregiver; Weight loss objective, clear deviation from a previous growth trajectory and/or documented crossing of centile lines in the preceding 3months, usingthe Road-to-Health growth chart and accurate weight-for-age centiles28; TST positive, tuberculin skin test size �10 mm (or �5 mm if HIV infected); Household contact, household exposure to anadult index case in the preceding 12 months.a Definition of individual variables.

TABLE 4 The value of Combined Variables, Documented atPresentation in Relevant Risk Groups, to DiagnosePulmonary Tuberculosis in Children

Combined Variables at Presentationa Value

Sensitivity Specificity PPV

Low-risk children�3 years and HIV uninfected 82.3 90.2 82.3High-risk children�3 years and HIV uninfected 51.8 92.5 90.1HIV infected (irrespective of age) 56.2 61.8 61.9

a We combined the 3 variables, documented at presentation, that statistically contributed themost diagnostic value on multivariate tree regression analysis. The 3 variables included were:(1) persistent, non-remittent coughing for�2 weeks; (2) objective weight loss (clear deviationfrom a previous growth trajectory and/or documented crossing of the centile lines) in thepreceding 3months; and (3) reported fatigue (perceived decrease in playfullness/activity sincethe onset of coughing).

e1356 MARAIS et al by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

The combined presence of 3 well-defined symptomsat presentation: (1) a persistent, nonremitting cough of�2 weeks’ duration; (2) objective weight loss (docu-mented failure to thrive) during the preceding 3 months;and (3) reported fatigue, provided good diagnostic accu-racy in HIV-uninfected children �3 years of age. How-ever, a sensitivity of 81% implies that �20% of pulmo-nary tuberculosis cases would have been missed,because they failed to meet these strict diagnostic crite-ria. Fortunately, clinical follow-up proved to be a valu-able additional diagnostic tool that differentiated tuber-culosis from other common conditions, particularly inthis low-risk group.

The symptom-based diagnostic approach performedless well in HIV-uninfected children �3 years of age.Fatigue added little diagnostic value, because it was dif-ficult for parents/legal guardians to comment on its pres-ence. However, the presence of a persistent, nonremit-ting cough together with documented failure to thriveprovided a fairly accurate diagnosis on its own (sensitiv-ity: 68.3%; specificity: 80.1%; PPV: 82.1%), which illus-trates the importance of regular weight monitoring. In

our experience, documenting the response to simplemeasures, such as food and/or iron supplementationand/or deworming, further improves the diagnosticvalue of documented failure to thrive. The observationthat inclusion of a positive TST only provided improveddiagnostic performance in children �3 years of age maybe expected in highly endemic settings where a largepercentage of older children would have been exposedto tuberculosis before.20 As in older children, clinicalfollow-up offered additional diagnostic value, but morecaution is required, especially in very young childrenwho have a greatly increased risk of rapid disease pro-gression.

Symptom-based approaches offered limited diagnos-tic value in HIV-infected children. Three (15%) of the 20HIV-infected children diagnosed with pulmonary tuber-culosis failed to report symptoms of sufficient durationto warrant study inclusion, whereas 25% reported per-sistent, nonremitting symptoms in the absence of tuber-culosis. In addition, the TST was positive in �20% ofHIV-infected children diagnosed with pulmonary tuber-culosis. The poor sensitivity of the TST in HIV-infected

FIGURE 2The improved diagnostic accuracy achieved by cumulatively combining symptoms documented at presentation in HIV-uninfected children. This graph illustrates the improvedspecificity achieved (moving closer to 0 on the x-axis) and the sensitivity sacrificed (moving away from 1 on the y-axis) by cumulatively combining the following 3 optimal variablesidentified by multivariate tree regression analysis: (1) persistent nonremitting cough of �2 weeks, which was the entry criterion; (2) objective weight loss during the preceding 3months; and (3) reported fatigue, to diagnose pulmonary tuberculosis in HIV-uninfected children. The denominator used to calculate the sensitivity included all children withpulmonary tuberculosis identified during concurrent disease surveillance irrespective of study inclusion.

PEDIATRICS Volume 118, Number 5, November 2006 e1357 by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

children is well documented and correlates with thedegree of immune suppression.21 This indicates the im-portance of evaluating more sensitive tests to confirm orexclude M tuberculosis infection in immune-compro-mised children. New T-cell-based assays, particularly theELISPOT test, seem to offer improved sensitivity in HIV-infected children with tuberculosis,22 whereas the per-formance is unaffected by the degree of immune sup-pression in adults.23 In this study, recent householdcontact with an adult index case seemed to provide morediagnostic value than a positive TST, but the sensitivityand specificity achieved remained poor.

The variable diagnostic value offered by this symp-tom-based approach illustrates the importance of riskstratification,24 which provides the motivation behindcurrent World Health Organization guidelines to screenall children �5 years in household contact with an adultindex case.25 However, relevant risk stratification has notbeen widely incorporated into clinical diagnostic ap-proaches. The natural history of disease demonstrates asubstantial reduction in the risk to progress to disease ifprimary infection occurs after 2 years of age. As diseaseprogression usually occurs within the first 12 months,we defined HIV-uninfected children �3 years of age,and all HIV-infected children irrespective of age, as highrisk.15,26 The importance of risk stratification is demon-strated by the fact that 11 (91.7%) of 12 children withparenchymal consolidation and/or airway compressionand/or disseminated (miliary) disease who did not meetthe entry criteria were either �3 years or HIV infected.Particular emphasis should be placed on the provision ofpreventive chemotherapy after exposure and/or docu-mented infection in these high-risk children.

Study limitations include the small number of HIV-infected children enrolled, which limits our ability tocomment on this particular group. However, we wereable to demonstrate that a symptom-based approach,which does not include an accurate measure of M tuber-culosis infection, has little diagnostic value in HIV-in-fected children. On the positive side, the large number ofHIV-uninfected children provided an excellent opportu-nity to describe both the diagnostic performance andpotential disadvantages of this novel symptom-based ap-proach in this group of children. HIV-uninfected chil-dren remain an important group, because they consti-tute the majority of tuberculosis cases, even in settingswhere most adults with tuberculosis are HIV infected.3,27

The study emphasizes the importance of ascertaining achild’s HIV status before symptom-based diagnosis isattempted.

Because children were recruited at both the clinic andhospital level, hospital-based recruitment may have in-troduced some selection bias. However, the only signif-icant difference between the 2 groups was the propor-tion of HIV-infected children. Pulmonary tuberculosiswas diagnosed with different levels of certainty, but

there were no significant differences between the symp-toms recorded in children with bacteriologically con-firmed tuberculosis versus those with radiologically cer-tain tuberculosis or probable tuberculosis. Isolatedpleural effusion was excluded from the case definition,because it represents a different pathologic mechanismand has a different prognosis. All 17 of the children withisolated pleural effusion presented with a distinct clinicalpicture: localized unilateral chest pain, intermittent fe-ver, the absence of acute illness or respiratory distress,and the presence of extensive unilateral dullness onpercussion.

The diagnostic accuracy achieved in this highly en-demic setting, using a combination of symptoms at pre-sentation and/or follow-up, may be much reduced innonendemic areas where the pretest probability of tu-berculosis is lower. In addition, all of the assessment,evaluation, and follow-up were performed by the inves-tigator, and it is presumed that this was done in a morerigorous manner than would be achieved at primaryhealth care clinics in general. It would be helpful if thediagnostic value of this symptom-based approach wasconfirmed in a different setting.

CONCLUSIONSPulmonary tuberculosis can be diagnosed with a reason-able degree of accuracy in HIV-uninfected children (witha high degree of accuracy in the low-risk group), using asimple symptom-based approach. This offers the excitingprospect of improving access to antituberculosis treat-ment for children in resource-limited settings.

ACKNOWLEDGMENTSThe research was funded by the Medical Research Coun-cil of South Africa and the US Agency for Aid andInternational Development. The study was done in par-tial fulfillment of a PhD thesis.

We thank the primary health care clinics involved,the City of Cape Town Health Department, and, in par-ticular, the patients and their parents for their participa-tion.

REFERENCES1. Walls T, Shingadia D. Global epidemiology of paediatric tuber-

culosis. J Infect. 2004;48:13–222. Nelson LJ, Wells CD. Global epidemiology of childhood tuber-

culosis. Int J Tuberc Lung Dis. 2004;8:636–6473. Chintu C, Mudenda V, Lucas S, et al. Lung diseases at necropsy

in African children dying from respiratory illnesses: a descrip-tive necropsy study. Lancet. 2002;360:985–990

4. Starke JR. Pediatric tuberculosis: time for a new approach.Tuberculosis. 2003;83:208–212

5. Osborne CM. The challenge of diagnosing childhood tubercu-losis in a developing country. Arch Dis Child. 1995;72:369–374

6. Eamranond P, Jaramillo E. Tuberculosis in children: reassess-ing the need for improved diagnosis in global control strategies.Int J Tuberc Lung Dis. 2001;5:594–603

7. Zar HJ, Hanslo D, Apolles P, Swingler G, Hussey G. Induced

e1358 MARAIS et al by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

sputum versus gastric lavage for microbiological confirmationof pulmonary tuberculosis in infants and young children: aprospective study. Lancet. 2005;365:130–134

8. Weismuller MM, Graham SM, Claessens NJ, Meijnen S, Sala-niponi FM, Harries AD. Diagnosis of childhood tuberculosis inMalawi: an audit of hospital practice. Int J Tuberc Lung Dis.2002;5:432–438

9. Theart AC, Marais BJ, Gie RP, Hesseling AC, Beyers N. Criteriaused for the diagnosis of childhood tuberculosis at primaryhealth care level in a high-burden, urban setting. Int J TubercLung Dis. 2005;9:1210–1214

10. Du Toit G, Swingler G, Iloni K. Observer variation in detectinglymphadenopathy on chest radiography. Int J Tuberc Lung Dis.2002;6:814–817

11. Hesseling AC, Schaaf HS, Gie RP, Starke JR, Beyers N. A criticalreview of diagnostic approaches used in the diagnosis of child-hood tuberculosis. Int J Tuberc Lung Dis. 2002;6:1038–1045

12. Salazar GE, Schmitz TL, Cama R, et al. Pulmonary tuberculosisin children in a developing country. Pediatrics. 2001;108:448–453

13. Schaaf HS, Beyers N, Gie RP, et al. Respiratory tuberculosis inchildhood: the diagnostic value of clinical features and specialinvestigations. Pediatr Infect Dis J. 1995;14:189–194.

14. Marais BJ, Obihara CC, Gie RP, et al. The prevalence of symp-toms associated with pulmonary tuberculosis in randomly se-lected children from a high-burden community. Arch Dis Child.2005;90:1166–1177

15. Marais BJ, Gie RP, Schaaf HS, et al. The natural history ofchildhood intra-thoracic tuberculosis: a critical review of thepre-chemotherapy literature. Int J Tuberc Lung. Dis. 2004;8:392–402

16. Marais BJ, Gie RP, Obihara CC, Hesseling AC, Schaaf HS,Beyers N. Well-defined symptoms have improved value in thediagnosis of childhood pulmonary tuberculosis. Arch Dis Child.2005;90:1162–1165

17. Marais BJ, Hesseling AC, Gie RP, Schaaf HS, Beyers N. Theburden of childhood tuberculosis and the accuracy of routinesurveillance data in a high-burden setting. Int J Tuberc Lung Dis.2006;10:259–263

18. Marais BJ, Gie RP, Schaaf HS, et al. A proposed radiologicclassification of childhood intra-thoracic tuberculosis. PediatrRadiol. 2004;33:886–894

19. de Wit D, Steyn L, Shoemaker S, Sogin M. Direct determina-tion of Mycobacterium tuberculosis in clinical specimens by DNAamplification. Clin Microbiol. 1990;28:2437–2441

20. Obihara CC, Kimpen JLL, Gie RP, et al. Mycobacterium tuber-culosis infection may protect against allergy in a tuberculosisendemic area. Clin Exp Allergy. 2006;36:70–76.

21. Madhi SA, Gray GE, Huebner RE, et al. Correlation betweenCD4� lymphocyte counts, concurrent antigen skin test andtuberculin skin test reactivity in human immunodeficiencyvirus type 1-infected and uninfected children with tuberculo-sis. Pediatr Infect Dis J. 18:800–805.

22. Liebeschuetz, Bamber S, Ewer K ,et al. Diagnosis of tubercu-losis in South African children with a T-cell-based assay: Aprospective cohort study. Lancet. 2004;364:2196–2203

23. Dheda K, Lalvani A, Miller RF, et al. Performance of a T-cell-based diagnostic test for tuberculosis infection in HIV-infectedindividuals is independent of CD4 cell count. AIDS. 2005;18:2038–2041

24. Marais BJ, Gie RP, Schaaf HS, Donald PR, Beyers N, StarkeJ. Childhood pulmonary tuberculosis: old wisdom and newchallenges. Am J Resp Crit Care Med. 2006;173:1078–1090

25. World Health Organization. Tuberculosis in children. In: Treat-ment of Tuberculosis; Guidelines for National Tuberculosis Pro-grammes. 3rd ed. Geneva, Switzerland: World Health Organi-zation; 2003

26. Marais BJ, Donald PR, Gie RP, Schaaf HS, Beyers N. Diversityof disease manifestations in childhood pulmonary tuberculosis.Ann Trop Paed. 2005;25:79–86

27. Chintu C, Bhat G, Luo C, et al. Seroprevalence of humanimmunodeficiency virus type 1 infection in Zambian childrenwith tuberculosis. Pediatr Infect Dis J. 1993:499–504

28. National Center for Health Statistics in collaboration with theNational Center for Chronic Disease Prevention and HealthPromotion. Growth charts (2000). Available at: www.cdc.gov/growthcharts. Accessed September 5, 2006

PEDIATRICS Volume 118, Number 5, November 2006 e1359 by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from

DOI: 10.1542/peds.2006-0519 2006;118;e1350Pediatrics

Donald A. Enarson and Nulda BeyersBen J. Marais, Robert P. Gie, Anneke C. Hesseling, H. Simon Schaaf, Carl Lombard,

ChildrenA Refined Symptom-Based Approach to Diagnose Pulmonary Tuberculosis in

  

ServicesUpdated Information &

htmlhttp://pediatrics.aappublications.org/content/118/5/e1350.full.including high resolution figures, can be found at:

References

html#ref-list-1http://pediatrics.aappublications.org/content/118/5/e1350.full.at:This article cites 24 articles, 5 of which can be accessed free

Citations

html#related-urlshttp://pediatrics.aappublications.org/content/118/5/e1350.full.This article has been cited by 27 HighWire-hosted articles:

Subspecialty Collections

diseasehttp://pediatrics.aappublications.org/cgi/collection/infectious_Infectious Disease & Immunitythe following collection(s):This article, along with others on similar topics, appears in

Permissions & Licensing

mlhttp://pediatrics.aappublications.org/site/misc/Permissions.xhttables) or in its entirety can be found online at: Information about reproducing this article in parts (figures,

Reprints http://pediatrics.aappublications.org/site/misc/reprints.xhtml

Information about ordering reprints can be found online:

rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.Grove Village, Illinois, 60007. Copyright © 2006 by the American Academy of Pediatrics. All and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elkpublication, it has been published continuously since 1948. PEDIATRICS is owned, published, PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly

by guest on June 1, 2013pediatrics.aappublications.orgDownloaded from