INT J TUBERC LUNG DIS 14(3):324–331 © 2010 The Union
Association of chest radiographic abnormalities with tuberculosis disease in asymptomatic HIV-infected adults T. Agizew,* M. A. Bachhuber,* S. Nyirenda,* V. Z. S. A. M. Makwaruzi,† Z. Tedla,* R. J. Tallaksen,‡ J. E. Parker,‡ J. J. Mboya,§ T. Samandari*¶ * BOTUSA, Gaborone, † Nyangabgwe Referral Hospital, Francistown, Botswana; ‡ School of Medicine, West Virginia University, Morgantown, West Virginia, USA; § Disease Control Unit, Ministry of Health, Gaborone, Botswana; ¶ Division of Tuberculosis Elimination, Centers for Disease Control and Prevention, Atlanta, Georgia, USA SUMMARY
Francistown and Gaborone, Botswana. Chest radiography is used to screen for tuberculosis (TB) in asymptomatic persons living with the human immunodeficiency virus (PLWH) seeking isoniazid preventive therapy (IPT). We describe radiographic features in PLWH in a TB-endemic setting and identify features associated with TB disease. D E S I G N : Asymptomatic PLWH seeking IPT under program conditions for a clinical trial between 2004 and 2006 received chest radiographs (CXRs) that were read using the standardized Chest Radiograph Reading and Recording System (CRRS). Clinical characteristics, including TB disease, were compared with the radiographic findings. R E S U LT S : From 2732 screening CXRs, 183 had one or
more abnormalities and were scored using CRRS, with 42% having infiltrates (36% upper lobes), 35% parenchymal fibrosis and 32% adenopathy. TB disease status was determined in 129 (70%) PLWH, of whom 22 (17%) had TB disease. TB disease was associated with upper lobe infiltrates (relative risk [RR] 3.0, 95%CI 1.5–6.2) and mediastinal adenopathy (RR 3.9, 95%CI 1.8–8.4). The sensitivity and specificity of either upper lobe infiltrates or mediastinal lymphadenopathy for TB disease were respectively 64% and 82%. C O N C L U S I O N : A combination of CXR features was useful for predicting TB disease in asymptomatic PLWH. CRRS should be used more frequently in similar studies. K E Y W O R D S : tuberculosis; HIV; screening; chest radiograph
THE CHEST RADIOGRAPH (CXR) has long been utilized in the screening and diagnosis of tuberculosis (TB), but the current human immunodeficiency virus (HIV) epidemic complicates its use. The World Health Organization recommends expansion of programs for TB prevention in persons living with HIV infection (PLWH),1 and an essential part of such programs is screening for TB disease to avoid providing singledrug therapy to persons with TB disease. It remains controversial whether CXRs should be required to exclude TB disease in asymptomatic individuals, as radiography may be costly for resource-constrained countries and some studies show the yield to be low.2–5 If CXRs are routinely used in screening asymptomatic PLWH, as advocated by some,3,6 correlating radiographic patterns with active TB disease is a valuable undertaking. Radiographic features of TB in PLWH are frequently ‘atypical’ (i.e., mid or lower zone infiltrates, adenopathy, effusions and a paucity of cavitary lesions), as opposed to ‘typical’ (i.e., upper lobe infil-
trates and cavitation),7–10 a phenomenon that varies with the degree of immunosuppression.11–15 PLWH are also vulnerable to a number of non-tuberculous acute and chronic pulmonary diseases that may manifest as atypical radiographic features.16 TB screening studies in sub-Saharan African settings report a TB disease rate of 5–15% among PLWH; however, these studies were reported from mixed populations of symptomatic and asymptomatic PLWH.2,3,17,18 Three previous studies in the United States assessed screening CXRs in asymptomatic PLWH,19–21 but only one documented the spectrum of abnormalities seen.20 To the authors’ knowledge, no study has described CXR findings in an asymptomatic HIV-infected African population and correlated findings with TB diagnosis. To address this need, screening CXRs and other clinical information were obtained in a large cohort of asymptomatic PLWH during screening for a programmatically based clinical trial of isoniazid preventive therapy (IPT) in Botswana, where TB and HIV are endemic. The CXRs were ex-
Correspondence to: Taraz Samandari, Division of Tuberculosis Elimination, Centers for Disease Control and Prevention, 1600 Clifton Rd N E, Mailstop E-10, Atlanta, GA 30333, USA. Tel: (+1) 404 639 6014. Fax: (+1) 404 639 1566. e-mail: [email protected]
Article submitted 24 April 2009. Final version accepted 6 October 2009.
CXR in asymptomatic TB-HIV patients
amined in detail using the standardized Chest Radiograph Reading and Recording System (CRRS),22 and then analyzed by TB status and history of TB.
STUDY POPULATION AND METHODS PLWH seeking IPT through the National IPT Program who were interested in the IPT Trial were referred to research nurses located at eight local clinics in the two major cities of Gaborone and Francistown. Nearly all candidates referred for evaluation were recently diagnosed as HIV-infected and were receiving their initial assessment for HIV care services. As per the screening procedures of the National IPT Program, the first stage of screening excluded candidates if they had no documentation of HIV infection, illness (e.g., fever, cough, malaise, weight loss or, on physical examination, lymphadenopathy, jaundice or pulmonary disease), TB disease in the past 3 years, or clinical AIDS.23 Details of the screening results are published elsewhere.24 CXRs were obtained under public health program conditions for all asymptomatic candidates not excluded at the first stage of screening. CXRs were interpreted by study physicians and later independently reviewed by one of two expert readers, a pulmonary specialist (JEP) or a radiologist (RJT). The expert readers, whose interpretations were considered final for this analysis, were blinded to clinical variables, including TB status and history of TB. A CXR was determined to be normal or abnormal if either expert reader agreed with the study physicians’ interpretation. If one expert disagreed with the study physician, the other expert read the film to break the tie. CXRs were categorized as potentially compatible with TB if they had any of the following findings: cavities, fibrosis, infiltrates, nodules, pleural effusion or thickening, miliary disease, hilar or mediastinal adenopathy or pericardial effusion. CXRs with features potentially compatible with TB were then scored using the CRRS. PLWH with abnormal CXRs were further evaluated for TB through a follow-up evaluation, physical examination, a repeat CXR as necessary, and, where possible, sputum smears and cultures. Individuals were categorized as not having TB disease (status ‘Not TB’) if a repeat CXR became normal at the time of the follow-up interview without anti-tuberculosis treatment, if the person had no symptoms of TB for ⩾6 months and never received IPT, had no symptoms of TB for ⩾6 months after stopping IPT, or had an alternative diagnosis to explain CXR abnormalities. A TB case (status ‘TB’) was a candidate with an abnormal CXR and a positive TB culture, positive smear for acid-fast bacilli, a clinical response to anti-tuberculosis treatment, or initiation of anti-tuberculosis treatment with no alternative diagnosis. The standard methods used to process smears and
cultures of sputum and blood specimens are cited elsewhere.25 Tuberculin skin tests (TST) were performed using 5 tuberculin units (TU) of RT23 tuberculin (Statens Serum Institut, Copenhagen, Denmark), and were read by study nurses within 48–72 h. Data were collected using standardized case report forms and double-entered into a Clindex database (Fortress Medical Systems, Minneapolis, MN, USA). Two-tailed Fisher’s exact test for P and relative risks comparing proportions were performed using SAS (SAS Institute, version 9.1, Cary, NC, USA). P values < 0.05 were considered significant. The protocol was approved by the Health Research Development Committee in Botswana and the Institutional Review Board at the Centers for Disease Control and Prevention, Atlanta, GA, USA, and written informed consent was obtained from all participants. The clinical trial registration number is NCT00164281 and may be viewed at http://www.clinicaltrials.gov.
RESULTS Between November 2004 and June 2006, 4331 PLWH were referred for participation in the IPT trial; 37% (n = 1286) were excluded or declined participation in the clinical trial (n = 313) during the first stage of screening (Figure). Most candidates who were excluded had symptomatic illness (57%) and were referred to local clinicians for further evaluation. CXRs were obtained on 2732 asymptomatic PLWH; of these, 11% (302/2732) had abnormal CXRs potentially compatible with TB. The overall kappa statistic for agreement between study physicians and expert readers in assignment of normal/abnormal was 0.92, given the very large number of normal CXRs (data not shown). However, 29% (86/295, study physician interpretation was unavailable for seven) of the abnormal CXRs were initially read as normal by study physicians and were later determined to be abnormal by both expert readers. Details of the outcomes of these screened candidates have been published elsewhere.25 In the early period of the study, CXRs that were abnormal were returned to PLWH to assist them in seeking medical care. Thus, 64% (194/302) of abnormal films were available for scoring using the CRRS. Eleven films (6%) were re-read as normal using the CRRS and were excluded, yielding 183 CRRS forms for analysis. CXR quality was recorded on the CRRS for 98% (180/183) of these radiographs: respectively 71%, 28% and 1% were of high, acceptable and poor but interpretable quality. Of those with an abnormal CXR, candidates whose CXR was scored using the CRRS were more likely to have a positive TST than those whose CXR was not scored using the CRRS (respectively 43% [79/183] vs. 30% [36/119], P = 0.029). No other significant differences were found in baseline characteristics between these groups (data not shown).
The International Journal of Tuberculosis and Lung Disease
Table 1 Characteristics of asymptomatic HIV-infected adults with abnormal chest radiographs, Botswana (n = 183) Characteristic
Demographics (n = 183) Female 110 (60) Age, years, median [range] 34 [18–67] Clinical Body mass index