Embed Size (px)
Citation preview
CHEST
A high-resolution computed tomography-based scoring systemto differentiate the most infectious active pulmonary tuberculosisfrom community-acquired pneumonia in elderlyand non-elderly patients
Jun-Jun Yeh & Solomon Chih-Cheng Chen & Cheng-Ren Chen &
Ting-Chun Yeh & Hsin-Kai Lin & Jia-Bin Hong & Bing-Tsang Wu & Ming-Ting Wu
Received: 17 December 2013 /Revised: 23 April 2014 /Accepted: 6 June 2014 /Published online: 28 June 2014# European Society of Radiology 2014
AbstractObjective The objective of this study was to use high-resolution computed tomography (HRCT) imaging to predictthe presence of smear-positive active pulmonary tuberculosis(PTB) in elderly (at least 65 years of age) and non-elderlypatients (18–65 years of age).Methods Patients with active pulmonary infections seen fromNovember 2010 through December 2011 received HRCTchest imaging, sputum smears for acid-fast bacilli and sputumcultures forMycobacterium tuberculosis. Smear-positive PTBwas defined as at least one positive sputum smear and apositive culture for M. tuberculosis. Multivariate logistic re-gression analyses were performed to determine the HRCTpredictors of smear-positive active PTB, and a prediction
score was developed on the basis of receiver operating char-acteristic curve analysis.Results Of 1,255 patients included, 139 were diagnosed withsmear-positive active PTB. According to ROC curve analysis,the sensitivity, specificity, positive predictive value, negativepredictive value, false positive rates and false negative rateswere 98.6 %, 95.8 %, 78.5 %, 99.8 %, 4.2 % and 1.4 %,respectively, for diagnosing smear-positive active PTB inelderly patients, and 100.0 %, 96.9 %, 76.5 %, 100.0 %,3.1 % and 0.0 %, respectively, for non-elderly patients.Conclusions HRCT can assist in the early diagnosis ofthe most infectious active PTB, thereby preventingtransmission and minimizing unnecessary immediate re-spiratory isolation.
Electronic supplementary material The online version of this article(doi:10.1007/s00330-014-3279-6) contains supplementary material,which is available to authorized users.
J.<J. Yeh (*)Section of Thoracic Imaging, Department of Chest Medicine andFamily Medicine, Ditmanson Medical Foundation Chia-Yi ChristianHospital, No. 539, Zhongxiao Rd, Chiayi City, Taiwan 600,Republic of Chinae-mail: [email protected]: [email protected]
S. C.<C. Chen : C.<R. ChenDepartment of Medical Research, Ditmanson Medical FoundationChia-Yi Christian Hospital, No. 539, Zhongxiao Rd, Chiayi City,Taiwan 600, Republic of China
T.<C. Yeh :H.<K. Lin : J.<B. Hong : B.<T. WuDepartment of Family Medicine, Ditmanson Medical FoundationChia-Yi Christian Hospital, No. 539, Zhongxiao Rd, Chiayi City,Taiwan 600, Republic of China
J.<J. YehChia Nan University of Pharmacy and Science, Tainan, Taiwan,Republic of China
J.<J. YehMeiho University, Pingtung, Taiwan, Republic of China
J.<J. YehPingtung Christian Hospital, Pingtung, Taiwan, Republic of China
M.<T. Wu (*)Section of Thoracic and Circulation Imaging, Department ofRadiology, Kaohsiung Veterans General Hospital, Kaohsiung,Taiwan 813-62, Republic of Chinae-mail: [email protected]
M.<T. WuFaculty of Medicine, School of Medicine, National Yang MingUniversity, Taipei, Republic of China
Eur Radiol (2014) 24:2372–2384DOI 10.1007/s00330-014-3279-6
Key Points• HRCT can assist in the early diagnosis of the infectiousactive PTB
• HRCT imaging is useful to predict the presence of smear-positive active PTB
• Predictions from the HRCT imaging are valid even beforesputum smear or culture results
Keywords High-resolution CT . Patient isolation . Sputumsmear . Tuberculosis
AbbreviationsAFB Acid-fast bacilliAUC Area under the curveCAP Community-acquired pneumoniaCI Confidence intervalCXR Chest x-rayED Emergency departmentFNR False negative rateEPR False positive rateHRCT High-resolution computed tomographyHU Hounsfield unitMDCT Multi-detector CTMDR Multi-drug resistanceNAA Nucleic acid amplificationNPV Negative predictive valueOPD Outpatient departmentOR Odds ratioPACS Picture archiving and communication systemPPV Positive predictive valueROC Receiver-operating characteristicSD Standard deviationSE Standard errorTB Tuberculosis
Introduction
A positive sputum smear predicts a poor outcome in patientswith active pulmonary tuberculosis (PTB), and is a majordeterminant as to whether bacilli will be transmitted to acontact [1]. In addition, infectiousness of multi-drug resis-tance (MDR) tuberculosis is associated with acid-fast bacilli(AFB) sputum smear-positivity [2, 3]. The incidence ofsmear-positive active PTB in elderly men with atypical radio-logic findings at community hospitals has increased in recentyears [4]. Active PTB (73.7 % with AFB sputum smear-positivity), which can masquerade as community-acquiredpneumonia (CAP) and escape initial diagnosis by emergencydepartment (ED) or outpatient department (OPD) physicians,increases the risk of nosocomial infection and further trans-mission to the general population [5]. The early detection of
smear-positive active PTB is the best method of preventingtransmission in-hospital and in the community [3].
Nucleic acid amplification (NAA) assays provide a reliablemethod of increasing the specificity of diagnosis (ruling indisease), but the sensitivity is too poor to rule out disease [6].Furthermore, if the patient is not able to provide a sputumsample, NAA cannot be performed. Studies by Hsieh et al. [5]and Hatipoglu et al. [7] indicated that in only 37.7% and 53%of cases with suspected PTB, respectively, were sputum sam-ples obtained from spontaneous cough and bronchial wash-ings. While chest x-ray (CXR) images are frequently used inlow-resource settings to diagnose PTB, their sensitivity andspecificity are not high [8]. Pinto et al. [9] showed that using asimple radiographic scoring system can result in a sensitivityof 85.5 % and a specificity of 63.9 % for a diagnosis of PTB.
High-resolution computed tomography (HRCT) chest im-aging is useful for diagnosing PTB, and Kowada [10] reportedthat it was a cost-effective addition to the evaluation andmanagement of tuberculosis contacts in public health policy.However, aging can modify the appearance of PTB on CTimaging, and these changes are related to nutrition and im-mune status [4, 11–13].
The purpose of this study was to develop models based onHRCT imaging data to predict the presence of smear-positiveactive PTB in both elderly and non-elderly patients. Predictionof smear-positive active PTB may reduce the risk of transmis-sion as well as unnecessary immediate respiratory isolation.
Patients and methods
Patients
We initially reviewed the medical records of all patients ad-mitted to three teaching and branch hospitals from November2010 through December 2011 to identify those with activepulmonary infections. All patients were seen in the ED orOPD in these teaching hospitals, with about 50 % of thepatients transferred from other regional hospitals for the eval-uation the chest lesions. Initial inclusion criteria included thepresence of an infiltrate on CXR or symptoms or other clinicalmanifestation (such as cough, anaemia, fever and weakness)suspected to be PTB or CAP, but with equivocal lesions ornormal finding on CXR [14]. CAP was defined as the pres-ence of (a) one or more symptoms including body temperaturegreater than 38.5 °C, confusion, sweating and body aches; (b)two or more symptoms including increasing cough, sputumproduction, shortness of breath, wheeze and new focal ordiffuse signs on physical examinations; (c) CXR and/orHRCT consistent with an infection which was neither pre-existing nor of other known causes or (d) treatment withantibiotics for pneumonia by a pulmonologist or infectiousdisease specialist. Patients were treated by the ED or OPD
Eur Radiol (2014) 24:2372–2384 2373
physician, and subsequently followed by either a pulmonolo-gist or infectious disease specialist [15].
Patients with suspected PTB or CAPwhomet the inclusioncriteria received HRCTchest imaging, sputum spears for AFBand sputum cultures for Mycobacterium tuberculosis. Smear-positive PTB was defined as at least one positive sputumsmear and a positive culture for M. tuberculosis [16]. Themedical records of included patients were examined for theidentification of underlying disease such as anaemia, priorgastrectomy, diabetes mellitus, alcoholism and steroid usage.Elderly was defined as at least 65 years of age and non-elderlywas defined as 18–65 years of age. Each age group was thencategorized as those with smear-positive active PTB and thosewith other pulmonary infections. The grouping designationswere as follows: elderly with smear-positive active PTB, G1;elderly with other pulmonary infections, G2; non-elderly withsmear-positive active PTB, G3; non-elderly with other pul-monary infections, G4. This study was approved by the ethicsreview boards of the participating institutions, and all partic-ipants were fully informed of the nature of this study andprovided written informed consent.
Microbiological evaluation
Regular sampling included at least three sputum specimens,two or more blood cultures and paired serologic specimens (atadmission and between the 4th and 8th week thereafter).Additional diagnostic techniques utilized included pleuralpuncture, transbronchial aspirations, flexible bronchoscopywith a protected specimen brush and urine samples forLegionella species antigen [17].
Sputum samples were gram-stained, and sputum appropri-ate for analysis and originating from the lower respiratory tractwas defined as containing at least 25 granulocytes and lessthan 10 epithelial cells per low-power field. Validated sputumsamples, blood culture samples, pleural fluid and protectedspecimen brush samples were plated on sheep blood agar.Microorganism identification was performed according tostandard protocols [17].
HRCT imaging and evaluation
Chest CT (0.625 mm collimation, 100–120 kV, 250 mAs, tablespeed of 57.5 mm/s, rotation time of 0.75 s, pitch of 1.07) usinga 64-multi-detector CT (MDCT) system (Brilliance, PhilipsMedical Systems, Cleveland, OH, USA) was performed onincluded patients. Images were acquired during a singlebreath-hold lasting 5–8 s, thus making respiratory motion arte-facts very uncommon. The images were reconstructed with a 1-mm slice thickness in the axial plane (no gap), and in thecoronal plane (5 mm apart) using a high spatial frequencyalgorithm, and then sent to the picture archiving and commu-nication system (PACS) for review. All thin-section MDCT
images were displayed on a monitor at the pulmonary windowlevel setting (level, −600 HU; width, 1,200 HU).
The HRCT images were evaluated by two radiologists anda chest physician. Each specialist had more than 15 years ofexperience reading thoracic radiology studies, and was un-aware of the sputum smear and clinical examination results.Findings were reached by consensus. The HRCT images wereassessed for the presence and distribution of parenchymaabnormalities. The HRCT morphological patterns of PTBare illustrated in Figs. 1 and 2. The location of lung involve-ment was reported as one or more of 18 designated segmentsin the lungs. The kappa value of inter-observer and intra-observer variation in HRCT interpretation were both greaterthan 0.9, as in a prior study [18].
Statistical analyses
Statistical analyses were performed using SPSS 19.0 statisticssoftware (SPSS Inc., Chicago, IL, USA). Continuous datasuch as age, lag time of TB smear and lag time of TB culturewere presented as mean±standard deviation (SD), and thetwo-sample t test was used to compare the differences be-tween groups. Categorical data were presented as num-ber with percentage (%), and Pearson’s chi-square testor Fisher’s exact test was used to compare the differences inthe distribution of the categorical data between two groups(G1 and G2, G3 and G4).
Multivariate logistic regression analyses were performed todetermine the predictors of smear-positive active PTB.Potential predictors for smear-positive active PTB were iden-tified in univariate analysis, and those for which the p valuewas at most 0.10 were entered into multivariate models [19,20]. A backward elimination process was then used, andvariables with p<0.05 were used to derive an index-basedscoring system [20]. Multivariate logistic regression modelresults were presented as the estimated beta (β) with standarderror (SE) and odds ratio (OR) with 95 % confidence interval(CI). In addition, a relative score was determined by using theestimated β as a base, as described by Kanaya et al. [19]. Therelative score was defined as 1, 2 or 3 if the relative ratio as abase was 1, between 1 and 2, and greater than 2, respectively.If the variables were not included in the predictive scoresystem, they were assigned a score of 0. The area under thereceiver operating characteristic (ROC) curve was deter-mined, and the Youden index (sensitivity + specificity − 1)was used to indicate the best cut-off point. All statisticalanalyses were considered significant at p<0.05.
After summing up the score based on the relative score, ifpatients had a total score less than 5 in the elderly group or hada total score less than 2 in the non-elderly group, they wouldnot be isolated, would have blood and sputum cultures per-formed and would be treated with antibiotics. Patients with atotal score of at least 5 in the elderly group or a total score of at
2374 Eur Radiol (2014) 24:2372–2384
least 2 in the non-elderly group were placed in respiratoryisolation, at least three sputum specimens would be collectedat different times for AFB staining and M. tuberculosis cul-ture, and antituberculosis drugs would be initiated. Owing tothe interaction between TB prevalence and predictive ability,we utilized the pretest probability in situations with differ-ences in prevalence with a positive likelihood ratio in order toidentify the post-test probability [19, 21–23]. The post-testprobability of smear-positive PTB given its prevalence inthree different geographical settings was calculated using thescoring system [19, 21–23].
Results
From November 2010 through December 2011, 1,465 adultpatients (aged over 18) who met the inclusion criteria werescreened. After excluding the 210 patients, the remaining
1,255 patients suspected of CAP or active PTB receivedHRCT chest imaging and were included in the study. Aflowchart of patient enrolment is shown in Fig. 3.
A total of 139 subjects had at least one positive sputumsmear and a positive culture for M. tuberculosis (smear-posi-tive active PTB; G1, n=74; G3, n=65). The remaining 1,116patients had negative sputum smears (G2, n=478; G4, n=638). Patient characteristics are summarized in Table 1. Themajority of subjects in each study group were male. Co-morbid conditions were noted in the study subjects. Thedistribution of diabetes (p=0.003) differed significantly be-tween groups G1 and G2, and the distribution of
Fig. 1 A 63-year-old male with smear-positive active pulmonary tuber-culosis presented with cough, fever, weight loss and dyspnoea. The high-resolution computed tomography (HRCT) score was 2. a The chest x-rayshowed consolidation of the right upper and right middle lung fields(black arrowhead). b Coronal HRCT showed consolidation of the apicalsegment (s1, upper black arrowhead) and posterior segment (s2, middleblack arrowhead) in the right upper lobe (score = 2). Consolidation of thesuperior segment (s6, lower black arrowhead) in right lower lobe wasalso found
Fig. 2 A 46-year-old male with smear-positive active pulmonary tuber-culosis presented with cough, fever and weight loss. The high-resolutioncomputed tomography (HRCT) score was 4. a Chest x-ray showedconsolidation/mass with patchy lesions of both upper lobes (upper blackarrowhead, lower black arrow). b Transverse HRCT showed clusters ofnodules/mass with speculatedmargins and satellite nodules in the anteriorsegment (s3, black arrow) of both upper lobes (score = 3). c The softtissue window of HRCT showed paratracheal adenopathy (white arrow)and clusters of nodules/mass in the apical segment (s1, white arrowhead)in the right upper lobe and the apico-posterior segment (s1+s2, whitearrowhead) in the left upper lobe (score = 1)
Eur Radiol (2014) 24:2372–2384 2375
hypoalbuminemia (p=0.025) and alcoholism (p=0.045) dif-fered significantly between groups G3 and G4. Among clin-ical symptoms and signs, only cough was significantly differ-ent between groups G1 and G2 (p=0.003) and groups G3 andG4 (p=0.047).
Comparison of HRCT imaging findings between groups G1and G2 showed that consolidation of the apical segment orposterior segment of the right upper lobe, apico-posterior seg-ment of the left upper lobe, or superior segment of the rightlower lobe or left lower lobe; cavitation of the apical segment or
posterior segment of the right upper lobe, apico-posterior seg-ment of the left upper lobe, lateral segment of the right middlelobe, superior segment of the left lingual lobe, medial segmentof the right middle lobe, inferior segment of the left linguallobe, or superior segment of right lower lobe or left lower lobe;clusters of nodules/mass in any segment of right lower lobe orleft lower lobe, tree-in-bud, centri-lobular nodules, ground-glass opacity (GGO), and fibrosis and calcification were allgreater in group G1 (all, p<0.05). Comparison of HRCT im-aging findings between groups G3 and G4 showed that
Fig. 3 Flow chart of patient enrolment
2376 Eur Radiol (2014) 24:2372–2384
consolidation of the apical segment or posterior segment of theright upper lobe, apico-posterior segment of the left upper lobe,lateral segment of the right middle lobe, superior segment of theleft lingual lobe, medial segment of the right middle lobe,inferior segment of the left lingual lobe, or superior segmentof the right lower lobe or left lower lobe were all greater ingroup G3 (all, p<0.05). Consolidation of the medial basalsegment, anterior basal segment, lateral basal segment, poste-rior basal segment of the right lower lobe, or medial-anteriorbasal segment, lateral basal segment, or posterior basal segmentof left lower lobe were all greater in group G4, and cavitation ofthe apical segment or posterior segment of right upper lobe,apico-posterior segment of the left upper lobe, lateral segmentof the right middle lobe, superior segment of the left linguallobe, medial segment of the right middle lobe, inferior segmentof the left lingual lobe, or superior segment of the right lowerlobe or left lower lobe, and clusters of nodules/mass in anysegment of the right lower lobe or left lower lobe were greaterin group G3 (all, p<0.05). All HRCT findings are summarizedin Table 2.
Multivariate logistic regression analysis showed that con-solidation, cavitation and clusters of nodules of the apical
segment or posterior segment of the right upper lobe orapico-posterior segment of the left upper lobe were positiveindexes of groups G1 and G3 (Table 3). Consolidation of thesuperior segment of the right lower lobe or left lower lobe andtree-in-bud were positive indexes of group G1. Clusters ofnodules/mass of the anterior segment of the right upper lobe orleft upper lobe, lateral segment of the right middle lobe,superior segment of the left lingual lobe, medial segment ofthe right middle lobe or inferior segment of left lingual lobewere a positive index. In addition, consolidation of the medialbasal segment, anterior basal segment, lateral basal segment orposterior basal segment of the right lower lobe, or consolida-tion of the medial-anterior basal segment, lateral basal seg-ment or posterior basal segment of the left lower lobe was anegative index for non-elderly patients. Centri-lobular nod-ules, GGO, adenopathy, bronchial thickening, interlobularthickening, fibrosis and calcification were not indexes forpredicting smear-positive active PTB in the elderly and non-elderly groups. Thus, HRCT criteria were useful for differen-tiating group G1 from G2 and G3 from G4.
The five independent variables that were associated withgroups G1 and G3 were used to generate a prediction score to
Table 1 Demographic and clinical characteristics of elderly and non-elderly patients with smear-positive active pulmonary tuberculosis (PTB) and otherpulmonary infections without PTB
Elderly (n=552) Non-elderly (n=703)
G1 (n=74) G2 (n=478) p value G3 (n=65) G4 (n=638) p value
Age (years) 72.81±5.53 73.07±5.04 0.686† 43.25±12.35 47.00±10.84 0.007†
Gender (male) 52 (70.3) 306 (64.0) 0.294‡ 44 (67.7) 390 (61.1) 0.300‡
Underlying disease
Diabetes mellitus 23 (31.1) 80 (16.7) 0.003‡* 21 (32.3) 247 (38.7) 0.311‡
Steroid usage 20 (27.0) 125 (26.2) 0.873‡ 15 (23.1) 98 (15.4) 0.107‡
Anemia 42 (56.8) 218 (45.6) 0.074‡ 31 (47.7) 259 (40.6) 0.268‡
Hypoalbuminemia 44 (59.5) 226 (47.3) 0.051‡ 35 (53.9) 252 (39.5) 0.025‡*
Gastrectomy 9 (12.2) 51 (10.7) 0.701‡ 11 (16.9) 77 (12.1) 0.260‡
Uremia 12 (16.2) 93 (19.5) 0.509‡ 13 (20.0) 86 (13.5) 0.150‡
Alcoholism 31 (41.9) 215 (45.0) 0.619‡ 28 (43.1) 197 (30.9) 0.045‡*
Malignancy 8 (10.8) 41 (8.6) 0.530‡ 6 (9.2) 39 (6.1) 0.293¶
Previous PTB 12 (16.2) 49 (10.3) 0.128‡ 3 (4.6) 71 (11.1) 0.103‡
Symptoms and signs
Cough 47 (63.5) 378 (79.1) 0.003‡* 39 (60.0) 458 (71.8) 0.047‡*
Fever 39 (52.7) 304 (63.6) 0.072‡ 41 (63.1) 457 (71.6) 0.148‡
Body weight loss 58 (78.4) 354 (74.1) 0.427‡ 43 (66.2) 421 (66.0) 0.979‡
Weakness 57 (77.0) 327 (68.4) 0.134‡ 42 (64.6) 383 (60.0) 0.472‡
Data are presented as mean±standard deviation or number (percentage). Comparisons between groups were performed using † independent t test, ‡Chi-square test or ¶ Fisher’s exact test
G1 smear-positive active PTB in elderly group, G2 other pulmonary infection without smear-positive active PTB in elderly group, G3 smear-positiveactive PTB in non-elderly group, G4 other pulmonary infection without smear-positive active PTB in non-elderly group
*p<0.05
Eur Radiol (2014) 24:2372–2384 2377
differentiate between smear-positive active PTB and otherpulmonary infections in both elderly and non-elderly patients.The ROC curve analysis is shown in Fig. 4. The area under thecurve (AUC) was 0.983 (95 % CI 0.963–1.000) for the pre-diction model for elderly patients (Fig. 4a) and 0.993 (95%CI0.989–0.998) for the prediction model for non-elderly patients(Fig. 4b). The optimal cut-off value of the prediction score tobest differentiate smear-positive active PTB and other pulmo-nary infections (based on maximizing the Youden index) was5 in elderly patients (sensitivity 98.6 %, specificity 95.8 %),and 2 in non-elderly patients (sensitivity 100.0 %, specificity96.9 %).
At these optimal cut-off values of prediction score todifferentiate smear-positive active PTB (score at least 5 forelderly patients, score at least 2 for non-elderly patients)and other pulmonary infections (score less than 5 forelderly patients, score less than 2 for non-elderly patients),the sensitivity, specificity, positive predictive value (PPV),negative predictive value (NPV), false positive rates (FPR)and false negative rates (FNR) were 98.6 % (73/74),95.8 % (458/478), 78.5 % (73/93), 99.8 % (458/459),4.2 % (20/478) and 1.4 % (1/74), respectively, forpredicting smear-positive active PTB in elderly patients,and 100.0 % (65/65), 96.9 % (618/638), 76.5 % (65/85),
Table 2 High-resolution computed tomography (HRCT) findings of elderly and non-elderly patients with smear-positive active pulmonary tuberculosis(PTB) and other pulmonary infections without smear-positive active PTB
HRCT parameters Elderly (n=552) Non-elderly (n=703)
G1 (n=74) G2 (n=478) p value G3 (n=65) G4 (n=638) p value
Consolidation 71 (96.0) 335 (70.1) <0.001†* 61 (93.9) 548 (85.9) 0.073†
s1, s2, s1+s2 67 (90.5) 127 (26.6) <0.001†* 58 (89.2) 188 (29.5) <0.001†*
s3, s4, s5 24 (32.4) 116 (24.3) 0.133† 21 (32.3) 136 (21.3) 0.043†*
s6 58 (78.4) 71 (14.9) <0.001†* 26 (40.0) 64 (10.0) <0.001†*
s7, s8, s7+8, s9, s10 33 (44.6) 283 (59.2) 0.018†* 17 (26.2) 508 (79.6) <0.001†*
Cavitation 54 (73.0) 47 (9.8) <0.001†* 45 (69.2) 64 (10.0) <0.001†*
s1, s2, s1+2 52 (70.3) 17 (3.6) <0.001†* 43 (66.2) 12 (1.9) <0.001†*
s3, s4, s5 12 (16.2) 14 (2.9) <0.001‡* 10 (15.4) 8 (1.3) <0.001‡*
s6 33 (44.6) 3 (0.6) <0.001‡* 20 (30.8) 3 (0.5) <0.001‡*
s7, s8, s7+8, s9, s10 14 (18.9) 56 (11.7) 0.083† 10 (15.4) 61 (9.6) 0.138
Clusters nodules/mass 53 (71.6) 10 (2.1) <0.001†* 58 (89.2) 29 (4.6) <0.001†*
s1, s2, s1+s2 50 (67.6) 8 (1.7) <0.001†* 57 (87.7) 27 (4.2) <0.001†*
s3, s4, s5 9 (12.2) 2 (0.4) <0.001‡* 46 (70.8) 5 (0.8) <0.001‡*
s6 23 (31.1) 3 (0.6) <0.001‡* 22 (33.9) 3 (0.5) <0.001‡*
s7, s8, s7+8, s9, s10 12 (16.2) 1 (0.2) <0.001‡* 18 (27.7) 3 (0.5) <0.001‡*
Tree-in-bud 69 (93.2) 118 (24.7) <0.001†* 43 (66.2) 327 (51.3) 0.022†*
Centri-lobular nodules 59 (79.7) 271 (56.7) <0.001†* 49 (75.4) 384 (60.2) 0.016†*
Ground-glass opacitya 69 (93.2) 395 (82.6) 0.020†* 60 (92.3) 517 (81.2) 0.025†*
Adenopathyb 50 (67.6) 271 (56.7) 0.078† 47 (72.3) 344 (54.0) 0.005†*
Bronchial thickening 49 (66.2) 295 (61.7) 0.457† 55 (84.6) 445 (69.8) 0.012†*
Interlobular thickening 41 (55.4) 249 (52.1) 0.595† 38 (58.5) 303 (47.5) 0.092†
Fibrosis 21 (28.4) 81 (17.0) 0.018†* 1 (1.5) 50 (7.8) 0.075‡
Calcificationc 17 (23.3) 66 (13.8) 0.035†* 2 (3.1) 42 (6.6) 0.418‡
Data are presented as numbers with percentage (%). Comparison between groups was performed using †Chi-square test or ‡ Fisher’s exact test
G1 smear-positive active PTB in elderly group, G2 other pulmonary infection without smear-positive active PTB in elderly group, G3 smear-positiveactive PTB in non-elderly group,G4 other pulmonary infection without smear-positive active PTB in non-elderly group, s1 apical segment, s2 posteriorsegment of right upper lobe, s1+s2 apico-posterior segment of left upper lobe, s3 anterior segment of right or left upper lobe, s4 lateral segment of rightmiddle lobe or superior segment of left lingual lobe, s5medial segment of right middle lobe or inferior segment of left lingual lobe, s6 superior segmentof right or left lower lobe, s7 medical basal segment of right lower lobe, s8 anterior basal segment of right lower lobe, s7+s8 medial-anterior basalsegment of left lower lobe, s9 lateral basal segment of right or left lower lobe, s10 posterior basal segment of right or left lower lobe* p<0.05aOne missing value of ground-glass opacity in the non-elderly groupbOne missing value of adenopathy in the non-elderly groupcOne missing value of calcification in the elderly group
2378 Eur Radiol (2014) 24:2372–2384
100.0 % (618/618), 3.1 % (20/638) and 0.0 % (0/65),respectively, for non-elderly patients (Table 4).
According to the model, unnecessary immediate isolationwas reduced to 20 out of 478 patients in the elderly group and20 out of 368 patients in the non-elderly group. Meanwhile,we detect the unnecessary immediate isolation for 458 out of459 patients in the elderly group and 618 out of 618 patients inthe non-elderly group on scoring system (Table 4 andAppendix Table 1). The post-test probability was high (up togreater than 70 %) in this study. Furthermore, the post-testprobability in moderate-to-low and very low prevalence areaswas greater than 5 % (Appendix Table 2).
In order to interpret the differences of the two models forpredicting smear-positive active PTB in elderly and non-elderly patients, we compared the CT finding of groups G1and G3. We found that consolidation in the right lower lobe orleft lower lobe, tree-in-bud, fibrosis and calcification occurredwith higher frequency in groupG1. On the other hand, clustersof nodules of the right upper lobe or left upper lobe, or rightmiddle lobe or left lingual lobe, and bronchial thickeningoccurred with higher frequency in group G3 (all, p<0.05)(Table 5).
A summary of the scoring system use is shown inAppendix Table 3 and the distribution of scores in the elderlyand non-elderly groups is shown in Appendix Table 4. If an
elderly patient had consolidation of the apical segment orposterior segment of the right upper lobe or apico-posteriorsegment of the left upper lobe (score 2), consolidation of thesuperior segment of the right lower lobe or left lower lobe(score 1), cavitation of the apical segment or posterior seg-ment of the right upper lobe or apico-posterior segment of theleft upper lobe (score 2), tree-in-bud in any segment (score 2)andGGO in any segment (score 0), the patient would receive atotal score of 7 (2+1+2+2+0). If a non-elderly patienthad tree-in-bud (score 0) and GGO (score 0) in thesame segment or in different segments, the total scoreis 0 (0+0); because tree-in-bud and GGO are not in-dexes in the non-elderly group.
Among the 74 patients in the elderly group who had smear-positive active PTB, five patient received a total score of 0,five patients received a total score of 5, two patients received atotal score of 6, 33 patients received a total score of 7, sevenpatients received a total score of 8, and 26 patients received atotal score of 10. Among the 65 patients in the non-elderlygroup who had smear-positive active PTB, one patient re-ceived a total score of 2, 10 patients received a total score of3, six patients received a total score of 4, three patientsreceived a total score of 5, 14 patients received a total scoreof 6, 14 patients received a total score of 7, and 17 patientsreceived a total score of 9 (Appendix Table 1).
Table 3 Multivariate logistic regression analysis for identification of patients with smear-positive active pulmonary tuberculosis (PTB) using high-resolution computed tomography (HRCT) findings
HRCT parameters Estimated β (SE) OR (95 % CI) p value Relative score†
Elderly group
Consolidation s1, s2, s1+2 2.649 (0.735) 14.14 (3.35, 59.67) <0.001* 2
Consolidation s6 2.100 (0.835) 8.17 (1.59, 41.93) 0.012* 1
Cavitation s1, s2, s1+s2 2.431 (0.615) 11.38 (3.41, 37.97) <0.001* 2
Tree-in-bud 2.738 (0.692) 15.46 (3.98, 60.03) <0.001* 2
Clusters nodules/mass s1, s2, s1+s2 4.323 (0.825) 75.40 (14.95, 380.14) <0.001* 3
Non-elderly group
Consolidation s1, s2, s1+s2 2.597 (0.727) 13.43 (3.23, 55.82) <0.001* 2
Consolidation s7, s8, s7+8, s9, s10 −2.231 (0.747) 0.11 (0.03, 0.46) 0.003* −2Cavitation s1, s2, s1+s2 3.325 (0.698) 27.81 (7.08, 109.27) <0.001* 3
Clusters nodules/mass s1, s2, s1+s2 1.486 (0.754) 4.42 (1.01, 19.35) 0.049* 1
Clusters nodules/mass s3, s4, s5 3.068 (0.749) 21.50 (4.96, 93.31) <0.001* 3
s1 apical segment, s2 posterior segment of right upper lobe, s1+s2 apico-posterior segment of left upper lobe, s3 anterior segment of right or left upperlobe , s4 lateral segment of right middle lobe or superior segment of left lingual lobe, s5medial segment of right middle lobe or inferior segment of leftlingual lobe, s6 superior segment of right or left lower lobe, s7medical basal segment of right lower lobe, s8 anterior basal segment of right lower lobe,s7+s8 medial-anterior basal segment of left lower lobe, s9 lateral basal segment of right or left lower lobe, s10 posterior basal segment of right or leftlower lobe, OR odds ratio, CI confidence interval
*p<0.05† In the elderly group, relative score was determined according to the ratio of each estimated β value, using the estimated β of 2.1 from consolidation ofs6 as the base. The relative score given is 2 if the ratio is between 1 and 2, 3 if the ratio is greater than 2. In the non-elderly group, using the estimatedβ of1.486 from cluster nodules of s1, s2, s1+2 as the base, the relative score given is 2 if the ratio is between 1 and 2, 3 if the ratio is greater than 2. Since theeffect of consolidation of s7, s8, s7+8, s9, s10 is negative with a ratio between 1 and 2, the relative score is set as −2
Eur Radiol (2014) 24:2372–2384 2379
Discussion
The results of this study showed that using the models devel-oped on the basis of HRCT imaging data, it was possible topredict the presence of smear-positive active PTB in elderlyand non-elderly patients. The goal of developing the models inthis study is to allow for respiratory precautions upon hospitaladmission, for all potentially contagious patients with activePTB, while minimizing the isolation of patients without PTB.Had the decision to isolate patients been made using thisHRCT prediction model, 73 out of 74 smear-positive activePTB in the elderly group and 64 out of 65 patients with smear-positive active PTB in the non-elderly group would have beenidentified and received respiratory isolation. Furthermore,unnecessary immediate isolation of 458 out of 459 patientsin the elderly group and 618 out of 618 patients in the non-elderly group would have been prevented (Table 4).
Differences in HRCT findings between patients withsmear-positive active PTB and those with other pulmonaryinfections allow the prediction of smear-positive active PTB;as the two models indicate, there are differences betweenelderly and non-elderly patients. The first difference notedbetween the two models is that consolidation of the basalsegments of lower lobe (medial basal segment, anterior basalsegment, medial-anterior basal segment, lateral basal segment,posterior basal segment) was a negative index for group G3,whereas it was not for group G1. We found that the frequencyof consolidation in the basal segments of the right lower lobeor left lower lobe among elderly patients was significantlygreater than in non-elderly patients (44.6 % in group G1 and26.2 % in group G3). In a study of 78 non-elderly (youngadult) patients with active PTB, Wong et al. [24] reported thatin smear-positive patients the frequency of consolidation inthe basal segments was 36.7 %. The low frequency of consol-idation of the basal segments (26.2 %) of the right lower lobeor left lower lobe in group G3 in our study is compatible withtheir findings. The frequency of consolidation of the basalsegments of the right lower lobe or left lower lobe in group G3was lower than in group G4 (26.2 % vs. 79.6 %, respectively),which contributed to this finding being a negative index forG3. Progressive age-related changes in pulmonary tuberculo-sis imaging, such as increased lower lobe lesions, have beenreported in several studies [4, 12, 25, 26].
The second difference between the two models is thatconsolidation of the superior segment of the right lower lobeor the left lower lobe was a positive index for group G1, butnot a positive index for group G3. The frequency of consol-idation of the superior segment of the right lower lobe or leftlower lobe in group G1 was higher than that in group G2(78.4 % vs. 14.5 %, respectively, p<0.001), and this contrib-uted to consolidation of the superior segment of the rightlower lobe or left lower lobe being a positive index for groupG1. Furthermore, there was a significant difference in
consolidation of the superior segment of the right lower lobeand left lower lobe between groups G1 and G3 (78.4 % vs.40.0 %, respectively, p<0.01). Wong et al. [24] reported thatconsolidation of the superior segment of the right lower lobeor left lower lobe occurred in 46.7 % of non-elderly (youngadult) patients with smear-positive active PTB, which is com-patible with a value of 40 % in group G3 in our study.
The third difference between the two models is that clustersof nodules/mass in the anterior segment of the right upper lobeor left upper lobe, medial segment of the right middle lobe,superior segment of the left lingual lobe, lateral segment of theright middle lobe or inferior segment of the left lingual lobewere a positive index for group G3, and the frequency of
Fig. 4 Receiver operating characteristic (ROC) curve analysis of predic-tion score to discriminate between patients with smear-positive activepulmonary tuberculosis (PTB) and those with other pulmonary infectionsfor the a elderly group (n=552) and b the non-elderly group (n=703). Forthe elderly group, the optimal cut-off point was 5 with a sensitivity of98.6 % and specificity of 95.8 %. For the non-elderly group, the optimalcut-off point was 2 with sensitivity of 100.0 % and specificity of 96.9 %.AUC area under the ROC curve, CI confidence interval
2380 Eur Radiol (2014) 24:2372–2384
clusters of nodules in the anterior segment of the right upperlobe or left upper lobe, medial segment of the right middlelobe, superior segment of the left lingual lobe, or lateralsegment of the right middle lobe, or inferior segment of theleft lingual lobe was greater in group G3 than G4 (70.8 % vs.0.8 %). Furthermore, a comparison of groups G1 and G3showed that there was a significant difference in the frequencyof clusters of nodules/mass in the anterior segment of the rightupper lobe or left upper lobe, medial segment of the rightmiddle lobe, superior segment of the left lingual lobe, lateralsegment of the right middle lobe or inferior segment of left thelingual lobe between the two groups (12.2 % vs. 70.8 %,respectively, p<0.001).
The fourth difference between the two models is that tree-in-bud was a positive index for group G1. The frequency oftree-in-bud was greater in group G1 than G2 (93.2 % vs.24.7 %, respectively, p<0.001) and greater in group G3 thanG4 (66.2 % vs. 51.3 %, respectively, p=0.22). Furthermore,the frequency of tree-in-bud in groups G1 and G3 was 93.2 %and 66.2 %, respectively (p<0.001). The high frequency oftree-in-bud in elderly patients with smear-positive active PTBis consistent with the results of a recent study by Okamuraet al. [13]. Studies have shown that tree-in-bud appearance islow in cases of smear-negative active PTB [27] and myco-plasma infection [28], and is rarely seen in cases of bacterialpneumonia [29].
Our results showed that consolidation in the apical segmentor posterior segment of the right upper lobe or apico-posteriorsegment of the left upper lobe occurred with greater frequencyin group G1 than G2 (90.5 % vs. 26.6 %, respectively,p<0.001), and with greater frequency in group G3 than G4(89.2 % vs. 29.5 %, respectively, p<0.001). This finding is in
accord with those of a prior study by Lee et al. [30]. Anotherstudy has shown that the frequency of consolidation in theapical segment or posterior segment of the right upper lobe orapico-posterior segment of the left upper lobe is lower insmear-negative than in smear-positive patients [18]. In addi-tion, the frequency of consolidation of the apical segment orposterior segment of the right upper lobe or apico-posteriorsegment of the left upper lobe is lower in pulmonary infectionwithout smear-positive active PTB [27, 30, 31]. We found nosignificant difference in the frequency of consolidation ofbetween groups G1 and G3 (90.5 % vs. 89.2 %, respectively,p=0.798).
Cavitation in the apical segment or posterior segment of theright upper lobe or apico-posterior segment of the left upperlobe occurred with higher frequency in group G1 than G2(73.0 % vs. 3.6 %, respectively, p<0.001), and with higherfrequency in group G3 than G4 (66.2 % vs. 1.9 %, respec-tively, p<0.001). Prior studies have shown that the frequencyof cavitation in the apical segment or posterior segment of theright upper lobe or apico-posterior segment of the left upperlobe in smear-negative PTB and in non-TB pulmonaryinfections is lower than in smear-positive PTB [18, 27,30, 31]. Our results also showed that there was nosignificant difference in the frequency of cavitation ofthe apical segment or posterior segment of the rightupper lobe or apico-posterior segment of the left upperlobe between groups G1 and G3 (70.3 % vs. 66.2 %,respectively, p=0.603). This finding is different fromthat of Korzeniewska-Kosela et al. [12] who reported ahigher frequency of cavitation in non-elderly PTB pa-tients than in elderly PTB patients. The difference infindings between our study and that of Korzeniewska-Kosela
Table 4 Predictive ability of the prediction score for elderly group and non-elderly patients
Elderly (n=552) Non-elderly (n=703)
G1 (n=74) G2 (n=478) G3 (n=65) G4 (n=638)
Predictive results†
Smear-positive active PTB 73 20 65 20
Pulmonary infection without smear-positive PTB 1 458 0 618
Diagnostic performance
Sensitivity 98.6 % (73/74) 100.0 % (65/65)
Specificity 95.8 % (458/478) 96.9 % (618/638)
PPV 78.5 % (73/93) 76.5 % (65/85)
NPV 99.8 % (458/459) 100.0 % (618/618)
FPR 4.2 % (20/478) 3.1 % (20/638)
FNR 1.4 % (1/74) 0.0 % (0/65)
G1 smear-positive active PTB in elderly group, G2 other pulmonary infection without smear-positive active PTB in elderly group, G3 smear-positiveactive PTB in non-elderly group, G4 other pulmonary infection without smear-positive active PTB in non-elderly group, PPV positive predictive value,NPV negative predictive value, FPR false positive rate (1 − specificty), FNR false negative rate (1 − sensitivity)†The cut-off value of the prediction score classified patients as smear-positive active PTB (score at least 5 for elderly, score at least 2 for non-elderly) orother pulmonary infection without smear-positive active PTB (score less than 5 for elderly and score less than 2 for non-elderly)
Eur Radiol (2014) 24:2372–2384 2381
et al. is likely because HRCT may early detect small or occultcavitation [18].
Clusters of nodules/mass in the apical segment or posteriorsegment of the right upper lobe or apico-posterior segment ofthe left upper lobe occurred with higher frequency in groupG1 than G2 (67.6 % vs. 1.7 %, respectively, p<0.001), andwith higher frequency in group G3 than in G4 (87.7 % vs.4.2 %, respectively, p<0.001), a finding that is in accordancewith a prior study [31]. Studies have shown that clusters ofnodules are less frequent in smear-negative active PTB [27,31] and in pulmonary infection without active PTB [32] thanin smear-positive active PTB [18].
An ideal test would have both 100 % sensitivity to avoidmissing any patients with active disease and 100 % NPV toensure that negative results truly represented a disease-freestatus. This is extremely important in public health situationsthat deal with highly infectious diseases like smear-positiveactive PTB. Our study demonstrates the effective use ofHRCT to predict smear-positive active PTB in elderly(98.6 % sensitivity and 99.8 % NPV) and non-elderly(100 % sensitivity and 100 % NPV) and pulmonary infectionwithout smear-positive active PTB in advance of serial spu-tum results or in cases without sputum. Using HRCT resultscan lead to appropriate respiratory isolation and presumablydecrease the transmission rate. We tested this model in high tovery low prevalence areas, and the post-test probability wasgreater than 5 %; thus, providing broad-spectrum antitubercu-losis medications is acceptable [19, 22, 23]. The smear-positive active PTB prediction score was created with wideranging likelihood ratios that could affect the post-test proba-bility of smear-positive active PTB by up to 23-fold and 35-fold in the elderly and non-elderly groups, respectively [19,21, 22]. This scoring system model may be applicable inregions with high to very low prevalence, but validation inother areas will be necessary.
There are a number of unique points of this study, notaddressed in other works [33, 34], that should be considered:(1) The data were from Taiwan, where PTB is endemic;however, the healthcare system is relatively well establishedcompared to that in previous reported areas. (2) Our analysisincludes a model with differentiation of patients with smear-positive active PTB from CAP. (3) We analysed the differentCT findings between the elderly and non-elderly patients. (4)We tested the ability of this model in high to very lowprevalence areas, and the post-test probability was high.
There are a number of limitations of this study that shouldbe noted. We only performed validation of the reliability andreproducibility of the five variables at only one centre and itsbranches. The demographic and clinical characteristics ofelderly and non-elderly adult patients with PTB may differin other geographic areas, especially in areas with variabletuberculosis prevalence. Therefore, the model needs to bevalidated in other settings before it can be implemented in
clinical practice. Furthermore, the less infectious smear-negative active PTB also requires intervention, owing to italso being a source of infection in hospital. Elderly wasarbitrarily defined with an age cut-off of 65 years; however,this cut-off point has been used in prior reports [35–37].Lastly, HRCT is relatively expensive for poor areas, and isnot likely to be available or radiologists trained in the inter-pretation of images with respect to PTB may not be available.However, the method described is valuable for areas likeSouth Korea, Japan and certain major cities in China [10, 38].
Table 5 Comparison of high-resolution computed tomography (HRCT)findings of elderly and non-elderly patients with smear-positive activepulmonary tuberculosis (PTB)
HRCT parameters a G1, Elderly(n=74)
G3, Non-elderly(n=65)
p value
Consolidation 71 (96.0) 61 (93.9) 0.705†
s1, s2, s1+s2 67 (90.5) 58 (89.2) 0.798†
s3, s4, s5 24 (32.4) 21 (32.3) 0.988†
s6 58 (78.4) 26 (40.0) <0.001†
s7, s8, s7+8, s9, s10 33 (44.6) 17 (26.2) 0.024†
Cavitation 54 (73.0) 45 (69.2) 0.627†
s1, s2, s1+s2 52 (70.3) 43 (66.2) 0.603†
s3, s4, s5 12 (16.2) 10 (15.4) 0.893†
s6 33 (44.6) 20 (30.8) 0.094†
s7, s8, s7+8, s9, s10 14 (18.9) 10 (15.4) 0.582†
Clusters nodules/mass 53 (71.6) 58 (89.2) 0.010†
s1, s2, s1+s2 50 (67.6) 57 (87.7) 0.005†
s3, s4, s5 9 (12.2) 46 (70.8) <0.001†
s6 23 (31.1) 22 (33.9) 0.728†
s7, s8, s7+8, s9, s10 12 (16.2) 18 (27.7) 0.101†
Tree-in-bud 69 (93.2) 43 (66.2) <0.001†
Centri-lobular nodules 59 (79.7) 49 (75.4) 0.539†
Ground-glass opacity 69 (93.2) 60 (92.3) 1.000‡
Adenopathy 50 (67.6) 47 (72.3) 0.544†
Bronchial thickening 49 (66.2) 55 (84.6) 0.013†
Interlobular thickening 41 (55.4) 38 (58.5) 0.717†
Fibrosis 21 (28.4) 1 (1.5) <0.001†
Calcificationb 17 (23.3) 2 (3.1) 0.001†
s1 apical segment, s2 posterior segment of right upper lobe, s1+s2 apico-posterior segment of left upper lobe, s3 anterior segment of right or leftupper lobe, s4 lateral segment of right middle lobe or superior segment ofleft lingual lobe, s5 medial segment of right middle lobe or inferiorsegment of left lingual lobe, s6 superior segment of right or left lowerlobe, s7 medical basal segment of right lower lobe, s8 anterior basalsegment of right lower lobe, s7+8 medial-anterior basal segment of leftlower lobe, s9 lateral basal segment of right or left lower lobe, s10posterior basal segment of right or left lower lobe
Data are presented as number and percentage (%). Comparisons betweengroups were performed using †Chi-square test or ‡Fisher’s exact test
*p<0.05
2382 Eur Radiol (2014) 24:2372–2384
Conclusions
The models developed on the basis on HRCT imaging dataallow one to predict the presence of smear-positive active PTBin elderly and non-elderly patients before sputum smear orcultures results have been obtained. This HRCT-based scoringsystemmay assist physicians in the early diagnosis of the mostinfectious smear-positive active PTB, thereby reducing thespatial and financial burden to hospitals, as well as minimizingunnecessary immediate respiratory isolation and infectiousrisk to patients and healthcare workers.
Acknowledgements The scientific guarantor of this publication is Jun-Jun Yeh. The authors of this manuscript declare no relationships with anycompanies whose products or services may be related to the subjectmatter of the article. The authors state that this work has not receivedany funding. Jun-Jun Yeh, Solomon Chih-Cheng Chen, Cheng-RenChen, Ming-Ting Wu kindly provided statistical advice for this manu-script. Institutional review board approval was obtained. Written in-formed consent was obtained from all subjects (patients) in this study.Methodology: retrospective, observational, multicentre study
References
1. Horne DJ, Royce SE, Gooze L et al (2010) Sputum monitoringduring tuberculosis treatment for predicting outcome: systematicreview and meta-analysis. Lancet Infect Dis 10:387–394
2. Beck-SagueC,Dooley SW,HuttonMDet al (1992)Hospital outbreak ofmultidrug-resistant Mycobacterium tuberculosis infections. Factors intransmission to staff and HIV-infected patients. JAMA 268:1280–1286
3. Dye C, Watt CJ, Bleed DM, Hosseini SM, Raviglione MC (2005)Evolution of tuberculosis control and prospects for reducing tuberculo-sis incidence, prevalence, and deaths globally. JAMA 293:2767–2775
4. Rawat J, Sindhwani G, Juyal R (2008) Clinico-radiological profile ofnew smear positive pulmonary tuberculosis cases among young adultand elderly people in a tertiary care hospital at Deheradun(Uttarakhand). Indian J Tuberc 55:84–90
5. Hsieh MJ, Liang HW, Chiang PC et al (2009) Delayed suspicion,treatment and isolation of tuberculosis patients in pulmonology/infectious diseases and non-pulmonology/infectious diseases wards.J Formos Med Assoc 108:202–209
6. Dinnes J, Deeks J, Kunst H et al (2007) A systematic review of rapiddiagnostic tests for the detection of tuberculosis infection. HealthTechnol Assess 11:1–196
7. Hatipoglu ON, Osma E, Manisali M et al (1996) High resolutioncomputed tomographic findings in pulmonary tuberculosis. Thorax51:397–402
8. Waitt CJ, Joekes EC, Jesudason N et al (2013) The effect of atuberculosis chest X-ray image reference set on non-expert readerperformance. Eur Radiol 23:2459–2468
9. Pinto LM, Dheda K, Theron G et al (2013) Development of a simplereliable radiographic scoring system to aid the diagnosis of pulmo-nary tuberculosis. PLoS ONE 8:e54235
10. Kowada A (2013) Cost effectiveness of high resolution computedtomography with interferon-gamma release assay for tuberculosiscontact investigation. Eur J Radiol 82:1353–1358
11. Kobashi Y, Mouri K, Fukuda M, Yoshida K, Miyashita N, Oka M(2008) Transitional change in the clinical features of pulmonarytuberculosis. Respiration 75:304–309
12. Korzeniewska-Kosela M, Krysl J, Muller N, Black W, Allen E,FitzGerald JM (1994) Tuberculosis in young adults and the elderly.A prospective comparison study. Chest 106:28–32
13. Okamura K, Nagata N, Kumazoe H et al (2011) Relationshipbetween computed tomography findings and nutritional statusin elderly patients with pulmonary tuberculosis. Intern Med50:1809–1814
14. American College of Radiology (ACR) (2010) ACR practice guide-line for the performance of high-resolution computed tomography(HRCT) of the lungs in adults. ACR, Reston, p 5, online publication
15. Fung HB, Monteagudo-Chu MO (2010) Community-acquired pneu-monia in the elderly. Am J Geriatr Pharmacother 8:47–62
16. Centers for Disease Control and Prevention. Menu of SuggestedProvisions For State Tuberculosis Prevention and Control Laws.2012. http://www.cdc.gov/TB/programs/Laws/menu/TBLawMenu.pdf. Accessed 20 Oct 2013
17. Lim WS, Macfarlane JT, Boswell TC et al (2001) Study of commu-nity acquired pneumonia aetiology (SCAPA) in adults admitted tohospital: implications for management guidelines. Thorax 56:296–301
18. Yeh JJ, Chen SC, Teng WB et al (2010) Identifying the most infec-tious lesions in pulmonary tuberculosis by high-resolution multi-detector computed tomography. Eur Radiol 20:2135–2145
19. Kanaya AM, Glidden DV, Chambers HF (2001) Identifying pulmo-nary tuberculosis in patients with negative sputum smear results.Chest 120:349–355
20. Wang JH, Pappas D, De Jager PL et al (2011) Modeling the cumu-lative genetic risk for multiple sclerosis from genome-wide associa-tion data. Genome Med 3:3
21. Akobeng AK (2007) Understanding diagnostic tests 2: likelihoodratios, pre- and post-test probabilities and their use in clinical practice.Acta Paediatr 96:487–491
22. Moreira J, Alarcon F, Bisoffi Z et al (2008) Tuberculous meningitis:does lowering the treatment threshold result in many more treatedpatients? Trop Med Int Health 13:68–75
23. Van den Ende J, Mugabekazi J, Moreira J et al (2010) Effectof applying a treatment threshold in a population. An exampleof pulmonary tuberculosis in Rwanda. J Eval Clin Pract 16:499–508
24. Wong KS, Huang YC, Lai SH, Chiu CY, Huang YH, Lin TY (2010)Validity of symptoms and radiographic features in predicting positiveAFB smears in adolescents with tuberculosis. Int J Tuberc Lung Dis14:155–159
25. Perez-Guzman C, Torres-Cruz A, Villarreal-Velarde H, Vargas MH(2000) Progressive age-related changes in pulmonary tuberculosisimages and the effect of diabetes. Am J Respir Crit Care Med 162:1738–1740
26. Wang CS, Chen HC, Yang CJ et al (2008) The impact of age on thedemographic, clinical, radiographic characteristics and treatment out-comes of pulmonary tuberculosis patients in Taiwan. Infection 36:335–340
27. Nakanishi M, Demura Y, Ameshima S et al (2010) Utility of high-resolution computed tomography for predicting risk of sputumsmear-negative pulmonary tuberculosis. Eur J Radiol 73:545–550
28. Miyashita N, Sugiu T, Kawai Yet al (2009) Radiographic features ofMycoplasma pneumoniae pneumonia: differential diagnosis and per-formance timing. BMC Med Imaging 9:7
29. Reittner P, Ward S, Heyneman L, Johkoh T, Muller NL (2003)Pneumonia: high-resolution CT findings in 114 patients. Eur Radiol13:515–521
30. Lee KS, Hwang JW, Chung MP, Kim H, Kwon OJ (1996) Utility ofCT in the evaluation of pulmonary tuberculosis in patients withoutAIDS. Chest 110:977–984
31. Yeh JJ, Yu JK, TengWB et al (2012) High-resolution CT for identifypatients with smear-positive, active pulmonary tuberculosis. Eur JRadiol 81:195–201
Eur Radiol (2014) 24:2372–2384 2383
32. Tanaka N, Matsumoto T, Kuramitsu T et al (1996) High resolutionCT findings in community-acquired pneumonia. J Comput AssistTomogr 20:600–608
33. Ors F, Deniz O, Bozlar U et al (2007) High-resolution CTfindings in patients with pulmonary tuberculosis: correlationwith the degree of smear positivity. J Thorac Imaging 22:154–159
34. Jamzad A, Shahnazi M, Khatami A et al (2009) Radiographic find-ings of pulmonary tuberculosis in Tehran in comparison with otherinstitutional studies. Iran J Radiol 6:131–136
35. Lee JH, Han DH, Song JW, Chung HS (2005) Diagnostic andtherapeutic problems of pulmonary tuberculosis in elderly patients.J Korean Med Sci 20:784–789
36. Yu MC, Bai KJ, Chang JH, Lee CN (2006) Age transition of tubercu-losis patients in Taiwan, 1957-2001. J Formos Med Assoc 105:25–30
37. Stead WW, Dutt AK (1991) Tuberculosis in elderly persons. AnnuRev Med 42:267–276
38. Fujikawa A, Fujii T, Mimura S et al (2014) Tuberculosis contactinvestigation using interferon-gamma release assay with chest x-rayand computed tomography. PLoS ONE 9:e85612
2384 Eur Radiol (2014) 24:2372–2384
