Received: 16 June 2018 | Accepted: 8 December 2018

DOI: 10.1002/ppul.24230

ORIGINAL ARTICLE: IMAGING

Chest ultrasound findings in children with suspected

pulmonary tuberculosis

Charlotte C. Heuvelings MD1,2 | Sabine Bélard MD1,2,3,4 |

Savvas Andronikou PhD2,5 | Norme Jamieson-Luff BSc2 |
Martin P. Grobusch FRCP1 | Heather J. Zar PhD2

1 Center of Tropical Medicine and Travel

Medicine, Amsterdam University Medical Abstract
Centers, University of Amsterdam,
Introduction: Chest ultrasound is increasingly used for the diagnosis of pediatric
Amsterdam, The Netherlands
2 Department of Paediatrics and Child Health
lung disease but there are limited data for its use in pediatric pulmonary tuberculosis
and SA-MRC Unit on Child and Adolescent (PTB).
Health, Red Cross War Memorial Children's
Hospital, University of Cape Town, South
Aim: To describe chest ultrasound findings in children with suspected PTB.
Africa Methods: Consecutive children, presenting with suspected PTB to a tertiary children's
3 Department of Pediatric Pneumology and
hospital in Cape Town between July 2014 and March 2016, were enrolled in this cohort
Immunology, Charité-Universitätsmedizin
Berlin, Berlin, Germany study. Children were categorized into three groups based on microbiological and
4 Berlin Institute of Health, Berlin, Germany clinical features; confirmed PTB (microbiologically confirmed), unconfirmed PTB
5 Department of Paediatric Radiology, (clinical diagnosis only), and unlikely PTB (respiratory disease not due to PTB). A
Children's Hospital Philadelphia, Philadelphia,
Pennsylvania clinician, blinded to categorization, performed chest and mediastinal ultrasound for
consolidation, pleural gaps, pleural effusions, B-lines or enlarged mediastinal lymph
Correspondence
Charlotte C. Heuvelings, MD, Center of nodes at enrolment and 1, 3, and 6 months thereafter. Two readers interpreted the
Tropical Medicine and Travel Medicine, ultrasounds independently.
Academic Medical Center Amsterdam,
University of Amsterdam, The Netherlands. Results: One hundred seventy children (median age 26.6 months) were enrolled; 40
Email: c.c.heuvelings@amc.uva.nl (24%) confirmed PTB, 85 (50%) unconfirmed PTB, and 45 (26%) unlikely PTB. In
Funding information children with confirmed PTB, pleural effusion was more common (30% vs 9% in unlikely
NIH, Grant number: RO1 HD058971; Marie PTB, P = 0.024), mediastinal lymph nodes were larger (median size 1.5 cm vs 1.0 cm in
Curie People grant; MRC South Africa; NRF
South Africa unlikely PTB, P = 0.027), resolution of consolidation occurred less commonly at 1-
month follow-up (24% vs 67% unlikely TB, P = 0.014) and the proportional size
reduction of a consolidation was lower (44% vs 80% in unlikely PTB, P = 0.009). Inter-
reader agreement was perfect to moderate.
Conclusion: Chest ultrasound identified abnormalities suggestive of PTB with a high
inter-reader agreement. Consolidation showed slower resolution in children with
confirmed PTB.

KEYWORDS
chest ultrasound, diagnostics, mediastinal ultrasound, pediatric pulmonary tuberculosis

Primary institution where research was conducted: Department of Paediatrics and Child Health and SA-MRC Unit on Child and Adolescent Health, Red Cross War Memorial Children's
Hospital, University of Cape Town, South Africa.

Pediatric Pulmonology. 2019;1–8. wileyonlinelibrary.com/journal/ppul © 2019 Wiley Periodicals, Inc. | 1

2 | HEUVELINGS ET AL.

1 | INTRODUCTIO N Children with extra-pulmonary TB (EPTB) only, or children on TB

treatment or TB prophylaxis for more than 72 h, or unable to provide
Tuberculosis (TB) causes substantial morbidity and mortality in an adequate sputum sample for microbiological testing were excluded.
children, with approximately 250 000 child deaths in 2016.1 Diagnos- For microbiological diagnosis, two induced sputum samples were
ing pulmonary TB (PTB) in children is challenging, as signs and obtained and subjected to liquid mycobacterial culture (BACTEC
symptoms are non-specific, and microbiological confirmation can be MGIT; Becton Dickinson, Sparks, MD) and Xpert MTB/RIF (Xpert,
difficult due to pauci-bacillary disease and limited expertise. Therefore, Cepheid, Sunnyvale, CA).12 Drug sensitivity testing was done by using
diagnosis of childhood PTB mainly relies on clinical and radiological Hain genotype MTBDR test and MGIT culture. Children were
findings. Radiological features of PTB in children are hilar or categorized as confirmed PTB (Mycobacterium tuberculosis detected
mediastinal lymphadenopathy, miliary pattern, consolidation, or by either culture or GeneXpert), unconfirmed PTB (clinical diagnosis
unilateral pleural effusion.2 Older children may exhibit radiological for PTB but negative microbiological test result) or unlikely PTB
features of adult-type PTB.2 However, the inter- and intra-reader (respiratory disease due to other organisms; symptoms improved
agreement for the detection of lymphadenopathy on chest X-ray (CXR) without TB treatment) according to revised National Institute of
3,4
is low. Radiography also exposes the child to ionizing radiation and in Health consensus recommendations.13 The treating clinician was
resource-limited settings, may be unavailable. responsible for commencing TB treatment based on the national TB
Chest ultrasound is an emerging imaging modality with good guidelines.14 All children underwent chest and mediastinal ultrasound
5
diagnostic accuracy for childhood pneumonia. The advantages of and were followed-up at 1 and 3 months after enrolment to repeat
chest ultrasound are non-exposure to ionizing radiation; bedside chest and mediastinal ultrasound; those with confirmed or uncon-
performance by the treating clinician reducing time to diagnosis; firmed PTB were seen 6 months after enrolment as well. We used a
repeatability, thus rendering it a good tool to monitor treatment convenience sample of 140 children to longitudinally investigate chest
response; and with the recent development of affordable, portable ultrasound changes. We anticipated a 10% lost-to-follow-up and
ultrasound machines this investigation can be cost-effective, especially therefore aimed to enroll at least 154 children.
in resource-limited settings.
A small number of studies have reported on chest ultrasound for
2.3 | Chest ultrasound
diagnosis of PTB in adults. Chest ultrasound of a HIV infected patient
with PTB showed a shred sign representing consolidation.6 Another A bedside chest ultrasound was performed at the pediatric ward or the
study, comprising ten adults with miliary PTB, described that TB research unit, either by a clinician (CCH) who had attended a 4-day
presents as bilateral B-lines in multiple lung zones plus sub-pleural ultrasound training or a trained sonographer (NJL) with 11 years of
lesions on chest ultrasound.7 A cohort study including 60 adults with echocardiography experience. The clinician performed most (85%) of
PTB found hypo-echoic sub-pleural nodules (smaller than 1 cm in the scans. A portable gray scale ultrasound machine (Mindray DP10,
diameter) as the commonest ultrasound finding.8 Three small pediatric Mindray, Shenzhen, China) with a 5-10 MHz linear probe was used.
studies showed that mediastinal lymphadenopathy can be demon- The scanning protocol described by Copetti and Cattarossi15 was used
9–11
strated on ultrasound in children with PTB. We describe the chest to scan the anterior (mid-clavicular line), posterior (mid-scapular line),
ultrasound findings in children with PTB compared to those with a and lateral (mid-axillar line) chest with the patient in sitting or supine
lower respiratory tract infection caused by other pathogens. position. The chest was scanned in two planes, longitudinal and
transverse (intercostal view) plane, moving the probe from the apex
down to the diaphragm. The apices were scanned, placing the probe
2 | METHODS
transversely in the supraclavicular region, tilting the probe caudally.
Additionally, a mediastinal ultrasound scan was done to identify
2.1 | Study design
enlarged lymph nodes. The patient was placed in a supine position with
This prospective cohort study was conducted at a tertiary children's the neck slightly extended to improve access to the suprasternal notch.
hospital in Cape Town, South Africa, between July 2014 and A micro-convex probe (5.0-8.5 MHz) was used. For the transverse view,
March 2016. Ethical approval was granted by the Research Ethics the probe was positioned transversely at the suprasternal notch, tilting
Committee of the Faculty of Health Sciences, University of Cape the probe cranially and caudally to obtain a full view. For the oblique
Town, and written informed consent was provided by the parent or view, the probe was positioned diagonally at the suprasternal notch,
legal guardian. with the superior tip on the patient's left, pointing the probe toward the
aortic arch and tilting from left to right to obtain a full view.16
The clinician (CCH) recorded the findings of all ultrasounds (also
2.2 | Inclusion criteria
the ultrasound scans performed by NJL) on a standardized reporting
Children up to the age of 13 years were consecutively enrolled if they form, reporting on 14 regions in the chest; namely the left and right,
presented with a cough plus one of the following symptoms; (i) weight upper and lower, anterior, lateral and posterior chest and left and right
loss or failure to thrive over the past 3 months; (ii) a positive tuberculin apex. Box 1 presents the definitions of the reported chest ultrasound
skin test; (iii) CXR suggestive of PTB; and/or (iv) close TB contact. findings.
HEUVELINGS ET AL..
| 3

agreement for each sonographic finding per patient was compared. In

case of discordant readings, the saved clips were re-reviewed and the
Box 1. Chest ultrasound findings findings discussed to achieve a concordant final decision.
The sonographer and second clinician reporting on the ultrasounds
1. Lung sliding: the presence or absence of lung sliding, the
were blinded to clinical data, PTB category, and to each other's findings.
movement of the pleural line during inspiration. Absence
of lung sliding is a sign of pneumothorax.
2. A-lines: a normal finding, these are horizontal reverbera- 2.4 | Analysis
tion artefacts of the pleural line (E-image S1, Supporting
Information). STATA 14.2 (StataCorp; 2015, TX) was used for analysis. Descriptive

3. Interruption of the pleural line: any interruption of the statistics were used to describe the study characteristics, mean and standard
deviations (SD) for normally distributed continuous data, median, and
pleural line was recorded and divided into:
interquartile range (IQR) for non-normally distributed continuous data.
a. Interruption caused by a consolidation—presents as a
sub-pleural echo-poor area or a tissue-like area Numbers and proportions were used for categorical data. The chi-square
test (Fisher's exact test if number <5) was used to evaluate statistical
>0.5 cm with or without air-bronchograms interrupt-
significance of a chest ultrasound finding between two PTB categories, and
ing the pleural line (E-image S2, Supporting
Information). the Mann-Whitney comparison was used to compare medians. We used the

b. Interruption caused by a pleural gap—a small (<0.5 cm) Kappa Cohen coefficient (κ) to evaluate inter-reader agreement.

sub-pleural nodular consolidations interrupting the

pleural line (E-image S3, Supporting Information). 3 | RESULTS
c. Interruption of the pleural line without consolidation that
can be caused by B-lines which are longitudinal lines, 3.1 | Characteristics
originating from the pleural line to the edge of the
ultrasound screen; more than three B-lines per intercos- We included 170 children, median age 26.6 months [IQR 15.2-59.3],
96 (57%) boys. Forty (24%) children were categorized as confirmed
tal space in the longitudinal plane in two or more lung
PTB, 85 (50%) as unconfirmed PTB and 45 (26%) as unlikely PTB
regions is a pathological finding for lung interstitial
syndrome (E-image S4, Supporting Information). (Figure 1). All children with confirmed PTB had drug sensitive TB.
Twenty-three children (14%) were HIV infected. Table 1 describes the
4. Pleural effusion: presents as an anechoic collection
baseline characteristics by PTB category.
between the pleural line and the chest wall. Fibrin strands
can be visible within the effusion (E-image S5, Supporting
Information).
5. Enlarged mediastinal lymph nodes: lymph nodes are seen as
hypoechoic, round or oval structures between the
mediastinal blood vessels16 (E-image S6, Supporting
Information). We recorded enlarged lymph nodes as
lymph nodes ≥1 cm in diameter, in the longest dimension.

Both clinicians (CCH and SB) reported the overall quality of the
chest ultrasound images as “good” (pleural line, A-lines and/or findings
clearly visible in all views), “moderate” (pleural line, A-lines and/or
findings clearly visible in most views but not all views), or “poor”
(pleural line, A-lines and/or findings not clearly visible in most views).
The quality of the mediastinal ultrasound was reported per view as
“good” (all landmarks16 clearly visible), “moderate” (most landmarks16
visible) and “poor” (landmarks16 not clearly visible). The sonographer
reported on the compliance of the patient as “good” (patient was
cooperative and calm during the examination), “moderate” (patient
cried during the examination but remained calm) or “poor” (patient was
crying and moving during the examination).

A second clinician (SB) with one year of pediatric ultrasound

experience also reported on the saved ultrasound clips. The inter-reader FIGURE 1 Flow diagram
4 | HEUVELINGS ET AL.

TABLE 1 Baseline characteristics

Total Confirmed PTBa Unconfirmed PTBb Unlikely PTBc
n = 170 n = 40 n = 85 n = 45
Male, n (%) 96 (57) 26 (65) 50 (59) 20 (44)
Median age in months [IQR] 26.6 [15.2-59.3] 48.5 [18.3-71.0] 23.9 [13.3-43.0] 23.9 [17.3-56.2]
HIV-infected, n (%) 23 (14) 6 (15) 12 (14) 5 (11)
CD4 countn = 22, median [IQR] 334 [157-626] 310 [206-358] 557 [225-649] 151 [46-157]
n = 19
CD4% , median [IQR] 11.4 [7.5-20.7] 18.6 [14.4-20.7] 10.3 [6.5-18.2] 9.6 [5.1-15.8]
Positive tuberculin skin testn = 135, n (%) 77 (57) 26 (96) 50 (59) 1 (3)
Clinical signs and symptoms
Cough, n (%) 143 (84) 32 (80) 70 (82) 41 (91)
Night sweats, n (%) 107 (63) 26 (65) 56 (66) 25 (56)
Weight loss, n (%) 114 (67.1) 27 (68) 53 (62) 34 (76)
Reduced breath sounds, n (%) 9 (5) 3 (8) 3 (4) 3 (7)
Crepitations, n (%) 25 (15) 12 (30) 13 (15) 17 (38)
Wheeze, n (%) 34 (20) 5 (13) 17 (20) 12 (27)
a
Confirmed PTB: Mycobacterium tuberculosis detected by either culture or GeneXpert.
b
Unconfirmed PTB: clinical diagnosis for PTB but negative microbiological test result.
c
Unlikely PTB: respiratory disease due to other organisms; symptoms improved without TB treatment.

One-hundred and fifty-four (91%) children (39/40 [98%] con- moderate, and 1% poor quality. Mediastinal ultrasound could only be
firmed PTB, 73/85 [86%] unconfirmed PTB, and 41/45 [91%] unlikely evaluated in 69% of the children, in 31% mediastinal ultrasound clips
PTB) returned for at least one follow-up chest ultrasound (Figure 1). were of poor quality or could not be performed at all due to poor
Of the children that returned for follow-up, 68 (16/39 [41%] compliance. Of the 117 mediastinal ultrasounds that could be
confirmed PTB, 52/73 [71%] unconfirmed PTB) received 3-drug TB evaluated, 62% had a good quality. The image quality of chest and
treatment (rifampicin, isoniazid, and pyrazinamide [RHZ]), 44 (23/39 mediastinal ultrasound improved with age.
[59%] confirmed PTB and 21/73 [30%] unconfirmed PTB, P = 0.002)
received 4-drug TB treatment (RHZ plus ethambutol or ethionamide
3.3 | Chest ultrasound findings at enrolment
[RHZE/RHZEto]), and one child in the unlikely PTB category received
seven days RHZ. Children with HIV were more likely to receive 4-drug The chest ultrasound findings at enrolment are presented in Table 3.
therapy than children without HIV (15/18 [83%] vs 28/94 [30%], Pleural effusion was seen in 25 (15%) children and was
P = 0.000). Nine children (2/39 [5%] confirmed PTB and 7/85 [9%] significantly more common in the confirmed PTB category (30%)
unconfirmed PTB) reported poor adherence to TB treatment, 117 children than in the unlikely PTB category (9%, P = 0.024). In 13/25 (52%)
(29/39 [74%] confirmed PTB, 56/74 [76%] unconfirmed PTB, and 32/41 children (8/12 [67%] confirmed PTB, 3/9 [33%] unconfirmed PTB and
[78%] unlikely PTB) received antibiotics other than anti-TB medication 2/4 [50%] unlikely PTB) the effusion was seen adjacent to a
and 15 children (6/39 [15%] confirmed PTB, 8/74 [11%] unconfirmed consolidation. The median age of the children with pleural effusion
PTB and 1/41 [2%] unlikely PTB) received prednisone (Table 2). was 64.6 months [IQR 33.4-104.8] vs 23.9 months [IQR 14.9-48.3] of
children without pleural effusion (P = 0.001).
Enlarged lymph nodes were seen in 27 (22.9%) children; the lymph
3.2 | Quality chest ultrasound
nodes were significantly larger in the confirmed (median size 1.50 cm
The quality of the chest ultrasound images was similar in all three [IQR 1.00-1.58 cm]) and unconfirmed PTB category (median size
categories (E-Table S1, Supporting Information), 83% good, 16% 1.35 cm [IQR 1.30-1.50 cm]) compared to unlikely PTB category

TABLE 2 Treatment per TB category

Returning for at least Poor
one follow-up scan RHZ RHZE adherence Antibiotics Prednisone
TB category n (%) n (%) n (%) n (%) n (%) n (%)
Confirmed PTB 39 (98) 17 (44) 22 (56) 2 (5) 29 (74) 6 (15)
Unconfirmed 74 (87) 49 (66) 25 (34) 7 (9) 56 (76) 8 (11)
PTB
Unlikely PTB 41 (91) 1 (3) 0 - 32 (78) 1 (2)

RHZ (Rifampicin, Isoniazid, Pyrazinamide), RHZE (RHZ + Ethambutol or Ethionamide).

HEUVELINGS ET AL..
| 5

TABLE 3 Chest ultrasound findings per TB category

P-value P-value P-value confirmed and
Confirmed PTB Unconfirmed PTB Unlikely PTB confirmed vs unconfirmed vs unconfirmed vs unlikely
n = 40 n = 85 n = 45 unlikely PTB unlikely PTB PTB

Interrupted 31 (78) 67 (79) 33 (73) 0.657 0.547 0.488

pleural line, n
(%)

Consolidation, n 22 (55) 35 (41) 21 (47) 0.443 0.661 0.902

(%)

Median size 2.78 [1.50-4.00] 2.00 [1.08-3.50] 2.74 [1.36-4.20] 0.361 0.286 0.774
in cm [IQR]
Pleural gap, n (%) 17 (43) 49 (58) 19 (42) 0.979 0.136 0.224

>3 B-lines, n (%) 13 (33) 22 (26) 10 (22) 0.287 0.573 0.451

Pleural effusion, 12 (30) 9 (11) 4 (9) 0.024 1.000 0.230

n (%)

Enlarged lymph 7/25a (28) 9/59 (15) 9/32 (28) 0.992 0.176 0.288
nodes
(n = 118), n (%)

Median size 1.50 [1.00-1.58] 1.38 [1.30-1.50] 1.00 [1.00-1.01] 0.028 0.001 0.001
in cm [IQR]

Normal chest 3 (8) 9 (10) 7 (17) 0.250 0.455 0.277

ultrasound, n
(%)

a
Enumerator = number of children with enlarged lymph nodes on mediastinal ultrasound; denominator = number of children that had a mediastinal
ultrasound of sufficient quality to be evaluated.

(median size 1.00 cm [IQR 1.00-1.01 cm]; P = 0.028 and P = 0.001, remained the same, and in three (12%) children (2/13 [15%] confirmed
respectively). Detection of enlarged lymph nodes was not age- PTB and 1/8 [13%] unconfirmed PTB) the consolidation increased in
dependent. The youngest child with enlarged lymph nodes was size.
13 months old, and the oldest child was 8.5 years old. At 3 months, consolidation was still more commonly seen in the
There was no statistically significant difference between the three confirmed PTB category (15/34 [44%]) compared to the unlikely
PTB categories for the other chest ultrasound findings, that is, category (5/33 [17%], P = 0.010).
consolidation (>0.5 cm), pleural gaps (consolidations <0.5 cm) or more In six of the 18 (35%) children (5/8 [63%] confirmed PTB, 1/9
than three B-lines per intercostal space in more than two lung regions [11%] unconfirmed PTB, 0/1 [0%] unlikely PTB) with pleural effusion
(Table 3). at enrolment, the effusion resolved at one-month. One of the
Two or more and three or more findings were more commonly 14 children with a pleural effusion at enrolment seen at 6-month
detected in children with confirmed PTB (58% and 40% respectively) follow-up, still had a small residual pleural effusion visible at the left
than in children with unlikely PTB (38%, P = 0.107 and 18%, P = 0.070, lower lobe; at enrolment this child had a very large effusion, with
respectively; E-table S2, Supporting Information). septae, compromising the complete left lung.
In 12/20 (60%) children (2/5 [40%] confirmed PTB, 5/9 [56%]
unconfirmed PTB, and 5/6 [56%] unlikely PTB) with enlarged
3.4 | Chest ultrasound findings at follow-up
mediastinal lymph nodes at enrolment, the lymph nodes were no
Follow-up findings are shown in Table 4. longer visible at month one.
At 1-month follow-up, consolidation resolved in 30/56 (53%) In two children with unconfirmed PTB the size of the lymph nodes
children with consolidation at enrolment. Resolution occurred reduced, in one child with confirmed PTB the lymph node size
significantly less commonly in children with confirmed PTB (4/17 remained the same and in five children (2/3 confirmed PTB, 2/4
[24%]) than in unconfirmed PTB (16/24 [67%]) or unlikely PTB (10/15 unconfirmed PTB, and 1/1 unlikely PTB) the lymph nodes increased in
[67%], P = 0.014). size.
In 18 (69%) of the 26 children (8/13 [62%] confirmed PTB, 5/8 In another six children, enlarged mediastinal lymph nodes were
[63%] unconfirmed PTB, and 5/5 [100%] unlikely PTB) who still had a seen for the first time at month one; of which three children had a poor
consolidation at 1-month follow-up, the consolidation had reduced in quality mediastinal ultrasound scan and one child had lymph nodes
size. The least proportional size reduction was seen in the confirmed smaller than 1 cm on enrolment.
PTB category (44% reduction vs 76% unconfirmed PTB and 80% At month three, mediastinal lymphadenopathy was no longer seen
unlikely PTB, P = 0.009). In five (19%) children (3/13 [23%] confirmed in the unlikely PTB category, 8/9 with enlarged lymph nodes at
PTB and 2/8 [25%] unconfirmed PTB) the size of the consolidation enrolment returned at 3-month follow-up.
6 | HEUVELINGS ET AL.

In 3/8 (38%) children (2/4 [50%] confirmed PTB and 1/4 [25%]

Unconfirmed PTB

2.00 [0.83-4.00]

1.26 [1.26-1.26]
unconfirmed PTB) mediastinal lymphadenopathy was still visible at
month six, the clinical symptoms in all three children were resolved at

n = 56/85

Enumerator = number of children with enlarged lymph nodes on mediastinal ultrasound; denominator = number of children that had a mediastinal ultrasound of sufficient quality to be evaluated.
1/35 (3)
month six but in the two confirmed PTB cases ultrasound showed a

33 (59)

30 (54)
7 (13)

1 (2)
2 (4)
consolidation as well.

1.49 [1.00-5.50]

1.00 [1.00-1.00]
3.4.1 | Follow-up findings and treatment
Confirmed PTB
n = 24/40

2/18 (11)
After 3 months of treatment a pleural effusion was only seen in
Month 6

14 (58)

10 (42)
7 (29)

0 (0) children treated with RHZ (n = 6) and no longer seen in children with
1 (4) RHZE/RHZEto (P = 0.011, E-Table S3, Supporting Information).
Treatment with RHZ or RHZE/RHZEto, adherence to TB treatment,
0.88 [0.63-1.10]

antibiotics other than anti-TB medication, prednisone or HIV infection

Unlikely PTB

did not have any influence on the appearance or disappearance of a

n = 33/45

0/30 (0)
11 (37)

consolidation or enlarged mediastinal lymph nodes at any time of

5 (17)

7 (23)

3 (10)
2 (6)

Enumerator = number of children returning for follow up per TB category per month; denominator = the total of children enrolled per TB category.

follow-up (E-Table S3, Supporting Information).

-
Unconfirmed PTB

3.5 | Inter-reader agreement

3.00 [1.50-5.00]

1.54 [1.00-2.00]

There was a perfect agreement between the sonographer and the

n = 64/85

3/67 (5)
29 (43)

27 (40)

second reader for lung sliding and A-lines, and the inter-reader
7 (10)

4 (6)
5 (8)

agreement for consolidation and pleural effusion was almost perfect

(κ = 0.84 and κ = 0.89, respectively). The inter-reader agreement for
2.48 [1.21-7.00]

1.39 [1.00-1.50]
Confirmed PTB

more than three B-lines per intercostal space in more than two lung
areas and for interruption of the pleural line were substantial (κ = 0.73
10/34 (29)
n = 34/40
Month 3

and κ = 0.62, respectively). There was a moderate agreement on

24 (71)
15 (44)

14 (41)

4 (12)
2 (7)

mediastinal lymphadenopathy (κ = 0.56).

1.64 [0.74-3.10]

1.24 [1.2-1.27]

4 | DISCUSSIO N
Unlikely PTB
n = 35/45

2/29 (7)
15 (47)

15 (47)
5 (16)

4 (13)

Our study has shown that chest ultrasound may be useful to detect
2 (6)

findings associated with PTB and for follow-up in children. At

enrolment, pleural effusion was associated with confirmed PTB as
Unconfirmed PTB

were larger lymph nodes. Slower, or no resolution, of a consolidation at

1.87 [0.89-2.64]

1.20 [1.11-1.68]

follow-up was associated with PTB and 4-drug TB treatment had a

n = 61/85

7/64 (11)

positive association on the disappearance of pleural effusions after

27 (42)
11 (17)

27 (42)

11 (17)
8 (13)

3 months of treatment. The inter-reader agreement was very high and

good quality ultrasounds were obtained in most children.
Chest ultrasound findings at follow-up

Pleural effusion was more common in children with PTB, occurring

2.48 [1.50-8.00]

1.49 [1.30-1.50]
Confirmed PTB

in 30% of the confirmed PTB cases, and particularly in older children.

n = 33/40a

5/31 (16)

This is consistent with prior reports documenting pleural effusion as a

Month 1

20 (61)
15 (46)

20 (61)

complication of PTB in 2-38% of pediatric cases, and more commonly

b
7 (21)
3 (9)

in older children.17 Pleural effusion may develop as an immune

reaction to Mycobacterium tuberculosis but could also be caused by
Interrupted pleural line, n (%)

contiguous spread of PTB.17 A pleural effusion is not diagnostic for

Enlarged lymph nodes, n (%)

PTB, as it may occur as a complication of pneumonia due to other

Median size in cm [IQR]

Median size in cm [IQR]

Pleural effusion, n (%)

bacteria or viruses.18 Nevertheless, the occurrence of pleural effusion

Consolidation, n (%)

Pleural gap, n (%)

in an older child, in a TB-endemic setting, should be considered as

>3 B-lines, n (%)

possible PTB. Treatment with four TB drugs was associated with a

TABLE 4

better resolution of pleural effusion, possibly suggesting that four TB

drugs could be the preferred choice of treatment in children with a
pleural effusion.
b
a
HEUVELINGS ET AL..
| 7

Larger lymph nodes were associated with PTB (confirmed and results from two interpreting sonographers would be desirable but was
unconfirmed PTB), although the proportion of enlarged lymph nodes not feasible in this setting.
was similar in all three categories. Therefore, the presence of enlarged Another limitation is a small sample size for the confirmed and
lymph nodes per se is not useful for identifying PTB, but the size of the unlikely PTB categories, and for the follow-up findings; larger studies
lymph nodes was helpful in discriminating children with PTB from are needed. Nevertheless, our study has shown that there are key
those with other respiratory diseases. Further studies evaluating ultrasound findings that are associated with PTB.
different size cut-offs stratified by age are needed. Additionally, a third of mediastinal ultrasounds could not be
Previous studies suggested that mediastinal ultrasound was useful evaluated due to low compliance or low-quality scans. Reasons for this
to detect PTB cases that were missed by CXR,9,10 and that if might be that the window to the mediastinum is limited especially in
mediastinal ultrasound was negative, further radiographic investiga- young children, the technique to perform mediastinal ultrasound is
tion for PTB could be avoided.10 However, we found that mediastinal more complex, and the detection of lymph nodes requires more
ultrasound has a poor negative predictive value as the proportion of experience.
children with enlarged lymph nodes visualized by mediastinal We did not compare ultrasound findings with a gold standard
ultrasound was very low in the confirmed (25%) and unconfirmed imaging technique like low-dose CT chest or MRI chest, nor did we
PTB category (15%). This was lower than in the study by Moseme compare our findings with the commonest imaging technique for PTB,
et al11 who found enlarged lymph nodes in 40% of the children with CXR. However, CXR is not the gold standard for the diagnosis of PTB
confirmed or unconfirmed PTB. and has many limitations with wide interobserver variability in
As mediastinal lymph nodes ≥1 cm were also seen in children with interpretation.3,4 However, comparing these two imaging modalities
other respiratory diseases, further studies should evaluate if it will be may be useful to investigate which is a better tool to use as a first-line
useful to included mediastinal ultrasound in routine lung ultrasound imaging modality for the diagnosis of PTB.
protocol for children with lower respiratory tract infection (LRTI). Furthermore, only a small number of children received prednisone
Enlarged lymph nodes were no longer seen in children with other treatment, none of whom had enlarged lymph nodes on mediastinal
respiratory infections at 3 months follow-up. ultrasound at enrolment. Therefore, we could not evaluate the
We found a moderate inter-reader agreement for enlarged influence of prednisone on mediastinal lymph nodes. We also did
mediastinal lymph nodes; this is similar to the overall inter-reader not investigate the role of co-infection with other pathogens on chest
agreement for CXR findings consistent with PTB but higher than the ultrasound findings.
inter-reader agreement for mediastinal lymphadenopathy on CXR Strengths of our study are the categorization in three PTB
(κ = 0.3).3,4 However, enlarged mediastinal lymph nodes on ultrasound categories, especially the confirmed PTB category and the control
still needs to be interpreted with caution. group of children with a lower respiratory tract infection not caused by
Consolidation was the most common finding detected by chest TB, the prospective design and careful follow-up and documentation
ultrasound. A study from Mozambique found that a consolidation of treatment. A further advantage of ultrasound is that it can
was seen on CXR in 65% of TB cases (confirmed and unconfirmed distinguish consolidation of the thymus which, on CXR, can be
TB) in young children,19 slightly higher than we found on chest confused with pneumonia or other pathology.23
ultrasound, with 55% in the confirmed PTB category and 41% in Chest ultrasound has the advantages of not exposing the child to
the unconfirmed PTB category. An important finding from our ionizing radiation, bedside performance by the treating clinician, and
study was that consolidation did not resolve, or resolved less the use to monitor treatment response. Therefore, chest ultrasound
quickly, with the least size reduction in the confirmed PTB should be considered as a first line imaging modality in children with
category, even in children adhering to the correct treatment. suspected PTB especially in settings were access to other imaging
Therefore, the speed of the reduction of a consolidation may be techniques is lacking.
useful ancillary information for the diagnosis of PTB and for
monitoring treatment response.
ACKNOWLEDGMENTS
Comparison of our findings with previous studies6–11 is limited, as
previous studies evaluated paediatric patients with a positive Mantoux We would like to thank all the study participants and their parents/
test9–11 and not children with confirmed PTB, or evaluated adults6–8 in legal guardians for their participation in this study. We are very grateful
whom PTB has a different pathophysiology. for the staff at the SA-MRC Research Unit for Child and Adolescent
A limitation of our study is that ultrasounds were performed and Health and the hospital staff at Red Cross War Memorial Children's
interpreted by a clinician without prior ultrasound experience but with Hospital in Cape Town. We would like to thank Dr. Tom Heller for the
bedside ultrasound training, which might have influenced the findings. ultrasound training. National Institute of Health (NIH) RO1
However, bedside chest ultrasound is intended to be used by clinicians, HD058971, MRC South Africa, NRF South Africa. SB and CCH were
and many other studies on chest ultrasound have been performed by funded by a Marie Curie People grant and SB is currently a participant
inexperienced operators,20–22 with the inter-reader agreement being in the BIH-Charité Clinician Scientist Program funded by Charité-
high. The second clinician interpreted the clips taken by the first Universiätsmedizin Berlin and the Berlin Institute of Health. HZ is
clinician, limiting a direct comparison of the two readers; comparing funded by the MRC South Africa.
8 | HEUVELINGS ET AL.

PRESENTATION 12. Zar HJ, Workman L, Isaacs W, Dheda K, Zamanay W, Nicol MP. Rapid
diagnosis of pulmonary tuberculosis in African children in a primary
Preliminary data were presented at 10th European Congress on care setting by use of Xpert MTB/RIF on respiratory specimens: a
Tropical Medicine and International Health in Antwerp, Belgium (16th- prospective study. Lancet Glob Health. 2013;1:e97–104.
20th October 2017). 13. Graham SM, Cuevas LE, Jean-Philippe P, et al. Clinical case definitions
for classification of intrathoracic tuberculosis in children: an update.
Clin Infect Dis. 2015;61:S179–S187.
14. Department of Health, South Africa. The National Guidelines for the
ORCID
Management of Tuberculosis in Children 2013.
Charlotte C. Heuvelings http://orcid.org/0000-0003-1857-4849 15. Copetti R, Cattarossi L. Ultrasound diagnosis of pneumonia in children.
Radiologia Medica. 2008;113:190–198.
16. Pool KL, Heuvelings CC, Bélard S, et al. Technical aspects of
mediastinal ultrasound for pediatric pulmonary tuberculosis. Pediatr
Radiol. 2017;47:1839–1848.
REFERENCES
17. Zampoli M, Zar HJ. Empyema and parapneumonic effusions in
1. World Health Organization (WHO). Global tuberculosis report 2017. children: an update. SAJCH. 2007;1:121–128.
Geneva: WHO, 2017. 18. Alkrinawi S, Chernick V. Pleural infection in children. Semin Respir
2. Marais BJ, Gie RP, Schaaf HS, et al. A proposed radiological Infect. 1996;11:148–154.
classification of childhood intra-thoracic tuberculosis. Pediatr Radiol. 19. García-Basteiro AL, López-Varela E, Augusto OJ, et al. Radiological
2004;34:886–894. findings in young children investigated for tuberculosis in Mozambi-
3. du Toit G, Swingler G, Iloni K. Observer variation in detecting que. PLoS ONE. 2015;10:e0127323.
lymphadenopathy on chest radiography. Int J Tuberc Lung Dis. 20. Esposito S, Papa SS, Borzani I, et al. Performance of lung ultrasonog-
2002;6:814–817. raphy in children with community-acquired pneumonia. Italian J
4. Swingler GH, du Toit G, Andronikou S, van der Merwe L, Zar HJ. Pediatr. 2014;40:37.
Diagnostic accuracy of chest radiography in detecting mediastinal 21. Shah VP, Tunik MG, Tsung JW. Prospective evaluation of point-of-
lymphadenopathy in suspected pulmonary tuberculosis. Arch Dis Child. care ultrasonography for the diagnosis of pneumonia in children and
2005;90:1153–1156. young adults. JAMA Pediatr. 2013;167:119–125.
5. Pereda MA, Chavez MA, Hooper-Miele CC, et al. Lung ultrasound for 22. Samson F, Gorostiza I, González A, Landa M, Ruiz L, Grau M.
the diagnosis of pneumonia in children: a meta-analysis. Pediatrics. Prospective evaluation of clinical lung ultrasonography in the
2015;135:714–722. diagnosis of community-acquired pneumonia in a pediatric emergency
6. Heuvelings CC, Bélard S, Janssen S, et al. Chest ultrasonography in department. Eur J Emerg Med. 2016;25:65–70.
patients with HIV: a case series and review of the literature. Infection. 23. Gravel CA, Bachur RG. Point-of-care ultrasound differentiation of lung
2016;44:1–10. consolidation and normal thymus in pediatric patients: an educational
7. Hunter L, Bélard S, Janssen S, van Hoving DJ, Heller T. Miliary case series. J Emerg Med. 2018;55:235–239.
tuberculosis: sonographic pattern in chest ultrasound. Infection.
2016;44:243–246.
8. Agostinis P, Copetti R, Lapini L, Badona Monteiro G, N’Deque A,
Baritussio A. Chest ultrasound findings in pulmonary tuberculosis. SUPPORTING INFORMATION
Trop Doct. 2017;47:320–328.
Additional supporting information may be found online in the
9. Bosch-Marcet J, Serres-Creixams X, Zuasnabar-Cotro A, Codina-Puig
X, Catala-Puigbo M, Simon-Riazuelo JL. Comparison of ultrasound Supporting Information section at the end of the article.
with plain radiography and CT for the detection of mediastinal
lymphadenopathy in chiuldren with tuberculosis. Pediatr Radiol.
2004;34:895–900.
10. Bosch-Marcet J, Serres-Creixams X, Borras-Perez V, Coll-Sibina MT,
How to cite this article: Heuvelings CC, Bélard S,
Guitet-Julia M, Coll-Rosell E. Value of sonography for follow-up of
mediastinal lymphadenopathy in children with tuberculosis. J Clin Andronikou S, Jamieson-Luff N, Grobusch MP, Zar HJ. Chest
Ultrasound. 2007;35:118–124. ultrasound findings in children with suspected pulmonary
11. Moseme T, Andronikou S. Through the eye of the suprasternal tuberculosis. Pediatric Pulmonology. 2019;1–8.
notch: point-of-care sonography for tuberculous mediastinal
https://doi.org/10.1002/ppul.24230
lymphadenopathy in children. Pediatr Radiol. 2014;44:
681–684.