Downloaded from http://thorax.bmj.com/ on September 14, 2016 - Published by group.bmj.

com

Tuberculosis

A simple, valid, numerical score for grading chest

x-ray severity in adult smear-positive
pulmonary tuberculosis
Anna P Ralph,1,2 Muhamed Ardian,3,4,5 Andri Wiguna,4,5 Graeme P Maguire,6
Niels G Becker,2 Glen Drogumuller,7 Michael J Wilks,8 Govert Waramori,4
Emiliana Tjitra,9 Sandjaja,9 Enny Kenagalem,3,10 Gysje J Pontororing,10
Nicholas M Anstey,1,11 Paul M Kelly1,2
1

Global Health Division, Menzies

School of Health Research and
Charles Darwin University,
Australia
2
National Centre for
Epidemiology and Population
Health Research, College of
Medicine, Biology and
Environment, Australian
National University, Australia
3
District Health Authority,
Timika, Papua Province,
Indonesia
4
International SOS, Timika,
Papua Province, Indonesia
5
Public Health & Malaria Control
Department, PT Freeport
Indonesia, Papua Province,
Indonesia
6
School of Medicine and
Dentistry, James Cook
University, Queensland,
Australia
7
Department of Radiology, Royal
Darwin Hospital, Northern
Territory, Australia
8
Radiology Department, Royal
Adelaide Hospital, North
Terrace, Adelaide, South
Australia
9
National Institute of Health
Research and Development,
Jakarta, Indonesia
10
Menzies School of Health
Research-National Institute of
Health Research and
Development Research
Program, and District Ministry of
Health, Timika, Papua Province,
Indonesia
11
Division of Medicine, Royal
Darwin Hospital, Darwin,
Australia
Correspondence to
Dr Anna Ralph, International
Health Division, Menzies School
of Health Research, PO Box
41096, Casuarina, Northern
Territory, 0811, Australia;
[email protected]
Received 8 February 2010
Accepted 30 June 2010

ABSTRACT
Background The grading of radiological severity in
clinical trials in tuberculosis (TB) remains unstandardised.
The aim of this study was to generate and validate
a numerical score for grading chest x-ray (CXR) severity
and predicting response to treatment in adults with
smear-positive pulmonary TB.
Methods At a TB clinic in Papua, Indonesia, serial CXRs
were performed at diagnosis, 2 and 6 months in 115
adults with smear-positive pulmonary TB. Radiographic
findings predictive of 2-month sputum microscopy status
were used to generate a score. The validity of the score
was then assessed in a second data set of 139
comparable adults with TB, recruited 4 years later at the
same site. Relationships between the CXR score and
other measures of TB severity were examined.
Results The estimated proportion of lung affected and
presence of cavitation, but not cavity size or other
radiological findings, significantly predicted outcome and
were combined to derive a score given by percentage of
lung affected plus 40 if cavitation was present. As well
as predicting 2-month outcome, scores were
significantly associated with sputum smear grade at
diagnosis (p<0.001), body mass index, lung function,
haemoglobin, exercise tolerance and quality of life
(p<0.02 for each). In the validation data set, baseline
CXR score predicted 2-month smear status significantly
more accurately than did the proportion of lung affected
alone. In both data sets, CXR scores decreased over time
(p<0.001).
Conclusion This simple, validated method for grading
CXR severity in adults with smear-positive pulmonary TB
correlates with baseline clinical and microbiological
severity and response to treatment, and is suitable for
use in clinical trials.

INTRODUCTION
Sputum smear microscopy, and culture where
available, are standardised modalities for diagnosing
and monitoring treatment response in pulmonary
tuberculosis (TB). Chest radiography (CXR)
provides useful information regarding disease extent
and progress, but there is no agreed-upon, validated
system for grading the severity of CXR abnormalities in bacteriologically proven pulmonary TB.
Several methods were devised for this purpose at the
time of early TB treatment trials, such as those
described by the Madras TB Chemotherapy Centre

Thorax 2010;65:863e869. doi:10.1136/thx.2010.136242

in 1960,1 Simon in 19662 and the National TB and

Respiratory Disease Association of the USA in 1969.3
Despite this, no system has been validated in
predicting outcome in more than one patient population. Recent randomised controlled trials
(RCTs)4e7 and observational studies in adults with
TB8e13 illustrate this lack of standardisation in the
grading of radiological severity, with each of these
studies utilising different non-validated investigatorgenerated systems to grade CXR severity.
The same problem of non-standardised radiological reporting has been recently articulated by
Dawson et al relating to TB screening, who evaluated and recommended the Chest Radiographic
Reading and Reporting System14 for TB screening
in HIV-positive people.15 However, this and other
screening tools16e18 seek to identify the presence of
latent TB infection or active disease, and are not
useful for researchers wishing to accurately document
severity or response to treatment in active TB.
Problems in CXR reporting arise from the heterogeneous CXR manifestations of pulmonary TB (eg,
in primary vs postprimary disease, adults vs children,
immunocompetent vs immunocompromised)19e21
and to inaccuracies inherent in CXR performance
and interpretation,2 including limited interobserver
agreement on CXR ndings.22 23 Despite these
shortcomings, the utility of CXR is well established
in TB diagnosis and clinical monitoring.
Associations between radiological extent and
other measures such as forced expiratory volume in
1 s (FEV1), age or multidrug-resistant (MDR)-TB
have previously been identied,8 24 but a standard,
simple, numerical score, validated against TB
outcome, in repeated data sets, is lacking. We
therefore aimed to devise a simple CXR score for
use in adults with smear-positive pulmonary TB,
which predicts outcome and correlates with
bacteriological and clinical severity markers, for the
purpose of grading severity and monitoring treatment response in the context of TB clinical trials.
We then determined the utility of the score in
a separate, comparable patient population.

METHODS
Study setting
The study was conducted at a community-based
TB clinic in Timika, Papua Province, Indonesia.
Timika has population of w200 000 and an
estimated TB incidence of 311/100 000.25
863

Downloaded from http://thorax.bmj.com/ on September 14, 2016 - Published by group.bmj.com

Tuberculosis
Participants
Adults (>15 years) diagnosed with sputum smear-positive
pulmonary TB who gave written informed consent were eligible
for enrolment in the study. Study participants were recruited
during two time periods: 2003e2004 (the derivation data set)
and 2008e2009 (the validation data set). The demographic,
clinical and microbiological ndings and outcomes in the rst
data set have been reported previously.26 27

Chest radiography
Standard full-size posteroanterior CXR were performed at the
time of TB diagnosis and 2 and 6 months thereafter, with
reports provided by a clinician at the eld site (rst data set,
PMK; second data set, APR) and, additionally for the rst data
set, by one of two radiologists (MJW or GD). During the rst
data collection period, the presence of small (1e2 mm) or large
(>2 mm) nodules, patchy or conuent consolidation, cavitation,
bronchial lesions or brosis was reported for each of three zones
(upper, mid or lower zones) in each lung. The presence of effusion or lymphadenopathy was reported, the total percentage of
each lung affected by any pathology was estimated, total cavity
size in millimetres was recorded and the effusion volume
(percentage of lung eld) was estimated. To grade the percentage
of affected lung, visual estimation of the extent of opacication,
cavitation or other pathology as a percentage of visible lung was
made; dense opacication of a zone was graded as 100% of that
zone, while patchy opacication within a zone attracted scores
<100% depending on the extent of opacication. Other remarks
including presence of miliary disease were recorded. During the
second data collection period, a simplied CXR report method
was used (percentage lung affected, cavitation (0, <4 cm,
$4 cm), effusion (0, <25%, $25% of hemithorax), presence of
consolidation, brosis, nodules, miliary disease). Reporters were
blinded to HIV status, bacteriological and clinical parameters
and treatment outcome.

Sputum microscopy and clinical evaluations

Baseline sputum microscopy was performed at the onsite laboratory and repeated at the reference laboratory on samples
collected at 0, 2 and 6 months, and the density of acid-fast bacilli
(AFB) was graded as 1, 2 or 3+ according to standard protocols.26 27 Baseline and follow-up evaluations included: body mass
index (BMI), FEV1 (spirometry performed using ML3535C,
MicroLoop, MicroMedical, Chatham, UK), haemoglobin (Hb),
measured using point-of-care HemoCue (ngelholm, Sweden) or
iSTAT (Abbott Park, Illinois, USA) tests, 6 min walk test
(distance walked in 6 min on a straight walking track), measured
according to American Thoracic Society guidelines, and
St Georges Respiratory Questionnaire (SGRQ) modied to
reect local conditions and translated into Indonesian.27 28
Standard denitions were used for nutritional category (normal,
mild malnutrition, moderate malnutrition or severe malnutrition)
according to BMI,29 and for TB treatment outcome at 6 months
(cured, completed, transferred, defaulted, failed or died).30
Impairment in FEV1 as a percentage of predicted values was
calculated using previously established local reference ranges.31

Outcome measure
The outcome measure used in this study is 2-month sputum
AFB microscopy status. Two-month smear positivity has been
previously shown to predict unfavourable outcomes including
treatment failure and death,32e34 and determines the need for
continued intensive-phase treatment versus switching to
continuation-phase therapy.30 Although an imperfect predictor
864

of outcome,35 in the absence of suitable alternatives, it remains

a commonly used surrogate end point.

Data analysis
Statistical calculations were performed using Intercooled Stata
10.1 (StataCorp, College Station, Texas, USA); graphs were
created in GraphPad Prism 5 (GraphPad, La Jolla, California, USA).
Statistical tests were two sided, with a p value of <0.05 indicating
statistical signicance. Intergroup differences in means or medians
were compared using two-sample t tests, Wilcoxon rank sum
tests, analysis of variance or KruskaleWallis tests as appropriate.
Agreement between reporters in the derivation data set was
tested using the concordance coefcients, rc for continuous
variables or the kappa statistic for categorical variables. Prevalence-adjusted, bias-adjusted kappa values were calculated
according to the method described by Byrt et al.36 Kappa values
were interpreted according to guidelines given by Landis and
Koch37 (kappa #0.00, poor; 0.00e0.20, slight; 0.21e0.40, fair;
0.41e0.60, moderate; 0.61e0.80, substantial; 0.81e1.00, almost
perfect).
The relationships between radiographic ndings and clinical
outcome were examined by multivariable regression analysis,
using a forward stepwise approach in which any radiological
variable found to be signicant (p<0.05) in univariate analysis
was included in the initial model. Goodness of t of nal models
was assessed using the HosmereLemeshow test and compared
using the likelihood ratio test. The weighting for a numerical
radiological score was derived from the regression coefcients.
Its ability to predict outcome in the validation data set was
determined using receiver operator characteristics (ROC; area
under the curve (AUC)). The relationships between this score
and demographic, biological and clinical variables were determined in data sets 1 and 2 using regression models using the
same principles.

Ethics
Approval was granted by the ethics committees of the National
Institute of Health Research and Development (Jakarta,
Indonesia), Menzies School of Health Research (Darwin,
Australia) and the Australian National University (Canberra,
Australia). Written informed consent was obtained from
participants in Indonesian or an appropriate Papuan language.

RESULTS
Characteristics of study participants in the two data collection
phases are shown in table 1. All participants had smear-positive
pulmonary TB ($2 AFB smear-positive sputum samples); the
result of an additional sample provided for microscopy and
culture on the day of treatment commencement is reported here.
This was negative in 5.7% and 7.2% of participants in the two
data sets, respectively despite their previous samples being
positive. Initial smear grade predicted the likelihood of smear
conversion by 2 months. In the derivation data set, failure to
convert to smear negative by 2 months was observed in 60.9% of
patients with a baseline smear grade of 3 and in 38.7% of
patients with a baseline smear grade of <3+ (p0.051). In the
validation data set, failure to convert to smear negative by
2 months was observed in 48.4% of patients with a baseline
smear grade of 3 and in 11.8% of patients with a baseline smear
grade of <3+ (p<0.001).
CXR reports were available at baseline, 2 and 6 months for
112, 76 and 76 study participants in the rst data set, and 136,
93 and 76 study participants in the second data set (incomplete
in the second data set as 30 of 139 had not yet completed
Thorax 2010;65:863e869. doi:10.1136/thx.2010.136242

Downloaded from http://thorax.bmj.com/ on September 14, 2016 - Published by group.bmj.com

Tuberculosis
Table 1

Study participant characteristics

Demographic details
Number
Age in years: median (range)
Female gender, n (%)
Papuan ethnicity, n (%)
Smokers, n (%)
HIV positive: no./no. tested (%)
MDR-TB, n (%)
Baseline clinical findings
BMI, median (range), kg/m2
Haemoglobin, median (range), g/dl
FEV1: median (range), litres
SGRQ total score, median (range)
6 min walk distance, median (range) m
Sputum AFB smear grade at diagnosis n
0*
Scanty or 1+
2+
3+
2-month smear status n (%)
Positive
Negative
No result available
6-month outcome n (%)
Cured/completed
Died
Failed
Default
Transferred
6 months not yet completed

Table 2

Derivation data set

Validation data set

115
30 (17e69)
33 (28.7)
57 (49.6)
38 (33.0)
5/112 (4.5)
2 (1.7)

139
27 (15e65)
48 (34.5)
66 (47.5)
41 (29.5)
16/121 (13.2)
2 (1.4)

18.6 (14.2e25.2)
11.2 (6.8e18.0)
1.76 (0.49e4.12)
45.3 (2.5e83.5)
405 (185e625)
(%)
6 (5.7)
25 (23.6)
28 (26.4)
47 (44.3)

19.0 (12.9e32.5)
12.2 (7.1e16.0)
1.70 (0.59e3.56)
37.8 (5.2e91.9)
410 (20e612)

25 (21.8)
81 (70.4)
9 (7.8)

31 (22.3)
95 (68.4)
13 (9.3)

88 (76.5)
3 (2.6)
2 (1.7)
13 (11.3)
9 (7.8)
0

92 (66.2)
2 (1.4)
1 (0.7)
8 (5.8)
6 (4.3)
30 (21.6)

10
65
35
29

(7.2)
(45.7)
(25.2)
(20.9)

*All study participants had at least two prior smear-positive sputum samples; some are
reported as negative since this result pertains to the additional spot specimen provided at
enrolment into the study (see the methods section).
AFB, acid-fast bacilli; BMI, body mass index; FEV1, forced expiratory volume in 1 s;
MDR-TB, multidrug-resistant tuberculosis; SGRQ, St Georges Respiratory Questionnaire.

6 months) (table 2). Reasons for missing CXR included patient

failure to attend (died, defaulted or transferred prior to
appointment), inability to obtain CXR (eg, electricity failure),
CXR date >3 weeks before or after the due follow-up date, or
CXR unavailable for reporting.

Agreement on radiological abnormalities

Agreement between reporters on radiological abnormalities in
the derivation data set is shown in table 3. Agreement was
relatively low overall. More substantial agreement was achieved
for some variables after adjusting kappa values for variable
prevalence and reporter bias.

Development of score using the training data set

Two-month sputum smear status in the initial data set (n115)
was signicantly predicted in univariate logistic regression
models by the presence of baseline cavitation (OR 3.26, 95% CI
1.11 to 9.56) and the total percentage of lung affected (OR 1.9,
95% CI 1.3 to 2.7, calculated per each 20% increment of affected
lung), but not by cavitation size, presence or number of nodules,
brosis, effusion or lymph nodes (table 4).
The relationships between those radiological ndings which
were independently predictive of 2-month outcome (cavitation
and percentage of lung affected) and baseline clinical and
bacteriological measures were then examined. Cavitary disease
on CXR at TB diagnosis was signicantly associated with higher
baseline AFB density in sputum (ie, smear microscopy grade)
(p0.007, c2 test for trend), and people with cavitary disease
Thorax 2010;65:863e869. doi:10.1136/thx.2010.136242

Chest radiograph (CXR) results

Derivation
data set

Number of CXR available for report

Baseline
112
2 months
76
6 months
76
Baseline radiological findings
Consolidation, n (%)
106 (94.6)
Cavitation, n (%)
71 (63.4)
Effusion, n (%)
37 (33.0)
Miliary, n (%)
2 (1.7)
Nodules, n (%)
83 (72.2)
Fibrosis, n (%)
62 (55.3)
Proportion of lung affected, median % (range)
35 (0e100)
CXR weighted score characteristics
Baseline score, median (range)
65 (5e130)
2 month score, median (range)
35 (0e130)
6 month score, median (range)
10 (0e105)
Median score decrease in 2 months, %
27.3
Median score decrease in 6 months, %
73.1
Baseline score by smear grade at diagnosis median (range)
0
15 (0.5e70)
Scanty or 1+
50 (0e110)
2+
70 (6.5e135)
3+
80 (10e140)

Validation
data set
136
93
76*
104 (76.5)
77 (56.6)
24 (17.3)
5 (3.7)
29 (21.3)
10 (7.4)
41.5 (0e100)
69 (0e140)
29 (0e140)
10 (0e115)
38.8
80.2
10
66
76
85

(0e133)
(0e140)
(11e140)
(4e140)

*30 patients still under follow up.

had worse lung function, with a mean percentage predicted

FEV1 of 59.0 (95% CI 54.4 to 63.6) in cavitary disease versus 68.7
(95% CI 60.6 to 76.7) in non-cavitary disease (p0.03,
two-sample t test). People with cavitary disease had slightly
lower BMI (18.5, 95% CI 18.0 to 19.1) compared with those
with non-cavitary disease (19.2 kg/m2, 95% CI 18.5 to 20.0), but
this difference was not statistically signicantly (mean betweengroup difference 0.70 kg/m2, 95% CI 0.18 to 1.59). No significant associations were identied between cavitary disease and
exercise tolerance (6 min walk distance), quality of life (SGRQ
total or individual domain scores) or Hb.
The amount (%) of lung affected signicantly predicted all
clinical and laboratory variables. Specically, greater proportions
of affected lung were signicantly associated with decreasing BMI
category (p0.002, KruskaleWallis test), lung function category
Table 3

Agreement on radiological findings, derivation data set

Inter-rater agreement
among dichotomous
variables

Kappa

Prevalenceadjusted,
bias-adjusted
kappa

Presence
Presence
Presence
Presence
Presence
Presence
Presence
Presence
Presence
Presence

0.20
0.33
0.24
0.12
0.09
0.19
0.08
0.33
-0.35
0.01

0.70
0.31
0.81
0.06
0.09
0.19
0.06
0.37
-0.61
-0.09

of
of
of
of
of
of
of
of
of
of

patchy consolidation
confluent consolidation
any consolidation
small nodules
large nodules
any nodules
fibrosis
cavitation
effusion
lymphadenopathy

Interpretation
of prevalenceadjusted,
bias-adjusted
kappa
Substantial
Fair
Almost perfect
Slight
Slight
Slight
Slight
Fair
Poor
Poor

Concordance among
continuous variables

rc

95% limits of agreement

(Bland and Altman)

Total amount (%) of lung affected

Extent of cavity size (mm)

0.85
0.69

28.20 to 22.46 %
56.62 to 50.66 mm

865

Downloaded from http://thorax.bmj.com/ on September 14, 2016 - Published by group.bmj.com

Tuberculosis
Table 4 Relationship between radiological and biological parameters
and 2-month sputum acid-fast bacilli density in the derivation data set,
showing results of univariate logistic regression analyses*
Radiological parameters
Presence of any consolidation
Per 1% increments
Per 20% increments
Cavitary disease
Effusion
Nodules
Fibrosis
Lymphadenopathy

Odds ratio

95% CI

1.30
1.90
3.26
1.50
1.17
2.23
1.43

0.14
1.30
1.11
0.59
0.41
0.86
0.57

to
to
to
to
to
to
to

p Value

12.18
2.68
9.56
3.82
3.32
5.79
3.56

0.001
0.001
0.032
0.590
0.773
0.097
0.448

*OR was unable to be calculated for the two instances of military disease (both had
negative sputum at 2 months).

(p<0.001, KruskaleWallis test), 6 min walk distance (00.001,

linear regression) and Hb in males (p0.003, linear regression),
though not in females (p0.4). A greater proportion of lung
affected on the baseline CXR was also signicantly associated
with SGRQ total scores (p<0.001, linear regression) and with
sputum smear grade at diagnosis (p<0.001, KruskaleWallis test).
To create a CXR score, cavitation and percentage of lung
affected were included as independent variables in a logistic
regression model for 2-month sputum smear status. The model
containing both variables was signicantly better than the model
containing cavitaion alone (likelihood ratio test p<0.001) or
proportion of lung affected alone (likelihood ratio test p0.016)
at predicting 2-month sputum smear status. Regression coefcients were 0.03167 for proportion of lung affected and 1.26151
for presence/absence of cavitation, indicating a relative weighting
of 40.27 for cavitation (1.26151O0.03167), thereby generating an
equation for the weighted score as follows:
CXR score proportion of total lung affected%
40 if cavitation present

CXR score results

CXR score characteristics are shown in table 2 and gures 1e4.
Scores did not signicantly differ according to sex, ethnicity or
smoking status (p>0.05, two-sample t tests), and were not

Figure 1 Chest x-ray (CXR) score at enrolment according to sputum

acid-fast bacilli (AFB) grade at diagnosis and 2 months (derivation data set).
866

Figure 2 Chest x-ray (CXR) score according to body mass index (BMI),
percentage of predicted forced expiratory volume in1 s (FEV1) and
haemoglobin (Hb) at diagnosis (derivation data set).
signicantly associated with age in univariate or multivariate
analyses. Mean baseline CXR score in people with unfavourable
(positive) 2-month outcomes was signicantly higher (88.2; 95%
CI 76.5 to 99.9) than in those with a favourable outcome (56.8;
95% CI 49.7 to 64.0), but the range of scores in each smear grade
was wide (gure 1). Scores were also signicantly associated
with baseline microscopy grade (gure 1). CXR scores were
inversely related to BMI, FEV1, Hb and 6 min walk distance,
were directly related to SGRQ total score (higher SGRQ scores
indicate worse quality of life) and signicantly decreased over
time (gures 2e4).

Performance of score using the validation data set

The weighted score calculated for the new data set showed
similar characteristics (table 2), including a median baseline
score of 69, no signicant relationship with demographic factors,
signicant positive association with baseline smear grade
(p0.009, KruskaleWallis test) and the same relationships as

Figure 3 Chest x-ray (CXR) score according to St Georges Respiratory

Questionnaire (SGRQ) (total score and 6 minute walk distance at
tuberculosis diagnosis (derivation data set).
Thorax 2010;65:863e869. doi:10.1136/thx.2010.136242

Downloaded from http://thorax.bmj.com/ on September 14, 2016 - Published by group.bmj.com

Tuberculosis
Figure 4 Chest x-ray (CXR) score at
diagnosis, 8 weeks (end of intensive
treatment phase) and 24 weeks (end of
treatment).

were found in the initial data set between CXR score and each of
the clinical/laboratory measures (BMI, Hb, FEV1, SGRQ total
score and 6 min walk distance; p<0.05 in each case).
Comparing ROC scores to predict outcome, the weighted
CXR score (AUC 0.75) was signicantly better at predicting
2-month smear status than the percentage lung affected alone
(ROC 0.69; p0.013, c2 test; gure 5). The optimal cut-off point
for weighted CXR score (value furthest from the diagonal) was
71, at which value the sensitivity for predicting a positive
sputum smear status at 2 months was 80% (95% CI 61.4 to
92.3) and specicity 67.7% (95% CI 57.3 to 77.1). Comparative
sensitivity and specicity values are shown in table 5.

DISCUSSION

0.50
0.00

0.25

Sensitivity

0.75

1.00

The current need for a universal and standard system for

reporting CXR in pulmonary TB is acknowledged.38 In order to
grade CXR severity and assess radiological treatment response,
we have derived a simple equation from radiographic parameters
from adults with smear-positive pulmonary TB that predicts
smear positivity at 2 months and provides a single numerical

0.00

0.25

0.50

0.75

1.00

1-Specificity
%Lung affected ROC area: 0.6867

Reference

CXR Score ROC area: 0.7471

Optimal cut-point

Figure 5 Receiver operator characteristics (ROC) for weighted chest

x-ray (CXR) score (% lung affected + 40 if cavitation present) (solid line)
compared with percentage of lung affected (dashed line) as predictors of
2-month smear status (positive or negative). Optimal cut-off point for
score (circle) 71.
Thorax 2010;65:863e869. doi:10.1136/thx.2010.136242

score for each CXR. The score shows good correlation with
baseline bacteriological and clinical severity markers, and is
sensitive to changes over time. The score performs better than
its individual components: it was signicantly better at
predicting outcome than was the percentage of lung affected
alone, and was signicantly associated with a broader range of
baseline severity measures (BMI, |Hb, exercise tolerance and
quality of life) than presence of cavitation alone. Advantages of
this method are that CXR assessment does not require aids, grids
or rulers, and it is derived by tting a statistical model to
outcome data rather than by assigning points based on assumed
relative importance of radiographic pathologies. It has been

Table 5 Sensitivity and specificity of baseline radiological and

bacteriological findings in predicting outcome (2-month sputum acid-fast
bacilli (AFB) status)
Sensitivity
Radiological parameters at diagnosis
Derivation data set
Total amount of lung affected
20%
96
37.5% (optimal cut-off
84
point)
80%
20
Cavitation
80
Score
20
100
72.5 (optimal cut-off point)
80
130
8
Validation data set
Total amount of lung affected
20%
96.7
42.0% (optimal cut-off
73.3
point)
80%
30
Cavitation
83.3
Score
20
96.7
71 (optimal cut-off point)
80
130
10
Bacteriological parameters: baseline smear grade
Derivation data set
AFB density*
60.9
Validation data set
AFB density *
48.4

Specificity

35.9
68.0
89.7
44.9
18.0
71.2
98.7

22.6
62.4
87.1
52.7
22.6
67.7
91.4

61.3
88.2

*Baseline AFB density (smear grade) dichotomised as #2 or 3.

867

Downloaded from http://thorax.bmj.com/ on September 14, 2016 - Published by group.bmj.com

Tuberculosis
validated in an independent data set, and offers a single, standardised solution where there are currently multiple unvalidated
methods in use.
The proportion of lung affected and/or cavitation feature as
the most important measures in many TB CXR grading methods.1e5 7 Cavitation is well recognised to correlate with bacillary
load.7 39 We conrmed the association between cavitation and
bacteriological measures (baseline and 2-month sputum smear
status), and additionally showed cavitary disease to be predictive
of worse lung function. The proportion of lung affected was
associated with both bacteriological and a range of clinical
measures.
This score was derived in adult patients with TB with smearpositive pulmonary disease, in a setting with relatively low rates
of HIVeTB co-infection and MDR-TB. The score requires
further evaluation in populations with high HIV prevalence, in
whom CXR ndings characteristic of HIVeTB co-infection
(subtle or absent pathology, non-cavitary disease, lower lobe
inltrates, hilar lymphadenopathy and pleural effusion)20 40 may
mean that a differently weighted score is needed. Nevertheless,
the score remained valid and applicable in the newer data set in
which HIVeTB co-infection rates were higher (13%); the rise in
HIV prevalence may account for some of the differences
observed between the two data sets. The presence of MDR-TB
would not be expected to alter radiographic patterns, other than
being associated potentially with higher scores and smaller
incremental improvements over time.
Potential limitations of the study include the use of 2 month
smear status as an outcome measure (rather than a longer term
measure such as 6-month outcome or recurrence).
The absence of suitable biomarkers or other surrogate end
points in TB research is readily acknowledged, and recent
estimates derived from meta-analysis found a sensitivity of only
57% and specicity of 81% for 2-month smear status in
predicting treatment failure.35 Nevertheless, until more suitable
measures become available, 2-month smear status remains
a suitable outcome measure.30 32e34
Another limitation was the inherent problem of limited
inter-rater agreement in CXR assessment. The low rates of
clinicianeradiologist agreement between reporters on CXR
ndings identied in the derivation data set are not unusual, with
only fair or poor agreement between radiologists and clinicians
also being reported elsewhere.22 23 This emphasises the importance of using simple rather than complex scores and ensuring
individuals allocating CXR scores participate in continuing
education to maximise agreement. The score derived from radiologist CXR evaluation in the rst data set is simple. Moreover, it
was shown to be valid in the second data set when used by an
independent TB clinician, rather than a radiologist, conrming its
practical utility in a clinical and trial setting. Some systematic
differences in CXR results were noted between the two data sets;
while this may represent systematic difference in reporting
styles, the ndings are in keeping with the possibility of less
severe disease in the validation data set, as indicated by their
lower bacillary burden (with slides read by the same senior
laboratory technician during both data collection periods).

data did not indicate this. This method can be used where
a numerical score is required for the purpose of comparing
radiographic severity between adults with smear-positive
pulmonary TB, and to monitor an individuals improvement
over time, such as in clinical trials of drug efcacy in TB.
Acknowledgements We thank the following for their support and assistance:
Dr M Okoseray, Pak Penias and Pak E Meokbun and the Timika District Health
Authority; Dr Dina Bisara Lolong and Ibu Meryani Girsang and the National Institute
of Health Research and Development, Jakarta; Dr P Penttinen, Dr M Bangs and
Dr M Stone, Public Health & Malaria Control (PHMC) and International SOS; Pak
Istanto and PHMC laboratory staff; Pak J Lempoy and Timika TB clinic staff;
Dr P. Sugiarto and Mimika Community Hospital (RSMM); Natalia Dwi Haryanti,
Sri Hasmunik, Sri Rahayu, G Bellatrix and clinical and laboratory staff, NIHRD-MSHR
Timika research programme; Mr R Lumb and Dr I Bastian at the Institute of Medical
and Vetinerary Science; and Associate Professor R Price, MSHR.
Funding Australian Respiratory Council, the Royal Australasian College of Physicians
(Covance award), Australian National Health and Medical Research Council.
Competing interests None.
Ethics approval This study was conducted with the approval of the Human Research
Ethics Committees of the NT Department of Health & Families and Menzies School of
Health Research, Australia, the Australian National University, and the National
Institute for Health Research and Development, Indonesia.
Provenance and peer review Not commissioned; externally peer reviewed.

REFERENCES
1.
2.
3.

4.
5.
6.
7.

8.
9.
10.
11.
12.
13.
14.
15.

CONCLUSION
In summary, we have derived and validated a simple method for
grading CXR severity in adults with smear-positive pulmonary
TB that predicts baseline clinical and microbiological severity
and response to treatment in two separate patient populations.
Although ner discriminatory accuracy might be achieved by
collecting more detailed CXR ndings (such as cavity size), our
868

16.
17.
18.

Tuberculosis Chemotherapy Centre Madras. A concurrent comparison of

isoniazid plus PAS with three regimens of isoniazid alone in the domiciliary treatment
of pulmonary tuberculosis in South India. Bull World Health Org 1960;23:535e85.
Simon G. Radiology in epidemiological studies and some therapeutic trials. BMJ
1966;2:491e4.
Falk A, OConnor J, Pratt P. Classification of pulmonary tuberculosis. In: Falk A,
OConnor JB, Pratt PC, et al, eds. Diagnostic standards and classification of
tuberculosis. Vol 12. New York, NY: National Tuberculosis and Respiratory Disease
Association, 1969:68e76.
Diacon AH, Pym A, Grobusch M, et al. The diarylquinoline TMC207 for
multidrug-resistant tuberculosis. N Engl J Med 2009;360:2397e405.
Burman WJ, Goldberg S, Johnson JL, et al. Moxifloxacin versus ethambutol in the
first 2 months of treatment for pulmonary tuberculosis. Am J Respir Crit Care Med
2006;174:331e8.
Katiyar SK, Bihari S, Prakash S, et al. A randomised controlled trial of high-dose
isoniazid adjuvant therapy for multidrug-resistant tuberculosis. Int J Tuberc Lung Dis
2008;12:139e45.
Benator D, Bhattacharya M, Bozeman L, et al. Rifapentine and isoniazid once
a week versus rifampicin and isoniazid twice a week for treatment of drugsusceptible pulmonary tuberculosis in HIV-negative patients: a randomised clinical
trial. Lancet 2002;360:528e34.
Heo EY, Chun EJ, Lee CH, et al. Radiographic improvement and its predictors in
patients with pulmonary tuberculosis. Int J Infect Dis 2009;13:e371e6.
Mankatittham W, Likanonsakul S, Thawornwan U, et al. Characteristics of
HIV-infected tuberculosis patients in Thailand. Southeast Asian J Trop Med Public
Health 2009;40:93e103.
Lawson L, Yassin MA, Thacher TD, et al. Clinical presentation of adults with
pulmonary tuberculosis with and without HIV infection in Nigeria. Scand J Infect Dis
2008;40:30e5.
Idh J, Westman A, Elias D, et al. Nitric oxide production in the exhaled air of patients
with pulmonary tuberculosis in relation to HIV co-infection. BMC Infect Dis
2008;8:146.
Thorson A, Long NH, Larsson LO. Chest X-ray findings in relation to gender and
symptoms: a study of patients with smear positive tuberculosis in Vietnam.
Scand J Infect Dis 2007;39:33e7.
Liam CK, Lim KH, Srinivas P, et al. Hypercalcaemia in patients with newly diagnosed
tuberculosis in Malaysia. Int J Tuberc Lung Dis 1998;2:818e23.
Den Boon S, Bateman ED, Enarson DA, et al. Development and evaluation of a new
chest radiograph reading and recording system for epidemiological surveys of
tuberculosis and lung disease. Int J Tuberc Lung Dis 2005;9:1088e96.
Dawson R, Masuka P, Edwards DJ, et al. Chest radiograph reading and recording
system: evaluation for tuberculosis screening in patients with advanced HIV.
Int J Tuberc Lung Dis 2010;14:52e8.
Gatner EM, Burkhardt KR. Correlation of the results of X-ray and sputum culture in
tuberculosis prevalence surveys. Tubercle. 1980;61:27e31.
Linh NN, Marks GB, Crawford AB. Radiographic predictors of subsequent
reactivation of tuberculosis. Int J Tuberc Lung Dis 2007;11:1136e42.
World Health Organization. Assessing tuberculosis prevalence through
population-based surveys. WHO, Western Pacific Region 2007 [October 2009].
http://www.wpro.who.int/publications/PUB_978+92+9061+314+5.htm.

Thorax 2010;65:863e869. doi:10.1136/thx.2010.136242

Downloaded from http://thorax.bmj.com/ on September 14, 2016 - Published by group.bmj.com

Tuberculosis
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.

Andreu J, Caceres J, Pallisa E, et al. Radiological manifestations of pulmonary

tuberculosis. Eur J Radiol 2004;51:139e49.
Raviglione MC, Narain JP, Kochi A. HIV-associated tuberculosis in developing
countries: clinical features, diagnosis, and treatment. Bull World Health Organ
1992;70:515e26.
Marais BJ, Gie RP, Schaaf HS, et al. A proposed radiological classification of
childhood intra-thoracic tuberculosis. Pediatr Radiol 2004;34:886e94.
Balabanova Y, Coker R, Fedorin I, et al. Variability in interpretation of chest
radiographs among Russian clinicians and implications for screening programmes:
observational study. BMJ 2005;331:379e82.
Zellweger JP, Heinzer R, Touray M, et al. Intra-observer and overall agreement in
the radiological assessment of tuberculosis. Int J Tuberc Lung Dis 2006;10:1123e6.
Plit ML, Anderson R, Van Rensburg CE, et al. Influence of antimicrobial
chemotherapy on spirometric parameters and pro-inflammatory indices in severe
pulmonary tuberculosis. Eur Respir J 1998;12:351e6.
Ardian M, Meokbun E, Siburian L, et al. A public-private partnership for TB control in
Timika, Papua Province, Indonesia. Int J Tuberc Lung Dis 2007;11:1101e7.
Kelly PM, Ardian M, Waramori G, et al. A community-based TB drug susceptibility
study in Mimika District, Papua Province, Indonesia. Int J Tuberc Lung Dis
2006;10:167e71.
Maguire GP, Anstey NM, Ardian M, et al. Pulmonary tuberculosis, impaired lung
function, disability and quality of life in a high-burden setting. Int J Tuberc Lung Dis
2009;13:1500e6.
Jones PW, Quirk FH, Baveystock CM, et al. A self-complete measure of health
status for chronic airflow limitation. The St. Georges Respiratory Questionnaire.
Am Rev Respir Dis 1992;145:1321e7.
World Health Organization. Management of severe malnutrition: a manual for
physicians and other senior health workers. Geneva: 1999[December 2009].
http://whqlibdoc.who.int/hq/1999/a57361.pdf.

Thorax 2010;65:863e869. doi:10.1136/thx.2010.136242

30.
31.
32.
33.
34.

35.
36.
37.
38.
39.
40.

World Health Organization. Treatment of tuberculosis: guidelines for national

programmes. WHO/CDS/TB/2003.313. Geneva, Switzerland: WHO, 2003.
Handojo T, Anstey N, Kelly P, et al. Normal spirometry, gas transfer and lung volume
values in Papua, Indonesia. Southeast Asian J Trop Med Public Health
2006;37:571e7.
Wang JY, Lee LN, Yu CJ, et al. Factors influencing time to smear conversion in
patients with smear-positive pulmonary tuberculosis. Respirology 2009;14:1012e19.
Rieder HL. Sputum smear conversion during directly observed treatment for
tuberculosis. Tuber Lung Dis. 1996;77:124e9.
Ramarokoto H, Randriamiharisoa H, Rakotoarisaonina A, et al. Bacteriological
follow-up of tuberculosis treatment: a comparative study of smear microscopy and
culture results at the second month of treatment. Int J Tuberc Lung Dis
2002;6:909e12.
Horne DJ, Royce SE, Gooze L, et al. Sputum monitoring during tuberculosis
treatment for predicting outcome: systematic review and meta-analysis.
Lancet Infect Dis 2010;10:387e94.
Byrt T, Bishop J, Carlin JB. Bias, prevalence and kappa. J Clin Epidemiol
1993;46:423e9.
Landis JR, Koch GG. The measurement of observer agreement for categorical data.
Biometrics 1977;33:159e74.
Daley C. Chest radiography in the diagnosis of tuberculosis: specificity and accuracy.
39th World Conference on Lung Health of the International Union Against TB and Lung
Disease. Paris, France: 2008:16e20.
Palaci M, Dietze R, Hadad DJ, et al. Cavitary disease and quantitative sputum
bacillary load in cases of pulmonary tuberculosis. J Clin Microbiol 2007;45:4064e6.
Aderaye G, Bruchfeld J, Assefa G, et al. The relationship between disease
pattern and disease burden by chest radiography, M. tuberculosis load, and HIV
status in patients with pulmonary tuberculosis in Addis Ababa. Infection
2004;32:333e8.

869

Downloaded from http://thorax.bmj.com/ on September 14, 2016 - Published by group.bmj.com

A simple, valid, numerical score for grading

chest x-ray severity in adult smear-positive
pulmonary tuberculosis
Anna P Ralph, Muhamed Ardian, Andri Wiguna, Graeme P Maguire, Niels
G Becker, Glen Drogumuller, Michael J Wilks, Govert Waramori, Emiliana
Tjitra, Sandjaja, Enny Kenagalem, Gysje J Pontororing, Nicholas M
Anstey and Paul M Kelly
Thorax 2010 65: 863-869

doi: 10.1136/thx.2010.136242
Updated information and services can be found at:
http://thorax.bmj.com/content/65/10/863

These include:

References
Email alerting
service

Topic
Collections

This article cites 36 articles, 4 of which you can access for free at:
http://thorax.bmj.com/content/65/10/863#BIBL
Receive free email alerts when new articles cite this article. Sign up in the
box at the top right corner of the online article.

Articles on similar topics can be found in the following collections

Radiology (diagnostics) (737)
TB and other respiratory infections (0)
Tuberculosis (47)

Notes

To request permissions go to:

http://group.bmj.com/group/rights-licensing/permissions
To order reprints go to:
http://journals.bmj.com/cgi/reprintform
To subscribe to BMJ go to:
http://group.bmj.com/subscribe/