Content uploaded by William D-C Man
Author content
All content in this area was uploaded by William D-C Man on Nov 12, 2014
Content may be subject to copyright.
www.thelancet.com/respiratory Vol 2 March 2014
195
Articles
Minimum clinically important diff erence for the COPD
Assessment Test: a prospective analysis
Samantha S C Kon, Jane L Canavan, Sarah E Jones, Claire M Nolan, Amy L Clark, Mandy J Dickson, Brigitte M Haselden, Michael I Polkey,
William D-C Man
Summary
Background The COPD Assessment Test (CAT) is responsive to change in patients with chronic obstructive pulmonary
disease (COPD). However, the minimum clinically important diff erence (MCID) has not been established. We aimed
to identify the MCID for the CAT using anchor-based and distribution-based methods.
Methods We did three studies at two centres in London (UK) between April 1, 2010, and Dec 31, 2012. Study 1
assessed CAT score before and after 8 weeks of outpatient pulmonary rehabilitation in patients with COPD who
were able to walk 5 m, and had no contraindication to exercise. Study 2 assessed change in CAT score at discharge
and after 3 months in patients admitted to hospital for more than 24 h for acute exacerbation of COPD. Study 3
assessed change in CAT score at baseline and at 12 months in stable outpatients with COPD. We focused on
identifying the minimum clinically important improvement in CAT score. The St George’s Respiratory
Questionnaire (SGRQ) and Chronic Respiratory Questionnaire (CRQ) were measured concurrently as anchors.
We used receiver operating characteristic curves, linear regression, and distribution-based methods (half SD, SE
of measurement) to estimate the MCID for the CAT; we included only patients with paired CAT scores in the
analysis.
Findings In Study 1, 565 of 675 (84%) patients had paired CAT scores. The mean change in CAT score with pulmonary
rehabilitation was –2·5 (95% CI –3·0 to –1·9), which correlated signifi cantly with change in SGRQ score (r=0·32;
p<0·0001) and CRQ score (r=–0·46; p<0·0001). In Study 2, of 200 patients recruited, 147 (74%) had paired CAT
scores. Mean change in CAT score from hospital discharge to 3 months after discharge was –3·0 (95% CI –4·4 to –1·6),
which correlated with change in SGRQ score (r=0·47; p<0·0001). In Study 3, of 200 patients recruited, 164 (82%) had
paired CAT scores. Although no signifi cant change in CAT score was identifi ed after 12 months (mean 0·6,
95% CI –0·4 to 1·5), change in CAT score correlated signifi cantly with change in SGRQ score (r=0·36; p<0·0001).
Linear regression estimated the minimum clinically important improvement for the CAT to range between
–1·2 and –2·8 with receiver operating characteristic curves consistently identifying –2 as the MCID. Distribution-
based estimates for the MCID ranged from –3·3 to –3·8.
Interpretation The most reliable estimate of the minimum important diff erence of the CAT is 2 points. This estimate
could be useful in the clinical interpretation of CAT data, particularly in response to intervention studies.
Funding Medical Research Council and UK National Institute of Health Research.
Introduction
The COPD Assessment Test (CAT) is a simple, eight item,
health status instrument for patients with chronic
obstructive pulmonary disease (COPD), which is highly
practical,1 has good psychometric properties, and has been
shown to be responsive to pulmonary rehabilitation and
recovery from exacerbation.2–8 A decrease in CAT score
represents an improvement in health status, whereas an
increase in CAT score represents a worsening in health
status. The CAT has been incorporated into the Global
Initiative for Chronic Obstructive Lung Disease (GOLD)
combined assessment of COPD as a means of establishing
a symptomatic threshold to guide initiation of regular
pharmacological treatment.9 Because it takes only 2–3 min
to complete, the CAT has practical advantages over longer-
established health status questionnaires such as the St
George’s Respiratory Questionnaire (SGRQ) and the
Chronic Respiratory Questionnaire (CRQ).
The minimum clinically important diff erence (MCID)
is the smallest change in score that patients perceive as
benefi cial or detrimental, and is useful to aid the
clinical interpretation of health status data, particularly
in response to intervention. By contrast with other
health status questionnaires commonly used in COPD
such as the SGRQ and CRQ,10–12 to our knowledge the
MCID of the CAT has not been described.
We aimed to estimate the MCID for the CAT using a
range of anchor-based and distribution-based methods
in three diff erent clinical scenarios: response to
pulmonary rehabilitation, recovery after admission to
hospital for acute exacerbation of COPD, and
longitudinal change with time. We postulated that
the CAT score would improve with pulmonary
rehabilitation and recovery from admission to hospital
(ie, a decrease in CAT score), but worsen with time in
stable patients (ie, an increase in CAT score), and that
Lancet Respir Med 2014;
2: 195–203
Published Online
February 4, 2014
http://dx.doi.org/10.1016/
S2213-2600(14)70001-3
See Comment page 167
NIHR Respiratory Biomedical
Research Unit (S S C Kon MBBS,
J L Canavan PhD, S E Jones MSc,
C M Nolan BSc,
Prof M I Polkey PhD,
W D-C Man PhD), and Harefi eld
Pulmonary Rehabilitation Unit
(C M Nolan, A L Clark BSc,
W D-C Man), Royal Brompton &
Harefi eld NHS Foundation
Trust, London, UK; Imperial
College, London, UK (S S C Kon,
J L Canavan, S E Jones, C M Nolan,
Prof M I Polkey, W D-C Man);
and The Hillingdon Hospital
NHS Foundation Trust,
Uxbridge, UK (M J Dickson RGN,
B M Haselden PhD)
Correspondence to:
Dr Samantha S C Kon,
Department of Respiratory
Medicine, Harefi eld Hospital,
Harefi eld UB9 6JH, UK
s.kon@rbht.nhs.uk
Articles
196
www.thelancet.com/respiratory Vol 2 March 2014
change in CAT score would correlate signifi cantly with
change in other well-established COPD health status
questionnaires.
Methods
Participants
In Study 1, we recruited stable patients with COPD from
the Harefi eld Pulmonary Rehabilitation Unit (Harefi eld
Hospital, London, UK). Inclusion criteria included a
diagnosis of COPD, ability to walk 5 m, and no
contraindication to exercise.13 For Study 2, we recruited
patients who had an acute exacerbation of COPD
diagnosed by a physician and an admission longer than
24 h to acute wards at Hillingdon Hospital (London, UK).
In Study 3, we recruited stable patients with COPD from
outpatient clinics at Harefi eld Hospital; patients were
deemed stable if they had not had an exacerbation in the
4 weeks before the baseline measurement. Patients were
not excluded from Study 3 if they had an excerbation
during the 12 month follow-up period. Age 35 years or
older was an inclusion criterion for all studies. Patients
unable to read or understand English were excluded
from all studies. All participants gave written informed
consent and all studies received local ethics committee
approval.
Procedures
Study 1 assessed response of the CAT to pulmonary
rehabilitation. We measured the CAT, SGRQ, and CRQ
before and after an 8 week outpatient pulmonary
rehabilitation programme, consisting of twice weekly
supervised exercise and education sessions.6,12,14 We
measured the incremental shuttle walk, fi ve-repetition
sit-to-stand, and 4 m gait speed to assess change in
physical performance.15–17 Patients were kept masked to
their performance and were asked to rate their change in
health status after pulmonary rehabilitation using an
adapted fi ve point Global Rating of Change
Questionnaire (GRCQ).18 Patients were asked to classify
how they felt after pulmonary rehabilitation according to
fi ve responses: “1: much better”; “2: a little better”;
“3: no change”; “4: a little worse” and “5: much worse”.
Data from 255 of the 565 patients in Study 1 have been
used as a comparator group in a previous study.19
Study 2 assessed change in CAT score after admission
to hospital for acute exacerbation of COPD. Forced
expiratory volume in one second (FEV1), CAT, SGRQ, and
4 m gait speed were measured on the day of hospital
discharge and about 3 months later.
Study 3 assessed longitudinal change in CAT score
with time. Patients with COPD attending outpatient
respiratory clinics were asked to complete FEV1,
incremental shuttle walk, 4 m gait speed, CAT, and
SGRQ, and again at 12 months.
Statistical analysis
Study 1 was a pragmatic observational study in which we
aimed to prospectively recruit a minimum of 500 patients
with paired CAT measurements. We anticipated
25% dropout on the basis of our experience with the
pulmonary rehabilitation programme and therefore
Baseline Change after pulmonary
rehabilitation
p value
Mean age (years [SD]) 70 (9) ·· ··
Sex
Men 327 (58%) ·· ··
Women 238 (42%) ·· ··
Mean BMI (kg/m² [SD]) 27·6 (6·0) ·· ··
FEV1 (% predicted) 47·6 (45·9 to 49·3) ·· ··
MRC 3·4 (3·3 to 3·5) ·· ··
ISW (m) 210 (199 to 222) 53 (47 to 59) <0·0001
5STS (s) 15·3 (14·6 to 16·0) –2·4 (–3·9 to –1·9) <0·0001
4MGS (m s–1) 0·90 (0·88 to 0·93) 0·07 (0·06 to 0·08) <0·0001
SGRQ
Total 51·0 (49·3 to 52·6) –5·0 (–6·1 to –3·8) <0·0001
Symptoms 64·7 (62·7 to 66·6) –3·8 (–5·4 to –2·3) <0·0001
Activities 68·8 (66·7 to 70·8) –3·7 (–5·3 to –2·1) <0·0001
Impact 36·5 (34·7 to 38·3) –5·9 (–7·2 to –4·5) <0·0001
CRQ
Total 75·9 (74·1 to 77·7) 14·5 (11·1 to 17·9) <0·0001
Dyspnoea 13·9 (13·4 to 14·5) 4·7 (3·9 to 5·5) <0·0001
Fatigue 13·4 (13·0 to 13·8) 3·1 (2·5 to 3·7) <0·0001
Emotion 30·6 (29·9 to 31·4) 4·2 (3·2 to 5·3) <0·0001
Mastery 18·0 (17·5 to 18·4) 2·7 (2·1 to 3·4) <0·0001
CAT 21·4 (20·8 to 22·0) –2·5 (–3·0 to –1·9) <0·0001
Data are mean (95% CI) or n (%) unless otherwise specifi ed. BMI=body-mass index. FEV1=forced expiratory volume in
1 s. MRC=Medical Research Council dyspnoea score. ISW=incremental shuttle walk. 5STS=fi ve-repetition sit-to-stand.
4MGS=4 m gait speed. SGRQ=St George’s Respirator y Questionnaire. CRQ=Chronic Respiratory Questionnaire.
CAT=COPD Assessment Test.
Table 1: Baseline characteristics and response to pulmonary rehabilitation (Study 1; n=565)
Slope (SE) y intercept (SE) rp value
Change in SGRQ
Total 0·18 (0·03) –1·53 (0·35) 0·32 <0·0001
Symptoms 0·06 (0·02) –2·21 (0·34) 0·14 0·0064
Activities 0·10 (0·02) –2·07 (0·33) 0·25 <0·0001
Impact 0·13 (0·02) –1·65 (0·35) 0·29 <0·0001
Change in CRQ
Total –0·15 (0·01) –0·29 (0·30) –0·46 <0·0001
Dyspnoea –0·34 (0·04) –0·90 (0·30) –0·36 <0·0001
Fatigue –0·46 (0·05) –1·03 (0·30) –0·36 <0·0001
Emotion –0·29 (0·03) –1·26 (0·28) –0·37 <0·0001
Mastery –0·40 (0·05) –1·39 (0·28) –0·33 <0·0001
Change in ISW –0·01 (0·00) –1·69 (0·35) –0·14 0·0008
Change in 4MGS –6·69 (2·17) –2·03 (0·35) –0·17 0·0022
Change in 5STS 0·03 (0·03) –2·39 (0·33) 0·06 0·40
SGRQ=St George’s Respiratory Questionnaire. CRQ=Chronic Respiratory Questionnaire. ISW=incremental shuttle walk.
4MGS=4 m gait speed. 5STS=fi ve-repetition sit-to-stand.
Table 2: Change in COPD Assessment Test score with pulmonary rehabilitation compared with external
anchors (Study 1; n=565)
Articles
www.thelancet.com/respiratory Vol 2 March 2014
197
recruited 675 patients. For studies 2 and 3, 124 paired
measurements were required to detect a correlation
coeffi cient above 0·3 between change in CAT and change
in anchor with 80% power at the 0·01 signifi cance level.
We anticipated a minimum dropout of 30% on the basis
of data from other longitudinal cohorts, therefore we
recruited 200 patients in each study.
We did data analyses and graphical presentations using
SPSS (version 21) and Prism (version 5). When
estimating MCID in change in CAT score, only data from
participants who completed paired CAT measurements
were included for analysis. We used the paired t test to
compare paired measurements. We used Pearson’s
correlation (in which the null hypothesis was defi ned as
no correlation) and linear regression to compare change
in CAT score with other outcome measurements.
Estimation of MCID
For anchor-based estimation of MCID, predefi ned criteria
for establishing the validity of external anchors were: a
signifi cant correlation between change in CAT score and
change in anchor, and a correlation coeffi cient of more
than 0·3 as previously recommended.20 In Study 1,
change in CAT score with pulmonary rehabilitation was
anchored against change in SGRQ total score, CRQ total
score, and CRQ domain scores. For the GRCQ, we
calculated the mean (95% CI) change in CAT score with
pulmonary rehabilitation in those reporting feeling “a
little better”. We did not include those reporting feeling
“much better” because we believed that including these
patients might lead to an overestimation of the MCID. In
studies 2 and 3, change in CAT score was anchored
against change in SGRQ.
In studies 1 and 2, the focus was to establish the
minimum clinically important improvement because of
the small numbers of patients reporting signifi cant
worsening of health status. For MCID at the individual
patient level, we used receiver operating characteristic
curves. The change in CAT score cutoff that best
discriminated between patients who improved their health
status by the established MCID in the SGRQ total score
(–4 point change) or CRQ total score (+10 point change)
was defi ned as the MCID, with equal weighting given to
sensitivity and specifi city.10,12 For MCID at the population
level, we used linear regression analysis to estimate change
in CAT score corresponding to the minimum clinically
important improvement for the SGRQ and CRQ total
scores, and CRQ domain scores (+2·5 dyspnoea,
+2·0 fatigue, +3·5 emotion, +2·0 mastery).10 In Study 3,
because there were much the same numbers of patients
showing improvement and worsening of health status, we
applied receiver operating characteristic curves and linear
regression to investigate both minimum clinically
important improvement and deterioration.
For distribution-based methods, we calculated half the
SD (0·5 SD)21 and the SE of measurement (SEM),22 given
by the equation: SEM = SD × √ (1 – [test-retest reliability]).
On the basis of previous data, we assumed the test-retest
reliability of the CAT to be 0·8.2,4
Role of the funding source
The sponsors of the study had no role in study design,
data collection, data analysis, data interpretation, or
writing of the report. All authors had full access to all the
data in the study. WD-CM made the fi nal decision to
submit for publication.
Results
Study 1 took place between April 1, 2010, and Dec 31,
2012. Of 675 patients with COPD referred for pulmonary
rehabilitation, 565 (84%) completed pulmonary
Figure 1: Association between change in CAT score and change in (A) SGRQ score and (B) CRQ score with
pulmonary rehabilitation (Study 1)
CAT=COPD Assessment Test. SGRQ=St George’s Respiratory Questionnaire. CRQ=Chronic Respiratory Questionnaire.
40
Change in CAT score
20
0
–20
–40
40
Change in CAT score
20
0
–20
–40
–50
Change in CRQ total score
050–100 100
–40
Change in SGRQ total score
–20 20–60 400
A
B
r=0·32
r=–0·46
Articles
198
www.thelancet.com/respiratory Vol 2 March 2014
rehabilitation and had paired CAT results. The reasons
for non-completion were exacerbation or admission to
hospital (46 patients, 7%), poor adherence to pulmonary
rehabilitation (29, 4%), work reasons (11, 2%), family
illness (ten, 1%), holiday (seven, 1%), declined end of
course assessment (four, 1%), and death (three, <1%).
Baseline characteristics of the non-completers are
shown in the appendix; patients who did not complete
pulmonary rehabilitation had signifi cantly worse
baseline health status as determined by their SGRQ
Impact and CRQ total scores. Table 1 shows the baseline
characteristics of the completers, and the changes in
health status and physical performance with pulmonary
rehabilitation. Mean change in CAT with pulmonary
rehabilitation was –2·5 (95% CI –3·0 to –1·9).
Table 2 shows the slope, y intercept, and correlation
coeffi cient between change in CAT score and change in
other outcome measures. Only change in SGRQ total and
change in CRQ total and CRQ domain scores correlated
with change in CAT score with a correlation coeffi cient
greater than 0·3 (fi gure 1). Figure 2 shows the receiver
operating characteristic curves for change in CAT in
identifying patients who improved their SGRQ and CRQ
total scores by more than the established MCID. Both
curves were consistent in identifying –2 as the best
discriminating CAT change cutoff with an area under the
curve C statistic of 0·65 for SGRQ and 0·70 for CRQ. The
appendix shows the sensitivity and specifi city for
alternative estimates of the minimum clinically important
improvement in CAT score.
With linear regression, by use of a change in CRQ total
score of +10 as the cutoff for minimum clinically
important improvement, the estimated minimum
clinically important improvement in CAT score was
–1·8 (95% CI –2·6 to –1·0). By use of the established
MCID for the CRQ domains as cutoff s, the estimates for
minimum clinically important improvement in CAT
score were –1·7 (95% CI –2·5 to –1·0) for dyspnoea,
–2·0 (–2·7 to –1·2) for fatigue, –2·3 (–3·0 to –1·5) for
emotion, and –2·2 (–2·9 to –1·5) for mastery. With
SGRQ total change of –4 as the cutoff , the estimated
minimum clinically important improvement for CAT
score was –2·3 (95% CI –2·7 to –1·8). There were few
patients reporting signifi cant worsening in health status
Figure 2: Receiver operator characteristic curves using a change in COPD Assessment Test score of –2 to
best predict achievement of minimum clinically important improvement in (A) SGRQ score and (B) CRQ
score (Study 1)
ΔSGRQ=change in St George’s Respiratory Questionnaire score. ΔCRQ=change in Chronic Respiratory
Questionnaire score. AUC=area under the curve C statistic.
100
Sensitivity (%)
80
60
40
20
0
100
Sensitivity (%)
80
60
40
20
0
A
B
0
100-Specificity (%)
4020 60 80 100
ΔSGRQ≤–4
AUC=0·65
ΔCRQ≥10
AUC=0·70
At hospital
discharge
Change at 90 days p value
Mean age (years
[SD])
71 (11) ·· ··
Sex
Men 88 (60%) ·· ··
Women 59 (40%) ·· ··
Mean BMI (kg/m²
[SD])
27·1 (5·2) ·· ··
FEV1 (% predicted) 42 (39 to 46) 5·7 (1·7 to 9·8) 0·19
MRC 3·9 (3·7 to 4·1) –0·6 (–0·8 to –0·4) <0·0001
SGRQ
Total 57·1 (54·6 to 59·6) –6·8 (–8·8 to –4·9) <0·0001
Symptoms 67·3 (64·2 to 70·3) –0·3 (–2·9 to 2·3) 0·84
Activities 74·3 (70·9 to 77·7) –6·0 (–8·8 to –3·1) <0·0001
Impact 44·1 (41·4 to 46·8) –9·5 (–11·8 to –7·2) <0·0001
4MGS (m s–1) 0·66 (0·62 to 0·70) 0·22 (0·18 to 0·26) <0·0001
CAT 23·5 (22·3 to 24·8) –3·0 (–4·4 to –1·6) <0·0001
Data are mean (95% CI) or n (%) unless otherwise specifi ed. BMI=body-mass index.
FEV1=Forced expiratory volume in 1 s. MRC=Medical Research Council dyspnoea
score. SGRQ=St George’s Respiratory Questionnaire. 4MGS=4 m gait speed.
CAT=COPD Assessmen t Test.
Table 3: Baseline characteristics (at hospital discharge) of patients with
acute exacerbation of COPD and change over 3 months (Study 2; n=147)
See Online for appendix
Articles
www.thelancet.com/respiratory Vol 2 March 2014
199
(fi gure 1); comparison of the slopes of CAT score against
CRQ and SGRQ scores in patients who improved and
those who deteriorated showed no signifi cant diff erence
(ANCOVA p=0·67 and p=0·99, respectively).
With the GRCQ, 299 of 565 (53%) patients reported
feeling much better, 209 (37%) patients reported feeling a
little better, 40 (7%) patients reported no change, and
17 (3%) patients reported feeling a little or much worse.
The mean change in CAT score in those reporting feeling
“a little better” after pulmonary rehabilitation was
–1·6 (95% CI –2·6 to –0·8). For those feeling “much
better”, the mean change in CAT score was
–3·2 (95% CI –4·0 to –2·5).
With distribution-based methods, the estimate for
signifi cant improvement in CAT score using 0·5 SD was
–3·8 and using SEM was –3·3.
Study 2 took place between Nov 16, 2011, and Dec 31,
2012. Of 200 patients with acute exacerbations of COPD
recruited, 147 (74%) had paired CAT measurements.
Reasons for dropout were death (14 patients, 7%), declined
attendance (26, 13%), acutely unwell in hospital (ten, 5%),
and moved out of area (three, 2%). The mean duration
between baseline (discharge from hospital) and follow-up
measurement was 90·2 days (SD 7·7; range 79–102). The
appendix shows baseline characteristics of patients who
did not attend follow-up; baseline characteristics of those
who attended follow-up and those who did not attend did
not diff er. Table 3 shows clinical characteristics of this
group and the change in outcomes after hospital dis-
charge. Mean CAT score at hospital discharge (23·5
[95% CI 22·3–24·8]) was signifi cantly higher than baseline
CAT score in patients referred for pulmonary rehabilitation
in Study 1 (21·4 [20·8 to 22·0]) and stable outpatients in
Study 3 (20·1 [19·1 to 21·2]; ANOVA p=0·0002).
Mean change in CAT score from hospital discharge to
3 months after discharge was –3·0 (95% CI –4·4 to –1·6).
Change in CAT score correlated signifi cantly with
change in SGRQ (r=0·47; p<0·0001; fi gure 3) and
change in FEV1 (r=–0·26; p=0·0021), but not change in
4 m gait speed (r=–0·10; p=0·24). By use of receiver
operating characteristic curves, a –2 change in CAT
score best discriminated patients who improved by the
MCID or more in SGRQ (n=84) with an area under the
curve C statistic of 0·66. Sensitivity and specifi city data
for alternative estimates are in the appendix. Linear
regression analysis, using an SGRQ change of –4 as the
cutoff , estimated the minimum clinically important
improvement of the CAT as –2·8 (95% CI –3·7 to –1·9).
The slopes of CAT against SGRQ in patients who
improved and in those who deteriorated were not
signifi cantly diff erent (ANCOVA p=0·21).
With distribution-based methods, the estimate for
signifi cant improvement in CAT score using 0·5 SD was
–3·7 and using SEM was –3·3.
Study 3 recruited between Jan 1, 2012, and Aug 31,
2012. Of 200 stable patients recruited, 164 (82%) returned
for measurements 12 months later. Reasons for dropout
included: death (six patients, 3%), current admission to
hospital for exacerbation (seven, 4%), current admission
to hospital for another reason (four, 2%), declined (ten,
5%), unable to contact (six, 3%), and moved out of area
(three, 2%). The mean duration between baseline and
follow-up measurement was 364·6 days (SD 20·7, range
332–401 days). The appendix shows baseline characteristics
of patients who did not attend follow-up; patients who did
not attend follow-up were younger and had better CAT
scores than those who attended follow-up. Table 4 shows
clinical characteristics of this cohort and change in FEV1,
incremental shuttle walk, 4 m gait speed, CAT, and
SGRQ with time. There was no signifi cant change in
CAT over time, but change in CAT correlated signifi cantly
with change in SGRQ (r=0·36; p<0·0001; fi gure 4),
Figure 3: Association between CAT score and change in SGRQ score at hospital discharge for acute
exacerbation of COPD to 90 days after discharge (Study 2)
CAT=COPD Assessment Test. SGRQ=St George’s Respiratory Questionnaire.
–40
Change in SGRQ total score
–20 40–60 60020
r=0·47
30
Change in CAT score
10
0
–20
–30
20
–10
Baseline Change over 12 months p value
Mean age (years [SD]) 70 (8) ·· ··
FEV1 (% predicted) 47·6 (44·4 to 50·8) –1·7 (–6·9 to 3·4) 0·51
MRC 3·1 (3·0 to 3·3) 0·1 (–0·1 to 0·2) 0·35
ISW (m) 227 (208 to 247) –15 (–43 to 12) 0·28
4MGS (m s–1) 0·92 (0·89 to 0·95) –0·04 (–0·06 to –0·02) <0·0001
SGRQ
Total 50·6 (48·0 to 53·1) –0·3 (–4·4 to 3·9) 0·91
Symptoms 63·9 (60·9 to 67·0) –2·5 (–7·4 to 2·4) 0·32
Activities 68·7 (65·4 to 72·0) 1·3 (–3·8 to 6·3) 0·62
Impact 35·9 (33·1 to 38·7) 0·0 (–4·3 to 4·3) 0·99
CAT 20·1 (19·1 to 21·2) 0·6 (–0·4 to 1·5) 0·25
Data are mean (95% CI) unless otherwise specifi ed. FEV1=forced expiratory volume in 1 s. MRC=Medical Research Council
dyspnoea score. ISW=incremental shuttle walk. SGRQ=St George’s Respiratory Questionnaire. 4MGS=4 m gait speed.
CAT=COPD Assessment Test.
Table 4: Baseline characteristics in patients with stable COPD and change over 12 months (Study 3; n=164)
Articles
200
www.thelancet.com/respiratory Vol 2 March 2014
incremental shuttle walk (r=–0·25; p=0·0005), and 4 m
gait speed (r=–0·18; p=0·0142). For those who improved
their SGRQ (n=64), the receiver operating characteristic
curve showed that a –1 point change in CAT score was
the best discriminant with an area under the curve C
statistic of 0·66, and that a +1 point change in CAT score
best identifi ed those who had a clinically signifi cant
worsening in health status as measured by the SGRQ
(n=51) with an area under the curve C statistic of 0·69.
With SGRQ change of –4 as an anchor, linear regression
analysis estimated the minimum important improvement
in CAT score as –1·2 (95% CI –2·5 to –0·0). The slopes
of CAT score against SGRQ score in patients who
improved and those who deteriorated were not
signifi cantly diff erent (ANCOVA p=0·20).
With distribution-based methods, the estimate for
signifi cant improvement in the CAT score using 0·5 SD
was 3·8 and using SEM was 3·4. Table 5 summarises all
estimates for the MCID for the CAT.
Discussion
Our studies show that the CAT is responsive to change
after pulmonary rehabilitation and recovery from
hospital admission for acute exacerbation of COPD, and
that change in CAT correlates signifi cantly with change
in other health status measures and physical
performance measures. Furthermore, as far as we are
aware, these studies are the fi rst to prospectively and
purposively estimate the MCID for the CAT. With
anchors measuring similar constructs in diff erent
cohorts, the estimates for the minimum important
improvement in CAT score ranged from –1·2 to –2·8
with –2 being the most consistent estimate from
sensitivity and specifi city analyses.
Health status is recommended as an essential outcome
measure by pulmonary rehabilitation guidelines.23 As far
as we are aware, Study 1 was the largest study so far to
use the CAT during pulmonary rehabilitation (panel). We
showed longitudinal validity of the CAT by identifying
signifi cant correlations between change in CAT and
change in SGRQ, CRQ, and physical performance
measures. We recorded a mean change in CAT score
of –2·5 with an 8 week outpatient pulmonary
rehabilitation programme, in line with previous studies
from the UK and the USA and Canada.3,7 The CAT has
also been studied in an unselected chronic respiratory
disease population undergoing pulmonary rehabilitation,
with 110 non-COPD patients showing a mean CAT
change of –2·1, much the same as the mean change
measured in patients with COPD.19
Several studies have used the CAT as a measure of
health status during hospital and community-based
treatment of acute exacerbations of COPD, showing
changes in CAT ranging from –1·4 to –9·9.4,5 Study 2
focused exclusively on recovery of the CAT score after
admission to hospital, with the baseline CAT measured
on the day of hospital discharge. The recorded mean
Figure 4: Association between change in CAT score and change in SGRQ score over 12 months in stable
patients with COPD (Study 3)
CAT=COPD Assessment Test. SGRQ=St George’s Respiratory Questionnaire.
–40
Change in SGRQ total score
–20 40020
r=0·36
30
Change in CAT score
10
0
–20
20
–10
Anchor or method MCID
Study 1: Pulmonary rehabilitation
ROC SGRQ –2
ROC CRQ –2
Linear regression SGRQ –2·3
Linear regression CRQ –1·8
Linear regression CRQ-D –1·7
Linear regression CRQ-F –2·0
Linear regression CRQ-E –2·3
Linear regression CRQ-M –2·2
Linear regression GRCQ* –1·6
Distribution 0·5 SD –3·8
Distribution SEM –3·3
Study 2: At discharge from hospital
ROC SGRQ –2
Linear regression SGRQ –2·8
Distribution 0·5 SD –3·7
Distribution SEM –3·3
Study 3: Longitudinal
ROC SGRQ –1
Linear regression SGRQ –1·2
Distribution 0·5 SD –3·8
Distribution SEM –3·4
MCID=minimum clinically important diff erence. ROC=receiver operating
characteristic curves. SGRQ=St George’s Respiratory Questionnaire. CRQ=Chronic
Respiratory Questionnaire total score. CRQ-D=dyspnoea domain. CRQ-F=fatigue
domain. CRQ-E=emotion domain. CRQ-M=mastery domain. GRCQ=Global Rating
of Change Questionnaire. 0·5 SD=half SD. SEM=SE of measurement. *In patients
feeling “a little better”.
Table 5: Anchor-based and distribution-based estimates of the MCID of
the CAT in the three studies
Articles
www.thelancet.com/respiratory Vol 2 March 2014
201
change in CAT was –3·0, and this outcome correlated
signifi cantly with change in SGRQ (r=0·47). This fi nding
could have implications when designing clinical
intervention trials in the period after admission to
hospital, particularly with regard to sample size
calculation.
Until now, there have been few data regarding
longitudinal change in CAT with time. In the original
description of the CAT, test-retest reproducibility over
7 days was good (intraclass correlation coeffi cient 0·8),
while in stable patients over 4 weeks, the CAT showed a
test-retest intraclass correlation coeffi cient of 0·83.2,4
Over a 6 month period, Dodd and colleagues6 showed an
increase in CAT from 19·2 to 20·7, although importantly,
baseline measurements were recorded immediately after
completion of a pulmonary rehabilitation programme.
We recorded paired measurements 12 months apart. Our
prespecifi ed hypothesis that health status would
deteriorate signifi cantly over 12 months (ie, increase in
CAT score) was not supported by our data, although
deteriorating health status, including death, was a reason
for loss to follow-up. Although the 12 month recall rate
was good (higher than 80%), there was likely to be an
element of selection bias in that paired CAT scores were
not obtained in those who had died or were receiving
inpatient treatment at the recall timepoint. Nevertheless,
it was reassuring to show that change in the CAT
correlated signifi cantly with change in SGRQ and
physical performance measures over 12 months.
The determination of the MCID remains controversial
with no fi rm consensus, but is important in the validation
of clinical instruments and the assessment of clinical
studies. Two main approaches are generally used: anchor
and distribution-based methods. Anchor-based methods
rely on comparison of the change in outcome of interest
with another outcome measure of change, known as the
anchor or external criterion. However, this comparison is
only relevant if there is an established association
between the outcome of interest and the anchor. No
consensus exists regarding the threshold strength of the
association: some investigators have suggested an
arbitrary minimum correlation coeffi cient of greater
than 0·50, although others have suggested 0·30.11,20
Although cross-sectional studies have shown strong
correlations of the CAT and SGRQ, with four units on
the SGRQ corresponding to 1·6 on the CAT, this
association might diff er when assessing change in these
instruments.26 The MCID is most useful for clinicians or
researchers in the “change” setting, and therefore we
made great eff orts to assess change in CAT in three
independent cohorts, including longitudinal follow-up.
The candidate anchors with strongest correlations were
COPD-specifi c health status questionnaires (CRQ and
SGRQ), presumably because these instruments measure
similar constructs to the CAT, with correlation coeffi cients
ranging from 0·32 to 0·47. Previous guidance has
recommended the use of several approaches and
triangulation of methods.20 We used several anchors,
adopted diff erent methods of anchor-based estimations
(linear regression analysis, sensitivity and specifi city
analysis using receiver operating characteristic curves),
and presented change in CAT in three diff erent clinical
scenarios to provide clinical context.
The estimates of the MCID of the CAT were consistent
across diff erent cohorts in diff erent scenarios over
diff erent timeframes (table 5), providing some degree of
corroboration and credibility, although the correlations
with the external anchors were only moderate and
caution is needed in the interpretation. At the individual
level (the CAT score permits only integers), the ROC
analysis and the anchor responses to the GRCQ
estimated the MCID to be –2. From the linear regression
analysis, the population-level MCID estimates ranged
from –1·2 to –2·8, with the mean of these estimates
being –2·0. Therefore, both the population-level estimate
(used to assess group eff ects of treatments or diff erences
between populations) and the individual patient-level
estimate (used for responder analysis) was –2.
Anchor-based methods to estimate MCID are often
preferred to distribution-based approaches because they
take into account patient-reported benefi t or deterioration.
However, there are limitations. As mentioned previously,
anchor-derived estimates are only valid if the outcome of
interest correlates with the anchor. Any patient-reported
outcome recorded before and after a period of time is also
subject to recall bias. Anchors are designed to detect
change in outcome but rarely take into account costs to the
Panel: Research in context
Systematic review
We searched PubMed for studies in English focusing on estimates of the minimum
clinically important diff erence (MCID) of the COPD Assessment Test (CAT) from
inception, to Oct 21, 2013. We used the terms “MID” or “minimal important diff erence”
or “minimum important diff erence” or “MCID” or “minimal clinically important
diff erence” and “CAT” or “COPD Assessment Test” or “Chronic Obstructive Pulmonary
Disease Assessment Test”. We identifi ed three studies7,24,25 that provided estimates of the
MCID of the CAT, although through retrospective opportunistic post-hoc analyses.
These studies looked at improvement in stable populations using only one
methodological approach for estimation; the resulting estimates ranged from a
–1·3 to –3·76 point change.
Interpretation
As far as we are aware, this is the fi rst report to provide prospective data to estimate the
MCID of the CAT. Our studies provide 19 separate estimates of the MCID using both
distribution-based and anchor-based methods, from three separate clinically relevant
cohorts, including stable patients having pulmonary rehabilitation and longitudinal
follow-up, and also in those recovering from exacerbation. To our knowledge, Study 1 is
the largest study using the CAT in pulmonary rehabilitation. Our fi ndings suggest that a
decrease in CAT of 2 points is the most reliable estimate of the MCID at the individual
and population level. This estimate will allow clinicians to interpret clinically important
change in individual patients, and help researchers with sample size calculations and the
interpretation of CAT data in response to intervention studies.
Articles
202
www.thelancet.com/respiratory Vol 2 March 2014
patient, for example side-eff ects of therapy. Another
limitation is that changes in an outcome measure might
be associated with baseline level. Potential reasons for this
association include so-called fl oor and ceiling eff ects. The
CAT score ranges from 0 to 40 with a decrease in score
showing improvement in health status. At extreme values
(eg, a baseline CAT of 0), the instrument cannot detect an
improvement in health status. Similarly, if the baseline
CAT is 40, there is no room for further worsening in health
status. Although we used three independent cohorts of
patients with COPD, with varying baseline mean values,
these were generally recruited from a secondary care
setting. The present studies cannot address whether our
estimates of MCID for the CAT hold true in populations at
extremes of the health spectrum (for example, community-
dwelling asymptomatic patients with COPD, or highly
symptomatic patients receiving palliative care).
To provide a comprehensive approach, we also used
distribution-based methods to estimate the MCID of the
CAT. Distribution-based approaches use statistical
methods, and are based on the distribution of the cohort
and the reliability of the measure; as such they do not take
into account whether the recorded change is important
from the patient’s perspective. Previous investigators
have noted that 0·5 SD and the SEM might approximate
the MCID in some patient-reported outcome measures,21,22
although other measures have also been described
including 1·96 SEM and minimum detectable change;
depending on which distribution-based approach is used,
a range of diff erent estimates of the MCID can be
generated which might signifi cantly limit their
interpretation.27 Another specifi c limitation to the present
studies was that the test-retest reliability of the CAT was
assumed on the basis of previous studies of much shorter
duration. Our estimates of the MCID for the CAT, using
0·5 SD and SEM, were consistent across the three clinical
scenarios, suggesting that the distributions of our three
cohorts were much the same. However, it was also
noticeable that the distribution-based methods
consistently estimated the MCID to be greater than the
anchor-based estimates, suggesting that the CAT had a
wide distribution in our cohorts. This situation is not
unique in patients with COPD. For example, Puhan and
colleagues28 showed that the MCID for maximal cycle
exercise capacity ranged from 2·2 W to 3·3 W, whereas
distribution-based estimates were 5·3–5·5 W.28
Although we provide data from both approaches we
have chosen to place greater emphasis on the anchor-
based estimates for two reasons. First, although
distribution-based estimates provide supportive infor-
mation regarding a signifi cant change, they do not
provide a direct measure of minimum clinical importance,
a view shared by other researchers.28,29 Second, the MCIDs
estimated by distribution-based methods were greater
than the mean change identifi ed after pulmonary
rehabilitation, widely accepted as a highly eff ective
intervention that signifi cantly improves health-related
quality of life in COPD, and the mean change identifi ed
during recovery after admission to hospital for severe
exacerbation of COPD. In view of these clinical contexts,
we believe the data derived from distribution-based
methods provide information about clinical signifi cance
but might overestimate the true MCID.
In the present studies, our focus was on establishing
the minimum clinically important improvement. In
studies 1 and 2, the number of patients with improving
health status far outweighed those who had worsening
health status (90% vs 3% according to the GRCQ), and
we did not report estimates of minimum important
clinical deterioration for fear of imprecision. We
believe that the magnitude of the minimum clinically
important deterioration is likely to be much the same
as for improvement. For example, when we compared
the linear regression slopes of those who improved
and deteriorated according to their anchor, we
identifi ed no signifi cant diff erence. In Study 3, in
which there were more equivalent numbers of patients
improving and worsening their health status, the
receiver operating characteristic curves estimated
similar magnitude cutoff s. Further studies focusing
on worsening health status are needed to confi rm
whether patients perceive size of deterioration
diff erently to size of improvement.
The main aim of the present studies was to estimate the
MCID in change in CAT, and hence paired measurements
were used for the analyses. In all three studies, there were
missing data (usually because of exacerbation) and it
could be argued that these missing data might bias our
estimates of the MCID. If a patient deteriorated because
of exacerbation or admission to hospital, we would expect
not only the CAT to worsen but also the external anchors.
Because there was no diff erence in the slopes of change
in patients who improved compared with those who
deteriorated, we do not believe that missing data were a
signifi cant source of bias in our estimates of the MCID
for the CAT.
In summary, the present studies show that the CAT is
responsive to the eff ects of pulmonary rehabilitation and
recovery from admission to hospital for acute exacerbation
of COPD. By use of various health status and global rating
of change questionnaires as external anchors, we estimate
the minimum important improvement of the CAT to be a
two point decrease at both the individual and population
level. This information could be useful in the clinical
interpretation of CAT data, particularly in response to
intervention studies.
Contributors
All authors contributed to the analysis and interpretation of data, and
the preparation of the report. WD-CM conceived the idea and is the
guarantor of the paper, taking responsibility for the integrity of the work
as a whole, from inception to published Article.
Confl icts of interest
MIP has received personal reimbursement for lecturing or
consultancyregarding muscle function in COPD from Novartis and
Philips Respironics; he discloses institutional reimbursement for
Articles
www.thelancet.com/respiratory Vol 2 March 2014
203
consultancy from GlaxoSmithKline, Novartis, Regeneron, Lilly,
Biomarin, and Boehringer Ingelheim and institutional agreements to
do research with GlaxoSmithKline, Novartis, AstraZeneca, and Philips
Respironics. All other authors declare that they have no confl icts of
interest.
Acknowledgments
SSCK is supported by the Medical Research Council. WD-CM is
supported by a National Institute for Health Research Clinician Scientist
Award, a Medical Research Council (UK) New Investigator Research
Grant, and a National Institute for Health Research Clinical Trials
Fellowship. This project was done at the NIHR Respiratory Biomedical
Research Unit at the Royal Brompton and Harefi eld NHS Foundation
Trust and Imperial College London; the salaries of JLC, SEJ, and MIP are
wholly or partly funded by the NIHR Biomedical Research Unit. The
views expressed in this publication are those of the authors and not
necessarily those of the NHS, the National Institute for Health Research,
nor the Department of Health. We thank the Harefi eld Pulmonary
Rehabilitation Team and the Hillingdon COPD Outreach Team for their
help in recruiting participants.
References
1 Ringbaek T, Martinez G, Lange P. A comparison of the assessment
of quality of life with CAT, CCQ, and SGRQ in COPD patients
participating in pulmonary rehabilitation. COPD 2012; 9: 12–15.
2 Jones PW, Harding G, Berry P, Wiklund I, Chen WH, Kline Leidy N.
Development and fi rst validation of the COPD Assessment Test.
Eur Respir J 2009; 34: 648–54.
3 Jones PW, Harding G, Wiklund I, et al. Tests of the responsiveness
of the COPD assessment test following acute exacerbation and
pulmonary rehabilitation. Chest 2012; 142: 134–40.
4 Agusti A, Soler JJ, Molina J, et al. Is the CAT questionnaire
sensitiveto changes in health status in patients with severe
COPDexacerbations? COPD 2012; 9: 492–98.
5 Miravitlles M, Garcia-Sidro P, Fernandez-Nistal A, Buendia MJ,
Espinosa de Los Monteros MJ, Molina J. Course of COPD
assessment test (CAT) and clinical COPD questionnaire (CCQ)
scores during recovery from exacerbations of chronic obstructive
pulmonary disease. Health Qual Life Outcomes 2013; 11: 147.
6 Dodd JW, Marns PL, Clark AL, et al. The COPD Assessment
Test(CAT): short- and medium-term response to pulmonary
rehabilitation. COPD 2012; 9: 390–94.
7 Dodd JW, Hogg L, Nolan J, et al. The COPD assessment test (CAT):
response to pulmonary rehabilitation. A multicentre, prospective
study. Thorax 2011; 66: 425–29.
8 Mackay AJ, Donaldson GC, Patel AR, Jones PW, Hurst JR,
Wedzicha JA. Usefulness of the Chronic Obstructive Pulmonary
Disease Assessment Test to evaluate severity of COPD
exacerbations. Am J Respir Crit Care Med 2012; 185: 1218–24.
9 Vestbo J, Hurd SS, Agusti AG, et al. Global strategy for the
diagnosis, management, and prevention of chronic obstructive
pulmonary disease: GOLD executive summary.
Am J Respir Crit Care Med 2013; 187: 347–65.
10 Schunemann HJ, Puhan M, Goldstein R, Jaeschke R, Guyatt GH.
Measurement properties and interpretability of the Chronic
respiratory disease questionnaire (CRQ). COPD 2005; 2: 81–89.
11 Schunemann HJ, Griffi th L, Jaeschke R, Goldstein R, Stubbing D,
Guyatt GH. Evaluation of the minimal important diff erence for
thefeeling thermometer and the St. George’s Respiratory
Questionnaire in patients with chronic airfl ow obstruction.
J Clin Epidemiol 2003; 56: 1170–76.
12 Jones PW. St George’s respiratory questionnaire: MCID.
COPD2005; 2: 75–79.
13 Pauwels RA, Buist AS, Calverley PM, Jenkins CR, Hurd SS.
Globalstrategy for the diagnosis, management, and prevention
ofchronic obstructive pulmonary disease. NHLBI/WHO Global
Initiative for Chronic Obstructive Lung Disease (GOLD) Workshop
summary. Am J Respir Crit Care Med 2001; 163: 1256–76.
14 Williams JE, Singh SJ, Sewell L, Guyatt GH, Morgan MD.
Development of a self-reported Chronic Respiratory Questionnaire
(CRQ-SR). Thorax 2001; 56: 954–59.
15 Kon SS, Patel MS, Canavan JL, et al. Reliability and validity of
4-metre gait speed in COPD. Eur Respir J 2013; 42: 333–40.
16 Jones SE, Kon SS, Canavan JL, et al. The fi ve-repetition sit-to-stand
test as a functional outcome measure in COPD. Thorax 2013;
68: 1015–20.
17 Singh SJ, Morgan MD, Scott S, Walters D, Hardman AE.
Development of a shuttle walking test of disability in patients
withchronic airways obstruction. Thorax 1992; 47: 1019–24.
18 Juniper EF, Guyatt GH, Willan A, Griffi th LE. Determining a
minimal important change in a disease-specifi c Quality of Life
Questionnaire. J Clin Epidemiol 1994; 47: 81–87.
19 Kon SS, Clark AL, Dilaver D, et al. Response of the COPD
Assessment Test to pulmonary rehabilitation in unselected
chronicrespiratory disease. Respirology 2013; 18: 974–77.
20 Revicki D, Hays RD, Cella D, Sloan J. Recommended methods for
determining responsiveness and minimally important diff erences
for patient-reported outcomes. J Clin Epidemiol 2008; 61: 102–09.
21 Norman GR, Sloan JA, Wyrwich KW. Interpretation of changes in
health-related quality of life: the remarkable universality of half a
standard deviation. Med Care 2003; 41: 582–92.
22 Wyrwich KW, Tierney WM, Wolinsky FD. Further evidence
supporting an SEM-based criterion for identifying meaningful
intra-individual changes in health-related quality of life.
J Clin Epidemiol 1999; 52: 861–73.
23 Bolton CE, Bevan-Smith EF, Blakey JD, et al. British Thoracic
Society guideline on pulmonary rehabilitation in adults. Thorax
2013; 68 (suppl 2): ii1–30.
24 Tsiligianni IG, van der Molen T, Moraitaki D, et al. Assessing health
status in COPD. A head-to-head comparison between the COPD
assessment test (CAT) and the clinical COPD questionnaire (CCQ).
BMC Pulm Med 2012; 12: 20.
25 Jones PW, Price D, van der Molen T. Role of clinical questionnaires
in optimizing everyday care of chronic obstructive pulmonary
disease. Int J Chron Obstruct Pulmon Dis 2011; 6: 289–96.
26 Jones PW, Brusselle G, Dal Negro RW, et al. Properties of the
COPD assessment test in a cross-sectional European study.
Eur Respir J 2011; 38: 29–35.
27 Copay AG, Subach BR, Glassman SD, Polly DW Jr, Schuler TC.
Understanding the minimum clinically important diff erence:
a review of concepts and methods. Spine J 2007; 7: 541–46.
28 Puhan MA, Chandra D, Mosenifar Z, et al. The minimal important
diff erence of exercise tests in severe COPD. Eur Respir J 2011;
37: 784–90.
29 Rennard SI. Minimal clinically important diff erence, clinical
perspective: an opinion. COPD 2005; 2: 51–55.