Severe exacerbations and BODE index: Two

independent risk factors for death in male COPD

patients
Juan Jose Soler-Cataluna
a,
*, Miguel A

ngel Mart nez-Garc a

a
,
Lourdes Sa nchez Sa nchez
b
, Miguel Perpina Tordera
c
, Pilar Roma n Sa nchez
d
a
Hospital General de Requena, Unidad de Neumologa, Servicio de Medicina Interna, Paraje Casablanca s/n., 46340
Requena, Valencia, Spain
b
Centro de Salud de Lliria, Valencia, Spain
c
Hospital La Fe, Servicio de Neumologa, Valencia, Spain
d
Hospital General de Requena, Servicio de Medicina Interna, Requena, Valencia, Spain
Received 1 August 2008; accepted 4 December 2008
Available online 7 January 2009
KEYWORDS
Chronic obstructive
pulmonary disease
(COPD);
Exacerbations;
Hospitalizations;
Mortality;
Prognostic value
Summary
Objectives: 1) To determine whether severe exacerbation of COPD is a BODE index indepen-
dent risk factor for death; 2) whether the combined application of exacerbations and BODE
(e-BODE index), offers greater predictive capacity than BODE alone or can simplify the model,
by replacing the exercise capacity (BODEx index).
Methods: A prospective study was made of a cohort of COPD patients. In addition to calcula-
tion of the BODE index we register frequency of exacerbations. An analysis was made of all-
cause mortality, evaluating the predictive capacity of the exacerbations after adjusting for
the BODE. These variables were also used to construct two new indexes: e-BODE and BODEx.
Results: The study included 185 patients with a mean age of 71 9 years, and FEV
1
% 47 17%.
Severe exacerbation appeared as an independent adverse prognostic variable of BODE index.
For each new exacerbation the adjusted mortality risk increased 1.14-fold (95% CI: 1.04e1.25).
However, the e-BODE index (C statistic: 0.77, 95% CI: 0.67e0.86) didnt improve prognostic
capacity of BODE index (C statistic: 0.75, 95% CI: 0.66e0.84) (p ZNS). An interesting nding
was that BODEx index (C statistic: 0.74, 95% CI: 0.65e0.83) had similar prognostic capacity
than BODE index.
* Corresponding author. Tel.: 34 96 233 96 88; fax: 34 96 233 97 88.
E-mail address: [email protected] (J.J. Soler-Cataluna).
0954-6111/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.rmed.2008.12.005
avai l abl e at www. sci encedi r ect . com
j our nal homepage: www. el sevi er . com/ l ocat e/ r med
Respiratory Medicine (2009) 103, 692e699
Conclusions: Severe exacerbations of COPD imply an increased mortality risk that is indepen-
dent of baseline severity of the disease as measured by the BODE index. The combined appli-
cation of both parameters (e-BODE index) didnt improve the predictive capacity, but on
replacing exacerbation with exercise capacity the multidimensional grading system is simpli-
ed without loss of predictive capacity.
2008 Elsevier Ltd. All rights reserved.
At present, chronic obstructive lung disease (COPD) is the
fourth most common cause of death in the world and the
future perspectives are equally discouraging.
1,2
Therefore,
we urgently need to adopt preventive and therapeutic strat-
egies designed torevert this trend. Akey element inthis sense
is the study of the causes of death and particularly the asso-
ciated risk factors. Traditionally it has been accepted that the
forced expiratory volume in one second (FEV
1
) and its accel-
erated decrease over time is one of the best predictors of
mortality.
3,4
For decades, this fact conditioned the thera-
peutic objective in patients with COPD. However, with the
exception of smoking cessation, very few interventions have
been able to slow the loss of lung function and improve the
prognosis.
5,6
Incontrast, wenowknowthat COPDis a complex
chronic inammatory disease with multiple dimensions e
some of which also have important prognostic implications.
Celli et al.
7
developed a multidimensional index that inte-
grates these principal prognostic determinants: the BODE
[body mass index (BMI), airow obstruction, dyspnea and
exercise capacity] index. This score showed to be more
effective than FEV
1
as a prognostic variable,
7
and unlike
the latter, it contains a number of dimensions that can be
modiedesuchas dyspnea, exercisetoleranceandevenBMI.
8
Acute exacerbations (AECOPD) are among the prognostic
factors that have generated most interest in recent months.
9
However, the BODE index does not include such events. In an
observational study of 304 patients, our group found severe
exacerbations to be an independent prognostic factor of the
baseline severity of the disease. The frequency of exacer-
bations was adjusted for different confounding variables,
including FEV
1
, BMI, age, gas exchange or comorbidity.
However, the 6-minute walking distance test (6MWD) was not
used in this cohort, and no measurements were made of
dyspnea; as a result, the model could not be adjusted for the
BODE index, and this raised questions regarding interaction
between the two prognostic factors.
The main objective of the present study was to deter-
mine whether severe AECOPD is a BODE index independent
risk factor for death. In turn, the secondary objectives
were: 1) to determine whether the combined application of
both parameters, exacerbations and BODE (e-BODE index),
offers greater predictive capacity than BODE alone; and 2)
to explore whether exacerbations can simplify the model,
by replacing the 6MWD with the registry of severe exacer-
bations (BODEx index).
Method
Patients
A prospective study was made of a cohort of 185 COPD
outpatients. The subjects were included between January
1999 and June 30, 2004. The diagnosis and classication of
COPD was carried out according to the latest Global
Initiative for Chronic Obstructive Lung Disease (GOLD)
criteria, and was based on the current or past smoking
history (>10 pack-year) and postbronchodilator FEV
1
/
forced vital capacity (FVC) ratio <70%.
1
Patients previ-
ously diagnosed with bronchial asthma, bronchiectasis,
cystic brosis, upper airways obstruction or bronchiolitis
related to systemic pathology were excluded. All included
patients were required to be in a stable phase of the
disease, i.e., without exacerbation in the month preceding
the study. The study was approved by local ethic
committee.
Protocol
Age, sex, smoking history, comorbidity, BMI, 6MWD, forced
spirometry and arterial blood gases data were collected in
all patients. The BODE index was calculated in all patients
using the score proposed by Celli et al.
7
BMI was calculated
by dividing patient body weight (kg) by the square of height
(m
2
). Dyspnea was assessed using the modied Medical
Research Council scale.
10
FEV
1
was determined by forced
spirometry (Vmax Spectra, SensorMedics Corporation, USA),
following the guidelines established by the Spanish Society
of Pneumology and Chest Surgery (SEPAR).
11
The FEV
1
and
FVC results are expressed as percentages of the adult
reference values.
12
Postbronchodilator FEV
1
was used as
airow limitation index, because it is regarded as a better
predictor of mortality than prebronchodilator FEV
1
.
3
The
6MWD was performed according to ATS guidelines.
13
We
used the best of two 6MWD separated by at least 30 min.
Comorbidity was quantied according to the index of
Charlson et al.
14
Exacerbations
All exacerbation episodes requiring hospital management
(emergency visits or admissions) (AECOPD) during the
previous year to inclusion were collected. We have an
informatic system where all hospital contacts are regis-
tered. A prospective collection was also made of all exac-
erbation episodes requiring hospital management during
the rst year of follow-up. The patients were divided into
three groups according to the recorded frequency of severe
AECOPD: Group A (patients with no severe AECOPD); Group
B (patients with one or two severe AECOPD); Group C
(patients with 3 or more severe AECOPD). Severe AECOPD
was dened as any sustained increase in respiratory symp-
tomatology vs the patient baseline situation, requiring
modication of habitual medication and hospital care
(emergency visit or admission).
15
Exacerbation and BODE index in COPD 693
e-BODE index (BODE plus exacerbation)
To construct the e-BODE index we used the 10 categories
previously described by Celli et al.
7
adding three further
possibilities according to the frequency of severe exacer-
bations. The absence of severe AECOPD was scored as 0, the
presence of 1e2 severe AECOPD episodes received 1 point,
and cases with three or more severe AECOPD were scored as
2. Thus, the e-BODE presents a score range of between
0 and 12 points. This score was divided into quartiles as
follows: quartile 1 Z0e2 points, quartile 2 Z3e4 points;
quartile 3 Z5e6 points; and quartile 4 Z7e12 points.
BODEx index
The methodology for developing this simplied index involved
replacing the four categories of the walking test (0e3 points)
with the three categories generated by the exacer-
bations (0, 1 or 2 points). The score range for this index is
therefore between 0 and 9 points. The following quartiles
were considered: quartile 1 Z0e2 points; quartile 2 Z3e4
points; quartile 3 Z5e6 points; and quartile 4 Z7e9 points.
Statistical analysis
Descriptive statistics were used to describe the study
population at baseline. The comparison of means among
the three study groups was based on analysis of variance
(ANOVA), with chi-square testing and Bonferroni correction
for the comparison of proportions. All-cause mortality was
evaluated. We rst conducted univariate analyses based on
the Cox proportional hazards model using each of the
potential predictors of respiratory mortality as indepen-
dent variables, and survival status as the dependent vari-
able.
16
Survival curves for AECOPD groups were estimated
by the KaplaneMeier product limit method and compared
with the log-rank test.
17
Multivariate analysis was also
based on the Cox proportional model. An interaction term
between the variables and time was introduced in the
model to analyze risk proportionality. To compare the
capacity of predicting all-cause mortality of the exacer-
bations, BODE index, e-BODE index and BODEx index, we
obtained ROC Type II curves and estimated the C statistics
for each one in those patients with a follow-up of over 5
years. The null value for the C statistic is 0.5, with
a maximum of 1.0 (higher values being better).
18
A method
developed by Halney and McNeil was used to compare ROC
curves derived from the same cases.
19
All statistical anal-
yses were carried out using a statistical software package
(SPSS for Windows, version 11.5; SPSS Inc., Chicago, IL,
USA). A p-value of <0.05 was considered to be signicant.
Results
Subject characteristics
A total of 185 men were included in the study. The baseline
characteristics of the patients are shown in Table 1. An
increasing frequency of exacerbations was associated with
older age, a higher dyspnea score, a lower predicted FEV
1
%,
shorter distance in the 6MWD and higher BODE score.
Comorbidity and BMI were similar in all groups. The mean
follow-up was 36 24 months. In 110 cases (59.5%), follow-
up lasted over 5 years.
Table 1 Baseline characteristics of the patients (n Z185).
Global sample (n Z185) Groups of AECOPD
A (n Z102, 55.1%) B (n Z48, 25.9%) C (n Z35, 18.9%) p-value
Age (years) 71 9 69 9 72 8 74 8 0.027
BMI (kg/m
2
) 28.1 5.2 28.4 5.6 27.8 4.6 27.8 4.8 0.738
Current smoking (%) 33 (17.8) 16 (15.7) 10 (20.8) 7 (20.0) 0.337
Pack-year 62 36 58 31 71 41 60 40 0.121
Charlson index 0.80 0.96 0.67 0.87 1.00 1.23 0.91 0.74 0.105
Dyspnea (MMRC) 2.12 1.01 1.92 0.93 2.10 1.09 2.71 0.92 <0.001
PaO
2
64 12 66 10 62 15 61 11 0.022
PaCO
2
44 7 43 6 44 8 44 6 0.322
FEV
1
(I) 1.17 0.44 1.25 0.52 1.15 0.32 0.99 0.26 0.072
FEV
1
% (predicted) 47.9 15.5 50.0 18.2 45.8 11.5 44.2 9.6 0.003
GOLD classic, n (%) 0.002
Stage I 10 (5.4) 9 (8.8) 1 (2.1) e
Stage II 53 (28.6) 37 (36.3) 12 (25.0) 4 (11.4)
Stage III 60 (32.4) 34 (33.3) 13 (27.1) 13 (37.1)
Stage IV 62 (33.5) 22 (21.6) 22 (45.8) 18 (51.4)
6MWD (m) 379 111 398 117 361 106 348 91 0.032
BODE index 3.5 2.1 3.0 2.0 3.7 2.4 4.5 1.6 <0.001
AECOPD (n) 1.43 2.41 e 1.42 0.50 5.63 2.65 <0.001
Admissions (n) 0.56 1.03 e 0.60 0.64 2.05 1.41 <0.001
AECOPD: acute exacerbations of COPD requiring hospital care. BMI: body mass index. MMRC scale: score on the modied Medical
Research Council (MMRC) dyspnea scale; Stage IeIV: GOLD criteria stratication. 6MWD: 6-minute walking distance test. Values are
expressed as the mean standard deviation.
694 J.J. Soler-Cataluna et al.
Univariate survival analysis
The overall median survival was 63 months (95% CI: 48e77
months). A total of 71 (38.4%) deaths were recorded. The
survival probability after 3, 5 and 7 years was 67%, 52% and
42%, respectively. Forty-three deaths (60.6%) were due to
respiratory causes, 15 patients (21.1%) died of cardiovas-
cular disease, 4 (5.6%) of malignant disease, and 7 (9.9%) of
other diseases. In two cases (2.8%) the cause of death was
not known. Nineteen patients were lost in the course of the
5-year follow-up period (follow-up rate: 89.7%).
Table 2 reects the prognostic inuence of the variables
included in the univariate analysis. All of the BODE
components, with the exception of BMI, were associated to
an increased mortality risk. For each unit increase in the
BODE index, the mortality risk increased 33% (HR: 1.33, 95%
CI: 1.20e1.48), and for each new severe AECOPD the risk
increased 17% (HR: 1.17, 95% CI: 1.09e1.26). There was
a moderate but signicant correlation between AECOPD
during the rst year of follow-up and the previous year to
inclusion into the study (r Z0.37, p <0.001). The C
statistic for the BODE index in quartiles was 0.75 (95% CI:
0.66e0.84) vs 0.70 (95% CI: 0.60e0.80) for the AECOPD
during the rst year of follow-up and 0.67 (95% CI: 0.57e
0.78) for the severe exacerbations that were registered
during the previous year.
Multivariate survival analysis
Table 3 shows the regression model for exacerbations
adjusted for age, comorbidity, blood gases and BODE index.
Severe AECOPD suffered during the previous year to inclu-
sion appeared as an independent adverse prognostic vari-
able e with an adjusted mortality risk 2.24-fold greater
who presented one or two severe exacerbations and 2.80-
fold greater for patients three or more severe AECOPD,
than for patients without AECOPD.
The severe exacerbations during the rst year of follow-
up appeared as an independent adverse prognostic factor
(HR: 1.04, 95% CI: 0.54e2.01 for group B and HR: 2.66, 95%
CI: 1.37e5.16 for group C, p Z0.007) only if we excluded of
the previous exacerbations of the model. For each new
exacerbation the adjusted mortality risk increased 1.14-
fold (95% CI: 1.04e1.25). No signicant interaction was
observed between the BODE index and the prognostic
effect of AECOPD. The risk proportionality test proved
nonsignicant. Hazard ratios were unchanged over time.
e-BODE index
Fig. 1 shows the survival curves according to e-BODE
quartiles adjusted by the other variables. Forty-seven
(25.3%) of the patients were allotted to quartile 1 (0e2
points), 63 (33.9%) to quartile 2 (3e4 points), 40 (21.5%) to
quartile 3 (5e6 points), and 36 (19.3%) to quartile 4 (7e12
points). Table 4 shows the death hazard ratios for e-BODE,
after adjusting the model for age, comorbidity and blood
gases. For each unit increase in this combined index, the
adjusted mortality risk increased 35% (HR: 1.35, 95% CI:
1.21e1.51, p <0.0001). The C statistic for the e-BODE
index was 0.77 (95% CI: 0.67e0.86).
BODEx index
The index combining BMI, FEV
1
%, dyspnea and exacerba-
tions also showed an independent predictive value for
mortality, after adjusting the model for age, Charlson index
and blood gases (Table 5, Fig. 2). For each point increase in
the BODEx index, the adjusted mortality risk increased 44%
(HR: 1.44, 95% CI: 1.25e1.66, p <0.001). The C statistic for
the BODEx index was 0.74 (95% CI: 0.65e0.83).
Table 2 Predictors of mortality: univariate analysis.
Hazard ratio
(crude)
95% CI p-value
Age (years) 1.07 1.04e1.11 <0.001
Cumulative smoking
(pack-year)
0.99 0.99e1.01 0.430
Comorbidity index 1.35 1.04e1.76 0.024
BMI (kg/m
2
)
21 1.24 0.54e2.88 0.606
Dyspnea (MMRC scale) <0.001
0e1 e e
2 1.52 0.63e3.62
3 3.33 1.47e7.57
4 5.10 1.97e13.17
FEV
1
(% predicted) 0.031
65 e e
50e64 1.47 0.58e3.74
36e49 2.23 0.96e5.19
35 2.99 1.30e6.87
6MWD (m) <0.001
350 e e
250e349 2.75 1.58e4.78
150e249 2.44 1.26e4.72
<149 8.18 3.13e21.38
BODE quartiles <0.001
Quartile 1 e e
Quartile 2 1.62 0.80e3.25
Quartile 3 3.28 1.66e6.49
Quartile 4 5.93 2.82e12.47
AECOPD groups
(previous year
to inclusion)
<0.001
Group A e e
Group B 1.96 1.05e3.63
Group C 2.96 1.70e5.13
AECOPD groups
(rst year of
follow-up)
<0.001
Group A e e
Group B 1.77 0.99e3.17
Group C 3.76 2.14e6.50
BMI: body mass index. 6MWD: 6-minute walking distance test.
AECOPD: acute exacerbations of COPD requiring hospital care.
Group A: No AECOPD; Group B: patients with one or two AECOPD
(emergency visits or hospitalizations); Group C: patients with 3
or more AECOPD.
Exacerbation and BODE index in COPD 695
Comparison among all indexes
Fig. 3 shows the ROC Type II curves corresponding to
exacerbations and the BODE, e-BODE and BODEx indexes.
The best result corresponded to the combined e-BODE
index, followed by the original BODE index, BODEx index
and nally the frequency of exacerbations. However, we
didnt observe statistical differences in C statistic when
comparing BODE index vs exacerbation alone during the
previous year or rst year of follow-up, e-BODE index or
BODEx index.
Discussion
The present study conrms the idea that severe exacer-
bations of COPD (emergency visits or admission) are an
important and independent indicator of poor patient
prognosis. Indeed, even after adjusting for a multidimen-
sional severity scale such as the BODE index, the mortality
risk is maintained if the patient suffers severe exacerba-
tions. In the case of patients with a recent history of
frequent exacerbations the mortality risk were near triple
(HR: 2.80, 95% CI: 1.43e5.48), after adjusting the model for
Table 3 Predictors of mortality: multivariate analysis for
exacerbation frequency. Adjusted model for age, comor-
bidity, blood gases and BODE index.
Hazard ratio
(adjusted)
95% CI p-value
Age (years) 1.07 1.03e1.12 <0.001
Charlson index 1.04 0.77e1.41 NS
PaO
2
(mmHg) 1.02 0.99e1.04 NS
PaCO
2
(mmHg) 1.07 1.02e1.11 0.003
BODE quartiles 0.001
Quartile 1 e e
Quartile 2 1.15 0.48e2.76
Quartile 3 2.32 0.98e5.50
Quartile 4 4.30 1.72e10.75
AECOPD groups
(previous year
to inclusion)
0.010
Group A e e
Group B 2.24 1.05e4.79
Group C 2.80 1.43e5.48
AECOPD groups
(rst year of
follow-up)
NS
Group A e e
Group B 0.86 0.44e1.70
Group C 1.78 0.87e3.65
CI: condence interval. p-value: level of signicance. AECOPD:
acute exacerbations of COPD requiring hospital care. Group A:
No AECOPD; Group B: patients with one or two AECOPD
(emergency visits or hospitalizations); Group C: patients with 3
or more AECOPD.
Figure 1 Survival curves by e-BODE index (BODE index plus
exacerbation frequency) in patients with COPD, adjusted by
confounding variables. The scores range from 0 to 12 points,
the highest quartile the highest mortality.
Table 4 Predictors of mortality: multivariate analysis for
e-BODE index. Adjusted model for age, comorbidity and
blood gases.
Hazard ratio
(adjusted)
95% CI p-value
Age (years) 1.07 1.03e1.11 <0.001
Charlson index 1.10 0.83e1.44 NS
PaO
2
(mmHg) 1.00 0.98e1.02 NS
PaCO
2
(mmHg) 1.05 1.01e1.09 0.016
e-BODE quartiles <0.001
Quartile 1 e e
Quartile 2 1.59 0.56e4.50
Quartile 3 3.22 1.22e8.48
Quartile 4 9.71 3.36e28.10
CI: condence interval. p-value: level of signicance.
Table 5 Predictors of mortality: multivariate analysis for
BODEx index. Adjusted model for age, comorbidity and
blood gases.
Hazard ratio
(adjusted)
95% CI p-value
Age (years) 1.09 1.05e1.13 <0.001
Charlson index 1.12 0.83e1.51 NS
PaO
2
(mmHg) 1.00 0.98e1.03 NS
PaCO
2
(mmHg) 1.05 1.01e1.10 0.016
BODEx quartiles <0.001
Quartile 1 e e
Quartile 2 1.52 0.66e3.53
Quartile 3 3.16 1.37e7.30
Quartile 4 5.86 2.42e14.17
CI: condence interval. p-value: level of signicance.
696 J.J. Soler-Cataluna et al.
age, comorbidity, blood gases, exacerbations during the
rst year of follow-up and BODE index (Table 3). On
combining both prognostic factors (frequency of exacer-
bations and BODE index) in the form of the e-BODE index,
the mortality risk predictive capacity didnt show a signi-
cant improvement (C statistic for BODE index 0.75 vs 0.77
for e-BODE index, p ZNS). The most remarkable nding
was that registry of the severe exacerbations allows
simplication of the BODE index, since on replacing the
6MWD with the frequency of exacerbations (thus forming
the BODEx index), the predictive capacity was similar to
BODE index (C statistic of the BODEx index: 0.74 vs 0.75 for
the BODE index, p ZNS) (Fig. 3).
Severe exacerbations, i.e., those requiring hospital care
(visits to the emergency service or admissions to hospital)
are an independent mortality risk factor. In a recent study
involving 304 male patients with a mean predicted FEV
1
% of
46 17%, our group found that as the frequency and
severity of exacerbations increases, mortality risk also
increases e this effect moreover being independent of
other classical prognostic factors such as patient age, FEV
1
,
BMI, PaO
2
, PaCO
2
or comorbidity.
9
The consequences of this
observation are relevant, since the prevention of severe
exacerbations could have benecial effects the survival of
these patients. However, before accepting these results, it
is necessary to contrast the prognostic importance of COPD
exacerbations with new studies involving adjustment of the
predictive model for new and potent prognostic dimensions
such as the BODE index.
7
Different studies have revealed
a somewhat bidirectional relationship between this index
and exacerbations.
20e22
In effect, as the severity of the
disease increases, and the BODE index therefore rises, so
does the frequency of exacerbations. Ong et al.,
20
in
a study of 127 COPD patients with a mean FEV
1
of 43.7%,
found the BODE index to be useful for predicting hospital
admission risk e its predictive capacity being even greater
than that of the COPD staging system as dened by the
GOLD. Similar results have been obtained recently by
a Spanish group.
21
In the opposite sense, it has been
reported that with the appearance of new exacerbations,
the BODE index undergoes longitudinal deterioration. Cote
et al.,
22
in a study of 205 consecutive patients prospectively
followed-up during two years, showed that moderate
exacerbations induce longitudinal deterioration of the
BODE index e the latter worsening an average of 1.38
points during exacerbation, and remaining altered an
Figure 2 Survival curves by BODEx index (BODE index with
exercise capacity replaced by exacerbation frequency) in
patients with COPD, adjusted by confounding variables. The
scores range from 0 to 9 points, the highest quartile the highest
mortality.
1,0
0,8
0,6
0,4
0,2
0,0
0,0 0,2 0,4 0,6 0,8 1,0
S
e
n
s
i
t
i
v
i
t
y
1- Specificity
ROC Type II curves
0.60 0.80 0.000 0.051 0.70 Exacerbation
(1st year)
0.67 0.86 0.000 0.046 0.77 e-BODE
0.66 0.84 0.000 0.048 0.75 BODE
0.65 0.83 0.000 0.048 0.74 BODEx
0.57 0.78 0.002 0.052 0.67 Exacerbation
(previous year)
95%CI p-value SE C Parameter
C-statistics
Exacerbation frequency (previous year)
BODE index
BODEx index
e-BODE index
Exacerbation frequency (1st year)
Figure 3 Comparative ROC Type II curves and C statistics value for exacerbation frequency (during the previous year to
inclusion and during the rst year of follow-up), BODE index, BODEx index and e-BODE index as predictors of mortality in patients with
COPD. The sensitivity and specicity of e-BODE was the highest, but there were not statistical differences between ROC curves.
Exacerbation and BODE index in COPD 697
average of 0.8 and 1.1 points above the baseline score after
one and two years, respectively. These changes are not
seen in patients without exacerbations. These results point
to an important association between the BODE index and
COPD exacerbations; consequently, any prognosis predic-
tive model should include both variables. This was done in
the present study, adjusting the predictive capacity of
exacerbations to the BODE index. Here again, the
frequency of severe exacerbations was seen to constitute
an independent adverse prognostic factor, even with
respect to the above multidimensional severity index
(Table 3). These results conrm the prognostic importance
of COPD exacerbations, and thus stress the need for
preventive measures. Only male patients were included in
our study and therefore we cannot advance the same
results for a female COPD populations.
The BODE index includes four classical prognostic factors
of the disease, i.e., BMI, FEV
1
, dyspnea and the 6MWD test.
This multidimensional approach has been shown to offer
a greater predictive capacity than the classical criterion
still in use for classifying patients, i.e., FEV
1
.
7
Given the
BODE index independent adverse prognostic capacity of
COPD exacerbations, our second objective was to deter-
mine whether the combined application of both parameters
could improve the predictive capacity. We failed to show it
perhaps due to the limited sample of our study. The e-BODE
index, which incorporates all the BODE index components
together with the frequency of recent severe exacerba-
tions, yielded a C statistic of 0.77 (95% CI: 0.67e0.86),
which is slightly superior to that of the BODE index alone (C
statistic of 0.75, 95% CI: 0.66e0.84) and the frequency of
past exacerbations considered isolatedly (C statistic of
0.67, 95% CI: 0.57e0.78). However, we didnt nd statis-
tical differences when comparing ROC curves using Hanley
method.
19
BODE index proposed by Celli et al.
7
showed
a similar C statistic of 0.74, which was superior to FEV
1
.
Casanova et al., by contrast, showed a C statistic of 0.795
for BODE which also was superior to FEV
1
and inspiratory-
to-total lung capacity ratio.
23
In both studies authors didnt
use statistical method to compare ROC curves and they
accepted absolute values of C statistic as a superiority
model for predictive capacity.
To assess severe exacerbations frequency we consid-
ered three threshold values (0: none; 1: one or two severe
exacerbations; and 2: three or more severe exacerba-
tions). We feel that this decision is arbitrary but are based
on a previous study where patients were divided into the
same 3 categories.
9
Other possible approach could be
weighted on the hazard ratio for predicting death as
previously reported
9
(0: none exacerbations; 2: one or two
severe exacerbations; and 4: three or more severe exac-
erbations). However, with these scores we obtained similar
results. Finally, we decided to employ the present
threshold values for severe exacerbations because of
similar approach to threshold values proposed by Celli
et al.
7
To construct e-BODE index we used previous year
exacerbations due to two reasons. First, we considered
more easy for clinical practice to collect information about
severe AECOPD appeared during last year, and to calculate
e-BODE index. If we should use new exacerbations
appeared during the rst year of follow-up it would only
calculate the score after this follow-up. And second, when
both variables were included into the model, only previous
year exacerbations were considered signicant. The
e-BODE index with new exacerbations was slightly superior
(C statistic of 0.80, 95% CI; 0.72e0.88) but not applicable
to clinical practice.
Although the BODE index has been shown to be superior
to FEV
1
, its implantation is proving to be slow and gradual.
One of the possible factors limiting its implantation is the
need to perform a 6MWD test. Although the test is simple
and inexpensive, it requires personnel and time. For this
reason we have tried to develop a simplied index (the
BODEx index) in which the 6MWD has been replaced by the
frequency of exacerbations. The C statistic of this test is
similar to BODE index (0.75 vs 0.74) e a fact that opens the
possibility of applying this simplied index. We feel that
further studies are needed to prospectively validate the
predictive capacity of this new index.
In conclusion, the frequency of severe exacerbations
(AECOPD) is a rst-order prognostic factor that has been
shown to be independent even of the BODE index. As the
frequency of severe exacerbations increases, the patient
prognosis worsens. The adoption of preventive measures to
reduce the impact of AECOPD is therefore a priority
concern. The combined index of BODE plus exacerbations
(the e-BODE index) didnt show to improve the capacity to
predict mortality risk in COPD patients, but probably our
sample was unpowered to nd statistical differences.
Finally, we like to remark that a simplied multidimen-
sional version in which the 6-minute walking distance test is
replaced by the frequency of exacerbations (i.e., the
BODEx index), offers a good predictive capacity. This
simplication could improve the implementation of
a multidimensional score.
Conict of interest statement
The authors have no conicts to disclosure.
References
1. Rabe KF, Hurd S, Anzueto A, et al. Global strategy for the
diagnosis, management, and prevention of chronic obstructive
pulmonary disease. GOLD executive summary. Am J Respir Crit
Care Med 2007;176:532e55.
2. Murray CJL, Lopez AD. Mortality by cause for eight regions of
the world: global burden of disease study. Lancet 1997;349:
1269e76.
3. Anthonisen NR, Wright EC, Hodgkin JE. Prognosis in chronic
obstructive pulmonary disease. Am Rev Respir Dis 1986;133:
14e20.
4. Mannino DM, Buist AS, Petty TL, Enright PL, Redd SC. Lung
function and mortality in the United States: data from the rst
national health and nutrition examination survey follow up
study. Thorax 2003;58:388e93.
5. Anthonisen NR, Connett JE, Kiley JP, et al. Efects of smoking
intervention and the use of an inhaled anticholinergic bron-
chodilator on the rate of decline of FEV1. The lung health
study. JAMA 1994;272:1497e505.
6. Anthonisen NR, Skeans MA, Wise RA, et al. The effects of
a smoking cessation intervention on 14.5-year mortality:
a randomized clinical trial. Ann Intern Med 2005;142:233e9.
7. Celli BR, Cote CG, Mar n JM, et al. The body-mass index,
airow obstruction, dyspnea, and exercise capacity index in
698 J.J. Soler-Cataluna et al.
chronic obstructive pulmonary disease. N Engl J Med 2004;350:
1005e12.
8. Cote CG, Celli BR. Pulmonary rehabilitation and the BODE
index in COPD. Eur Respir J 2005;26:630e6.
9. Soler-Cataluna JJ, Mart nez-Garc a MA, Roman P, Salcedo E,
Navarro M, Ochando R. Severe acute exacerbations and
mortality in patients with chronic obstructive pulmonary
disease. Thorax 2005;60:925e31.
10. Mahler DA, Rosiello RA, Harver A, Lentine T, McGovern JF,
Daubenspeck JA. Comparison of clinical dyspnea ratings and
psychophysical measurements of respiratory sensation in
obstructive airway disease. Am Rev Respir Dis 1987;165:
1229e33.
11. Sanchis J, Casan P, Castillo J, Gonzalez N, Palenciano L,
Roca J. Normativa para la practica de la espirometr a forzada.
Arch Bronconeumol 1989;25:132e42.
12. Roca J, Sanchis J, Agust -Vidal A, et al. Spirometric reference
values for a Mediterranean population. Bull Eur Physiopathol
Respir 1986;22:217e24.
13. American Thoracic Society Committee on Prociency Stan-
dards for Clinical Pulmonary Function Laboratories. ATS
statement: guidelines for the six-minute walk test. Am J Respir
Crit Care Med 2002;166:111e7.
14. Charlson ME, Pompei P, Ales KL, MacKenzie CRL. A new method
of classifying prognostic comorbidity in longitudinal studies:
development and validation. J Chronic Dis 1987;40:373e83.
15. Rodr guez-Roisin R. Toward a consensus denition for COPD
exacerbation. Chest 2000;117:398se401s.
16. Cox DR. Regression models and life tables. J R Stat Soc 1972;
B34:187e220.
17. Kaplan EL, Meier P. Nonparametric estimation from incomplete
observation. J Am Stat Assoc 1969;53:457e81.
18. Nam B-H, DAgostino R. Discrimination index, the area under
the ROC curve. In: Huber-Carol C, Balakrishnan N, Nikulin MS,
Mesbah M, editors. Goodness-of-t test and model validity.
Boston: Birkhauser; 2002. p. 273e7.
19. Hanley JA, McNeil BJ. A method of comparing the areas under
receiver operating characteristic curves derived from the same
cases. Radiology 1983;148:839e43.
20. Ong KC, Earnest A, Lu SJ. A multidimensional grading system
(BODE index) as predictor of hospitalization for COPD. Chest
2005;128:3810e6.
21. Marin JM, Carrizo SJ, Casanova C, Martinez-Camblor P,
Soriano JB, Agust AGN, et al. Prediction of risk of COPD
exacerbations by the BODE index. Respir Med 2008;103(3):
373e8.
22. Cote CG, Dordelly LJ, Celli B. Impact of COPD exacerbations on
patient-centered outcomes. Chest 2007;131:696e704.
23. Casanova C, Cote C, de Torres JP, et al. Inspiratory-to-total
lung capacity ratio predicts mortality in patients whit chronic
obstructive pulmonary disease. Am J Respir Crit Care Med
2005;171:591e7.
Exacerbation and BODE index in COPD 699