CHEST Original Research

COPD

Bidirectional Associations Between

Clinically Relevant Depression
or Anxiety and COPD
A Systematic Review and Meta-analysis
Evan Atlantis, PhD; Paul Fahey, MMedStat; Belinda Cochrane, MD;
and Sheree Smith, PhD

Background: The longitudinal associations between depression or anxiety and COPD, and their
comorbid effect on prognosis, have not been adequately addressed by previous reviews. We aimed
to systematically assess these associations to inform guidelines and practice.
Methods: We searched electronic databases for articles published before May 2012. Longitudinal
studies in adult populations that reported an association between clinically relevant depression
or anxiety and COPD, or that reported their comorbid effect on exacerbation and/or mortality,
were eligible. Risk ratios (RRs) were pooled across studies using random-effects models and were
verified using fixed-effects models. Heterogeneity was explored with subgroup and metaregression
analyses.
Results: Twenty-two citations yielded 16 studies on depression or anxiety as predictors of COPD
outcomes (incident COPD/chronic lung disease or exacerbation) and/or mortality, in 28,759 par-
ticipants followed for 1 to 8 years, and six studies on COPD as a predictor of depression in
7,439,159 participants followed for 1 to 35 years. Depression or anxiety consistently increased the
risk of COPD outcomes (RR, 1.43; 95% CI, 1.22-1.68), particularly in higher-quality studies and
in people aged ⱕ 66 years. Comorbid depression increased the risk of mortality (RR, 1.83; 95%
CI, 1.00-3.36), particularly in men. Anxiety (or psychologic distress) increased the risk of COPD
outcomes/mortality in most studies (RR, 1.27; 95% CI, 1.02-1.58). Finally, COPD consistently
increased the risk of depression (RR, 1.69; 95% CI, 1.45-1.96).
Conclusions: Depression and anxiety adversely affect prognosis in COPD, conferring an increased
risk of exacerbation and possibly death. Conversely, COPD increases the risk of developing depres-
sion. These bidirectional associations suggest potential usefulness of screening for these disease
combinations to direct timely therapeutic intervention. CHEST 2013; 144(3):766–777

Abbreviations: GOLD 5 Global Initiative for Chronic Obstructive Lung Disease; RR 5 risk ratio

COPD is a leading and growing cause of the global

burden of disease. Worldwide surveys estimated
1
management and prevention planning, resulting in sig-
nificant health and economic benefits.
that the prevalence of COPD, defined according to the Patients with COPD often have comorbid depres-
GOLD (Global Initiative for Chronic Obstructive Lung sion or anxiety that may worsen COPD prognosis,
Disease) criteria as ⱖ stage II, was 10% in populations interfere with effective COPD management, or both.
aged ⱖ 40 years in 2006.2 This estimate increased mark-
edly with age, independent of smoking history. Because For editorial comment see page 726
the world’s population is aging, the projected increase
in COPD burden will undoubtedly cause significant Poor treatment adherence or an enhanced percep-
stress on health-care systems in high-income countries tion of COPD symptoms may herald despair and create
and may threaten economic growth in developing coun- a pathologic cycle of deteriorating health status. On
tries. Reliable information about the prognostic fac- average, it is estimated that the global prevalence of
tors associated with COPD could lead to more effective clinically relevant depression and anxiety (defined by

766 Original Research

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 diagnostic criteria or correlated cutoff scores for self- and/or mortality were eligible. Risk ratios (RRs) were pooled across
rated symptoms) is 40% in patients with COPD3 com- studies using random-effects models and verified using fixed-
effects models. Relative risks, ORs, incident rate ratios, and hazard
pared with , 10% in the general population across most ratios were included as eligible RR, without distinction because
countries surveyed.4 Depression or anxiety comorbid- each provides effect sizes of similar magnitude.18 Heterogeneity
ity in patients with COPD predicts poor adherence to was explored with subgroup and metaregression analyses. Detailed
pulmonary rehabilitation5 and COPD-related medi- sections on search strategy, study selection, data extraction, quality
cations6; decreased exercise capacity7,8 and health- assessment, and data synthesis appear in e-Appendix 1.
related quality of life9; lost productivity; and increased
health resource use, functional disability,10,11 and risk Primary Outcomes
of exacerbation12 and mortality.13,14 The primary outcomes were COPD exacerbation (as defined in
Moreover, there are complex associations between the original study to establish diagnosis) or incidence of chronic
nicotine dependence, depression and anxiety disorders, lung diseases included within the spectrum of COPD-related diag-
and smoking cessation. Prospective cohort studies noses (“emphysema” or “chronic bronchitis”; a detailed search strat-
egy describes other terms and phrases [e-Appendix 1]), mortality,
show that depression predicts smoking initiation15 and and new depression or anxiety events. According to the GOLD
increases in smoking behavior16 and decreases in phys- guidelines, an exacerbation is an acute event characterized by a
ical activity.17 Because smoking is a key lifestyle risk worsening of the patient’s respiratory symptoms that is beyond nor-
factor for COPD and reduced exercise capacity is a mal day-to-day variations and that often necessitates a change in
marker of poor prognosis in COPD, these associa- medication.19
tions have obvious mechanistic implications.
Thus, it is important to know whether the associ- Statistical Methods
ations between depression or anxiety and COPD are In examining the associations between depression or anxiety
causal or coincidental. Longitudinal associations between and COPD/mortality, RRs were pooled across studies using random-
depression or anxiety and COPD, and their comorbid effects meta-analysis models and were verified using fixed-effects
effect on prognosis, could strengthen the overall evi- models.20 For each meta-analysis model, the degree of heteroge-
neity in individual effect estimates was assessed by visual inspec-
dence of causality but have not been adequately assessed tion and by using the I2 statistic (I2 values . 40%) and the x2 test
by previous reviews.3,12 We aimed to determine whether of goodness of fit.21 When evidence of heterogeneity was observed,
(1) depression or anxiety predicts COPD outcomes we checked the data extracted from individual outlier studies,
(see the “Primary Outcomes” section), (2) depression qualitatively investigated reasons for their different results, and
or anxiety comorbidity in COPD predicts mortality, explored the effects of study exclusion in sensitivity analyses. We
explored potential sources of heterogeneity with post hoc subgroup
and (3) COPD (or chronic lung disease) predicts new and metaregression analyses using the following study-level explana-
depression or anxiety events to inform guidelines and tory variables: (1) length of follow-up, (2) sex, (3) mean age, (4) study
practice. country/region, and (5) quality score.

Materials and Methods

We searched electronic databases for articles published before
May 2012. Longitudinal studies in adult populations that reported
an association between clinically relevant depression or anxiety
and COPD or that reported their comorbid effect on exacerbation

Manuscript received July 31, 2012; revision accepted January 12,

2013.
Affiliations: From the School of Nursing and Midwifery (Drs Atlantis,
Cochrane, and Smith), the School of Science and Health (Dr Fahey),
and the School of Medicine (Dr Cochrane), University of Western
Sydney, Campbelltown Campus, NSW, Australia; Campbelltown Hos-
pital (Dr Cochrane), Department of Medicine, SLHD/SWSLHD,
Campbelltown, NSW, Australia; and the Centre for Pharmacology
and Therapeutics (Dr Smith), Division of Experimental Medicine,
Imperial College, South Kensington, London, England.
Funding/Support: The authors have reported to CHEST that
no funding was received for this study.
Correspondence to: Evan Atlantis, PhD, Family and Community
Health Research Group, School of Nursing and Midwifery, Univer-
sity of Western Sydney, Bldg 17, Room 17.1.14, Campbelltown
Campus, Locked Bag 1797, Penrith, NSW 2751 Australia; e-mail:
[email protected]
© 2013 American College of Chest Physicians. Reproduction
of this article is prohibited without written permission from the
American College of Chest Physicians. See online for more details. Figure 1. Flowchart summarizing identification of studies included
DOI: 10.1378/chest.12-1911 for review.

journal.publications.chestnet.org CHEST / 144 / 3 / SEPTEMBER 2013 767

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 We used sensitivity analysis techniques to investigate the robust- and one additional relevant article was sourced by
ness of the meta-analysis models. Studies in non-COPD popula- the author (E. A.). After further assessment of these
tions were excluded in the meta-analyses, where relevant. We also
excluded studies of anxiety or psychologic distress (rather than
50 citations, 28 were excluded for reasons listed in
depression), studies of lower quality (score , 6.0), and studies using Figure 1, leaving 22 for final inclusion in the system-
questionable (possible rather than probable) criteria to define atic review. Most studies were excluded because of
depression or anxiety. Publication bias, which reflects the tendency inadequate predictor or outcome variables or insuffi-
for smaller studies to be published in the literature only when find- cient data (e-Appendix 1).
ings are positive, was assessed visually using funnel plots.22 All cal-
culations were performed in Stata version 12 (StataCorp LP) using
the “metan,” “metareg,” and “metafunnel” commands. A two-tailed Descriptive Data Synthesis
P value , .05 was considered statistically significant throughout
the analyses.
The study characteristics of the 22 citations included
for review appear in Tables 1-3. Of these, 16 citations
yielded 13 studies assessing whether depression or
Results anxiety predicts COPD outcomes and seven studies
assessing whether comorbid depression or anxiety in
Figure 1 presents a flowchart summarizing the patients with COPD predicts mortality, published
identification of potentially relevant studies, and those between 2002 and 2012. Five studies were conducted
included and excluded. Our search strategy identified in the United States10,13,23-25; three in Asian countries14,26,27;
899 citations after duplicates were removed. Of these, two each in Nordic countries,28,29 Canada,30,31 and the
855 citations were excluded after the first screening United Kingdom32,33; and one each in Spain34 and The
of titles and/or abstracts for inclusion and exclusion Netherlands.35 All except two studies31,33 were con-
criteria, leaving 44 citations for a second full text screen- ducted in patients with COPD, and four studies were
ing. Hand searching the reference lists of these articles in populations without comorbid chronic lung or other
identified five additional potentially relevant citations, diseases at baseline.26,31,33,34

Table 1—Characteristics of Prospective Cohort Studies Reviewed

Study/Year Country Cohort Type Sample Size Female, %

Depression or anxiety exposure → COPD
or mortality outcome
Almagro et al34/2002 Spain COPD population, without comorbid lung 135 8
diseases
Coventry et al32/2011 United Kingdom COPD population 79 44
Dalal et al10/2011 United States COPD population 7,522 63
de Voogd et al35/2009 The Netherlands COPD population 122 52
Eisner et al23/2010 United States COPD population 1,202 57
Fan et al13/2007 United States COPD population 610 36
Gudmundsson et al28/2005 Nordic countries COPD population 406 51
Ito et al26/2012 Japan COPD population, without comorbid 76 9
lung/diseases
Jennings et al24/2009 United States COPD population 194 51
Laurin et al30/2009 Canada COPD population 110 51
Ng et al14/2007 Singapore COPD population 376 15
Patten et al31/2008 Canada Population, without chronic lung diseases 14,682 54
Pembroke et al33/2006 United Kingdom Population, without chronic bronchitis/diseases 2,653 59
Stage et al29/2005 Denmark COPD population 49 67
Walke et al25/2007 United States COPD population 74 50
Xu et al27/2008 China COPD population 469 31
COPD exposure → depression or
anxiety outcome
Atlantis et al40/2011 Australia Population, without depression (PAS score , 5 721 53
or ADM)
Giltay et al38/2010 Finland/Italy Population, without depressive symptoms 499 0
(Zung SDS score ⱖ 50)
Schneider et al41/2010 United Kingdom Population, without depression 10,135 57
Sode et al39/2011 Denmark Population, without depression 7,419,791 50
van den Bemt et al37/2009 The Netherlands Population, without depression 3,493 35
Wagena et al36/2005 The Netherlands Population, without depressive/anxious 4,520 27
symptoms (HADS score . 7)
ADM 5 antidepressant medication; HADS 5 Hospital Anxiety and Depression Scale; PAS 5 Psychogeriatric Assessment Scale; Zung SDS 5 Zung
Self-rating Depression Scale.

768 Original Research

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 Table 2—Characteristics of Prospective Cohort Studies Reviewed

Mean Predictor
Study/Year Age, y Predictor, Assessment Prevalence, % Outcome, Assessment
Depression or anxiety
exposure → COPD
or mortality outcome
Almagro et al34/2002 72 Depressive symptoms, NR Mortality
Yesavage score ⱖ 11 (PDA)
Coventry et al32/2011 65 Depressive symptoms, 43 COPD exacerbation, hospital
HADS score ⱖ 15 (PDA) readmission
Dalal et al10/2011 64 Depression or anxiety, prescriptions & 50 COPD exacerbation, hospital ED
ICD-9 criteria (PDA) presentation or readmission
de Voogd et al35/2009 61 Depressive symptoms, HADS score ⱖ 8 33 Mortality
Eisner et al23/2010 57 Anxious symptoms, HADS score ⱖ 8 15 COPD exacerbation, hospital readmission
Fan et al13/2007 67 Depressive symptoms, NR COPD exacerbation, hospital presentation
BDI score ⱖ 15 (PDA) (respiratory related)/mortality
Gudmundsson et al28/2005 69 Depressive/anxious symptoms, 29/41 COPD exacerbation, hospital readmission
HADS score ⱖ 8
Ito et al26/2012 70 Depressive symptoms, 18 COPD exacerbation, hospital
CES-D score ⱖ 16 readmission/mortality-hospitalization
Jennings et al24/2009 67 Depressive symptoms, MH score ⱕ 20 17 COPD exacerbation, patient records
Laurin et al30/2009 66 Depression or anxiety, DSM-IV 49 COPD exacerbation (consistent with
criteria (PDA) GOLD), hospital readmission or
additional treatment
Ng et al14/2007 72 Depressive symptoms, HADS score ⱖ 8 44 COPD exacerbation, hospital
readmission/mortality
Patten et al31/2008 44 Probable depression, CIDI-SF (PDA) 6 Chronic bronchitis/emphysema,
questionnaire
Pembroke et al33/2006 54 Psychologic distress symptoms, 14/20 Chronic bronchitis, questionnaire
GHQ score ⱖ 4
Stage et al29/2005 71 Depression, ICD-10 criteria (PDA) 47 Mortality
Walke et al25/2007 72 Depressive/anxious symptoms, 30 (49) Mortality
ESAS moderate or severe
Xu et al27/2008 66 Depressive/anxious symptoms, 10/5 COPD exacerbation (consistent with
HADS score ⱖ 11 vs ⱕ 7 (PDA) GOLD), hospital readmission
COPD exposure → depression
or anxiety outcome
Atlantis et al40/2011 73 Chronic bronchitis, emphysema or 3 Depressive symptoms, PAS score ⱖ 5
asthma, questionnaire
Giltay et al38/2010 57 Chronic lung diseases (bronchial 26 Depressive symptoms, Zung SDS
asthma, pulmonary emphysema, score ⱖ 50
chronic bronchitis), spirometry
Schneider et al41/2010 66 COPD, GP diagnosis 46 Depression, GP
Sode et al39/2011 55 case/26 COPD, ICD-8/10 criteria 4 Depression, ICD-8/10 criteria
controls
van den Bemt et al37/2009 62 COPD, GP diagnosis 29 Depression, GP
Wagena et al36/2005 42 Chronic bronchitis, questionnaire 2 Depressive/anxious symptoms,
HADS score ⱖ 11
BDI 5 Beck Depression Inventory; CES-D 5 Centre for Epidemiologic Studies Depression scale; CIDI-SF 5 Composite International Diagnostic
Interview Short Form; DSM-IV 5 Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition); ESAS 5 Edmonton Symptom
Assessment Scale; GHQ 5 General Health Questionnaire; GOLD 5 Global Initiative for Chronic Obstructive Lung Disease; GP 5 general
practitioner; ICD 5 International Classification of Diseases; MH 5 Mental Health item from the 36-Item Short-Form Health Survey; NR 5 not
reported; PDA 5 probable depression or anxiety (good evidence based on clinical interview or valid instrument using a high cutoff point for
clinically relevant symptoms). See Table 1 for expansion of other abbreviations.

The sample sizes ranged from 49 to 14,682, resulting and by psychometrics using various cutoff points in
in a total of 28,759 participants across studies. The the remaining 13 studies. There was good evidence
mean age of the samples ranged from 44 to 72 years. of probable depression or anxiety (clinical interview or
All except one study sampled both sexes (overall, valid instrument using a high cutoff point for clinically
54% men), and reported effect estimates in men and relevant symptoms) in only eight of 16 studies.10,13,27,29-32,34
women separately.33 Depression or anxiety predictors The prevalence of depression or anxiety predictors
were defined by physician diagnosis or the use of anti- (exposure category) reported in 14 studies ranged from
depressant/antianxiety medications in three studies,10,29,30 5% to 50%. The primary outcomes consisted of COPD

journal.publications.chestnet.org CHEST / 144 / 3 / SEPTEMBER 2013 769

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 exacerbation by hospital readmission in nine studies, out depression or anxiety at baseline. The sample sizes
only two of which reported definitions that were ranged from 499 to 7,419,791, resulting in a total of
consistent with GOLD guidelines,27,30 and by patient 7,439,159 participants across studies. The mean age
records in one study,24 chronic bronchitis or emphy- of the samples ranged from 26 to 73 years. All except
sema by self-report in two studies,31,33 and mortality in one study sampled both sexes (overall, 50% men), and
seven studies.13,14,25,26,29,34,35 Total or median follow-up this study reported effect estimates in men only.38
periods ranged from 1 to 8 years. Mean quality scores COPD predictors consisted of COPD by physician
ranged from 4.8 to 6.6, and four studies received a diagnosis in three studies37,39,41; by spirometry in one
score of ⱖ 6.0.10,24,31,33 study38; and by self report for chronic bronchitis in one
Six of the 22 studies included for review published study36 and chronic bronchitis, emphysema, or asthma
between 2005 and 2011 assessed whether COPD or in another study.40 The prevalence of COPD predic-
chronic lung disease predicts depression. Only one of tors (exposure category) reported in six studies ranged
the six studies also assessed whether chronic bronchi- from 2% to 46%. The primary outcomes consisted of
tis predicts anxiety.36 Two studies were conducted in depression defined by physician diagnosis or use of
The Netherlands,36,37 and one each in Finland/Italy,38 antidepressant medications in three studies,37,39,41 and
Denmark,39 Australia,40 and the United Kingdom.41 by psychometrics using various cutoff points in the
All the studies were conducted in populations with- other three studies.36,38,40 The total follow-up period

Table 3—Characteristics of Prospective Cohort Studies Reviewed

Follow-up,
Total or Risk Quality
Study/Year Median, y Expression Covariate Adjustments/Considerations Score
Depression or anxiety
exposure → COPD
or mortality outcome
Almagro et al34/2002 2.3, median OR Age, sex, FEV1, functional status, medications, dyspnea, SGRQ, medical 5.2
conditions, past-year readmission, marital status, discharge Paco2
Coventry et al32/2011 1 OR Age, sex, FEV1, smoking, comorbidity, previous COPD admission 5.7
Dalal et al10/2011 1 OR Age, sex, comorbidity, medications, medical costs 6.4
de Voogd et al35/2009 7, median HR Age, sex, partner status, smoking, FEV1, BMI, depressive symptoms, 5.2
Type D personality, exercise capacity
Eisner et al23/2010 2.1, median HR Age, sex, race, education, income, smoking, BODE index, COPD severity 5.7
Fan et al13/2007 1/3 OR Age, sex, lung function/capacity, BODE index, comorbidity, antidepressants 5.1
Gudmundsson et al28/2005 1 HR Age, smoking, FEV1, SGRQ 4.8
Ito et al26/2012 1 RR BMI, GOLD stage, Pao2, Paco2, treatments (long-term oxygen therapy; 5.3
noninvasive positive pressure ventilation; corticosteroids)
Jennings et al24/2009 1 OR Age, sex, FEV1, exercise capacity, BMI, treatments (corticosteroids; use 6.6
of home oxygen), comorbidities
Laurin et al30/2009 2 RR Age, sex, length of COPD diagnosis, disease severity, smoking, setting, 5.7
comorbidity, follow-ups, past hospitalizations
Ng et al14/2007 1, median HR Age, sex, marital status, hospital length of stay, BMI, FEV1, FVC, Pao2, 5.8
Paco2, COPD duration, dyspnea, smoking, past readmissions, SGRQ
Patten et al31/2008 8 HR Age, sex, health service use, smoking 6.4
Pembroke et al33/2006 3, median OR Age, sex stratification, socioeconomic status, BMI, FEV1 6.2
Stage et al29/2005 2.2 HR Age, sex, smoking, antidepressants, FEV1, corticosteroids 5.9
Walke et al25/2007 2 OR None 5.0
Xu et al27/2008 1 IRR Age, sex, smoking, marital status, education/employment, FEV1, 5.3
dyspnea, walking capacity, social support, COPD-related self-efficacy,
comorbidities, hospital type, respiratory medications, oxygen
therapy, past readmissions
COPD exposure → depression
or anxiety outcome
Atlantis et al40/2011 2 RR None 5.5
Giltay et al38/2010 5 OR Age, country, socioeconomic status, marital status, smoking, BMI 6.7
Schneider et al41/2010 10 OR Smoking, BMI, infections, sleep disorders, medical conditions 6.9
Sode et al39/2011 26 OR Age, sex, descent, geographical residency, education 8.4
van den Bemt et al37/2009 35 HR Age, sex, socioeconomic status, GP setting, comorbidity 7.1
Wagena et al36/2005 1.5 OR Age, sex, education, smoking 6.6
BODE index 5 based on BMI, degree of airflow obstruction measured by FEV1, grade of dyspnea assessed by the modified Medical Research
Council dyspnea scale, and exercise capacity measured by the 6-min walk test; HR 5 hazard ratio; IRR 5 incident rate ratio; RR 5 relative risk;
SGRQ 5 St. George’s Respiratory Questionnaire.

770 Original Research

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 ranged from 1.5 to 35 years. The mean quality scores were larger in the higher- vs lower-quality study group
ranged from 5.5 to 8.4, and all but one study40 received (1.78 vs 1.21, P , .01) and in the younger vs older
a score of ⱖ 6.0. population group (1.54 vs 1.15, P 5 .07). Similarly, the
only variables found to be predictive of the log RR in
metaregressions were quality score (1.35 [95% CI,
Quantitative Data Synthesis
1.00-1.81] per 1-numerical point increment) and mean
Depression or Anxiety Exposure and Risk of COPD age (0.98 [95% CI, 0.96-1.00] per 1-year increment).
Outcomes: Figure 2 presents RRs with 95% CIs for In addition, a funnel plot was produced and showed
all 13 studies assessing the association between depres- only slight evidence of publication bias, because there
sion or anxiety and risk of COPD outcomes. The pooled were only two small studies (sample sizes n , 100),24,26
RR of 1.43 (95% CI, 1.22-1.68) shows that depression both of which had RR estimates that were larger and
or anxiety was associated with increased risk of COPD aligned with the pooled RR (e-Fig 1).
outcomes. There was a high degree of heterogeneity
among studies (I2 5 56.5%, P 5 .06) that was mostly a Comorbid Depression Exposure in COPD and Risk
result of variation in degree of deference rather than of Mortality: Figure 4 presents RRs (with 95% CIs)
in direction of effect estimates to the null. The sensi- for all seven studies assessing the association between
tivity analyses presented in Table 4 show that the pooled comorbid depression in COPD and risk of mortality.
RR was similar after exclusion of anxiety or psycho- The pooled RR of 1.83 (95% CI, 1.00-3.36) shows that
logic distress studies (1.45; 95% CI, 1.21-1.74) and comorbid depression was associated with increased
one study with extreme CIs (1.43; 95% CI, 1.21-1.68), risk of mortality in patients with COPD. There was a
but was substantially changed after exclusion of lower- high degree of heterogeneity among studies (I2 5 70.2%,
quality studies (increased to 1.78; 95% CI, 1.50-2.11), P 5 .003), which was mostly due to a single outlier
studies using possible rather than probable depression study29 that had an RR opposite in direction to the
or anxiety criteria (increased to 1.58; 95% CI, 1.35-1.86), pooled RR (showing depression was protective against
and studies in non-COPD populations (decreased to mortality). After exclusion of the single outlier study,
1.31; 95% CI, 1.12-1.54). the pooled RR increased to 2.29 (95% CI, 1.53-3.43)
In subgroup analyses (Fig 3), effect estimates did and there was less heterogeneity evident (I2 5 32.0%,
not differ among studies by length of follow-up, sex, P 5 .196). The sensitivity analysis presented in Table 5
or study region explanatory variables. However, RRs shows that the pooled RR was substantially changed

Figure 2. Risk ratios for all 13 studies assessing the association between depression or anxiety and risk
of COPD outcomes.

journal.publications.chestnet.org CHEST / 144 / 3 / SEPTEMBER 2013 771

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 Table 4—Sensitivity Analysis of Depression or Anxiety Exposure → COPD Outcomes Meta-analysis

Sensitivity Analysis No. Studies Total Sample Size Risk Ratio (95% CI) P Value
Fixed-effects model 13 30,137 1.41 (1.29-1.55) .012
Exclusion of anxiety or psychologic distress studies 10 25,050 1.43 (1.20-1.71) .015
Exclusion of lower-quality studies (score , 6.0) 5 26,737 1.69 (1.48-1.94) .617
Exclusion of 3 studies in non-COPD populations 10 10,998 1.31 (1.12-1.54) .041
Exclusion of 7 studies using possible rather than probable 6 23,998 1.58 (1.35-1.86) .148
depression criteria
Exclusion of 1 study with extreme CIs 12 30,061 1.43 (1.21-1.68) .004
Pembroke et al men and women in Figure 2 have been counted as two studies. P value is for heterogeneity within each sensitivity analysis (not
33

for tests that risk ratio 5 1).

after exclusion of studies using possible rather than produced and showed only slight evidence of publi-
probable depression criteria (decreased to 1.47 [95% CI, cation bias, because there was only one small study26
0.37-5.78]) and one study with extreme CIs (decreased whose RR estimate was larger and aligned with the
to 1.65 [95% CI, 0.93-2.91]). pooled RR (e-Fig 2).
In subgroup analyses (Fig 5), there was a trend for
effect estimates that was larger in the group of studies Ancillary Meta-analysis: Anxiety or Psychologic
containing a predominance of men (2.82 vs 0.90, Distress Exposure and Risk of COPD Outcomes or
P 5 .08). Similarly, there was a trend for male sex pre- Mortality: Figure 6 presents RRs (with 95% CIs) for
dicting the log RR in metaregression (1.03; 95% CI, all six studies assessing the association between anxiety
1.00-1.07; P 5 .07 per 1 percentage point increment or psychologic distress exposure (rather than depres-
in percentage of men). In addition, a funnel plot was sion) and risk of COPD outcome or mortality (one

Figure 3. Results of subgroup analyses of 13 studies assessing the association between depression or
anxiety and risk of COPD outcomes. df 5 degrees of freedom.

772 Original Research

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 Figure 4. Risk ratios for all seven studies assessing the association between comorbid depression in
COPD and risk of mortality.

study25), in an ancillary meta-analysis. The pooled RR (I2 5 27.7%, P 5 .246). The sensitivity analysis pre-
of 1.27 (95% CI, 1.02-1.58) shows that anxiety or psy- sented in Table 6 shows that the pooled RR was sim-
chologic distress, like depression, was associated with ilar after exclusion of three studies in populations with
an increased risk of COPD outcomes or mortality. self-reported chronic lung disease38,40 (1.55 [95% CI,
There was a low degree of heterogeneity among studies 1.45-1.65]), and the single lower-quality study40 (1.65
(I2 5 21.8%, P 5 .270), which was mostly attributed to [95% CI, 1.43-1.91]). Limited data from one study
two large studies23,28 that had smaller RR estimates suggested that chronic bronchitis may also increase
aligned with the pooled RR. the risk of anxiety (RR, 5.09; 95% CI, 2.91-8.89).36
Effect estimates were not associated with explana-
COPD Exposure and Risk of Depression or Anxiety: tory variables in subgroup analyses (e-Fig 3) or metare-
Figure 7 presents RRs (with 95% CIs) for all six studies gressions. In addition, a funnel plot was produced
assessing the association between COPD and risk of and showed only slight evidence of publication bias,
depression. The pooled RR of 1.69 (95% CI, 1.45-1.96) because all studies had relatively large sample sizes
shows that COPD was associated with increased risk (e-Fig 4).
of depression. There was a high degree of heteroge-
neity among studies (I2 5 70.8%, P 5 .004), which was
mostly a result of variation in degree of deference, Discussion
because all six studies had RR estimates with 95% CIs
Summary of Evidence
farther from the null. After exclusion of two outlier
studies,36,40 the pooled RR decreased to 1.55 (95% CI, We have confirmed that depression or anxiety con-
1.46-1.66) and the degree of heterogeneity lessened fers an increased risk of COPD outcomes and possibly

Table 5—Sensitivity Analysis of Depression or Anxiety Exposure → Mortality

Sensitivity Analysis No. Studies Total Sample Size Risk Ratio (95% CI) P Value
Fixed-effects model 7 1,442 1.92 (1.42-2.61) .003
Exclusion of 4 studies using possible rather than probable 3 794 1.47 (0.37-5.78) .001
depression criteria
Exclusion of 1 study with extreme CIs 1 1,366 1.65 (0.93-2.91) .007
P value is for heterogeneity within each sensitivity analysis (not for tests that risk ratio 5 1).

journal.publications.chestnet.org CHEST / 144 / 3 / SEPTEMBER 2013 773

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 Figure 5. Results of subgroup analyses of seven studies assessing the association between comorbid
depression in COPD and risk of mortality. See Figure 3 legend for expansion of abbreviation.

death. Moreover, COPD increases the risk of devel- one.29 The conflicting result could have been due to
oping depression. People with depression or anxiety chance, because that outlier study was conducted in
had a 43% increased risk of COPD outcomes (Fig 2), only 49 patients.29 The subgroup analyses suggested
and patients with COPD and depression or anxiety that effect estimates were larger in studies with a male
had a 31% increased risk of exacerbation (Table 4). predominance. This could have been due to several
Compared with the effect estimates in a recently factors other than comorbid depression, such as the
reported systematic review,12 our results for COPD higher prevalence of COPD,2 ever smoking,2 and mor-
exacerbation risk are two- to threefold greater and more tality rates1 observed in men. However, the signifi-
conclusive. The effect estimate for COPD outcomes cant effect estimate for mortality was partially driven
increased in size for higher-quality studies, in studies by one study that had a very large and imprecise effect
with good evidence of clinically relevant depression estimate26 (Table 5). In addition, the results of our ancil-
or anxiety (probable depression/anxiety), and in those lary analysis suggested that people with anxiety or
containing people aged ⱕ 66 years (Fig 3); remained psychologic distress had an increased risk of COPD
robust after exclusion of anxiety or psychologic dis- outcome or mortality (Fig 6), but the size of the effect
tress studies; and decreased in size after exclusion of estimate (27%) was smaller when compared with our
three studies in non-COPD populations (Table 4). results for depression.
Thus, limited evidence based on subgroup analyses People with COPD had a 55% to 69% increased
suggests that depression or anxiety is a strong predictor risk of developing depression (Fig 7), and effect esti-
of incident chronic lung disease and a most useful mates were robust across all six reviewed studies.
prognostic indicator when there is good evidence of Furthermore, the limited data suggest that chronic
psychopathology, especially in younger patients with bronchitis alone may increase the risk of anxiety. Once
COPD. comorbidity is present, depression or anxiety worsens
Patients with COPD and comorbid depression had COPD prognosis and may interfere with effective
an 83% increased risk of mortality (Fig 4), and effect COPD management, as described in the introduc-
estimates were consistent across all studies except tion. The increased burden of comorbid depression or

774 Original Research

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 Figure 6. Risk ratios for all six studies assessing the association between anxiety or psychologic distress
and risk of COPD outcomes or mortality (in Walke et al25) in an ancillary meta-analysis.

anxiety in COPD likely rises with the degree of dis- oping depression, which should inform guidelines and
ease severity. Accordingly, airflow limitation, a conven- practice. Furthermore, we found evidence to suggest
tional measure of COPD severity, tended to be greater that future higher-quality studies will likely strengthen
among patients with, than without, depression or anx- rather than weaken this evidence base.
iety in several of the studies we reviewed.13,14,23,27
Collectively, the existing literature is sufficient to
Limitations
draw the conclusions that depression or anxiety confers
an increased risk of COPD outcomes and possibly Several limitations require consideration. First,
mortality too, and COPD increases the risk of devel- because only a small number of studies conducted in

Figure 7. Risk ratios for all six studies assessing the association between COPD and risk of depression.

journal.publications.chestnet.org CHEST / 144 / 3 / SEPTEMBER 2013 775

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 Table 6—Sensitivity Analysis of COPD Exposure → Depression Meta-analysis

Sensitivity Analysis No. Studies Total Sample Size Risk Ratio (95% CI) P Value
Fixed-effects model 6 7,439,159 1.58 (1.54-1.61) .004
Exclusion of a lower-quality study (score , 6.0) 5 7,438,438 1.65 (1.43-1.91) .006
Exclusion of 3 studies in non-COPD populations 3 7,433,419 1.55 (1.45-1.65) .215
P value is for heterogeneity within each sensitivity analysis (not for tests that risk ratio 5 1).

specific populations were included, the findings of this bidities, because timely diagnosis and treatment may
review may not be relevant to other countries and key favorably affect COPD prognosis.
groups; thus, further research is required. Second, only
five10,24,31,33 out of 13 studies of depression or anxiety
exposure and risk of COPD outcomes, and none of Acknowledgments
seven studies of comorbid depression exposure in Author contributions: Dr Atlantis is the guarantor of the paper
and takes responsibility for the integrity of the work as a whole,
patients with COPD and risk of mortality, received a from inception to published article.
score of ⱖ 6.0. This suggests there were several poten- Dr Atlantis: contributed to the conception and design of the review,
tial sources of bias that could have weakened our results, identification of studies for inclusion, extraction and interpretation
of data, drafting of the article, and approval of the final completed
because we found that effect estimates increased with article.
quality score. For instance, two lower-quality studies Dr Fahey: contributed to the extraction, analysis, and interpretation
had conflicting findings29 or very imprecise effect esti- of data; revision of the article; and approval of the final completed
article.
mates (extremely wide CIs) that biased the effect Dr Cochrane: contributed to the extraction and interpretation of data,
estimate for mortality,26 and there was substantial het- revision of the article, and approval of the final completed article.
erogeneity in assessments and definitions of COPD Dr Smith: contributed to the interpretation of data, revision of the
article, and approval of the final completed article.
exacerbation and depression or anxiety among stud- Financial/nonfinancial disclosures: The authors have reported
ies. Third, we acknowledge that our use of checklist- to CHEST the following conflicts of interest: Dr Cochrane has
derived summary scores to distinguish higher- and received pharmaceutical company sponsorship to attend national
and international symposia from GlaxoSmithKline plc, Novartis AG,
lower-quality studies is questionable and should be and Boehringer Ingelheim GmbH and has received honoraria as
interpreted with caution. However, we were unable a speaker from GlaxoSmithKline plc. Drs Atlantis, Fahey, and Smith
to assess each quality item individually using subgroup have reported that no potential conflicts of interest exist with any
companies/organizations whose products or services may be dis-
and metaregression analyses because of the high num- cussed in this article.
ber of quality items on our checklist (n 5 9) and low Other contributions: This research was performed at the Univer-
statistical power. Fourth, failure to explore and/or detect sity of Western Sydney. We are grateful to Geoffrey Lattimore, BA,
Dip Teaching, for his work in developing and conducting the elec-
significant sources of heterogeneity among studies using tronic database searches.
subgroup and metaregression analyses could have been Additional information: The e-Appendix and e-Figures can be
due to the small number of studies included, result- found in the “Supplemental Materials” area of the online article.
ing in low statistical power. Fifth, reviewer-level lim-
itations include incomplete retrieval of information
for several of the 28 citations excluded and the exis- References
tence of other relevant studies not identified with our 1. World Health Organization. The Global Burden of Disease: 2004
Update. Geneva, Switzerland: World Health Organization;
search strategy despite having consulted several experts 2008.
in the field, resulting in selection bias. However, this 2. Buist AS, McBurnie MA, Vollmer WM, et al; BOLD Col-
was an unlikely source of bias, because the results and laborative Research Group. International variation in the
conclusions reported in most of the excluded studies prevalence of COPD (the BOLD Study): a population-based
were in line with those reported herein. prevalence study. Lancet. 2007;370(9589):741-750.
3. Yohannes AM, Willgoss TG, Baldwin RC, Connolly MJ.
Depression and anxiety in chronic heart failure and chronic
obstructive pulmonary disease: prevalence, relevance, clin-
Conclusions ical implications and management principles. Int J Geriatr
Psychiatry. 2010;25(12):1209-1221.
We have confirmed that depression or anxiety con- 4. Demyttenaere K, Bruffaerts R, Posada-Villa J, et al; WHO
fers an increased risk of COPD outcomes and possibly World Mental Health Survey Consortium. Prevalence, severity,
death, and that COPD increases the risk of developing and unmet need for treatment of mental disorders in the
depression. Consideration of these bidirectional associ- World Health Organization World Mental Health Surveys.
ations and the adverse prognostic impact of comorbid JAMA. 2004;291(21):2581-2590.
5. Fan VS, Giardino ND, Blough DK, Kaplan RM, Ramsey SD;
COPD and depression or anxiety is an important step Nett Research Group. Costs of pulmonary rehabilitation and
in informing physician practice. Our results are suffi- predictors of adherence in the National Emphysema Treat-
cient to justify screening for this combination of comor- ment Trial. COPD. 2008;5(2):105-116.

776 Original Research

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013

 6. Khdour MR, Hawwa AF, Kidney JC, Smyth BM, McElnay JC. 24. Jennings JH, Digiovine B, Obeid D, Frank C. The association
Potential risk factors for medication non-adherence in patients between depressive symptoms and acute exacerbations of
with chronic obstructive pulmonary disease (COPD). Eur J COPD. Lung. 2009;187(2):128-135.
Clin Pharmacol. 2012;68(10):1365-1373. 25. Walke LM, Byers AL, Tinetti ME, Dubin JA, McCorkle R,
7. Spruit MA, Watkins ML, Edwards LD, et al; Evaluation Fried TR. Range and severity of symptoms over time among
of COPD Longitudinally to Identify Predictive Surrogate older adults with chronic obstructive pulmonary disease and
Endpoints (ECLIPSE) study investigators. Determinants heart failure. Arch Intern Med. 2007;167(22):2503-2508.
of poor 6-min walking distance in patients with COPD: the 26. Ito K, Kawayama T, Shoji Y, et al. Depression, but not sleep
ECLIPSE cohort. Respir Med. 2010;104(6):849-857. disorder, is an independent factor affecting exacerbations and
8. An L, Lin YX, Yang T, et al. Predictive validity of BODE hospitalization in patients with chronic obstructive pulmo-
index for anxious and depressive symptoms in patients with nary disease. Respirology. 2012;17(6):940-949.
chronic obstructive pulmonary disease. Chin Med J (Engl). 27. Xu W, Collet JP, Shapiro S, et al. Independent effect of depres-
2010;123(14):1845-1851. sion and anxiety on chronic obstructive pulmonary disease
9. Tsiligianni I, Kocks J, Tzanakis N, Siafakas N, van der Molen T. exacerbations and hospitalizations. Am J Respir Crit Care
Factors that influence disease-specific quality of life or health Med. 2008;178(9):913-920.
status in patients with COPD: a review and meta-analysis of 28. Gudmundsson G, Gislason T, Janson C, et al. Risk factors for
Pearson correlations. Prim Care Respir J. 2011;20(3):257-268. rehospitalisation in COPD: role of health status, anxiety and
10. Dalal AA, Shah M, Lunacsek O, Hanania NA. Clinical and depression. Eur Respir J. 2005;26(3):414-419.
economic burden of depression/anxiety in chronic obstruc- 29. Stage KB, Middelboe T, Pisinger C. Depression and chronic
tive pulmonary disease patients within a managed care popu- obstructive pulmonary disease (COPD). Impact on survival.
lation. COPD. 2011;8(4):293-299. Acta Psychiatr Scand. 2005;111(4):320-323.
11. Egede LE. Major depression in individuals with chronic med- 30. Laurin C, Labrecque M, Dupuis G, Bacon SL, Cartier A,
ical disorders: prevalence, correlates and association with health Lavoie KL. Chronic obstructive pulmonary disease patients
resource utilization, lost productivity and functional disability. with psychiatric disorders are at greater risk of exacerbations.
Gen Hosp Psychiatry. 2007;29(5):409-416. Psychosom Med. 2009;71(6):667-674.
12. Laurin C, Moullec G, Bacon SL, Lavoie KL. Impact of anx- 31. Patten SB, Williams JV, Lavorato DH, Modgill G, Jetté N,
iety and depression on chronic obstructive pulmonary disease Eliasziw M. Major depression as a risk factor for chronic dis-
exacerbation risk. Am J Respir Crit Care Med. 2012;185(9): ease incidence: longitudinal analyses in a general population
918-923. cohort. Gen Hosp Psychiatry. 2008;30(5):407-413.
13. Fan VS, Ramsey SD, Giardino ND, et al; National Emphy- 32. Coventry PA, Gemmell I, Todd CJ. Psychosocial risk factors
sema Treatment Trial (NETT) Research Group. Sex, depression, for hospital readmission in COPD patients on early discharge
and risk of hospitalization and mortality in chronic obstructive services: a cohort study. BMC Pulm Med. 2011;11:49.
pulmonary disease. Arch Intern Med. 2007;167(21):2345-2353. 33. Pembroke TP, Rasul F, Hart CL, Davey Smith G, Stansfeld
14. Ng TP, Niti M, Tan WC, Cao Z, Ong KC, Eng P. Depressive SA. Psychological distress and chronic obstructive pulmo-
symptoms and chronic obstructive pulmonary disease: effect nary disease in the Renfrew and Paisley (MIDSPAN) study.
on mortality, hospital readmission, symptom burden, func- J Epidemiol Community Health. 2006;60(9):789-792.
tional status, and quality of life. Arch Intern Med. 2007;167(1): 34. Almagro P, Calbo E, Ochoa de Echagüen A, et al. Mortality
60-67. after hospitalization for COPD. Chest. 2002;121(5):1441-1448.
15. Patton GC, Carlin JB, Coffey C, Wolfe R, Hibbert M, Bowes G. 35. de Voogd JN, Wempe JB, Postema K, et al. More evidence
Depression, anxiety, and smoking initiation: a prospective study that depressive symptoms predict mortality in COPD patients:
over 3 years. Am J Public Health. 1998;88(10):1518-1522. is type D personality an alternative explanation? Ann Behav
16. van Gool CH, Kempen GI, Penninx BW, Deeg DJ, Beekman Med. 2009;38(2):86-93.
AT, van Eijk JT. Relationship between changes in depressive 36. Wagena EJ, van Amelsvoort LGPM, Kant I, Wouters EF.
symptoms and unhealthy lifestyles in late middle aged and Chronic bronchitis, cigarette smoking, and the subsequent
older persons: results from the Longitudinal Aging Study onset of depression and anxiety: results from a prospective
Amsterdam. Age Ageing. 2003;32(1):81-87. population-based cohort study. Psychosom Med. 2005;67(4):
17. Patten SB, Williams JV, Lavorato DH, Eliasziw M. A longi- 656-660.
tudinal community study of major depression and physical 37. van den Bemt L, Schermer T, Bor H, et al. The risk for depres-
activity. Gen Hosp Psychiatry. 2009;31(6):571-575. sion comorbidity in patients with COPD. Chest. 2009;135(1):
18. Davies HTO, Crombie IK, Tavakoli M. When can odds ratios 108-114.
mislead? BMJ. 1998;316(7136):989-991. 38. Giltay EJ, Nissinen A, Giampaoli S, Zitman FG, Kromhout D.
19. Global Initiative for Chronic Obstructive Lung Disease. Global Low respiratory function increases the risk of depressive
strategy for the diagnosis, management, and prevention of symptoms in later life in men. Psychosom Med. 2010;72(1):
chronic obstructive pulmonary disease. Revised 2011. GOLD 53-60.
website. http://www.goldcopd.org/guidelines-global-strategy- 39. Sode BF, Dahl M, Nordestgaard BG. Myocardial infarction
for-diagnosis-management.html. Accessed May 2013. and other co-morbidities in patients with chronic obstructive
20. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control pulmonary disease: a Danish nationwide study of 7.4 million
Clin Trials. 1986;7(3):177-188. individuals. Eur Heart J. 2011;32(19):2365-2375.
21. Higgins JP, Thompson SG. Quantifying heterogeneity in a 40. Atlantis E, Grayson DA, Browning C, Sims J, Kendig H.
meta-analysis. Stat Med. 2002;21(11):1539-1558. Cardiovascular disease and death associated with depression
22. Egger M, Davey Smith G, Schneider M, Minder C. Bias in and antidepressants in the Melbourne Longitudinal Studies on
meta-analysis detected by a simple, graphical test. BMJ. 1997; Healthy Ageing (MELSHA). Int J Geriatr Psychiatry. 2011;
315(7109):629-634. 26(4):341-350.
23 . Eisner MD, Blanc PD, Yelin EH, et al. Influence of anxiety 41. Schneider C, Jick SS, Bothner U, Meier CR. COPD and the
on health outcomes in COPD. Thorax. 2010;65(3):229-234. risk of depression. Chest. 2010;137(2):341-347.

journal.publications.chestnet.org CHEST / 144 / 3 / SEPTEMBER 2013 777

Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 09/03/2013