Co-Morbidity, Body Mass Index and Quality of Life in COPD Using the Clinical COPD Questionnaire

Abstract

Introduction: Quality of life is an important patient-oriented measure in COPD. The Clinical COPD Questionnaire (CCQ) is a validated instrument for estimating quality of life. The impact of different factors on the CCQ-score remains an understudied area. The aim of this study was to investigate the association of co-morbidity and body mass index with quality of life measured by CCQ. Methods: A patient questionnaire including the CCQ and a review of records were used. A total of 1548 COPD patients in central Sweden were randomly selected. Complete data were collected for 919 patients, 639 from primary health care and 280 from hospital clinics. Multiple linear regression with adjustment for sex, age, level of education, smoking habits and level of care was performed. Subanalyses included additional adjustment for lung function in the subgroup (n = 475) where spirometry data were available. Results: Higher mean CCQ score indicating lower quality of life was statistically significant and independently associated with heart disease (adjusted regression coefficient (95%CI) 0.26; 0.06 to 0.47), depression (0.50; 0.23 to 0.76) and underweight (0.58; 0.29 to 0.87). Depression and underweight were associated with higher scores in all CCQ subdomains. Further adjustment for lung function in the subgroup with this measure resulted in statistically significant and independent associations with CCQ for heart disease, depression, obesity and underweight. Conclusion: The CCQ identified that heart disease, depression and underweight are independently associated with lower health-related quality of life in COPD.

Keywords::

• COPD
• Quality of life
• Co-morbidity
• Body Mass Index

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) contributes increasingly to the global burden of morbidity, and is predicted to be ranked fifth among diseases influencing disability-adjusted life years lost in 2020 (1). In addition to respiratory symptoms and physical incapacity, patients with COPD often also experience secondary problems including anxiety, depression and social isolation. Therefore, health related quality of life (HRQL) and its measurement are important for evaluating the consequences of the disease (2).

Previous studies have shown that HRQL in COPD is influenced by low physical capacity, a high level of dyspnoea and worsening of emotional state, but it is also associated with female sex, reduced lung function, older age, a history of heavy tobacco use and frequent exacerbations (3–12).

Heart disease, diabetes and depression are common among COPD patients (13–15), and heart disease is known to influence mortality in COPD (16). Body mass index (BMI) may also be relevant in COPD, as both low and high BMI are associated with adverse disease outcomes and poor HRQL in general populations (17, 18) and increased mortality among COPD patients (19). Several authors have concluded that the presence of one or more co-morbid conditions influences HRQL in COPD patients (20, 21).

Measurement of HRQL in COPD has most commonly been estimated either by generic instruments such as the Short-Form-36 (SF-36) and Short-Form-12 (SF-12) (22, 23), or disease-specific questionnaires including St George's Respiratory Questionnaire (SGRQ) (24) and the Chronic Respiratory Questionnaire (CRQ) (25). However, these instruments are often too complex or time-consuming for routine use in the majority of clinical settings. The Clinical COPD Questionnaire (CCQ) was developed recently and benefits from comprising only ten items (26, 27): its brevity and simplicity makes it particularly suitable for more routine use in clinical practice. The CCQ was originally created to measure clinical health status in patients, including status of the airways, limitation of physical activity and emotional dysfunction (26). Measurements of health-related quality of life are an attempt to quantify the impact of disease on a patient's well-being (2). Although not entirely synonymous, health status and HRQL share several characteristics. This common ground is demonstrated by the strength of the correlation for CCQ with SGRQ (26), SF-36 (26), and CRQ (28).

The CCQ has been used in observational studies (29) and to evaluate pharmacological treatment (30). A Spanish study investigated different disease-related and socioeconomic factors in COPD patients and reported that lower education, higher grade dyspnoea, airflow limitation, a higher number of exacerbations and hospital admissions were associated with higher CCQ score indicating poorer quality of life (31). Apart from this, little has been published on factors that are associated with CCQ score among COPD patients.

The aim of this study was to examine associations with HRQL estimated by CCQ for co-morbidity and BMI, in a multi-centre population with COPD patients from both primary and secondary health care settings in Sweden.

MATERIALS AND METHODS

Procedure

The sample examined was taken from a larger cohort of patients involved in an investigation of asthma and COPD in a region covered by 7 county councils in central Sweden (32, 33). Each Swedish county council has a central hospital and one or more district hospitals. In this study, each county council was represented by the department of respiratory medicine in their central hospital, the department of internal medicine from 1 randomly selected district hospital and 8 randomly selected primary health care centres (PHCC-s). In total the sample comprised 14 hospitals and 56 PHCC-s.

A list of all patients aged 18–74 years with a COPD diagnosis (ICD-10 code J44) in the medical records during the period of 2000–2003 was compiled for every participating primary and secondary care centre. A random selection recruited 35 COPD patients per hospital list and 22 COPD patients per PHCC list. In 12 PHCC-s, where the original list contained fewer than 22 patients, everyone on the list was included. After excluding a small number of patients with an incorrect diagnosis, the sample comprised a total of 1548 patients, including 1084 in primary care and 464 in hospital clinics.

Data collection

Data were collected in 2005 by self-completion questionnaire and by record review for the period 2000–2003. The response rate for the questionnaire was 75%. Of those who responded, 98% (1,111 patients), consented to a review of their medical records. Complete information for all variables used in the analysis was available from 919 records including 639 from PHCC-s and 280 from hospital clinics. Information on lung function from spirometry data was available for a subset of 475 patients (Table 1). Two research nurses entered the data from the patient questionnaire and medical records.

Table 1.  Patient characteristics in primary and secondary care

Characteristics	Primary care (n = 639)	Secondary care (n = 280)	P-value
Sex
Male	267 (41.8%)	125 (44.6%)
Female	372 (58.2%)	155 (55.4%)	0.42
Age
≤50	46 (7.2%)	16 (5.7%)	0.41
51–60	134 (21.0%)	51 (18.2%)	0.34
61–70	315 (49.3%)	150 (53.6%)	0.23
>70	144 (22.5%)	63 (22.5%)	0.99
Educational level
Lower	428 (67.0%)	187 (66.8%)
Higher	211 (33.0%)	93 (33.2%)	0.95
Smoking
Current	215 (33.6%)	47 (16.8%)	<0.0001
Sometimes	39 (6.1%)	19 (6.8%)	0.70
Ex	345 (54.0%)	198 (70.7%)	<0.0001
Never	40 (6.3%)	16 (5.7%)	0.75
Comorbidity
Heart disease	123 (19.2%)	85 (30.4%)	<0.0001
Diabetes mellitus	66 (10.3%)	25 (8.9%)	0.51
Depression	82 (12.8%)	27 (9.6%)	0.17
Body mass index
<20	57 (8.9%)	44 (15.7%)	0.002
≥20, <25	230 (36.0%)	92 (32.9%)	0.36
≥25, <30	238 (37.2%)	86 (30.7%)	0.06
≥30	114 (17.8%)	58 (20.7%)	0.30
CCQ^*
CCQ total	2.01 ± 1,29	2.71 ± 1.35	<0.0001
CCQ symptoms	2.21 ± 1.19	2.72 ± 1.29	<0.0001
CCQ mental state	2.01 ± 1.80	2.83 ± 1.89	<0.0001
CCQ functional state	1.82 ± 1.48	2.65 ± 1.63	<0.0001
Lung function^**
Stage 1	69 (25.3%)	25 (12.4%)	<0.0001
Stage 2	124 (45.4%)	82 (40.6%)	0.29
Stage 3	68 (24.9%)	62 (30.7%)	0.16
Stage 4	12 (4.4%)	33 (16.3%)	<0.0001

Data presented as n (%) or mean ± SD.

^*CCQ = Clinical COPD questionnaire

^**For subgroup with available spirometry data (n = 475).

Patient characteristics

Information on age, sex, smoking history and level of education were gathered from the patient questionnaires. Smoking history was categorized into current smoking, ex-smoking, occasional smoking and never smoked. Compulsory education at school in Sweden is for nine years. In our study, the dichotomous educational variable identified the most educated group as those who had continued in full-time education for at least 2 years beyond the compulsory school period.

HRQL

The patient questionnaire included a Swedish version of the CCQ, validated for studies on level of populations (34). The CCQ consists of ten questions distributed in three domains: symptoms, mental state and functional state. Observed symptoms are dyspnoea, cough and phlegm; mental state includes questions about feeling depressed and concerns about breathing; and functional state describes limitations in different activities of daily life due to the lung disease. The questions apply to the last week and are assessed by a 7-point scale from 0 to 6. The main outcome measure of HRQL is the mean CCQ value (26), a higher value indicating lower quality of life. The minimum difference in mean CCQ considered to be of clinical importance is 0.4 (27).

Co-morbidity

Information about co-morbid diagnoses was gathered from the patients' records. Heart disease was defined as presence of ischemic heart disease or heart failure, and diabetes as type 1 or type 2 diabetes mellitus recorded at anytime during the period of 2000–2003. Depression was defined as having a diagnosis of depression in combination with antidepressant drug treatment. Asthma was identified by the ICD-10 code J45. At study entrance 194 of the 919 patients had concomitant diagnoses of asthma and COPD.

BMI

Information on self-reported height and weight was provided by the patient questionnaire. Obesity was defined as body mass index (BMI) ≥30, overweight as BMI <30 and ≥25, and underweight as BMI below 20.

Spirometry

Lung function values were collected from the patients' records (33). In patients where spirometry data were available, their disease was graded based on percent of predicted value using FEV₁ (FEV₁%pred) (35).

Treatment

Data from the patient questionnaire were used, and the following treatment modalities were identified: no treatment; short-acting anticholinergics; long-acting beta agonists or anticholinergics; inhalation steroids; and combined long-acting beta agonists and inhalation steroids separately or in a fixed preparation.

Statistical analysis

The main analyses were performed using PASW version 18.0 (SPSS Inc, Chicago). Only patients with complete data for all variables were included (n = 919). Total mean CCQ score and the mean score of each separate domain were calculated and used as dependent variables in multiple linear regression (26). Heart disease, diabetes, depression, obesity, overweight and underweight were modeled as independent variables. In an intermediate step, the models were adjusted for sex, age, smoking habits, level of education and level of care. A final model simultaneously included all of the independent variables and the potential confounding factors. Sex, age, smoking habits, level of education, level of care, BMI, heart disease, diabetes and depression were modeled as series of binary dummy variables. Age was modeled by the age groups ≤50 years, 51–60 years, 61–70 years and >70 years.

Stratification and interaction analyses were used to investigate sex differences. The interaction analyses used interaction terms for sex with each relevant variable with adjustment for the main effects and the potential confounding factors. Potential effect modification by level of care was similarly investigated using stratification and interaction analyses.

Sub-analysis in the group with spirometry data (n = 475) allowed further adjustment for lung function. Lung function was modeled both as a continuous variable for FEV₁%pred and as a series of binary dummy variables based on the established GOLD stages of COPD (35).

To examine the influence of a concomitant asthma diagnosis, additional analyses were performed both with adjustment for a diagnosis of asthma and with exclusion of the patients with an asthma diagnosis. To assess the possible influence of pharmacological treatment on our results, the main analyses were repeated with further adjustment for treatment modality.

As the mean CCQ distribution was somewhat skewed, the main analyses were repeated using a log-transformed measure of mean CCQ to ensure the results were not influenced by the skewed distribution. The Inskew0 command provided by STATA software (StataCorp LP, Texas) was used to generate a normal distribution using the equation ln(meanCCQ) + 3.358. The main linear regression analyses were repeated using the log-transformed dependent variable.

Ethics

The study was approved by the Regional Ethical Review Board of Uppsala University (Dnr 2010/090). Written consent was obtained for all participating patients.

RESULTS

Patient characteristics

The characteristics of the sample were similar in primary and secondary care patients. However, a statistically significant larger proportion of current smokers and patients with COPD stage 1 were found in primary care, while a greater proportion of ex-smokers, patients with COPD stage 4, heart disease and underweight were found in secondary care. The mean total and domain values of CCQ were generally higher for secondary care patients (Table 1). Women were more common in the 2 oldest age groups, among current smokers, and in patients with depression and underweight. There were more men in the youngest age group, among ex-smokers, and in patients with heart disease and overweight (data not shown).

CCQ

In the unadjusted analyses heart disease, depression, obesity and underweight were statistically significantly associated with higher mean CCQ score indicating a lower quality of life. Higher magnitude associations, exceeding the reported threshold for clinical importance (27), were found for depression and underweight. The associations with CCQ for heart disease, depression and underweight remained statistically significant after simultaneously adjusting for each other as well as for the potential confounding factors in the final model. The statistically significant regression coefficients represent difference in mean values (Table 2, Figures 1–2).

Table 2.  Association between patient characteristics and CCQ

Characteristics	Mean CCQ ± SD	Adjusted regression coefficient^* (95% CI)	P-value
Level of care
Primary care	2.01±1.29	Ref^**
Secondary care	2.71±1.35	0.60 (0.41 to 0.78)	<0.0001
Sex
Male	2.29±1.33	Ref
Female	2.18±1.36	−0.12 (−0.30 to 0.06)	0.18
Age
31–50	1.98±1.36	−0.03 (−0.39 to 0.32)	0.86
51–60	2.06±1.33	−0.15 (−0.37 to 0.07)
61–70	2.25±1.38	Ref	0.19
71–80	2.38±1.26	0.12 (−0.10 to 0.34)	0.28
Educational level
Lower	2.33±1.36	Ref
Higher	2.03±1.29	−0.24 (−0.42 to 0.05)	0.01
Smoking
Current	2.06±1.30	Ref
Sometimes	2.32±1.56	0.06 (−0.31 to 0.43)	0.75
Ex	2.33±1.35	0.13 (−0.07 to 0.33)	0.20
Never	1.91±1.28	−0.25 (−0.63 to 0.14)	0.21
Heart disease
Absent	2.13±1.36	Ref
Present	2.55±1.26	0.26 (0.06 to 0.47)	0.01
Diabetes
Absent	2.23±1.35	Ref
Present	2.19±1.34	−0.04 (−0.32 to 0.24)	0.79
Depression
Absent	2.19±1.35	Ref
Present	2.53±1.26	0.50 (0.23 to 0.76)	<0.0001
Body mass index
<20	2.79±1.48	0.58 (0.29 to 0.87)	<0.0001
≥20, <25	2.14±1.31	Ref
≥25, <30	2.03±1.29	−0.13 (−0.33 to 0.07)	0.20
≥30	2.43±1.32	0.21 (−0.04 to 0.45)	0.09

^*Regression coefficients representing mean differences in CCQ, adjusted for all variables shown.

^**Reference.

Figure 1.  COPD co-morbidity and CCQ score. A higher CCQ score indicates lower quality of life.

Figure 2.  Body Mass Index and CCQ score. A higher CCQ score indicates lower quality of life.

In addition to the main results, higher level of education and better lung function were found to be statistically significant and independently associated with a higher quality of life in both the unadjusted and the adjusted regression analyses (Table 2, Figure 3).

Figure 3.  Lung function and CCQ score. A higher CCQ score indicates lower quality of life.

Neither adding adjustment for asthma nor excluding patients with a diagnosis of asthma changed the results notably (data not shown).

Domains

In the adjusted analyses, statistically significant associations were found for both depression and underweight with higher scores in all 3 CCQ domains. In addition, heart disease was associated with a statistically significantly higher score for the functional state domain (Table 3). The highest magnitude associations and therefore most clinically important were found for underweight in all domains and for depression in the mental and functional domain. In addition, better lung function and higher level of education were statistically significant and independently associated with better outcomes in all separate CCQ domains.

Table 3.  Association between patient characteristics and separate CCQ domains

Characteristics	CCQ; symptoms Adjusted regression coefficient^* (95%CI)	CCQ; mental state Adjusted regression coefficient (95%CI)	CCQ functional state Adjusted regression coefficient (95%CI)
Level of care
Primary care	Ref	Ref	Ref
Secondary care	0.48 (0.30 to 0.60)	0.74 (0.48 to 1.00)	0.65 (0.44 to 0.86)
Sex
Male	Ref	Ref	Ref
Female	−0.30 (−0.48 to −0.13)	0.15 (−0.10 to 0.40)	−0.08 (−0.28 to 0.13)
Level of education
Lower	Ref	Ref	Ref
Higher	−0.16 (−0.34 to 0.02)	−0.39 (−0.65 to −0.13)	−0.23 (−0.44 to −0.02)
Heart disease
Absent	Ref	Ref	Ref
Present	0.19 (−0.02 to 0.39)	0.15 (−0.15 to 0.44)	0.40 (0.16 to 0.63)
Diabetes
Absent	Ref	Ref	Ref
Present	−0.13 (−0.40 to 0.15)	−0.02 (−0.41 to 0.38)	0.04 (−0.29 to 0.36)
Depression
Absent	Ref	Ref	Ref
Present	0.35 (0.10 to 0.61)	0.69 (0.32 to 1.05)	0.54 (0.24 to 0.84)
BMI
BMI <20	0.54 (0.26 to 0.82)	0.48 (0.07 to 0.88)	0.68 (0.34 to 1.01)
20< BMI <25	Ref	Ref	Ref
25< BMI <30	−0.02 (−0.22 to 0.17)	−0.28 (−0.55 to 0.01)	−0.17 (−0.40 to 0.06)
BMI >30	0.14 (−0.10 to 0.37)	0.29 (−0.05 to 0.64)	0.24 (−0.04 to 0.52)

^*Regression coefficients representing mean differences in CCQ, adjusted for all variables from Table 2.

Sex

When stratified by sex, the adjusted associations for heart disease, depression, and underweight with higher mean CCQ score remained statistically significant in women (n = 527), with the highest magnitude and thus clinical relevance for depression and underweight (Table 4). In men (n = 392) only underweight was associated with lower HRQL (Table 4). Effect modification by sex was found for the association of overweight with HRQL, (p = 0.015) with a significant association between overweight and lower mean CCQ in men but not in women (Table 4).

Table 4.  Association between patient characteristics and CCQ, stratified for sex

Variable	Mean CCQ ± SD	Men Adjusted regression coefficient^* (95%CI)	Mean CCQ ± SD	Women Adjusted regression Coefficient (95%CI)
Level of care
Primary care	2.07 ± 1.34	Ref	1.97 ± 1.26	Ref
Secondary care	2.75 ± 1.21	0.52 (0.24 to 0.80)	2.68 ± 1.46	0.65 (0.40 to 0.90)
Level of education
Lower	2.41 ± 1.34	Ref	2.27 ± 1.38	Ref
Higher	2.07 ± 1.30	−0.22 (−0.49 to 0.06)	1.99 ± 1.29	−0.24 (−0.49 to 0.00)
Heart disease
Absent	2.17 ± 1.36	Ref	2.10 ± 1.35	Ref
Present	2.54 ± 1.23	0.24 (−0.05 to 0.53)	2.57 ± 1.32	0.34 (0.04 to 0.65)
Diabetes
Absent	2.29 ± 1.34	Ref	2.19 ± 1.35	Ref
Present	2.27 ± 1.28	0.10 (−0.32 to 0.51)	2.11 ± 1.40	−0.12 (−0.51 to 0.27)
Depression
Absent	2.26 ± 1.33	Ref	2.12 ± 1.37	Ref
Present	2.64 ± 1.30	0.34 (−0.18 to 0.86)	2.49 ± 1.25	0.52 (0.21 to 0.83)
BMI
BMI <20	3.07 ± 1.42	0.61 (0.10 to 1.12)	2.67 ± 1.50	0.64 (0.28 to 0.10)
20≤ BMI <25	2.40 ± 1.39	Ref	1.98 ± 1.24	Ref
25≤ BMI <30	1.98 ± 1.23	−0.41 (−0.71 to −0.11)	2.08 ± 1,36	0.12 (−0.15 to 0.39)
BMI ≥30	2.50 ± 1.25	0.01 (−0.38 to 0.39)	2.38 ± 1.38	0.30 (−0.01 to 0.62)

^*Regression coefficients representing mean differences in CCQ in, adjusted for all variables from Table 2.

Stratification for sex in the analysis of separate CCQ domains revealed that, similar to the total population, heart disease was statistically significant and independently associated with higher CCQ score in functional state in both sexes (data not shown). The associations with depression and underweight for all of the separate CCQ domains remained in women, but were statistically significant only for functional state and symptoms with underweight in men. Statistically significant and independent associations between overweight and lower mean CCQ in all separate domains were found in men but not in women. Educational level was statistically significant and independently associated with lower CCQ score in functional and mental state in women but not in men.

Level of care

Stratification for level of care showed a statistically significant adjusted regression coefficient (95%CI) for the association with CCQ for depression in primary care of 0.43 (0.12 to 0.74) and 0.61 (0.10 to 1.12) for secondary care; and for underweight in primary care of 0.64 (0.27 to 1.02) and 0.47 (0.06 to 0.93) in secondary care. The interaction analysis of level of care only found a significant modifying effect on the association with CCQ for obesity (p = 0.001), with a statistically significant adjusted regression coefficient (95%CI) for the association with CCQ for obesity of 0.59 (0.16 to 1.02) in secondary care but not in primary care (0.01, −0.28 to 0.31).

In the unadjusted analyses of the subgroup of 475 patients with spirometry data, the same independent and statistically significant associations between heart disease, depression, obesity and underweight with higher mean CCQ and between higher level of education and lower mean CCQ observed in the unadjusted analyses of the entire sample were found.

In the adjusted analyses the same pattern was found, although the association between heart disease and mean CCQ lost statistical significance. Additional adjustment for lung function in the subgroup revealed statistically significant and independent associations for heart disease, (regression coefficient (95%CI) 0.28, 0.02 to 0.54), depression (0.66, 0.35 to 0.98), obesity (0.60, 0.30 to 0.91), underweight (0.40, 0.06 to 0.75) and higher level of education (−0.28, −0.49 to −0.06). Changing the lung function variable for FEV₁%pred from categorical to continuous did not change the results notably.

A distinct independent and statistically significant association between worse lung function and higher mean CCQ was found both for total mean CCQ (Figure 3) and for separate domains (data not shown).

Treatment

The main results for associations between heart disease, depression and underweight with higher mean CCQ remained similar after further adjustment for treatment categorized into groups (data not shown).

Analysis of logtransformed outcome variables

Log transformation for mean CCQ scores eliminated the skewness of the dependent variables. None of the main findings differed when the log-transformed measures were used, indicating our results are not artifacts resulting from the skewed CCQ distribution.

DISCUSSION

The primary findings of this multi-centre study of a clinical population including COPD patients from both primary and secondary care are that heart disease, depression, and underweight are independently associated with lower HRQL estimated using CCQ. In addition, associations of level of education and lung function with CCQ were demonstrated. The CCQ is convenient to use in clinic due to its simplicity and brevity. The associations with HRQL reported here were not identified previously using CCQ and the results are potentially clinically relevant. The results of the study replicate previously shown associations with other HRQL instruments, but extend them to a multi-centre design in different care settings with simultaneous consideration of multiple patient characteristics.

The association between heart disease and mean CCQ was of both statistical significance and exceeded the reported clinically important threshold of 0.4 (27) in the unadjusted analyses. In the final adjusted model the magnitude of heart disease was attenuated although still being statistically significant, while the magnitude of the association between depression and mean CCQ increased. Thus a component of the association between heart disease and mean CCQ could probably be explained by a concurrent depressive state in COPD patients with heart disease. The association between heart disease and mean CCQ in COPD patients is consistent with the results in a HRQL study with SF-12 (36). In contrast to a previous study of COPD patients where diabetes was associated with worse outcomes in SF-12 (37), diabetes was not associated with CCQ in our study.

An association of magnitude exceeding that defined for minimal clinical importance was found for depression with both mean CCQ and two of the CCQ domains. This result is also consistent with a corresponding study using SF-12 (38). In the domain analyses, the association of mental state with depression was expected, as they could be seen as different measures of the same or a related characteristic. However, the other domains were significantly associated with depression, emphasizing the importance of mental health for the subjective burden of COPD.

Quality of life was lower in both underweight and obese patients compared with those of normal weight, creating a U-shaped association similar to the known connection between BMI and mortality (17, 19). The association between underweight and CCQ showed the highest of all magnitudes in the main analysis, and low BMI was the only variable with clinically significant regression coefficients for all separate domain CCQ scores. The association between underweight and HRQL is consistent with the results in a study estimating quality of life using SGRQ and SF-36 (39).

The associations for heart disease, depression and underweight with CCQ were similar to those found with other HRQL instruments. It is important that the same associations were identified using a shorter and clinically more convenient instrument. We did not identify an association between diabetes and HRQL, perhaps because the CCQ is designed to identify HRQL associated with lung disease, rather than other types of mortality. Previously an association for diabetes and quality of life measured with a generic instrument has been demonstrated.

Level of education could be seen as a marker of several cultural and material factors. An association between higher level of education and increased quality of life has been shown in general populations (40, 41), and to our knowledge only in one study of COPD patients (31). However, the impact of socioeconomic factors on HRQL in COPD patients has been studied without significant results (8). The association between worse lung function and lower quality of life estimated with CCQ in our study is consistent with both previous studies of CCQ (31) and other HRQL instruments (9).

In the entire group, the unadjusted analyses showed statistically significant associations with mean CCQ score for heart disease, depression, underweight and obesity. After adjustment, the pattern of associations was the same although the association with obesity was not statistically significant. In the subgroup with information on FEV₁, the unadjusted analyses also showed statistically significant associations with mean CCQ score for heart disease, depression, underweight and obesity with mean CCQ, but in the adjusted analyses heart disease lost statistical significance. The changes may be due to the reduced number in the subset or due to changes in the composition of patient characteristics.

With further adjustment for lung function, the magnitude of the association of underweight with mean CCQ was diminished while the magnitude of the association between obesity and depression increased. Thus, poorer lung function may possibly explain part of the association between underweight and HRQL, while obesity may have been confounded by lung function and still of clinical importance. However the adjustment for lung function in the subgroup with FEV₁ did not cause any loss of significance in the subgroup. Thus, we concluded that the associations with HRQL for co-morbidity and BMI in our study could not be explained by poorer lung function.

The result of stratifying for sex, points at the possibility that co-morbidity and BMI influences the burden of COPD more for women. As seen for the whole population, underweight and depression had the highest magnitudes of association and seem to be of most clinical significance for HRQL in women with COPD. The association with depression was not statistically significant among men, which could partly be due to smaller number of male patients. However, because of lower magnitudes of association it might also mean that depression is indeed a more notable problem in women with COPD.

The finding that HRQL is worse in women is consistent with previous studies of using other HRQL instruments (3–5). Do factors determining quality of life in men and women differ; with mental state and symptoms playing a more dominant role in HRQL among women with COPD (42)? The sex difference for the association between overweight and HRQL was striking. This is however consistent with a recent study of quality of life in a general population where the maximum HRQL score in men was among the overweight, but was in the normal BMI category among women: this suggests that overweight could be viewed as the most ‘normal’ category among men (17).

Stratification and interaction analyses for level of care only revealed a difference in how obesity is associated with CCQ, which means that the other main results are relevant to patients whether they are treated in primary or secondary care. Overlap syndrome with COPD and obstructive sleep apnea syndrome or obesity hypoventilation syndrome (43) may contribute to worse quality of life in the hospital patients.

A major strength of this study is that the population is taken from multiple centers representing both primary and secondary care. As patients tend to have more than one co-morbid condition, multivariate analysis sought to identify which of these was most notably associated with quality of life. Finally, the data recorded prospectively from the medical records should be very reliable and not subject to recall bias.

A limitation is that spirometry data are only available for some of the patients. This is likely to be because many COPD diagnoses are based on clinical findings rather than spirometry (33). This could be due to an unwillingness to investigate what is considered a very likely diagnosis or, especially in primary care, result from a lack of confidence in use and interpretation (44). The former explanation may result in selection bias, which could potentially influence the results. However, the main findings did not change when comparing results of the subgroup with spirometry based diagnosis in unadjusted analyses and after further adjustment for lung function.

A common co-morbidity in COPD is asthma, sometimes because establishing the correct diagnosis is difficult and sometimes because of coexisting diseases. A subset of our study population also had asthma, reflecting the clinical reality of a COPD population. However, taking simultaneous diagnosis of asthma into account did not influence the results.

By examining a more complete population of patients with a diagnosis of COPD, rather than selecting them on strict diagnostic criteria, we aimed to tackle the urgent need in respiratory research to conduct studies of representative clinical patient populations; “real-life studies” (45, 46). Performing the analysis using a log-transformed CCQ score and adding adjustment for treatment groups showed that neither a skewed distribution for CCQ nor the pharmacological COPD treatment could explain the results of our study.

The level of attrition is acceptable, and we speculate that if selection occurred, then patients with more severe disease may have been more likely to have been excluded. The definitions of co-morbid conditions were mainly based on finding the diagnoses in the medical records. The definition of depression required pharmacological treatment and may have resulted in a selected group. Our model does not take into account the duration of the different diagnoses before measuring the outcome, but since the diseases investigated as independent variables are not typically progressive we do not believe this has influenced the results notably.

Weight and height were self-reported, since too few records contained these data. Reported weight can be biased when not objectively measured, and may be differential by sex. Most studies of general populations show that people tend to underestimate weight and overestimate height (47), possibly producing erroneously low estimates of BMI. However, the association between underweight and quality of life in our study is consistent with the results of another study of underweight and HRQL using SGRQ and SF-36 where weight and height were measured objectively (39).

We conclude that co-morbidity and BMI are associated with lower HRQL estimated using CCQ in COPD patients. Amongst women, depression may be of particular importance in determining low quality of life. Our study indicates that the CCQ is a clinical useful instrument for evaluating HRQL in COPD, as it was possible to identify the same determinants for quality of life with this short questionnaire as with more complicated and time-consuming instruments.

ACKNOWLEDGMENTS

Thanks to Ulrike Spetz-Nyström and Eva Manell for reviewing the patient records, and to all participating centres. The study was supported by grants from the county councils of the Uppsala-Örebro Health Care region, the Swedish Heart and Lung Association, the Swedish Asthma and Allergy Association and the Bror Hjerpstedts Foundation, Uppsala.

Declaration of interest

The authors have no financial or other conflicts of interest related to the material of the present study. The authors alone are responsible for the content and writing of the paper.