Background
Chronic Obstructive Pulmonary Disease (COPD) is a common progressive disease that is characterized by persistent airflow limitation and related to an enhanced chronic inflammatory response in the airways and the lung to noxious particles or gases [
1]. Although preventable and treatable, COPD represents an important public health challenge which is projected to increase in coming years because of continued exposure to risk factors such as tobacco smoking, indoor and outdoor air pollution and the aging population. The World Health Organization predicted that COPD will become the fourth leading cause of death worldwide by 2030 [
2].
Patients with COPD usually experience a decrease in their health-related quality of life (HRQL): COPD symptoms such as cough, dyspnea, and sputum production, as well as acute exacerbations of the disease, and comorbidities which are ubiquitous in COPD patients contribute to the overall severity of the disease [
3‐
6].
HRQL as an important patient-reported outcome measure in COPD has gained attention in the last years both as an individual descriptive measure as well as an endpoint in clinical studies [
7]. A large variety of generic and disease-specific HRQL instruments with proven validity and reliability is available to assess HRQL of COPD patients [
8]. While disease-specific instruments focus on symptoms and functional impairments associated with COPD, generic instruments are more widely applicable irrespective of the underlying disease [
7]. Therefore, they can be used to compare HRQL of diseased and healthy subjects or to compare the burden of different diseases. Some generic questionnaires are used as utility instruments e.g. for the calculation of quality-adjusted life years (QALYs) in health economic evaluations. Nevertheless, they may be less sensitive to special problems of a certain disease than disease-specific instruments.
This paper describes generic as well as disease-specific HRQL in a large cohort of patients with COPD comprising all disease grades and compares assessment properties and correlations of different HRQL instruments. HRQL instruments are analyzed regarding their ability to differentiate between COPD grades. Special attention is given to the impact of comorbid conditions in COPD patients on generic and disease-specific HRQL assuming that generic instruments are more suitable to reflect the impact of comorbidities on health status than disease-specific instruments. Furthermore, we analyze whether COPD and comorbidities have additive effects on HRQL measures and whether this is different for generic and disease-specific instruments.
Methods
Patients
This cross-sectional analysis is based on data from the baseline visit of the German national COPD cohort COSYCONET (
German COPD and Systemic Consequences - Comorbidities Network). In brief, COSYCONET recruited 2,741 patients ≥ 40 years with physician-diagnosed COPD by outpatient and inpatient healthcare providers, patient groups, and media campaigns and examined them in 31 study centers all over Germany between September 2010 and December 2013. Exclusion criteria were previous lung transplantation or lung volume reduction surgery and lung malignancies. All participants were clinically stable defined as no moderate or severe exacerbations for at least 4 weeks at the time of enrolment. Details on the cohort have been published elsewhere [
9,
10].
Lung function test and definition of COPD
Standardized spirometry was performed in the COSYCONET cohort after bronchodilation with 400 μg salbutamol and 80 μg ipratropium bromide. COPD was defined as FEV
1/FVC < 0.7 according to the GOLD criteria [
1]. Patients were classified as grade 1 with FEV
1 % pred. ≥ 80, grade 2 with 50 ≤ FEV
1 % pred. < 80, grade 3 with 30 ≤ FEV
1 % pred. < 50 and grade 4 with FEV
1 % pred. < 30, with predicted values based on reference equations from the Global Lung Initiative (GLI) [
11]. 430 participants from the cohort reporting a physician-diagnosed COPD but with FEV
1/FVC ≥ 0.7 were excluded from this analysis as well as 20 participants with incomplete or missing lung function data.
The BODE index [
12] as established multidimensional grading system in COPD was calculated in addition based on body mass index (BMI, defined as weight in kilograms/height in squared meters), FEV
1 % pred., the 5-point modified Medical Research Council (mMRC) dyspnea scale and the six-minute walking test which was performed according to the criteria of the American Thoracic Society [
13]. The BODE index ranges from 0 to 10 points with higher values indicating worse health.
HRQL assessment
HRQL was assessed by self-administered questionnaires: The St. George’s Respiratory Questionnaire in its COPD-specific version (SGRQ-C) [
14] and the COPD Assessment Test (CAT) were used as disease-specific HRQL instruments, and the EuroQol 5 dimensions (EQ-5D-3 L) as a generic HRQL instrument.
The SGRQ-C with 40 items provides three component scores for symptoms, activity and impact, and a total score. Each score ranges from 0 (no impairment) to 100 (worst possible). A difference of 4 unit points is considered the minimum clinically important difference (MCID) [
15].
The CAT is an 8-item, short and validated tool for the assessment and monitoring of COPD [
16]. Symptoms are assessed on a scale from 0 to 5. The total score ranges from 0 to 40 with higher scores representing worse health status. A MCID of 2 points has been proposed [
17].
The EQ-5D-3 L questionnaire is a preference-based HRQL instrument with two parts. Part 1, the descriptive section, covers five dimensions of health: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression with three levels per item (no problems, some problems, and extreme problems). An index-based utility score (EQ-5D utility) ranging from 0 to 1 can be calculated via a scoring algorithm which is based on valuations of representative general population samples. We used the German time-trade-off tariff by Greiner et al. for scoring [
18]. A MCID for the utility score in COPD patients has not yet been established. Based on other diseases, it may range between 0.08-0.10 [
19,
20]. Part 2, the valuation section, comprises a Visual Analog Scale (VAS) for valuing health states on a rating scale from 0 (worst imaginable health state) to 100 (best imaginable health state). A MCID of 8 points has been suggested in COPD patients with moderate to severe disease [
21].
All three HRQL instruments are designed to assess current health status without specifying a recall period.
Comorbidities and covariables
Self-reported information on physician-diagnosed comorbid conditions was systematically obtained in semi-structured interviews. We assessed 33 of these comorbid conditions. The number of comorbid conditions was summarized as a simple count. This approach has been shown to be a good proxy for the burden of comorbidities [
22]. For defining groups with a low or a high number of comorbidities, the median number of comorbidities (≤3 vs. >3) was used as cut-off. A list of all comorbidities considered can be found in the (Additional file
1: Figure S1).
Information on age, sex, school education (basic (≤9 years), secondary (10–11 years), higher (≥12 years)), and smoking status (current, former, and never smoker) was collected by standardized questionnaires. BMI was measured at the study centers and classified as normal weight (18.5 ≤ BMI < 25), overweight (25 ≤ BMI < 30), obese (BMI ≥ 30), and underweight (BMI < 18.5).
Statistical analysis
Characteristics of COPD patients in different grades were compared using analysis of variance (ANOVA) for continuous variables and Chi2-tests for categorical variables. For each HRQL instrument, the proportion of participants with the worst or best possible health state was determined to assess the presence of floor or ceiling effects. Bivariate correlations between the HRQL instruments and with clinical COPD measures were quantified by Spearman’s rank coefficient.
To assess the association of COPD grades 1–4 with HRQL and the discriminative ability of HRQL instruments, linear regression models were performed adjusting for age, sex, education, smoking status, BMI, and low vs. high number of comorbidities. The number of comorbid conditions as a continuous variable was considered in a sensitivity analysis instead of the dichotomized variable. Adjusted mean HRQL scores resulting from the regression models were reported for COPD grade 1–4 as well as for the group with a low or high number of comorbidities. Standardized mean differences were calculated as the mean adjusted difference between two COPD grades divided by their pooled (unadjusted) standard deviation (SD) [
23] in order to assess the magnitude of the difference and to judge the ability of HRQL instruments for discrimination. In additional models, non-additive effects of COPD and comorbidity were checked by interaction terms between COPD grades and low or high number of comorbidities. Pseudo-R
2 according to Cox & Snell [
24] were calculated to compare the additional variance explained by COPD grade and/or low vs. high number of comorbidities in models for all HRQL instruments.
Further models were calculated controlling for all single comorbidities instead of a comorbidity count to identify the most important comorbid conditions for each HRQL instrument, again adjusted for age, sex, education, smoking status, and BMI.
Statistical analyses were performed using SAS software (SAS Institute Inc., Cary, NC, USA, version 9.3), and p-values of 0.05 or less were considered to be statistically significant.
Discussion
This study performed a comparison of the most frequently used generic and disease-specific HRQL instruments in a large cohort of COPD patients in all levels of airflow limitation with particular attention on the association with a comprehensive list of comorbid conditions.
Our study confirms that mean scores of SGRQ, CAT, EQ-5D utility, and EQ-5D VAS worsen with increasing COPD grade, despite high variability within each grade. EQ-5D utility and VAS as generic HRQL instruments and SGRQ and CAT as disease-specific instruments were found able to discriminate between COPD grades, with some limitations for the EQ-5D utility in mild COPD. Utility results which are based on population preferences and therefore introduce external HRQL valuation put the highest weight on comorbid conditions. However, disease-specific HRQL instruments also reflected the burden of comorbid conditions. Non-additive effects of COPD and comorbidity were not observed. When focusing on single comorbid conditions, psychiatric disorders and peripheral artery disease ranked among the top 5 comorbid conditions with negative associations with HRQL for all instruments. Sleep apnoea, brain deficiency, asthma, and heart disease were also important contributors to low HRQL.
Regarding mean HRQL scores per COPD grade or correlation of disease-specific and generic HRQL measures, our results are in line with previous studies. Despite the differences in underlying populations, the mean CAT scores of our study fit very well in the range of means reported in a systematic review on studies using the CAT by Gupta et al. [
25]. In a small sample of US veterans in COPD grades 1–4 (
n = 120), Pickard et al. reported correlations between EQ-5D utility or VAS and SGRQ total score similar to our findings [
26]. The EQ-5D VAS means, which were considerably higher than in our study, differed between COPD grades 1–4, while EQ-5D utilities did not. In line with our results, best discriminative properties between COPD grades were reported for the SGRQ in this study.
Our results showed clear HRQL differences between COPD grade 1 and 2 (except for EQ-5D utility), while some previous studies did not observe HRQL differences between milder COPD grades. E.g., Jones et al. did not find differences in SGRQ total score and the generic SF-12 questionnaire between primary care COPD patients in grade 1 and 2. This might be due to the fact that reported SGRQ means for grade 1 were more than 10 points higher than in our sample, while mean values for grades 2, 3 and 4 as reported in Jones et al. were comparable to our results [
27]. In the same population, Jones et al. did not find significant differences in CAT scores between COPD grade 1 and 2 [
28]. A possible explanation for this lack of differences between these grade 1 and 2 patients may be that symptoms and HRQL impairment are important drivers for the patients to see their primary care physician.
For health-economic evaluations in COPD, it is important to know the discriminative abilities of the EQ-5D questionnaire because it is frequently applied for determining utilities needed for the calculations of quality-adjusted life years (QALYs) [
29]. Based on data from UPLIFT, a large COPD trial covering 13 countries, Rutten-van Mölken et al. investigated discriminative abilities of the EQ-5D questionnaire in COPD patients [
29]. This study found better discriminative properties of the EQ-5D utility in higher COPD grades, but did not include COPD grade 1. When comparing EQ-5D utilities, it has to be kept in mind that utilities depend on the underlying country-specific tariff and methods (time trade off vs. standard gamble) used for their calculation. Furthermore, ceiling effects may limit the sensitivity of the utility.
There seems to be a consensus that COPD patients with a higher number of comorbid conditions have a worse HRQL [
27,
28,
30‐
34]. Several previous studies supplied evidence that comorbidities in patients with COPD are associated with worse COPD-specific HRQL scores [
6,
27,
28,
31,
35,
36]. In a hospital-based COPD cohort, Koskela et al. also found that a generic HRQL instrument better captured the effects of comorbid conditions but that these conditions were also associated with COPD-specific HRQL [
3]. For COPD patients with three or more comorbid conditions, Jones et al. reported significantly worse CAT (+2.5 units) [
28] and SGRQ scores (approx. +8 units) [
27] than for patients without or with 1–2 comorbidities. Despite referring to different comorbid conditions, these extra units correspond well to the estimates of high comorbidity found in our study.
When focusing on single comorbid conditions, especially mental health problems are often reported contributors to decreased HRQL in COPD [
3,
5,
6,
31,
34,
37]. This was confirmed by our analysis. In general, a direct comparison of the effects of single comorbidities on HRQL in COPD patients is often hampered by different conditions considered and diverging definitions. Depression, anxiety, and PAD which we identified as the most important comorbidities in our study, were also included in the recently published COMCOLD index that identified five comorbidities with the highest effect on the VAS [
38]. Further comparison with the two additional comorbidities of this index (symptomatic heart disease and cerebrovascular disease) was not possible due to differences in definitions.
Urff et al. reported that COPD patients with depression or heart failure had considerably lower disease-specific HRQL as measured by the Clinical COPD Questionnaire (CCQ) [
36] and explained this finding by overlapping symptoms of COPD and depression or heart failure. Having the items of EQ-5D, CAT, and SGRQ in mind, another reason could be that there is an overlap of these instruments with regard to psychological and mobility aspects of health. Therefore, it is not surprising that especially psychological disorders and PAD showed the strongest associations with the HRQL scores in our analysis. We did not consider dyspnea and exacerbation frequency in our multivariate analysis. As the CAT and SGRQ include dyspnea and symptoms of exacerbations already in their items, controlling for these factors as explanatory variables would cause problems of circularity. The finding that the multidimensional BODE index showed higher correlation with HRQL scores than COPD grades is in line with this reasoning because BODE includes mMRC as a measure of dyspnea and the 6-minute-walking-test as a measure of physical capacity.
Further explanations for the ability of disease-specific HRQL instruments to reflect the burden of comorbidity need to be addressed in future research.
Several strengths of our study have to be considered. Based on a large patient sample comprising all COPD grades, our analysis had sufficient power to detect differences between grades. A comprehensive list of comorbid conditions was available and possible interactions between COPD and comorbidity were analyzed. Therefore, the results of our observational study might be more generalizable than results from clinical trials where patients with comorbidities are often excluded.
There are, however, also potential limitations: First, since only cross-sectional data were available, we were not able to investigate the responsiveness of HRQL instruments or sensitivity to change over time. Second, data on physician-diagnosed comorbidities were self-reported which might limit their validity, and no information on the severity of comorbid conditions was available. Furthermore, comorbid conditions were summarized and dichotomized without weighting. However, this approach has been shown to perform just as well as weighted scores to quantify the impact of comorbidities on disease-specific HRQL in COPD patients [
22].
Nevertheless, from a research perspective, our direct comparison of HRQL instruments did not show a clear ranking. The decision for a generic or disease-specific instrument must be based on the purpose of HRQL measurement. Cross-sectional comparisons of COPD patients (with or without control subjects) or a longitudinal monitoring of disease progression pose different demands on HRQL instruments. Researchers are responsible to ensure that the chosen HRQL instrument addresses the problems of the research questions. While generic instruments allow comparisons of groups with different conditions, may support the identification of unexpected HRQL issues, and are need for the calculation of QALYs in economic evaluations, disease-specific instruments are more likely to detect small, but relevant changes [
39]. Furthermore, by focusing on relevant dimensions of the disease, disease-specific instruments may have a higher acceptability among patients and thus increased responsiveness.
However, we have shown that the effect of comorbid conditions on HRQL was both apparent in generic and to a smaller degree in COPD-specific HRQL measures. This finding is consistent with studies from other diseases [
40]. To cover all aspects of HRQL in COPD, both generic and disease-specific instruments should be used in the ideal case. With regard to economic evaluations, further research is needed towards the discriminative properties of the new EQ-5D 5-level version in COPD.
From a clinical perspective, our results underline the need for a holistic approach for COPD healthcare with a mandatory screening for comorbidities, assessment of HRQL and in consequence a multidisciplinary treatment as generic and also COPD-specific health status could be improved [
41].
Competing interests
CV reports personal fees from Almirall, AstraZeneca, Boehringer Ingelheim, Chiesi, GlaxoSmithKline, Grifols, Janssen, Mundipharma, Novartis, Takeda, and Cipla outside the submitted work. All other authors have no relevant conflicts of interest.
Authors’ contributions
MW and RH conceptualized the paper. MW, AKa, RH performed the statistical analysis and interpreted the data. MW, RH drafted the manuscript. RJ, HS, SK, JH, AKo, HW, RL, CV were involved in the coordination of the study and all commented on drafts of the paper. All authors critically reviewed each draft of the manuscript and approved the final manuscript.