Independent determinants of disease-related quality of life in COPD – scope for nonpharmacologic interventions?

Purpose Quality-of-life (QoL) scores in chronic obstructive pulmonary disease (COPD) have a weak relationship with physiologic impairment. We investigated factors associated with poor QoL, focusing on psychological measures potentially amenable to intervention. Patients and methods We utilized a pre-existing Birmingham (UK) COPD cohort to assess factors associated with QoL impairment (COPD Assessment Test [CAT] scores). Univariate and multivariate regression models were constructed from three categories of variables: demographic, lung function/COPD-related symptoms, and psychosocial/behavioral factors. Results Analyses were based on self-report questionnaire data from 735 participants. The multivariate model of variables independently associated with CAT included depression, dysfunctional breathing symptoms (Nijmegen score), and illness perception, in addition to COPD symptoms (wheeze, cough), exercise capacity, breathlessness, exacerbations, and deprivation; this model explained 72% of CAT score variation. In a dominance analysis assessing the relative contribution of variables, similar contributions were made by breathlessness (20.2%), illness perception (19.8%), dysfunctional breathing symptoms (17.5%), and depression (12.5%) with other variables contributing <5%. Conclusion Psychological factors significantly contribute to disease-specific QoL impairment in COPD, and potentially explain the mismatch between objective physiologic impairment and patients’ experience of their disease. Interventions targeting psychological factors, illness perception, and dysfunctional breathing should be assessed.


Introduction
Chronic obstructive pulmonary disease (COPD) is a major cause of morbidity and mortality and a global public health problem. 1 As a complex, multifaceted disease, it affects patients in many ways and results in significant quality-of-life (QoL) impairment. 2 However, QoL varies greatly between individuals, and is only weakly associated with physiologic factors such as percentage predicted forced expiratory volume in 1 second (FEV 1 ). 3 Anxiety and depression are common, although frequently unrecognized and untreated, 4 and are associated with poor COPD outcomes. [5][6][7] Previous research suggests that a variety of disease-related and patient-related factors may be associated with QoL impairment in COPD. 3 In addition to biologic factors (such as lung function and inflammation) and demographic/socioeconomic factors, there is evidence that QoL is affected by comorbidities, particularly psychological conditions, [8][9][10][11] and by psychological constructs such as illness perception. 12 The relative importance of these different factors is currently unclear.
Understanding the determinants of QoL impairment in COPD may highlight modifiable factors that could be targeted to minimize disease impact and to help patients to cope better with the consequences of having an incurable long-term condition. In particular, psychological and behavioral interventions may be appropriate for selected patients. Nonpharmacologic interventions are acceptable to many patients, particularly those with significantly impaired QoL and those with "disproportionate" QoL impairment in relation to their lung function impairment. 13 The importance of a "personalized medicine" strategy for managing airways disease has recently been emphasized, aiming to target appropriate treatments, both pharmacologic and nonpharmacologic, on potentially modifiable factors in wellcharacterized individual patients. 14 We hypothesized that a range of psychological, social, perceptual, and behavioral factors, measured using validated patient-reported outcome measures, would be independently associated with QoL in patients with COPD.

Methods study design
We report a cross-sectional analysis of data from the Birmingham COPD cohort study. 15 The current analysis investigated factors independently influencing disease-specific QoL (assessed by the COPD Assessment Test [CAT]) 16 and generic health-related (HR) QoL score using the 5 level EuroQoL questionnaire (EQ-5D 5L), 17 and estimated the magnitude of contribution of different contributory factors, with a particular focus on clarifying the relative contribution of psychosocial and behavioral factors.

subjects and setting
The Birmingham COPD cohort study is described fully elsewhere, 18 but briefly consists of three patient groups recruited from 71 primary care practices in the West Midlands, UK during the period May 2012-June 2014: 1) patients with diagnosed COPD on general practitioner registers; 2) newly identified COPD patients from a linked case-finding study; 15 and 3) participants from the case-finding study with chronic respiratory symptoms but without airflow obstruction. Cohort participants were characterized at baseline with a series of questionnaires and objective measurements (including the Medical Research Council [MRC] breathlessness scale, 19 assessments of lung function, muscle strength, and exercise capacity as sit/stand repetitions with Borg breathlessness scores 20 pre-and post-exercise). Cohort participants were invited to complete review questionnaires at 6-monthly intervals for 3 years including items regarding health, lifestyle, health-related quality of life (HRQoL), exacerbations, health care usage, and medical conditions; health resource use information was also collected. At the time of this analysis, the full cohort consisted of 2,188 patients. Ethical approval was provided by the National Research Ethics Service Committee, West Midlands, Solihull, UK (ref: 11/ WM/0304). All participants provided written informed consent for this study.
For the purposes of the current analysis, an additional questionnaire set was administered at one time point per patient, assessing symptoms of dysfunctional breathing (Nijmegen questionnaire), 21 depression (PHQ-9), 22 anxiety (GAD-7), 23 illness perceptions (Brief Illness Perception Questionnaire, IPQ), 24 and agoraphobic avoidance. 25 Questionnaires were posted to participants with an explanatory letter and a return envelope, with one reminder sent 2 weeks after the initial mailing.
Patients were included in our analysis if they 1) had an existing COPD diagnosis or were identified from the casefinding trial and 2) met the spirometric criteria for COPD, based on UK guideline definitions (FEV 1 /FVC [forced vital capacity] ,0.7) at the baseline assessment.

statistical methods
Descriptive statistics are reported on all variables. Univariate associations between disease-specific QoL (CAT score) and other variables were analyzed using linear regression. Initially we fitted univariate regression models to determine which variables were significantly associated with CAT at the 5% level. These variables were taken forward into multivariate regression models to determine which were independently predictive in mutually adjusted analyses. We considered variables within three categories: 1) demographic characteristics (age, sex, body mass index [BMI], socioeconomic status [Index of Multiple Deprivation {IMD}, as quintiles], 26 employment status, comorbidity); 2) COPD-related factors (FEV 1 % predicted, physical activity, chronic cough/phlegm, chronic wheeze, breathlessness, exacerbations, hospitalizations); and 3) psychosocial and behavioral factors (anxiety, depression, dysfunctional breathing symptoms, illness perception, agoraphobic avoidance).
Those variables that were independently predictive of CAT score at the 5% significance level within each of these categories were taken forward in a final regression model to

249
Determinants of Qol in COPD determine which variables were independently predictive overall, and to explore whether psychological measures had an association with CAT score independent of patient characteristics and COPD-related symptoms. In the multivariate models, only cases without missing data in any field were included and none of the missing data were imputed. Although this does result in a reduction in the sample size, as the purpose of this study was to explore the relationship between variables and to determine which were most highly predictive of HRQoL, it was important that participants contributed data on all variables. A multiple imputation model would have required some assumption of the likely distribution of missing values and might have introduced bias into our exploratory analysis by imposing a likely relationship between variables and between the variables and HRQoL.
The R-squared value was used in order to quantify the extent to which the variables explained the variation in CAT score. To determine the contribution that each predictor made to the overall variance, and therefore their relative importance, a dominance analysis was undertaken. 27 Dominance analysis examines the change in R-squared from adding a variable to all possible subset regression models and then averaging across all possible models. In this way, it is possible to obtain a general dominance weight and thereby partition the R-squared value among the predictors. All variables that were significant in their respective categories were included in the dominance analysis.
A further analysis was conducted using the same methodology with generic HR QoL score EQ-5D 5L rather than the disease-specific CAT score as the outcome variable.

eligible population
One thousand six hundred three cohort participants meeting entry criteria and with valid spirometry readings were posted the additional questionnaire sets, and 1,233 were returned (76.9%). Of these, 181 were excluded from analysis as they did not have obstructive spirometry (FEV 1 /FVC ratio ,0.70), and a further 317 as they did not have a valid CAT score, leaving an analysis sample of 735 participants (consisting of 599 previously diagnosed COPD subjects and 136 identified through case finding). The case-found patients had milder disease, with better-preserved lung function, less breathlessness, chronic cough and wheeze, and a history of fewer exacerbations (Table 1).
A large variation in CAT score was observed (mean, SD CAT score 17.9, 8.3). Case-found participants reported less COPD impact compared to those with diagnosed COPD (14.0, 7.6 vs 18.7, 8.2, respectively).

Univariate and multivariate associations with COPD-related Qol Demographic characteristics
Four patient characteristics were significantly associated with higher CAT score: higher level of deprivation (IMD status), being currently unemployed, younger age, and being a current smoker ( Table 2). Although these associations were statistically significant at the 5% level, the R-squared values were low. The highest R-squared value was for deprivation and even that explained only 5.7% of the variation in CAT scores.

lung function and COPD-related symptoms
The independent predictors in this category were chronic wheeze, chronic cough/phlegm, exercise capacity at baseline (sit/stand repetitions), MRC score, Borg breathlessness pre-and post-exercise scores, and having $2 exacerbations in the previous 12 months ( Table 3). The MRC score was the strongest predictor, explaining 46% of the variation in CAT score in this category of factors. The other variables explained no more than 16% each, with lung function impairment (% predicted FEV 1 ) independently explaining under 6% of CAT variation in this category group.

Psychosocial and behavioral factors
The PHQ-9 (measuring depression), Nijmegen questionnaire (measuring symptoms of dysfunctional breathing), and the brief IPQ (measuring illness perceptions) were all significantly associated with CAT in this group of factors, and explained 34%, 38%, and 45% of the variance in CAT, respectively. Anxiety (GAD-7) scores were not significantly independently associated and agoraphobic avoidance scores were only weakly associated (Table 4).

Full multivariate model and dominance analysis
A full multivariate model was constructed including all variables independently associated with CAT from each category ( Table 5). This model only included people without missing data for any variable (n=476). In this model, depression (PHQ9), dysfunctional breathing symptoms (Nijmegen score), illness perception (IPQ), symptoms of both chronic wheeze and chronic phlegm/cough, exercise capacity (sit/ stand repetitions), breathlessness (MRC score and Borg score post-exercise), having $2 exacerbations in the previous year, and deprivation (IMD quintile) were all significantly associated with CAT at the 5% level. This model explains 72.2% of the observed variation in CAT.
In the dominance analysis, the largest contributions to variation in CAT scores were made by functional breathlessness (MRC score, 20.2% of R-squared value), with illness perception (IPQ) providing only a slightly lower contribution analysis using generic hrQol (eQ-5D 5l) A further analysis was performed with generic QoL score (EQ-5D 5L) as the outcome measure (Table 6). Generally, the same variables emerged as important predictors, although measures of depression (PHQ-9) and anxiety (GAD-7) were more strongly associated with EQ-5D 5L than with CAT scores. This model predicted 60.4% of the variation in the EQ-5D 5L, with the largest contributions made by depression (PHQ-9, 23.1%), functional breathlessness (MRC score, 20.1%), illness perception (IPQ, 16.0%), anxiety (GAD-7, 14.0%), and symptoms of dysfunctional breathing (Nijmegen score, 13.3%). The other significant variables (exercise capacity, BMI, physical activity levels, breathlessness pre-exercise, smoking status, deprivation, chronic wheeze, and employment status) all made smaller contributions (5% or less).

Discussion
The stimulus for this study was the observation that there is a large discrepancy at an individual level between the objective, biologic severity of COPD and the impact of the disease on the patient, as assessed by functional impairment and effects on QoL. We aimed to assess and quantify the relative contributions of psychosocial and behavioral factors on QoL, in order to identify candidate targets for future interventions. In this cohort of community-based patients predominantly having physiologically mild-to-moderate COPD, we observed a wide variation in disease-related QoL scores, with a mean (SD) CAT score of 17.9 (8.3). The reference values for CAT scores suggest that a score of ,10 indicates low impact, 10-20 moderate impact, and .20 high impact, 28 showing that the patients in our cohort spanned a wide range of perceived QoL impairment from their COPD, ranging from low to very high impact. As in previous research, a weak relationship was observed between physiologically assessed lung function and QoL. 29 We found, unsurprisingly, that a major independent contribution to impaired QoL came from functional, activityrelated breathlessness, with the MRC breathlessness score explaining 20% of the variation in both disease-specific (CAT) and generic QoL scores (EQ-5D 5L). However, we also found that both pessimistic health beliefs (as measured by the validated IPQ) and depressed psychological state (assessed by the PHQ-9) have major independent impacts on patients' experience of their COPD, explaining 20% and 12%, respectively, of the variation in CAT scores and high proportions (16% and 23%, respectively) of the variation in EQ-5D 5L. In addition, we found that dysfunctional breathing (measured by the Nijmegen Questionnaire) independently explained a large proportion of QoL variability (17% of CAT and 13% of EQ-5D 5L). All these factors are potentially amenable to intervention. Overall, the models we constructed explained over 70% of the variation in disease-specific QoL and over 60% of that in generic QoL.
Our findings are in keeping with other smaller studies suggesting that psychological factors and illness perceptions are important determinants of well-being, overshadowing the influence of "harder" factors such as lung function in COPD. 3,8,12,30,31 Illness perception is a construct of the cognitive representations and beliefs that patients have about their illness, and has been found to be an important determinant of behavior and to be associated with a number of important outcomes, including treatment adherence and functional recovery. 32 The Brief IPQ is a widely used and validated

253
Determinants of Qol in COPD tool with good psychometric properties. 33 There is limited research linking poor illness perception with disability and impaired QoL in COPD, 34 and no interventional studies that we are aware of. There is emerging evidence that brief, straightforward psychoeducational interventions can help to modify negative illness beliefs and lead to improvements over a range of different health outcomes. 35 In view of the findings we report in this study, there is a case for developing interventional studies based on strategies to improve illness perception in COPD.
There is consistent evidence that depression and anxiety are common comorbidities in patients with COPD and are associated with poor outcomes. [3][4][5][6][7][8][9][10]36,37 There is, however, surprisingly little evidence to support the use of psychological interventions in managing COPD, or to clarify which interventions are most effective and acceptable. 38 Although there is some preliminary evidence to support interventions such as relaxation, cognitive behavioral therapy, self-management, and antidepressant medication in COPD, the data are limited and mainly consisted of small studies. 38 Qualitative evidence suggests that psychological interventions are acceptable to COPD patients with disproportionately impaired QoL in relation to lung function impairment, particularly interventions based on nonpharmacologic strategies. 13 Our finding that a large proportion of QoL variation is explained by psychological factors supports the need for interventional studies.
A novel contribution of this study was to include a measure of dysfunctional breathing, the Nijmegen Questionnaire,  which was found to be a major contributor to the variation in both generic and disease-specific QoL. Although this questionnaire has not been validated for use in COPD, it has been widely used in other airways diseases including asthma, 39 with breathing retraining interventions shown to improve QoL scores. 40 The significant independent association of Nijmegen Questionnaire score to QoL improvement in patients with COPD indicates that breathing retraining may also be a possible strategy to improve QoL in people with COPD. A strength of this study was the availability of data from a large well-characterized UK COPD cohort, containing a relatively representative sample of patients with mild and moderate COPD treated in primary care. The large sample size, compared with much smaller previous studies (two studies had samples of ,100), 12,30 and the collection of data on a wide panel of disease-related factors allowed us to examine the relative contribution of different factors relating to QoL impairment in COPD. We were able to include a wide range of potential explanatory variables in the models, which were more inclusive than those in previous studies. Also, unlike previous studies that used only generic QoL tools, 31 we used both generic and disease-specific instruments. Similar messages emerged from our study for disease-specific and generic QoL instruments.

255
Determinants of Qol in COPD A limitation of our study is that it is cross-sectional, describing associations between measured factors and QoL that cannot be assumed as causal, and there is no certainty that interventions targeting these factors will result in clinically important improvements in QoL. It does, however, provide a justification for future interventional studies targeted on improving illness perception, improving depression and correcting dysfunctional breathing in COPD patients, particularly in those with disproportionately impaired QoL in relation to lung function impairment.

Conclusion
COPD is a distressing and progressive condition that makes life miserable for many, and a holistic and multidimensional, personalized approach is needed. 14 The clinical focus in COPD is generally directed toward biologic factors such as physiologic impairment and airways inflammation, with a predominant focus on pharmacologic interventions. However, along with smoking cessation, nonpharmacologic approaches are also very important in improving patient outcomes, with pulmonary rehabilitation now recognized as being a key element in overall disease management. This study shows that there are a number of factors associated with QoL impairment that could potentially be modified through suitable nonpharmacologic interventions focused on appropriately characterized patients, and supports the need for future interventional studies.
The International Journal of COPD is an international, peer-reviewed journal of therapeutics and pharmacology focusing on concise rapid reporting of clinical studies and reviews in COPD. Special focus is given to the pathophysiological processes underlying the disease, intervention programs, patient focused education, and self management protocols.
This journal is indexed on PubMed Central, MedLine and CAS. The manuscript management system is completely online and includes a very quick and fair peer-review system, which is all easy to use. Visit http://www.dovepress.com/testimonials.php to read real quotes from published authors.