Comorbidities and COPD severity in a clinic-based cohort

COPD has emerged as the most important respiratory disease worldwide. The epidemiology of COPD had changed in recent years, with more women affected [1], fewer old subjects and more medications available for health providers.

To improve the management of COPD and take into account the heterogeneity of the disease, the Global Obstructive Lung Disease Initiatives [2] proposed a new classification in 2011 that takes into account respiratory symptoms, the burden of exacerbations and lung function.

Comorbidities in COPD have received considerable attention as COPD patients frequently suffer from comorbidities such as cardiovascular and cerebrovascular disease, lung cancer and diabetes, with a significant impact on mortality that was termed by Divo et al. known as the “comorbidome” [3]. They constructed a comorbidity index (COTE index) based on 12 comorbidities that seem to negatively influence survival. However, the use of indexes like the COTE and the BODE [4] in clinical practice needs to be clarified. The validity of the COTE has been questioned since patients with GOLD B seem to have worse survival than patients with GOLD C, because of the particular heart disease found in a very large population study in Copenhagen area [5]. Some authors suggest that the presence of comorbidities should influence the relationship between the GOLD score and lung function measurements, the former perhaps being more representative of morbidity than of COPD severity [6].

The distribution and the type of comorbidities seem to vary between studies, except for cardiovascular disease which seems to be stable across them [7]. The complexity of COPD was reported in the Eclipse cohort, suggesting that COPD includes several different phenotypes when taking into account clinical parameters, survival, hospitalization, comorbidities and systemic inflammation [8]. Recently, in a complex analysis using network analysis, Divo et al. showed that comorbidities in COPD do not occur by chance [9].

In 213 patients included in a rehabilitation center, Vanfleteren et al. identified 13 comorbidities and five comorbidity clusters: less comorbidity, cardiovascular, cachectic, metabolic and psychological [10]. However, little is known about the reproducibility of these comorbidity phenotypes in COPD patients in real life and their association with COPD severity. Recently, the ATS/ERS consensus statement recommended that studies be performed either to confirm or rule out an association between specific comorbidities and COPD [11]. In an ongoing prospective observational cohort of outpatients with COPD followed up by pulmonologists, the aim of our study was to determine the association between specific comorbidities and COPD severity.

This study sought to determine whether comorbidities were associated with COPD severity in a clinic-based cohort of COPD patients, mostly GOLD A and B, followed up by pulmonologists. Only 28.4% of patients had no associated comorbidities. Fourteen comorbidities were significantly different with COPD severity. In this large population of patients, the median number of comorbidities was two.

Hypertension, OSA, dyslipidemia, ischemic cardiopathy and heart rhythm disorder were more frequent in GOLD B and D patients than in the other groups, as were anxiety and depression. Undernutrition was the most frequent in GOLD D patients and osteoporosis was the most frequent in GOLD B and D subjects. The number of comorbidities was the highest in GOLD B and D patients. Even when the severity of symptoms was similar, the management of COPD seemed to be different according to whether patients had comorbidities or not. Finally, five clusters of comorbidities were established, the most frequent being the cluster with cardiovascular disease and obstructive apnea syndrome.

Our approach was to analyze the relationship between these comorbidities and COPD severity by using three different approaches: the impact of the number of comorbidities, the univariate association between the comorbidities and COPD severity and cluster analysis to determine the association between the comorbidities i.e. to establish the existence of different phenotypes. Our findings are consistent with previous publications reporting a high prevalence of comorbidity in COPD, particularly cardio-vascular disease. Chen et al. [7] in a large review reported that compared with the non-COPD population, patients with COPD were more likely to be diagnosed with cardiovascular disease (odds ratio [OR] 2·4; 95% CI 2·02–3·00; p < 0·0001), including a two- to five-fold higher risk of ischemic heart disease, cardiac dysrhythmia, heart failure, diseases of the pulmonary circulation, and arterial disease. Additionally, patients with COPD reported hypertension more often (OR 1·3, 95% CI 1·1–1·5; p = 0·0007), diabetes (1·3, 1·2–1·5; p < 0·0001], and ever smoking (4·2, 3·2–5·6; p < 0·0001). Divo et al. found in their cohort that cardiovascular disease was highly associated with the risk of mortality, but that the highest risk of mortality was associated with anxiety [3]. However, we found a high prevalence of OAS, probably owing to the overlap syndrome as reported by Soler et al. [14]. By contrast, OAS was the most frequent comorbidity is GOLD A and B patients although it seemed to be associated with moderate-to-severe COPD. It is essential to diagnose OAS in patients with COPD as patients with overlap syndrome who are not treated with CPAP have a higher mortality [15].

In our cohort, 14 comorbidities were significantly associated with COPD severity. Hypertension, OSA, dyslipidemia, ischemic cardiopathy, and heart rhythm disorder were more frequent in GOLD B and D compared to GOLD A and C. Anxiety and depression was higher in GOLD D and B, compared to the other groups of severity. These results are in line with the analysis performed in the Copenhagen cohort showing that GOLD B patients had more severe dyspnea and significantly poorer survival than group C ones, in spite of a higher FEV1 level [5]. The same trend concerned the number of comorbidities, with a prevalence of comorbidities (more than two) in GOLD B and D patients. At an equal level of severity, management of COPD seems to be different in severe COPD patients with comorbidities, with more LABA and SABA in severe COPD, suggesting that comorbidities could increase respiratory symptoms. Moreover, prevention of exacerbations requires interventions beyond the lungs, including treatment of comorbidities such as gastro-esophageal reflux disease, reduction of cardiovascular risks, and management of dyspnea and anxiety [16].

LABA were prescribed the most in GOLD A patients, which is not in agreement with the guidelines GOLD 2011. This could be due to the high prevalence of symptoms like cough in this group, as symptoms included in the GOLD classification are based on dyspnea and exacerbations but not on cough. We cannot rule out that symptomatic GOLD A patients could represent a specific phenotype. Recently, Woodruff et col. described a subgroup of symptomatic patients with no criteria for COPD regarding lung function [17]. In addition, management of COPD differed according to the comorbidities that patients had, even if those with or without had the same level of severity. This was particularly the case for rehabilitation and vaccination which were more prescribed in symptomatic GOLD B and D patients who had comorbidities than in those without.

We expected to have a gradient in COPD severity, perhaps patients GOLD B should be called differently, as they seemed to be more severe than GOLD C.

The cluster analysis revealed five clusters: The cluster analysis showed five phenotypes of comorbidities: cluster 1 included cardiac profile; cluster 2 included less comorbidities; cluster 3 included metabolic syndrome, apnea and anxiety-depression; cluster 4 included undernutrition and osteoporosis and cluster 5 included bronchiectasis. Vanfleteren found 13 comorbidities in a sample of 213 COPD patients [10] with five comorbid phenotypes: less comorbidity, cardiovascular, cachectic, metabolic, and psychological. Four of our clusters are concordant, i.e. cardiovascular, cachectic, metabolic and less comorbidities. Nevertheless, all the clusters were more significantly associated with GOLD D and in less manner with GOLD B. This finding could explain the higher risk of mortality in GOLD B and D patients, as previously reported elsewhere [5]. In the same way, Divo et al. [9] also identified a number of modules in the comorbidity network, including a cardiovascular one, and a module characterized by mil-moderate airflow limitation and metabolic syndrome with high BMI, these two modules are concordant with our findings.

Our results show that while comorbidity in COPD is a complex issue, comorbidities contributed prominently to the clinical severity of our patients, and that management of their severe COPD differed according to whether they had comorbidities or not, at the same level of obstruction.

Our study has some limitations. First, comorbidities were recorded by pulmonologists; previous studies showed that comorbidities are underdiagnosed in real life. This could also be the case in our study for most comorbidity except for OSA, as OSA was diagnosed by a polysomnography performed by the same pulmonologist who diagnosed COPD. However, we think that this bias is limited in our study, as we found a significant correlation between the comorbidities declared and compliance with the treatment given for them (data not shown). Second, we cannot generalize these findings to patients with COPD in the general population, as our population was managed both by a general practitioner and a pulmonologist. Third, we performed a multivariable exploratory analysis in order to better describe the associations between the different comorbidities [18]. This type of analysis uses a statistical method that processes a large amount of information from heterogeneous variables in homogeneous groups. It is well known that various factors can influence the analysis and therefore the results: the choice of the methods, hierarchical or nonhierarchical, the determination of the number of clusters before the analysis, the choice of the variables included in the analysis, the correlation between the selected variables and the clinical judgment of the investigators. To limit the impact of the specific correlation between the variables, we first performed a principal component analysis. Then we used the scree plot of the eigenvalue, the Kaiser-Guttman criterion and the percentage of variance explained to determine the number of clusters.

Lastly, we cannot validate our clusters in terms of survival as the study was performed with inclusion criteria, or with systemic inflammation. Further analysis with survival data from these COPD patients would provide important information for validating these clusters of comorbidity.

This study in a large clinic-based cohort shows that multimorbidity is common in patients with COPD, and that five comorbidity clusters can be identified. Patients GOLD B have more comorbidities than GOLD C.

The presence of comorbidities should therefore be included in any assessment of COPD severity. Further analysis is needed to validate these clusters in a future cohort.