Discussion
In this study, we identified three distinct subgroups of COPD through a cluster analysis of 272 patients with CODA cohort. We also demonstrated that the frequency of exacerbations requiring hospitalization, progress of respiratory symptoms, and changes in the mMRC and CAT scores in 1 year varied among these subgroups. Among the three subgroups, subjects with mild COPD were divided into two subgroups (subgroups 1 and 3), according to the number of symptoms. According to the GOLD classification of airflow limitation severity (based on post-bronchodilator FEV1), subgroup 1 (mild disease group) included younger patients who had fewer symptoms, and subgroup 3 (the other mild disease group) included a majority of female patients with more respiratory symptoms. However, subgroup 3 reported more exacerbations requiring hospitalization and more symptom progression during the 1-year follow-up than subgroup 1. Subgroup 2 (moderate disease group) included subjects with additional respiratory symptoms; this group had more frequent exacerbations requiring hospitalization during the 1-year follow up than subgroup 1 did.
There have been several reports on the various phenotypes of COPD in Western countries in order to identify more homogeneous subgroups [
12‐
16]; however, there have been few of such reports from Asian countries [
17]. In the Korean Obstructive Lung Disease (KOLD) cohort, three clusters with the following phenotypes were identified: cluster 1 included subjects with moderate-to-severe airflow obstruction and bronchodilator reversibility, cluster 2 included subjects with moderate airflow obstruction without bronchodilator reversibility, and cluster 3 included subjects with severe airflow obstruction without bronchodilator reversibility [
17]. In the KOLD cohort, in terms of risk factors, cluster 3 patients showed more severe airflow obstruction and hyperinflation, had greater emphysematous change in the CT scan, and smoked less [
17]. Conversely, in the present study, cumulative smoke exposure (pack-years) was the highest in subgroup 2 (moderate disease group with additional symptoms) compared with the milder subgroups. However, the three subgroups of the present study had a similar biomass exposure, which has previously been reported to result in phenotypic differences [
18]. Hong et al. suggested that the airway phenotype of COPD was more common in females, and females are more susceptible to the damaging effects of biomass smoke, thereby leading to the development of airway disease [
8]. In the present study, subgroup 3 included more female patients, a large number of never-smokers, and a small percentage of emphysema cases; however, they showed no differences in airway wall thickness or biomass smoking history. Cho et al. used clinical and genetic characteristics to cluster patients with COPD in the National Emphysema Treatment Trial Genetics Ancillary Study cohort with severe emphysema: 1) emphysema predominance, 2) milder severity and bronchodilator responsiveness, 3) discordant lung function/CT emphysema and airway severity, and 4) airway predominance [
14]. In the present study, subgroup 2 exhibited severe emphysema and the lowest FEV1, and approximately 20% of the patients were bronchodilator responders; this value was not different among the three subgroups. Furthermore, no difference in airway wall thickness was observed among the subgroups. Regarding the 1-year follow-up, an average of 60 mL of FEV1 decline was noted in subgroup 2, but this value was not significantly different among the subgroups. However, the mMRC and CAT scores improved only in subgroup 1, which exhibited fewer symptoms, such as cough, sputum production, and chronic bronchitis, during the 1-year follow-up. The PCA variables were the ones that changed differently according to subgroups. This may have influenced the results. For example, FEV1, was included in the PCA variables, and patient with better lung function showed more lung function decline in the previous report [
19]. However, subgroup 2, which have worse lung function showed more decline in the current study. We did not compare the treatment history; therefore, we could not identify which subgroup would benefit from bronchodilators and/or inhaled corticosteroid (ICS) treatment. Lee et al. suggested that the response to long-acting beta2-agonist and ICS treatment varied with the COPD subtype, and the obstruction-dominant COPD patients exhibited the best response compared with the emphysema-dominant patients who had the worst response [
20].
Most recently, Castaldi et al. evaluated 10,192 subjects from the COPD Gene cohort: (1) relatively smoking-resistant individuals, (2) individuals with mild upper zone-predominant emphysema and airflow obstruction, (3) individuals with airway-predominant disease, and (4) individuals with severe obstruction and emphysema [
4]. These clusters were strongly associated with known COPD-associated variants [
21]. The COPD Gene study reported that the severe subgroup had older and more male patients, and the severe emphysema group showed the most frequent exacerbations and the worst symptoms [
4]. This was similar to our study in that the subgroup with a relatively severe stage of disease and additional symptoms (subgroup 2) showed the most severe emphysema and the most frequent exacerbations. In addition, this subgroup had more male and relatively older patients, the lowest BMI, and the highest IL-6 and CRP values compared with the mild disease subgroups.
Garcia-Aymerich identified three clusters in 342 patients who were hospitalized for the first time because of an exacerbation of COPD and proposed clinically relevant COPD subtypes [
13]. Interestingly, these three clusters relatively correspond to our subgroups in terms of their clinical features and follow-up outcomes, such as subsequent hospitalizations. According to the ‘Phenotype and Course of COPD (PAC-COPD)’ study group [
13], one cluster displayed the worst status in most of the respiratory domains of the disease, such as exercise capacity, more frequent hospitalizations due to COPD, and the highest all-cause mortality; these features correspond well with our subgroup 2. The remaining two clusters of the PAC-COPD study group were characterized by a milder respiratory status, which closely resembles our subgroups 1 and 3; one subtype of the milder clusters had a higher prevalence of obesity, cardiovascular disease, diabetes and higher levels of systemic inflammatory markers. In the present study, however, the more severe stage group (subgroup 2) showed the highest IL-6 and CRP values but the lowest BMI, whereas the incidences of comorbidities was not different among the three clusters.
The present study included patients with relatively mild stages of COPD. The mild-severity group was divided into two subgroups according to the symptoms. The subgroup with additional symptoms, among the subjects with mild airway obstruction, experienced more exacerbations, requiring hospitalization during the 1-year follow-up. These findings may provide an important understanding of COPD phenotypes in terms of prognosis of symptoms and may also demonstrate the importance of the early management of COPD [
22]. Moreover, the results of the present study suggest that the COPD classification system of the Korean COPD guideline is reasonable for the prediction of disease prognosis in Korean COPD patients. The Korean guideline classified the COPD patients into three groups, combining the GOLD C and D groups into one group (group “da”) [
23]. In the present study, the subgroup with additional symptoms and a relatively severe stage of disease (subgroup 2) showed the most frequent exacerbations requiring hospitalization and the most progressive symptoms, according to the mMRC and CAT scores. Subgroup 2 mostly met the criteria for group B; however, 25% of them showed features of groups C and D, according to the GOLD classification. Subgroup 1 included the youngest patients with mild symptoms and mild severity; this group seldom had acute exacerbations or symptom progression. Subgroup 1 mostly met the criteria for group B, but 40% showed group A characteristics, according to the GOLD classification. Subgroup 3 included more female patients with a lower smoking history and showed the mild severity but with more symptoms and acute exacerbations than subgroup 1 did. Subgroup 3 exclusively included group B patients, according to the GOLD classification.
Some potential limitations of our study are as follows. First, Biomass exposure was only measured using self-reported questionnaires and this may explain the lack of differences in biomass exposure among the three subgroups. While biomass exposure may have effects on phenotypic differences of COPD, better measurement of exposure to wood smoke constituents using validated questionnaire instruments or home exposure-monitoring devices will be needed to detect them. Secondly, the study population was localized to several provinces in Korea and the sample size was relatively small. The present study included only patients with mild-to-moderate stage COPD, which might be another limitation. Therefore, our results should be extrapolated to heavy smokers and patients with severe COPD with caution. Thirdly, the population of this study included approximately 20% of bronchodilator responders; however, we did not analyze the proportion of airway disease, as measured by the methacholine reactivity test. Hence, some patients with bronchial asthma or asthma–COPD overlap might be included. However, there were no differences in the percentage of bronchodilator responders and airway thickness among the three subgroups; accordingly, we suggest that there was little confounding due to possible bronchial asthma or asthma–COPD overlap patients. We did not analyze whether the patients used bronchodilators or ICSs; the treatment outcome of the medications was also not analyzed. Lastly, we analyzed only the 1-year follow-up results; therefore, the long-term prognosis, such as mortality, could not be determined. Nevertheless, the possible clinical implication of this study is that our COPD cohort from dusty areas comprises discrete groups of subjects with different clinical characteristics associated with different outcomes as in other COPD cohorts. We identified three subgroups of COPD patients in this population. Although this study included several environmental factors for cluster analysis, additional long-term follow-up and multinational studies using exposure metrics are warranted.