Chronic obstructive pulmonary disease (COPD) is characterized by persistent respiratory symptoms and airflow limitation. Patients with COPD experience exacerbations and disease progression that lead to increased mortality risk. The Global Initiative for Chronic Obstructive Lung Disease 2011 (GOLD 2011) guidelines introduced a multidimensional assessment tool to guide therapy by classifying COPD patients into four groups: A (less symptoms and low risk), B (more symptoms and low risk), C (less symptoms and high risk), and D (more symptoms and high risk).1

The factors used in COPD assessment include the symptoms, which are evaluated by the modified Medical Research Council (mMRC) grade,2 COPD Assessment Test (CAT), or St. George Respiratory Questionnaire (SGRQ) score3,4; forced expiratory volume in 1 second (FEV1),3,5,6 and previous exacerbation history.3,7 It is well known that each of these factors is associated with future risks, including exacerbation and mortality, although the ABCD categorization was not intended to predict outcomes. Among these factors, prior history of exacerbation is known to be the strongest predictor of future exacerbation.3 However, it has not been investigated whether symptoms or FEV1 aside from prior exacerbation history are more likely to be associated with future exacerbation risks. Furthermore, although the GOLD 2017 update excluded FEV1 for risk group assessment,8 it has not been fully investigated whether GOLD 2017 has a weaker or stronger relationship with future exacerbation than prior GOLD grade and GOLD 2011.

This study investigates whether the GOLD 2017 classification excluding FEV1 as a criterion is associated with different future exacerbation risk from the GOLD grade and GOLD 2011 categorizations. The study also compares the symptoms and FEV1 without the previous exacerbation history in terms of their associations with future exacerbations.

Among 1511 patients who were enrolled from the three cohorts, 611 patients were included in the study (36 from SNUH, 257 from KOCOSS, and 318 from KOLD), as shown in Fig. 1. The baseline characteristics of patients are compared among the three cohorts in Table 1. The mean age was 67.6 years, and 593 patients (97.1%) were male. The average CAT and SGRQ scores were 15.4±7.4 and 35.7±19.8, respectively, and 311 (50.9%) patients had dyspnea with an mMRC grade ≥2. Of these, 22.1% had at least one exacerbation in the past year, and about half had experienced at least one severe exacerbation.

Fig. 1.

Flowchart for patients inclusion.

(0.09MB).

According to the GOLD 2017 classification, only 14 (2.3%) patients were categorized into group C. The proportion of patients classified into group D by GOLD 2017 was less than that by GOLD 2011 (15.6% vs. 37.3%; Table 2). The distribution of patients was significantly different in all comparisons between any two classification systems (P<0.001). The risks of moderate to severe exacerbation were not different among GOLD grades. Groups B–D had significantly higher exacerbation risk than group A according to GOLD 2011 and GOLD 2017 (except for group C by GOLD 2011 in the logistic model; adjusted odds ratio (aOR): 2.13; 95% CI: 0.88–5.16).

No model clearly showed better outcomes according to the order of groups A–D. According to the GOLD 2017 classification, group C had the highest exacerbation risk (aOR: 3.71; 95% CI: 1.12–12.3; adjusted hazard ratio (aHR): 3.02; 95% CI: 1.52–5.99). When groups C and D were combined into one category (group CD), the GOLD 2017 classification showed an increased risk of exacerbation in the order of A-B-CD (Table 2 and Fig. 2). The GOLD 2017 classification had lower AIC and BIC than the GOLD grade and GOLD 2011 classification. There were no statistically significant differences between GOLD 2017, GOLD 2011, and GOLD grade in c-statistics (AUROC: 0.635 vs. 0.651 vs. 0.659; P=0.492; Harrell's C, 0. 609 vs. 0.625 vs. 0.635 for GOLD grade, GOLD 2011 and GOLD 2017, respectively; P>0.05 for all comparisons; Table 3). The calibration of all models was fair (data not shown). We also made various other models by combining different symptom and risk criteria, which showed similar results (supplement tables 2 and 3).

Fig. 2.

Time to first exacerbation in each category of different classification systems. (A) GOLD grade, (B) GOLD 2011, (C) GOLD 2017, (D) GOLD 2017 (CD).

(0.36MB).

We also investigated the risk of exacerbation among 8 groups (Table 4) categorized by their previous exacerbation history, FEV1, and their symptoms. All models were adjusted by various covariates. Among those with a total exacerbation level <2 and no severe exacerbation in the past year (groups 1–4), the group with more symptoms had significantly higher future exacerbation risk than the group with less symptoms among those with FEV1 ≥50% (group 2 vs. group 1; P=0.009). We did not find a significantly higher exacerbation risk in the more-symptom group than in the less-symptom group in other comparisons (group 4 vs. group 3; group 6 vs. group 5; group 8 vs. group 7). A lower FEV1 (FEV1 <50%) was not associated with a higher future exacerbation risk than a higher FEV1 (FEV1 ≥50%), regardless of prior exacerbation history and symptom group (group 3 vs. group 1; group 4 vs. group 2; group 7 vs. group 5; group 8 vs. group 6). There were few participants in groups 3 (n=22), 5 (n=9), and 7 (n=5).

In our study, there were no differences in future exacerbation risks based on c-statistics between GOLD grade, GOLD 2011, and GOLD 2017 categorizations, although GOLD 2017 showed lower AIC and BIC. More symptoms were significantly associated with future exacerbation risks among those with no frequent exacerbation history and FEV1 ≥50%, but FEV1 <50% did not show an increased association with future exacerbations as compared to FEV1 ≥50%, regardless of prior exacerbation history and symptom categories. All statistical models were adjusted by various covariates, including demographic characteristics, smoking history, comorbidities, and treatment during the follow-up period.

The strength of this study is that we compared various classification systems, which were different in risk assessment: airflow limitation alone (GOLD stage), clinical presentation such as exacerbation history (GOLD 2017), and both of these (GOLD 2011). GOLD 2011 implemented a comprehensive assessment tool in order to guide therapy based on patient symptoms and risk of a future adverse event.1 However, the GOLD 2011 assessment tool applied heterogeneous grouping, which leads to mixed clinical outcomes within groups C and D.10,11

The COPDGene study showed heterogeneous exacerbation rates depending on whether the patients were categorized as group D according to lung function alone (0.89exacerbations/person-year), exacerbation history alone (1.34exacerbations/person-year), or both (1.86exacerbations/person-year). The patients categorized into high-risk groups based on FEV1 only (C1 or D1) showed a similar exacerbation risk to group B patients.12 Our study showed that he GOLD 2017 assessment tool without lung function shows no less association with future exacerbation risk compared with previous GOLD classifications. In addition, our study also suggested that categorization according to FEV1% is not associated with future exacerbation risk when considering prior exacerbation history and symptoms. The recently developed B-AE-D index includes exacerbation history but not lung function.13 Several published COPD guidelines also classify patients without considering lung function.14,15

It is well known that the history of previous exacerbation is the strongest predictor of the future exacerbation risk.3 However, it has not been clarified whether symptoms or FEV1 is more related to future exacerbation risk, aside from previous exacerbation risks. In our study, symptoms were significantly associated with future exacerbation, at least in patients with no previous frequent exacerbation history and FEV1 ≥50%. The more-symptom group showed a numerically higher risk of exacerbation in those with FEV1 <50%, but statistical significance was not detected, which could have been due to the small number of participants in some of the groups. However, FEV1 was not associated with exacerbation risk in all comparisons.

The influence on future exacerbation risk of the GOLD 2017 classification in which FEV1 was not considered in the categorization of patients was not less than that of the GOLD grade and GOLD 2011. This also suggests that FEV1 might not be an important predictor for future exacerbation risks. In addition, it might partly explain why GOLD group B paradoxically showed significantly poorer survival than group C in some cohort studies.16

In our study, the percentage of group C patients according to GOLD 2017 was only 2.3%, which was anticipated considering that one of the problems in the GOLD 2011 categorization is that only a small number of patients are classified into group C.5,17,18 The number of patients categorized into group C by the GOLD 2017 classification should be less than that by the GOLD 2011 classification because patients in group C according to GOLD 2011 with FEV1 <50% and <2 exacerbations in the past year were moved to group A. When groups C and D were combined into one category (group CD), GOLD 2017 showed an increased effect size of risks for exacerbation corresponding with the order of A-B-CD groups. Notably, there are already guidelines that use a classification similar to the A-B-CD system.19

This study has several limitations, including a short follow-up period of 1 year and a relatively small number of enrolled patients. Because these cohorts are still actively recruiting patients, many participants were followed up for less than 1 year at the date of our analysis. Excluding participants with a follow-up shorter than 1 year might lead to selection bias. However, international guidelines recommend that the study duration should be at least 1 year if the objective of the study is to investigate exacerbations.20 Thus, we included participants who were followed-up for at least 1 year. Further follow-up of these patients is needed to evaluate long-term outcomes, including exacerbations and mortality.

The generalizability of our study might be limited because it included only Korean patients from a predominantly male cohort. The possibility of including relatively mild patients (grade 4 represented only 4.5%) could be another limitation. Additionally, we did not consider other potential factors that might be related to exacerbations, such as COPD phenotypes (i.e., chronic bronchitis21), comorbid bronchiectasis,22,23 and exercise capacity24,25 because our study focused on the comparisons of items used in the GOLD classification. Instead, models were compared after adjusting for age, sex, body mass index, smoking status, medication use, and comorbidities.

Furthermore, most participants in our cohort were male (97.1%), which might affect the results because sex differences in the clinical course of COPD have been suggested.26 It should be stressed that patient characteristics lead to changes over time in the group to which patients belong, including symptoms, lung function, and exacerbation experience of those with COPD. However, we used only baseline characteristics. We applied SGRQ as symptom criteria, which is not a recommendation of the GOLD statement, as no participants underwent CAT or CCQ at baseline in one cohort that was initiated in 2005. However, GOLD states a symptom score equivalent to SGRQ ≥25 should be used as a threshold for considering regular treatment for symptoms.27 Finally, although GOLD assessment should also consider lung function severity, symptoms, and comorbidities in order to guide therapy, our analysis only focused on exacerbation. In addition, it should be considered that an association between the GOLD categorization and future outcomes might not only depend on the predictability of classifications but also on the adherence to the therapy according to the GOLD statement classification.

In conclusion, the GOLD 2017 classification in which FEV1 was not considered in the categorization of patients was not different from GOLD grade and GOLD 2011 in terms of the association with future exacerbation risk. FEV1 might be a less important factor for predicting exacerbation risk than prior exacerbation history and symptoms. These results partly support the GOLD 2017 assessment tool, although the size of group C was very small.

The authors declare no conflict of interest.