Risk Factors for Chronic Obstructive Pulmonary Disease Exacerbations among Individuals without a History of Recent Exacerbations: A COPDGene Analysis

Abstract

Rationale

Acute exacerbations of chronic obstructive pulmonary disease (AE-COPD) are detrimental events in the natural history of COPD, but the risk factors associated with future exacerbations in the absence of a history of recent exacerbations are not fully understood.

Objectives

To identify risk factors for COPD exacerbations among participants in the Genetic Epidemiology of COPD Study (COPDGene) without a history of exacerbation in the previous year.

Methods

We identified participants with a smoking history enrolled in COPDGene who had COPD (defined as forced expiratory volume in 1 second [FEV₁]/forced vital capacity < 0.70), no exacerbation in the year before their second study site visit, and who completed at least one longitudinal follow-up questionnaire in the following 36 months. We used univariable and multivariable zero-inflated negative binomial regression models to identify risk factors associated with increased rates of exacerbation. Each risk factor’s regression coefficient (β) was rounded to the nearest 0.25 and incorporated into a graduated risk score.

Results

Among the 1,528 participants with a smoking history and COPD enrolled in COPDGene without exacerbation in the year before their second study site visit, 508 participants (33.2%) had at least one moderate or severe exacerbation in the 36 months studied. Gastroesophageal reflux disease, chronic bronchitis, high symptom burden (as measured by Modified Medical Research Council Dyspnea Scale and COPD Assessment Test), and lower FEV₁% predicted were associated with an increased risk of exacerbation. Each 1-point increase in our graduated risk score was associated with a 25–30% increase in exacerbation rate in the 36 months studied.

Conclusions

In patients with COPD without a recent history of exacerbations, gastroesophageal reflux disease, chronic bronchitis, high symptom burden, and lower lung function are associated with increased risk of future exacerbation using a simple risk score that can be used in clinical practice.

Chronic obstructive pulmonary disease (COPD) is a prevalent and morbid condition associated with a high mortality rate. The disease affects nearly 174 million people worldwide and roughly 29 million people in the United States (1, 2). It is the eighth leading cause of disability worldwide, also accounting for 1.5 million emergency department visits and nearly 700,000 admissions in the United States (1, 3). A total of 3.2 million deaths worldwide in 2015 were a result of COPD, and it is the fourth leading cause of death in the United States (1, 4).

COPD exacerbations are detrimental events in the history of COPD. Exacerbations are associated with worse quality of life, higher mortality, accelerated lung function decline, and increased healthcare costs (5–8). As a result, research has focused on identifying those at risk of COPD exacerbations. Previously identified risk factors include chronic respiratory symptoms such as cough and chronic mucus hypersecretion, gastroesophageal reflux disease (GERD), cardiovascular comorbidities, reduced forced expiratory volume in 1 second (FEV₁) percent predicted, and blood eosinophil count (9–13). Perhaps the strongest risk factor identified has been prior history of exacerbation (14). However, at some point most patients without prior events do develop exacerbations. Among these patients without a recent history of exacerbation, risk factors associated with the occurrence of future exacerbations are not completely understood.

In this analysis, we aimed to identify routinely collected clinical parameters that could serve as risk factors for identification of future exacerbations in patients with COPD without a history of exacerbation in the previous year. Previous research has shown that the majority of patients with COPD are taken care of in primary care and only a minority of patients with COPD regularly see a pulmonologist (15). As a result, we aimed to identify risk factors that could be used in a variety of healthcare settings.

A portion of the work contained in this manuscript has been previously published in abstract form (16).

Methods

Design and Participants

This analysis used data from participants in the Genetic Epidemiology of COPD Study (COPDGene; ClinicalTrials.gov identifier: NCT00608764). COPDGene is an ongoing multicenter cohort study to identify genetic risk factors in smoking-related lung disease and define disease phenotypes. The protocol for COPDGene was approved at the University of Michigan, where this analysis was conducted (University of Michigan Health System Research Committee Institutional Review Board approval HUM000014973, July 16, 2010) and all other participating centers. Informed consent was provided by all participants in the study.

The design of COPDGene has been previously described (17). Briefly, COPDGene recruited non-Hispanic white and Black or African American individuals who had at least a 10 pack-year smoking history and were between age 45 and 80 years. A small group of never-smoker control subjects were also included. Site visits for COPDGene have occurred at 5-year intervals, with the initial site visits taking place between 2008 and 2012 and subsequent site visits occurring on a rolling basis every 5 years thereafter. At these site visits, demographic data, including self-reported medical history, medications, and health-related quality of life scores (Modified Medical Research Council [mMRC] Dyspnea Scale and COPD Assessment Test [CAT] score, among others) are collected. Spirometry is performed using an EasyOne spirometer (ndd Medical Technologies) with trained research staff. Inspiratory and expiratory chest computed tomography scans are completed and blood samples obtained. Longitudinal follow-up questionnaires are completed at 3- to 6-month intervals between site visits via an automated telephony or web-based system through which participants complete a branching questionnaire about respiratory exacerbations, new long-term medications, new comorbid conditions, current smoking status, and general health status since last contact (18).

Participants included in this analysis completed their second site visit as part of the COPDGene protocol, had COPD (defined as a post-bronchodilator FEV₁/forced vital capacity [FVC] < 0.70 and classified as Global Initiative for Chronic Obstructive Lung Disease [GOLD] grade 1–4) (19), had no COPD exacerbations in the year before their second site visit, and completed at least one longitudinal follow-up questionnaire in the subsequent 36 months after the second site visit. Although individuals with tobacco smoke exposure can develop respiratory symptoms and have abnormal computed tomography imaging, airflow obstruction defined as FEV₁/FVC < 0.70 remains mandatory for the diagnosis of COPD by GOLD. As such, we excluded participants who did not meet this criterion. For the purposes of this analysis, we treated the second COPDGene site visit (i.e., the participant’s second site visit as part of the COPDGene protocol, which occurred approximately 5 years after their initial visit) as our baseline visit. This gave us access to a broader set of variables collected at the second site visit (such as CAT score) as well as a greater level of granularity regarding exacerbations before and after the visit than had we used the first site visit as the baseline.

The potential risk factors considered as part of this analysis included age, sex, body mass index, smoking status, GERD, chronic bronchitis, coronary artery disease, congestive heart failure, CAT score, mMRC score, FEV₁% predicted, white blood cell count, blood eosinophil count, and blood neutrophil/lymphocyte ratio. These risk factors were based on previous research that has identified them as risk factors among patients with a history of COPD exacerbation and also because it was believed the majority are readily available or could be easily obtained in clinical practice.

COPD Exacerbations

For the purposes of this study, COPD exacerbations were defined as an episode of increased cough and phlegm or shortness of breath that lasted for 48 hours or more and required systemic steroids and/or antibiotics with or without an emergency department visit or hospital admission. Those that only required systemic medications were considered moderate exacerbations, and those that required an emergency department visit or hospitalization were considered severe. The outcome for this analysis consisted of the cumulative number of exacerbations reported by the participant via automated telephony or web-based longitudinal follow-up questionnaires during the 36 months after their second site visit. A summary of exacerbations per year of follow-up by baseline inhaled medication is included in the data supplement (Table E1).

Statistical Analysis

Exacerbation rates were modeled using univariable and multivariable zero-inflated negative binomial regression models with an offset for the logarithm of follow-up time. In all models, the logistic regression component, which models the probability of being susceptible to exacerbations, had FEV₁% predicted as the sole predictor. In all multivariable models, the negative binomial regression component, which models the mean exacerbation count, controlled for age, sex, and smoking status. Risk factors included in this model were based on previous research showing an effect on exacerbation rate. Risk factors that were associated with a statistically significant increase in exacerbation rate in univariable analysis were included in a multivariable model. Risk factors no longer statistically significant were removed from the multivariable model, and the remaining variables and their model coefficients were used to form graduated risk scores. We approximately rounded each risk factor’s regression coefficient (β) to the nearest 0.25 and then assigned 1 point in the risk score per 0.25 in the rounded coefficient (see Table E2 in the data supplement). Subgroup analyses were used to evaluate whether results differed by sex, smoking status, and GOLD grade (Figure E1). For a more detailed description of the development of this model, please see the Appendix E1. To assess the sensitivity of the risk scores and their predictive accuracy to our method for identifying risk factors and constructing the scores, we also fit LASSO Poisson regression models on the same candidate variables using cross validation for LASSO parameter tuning and unbiased assessment of predictive accuracy (Table E3), which resulted in similar coefficients and accuracy (20). Predictive (discriminative) accuracy was measured in terms of the concordance index.

Results

Description of Participants

A total of 1,528 COPDGene participants with airflow obstruction on spirometry, no exacerbation in the year before their second site visit, and who completed at least one longitudinal follow-up questionnaire in the subsequent 36 months after their second site visit were included in the final analysis (Figure 1). The mean age was 69 years (standard deviation [SD], 8 yr). Most were male (58.1%), white (80.7%), and former tobacco users (66.3%) (Table 1). The majority of participants (n = 856, 56.0%) were not on any inhaled medications at the time of their second study site visit. Most participants had GOLD grade 2 COPD (46.7%). The mean FEV₁% predicted was 65.5% (SD, 22%) and mean CAT score was 12.3 (SD, 7.9). Median longitudinal follow-up was 32 months and 4.5 questionnaires. A total of 69.3% of participants had their last follow-up contact within the last 6 months of the 36-month period studied. A total of 508 participants (33.2%) had at least one moderate or severe exacerbation during the period studied, with most experiencing one exacerbation (n = 222, 43.7%). Exacerbation frequency ranged from 0 (n = 1,020, 66.8%) to ⩾10 (n = 13, 0.9%).

Figure 1.

Participant selection. COPDGene = Genetic Epidemiology of COPD Study; FEV₁ = forced expiratory volume in 1 second; FVC = forced vital capacity.

Table 1.

Characteristics of participants at second site visit

	All Participants (N = 1,528)
Age, yr	69 ± 8.0
Female	640 (41.9)
White	1,233 (80.7)
BMI, kg/m2	27.9 ± 5.9
Active tobacco users	515 (33.7)
Pack-years	50.6 ± 25.3
FEV1% predicted	65.5 ± 22.0
GOLD grade
1	418 (27.4)
2	714 (46.7)
3	312 (20.4)
4	84 (5.5)
CAT score
<10	654 (42.8)
⩾10	874 (57.2)
mMRC score
0–1	897 (58.7)
2	217 (14.2)
3–4	414 (27.1)
Chronic bronchitis	289 (18.9)
GERD	493 (32.3)
Coronary artery disease	172 (11.3)
Congestive heart failure	58 (3.8)
Eosinophil count
<100 cells/μl	235 (15.4)
100–299 cells/μl	957 (62.6)
⩾300 cells/μl	306 (20)
Missing	30 (2)
Inhaled medications
None	689 (45.1)
Short-acting inhalers*	167 (10.9)
LAMA alone	144 (9.4)
ICS alone	38 (2.5)
LABA alone	13 (0.9)
LABA + LAMA	24 (1.6)
LABA + ICS	205 (13.4)
LAMA + ICS	14 (0.9)
LABA + LAMA + ICS	234 (15.3)

Definition of abbreviations: BMI = body mass index; CAT = Chronic Obstructive Pulmonary Disease Assessment Test; FEV₁ = forced expiratory volume in 1 second; GERD = gastroesophageal reflux disease; GOLD = Global Initiative for Chronic Obstructive Lung Disease; ICS = inhaled corticosteroid; LABA = long-acting β agonist; LAMA = long-acting muscarinic antagonist; mMRC = Modified Medical Research Council.

Continuous variables are presented as means ± standard deviation; categorical variables are presented as n (%).

^*Short-acting inhalers include short-acting β-agonists, short-acting muscarinic antagonists, and combined short-acting β-agonists and muscarinic antagonists.

Readily Available Risk Factors

We first focused on clinical risk factors that can be easily obtained in a variety of clinical settings. Self-reported history of GERD (rate ratio [RR], 1.26; 95% confidence interval [CI], 1.02–1.56; P = 0.03), self-reported history of chronic bronchitis (RR, 1.64; 95% CI, 1.30–2.08; P < 0.001), and mMRC ⩾ 2 (RR, 1.99; 95% CI, 1.62–2.46; P < 0.001) were associated with an increased rate of moderate or severe exacerbations in univariable analyses. All three remained significant in a multivariable model controlling for age, sex, and smoking status (GERD: RR, 1.30; 95% CI, 1.05–1.60; P = 0.01; chronic bronchitis: RR, 1.62; 95% CI, 1.27–2.05; P < 0.001; mMRC = 2: RR, 1.39; 95% CI, 1.04–1.88; P = 0.03; mMRC > 2: RR, 2.16; 95% CI, 1.72–2.72; P < 0.001).

Advanced Risk Factors

Next, we evaluated factors that may not be routinely obtained in primary care but are more frequently collected in secondary or tertiary care settings. A CAT score ⩾ 10 (RR, 2.28; 95% CI, 1.83–2.86; P < 0.001) and lower FEV₁% predicted (RR, 1.40 per −20%; 95% CI, 1.22–1.60; P < 0.001) were both associated with increased rates of moderate or severe exacerbations in univariable analyses. Both remained significant in a multivariable model controlling for age, sex, mMRC, GERD, chronic bronchitis, and smoking status (CAT ⩾ 10: RR, 1.59; 95% CI, 1.25–2.05; P < 0.001; FEV₁% predicted per −20%: RR, 1.25; 95% CI, 1.08–1.43; P = 0.002).

Graduated Model

We incorporated the above risk factors into a graduated risk score to predict the risk of moderate or severe exacerbations starting with readily available risk factors (mMRC, chronic bronchitis, GERD) and adding more time- or resource-intensive variables (CAT and spirometry) in a stepwise fashion (Table 2). All models were adjusted for age, sex, and smoking status. The first model incorporated clinical variables believed to be available in most treatment settings: mMRC (score: 2 and score: 3 or 4), chronic bronchitis, and GERD (risk score 1 in Table 2 and Figure 2A). A higher score in this model was associated with an increased risk of exacerbations (RR, 1.37; 95% CI, 1.29–1.45; P < 0.001), such that a 1-point increase was associated with a 29% higher rate of exacerbations in the subsequent 36 months.

Table 2.

Risk scores for COPD exacerbations

Predictor	Risk Score 1	Risk Score 2	Risk Score 3
mMRC score 2	1	1	1
mMRC score 3–4	3	2	2
Chronic bronchitis	2	2	2
GERD	1	1	1
CAT score ⩾ 10	—	2	2
FEV1% predicted 60–79%	—	—	1
FEV1% predicted 40–59%	—	—	2
FEV1% predicted <40%	—	—	3
Total possible points	6	7	10
Rate ratio for any exacerbation per 1-point increase (95% confidence interval)	1.37 (1.29–1.45)	1.34 (1.27–1.41)	1.30 (1.25–1.36)
Concordance with observed exacerbation rate	0.64	0.65	0.67

Definition of abbreviations: CAT = Chronic Obstructive Pulmonary Disease Assessment Test; COPD = chronic obstructive pulmonary disease; FEV₁ = forced expiratory volume in 1 second; GERD = gastroesophageal reflux disease; mMRC = Modified Medical Research Council.

Figure 2.

(A) Model 1 estimated exacerbation rate by risk score. (B) Model 2 estimated exacerbation rate by risk score. (C) Model 3 estimated exacerbation rate by risk score.

Next, we added CAT score ⩾ 10 to the model (risk score 2 in Table 2 and Figure 2B). A higher score in this model was also associated with a higher risk of exacerbations (RR per 1-point score increase, 1.34; 95% CI, 1.27–1.41; P < 0.001).

Last, we added FEV₁% predicted, categorized for ease of calculation, to the model (risk score 3 in Table 2 and Figure 2C). Similarly, a higher score in this model was associated with a higher risk of exacerbations (RR per 1-point score increase, 1.30; 95% CI, 1.25–1.36; P < 0.001).

As mMRC ⩾ 2, GERD, and chronic bronchitis were included in all three models, we analyzed these predictors among subgroups based on sex, smoking status, and GOLD grade (Figure E2). The importance of these predictors was consistent across nearly all subgroups; exceptions included participants with mMRC score ⩾ 2 and GOLD grade 1 COPD, those with chronic bronchitis and GOLD grade 1 or GOLD grade 3 or 4, and GERD among males, current tobacco users, and those with GOLD grade 2 COPD.

We performed two sensitivity analyses as part of this work. Moderate and severe exacerbations were combined in our initial risk scores. We subsequently evaluated the risk scores’ ability to predict severe exacerbations alone. The risk score again performed well with similar magnitude of effect and remained statistically significant (Figures E3A–E3C; risk score 1: RR per 1-point score increase, 1.43; 95% CI, 1.32–1.55; P < 0.001; risk score 2: RR, 1.40; 95% CI, 1.31–1.51; P < 0.001; risk score 3: RR, 1.34; 95% CI, 1.26–1.42; P < 0.001). As most participants were not using an inhaled medication at baseline, and long-acting inhaler use could have a protective effect against exacerbations, we also performed a sensitivity analysis adjusting for inhaler use among participants. The magnitude of effect again remained similar, and the risk scores remained statistically significant (Figures E4A–E4C; risk score 1: RR, 1.27; 95% CI, 1.20–1.35; P < 0.001; risk score 2: RR, 1.25; 95% CI, 1.19–1.32; P < 0.001; risk score 3: RR, 1.23; 95% CI, 1.18–1.29; P < 0.001).

Discussion

COPD exacerbations are detrimental events in the history of patients with COPD and are associated with increased morbidity and mortality. Previous studies evaluating risk factors associated with exacerbations have identified a history of recent exacerbation to be a particularly strong predictor of future exacerbations. The risk factors identified for future exacerbations in this study, namely GERD, chronic bronchitis, high symptom burden, and reduced lung function, are similar to those identified in previous studies. However, what makes this work unique is that we specifically excluded participants with a history of recent exacerbations. As such, we have shown that these variables are also applicable to a group of patients who may not have been fully represented in previous research. It is also interesting to note that one-third of participants had an exacerbation in the 36 months studied, despite a lack of recent exacerbations, which suggests that patients without a recent history remain at risk of exacerbation despite the absence of recent exacerbations.

We incorporated these risk factors into a graduated clinical risk score, which could be used in a variety of resource settings to care for patients with COPD. The variables listed in our first risk score, mMRC, history of chronic bronchitis, and history of GERD, are easily obtainable in the setting of a routine clinic visit and do not require a significant amount of time or resources. The next risk score also included CAT score. Although the CAT—a survey containing eight questions that the patient scores 0 to 5—can be somewhat time intensive, it has been translated and validated in more than 50 languages and is free to use in clinical practice (21). The CAT score encompasses questions about dyspnea and chronic cough and phlegm production but also inquires about a variety of other health markers (e.g., quality of sleep, energy level, and confidence leaving the home).

Our last risk score includes FEV₁% predicted as measured through spirometry. Airflow obstruction defined as FEV₁/FVC < 0.70 documented at least once is required for the diagnosis of COPD. Spirometry is a relatively time- and resource-intensive test, which is not obtained during routine primary care visits but is obtained in conjunction with many routine visits in pulmonary subspecialty clinics. Although all patients with a diagnosis of COPD need to have documented airflow obstruction, they may not have a recent FEV₁ if their COPD is managed by a primary care provider. As such, we considered a recent FEV₁% predicted to be a more advanced piece of data, which may not be available in all settings.

In all three risk scores, each 1-point increase was associated with a 25–30% increase in the rate of exacerbation over the subsequent 36 months. We noted a consistent dose–response relationship such that there was an increase in exacerbation risk with increasing score (Table E4). The addition of more complex clinical variables to risk scores led to a larger gradation in exacerbation rate. Subgroup analyses of the three models showed that all three continued to perform well when participants were grouped by sex, tobacco use, and GOLD grade (Figure E1).

As this is an analysis of a cohort and not a prospective study, it is not fully known if these findings can be extrapolated to a wider population. However, the risk factors identified could be potentially modifiable within the course of 36 months. Given the association with GERD, it would be reasonable to consider pharmacologic and nonpharmacologic interventions in patients with comorbid GERD and COPD. Previous research has shown a reduction in exacerbation rate among patients with COPD who are started on a proton pump inhibitor as part of a randomized controlled trial (22, 23). Reducing baseline symptom burden may be another method of reducing exacerbation risk. Many participants in this analysis were not on inhaled medications, which can reduce both symptom burden and exacerbation risk. Initiation of inhaled medication, change of current inhalers, or referral to pulmonary rehabilitation could be considered in those with a high baseline symptom burden. Similar interventions could be considered in patients with low lung function. Although our risk scores only show those at risk of exacerbation over a 36-month period, it does seem reasonable that these interventions could be implemented in that time frame and may confer some benefit beyond this time period. In addition, an increasing score in all risk scores resulted in higher exacerbations per year (Table E4). This information could be used for patient education or management, as a 1-year risk of exacerbation could result in more motivation for intervention.

Limitations

There are some limitations associated with our analysis. First, this is a group from a cohort study and is not a representative sample of the entire population. Previous research has shown socioeconomic status has an impact on symptoms and lung function (24). It is possible these could have affected exacerbation rate but were not specifically evaluated.

This cohort is also limited because all participants had at least a 10 pack-year smoking history and had COPD defined as an FEV₁/FVC < 0.70. Although tobacco smoke exposure is the most important risk factor for the development of COPD, population studies have shown that close to 30% of those with COPD have no tobacco smoke exposure (25). It is unknown if those with a history of COPD and no tobacco smoke exposure would have similar risk factors to those identified in this analysis. Similarly, those included in this cohort had documented airflow obstruction, and so these results should be interpreted with caution in those without FEV₁/FVC < 0.70. Tobacco-exposed individuals are often treated with bronchodilators, although there is not currently evidence that this improves symptoms or spirometry (26). Although the risk factors described here identify those at increased risk for exacerbation within the cohort studied, it is unclear if they would also identify those at risk of exacerbation without documented airflow obstruction. Given the risk factors identified and possible interventions suggested above, a physician could consider implementing similar interventions (e.g., initiation of inhaled medications, change of current inhaled medications, treatment for GERD, etc.) in those without known airflow obstruction, although it is important to note that this population would not have been included in this study or previous research regarding COPD, and it is currently unclear how best to manage these patients.

Another potential limitation is that only 55% of participants were on inhaled medications. The language of the survey used for COPDGene emphasizes medication use as opposed to what may be prescribed for the patient. Use of and adherence to inhaled medications among patients with COPD is known to be often poor and variable. Previous research has shown adherence rates with long-acting inhalers to range anywhere from 8.8% to 54%, depending on the cohort and the type of inhaled medication (27). As a result, we believed that this cohort was representative of a real-world sample and that controlling for medication use would potentially limit its generalizability. A sensitivity analysis showed that the risk scores remained significant while controlling for baseline maintenance inhalers.

The use of self-reported exacerbations as the outcome could be a limitation, although this schema was designed and validated by the COPDGene investigators to accurately record this outcome (18). Our definition of COPD exacerbation does not differentiate between participants treated with steroids or antibiotics, so it is possible some of these patients were actually treated for pneumonia. It also does not differentiate between those who required hospitalization and those who were treated as outpatients. Last, the risk scores only cover 36 months, and this is a retrospective study, but we do believe that risk factors identified could be modified within a 3-year period.

Conclusions

In patients with COPD but without a recent history of exacerbations, risk factors such as GERD, chronic bronchitis, high symptom burden, and lower lung function are associated with an increased risk of future exacerbation. These risk factors can be collated into a simple risk score, which can be used in clinical practice.