Published ahead of print on January 26, 2006, doi:10.1164/rccm.200508-1344OC
© 2006 American Thoracic Society doi: 10.1164/rccm.200508-1344OC
Lung Function Decline and Outcomes in an Adult PopulationDivision of Pulmonary and Critical Care Medicine, University of Kentucky Medical Center, Lexington, Kentucky; and GlaxoSmithKline Research and Development, Research Triangle Park, North Carolina Correspondence and requests for reprints should be addressed to David M. Mannino, M.D., Division of Pulmonary and Critical Care Medicine, University of Kentucky Medical Center, 800 Rose Street, MN 614, Lexington, KY 40536. E-mail: dmannino{at}uky.edu
Rationale: Chronic obstructive pulmonary disease (COPD) is an important cause of morbidity and mortality. Objectives: To determine risk factors for and outcomes of rapid lung function decline in a cohort of adults in the United States. Methods: We analyzed data from 15,536 adults aged 44–66 yr in the Atherosclerosis Risk in Communities study. We used Cox proportional hazard models to determine the risk of rapid lung function decline at 3 yr on mortality and COPD hospitalizations over the subsequent 8 yr. Measurements and Main Results: Of those in the baseline cohort, 13,756 (88.5%) had spirometry at the Year 3 visit. The strongest risk factors for not having a follow-up spirometry were as follows: having Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage 3 or 4 disease at baseline (adjusted odds ratio [OR] 2.8; 95% confidence interval [CI], 2.1–3.8), being black (adjusted OR, 2.4; 95% CI, 2.1–2.7), and being a current smoker (adjusted OR, 1.8; 95% CI, 1.5–2.0). Participants with GOLD stage 3 or 4 disease were also more likely to be in the most rapidly declining lung function quartile (adjusted OR, 3.7; 95% CI, 2.7–5.0). Overall, participants with the most rapidly declining lung function had a modestly increased risk of death (adjusted hazard ratio, 1.4; 95% CI, 1.2–1.7) and time to a COPD-related hospitalization (adjusted hazard ratio, 1.4; 95% CI, 1.2–1.8). Conclusion: Rapid lung function decline was independently associated with a modest increased risk of COPD hospitalizations and deaths.
Key Words: chronic obstructive pulmonary disease • lung function • mortality Chronic obstructive pulmonary disease (COPD) is an important cause of morbidity and mortality in the United States and worldwide (1–3). The magnitude of this disease spurred the Global Initiative for Chronic Obstructive Lung Disease (GOLD) in 2001 (4). From this initiative came a standardized staging system for COPD based on FEV1 and the FEV1/FVC ratio (5). The purpose of this staging system was to improve investigation and management of this complex disease. Many other parameters, such as patient age, resting oxygen saturation, body mass index, symptoms, and exercise capacity have also been assessed for their values as predictors of morbidity or mortality in COPD (6–9). Because lung function testing has been the standard for diagnosis and staging of COPD, other indices are often evaluated in comparison to FEV1 (10, 11). Previously, dyspnea symptoms and 6-min walk distance were both studied and found to be superior to FEV1 in predicting mortality (12–14). Recently, the body mass, obstruction, dyspnea, and exercise (BODE) index was evaluated by Celli and colleagues and found to be of greater predictive value than FEV1 alone (6). The original use of FEV1 decline rate as a marker of chronic respiratory disease progression was developed by Fletcher and coworkers nearly three decades ago (15). More recently, publications by Anthonisen and colleagues and Burrows and colleagues addressed prognostic factors in COPD, including age, FEV1, dyspnea symptoms, poor bronchodilator response, functional reserve capacity, exercise tolerance, exercise heart rate, continued smoking, and decline in FEV1 (16, 17). More contemporary work has examined additional factors, such as asthma, diet, and comorbid disease, which are important in the decline of lung function (18–20). Our purpose in this study was to evaluate lung function decline as a predictor of morbidity and mortality in a population both with and without COPD. We used data from 15,536 people involved in the Atherosclerosis Risk in Communities (ARIC) study. These participants underwent baseline spirometry, which was repeated at 3 yr, the second visit in this study. The cohort was followed for 8 yr after the second spirometry assessment. Data from this study have previously been used to evaluate the GOLD criteria's predictive value for morbidity and mortality (21). In this analysis, we hypothesized that lung function decline is an independent predictor of morbidity and mortality and that its predictive value is superior to FEV1 alone.
Study Population The ARIC study was initiated in 1986 as a longitudinal, population-based study of the etiology and clinical sequelae of atherosclerosis. Study protocols were approved for protection of human subjects. Details of the ARIC study are published elsewhere (22). Our analysis was limited to ARIC participants ages 44 to 66 yr, who provided baseline information on respiratory symptoms and diagnoses, who underwent pulmonary function testing at the baseline examination, and for whom follow-up data were available (n = 15,536).
Pulmonary Function Data Rapid decliners in lung function were determined on the basis of lung function values at the baseline evaluation and at the first follow-up examination at approximately 3 yr. We determined the quartiles of the change (determined as an annualized percent change from the baseline level) in the FEV1 and classified those in the most negative category as "rapid decliners" and the other three categories as the referent group.
Variable Definition
Analysis Our primary outcome of interest in the two separate Cox proportional hazard regression models was time to death or the time to the first COPD-related hospitalization that occurred during the 8 yr of follow-up. For deaths, the exit date was the date of death reported on the death certificate, and, for survivors, the exit date was the date the participant was last known to be alive. For COPD-related hospitalizations, the exit date was either the first COPD-related hospitalization, death, or the date the person was last known to be alive. Age, sex, race, smoking status, education level, body mass index, heart disease, education level, respiratory symptoms at the second visit, and physical activity level were included in the adjusted models. Analyses were also done on subsets of subjects stratified by lung function and stratified by sex. An expansion of the methods is available on the online supplement.
Of the original 15,792 subjects in the ARIC cohort, 146 did not have adequate pulmonary function testing at the baseline examination and another 110 were missing data on key variables in our analysis, leaving 15,536 (98.4%) subjects available for analysis. The demographic characteristics of the population are displayed in Table 1. Overall, 281 (1.8%) of subjects were in GOLD stage 3 or 4 at baseline, and an additional 1,486 (9.6%) were in GOLD stage 2. Spirometry at the follow-up visit was available on 13,756 (88.5%) of subjects in the cohort. The strongest predictors of not having a spirometry at 3 yr included having GOLD stage 3 or 4 COPD at baseline (odds ratio [OR], 2.8; 95% confidence interval [CI], 2.1–3.8), being black (OR, 2.4; 95% CI, 2.1–2.7), and being a current smoker at baseline (OR, 1.8; 95% CI 1.5–2.0). Data from all of the variables used in the analysis are available in the online supplement.
Mean annual loss of FEV1 in the overall cohort was 62 ml. The mean loss of FEV1 as a percentage of the baseline value was 1.5% annually. Participants in the most rapidly declining quartile of FEV1 had a mean annual loss of 171 ml, which was 4.7% of the baseline level per year (Table 2). All of the subjects in this group lost at least 2.7% of their baseline FEV1 annually. The strongest risk factors for being in the rapidly declining quartile of FEV1 included having GOLD stage 3 or 4 COPD at baseline (OR, 3.7; 95% CI 2.7–5.0) and having GOLD stage 2 COPD at baseline (OR, 2.6; 95% CI, 2.3–3.0), although being 50 or older, female, and of black race were also significant (Table 1).
Overall, rapid decline of lung function was a predictor of death (hazard ratio [HR], 1.4. 95% CI, 1.2–1.7; Table 3). In the unadjusted calculation of deaths per person-year, people who were classified as GOLD stage 3 or 4 at baseline who were "rapid decliners" died at a rate of 45.0 per 1,000 person-years, 10 times higher than those with normal lung function at baseline without rapid lung function decline, with a mortality rate of 4.5 per 1,000 person-years. Of the 720 deaths in the cohort, 251 (35%) were from cardiovascular disease, 95 (13%) were from lung cancer, and 24 (3%) were from COPD. A total of 273 (38%) of the deaths were in the 25% of the subjects we classified as "rapid decliners."
In the stratified analyses that adjusted for covariates, the risk of rapid lung function decline for death varied by baseline lung function, and the effect was strongest in subjects who were "restricted" at baseline (HR, 1.9; 95% CI, 1.2–3.0; Table 3). In the overall model examining the interaction between baseline level of lung function and rapid lung function decline (Figure 1), an effect of both baseline lung function and lung function decline could be seen. Results in the overall cohort, stratified by sex showed a large overlap between men (HR, 1.3; 95% CI, 1.1–1.7) and women (HR, 1.5; 95% CI, 1.2–1.9).
Similarly, rapid decline of lung function was a predictor of time to COPD-related hospitalization (HR, 1.4; 95% CI, 1.2–1.8; Table 4). In the unadjusted calculation of COPD-related hospitalizations per person-year, people who were classified as GOLD stage 3 or 4 at baseline who were rapid decliners were hospitalized at a rate of 62.9 per 1,000 person-years, more than 40 times higher than those with normal lung function at baseline without rapid lung function decline, with a rate of 1.4 per 1,000 person-years. In the stratified analyses that adjusted for covariates, the risk of rapid lung function decline for COPD-related hospitalization was strongest in subjects who were GOLD stage 0 at baseline (HR, 1.9; 95% CI, 1.2–3.0; Table 4, Figure 2). Results in the overall cohort, stratified by sex showed a large overlap between men (HR, 1.4; 95% CI, 1.03–1.9) and women (HR, 1.7; 95% CI, 1.2–2.3).
In this analysis of data from more than 15,000 adults in the United States, we found that FEV1 decline was a significant predictor of both COPD-related hospitalizations and all-cause deaths. However, the predictive value of being a rapid decliner over 3 yr, as defined in this study, varied by a person's baseline lung function and was influenced by missing data from people in whom we could not determine lung function decline. An interesting finding was that having a more rapid FEV1 decline was a stronger predictor of death or COPD-related hospitalizations among subjects who had GOLD stage 0 disease or were restricted at baseline than in people with more severe disease (Figures 1 and 2, Tables 3 and 4). Prior analysis of the ARIC cohort revealed that all GOLD stages are associated with increased risk of morbidity and mortality (21). In the baseline cohort, almost half (46%) of the 15,792 subjects were either GOLD stages 0 through 4 or had restrictive spirometry. Of this group, about 75% of the subjects were stage 0 or 1, or restrictive at baseline (21). Thus, these groups represent a large population that would likely benefit from further risk reduction. Addition of other indices such as body mass index and exercise capacity furthers the risk assessment by accounting for systemic manifestations of COPD as evidenced by Celli and colleagues (6). Our data suggest that rapid decline in FEV1 provides important information about COPD because it directly addresses disease progression, especially in those patients with either restrictive spirometry or GOLD stage 0 COPD. The mean loss of lung function in our cohort was 62 ml of FEV1 annually. This is higher than mean losses reported in other cohorts of 26 ml annually in the Honolulu Heart Cohort (28), 30 to 40 ml annually in the Busselton Health Study (20), 38 ml in the Nottingham study (19), and 22 to 38 ml annually in the Copenhagen City Heart Study (18). This difference is probably related to several factors, including the shorter interval between spirometry and the presence of only two measurements in the ARIC study. This analysis also demonstrated unique features of subjects classified as having restrictive spirometry. For example, people classified as "restrictive" at baseline were less likely to be rapid decliners (Table 1), yet people with restriction who were rapid decliners had one of the highest risks for death and COPD-related hospitalizations (Tables 3 and 4, Figures 1 and 2). This is probably related to the many comorbid conditions, such as diabetes, obesity, and congestive heart failure, present in people with restriction on spirometry (29). One possible explanation for this finding is that worsening lung function in people with restriction is mainly driven by a worsening of the underlying comorbid condition that then results in worse outcomes.
This analysis had certain limitations. The primary limitation was that spirometry was measured at only two points. It is possible that people may have had a really good day or really bad day at either the baseline or follow-up examination, influencing our results. Our main way of dealing with this limitation was the use of two categories to classify participants with regard to lung function decline. Thus, a person who may have, falsely, had a very rapid decline, would still just be classified as a "rapid decliner" as opposed to a "very rapid decliner." Somewhat reassuring is that, with this approach, people who one would expect to have a more rapid decline (i.e., the elderly, current smokers, people with impaired lung function) did. Second, ARIC participants underwent prebronchodilator studies only. This may have affected the staging of participants in that those with reversible disease could have a falsely low FEV1 based on their prebronchodilator lung function. Two prior studies used post-bronchodilator FEV1 for this reason (16, 17). Also, treatment was not taken into account in this study. Pharmacologic treatment for COPD often bases its efficacy on the ability to influence FEV1. Although there are contradictory studies, inhaled steroids, Another potential source of error is survival bias. For patients to be assessed for degree of decline in FEV1, they must survive and be healthy enough to attend the 3-yr follow-up and undergo spirometry. Thus participants with the most rapidly declining lung function may be underrepresented in this study because of the inability to obtain follow-up spirometry. However, because this source of error underestimates the predictive value of rapid decline in FEV1, it would bias the study against finding an effect. Of note, in both the Fletcher and Peto study (15) and the Busselton study (20), a large proportion of subjects were excluded from the analytic database because of missing data, either because the subjects died before follow-up or were too ill to participate and undergo spirometry. Interestingly, the groups in which rapid decline in FEV1 showed the greatest predictive value for death and hospitalization were also those least likely to be affected by the sources of error listed previously. These subsets were subjects with GOLD stage 0 and those with restrictive disease. Patients with falsely elevated GOLD staging based on lack of post-bronchodilator spirometry would be less likely to be included in these groups. Patients in these groups would also be less likely to be treated with medication overall. The GOLD recommendations for treatment in stage 0 and 1 are smoking cessation and possibly short-acting bronchodilator as needed. There is no recommendation for corticosteroids or anticholinergic agents in these stages. Therefore, separation based on medical care would presumably be least in these groups. In addition, because survival was highest these groups, survival bias should be least in these subsets. In the study by Anthonisen and colleagues, correlates with FEV1 were sought only in the least impaired group to minimize the effects of survival bias (16).
Conclusions
The authors thank the staff and participants in the Atherosclerosis Risk in Communities study for their important contributions. The authors also thank Dana Hazen, Kara Guiliani Cecil, and Rebecca Copeland for their valuable contributions to this work.
Supported by a research grant from GlaxoSmithKline. The ARIC study is conducted and supported by the National Heart Lung and Blood Institute in collaboration with the ARIC Investigators. This article was not prepared in collaboration with investigators of the ARIC and does not necessarily reflect the opinions or views of the ARIC or the NHLBI. This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Originally Published in Press as DOI: 10.1164/rccm.200508-1344OC on January 26, 2006 Conflict of Interest Statement: D.M.M. serves on advisory boards for Boehringer Ingelheim, GlaxoSmithKline (GSK), and Ortho Biotech; he is on the speaker's bureau for Boehringer Ingelheim, Pfizer, and Dey, and has received grants from GSK and Pfizer. M.M.R. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. K.J.D. is a current employee of GSK R&D and owns GSK stock options. Received in original form August 30, 2005; accepted in final form January 26, 2006
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
70% and FVC < 80% predicted). We defined a subject as having a respiratory symptom if he or she reported cough, phlegm, dyspnea, or wheeze. 
2 agonists, and anticholinergic agents have all been shown to have this effect to some degree (
