INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX: STUDY PARTICIPANTS REFERENCES

Patients with chronic obstructive pulmonary disease (COPD) frequently develop exacerbations that require unscheduled physician visits or hospitalization. Hospital care for these patients is expensive and long-term outcomes are relatively poor. For patients with severe disease the median length of hospitalization is 9 d and the median cost is estimated at $7,100 (1).

Infections are thought to trigger most COPD exacerbations (2), but other factors may also play a role. Bronchial inflammation, bronchospasm, and retained purulent secretions contribute to worsening airflow obstruction, which may be so severe as to cause acute respiratory failure. FEV₁ is a widely used, objective measure of airflow obstruction that correlates with clinical outcomes in COPD (3). Despite its wide availability, there are few studies in which FEV₁ has been measured systematically and prospectively during COPD exacerbations (7).

We previously reported the principal results of a large, multicenter trial that evaluated the effects of systemic corticosteroids on clinical outcomes in patients hospitalized for COPD exacerbations (11). As part of the study protocol, we measured FEV₁ serially in all subjects who were capable of performing the test. We found that systemic corticosteroids improve both clinical outcomes and FEV₁ to a modest degree.

In this communication we report further results of this study, with attention being paid to two specific questions. First, are FEV₁ measurements upon admission to the hospital and during the first few days of treatment predictive of clinical outcomes? Second, are baseline characteristics and treatments predictive of FEV₁ at entry and on subsequent study days?

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX: STUDY PARTICIPANTS REFERENCES

This trial was sponsored by the Veterans Affairs Cooperative Studies Program. Its Human Rights Committee and the institutional review board of all participating medical centers approved this study. All subjects gave written informed consent.

Details of the study protocol were described in previous publications (11, 12). Briefly, all patients admitted with a primary diagnosis of COPD to participating Veterans Affairs Medical Centers between November, 1994, and November, 1996, were candidates for the trial. Major inclusion criteria were a clinical diagnosis of COPD exacerbation, age of 50 yr or older, smoking history of 30 pack-yrs or more, and either an FEV₁ of 1,500 ml or less or inability to perform spirometry owing to dyspnea. Principal exclusion criteria were a clinical diagnosis of asthma, use of systemic glucocorticoids within the preceding 30 d, co-existing medical conditions making survival for 1 yr unlikely, and inability to give informed consent. We collected baseline data concerning respiratory diseases and other medical history by questionnaire (13).

We randomized eligible subjects to one of three treatment arms. The 8-wk corticosteroid arm received intravenous methylprednisolone (Solu-Medrol; Pharmacia and Upjohn, Bridgewater, NJ) (125 mg every 6 h for 72 h) followed by oral prednisone (60 mg on Study Days 4 to 7; 40 mg on Days 8 to 11; 20 mg on Days 12 to 43; 10 mg on Days 44 to 50; 5 mg on Days 51 to 57) taken once daily. The 2-wk corticosteroid arm was identical to 8-wk corticosteroid arm except that subjects received placebo capsules on Study Days 16 to 57. The placebo arm received an equal volume of intravenous 5% dextrose solution every 6 h for 72 h followed by placebo capsules on Study Days 4 to 57. All subjects received a broad spectrum antibiotic for at least 7 d. All subjects also received an inhaled -adrenergic agonist (two puffs from a metered dose inhaler or a nebulizer treatment at least four times daily), inhaled ipratropium bromide (two puffs from a metered dose inhaler or a nebulizer treatment at least 4 times daily), and, starting on Study Day 4, inhaled triamcinolone acetonide (eight puffs, total of 800 µg, daily in divided doses), or its equivalent, for the entire 6-mo period. Subjects were told to not take theophylline, high-dose inhaled corticosteroids (> 8 puffs of triamcinolone acetonide, or its equivalent), or open-label systemic corticosteroids for the 6-mo duration of the study, unless it was prescribed by their physicians.

First treatment failure was the primary study outcome. This was defined as death from any cause, intubation and mechanical ventilation, hospital readmission with a principal diagnosis of COPD, or intensification of pharmacologic therapy. We defined intensification of pharmacologic therapy as prescription of open-label systemic corticosteroids, of high-dose inhaled corticosteroids (> 8 puffs per day of triamcinolone acetonide, or its equivalent), or of theophylline. Primary physicians were instructed to prescribe these medications only if they made a judgment that the subject was not making satisfactory clinical improvement.

Study personnel evaluated all subjects at baseline, at Study Days 1, 2, and 3 and at Weeks 2, 8, and 26. If the subject was able to cooperate, FEV₁ was measured at each visit. All centers used a standard instrument (Model 922; SensorMedics, Yorba Linda, CA) and the test was performed according to American Thoracic Society recommendations (14). Two puffs of a -adrenergic agonist were administered from a metered dose inhaler before measurement of the FEV₁, and the best of a minimum of three efforts was used for analyses.

Although follow-up extended for 6 mo after enrollment, we restrict our analyses to the first 30 d in this communication. Maximal FEV₁ improvement occurred within 2 wk after enrollment and approximately one-half of all treatment failures occurred within the first 30 d (11). We have combined the 2-wk and 8-wk corticosteroid arms into a single active treatment group, because we found no significant differences between these two groups for any treatment outcome.

Descriptive statistics were determined for baseline variables. Chi-square was used for simple two-way comparisons. The primary outcome was time to first event (death, intubation and mechanical ventilation, hospital readmission for COPD, or intensification of pharmacologic therapy) in the first 30 d after randomization. Relative risks and confidence intervals (CI) were calculated by Cox proportional hazards regression. Univariate regressions analyses were done to explore potential risk factors. Multivariate regression analyses were also done to examine risk factors adjusted for potential confounding covariates. Multiple linear regression was performed to determine the relationship between baseline FEV₁, and the change in FEV₁ at Day 2 to selected baseline variables. All analyses were run using SAS software (version 6.12; Cary, NC).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX: STUDY PARTICIPANTS REFERENCES

The trial enrolled 271 subjects from 1,840 candidates who were screened at 25 Veterans Affairs Medical Centers. Eighty subjects were randomized to the 8-wk corticosteroid arm, 80 subjects to the 2-wk corticosteroid arm, and 111 subjects to the placebo arm. The subjects were predominantly male (268/3) with an average age of 68 yr and a mean lifetime cigarette consumption of 75 pack-years.

FEV₁ values from entry through Study Day 14 are shown in Table 1. Over the first 14 study days at least one FEV₁ value was obtained in 261 subjects. Including all available data, the FEV₁ increased from 767 ± 280 ml at entry to 927 ± 393 on Study Day 3 and to 1,056 ± 427 ml on Day 14 (Table 1). Entry FEV₁ did not predict subsequent changes; the correlation coefficients between entry FEV₁ and FEV₁ on Study Days 1, 2, 3, and 14 were 0.10, 0.01, 0.03, and 0.01, respectively.

Sixty-four of the 261 subjects (25%) included in these analyses experienced at least one treatment failure by 30 d. Therapy intensification comprised 70% of all treatment failures at 30 d and most of these subjects received open-label systemic corticosteroids. Hospital readmission (15%), mechanical ventilation (8%), and death (7%) occurred less commonly.

To determine their association with clinical outcome, we subjected FEV₁ data and selected baseline characteristics to univariate Cox analysis with treatment failure at 30 d as the dependent variable. In addition to entry FEV₁, we evaluated FEV₁ on Study Days 1, 2, 3, and 14, the largest FEV₁ over the first 3 d, and the last measured FEV₁ over the first 3 d. Baseline characteristics included in the analyses (mean or prevalence given in parentheses) were age (68 yr), current smoking (51%), history of chronic cough (55%), history of chronic phlegm (55%), history of frequent wheezing (89%), history of at least one chest cold in previous year (82%), inhaled -adrenergic agonist use at entry (86%), inhaled anticholinergic use at entry (69%), theophylline use at entry (34%), inhaled corticosteroid use at entry (47%), hospitalization for COPD in previous 2 yr (68%), and randomization to systemic corticosteroid treatment arm (59%).

We found that changes in the FEV₁ over the first 3 d were closely interrelated and all were highly predictive of treatment failure. For sake of simplicity we report only Day 2 FEV₁. Only six of the 64 treatment failures included in this analysis occurred before Study Day 2. FEV₁ at Study Day 14 was not associated with treatment failure at a significant level, and that may be due in part to corruption of the data. With increasing time on study there were more missing values and a substantially greater likelihood for missing values to occur in subjects who had already experienced a treatment failure.

Entry FEV₁, Day 2 FEV₁, theophylline use at entry, inhaled anticholinergic use at entry, and study treatment assignment were related to treatment failure with p < 0.10 in a univariate Cox model (Table 2). We included those five variables in a multivariate Cox model and the results of that analysis are shown in Table 3. Both entry FEV₁ and Day 2 FEV₁ independently predicted 30-d treatment failure at a highly significant level. An entry FEV₁ that was larger by 100 ml was associated with a 13% reduction in relative risk for treatment failure (odds ratio [OR] 0.87; 95% CI, 0.79 to 0.96), whereas a Day 2 FEV₁ that was larger by 100 ml was associated with a 20% reduction (OR, 0.80; 95% CI, 0.69 to 0.92). Theophylline use at entry also entered the model at a significant but weaker level. Patients who took theophylline before randomization were more likely to experience a treatment failure at 30 d than those patients who did not (OR, 1.74; 95% CI, 1.05 to 2.88).

We also searched for interactions, examining all possible combinations of the same five independent variables found significant in the univariate analyses. When entered into the multivariate model, no interaction was found to be significant at p  0.05.

The close relationship of FEV₁ to treatment failure is also evident from a simpler analysis. Subjects with an entry FEV₁ equal to or greater than the median of 750 ml had a lower treatment failure rate at 30 d than did subjects with a value less than 750 ml (18% versus 34%, p = 0.006). Similarly, subjects with a Day 2 FEV₁ greater than the median of 100 ml had a lower treatment failure rate than did subjects with a value equal to or less than 100 ml (13% versus 37%, p = 0.001). Subjects with an entry FEV₁ and a Day 2 FEV₁ that were both above the median had a treatment failure rate of 8% compared with a failure rate of 45% in subjects in whom both values fell below the median (p < 0.0001).

We attempted to identify subject characteristics that might explain variations in the entry FEV₁ and in the Day 2 FEV₁. In a multivariate linear regression model with entry FEV₁ as the dependent variable, we included age; histories of cough, phlegm, wheeziness, and chest colds; current smoking; inhaled -adrenergic agonist, inhaled anticholinergic, theophylline, and inhaled corticosteroid uses at entry; and previous hospitalization for COPD. None of the subject characteristics included in the model correlated with entry FEV₁ with p < 0.10 (results not shown).

The same variables were included in a linear regression model for Day 2 FEV₁, except that we added study drug assignment (placebo versus systemic corticosteroid) and entry FEV₁. This analysis showed a significant effect of treatment assignment (Table 4). As estimated from the slope coefficient, subjects receiving systemic corticosteroids had a Day 2 FEV₁ that was on average 80 ml larger than those subjects who received placebo. Theophylline entered the model at a marginally significant level (p = 0.06) with those subjects using theophylline at entry having a Day 2 FEV₁ that was on average 65 ml smaller than those subjects who had not been taking theophylline. However, adjusted coefficient of determination (R²) for this model was 0.04, meaning that we could explain only 4% of the total variability in Day 2 FEV₁.

The modest effect of corticosteroid treatment, relative to overall variability of Day 2 FEV₁, can be appreciated from the frequency distribution histograms in Figure 1. The range of values was greater than 1,500 ml in both treatment groups. The improvement associated with corticosteroid therapy, reflected as a small rightward shift of values in the histogram, is barely discernible against background variability. In addition, corticosteroid therapy caused no obvious change in the pattern of the frequency distribution, as might be expected if some very responsive patients had experienced unusually large improvements from corticosteroid therapy.

View larger version (25K): [in this window] [in a new window]   Figure 1.   Frequency distributions of changes in FEV1 from entry to Study Day 2 in placebo (n = 92) and corticosteroid-treated (n = 137) subjects. Values are shown as group proportions in 100-ml intervals. Mean values are 108 ml in the placebo arm and 198 ml in the corticosteroid arm.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX: STUDY PARTICIPANTS REFERENCES

This study demonstrated that FEV₁ measurements are strongly associated with a clinical outcome in patients with COPD exacerbations. The clinical outcome used in this trial, treatment failure, was a composite of death, intubation, rehospitalization, and therapy intensification. Intensification of drug therapy was the commonest form of treatment failure. Physicians were given the prerogative of intensifying therapy with one or more pharmacologic alternatives if at any time they made a judgment that the subject's overall clinical condition was unsatisfactory. We believe that was a clinically meaningful endpoint. We did not gather specific information as to how individual physicians came to their decision, and we did not attempt to assess the reproducibility of that judgment.

FEV₁ is a commonly used endpoint in respiratory disease trials, because the measurement is reproducible and easy to measure. It is less certain as to what constitutes a clinically meaningful change. In general, FEV₁ abnormalities tend to parallel clinical measures of respiratory health. However, cross-sectional surveys have shown relatively poor correlations of FEV₁ with dyspnea or exercise tolerance in patients with stable COPD (15, 16). Within-subject changes, as in response to a specific treatment, might be expected to show closer associations. Bronchodilator-induced FEV₁ improvements of 75 to 100 ml produced clear-cut evidence of improved quality of life in one large study of patients with stable COPD (17), but FEV₁ changes of the same magnitude were less clearly associated with improved clinical outcomes in another trial of similar size (18).

In the setting of a COPD exacerbation, this trial demonstrated that relatively small differences in FEV₁ have a substantial impact on clinical outcome. A 100-ml increase in FEV₁ measured at entry into the study reduced the relative risk of treatment failure at 30 d by 13%. A 100-ml improvement over the first 2 d of hospitalization reduced the relative risk of treatment failure by 20%.

Entry FEV₁'s and subsequent changes over the first few days of hospitalization were highly variable, and this trial provided little insight about their causes. We were unsuccessful in identifying any patient characteristics that predicted the magnitude of the entry FEV₁. Systemic corticosteroid therapy and theophylline use at entry into the study were the only factors that predicted Day 2 FEV₁ at significant or marginally significant levels, but these two factors accounted for less than 4% of the total variance. The information gathered by questionnaire related principally to the subjects' medical history and clinical condition before the exacerbation. A more focused assessment of events immediately preceding the hospitalization might have been more informative.

Thirty-four percent of the subjects were taking theophylline when they enrolled in the trial, and the protocol required that they stop taking the drug. The rationale for that requirement was described in a previous publication (12). Our analyses indicate that subjects who stopped taking theophylline experienced a significantly higher rate of treatment failure at 30 d and, at a marginal level statistically, a lower Day 2 FEV₁. These effects remained evident after adjustment for other variables. Stopping theophylline may have had a direct adverse effect on the clinical course of those subjects. Alternatively, theophylline use prior to hospital admission may have identified a sicker group of patients who were destined to do less well irrespective of whether they continued the drug.

The only randomized, controlled trial that evaluated theophylline use during COPD exacerbations found no statistically significant benefit in terms of FEV₁ improvement (7). Because of the small size of that trial, FEV₁ increases of approximately 100 ml could not be excluded. Those results are not inconsistent with the present trial, as we estimate that stopping theophylline at entry reduced the Day 2 FEV₁ by approximately 65 ml. Theophylline has a number of undesirable features, but our analyses suggest that it might improve clinical outcomes to a modest extent during COPD exacerbations. Confirmation of this observation would require a separate trial of substantial size.

Some, but not all, small trials suggest that a few patients with stable COPD experience unusually large FEV₁ improvements after a 1- or 2-wk course of systemic corticosteroids (19). Reference is sometimes made to patients with COPD as being either "corticosteroid-responsive" or "corticosteroid-nonresponsive," and the existence of these subtypes is implied in published guidelines for treating outpatients with COPD (23). The search for corticosteroid-responsive characteristics in patients with COPD remains a subject of clinical investigations (24, 25).

In this trial we found no evidence that selected subjects might have exhibited unusually large FEV₁ responses from systemic corticosteroids during a COPD exacerbation. As shown in Figure 1, there are no obvious differences in the distribution patterns for Day 2 FEV₁ when placebo-treated subjects are compared with corticosteroid-treated subjects. A heightened response to corticosteroids in selected patients should cause a skewed or bimodal distribution pattern in the active treatment arm. The few subjects with very large FEV₁ changes (> 700 ml) were as likely to have received placebo as active treatment. Within our limits to discern minor alterations in distribution patterns, our data are consistent with a scenario in which all subjects responded to systemic corticosteroids with a nearly constant 100-ml increase in FEV₁.

We conclude that relatively small changes in FEV₁ are associated with clinically meaningful differences in outcomes during exacerbations of COPD. Validation of FEV₁ as a surrogate endpoint in this setting may be useful for planning future trials. It is also possible that this information could be used to stratify risk and individualize therapy. Aside from relatively small treatment effects from systemic corticosteroids and possibly from theophylline, we were unable to explain the large variabilities associated with the entry FEV₁ or with the changes that occur in FEV₁ over the first few days of hospitalization.

	APPENDIX: STUDY PARTICIPANTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX: STUDY PARTICIPANTS REFERENCES

Planning Committee: M. Erbland (Co-chair), D. Niewoehner (Co-chair), R. Deupree (statistician), D. Collins (statistician), R. Light, S. Weiss, J. Stoller, I. Tager, M. Antonelli, M. Buchanan.

Data Monitoring Board: D. Dantzker (Chair), D. Tashkin, R. Simon, M. Lebowitz.

Pharmacy Coordinators: N. Morgan, J. Day.

Investigators at individual Veterans Affairs Medical Centers: J. Curtis (principal investigator) and C. Siegert, Ann Arbor, MI; G. Emmanuel (co-principal investigator), S. Carranza (co-principal investigator), and D. Johnson, Bay Pines, FL; P. Romano (co-principal investigator), P. Kaul (co-principal investigator) and R. Varano, Brooklyn, NY; S. Sethi (principal investigator) and P. DiMarzia, Buffalo, NY; N. Gross (principal investigator) and R. Keller, Hines, IL; M. Reinoso (principal investigator) and P. Guillet, Houston, TX; G. Bhaskar (principal investigator) and H. Hermesman, Lake City, FL; P. Anderson (co-principal investigator), G. San Pedro (co-principal investigator) and L. Frazier, Little Rock, AR; R. Light (principal investigator) and J. Despars, Long Beach, CA; K. Rice (principal investigator) and F. Lebahn, Minneapolis, MN; A. Fulambarker (principal investigator) and D. Ferguson, North Chicago, IL; P. Krumpe (principal investigator) and R. Weldermuth, Reno, NV; J. Liu (principal investigator) and T. Thompson, Salem, VA; M. Habib (principal investigator) and T. Vincent, Tucson, AZ; S. Santiago (principal investigator), D. Boyd, and L. Robinson, West Los Angeles, CA; J. Sampson (principal investigator), Alexandria, LA; F. Al-Bazzaz (principal investigator), Chicago Westside, IL; M. Nelson (principal investigator), Kansas City, MO; J. McCormick (principal investigator) and S. Shariaty, Lexington, KY; M. Tenholder (principal investigator), Memphis, TN; W. Davis (principal investigator) and Z. She, Augusta, GA; P. Caralis (principal investigator), Miami, FL; B. Gray (principal investigator) and K. Laughlin, Oklahoma City, OK; C. Atwood (principal investigator), Pittsburgh, PA.