INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The role of long-acting -agonists in the management of chronic asthma has evolved over recent years, and their use is now an integral part of asthma-treatment guidelines (1). In inadequately controlled asthma, addition of salmeterol to low or moderate doses of inhaled corticosteroid provided better symptom control and greater improvement of airway function than did increasing the dose of inhaled corticosteroid treatment by twofold or more (2, 3). Because concerns have been raised about possible adverse effects of long-term use of high-dose inhaled corticosteroid (4), achieving better control of asthma with lower doses of corticosteroid by adding salmeterol is likely to be widely accepted as appropriate treatment. On the other hand, concern has been expressed about whether long-acting -agonists may, by their bronchodilator and symptom-relieving effects, mask the development or persistence of airway inflammation and so put the asthmatic individual at risk of more severe asthma (7). This may be particularly true if improved symptom control leads to reduced use of inhaled corticosteroid.

We undertook a randomized placebo-controlled trial, using repeated measurements of noninvasive markers of inflammation (8), to determine whether salmeterol treatment could delay recognition of the occurrence of an exacerbation of asthma artificially induced by a progressive reduction of inhaled corticosteroid (9). Our hypothesis was that the symptom-relieving effects of a long-acting -agonist would mask the clinical effects of worsening inflammatory changes in the airway, and allow greater inflammation to develop before the patient reported an exacerbation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Asthmatic individuals considered to require high dose inhaled corticosteroid ( 1,500 µg beclomethasone dipropionate or budesonide) to maintain best results were recruited from the clinics of the Firestone Regional Chest and Allergy Unit (Table 1). All subjects had a typical history of asthma, with episodic wheezing, chest tightness, and dyspnea. Asthma was objectively confirmed by the presence of airway hyperresponsiveness (provocative concentration of methacholine required to reduce FEV₁ by 20% [PC₂₀ methacholine] < 8 mg/ml) or by evidence of variable airflow obstruction with an increase in FEV₁ of > 15% following 200 µg salbutamol taken by metered dose inhaler. In addition, subjects had to be able to produce sputum following saline induction. Asthmatic individuals with exacerbations within the 4 wk preceding enrollment in the study, those with current chest infections, or those with a need for regular oral corticosteroids were excluded, as were cigarette smokers and subjects with cardiac or nonasthmatic lung disease. Stability of asthma with the high-dose inhaled corticosteroid was judged by the need for no more than four puffs of salbutamol daily for symptom relief. Ethical approval for the study was obtained from the Research Committee of St. Joseph's Hospital, and all subjects gave signed consent.

Study Design

The study was a double-blind, two-period, randomized, controlled crossover trial comparing salmeterol with placebo, utilizing a stepwise reduction of inhaled corticosteroid, to determine the extent to which inflammation developed (as judged by sputum eosinophilia) before an exacerbation of asthma was clinically apparent (Figure 1).

View larger version (13K): [in this window] [in a new window]   Figure 1.   Study design. ICS = inhaled corticosteroid.

During an initial 1-wk run-in period ("initial" baseline), subjects continued their previous treatment in order to demonstrate stability when taking high-dose inhaled corticosteroid, and were then randomized to receive either salmeterol dry powder (Diskhaler Glaxo, UK) 50 µg twice daily or matching placebo. After 1 wk of this regimen to establish a new "treatment" baseline of symptoms and lung function, the daily dose of inhaled corticosteroid was progressively reduced. Beclomethasone dipropionate delivered from a metered dose inhaler with a valved holding chamber was reduced in 500-µg decrements each week until the daily dosage was 1,000 µg, and then by 250 µg at weekly intervals. Budesonide delivered from a Turbuhaler (Astra Draco, Sweden) was reduced each week in 400-µg steps to 800 µg daily, and then in 200 µg steps. Reduction in dosage continued until the subject met defined criteria for a mild exacerbation of asthma, or until the inhaled corticosteroid was totally withdrawn. Following restabilization for at least 4 wk at the original inhaled corticosteroid dose, with an initial 1-wk course of prednisone at 30 mg daily if the exacerbation was more than mild, the subject entered the second arm of the study, using the alternative blinded agent, and followed the same protocol with the same dose-reduction steps. Subjects were not entered into the second arm of the study until they demonstrated a return to their initial baseline values for symptoms, -agonist use, peak flow rates, and FEV₁, and had waited a minimum of 4 wk from the onset of the exacerbation terminating the first arm of the study.

At the initial visit, symptoms, medications, spirometric values, methacholine challenge response (if the prebronchodilator FEV₁  70% of predicted) or reversibility of FEV₁ with 200 µg inhaled salbutamol, allergy skin prick test results, sputum eosinophil count, and serum eosinophil cationic protein (ECP) concentration were determined. Subjects were instructed to continue their usual treatment and to record symptoms, medication use, and morning and evening peak expiratory flow (PEF) in a daily diary. All subsequent weekly visits were at the same time of day. Study medication was withheld for 12 h before each visit. Diary cards were reviewed, and symptoms, spirometric, methacholine challenge, blood, and sputum measurements were repeated. Subjects were asked to contact a study physician if symptoms or PEF deteriorated to levels reflecting an exacerbation, which was defined by the simultaneous occurrence of at least one subjective and one objective marker of deterioration. Subjective markers of exacerbation were either increased symptoms (a Likert symptom score that fell below 80% of the score determined during the initial baseline period for that subject) or increased use of short-acting -agonist (two or more puffs above the mean daily baseline use for that subject). Objective markers of exacerbation were either decreased PEF (morning PEF decreased to less than 85% of mean morning PEF during the initial baseline week) or decreased FEV₁ at the clinic visit (< 85% of the mean baseline FEV₁ determined at visits at either end of the initial baseline week).

Clinical Methods

Patient characteristics were documented with a questionnaire. Symptoms (chest tightness, shortness of breath, wheezing, and cough) were each graded on a Likert scale (10) and recorded in a daily diary. Each symptom score ranged from 1 (the most severe discomfort) to 9 (asymptomatic), giving a maximum score of 36 (fully asymptomatic). PEF was measured with a MiniWright peak flow meter (Armstrong Medical Industries, Scarborough, ON, Canada), and the best of three measurements was recorded in the diary. Spirometry was performed with a Collins 9L water spirometer (Warren E. Collins Inc., Braintree, MA). Spirometry, methacholine inhalation tests, and allergy skin tests with 12 common inhaled allergen extracts were performed according to standard procedures (11).

Sputum Induction

At every visit, sputum was induced by the inhalation of an aerosol of hypertonic saline in increasing concentrations (3%, 4%, and 5%) generated by a Fisoneb ultrasonic nebulizer (Canadian Medical Products, Ltd., Markham, ON, Canada) with an output of 0.87 ml/min and particle size of 5.58 µm aerodynamic mass median diameter (8).

Sputum Examination

Sputum was separated from saliva (14) and processed within 2 h as described by Pizzichini and coworkers (8). Briefly, sputum was treated by adding four volumes of 0.1% dithiothreitol (DTT) (Sputalysin 10%; Calbiochem Corp., San Diego, CA) followed by four volumes of Dulbecco's phosphate buffered saline (D-PBS). The suspension was filtered through a 48-µm nylon gauze (BBSH Thompson, Scarborough, ON, Canada), the filtrate centrifuged at 790 × g for 10 min, and the supernatant aspirated and stored in Eppendorf tubes at 70° C for later assay. The cell pellet was resuspended in 200 to 600 µl of D-PBS, depending on macroscopic size, and a total cell count of leukocytes was made and cell viability determined. The cell suspension was adjusted to 1.0 × 10⁶/ml and placed into cups of a Shandon III cytocentrifuge (Shandon Southern Instruments, Sewickley, PA), and two coded cytospin preparations were made, air dried, and stained with Wright's stain. A differential cell count was made on 400 nonsquamous cells.

Blood Examination

Venous blood was collected into a 5.0 ml ethylenediamine tetraacetic acid (EDTA)-containing tube (K₃ Vacutainer; Becton-Dickinson, Rutherford, NJ). Serum was collected after blood coagulation for 1 h at room temperature, centrifuged at 20° C at 1,500 rpm for 10 min, and stored at 20° C until analyzed. The concentration of serum ECP was determinated with a radioimmunoassay (RIA) (Kabi Pharmacia Diagnostics AB, Uppsala, Sweden). All blood analyses and sputum differential cell counts were performed with blinding to both the clinical situation and dose of inhaled corticosteroid taken.

Sample Size

Sample-size calculations were based on expected corticosteroid dose reduction rather than on markers of inflammation, for which no predictive data were available. Calculations assumed an expected difference of 20% in inhaled corticosteroid dose reduction (e.g., 70% dose reduction during salmeterol treatment versus 50% dose reduction during the placebo phase, and a uniform distribution of differences in steroid reduction between 0% and 100%). Significance was set at 5% and power at 95%, yielding a sample size of 32 subjects. Because of the complexity of the study and demands on the patients with weekly visits and sputum inductions, interim analyses of data were planned after 11 and 22 subjects, respectively, completed the study.

Data Analysis

The outcomes of interest in the study were the magnitude of reduction of inhaled corticosteroid during treatment with salmeterol versus placebo, and the degree of inflammation present before an exacerbation was recognized. This was reflected in the proportion of sputum eosinophils, levels of serum ECP, and measurements of airway responsiveness to methacholine in the week immediately before exacerbation and at the two visits prior to that week.

All data were analyzed with the statistical package SPSS for Windows, release 7.0 (SPSS Inc., Chicago, IL). Clinical data are reported as mean ± SD. Nonnormally distributed data were log transformed before analysis, and are reported as geometric mean and geometric SD. Significance was accepted at 5%. Two-tailed paired t tests with Bonferroni's correction for multiple comparisons were used for comparisons between treatments at the new treatment baseline during the first week after randomization, and at the final inhaled steroid dose in each arm. In order to examine events leading up to exacerbation, which occurred over a variable period (in different subjects and between treatments) after commencing the reduction of inhaled steroid, the "exacerbation visit" of each phase was taken as the reference point, and data obtained at 1, 2, and 3 wk before that time were compared. For example, the development of inflammation prior to exacerbation was determined by analysis of the weekly sputum eosinophil counts at the three visits (E-3, E-2, E-1) preceding the exacerbation (E) visit. In those cases in which inhaled corticosteroid was fully withdrawn and the subject did not experience an exacerbation in the subsequent week, the final visit for that phase was coded as being 1 wk before exacerbation (E-1), and the exacerbation (E) visit was coded as missing data (not achieved).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Seventeen subjects meeting inclusion and exclusion criteria were recruited and entered the randomization period of the study. Of these, 13 subjects provided evaluable data for the first planned interim analysis, at which time the study was halted. Characteristics of these 13 subjects are shown in Table 1. Four subjects were discontinued from participation in the study: one was noncompliant, one moved from the area, one needed prednisone for another condition, and one failed to regain stability after an exacerbation terminating the first study treatment period (salmeterol). In one included subject, an exacerbation occurred as defined by symptoms and -agonist use, both of which increased markedly, but because of a need for frequent -agonist to control these symptoms, it was not possible to obtain prebronchodilator PEF or FEV₁ measurements. Postbronchodilator measurements did not yield values below the cutoff-point used to define an exacerbation. However, because the clinical situation was very clearly an exacerbation, this episode was coded as such despite the nonavailability of objective criteria. This decision was made without knowledge of which treatment the patient was taking at the time of this exacerbation.

Initial baseline data for the 13 subjects reported, derived from the first week of records of each study period, were not significantly different for the different treatments (Table 2, Figure 2). Treatment with salmeterol for 1 wk, but not with placebo, resulted in different treatment baselines, with a reduction in ₂-agonist use (2.8 to 1.7 puffs/d, p = 0.06), and a significant increase in mean morning PEF (419 to 450 L/min, p < 0.001) and in FEV₁ (75.2 to 81.9% predicted, p < 0.001) during salmeterol treatment (Table 2, Figure 2).

View larger version (15K): [in this window] [in a new window]   Figure 2.   Treatment effect is shown by mean paired differences (95% CI) between treatments. The three panels represent differences at initial baseline (randomization), treatment baseline (1 wk of treatment with salmeterol or placebo), and pre-exacerbation visit (E-1). *p = 0.05, **p = 0.006.

Treatment with salmeterol permitted a significantly greater reduction in inhaled corticosteroid dose before exacerbation (87.0% ± 24.1 versus 69.2% ± 34.3, p = 0.04). The median dose of inhaled corticosteroid at the time of exacerbation (or termination of this phase of the study if no exacerbation occurred) was 277 ± 661 µg/d during salmeterol treatment and 612 ± 795 µg/d during placebo administration (p = 0.01), which corresponded to reductions of 91.3 ± 20.7% and 77.9 ± 28.3%, respectively (p = 0.01). The average number of dose- reduction steps was 6.9 ± 2.1 during salmeterol treatment, and 5.6 ± 2.6 during administration of placebo. During salmeterol treatment, six of 13 subjects reduced their steroid use to zero without exacerbation, three experienced exacerbations only in the last week of the study, when their corticosteroid dose was zero, and four experienced exacerbations during the reduction regimen. During the placebo arm of the study, four subjects reduced steroid use to zero without exacerbation, one exacerbation occurred in the week in which inhaled corticosteroid was zero, and eight exacerbations occurred during steroid reduction.

There was substantial scatter in the percentages of sputum eosinophils at all time points, but consistent trends toward increasing levels as corticosteroids were reduced (Figure 3). Sputum eosinophil counts were higher in the salmeterol arm than in the placebo arm at the last three visits before exacerbation (E-3, E-2, E-1), when the dose of inhaled corticosteroid was lower in the salmeterol arm, but the differences only reached statistical significance in the week before exacerbation (visit E-1), when the mean sputum eosinophil count was 19.9 ± 29.8% during salmeterol treatment versus 9.3 ± 17.6% during administration of placebo (p = 0.006). Despite these levels of eosinophilia, clinical variables and FEV₁ remained stable. Nine of 13 subjects had abnormal eosinophil counts above 2% at visit E-1 in the salmeterol phase, compared with six of 13 in the placebo phase; in five cases the counts were markedly elevated, at more than 10% eosinophils during salmeterol treatment as compared with only two cases with placebo. Hence the use of salmeterol allowed subjects to tolerate a greater degree of inflammation without increased symptoms or reduced lung function.

View larger version (12K): [in this window] [in a new window]   Figure 3.   Time-course of effects of inhaled steroid reduction associated with use of salmeterol (shaded squares) or placebo (closed circles) on mean (SEM) sputum eosinophil (percentage). IB = initial baseline; TB = treatment baseline; E-4 = four visits pre-exacerbation; E-3 = three visits pre-exacerbation; E-2 = two visits preexacerbation; E-1 = pre-exacerbation visit; Ex = exacerbation visit. *p < 0.05 for paired differences between treatments. Within salmeterol treatment p values for sputum eosinophils were < 0.01 at visits E-3, E-2, E-1, and Ex. Within placebo administration, p values for sputum eosinophils were < 0.01 at visits E-1 and Ex.

There was a trend toward increasing airway responsiveness in the salmeterol-treated group as compared with the placebo group before exacerbation of symptoms was clinically evident, but this did not achieve statistical significance. There were no differences in symptom scores, -agonist use, PEF, or FEV₁ leading up to the exacerbation, despite increasing eosinophilia (Figure 4).

View larger version (15K): [in this window] [in a new window]   Figure 4.   Time-course of effects of inhaled steroid reduction associated with use of salmeterol (shaded squares) or placebo (closed circles) on mean symptoms scores and FEV1 % predicted. IB = initial baseline; TB = treatment baseline; E-4 = four visits pre-exacerbation; E-3 = three visits pre-exacerbation; E-2 = two visits pre-exacerbation; E-1 = pre-exacerbation visit; Ex = exacerbation visit. There were no significant differences between treatments at any time point after the initial increase in FEV1 at TB with salmeterol (p < 0.001).

Of the 16 subjects who experienced exacerbations, four required more than 4 wk to restabilize (two after placebo, and two after salmeterol).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Many studies report the efficacy of long-acting -agonists in improving lung function and symptom control in asthma (2, 3, 15), both when added to inhaled corticosteroids and when used as monotherapy. These beneficial effects may lead the physician or the patient to reduce the dose of inhaled corticosteroid, especially as awareness of adverse effects of inhaled corticosteroids increases (4).

Using the model of stepwise reduction of inhaled corticosteroid to allow inflammation to gradually increase, we have confirmed our hypothesis that regular inhalation of salmeterol can delay the recognition of worsening asthma. In the salmeterol arm of our study, subjects had a twofold greater level of sputum eosinophils in the week before a clinically recognizable exacerbation of asthma than occurred in the placebo arm. Hence, salmeterol allowed the subjects to tolerate inflammation that would otherwise have led to symptoms and decreased lung function, a situation appropriately termed "masking." There was no clear evidence that salmeterol increased airway inflammation; rather, its use masked inflammation that was worsening because of steroid reduction. The study design resulted in early detection and treatment of exacerbations, and did not provide data indicating whether the use of salmeterol would have led to more severe exacerbations without treatment. We observed three instances of delayed recovery from exacerbation after salmeterol (including the subject who could not reenter the study for the second phase) as compared with two after placebo.

The steroid-reduction model that we used to induce an exacerbation of asthma has been validated in other clinical trials (9, 18), but clearly reflects an artificial situation in that inhaled corticosteroids would usually be reduced at intervals of a few weeks rather than every week. The degree of corticosteroid reduction, averaging about 75% over both arms of the study, is greater than would have been expected had the reduction been performed more slowly, as the benefits of corticosteroid may persist for many weeks. There may have been a degree of overtreatment of some subjects initially, but others had borderline increases in eosinophils at baseline despite the high doses of steroid they were using. In most subjects, an exacerbation occurred as the dose was reduced. It was not surprising that the dose of inhaled corticosteroid could be reduced more during salmeterol treatment than with placebo, as has been previously shown (19), nor that sputum eosinophils increased as inhaled corticosteroids were reduced. The significance of this study is that in this steroid-reduction model of induced exacerbation, salmeterol controlled symptoms and PEF for a longer period before an inflammatory exacerbation became clinically evident.

A previous study, using daily symptom scores, nocturnal awakenings, PEF, and -agonist use as markers of exacerbation, did not detect any evidence that salmeterol treatment masked worsening of asthma (20). However, the prolonged bronchodilator effect of salmeterol makes such clinical markers less useful in detecting worsening inflammation. We have used markers of airway inflammation not directly affected by the bronchodilator properties of salmeterol to look for evidence of increasing inflammation before clinical markers indicated deterioration.

Both a case report (7) and experimental studies have suggested that long-acting -agonists can mask deterioration of asthma. In a trial of regular formoterol, one subject discontinued inhaled corticosteroid while continuing formoterol and theophylline (7). When he then discontinued formoterol, he had a rapid and severe deterioration with a decreased response to short-acting -agonist, indicating that formoterol had effectively masked the worsening of asthma to that point. In a recent study, formoterol added to constant high (800 µg) or low (200 µg) doses of inhaled corticosteroid increased lung function and reduced the mean number of exacerbations (21). In a parallel study, sputum eosinophils decreased rapidly and remained low in subjects treated with budesonide at 800 µg or budesonide at 200 µg plus formoterol (22). However, there was a trend after 1 yr toward increasing sputum eosinophil numbers in the formoterol-treated group, which was using a low dose of inhaled corticosteroid, despite maintenance of better lung function (22), suggesting that formoterol may mask increasing inflammation. Studies of allergen inhalation have shown that the potent functional effects of even a single dose of salmeterol (23) or formoterol (24) can mask the clinical effects of airway inflammatory-cell influx following challenge, which is analogous to what occurred in the present study.

In summary, we have shown that the regular use of salmeterol has the potential to mask a worsening of airway inflammation and to delay awareness of an exacerbation of asthma, since symptoms and lung function during treatment with salmeterol remained well controlled until eosinophilic inflammation became significantly more advanced. Reduction of inhaled corticosteroid therapy in subjects using long-acting -agonists should be undertaken with caution, since worsening of inflammation may be less easily recognized in the presence of a long-acting bronchodilator agent.