Long-Acting β₂-Agonist Monotherapy vs Continued Therapy With Inhaled Corticosteroids in Patients With Persistent Asthma:  A Randomized Controlled Trial FREE

Regular use of long-acting β₂-agonists has been shown to be more effective than regular use of albuterol sulfate, a short-acting β₂-agonist, at improving peak expiratory flow (PEF) and reducing asthma symptoms.^1- 2 Asthma treatment guidelines³ recommend addition of a long-acting β₂-agonist for asthma inadequately controlled by low-to-moderate dosages of an inhaled corticosteroid (ICS) because addition of salmeterol xinafoate or formoterol fumarate, 2 long-acting β₂-agonists, was shown to be more effective than increasing the dose of the ICS.^4- 6 In the first half of 1999, more prescriptions were written in the United States for salmeterol for patients with asthma than for any other asthma medication except albuterol (IMS Health, Plymouth Meeting, Pa, unpublished data, September 2, 1999). The role of long-acting β₂-agonists as monotherapy, however, is unclear,^7- 9 since most studies of salmeterol have included a significant proportion of patients taking other controller medications, including ICSs.^1- 2 A recent study suggests that salmeterol may be equivalent to low-dose beclomethasone dipropionate in corticosteroid-naive patients with persistent asthma.¹⁰

Current asthma treatment guidelines suggest use of controller medications for patients who require rescue albuterol more than 2 to 3 times per week.^3,11 These recommendations are based on recognition of the inflammatory basis of asthma and on several studies suggesting that a delay in initiation of ICS therapy could lead to irretrievable loss of airway function, attributable to "airway remodeling" that results from untreated inflammation.^12- 15 Inhaled corticosteroids are widely used as controller therapy, but long-acting β₂-agonists are also classified as long-term controller medications³ even though they are generally believed to have little effect on inflammatory mediators or cells.^16- 18

How differences in the mechanisms of action of long-acting β₂-agonists and ICSs translate into differences in effectiveness of long-term treatment of adults with asthma is not known. For patients whose asthma is well controlled by low-to-moderate doses of an ICS, must the ICS be continued, or might monotherapy with a long-acting β₂-agonist be equally or more effective? To our knowledge, no evidence from clinical trials supports replacement of ICS therapy with salmeterol monotherapy in patients with persistent asthma, but this approach has begun to appear in clinical practice. To study these questions, the Asthma Clinical Research Network (ACRN) of the National Heart, Lung, and Blood Institute (NHLBI) undertook a 28-week, randomized, multicenter, blinded, placebo-controlled trial in patients with moderate persistent asthma that was well controlled with an ICS. The trial compared the effects of continuing the ICS, triamcinolone acetonide, after a 6-week run-in period; switching to a long-acting β₂-agonist, salmeterol; or switching to placebo on asthma symptoms, pulmonary function, bronchial reactivity, markers of airway inflammation, and frequency of exacerbations during treatment and after treatment was withdrawn. The trial tested the null hypothesis that in patients with moderate persistent asthma whose symptoms are well controlled with regularly scheduled triamcinolone and as-needed inhaled rescue albuterol, continued treatment with triamcinolone does not differ in efficacy from a change to monotherapy with salmeterol.

This 28-week, randomized placebo-controlled trial, the Salmeterol or Corticosteroids (SOCS) study, was conducted from February 1997 to January 1999 at the 6 clinical sites that comprise the ACRN. The study was approved by the NHLBI-ACRN Protocol Review Committee and by the committees on human research at each clinical site. Written informed consent was obtained from all enrolled patients. Patients with asthma as defined by the American Thoracic Society¹⁹ who met recommended criteria for treatment with an ICS²⁰ were recruited. Four hundred twenty-two patients who met the inclusion criteria (BOX 1) entered a 6-week run-in phase during which all patients received 400 µg (4 puffs) twice per day of open-label triamcinolone acetonide. Patients whose asthma was well controlled, defined objectively (BOX 1), following the 6-week run-in period were entered into the SOCS study. Patients whose asthma was not well controlled after the run-in were assigned to a concurrent study, the Salmeterol ± Inhaled Corticosteroids (SLIC) study.²¹

Patients eligible to continue in the SOCS study were randomly assigned to receive either 400 µg (4 puffs) twice per day by metered-dose inhaler (MDI) of triamcinolone; 42 µg (2 puffs) twice per day by MDI of salmeterol xinafoate; or 2 puffs twice per day by MDI of placebo for the next 16 weeks. After 16 weeks, all active scheduled medications were stopped for an additional 6-week, single-blind placebo run-out period.

Inclusion Criteria for Study Entry (Triamcinolone Run-in Period)

Inclusion Criteria for Allocation Into the SOCS Study After 6-Week Triamcinolone Run-in Period

Patient randomization was performed online via an Internet connection to the computer system at the data coordinating center. Patients were stratified according to clinical center, methacholine responsiveness (PC₂₀, the provocative concentration of methacholine required to decrease forced expiratory volume in 1 second [FEV₁] by 20%), race/ethnicity, sex, and age by a permuted-blocks scheme, with blocks of random size within each stratum. When a patient was deemed eligible for study entry, a clinical center staff member entered and verified the pertinent data and received a drug packet number to give the patient.

The study was triple-blinded: patients, clinical center personnel, and data analysts were all blinded to treatment identity and dosages. Each patient received 2 inhalers, either active triamcinolone plus placebo salmeterol, placebo triamcinolone plus active salmeterol, or placebo triamcinolone plus placebo salmeterol. All patients were given albuterol inhalers to use for rescue treatment throughout the study. Triamcinolone and triamcinolone placebo were administered through a built-in spacer; all other MDIs were used without spacers. All inhalers used chlorofluorocarbon propellant. Patients were trained in each technique, and techniques were assessed at each visit. Treatment medication for each patient was packaged together, labeled with a unique number, and distributed to the clinical centers. The contents of the drug packages were known only to administrative personnel at the data coordinating center.

Throughout the study, patients rated the severity of 5 asthma symptoms (shortness of breath, chest tightness, wheezing, cough, and phlegm/mucus) on a scale of 0 (none) to 3 (severe). Patients recorded daytime and nighttime asthma symptom scores, morning (AM) and evening (PM) PEF using an Airwatch device (Enact, Palo Alto, Calif), rescue albuterol use, intercurrent illnesses, and hospitalizations. Patients were evaluated every 2 to 4 weeks (after a 6-hour albuterol hold and 48-hour salmeterol hold) for interval history, physical examination, diary review, spirometry (Collins Eagle 2 spirometer, Quincy, Mass), and measurement of exhaled nitric oxide (NOA 280, Sievers, Boulder, Colo).²² An asthma-specific quality-of-life questionnaire²³ was administered at the beginning of the run-in period (baseline), at the end of the run-in period (week 6), at the end of the randomized treatment period (week 22), and 2 and 6 weeks after cessation of therapy (run-out period, study weeks 24 and 28). Methacholine responsiveness²⁴ and sputum induction for analysis of total and differential cell counts, eosinophil cationic protein (ECP), and tryptase²⁵ were performed at baseline; at the end of the run-in period; after 2, 8, and 16 weeks of randomized treatment; and 2 and 6 weeks after cessation of therapy (Figure 1).

All testing was performed at each site with standardized equipment and procedures. Network staff were trained and tested to ensure proficiency and uniformity in all procedures. All spirometric testing, including that for PC₂₀, was overread by a single member of the network. The function of the PEF measurement device was confirmed at each clinic visit, and the devices were replaced if they failed to meet predetermined standards. A distributed data entry system allowed each clinical center to submit data electronically to the data coordinating center. The data coordinating center entered the data a second time for verification.

The primary outcome variable was change in AM PEF from the final week of the run-in period (week 6) to the final week of the randomized treatment period (week 22). To examine the duration of benefit after treatment was stopped, we also compared the change in AM PEF from the final week of the run-in period (week 6) to the final week of the run-out period (week 28). Similar analyses were applied to other markers of asthma severity and asthma control, including FEV₁, asthma symptom scores, rescue albuterol use, quality-of-life scores, and PC₂₀. Other outcomes included the number of asthma exacerbations in the treatment and off-treatment periods.

To determine whether differences in efficacy or duration of benefit reflected differences in anti-inflammatory activity, we compared total and differential cell counts as well as concentrations of ECP and tryptase from induced sputum samples²⁵ and measured exhaled nitric oxide at each visit.²²

Specific criteria were established prospectively to define treatment failure, asthma exacerbation, and run-out failure (BOX 2). Patients who met treatment failure criteria were treated with a short burst of prednisone or 400 µg twice per day of open-label, inhaled triamcinolone acetonide, and continued open-label triamcinolone instead of study triamcinolone for the remainder of the study. Study inhalers of salmeterol or placebo were continued. Data collected during and after treatment failure were included in the primary intention-to-treat (ITT) analysis for all outcome variables. Patients who developed an asthma exacerbation during the run-in period (prior to randomization) were withdrawn from the study; asthma exacerbations following randomization were managed according to specific predefined rescue algorithms.³ Trial drugs were continued during exacerbations unless a physician believed it appropriate to suspend such therapy until the exacerbation had resolved. Patients with asthma exacerbations, as in the case of treatment failure, were included in the ITT analysis. Patients who met failure criteria during the run-out phase were dropped from further study participation. All data collected prior to study withdrawal were included in the analyses.

Criteria for Treatment Failure Status During the Randomized Treatment Period

Any of the following:

Criteria for Asthma Exacerbations

Increased cough, chest tightness, or wheezing in association with 1 or more of the following:

Criteria for Failure During Placebo Run-out Period

Any of the following:

*Baseline rescue albuterol use refers to the average from the last 2 weeks of the triamcinolone run-in period.

†Reference PEF level was the average prebronchodilator AM PEF from the last 2 weeks of the triamcinolone run-in period.

The primary analyses were conducted using longitudinal data analysis based on fitting a mean for each treatment group at each point.²⁶ Daily daytime and nighttime symptom scores for each patient were averaged, yielding a 15-point scale between 0 and 3, and then averaged over each week for each patient, yielding a 105-point score that was used in the longitudinal data analysis. Within-group differences were constructed from model-based estimates, eg, end of treatment minus end of run-in, and these within-group differences were compared across groups. Variables that displayed a high level of skewness or discreteness were analyzed via rank tests. All available data on all randomized patients were included in the primary statistical analysis and analyzed according to treatment group assignments at randomization. Values were not imputed for missing data.

Patients who were assigned treatment failure or asthma exacerbation status received protocol-defined treatment with prednisone, inhaled triamcinolone, or both. Because this rescue therapy was likely to affect outcomes, we planned a secondary analysis, performed by the last-observation-carried-forward (LOCF) method, carrying forward the last value prior to treatment failure or asthma exacerbation, or by excluding data collected after treatment failure or asthma exacerbation (truncation). The cumulative incidences of treatment failure and asthma exacerbation were analyzed using Kaplan-Meier survival curves.

Based on published data,¹³ we estimated that a total of 150 randomized patients were required to provide an 80% likelihood of detecting clinically significant changes in AM PEF. This sample size would detect a 15.7-L/min difference in AM PEF in a 3-armed trial, assuming a withdrawal rate of 20%. This sample size also yielded effect sizes for FEV₁ of 0.17 L and PC₁₅ of 0.68 dose steps (less than 1 doubling dose). Because we applied a Bonferroni correction to account for 3 pairwise comparisons and conducted an interim analysis at the trial midpoint based on the O'Brien-Fleming group sequential method,²⁷ a P value of less than .016 was required for statistical significance for all 3-way comparisons. Although descriptive statistics and P values are reported for within-group changes, the P values also should be considered as descriptive statistics because changes over time within a treatment group are not necessarily due to the effect of that particular treatment.

Four hundred twenty-two patients were enrolled, 389 (92%) of whom were already receiving an ICS at enrollment. Of the 422 patients enrolled, 361 completed the run-in period (Figure 1). Of these, 164 met entry criteria for the SOCS study and were randomly assigned to blinded treatment with placebo (n = 56), salmeterol (n = 54), or triamcinolone (n = 54). The groups were well matched with regard to age, sex, and race/ethnicity and did not differ by airway function, asthma symptoms, or inflammatory cells in induced sputum (Table 1). During the 22 weeks of randomized treatment and the placebo run-out period, 34 patients withdrew from the trial, 7 in the placebo group, 13 in the salmeterol group, and 14 in the triamcinolone group. Twelve of 34 withdrawals occurred during the randomized treatment period; 3 were for dissatisfaction with asthma control, 1 was physician-initiated, and the remainder were for personal or administrative reasons. One serious adverse event (not asthma-related) occurred in the salmeterol group. Of 792 scheduled visits in the triamcinolone run-in period, none were missed. During the treatment period, 1314 (99%) of 1327 scheduled visits were completed and 91% occurred within preassigned time windows. Over the duration of the study, patients reported their AM PEF on 94.7% of days and reported taking their inhaled medications as directed on 81.1% of days.

Even though 159 (97%) of the 164 patients who entered the SOCS study were already receiving an ICS at enrollment, AM PEF, FEV₁, PC₂₀, and sputum eosinophil concentration all improved significantly after 6 weeks of run-in therapy with 400 µg twice per day of open-label triamcinolone (Table 2).

Primary ITT Analysis. During the randomized treatment period, the primary outcome variable, AM PEF, increased in the salmeterol and triamcinolone groups and decreased initially then increased in the placebo group (Figure 2A). We found no statistically significant differences either within or among the 3 groups (Table 3).

In the Kaplan-Meier analysis, the treatment failure rate for the triamcinolone group was significantly lower than the treatment failure rates in the placebo (P<.001) and salmeterol (P = .004) groups (Figure 3A). Treatment failure occurred in 20 patients (36%) in the placebo group, 13 (24%) in the salmeterol group, and 3 (6%) in the triamcinolone group. Nineteen of these 36 patients were judged to have treatment failure because of an asthma exacerbation that met our preestablished criteria for prednisone therapy; in the remaining 17, the patient's asthma control was considered to be sufficiently unstable to warrant a change in therapy by physician clinical judgment for safety. The difference in treatment failure rates between the placebo and salmeterol groups was not significant (P = .18). The pattern of asthma exacerbations paralleled that of treatment failures (Figure 3B). Asthma exacerbations were experienced by 16 patients (29%) in the placebo group, 11 (20%) in the salmeterol group, and 4 (7%) in the triamcinolone group. The rate of asthma exacerbations was significantly lower in the triamcinolone group compared with the salmeterol (P = .04) and placebo (P = .003) groups.

Secondary LOCF Analysis: Outcomes Prior to Treatment Failure or Asthma Exacerbation. In the LOCF analyses, there were statistically significant within-group changes in every outcome in the placebo group during the randomized treatment period (Figure 2 and Figure 4); AM PEF, PM PEF, FEV₁, PC₂₀, and quality of life all decreased; rescue albuterol use, daily asthma symptom scores, sputum eosinophils, ECP, and tryptase, and exhaled nitric oxide all increased (P≤.03 for all). In contrast, no significant within-group change occurred in any outcome in the triamcinolone group. In the salmeterol group, AM PEF, PM PEF, rescue albuterol, asthma symptom scores, and quality of life did not change significantly within the group, but FEV₁ and PC₂₀ (both measured after β₂-agonist hold) decreased and exhaled nitric oxide and sputum eosinophils, ECP, and tryptase all increased (P<.04 for all).

During the randomized treatment period (end of week 6 through week 22), increases in markers of inflammation—sputum eosinophils (median [interquartile range {IQR}], 2.4% [0% to 10.6%] vs −0.1% [−0.7% to 0.3%]; P<.001), sputum ECP (71 [−2 to 430] U/L vs −4 [−31 to 56] U/L; P = .005), and sputum tryptase (3.1 [2.1 to 7.6] ng/mL vs 0 [0 to 0.7] ng/mL; P<.001)—were significantly greater in the salmeterol group than in the triamcinolone group. There were no significant differences in the salmeterol vs triamcinolone groups for interval changes in any of the other outcomes. Similar results were obtained for the salmeterol vs triamcinolone comparisons using ITT and truncation analyses, except that changes in ECP in the 2 groups were no longer significantly different. When we compared changes in the placebo and salmeterol groups during the randomized treatment period using the LOCF analysis, we found significant differences for AM PEF, rescue albuterol use, daily asthma symptom scores, and quality of life (P<.01 for all); there were no significant differences between placebo and salmeterol for PM PEF, FEV₁, PC₂₀, exhaled nitric oxide, sputum eosinophils, ECP, or tryptase. Changes during the randomized treatment period were significantly different between the placebo and triamcinolone groups for AM PEF, rescue albuterol use, daily asthma symptom scores, quality of life, sputum eosinophils, and tryptase (P<.01 for all); there were no significant differences for PM PEF, FEV₁, PC₂₀, exhaled nitric oxide, or ECP.

During the placebo run-out period (weeks 23-28), when all patients received placebo, there were 18 run-out failures, 2 in the placebo group, 6 in the salmeterol group, and 10 in the triamcinolone group. Sixteen patients met objective criteria for run-out failure; 2 others were identified because of unstable asthma by physician clinical judgment for safety. Four additional patients withdrew during the run-out period for personal or administrative reasons. The differences in failure rates between the salmeterol and triamcinolone groups and between the placebo and salmeterol groups were not significant (P = .05 for both); the difference between the placebo and triamcinolone groups was significant (P = .004).

When patients were switched from triamcinolone to placebo, there were statistically significant decreases in AM PEF, FEV₁, and PC₂₀, and increases in rescue albuterol use and daily asthma symptom scores in all analyses (within-group, ITT, and LOCF) (Figure 5). When patients were switched from salmeterol to placebo, there were significant within-group changes in FEV₁, rescue albuterol use, and daily asthma symptom scores. The placebo group, having already worsened during the randomized treatment period, demonstrated no further significant change in any outcome. When changes during the placebo run-out period were compared between groups, there were significant differences between the placebo and triamcinolone groups for FEV₁ (P = .015 by ITT), and PC₂₀ (P = .003 by ITT; P = .01 by LOCF). Percentage of eosinophils and tryptase concentration in sputum and exhaled nitric oxide all increased significantly (within group, P = .002 for all) when the triamcinolone group was switched to placebo, and the pattern mirrored that seen in the placebo and salmeterol groups when triamcinolone was stopped at the end of the triamcinolone run-in period.

To determine whether the duration of active treatment affected airway function after treatment was stopped, we compared ITT changes in AM PEF, FEV₁, and PC₂₀ between the end of the placebo run-out (week 28) and the end of the triamcinolone run-in (week 6) periods. There were no significant differences among groups, indicating no long-term carryover benefit from up to 22 weeks of triamcinolone treatment.

Although use of long-acting β₂-agonists as additive therapy in patients whose asthma is not controlled by inhaled corticosteroids is well supported by the results of clinical trials,^4- 6 their use as monotherapy in persistent asthma is not. Indeed, US guidelines for diagnosis and management of asthma recommend long-acting β₂-agonists only as additive therapy,³ but this recommendation was based on expert opinion, not on evidence from clinical trials. Our study demonstrates that patients with persistent asthma that is well controlled by low-dose triamcinolone monotherapy cannot be switched to salmeterol monotherapy without risk of clinically significant loss of asthma control.

Although most of the 164 patients in this study reported using an ICS prior to entry, significant improvements were still noted in AM PEF, FEV₁, PC₂₀, and sputum eosinophil concentration during the triamcinolone run-in period, presumably reflecting greater compliance with regular ICS treatment in the setting of a clinical trial. In the randomized treatment period, patients who continued treatment with triamcinolone had better asthma control than those treated with placebo or salmeterol. Treatment failure and asthma exacerbations were the most robust indicators of the difference in asthma control achieved by treatment with triamcinolone vs salmeterol. Thus, although patients who received salmeterol had fewer asthma symptoms, less need for rescue albuterol, and better airway function than patients who received placebo, they experienced treatment failures and asthma exacerbations at a rate similar to the placebo group. Similarly, methacholine responsiveness worsened significantly in the salmeterol and placebo groups but not in the triamcinolone group. In the placebo run-out period, all patients whose treatment had not failed were switched to inhaled placebo, and most elements of asthma control predictably worsened in both active treatment groups. Notably, the duration of treatment benefit was not significantly longer in the triamcinolone group than in the other 2 groups. Fourteen (25%) of the patients assigned to receive placebo at the end of the 6-week triamcinolone run-in period met criteria for treatment failure within 6 weeks; 12 (24%) of the patients who received triamcinolone for 22 weeks also qualified as failures during the 6-week placebo run-out period. We thus found no evidence that an additional 16 weeks of continuous corticosteroid treatment conferred persistent benefit.

This study illustrates some limitations of traditional asthma outcome measures in clinical trials. Although treatment with salmeterol resulted in greater improvements in AM and PM PEF, asthma symptoms, and rescue albuterol use than treatment with placebo, the differences in the rates of treatment failure and asthma exacerbations between these 2 groups were small and insignificant. A similar dissociation between symptom control and asthma exacerbations was observed when salmeterol was given as monotherapy to children for 1 year. In that study, salmeterol was less effective than beclomethasone in preventing exacerbations.²⁹

The effect of long-acting β₂-agonists on asthmatic inflammation is controversial. In this study, airway inflammation, as inferred from analysis of induced sputum, decreased during triamcinolone treatment in the run-in period and increased in both the placebo and salmeterol groups during the randomized treatment period when triamcinolone was stopped. The authors of a recent small study proposed that treatment with long-acting β₂-agonists may mask worsening airway inflammation and delay awareness of worsening asthma.⁹ We observed an increase in markers of inflammation when patients were switched from triamcinolone to salmeterol. This change appears to reflect cessation of triamcinolone treatment rather than a direct salmeterol effect, since we saw a similar increase in inflammatory markers when patients were switched from triamcinolone to placebo in the run-out period. Other recent studies have shown no increase in markers of inflammation and a reduction in some inflammatory cell types in airway mucosal biopsy specimens after combined treatment with salmeterol and an ICS³⁰ or with the long-acting β₂-agonist formoterol alone.³¹

Our findings indicate that salmeterol should not be used as monotherapy for treatment of persistent asthma. Although salmeterol was highly effective at maintaining improvement in some conventional asthma outcome measures, including PEF, asthma symptoms, and rescue albuterol use, salmeterol was no more effective than placebo at preventing treatment failures and asthma exacerbations or at suppressing airway inflammation. These data support the concept that an ICS is preferable to a long-acting bronchodilator as monotherapy in patients with persistent asthma.