Long-Acting β₂-Agonist Step-off in Patients With Controlled Asthma: Systematic Review With Meta-analysis

Background Because of concerns about the safety of long-acting β₂-agonist (LABA) use in patients with asthma, withdrawal of the LABA is recommended by the US Food and Drug Administration once asthma is controlled by combination therapy with a LABA and inhaled corticosteroid (ICS).

Objective To perform a systematic review and meta-analysis assessing evidence supporting the discontinuation of LABA therapy once asthma control has been achieved with a combination of ICS and LABA.

Data Sources MEDLINE, EMBASE, and Cochrane Central Register of Controlled Trials databases were searched (through August 2010), references of identified studies and selected narrative review articles were evaluated, registries of clinical trials were reviewed, and manufacturers of LABAs were contacted.

Study Selection Randomized controlled trials of discontinuation of LABA therapy in patients with asthma controlled with a combination of ICS and LABA.

Data Extraction Two reviewers independently screened each title and abstract in the initial searches and then the full text of each nominated article to extract data for analyses.

Results Of 1492 screened articles, only 5 trials involving patients aged 15 years or older fulfilled a priori–specified inclusion criteria. Results did not favor the LABA step-off approach compared with no change in treatment. The LABA step-off regimen increased asthma impairment, with worse Asthma Quality of Life Questionnaire score (mean difference [95% CI], 0.32 [0.14-0.51] points lower); worse Asthma Control Questionnaire score (0.24 [0.13-0.35] points higher); fewer symptom-free days (9.15% [1.62%-16.69%] less); and greater risk of withdrawal from study resulting from lack of efficacy or loss of asthma control (risk ratio, 3.27 [2.16-4.96]). Risk of exacerbations and deaths after LABA step-off were not evaluable because of the small number of events and short duration of follow-up.

Conclusions Evidence suggests that discontinuing LABA therapy in adults and older children with asthma controlled with a combination of ICSs and LABAs results in increased asthma-associated impairment. Additional trials measuring all long-term patient-important outcomes are needed.

Recommendations for asthma therapy have been promulgated through the National Asthma Education and Prevention Program (NAEPP) guidelines, the most recent of which were published in 2007.¹ The NAEPP Expert Panel Report 3 (EPR-3) emphasized the importance of assessing asthma control to guide long-term therapy, whether stepping up or stepping down therapy. Asthma severity and control have 2 domains: (1) impairment (ie, severity of symptoms, functional limitations, reduced quality of life, and other manifestations of asthma) and (2) risk of exacerbations or decreased lung function. The primary goal of asthma therapy is to eliminate impairment and reduce the risk.

One major class of medications for asthma therapy is the long-acting β₂-agonists (LABAs). Quiz Ref IDIn response to safety concerns, the guidelines recommend that the drugs not be used as monotherapy for long-term control of persistent asthma but should continue to be considered adjunctive therapy to inhaled corticosteroids (ICSs). In making recommendations for preferred medications, the NAEPP EPR-3 tried to balance efficacy and safety for patients with various levels of asthma severity or control.¹ For patients with asthma not adequately controlled by low-dose ICSs, the EPR-3 gave equal weight to increasing the ICS dose and adding a LABA. The EPR-3 continued to endorse the use of a combination of ICS and LABA as the most effective therapy for patients with severe persistent asthma or asthma that was not well controlled by ICS monotherapy.¹

The EPR-3 recommended that consideration be given to stepping down therapy after asthma control is achieved for several months.¹ The timing and strategies of stepping down asthma care have not been well defined and are not well substantiated by data. Most studies focused on reducing the dose of ICSs while continuing other controller medications, including LABAs. This strategy aims to reduce the potential adverse effects of higher dosages of ICSs for children and adults. Appropriate step-down recommendations must be individualized.

In February 2010, the US Food and Drug Administration (FDA) announced a class labeling change for LABAs based on safety concerns.² The labeling change was prompted primarily by data from the Salmeterol Multicenter Asthma Research Trial (SMART), which raised concerns that the use of salmeterol xinafoate alone has associated risks for severe asthma outcomes.³ The data from the SMART suggested a higher risk for catastrophic asthma events, including hospitalizations, intubations, and deaths, in patients receiving salmeterol. The FDA subsequently published 4 recommendations for labeling changes for LABAs. The fourth recommendation is perhaps the most controversial and states: once asthma control is achieved and maintained, patients should be assessed at regular intervals, and step-down therapy should begin (eg, discontinue the LABA), if possible without loss of asthma control, and the patient should continue to be treated with a long-term asthma control medication, such as an inhaled corticosteroid.²

Subsequent to publication of the FDA recommendations, members of the EPR-3 wrote an editorial addressing these recommendations.⁴ Commenting especially on the fourth recommendation, they voiced a concern that the precipitous withdrawal of LABAs might put patients at risk for worsening asthma and that step-down approaches had not been sufficiently studied.

The American Academy of Allergy Asthma and Immunology and the American Thoracic Society formed a joint task force to address the issue of step-off strategies for patients whose asthma was controlled by combination ICS and LABA treatment. We searched for systematic reviews that addressed step-off strategies but found none. We therefore performed a systematic review and meta-analysis assessing the evidence supporting discontinuation of LABAs once asthma control is achieved. The clinical question is: Should LABAs be discontinued after achieving control of symptoms in patients with asthma who required combined therapy with an ICS and a LABA because of inadequate control of symptoms with an ICS alone?

Our primary objective was to evaluate the effects of the discontinuation of LABA once asthma control has been achieved with a combination of ICS and LABA. We developed and followed a protocol that included several criteria.

We developed a protocol and determined a priori to include randomized controlled trials that enrolled children aged 4 or more years and adults with asthma who required treatment with a combination of an ICS and LABA to control asthma symptoms. The intervention of interest was the discontinuation of LABA therapy and unchanged dosage of ICSs (LABA step-off) compared with continued use of unchanged dosages of LABAs and ICSs (no change in treatment) in patients who initially received a drug from each class to control symptoms. We reviewed studies that used salmeterol or formoterol fumarate dihydrate. We specified a priori that a target population of interest would have used the combination of an ICS and a LABA twice daily for at least 3 months and that asthma would have been well controlled for at least 3 months before discontinuation of the LABA.

The outcome measures did not constitute criteria for including studies in this review. We selected outcomes of interest following the approach suggested by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group.⁵ Outcomes that we identified a priori as potentially important for patients are listed in eTable 1. We rated their relative importance on a 9-point scale: critical, 7 to 9; important, 4 to 6; and of limited importance for decision making, 1 to 3.

We developed search strategies that were a combination of subject terms (eAppendix 1). We searched OVID MEDLINE (1950 through August week 5, 2010), OVID EMBASE (1980 through 2010 week 36), and The Cochrane Library, including Cochrane Central Register of Controlled Trials; Clinical Trials; Database of Abstracts of Reviews of Effects (DARE); and NHS (National Health Service) Economic Evaluation Database (until August 2010).

We also searched registries of clinical trials provided by the manufacturers of formoterol (http://www.astrazenecaclinicaltrials.com) and salmeterol (http://www.gsk-clinicalstudyregister.com). We reviewed the references of identified studies and selected narrative review articles on the topic. We also asked experts in the field and the manufacturers of salmeterol and formoterol for studies that we might have missed.

One investigator (J.L.B.) carried out the initial searches, and 2 investigators (J.L.B. and J.A.K.) independently reviewed titles and abstracts of all identified citations and assessed them for eligibility. Decisions were recorded and compared; in cases of disagreement about whether to include a study, we accepted it for full-text screening. Agreement between investigators about inclusion for full-text screening was moderate because one of the reviewers was generally more inclusive than the other (κ = 0.46; 95% CI, 0.42-0.50). Overall, 43 potentially relevant reports of studies were identified from the initial searches. Subsequently, 2 review authors independently screened the full-text articles for eligibility, using a standardized form with explicit inclusion and exclusion criteria. Agreement between review authors about study eligibility was very good (κ = 0.81; 95% CI, 0.50-1.13). Disagreements were resolved by discussion or by consulting a third review author (J.L.B.). Studies that did not fulfill the criteria^6-31 were excluded, and the reasons for exclusion were recorded (eAppendix 2).

We developed, piloted, and used a standardized data extraction form. One author extracted data from each included study, and the other systematically checked the correctness of data extraction. Disagreements were resolved by discussion. Details of the process are reported in eAppendix 3.

We assessed the risk of bias at the outcome level, using The Cochrane Collaboration's tool for assessing risk of bias.³² Subsequently, we assessed the quality of evidence (ie, confidence in the estimated effects) for each of the outcomes of interest, following the GRADE approach.³³ The small number of studies did not allow use of a funnel plot or statistical tests to detect the possibility of publication bias.

We calculated the agreement between the 2 reviewing authors for the assessment of trial eligibility using the κ statistic. We calculated risk ratio (RR) of dichotomous outcomes and rate ratios for count data for outcomes that each participant could experience more than once.³⁴ We used mean difference meta-analysis to combine the results for continuous or ordinal outcome data. Commercial software (Review Manager, version 5.1.1)³⁵ was used to conduct the meta-analyses.

We assessed heterogeneity between trials by visual inspection of forest plots, by estimation of the percentage of heterogeneity between trials that cannot be ascribed to sampling variation, and by a formal statistical test of the significance of the heterogeneity.³⁶ If there was evidence of substantial heterogeneity (ie, I² ≥50%), we investigated and reported possible explanations. We intended to explore heterogeneity based on a priori–specified characteristics of participants (age, 4-11, 12-17, and ≥18 years; receiving ICS plus LABA because of inadequate control of symptoms with ICSs alone vs receiving the combination for an unspecified reason; receiving ICSs at a dose of ≥200 μg/d of fluticasone propionate or an equivalent vs a lower dose; and duration of good control of asthma before the step-off phase of 3-5 months vs ≥6 months) and type of the experimental intervention (combination ICS and LABA from a single inhaler vs separate inhalers).

Our electronic searches identified 1497 distinct records; we excluded 1457 on the basis of title and abstract screening (Figure 1). Full-text articles of the remaining 40 articles were assessed for eligibility and 23 were excluded (eAppendix 2). We found 5 randomized controlled trials that examined LABA step-off in patients with asthma compared with no change in use of LABAs and ICSs (Table 1).^37-51 Four studies were published in peer-reviewed journals^37,39,42,47 and 1 was available only as a conference abstract⁵⁰ and a report in the manufacturer's online register of studies.⁵¹

All studies included adult or older adolescent patients with a 6-month or longer history of mild to moderate asthma (Table 1). All studies used a combination of ICS and LABA from the same inhaler as study medication. Studies allowed short-acting β₂-agonists as rescue medication; no study reported using short-acting anticholinergic agents. Based on the reported information, we were uncertain whether a combination treatment with ICS and LABA was required to control symptoms at study enrollment in 4 of 5 included studies. Also, the duration of asthma that was well controlled with use of the combined treatment before LABA step-off was shorter than the recommendation of at least 3 months.¹

Our assessment of the risk of bias in the 5 randomized trials included in the review is reported in eTable 2. Quiz Ref IDThere was no indication of the serious risk of bias in the studies except for incompleteness of follow-up: 12% to 38% of participants discontinued treatment and were excluded from some analyses. All studies were funded by the manufacturers of study medications.Table 2 presents numerical values for all analyses for all outcomes. Our assessment of the quality of available evidence (ie, confidence in the estimates of the effects) is given in eTable 3.

No study measured or reported the number of emergency department visits or unscheduled consultations for asthma, days missed from work or school, resource use (cost), and complications associated with the systemic effects of ICSs. No patient died, was admitted to the hospital, or required intubation or mechanical ventilation during the observation periods of the included studies.

There were no statistically significant results for any of the reported asthma outcomes of interest showing a benefit from LABA step-off approach compared with continued use of the same dose of ICS and LABA (no change in treatment) (Table 2). The frequency of adverse events not related to asthma was similar between patients undergoing LABA step-off and those with no change in treatment (RR, 0.93; 95% CI, 0.86-1.02) (eFigure 1). Serious adverse events occurred in 7 of 660 patients (1.1%) in the LABA step-off groups and in 9 of 682 patients (1.3%) in the control groups (RR, 0.79; 95% CI, 0.29-2.09). Similarly, 17 of 660 patients (2.6%) in the LABA step-off groups and 19 of 682 patients (2.8%) in the control groups withdrew from the studies because of an adverse event (RR, 0.92; 95% CI, 0.47-1.8).

Quiz Ref IDThe LABA step-off approach reduced the quality of life of patients (mean difference [95% CI] in Asthma Quality of Life Questionnaire score, 0.32 [95% CI, 0.14-0.51] points lower) as well as reduced control of asthma (mean difference in Asthma Control Questionnaire score, 0.24 [95% CI, 0.13-0.35] points higher).Quiz Ref IDDiscontinuation of LABAs also increased symptom frequency as measured by the proportion of symptom-free days (mean difference, 9.15% [95% CI, 1.62%-16.69%] fewer symptom-free days) and the risk of withdrawing from the study because of lack of efficacy or loss of asthma control (RR, 3.27 [95% CI, 2.16-4.96]). Patients who discontinued LABAs required a mean of 0.71 (95% CI, 0.29-1.14) more puffs per day of a rescue bronchodilator. The risk of need for oral corticosteroids was increased, but the increase was not statistically significant (RR, 1.68 [95% CI, 0.84-3.38]). Similarly, there were nonsignificant trends for increased risk of loss of asthma control as defined by the investigators (RR, 1.24 [95% CI, 0.79-1.95]) and a reduced proportion of awakening-free nights (mean difference, 1.47% fewer [95% CI, 3.18% fewer to 0.23% more]) (Table 2).

Individual forest plots of some of the asthma impairment measures rated by us as the most important for decision making are shown in Figure 2 and Figure 3. The remainder of the forest plots are shown in eFigure 2.

To our knowledge, this is the first systematic review and meta-analysis of randomized controlled trials investigating the effects of stepping off LABA use in patients with asthma controlled by combination ICS and LABA therapy. An interesting and important finding is the paucity of studies evaluating this issue. The results of the available studies suggest that discontinuation of LABAs compared with continued use of LABAs and ICSs increases the risk of loss of asthma control in adults. However, there is uncertainty about estimated effects because of the risk of bias in the included studies, imprecision of the estimates, and indirectness of the evidence (Table 2).

Limitations of our review include the methodological quality of the included randomized trials (eTable 2). Potential issues identified included lack of verification of the need for combination ICS and LABA treatment for the patients studied, the short duration of the studies and trial phases, lack of information on treatment adherence, and the relatively high withdrawal rates of patients in some studies. In addition, we did not conduct an explicit search for observational studies to assess the information about outcomes for which there was no evidence from randomized trials or those for which the quality of evidence from randomized trials was very low. We could not exclude the possibility of publication bias. With only 5 studies being available, techniques of estimating the likelihood of publication bias are not reliable. One might suspect publication bias in uniformly small studies showing favorable results and sponsored by the for-profit institution with an interest in the study results. However, we identified one industry-sponsored study that remained unpublished despite showing the same effects as the published studies; we did not identify any other study that would be registered but not published. Quiz Ref IDBecause the quality of the evidence about many critical outcomes was low to very low, any additional downgrading for possible publication bias would have little influence on our confidence in the estimated effects.

The strengths of this systematic review are the very broad search criteria and the paucity of missing information that was obtained directly from the study sponsors, making it unlikely that we missed any relevant information. Thus, our findings likely represent the current best evidence about stepping off LABA therapy in patients with asthma.

Concern over the safety of LABAs was initially raised in 1993 from a study conducted in the United Kingdom.⁵² The Serevent Nationwide Surveillance Study was a 14-week double-blind, randomized study that enrolled 25 180 individuals, with 16 787 receiving salmeterol twice daily and 8393 receiving albuterol sulfate (salbutamol sulfate) 4 times daily. At entry, more than 80% of the participants had moderate or severe persistent asthma, but less than 70% of patients were receiving ICSs. More asthma deaths occurred in patients receiving salmeterol (12 of 16 787) vs albuterol (2 of 8393), representing a nonstatistically significant increase in mortality (RR, 3.0; 95% CI, 0.7-20). However, those results were very imprecise because of the small number of events in both groups. Twelve of 14 participants who died were using more than 2 short-acting β₂-agonist canisters per month. The authors concluded that overuse of β₂-agonists and inadequate use of ICSs in patients with severe or unstable asthma, rather than LABA therapy, placed patients at greatest risk of asthma mortality.⁵²

These findings prompted the SMART, which was designed to assess the safety of salmeterol added to usual therapy compared with usual therapy alone.³ In this 7-month placebo-controlled, double-blind study, patients were evaluated by the study physician only once and received a 7-month supply of study medication. Subsequent contact was monthly by telephone from a central office. An interim analysis at 50% of the randomization goal demonstrated an increase in asthma mortality in patients randomized to receive salmeterol. Thirteen of 13 176 patients receiving salmeterol died compared with 3 patients of 13 179 who had received placebo (RR, 4.37; 95% CI, 1.25-15.34). Seven of 13 fatalities occurred in African Americans, who represented 18% of the cohort, suggesting a possibly higher risk in this population. The potential contribution of the use of ICSs was not measured. Although the study was not designed to evaluate whether concurrent ICS use modified the risk of asthma death, 7 participants reporting ICS use at entry died: 4 randomized to receive salmeterol and 3 randomized to receive placebo. Nine patients who were not receiving ICSs died; all were randomized to receive salmeterol. There were many flaws in this study, and patient follow-up was not typical of most controlled trials. Nonetheless, the results prompted a number of varied opinions about the potential risk-benefit ratio of LABAs for asthma. Several authors^53-55 concluded that the increase in unadjusted risk associated with LABAs may be the result of confounding by severity. Others⁵⁶ concluded if LABAs were removed from the market in the United States, we might witness a reduction in asthma mortality rates.

The NAEPP EPR-3 had these data prior to formulating the latest asthma guidelines but did not conclude that LABAs put patients at risk for catastrophic asthma events.¹ The FDA, however, directed manufacturers to place a black-box warning on all products that contain salmeterol or formoterol and expressed a concern about the safety of LABAs.²

Few new data about the safety of LABAs have been forthcoming, but investigators, including some from the FDA, have pooled data from previous clinical trials to further study the potential risks of LABAs.^57-69 These analyses have provided contradictory conclusions, putting clinicians in an unsettling role of trying to decide how to best step down therapy. Indeed, the present study was prompted by members of the American Academy of Allergy Asthma and Immunology and American Thoracic Society appealing to the leadership of these professional societies for guidance, especially in regard to how best to step down asthma treatment once asthma is controlled.

We found few randomized controlled trials designed to explore the effects of stepping off LABAs in patients whose asthma was well controlled. Because of the paucity of data, we were unable to assess the critical issue of asthma risk and whether LABA use had any effect on catastrophic asthma events. This issue has been of concern to the FDA and others.^2,56 Indeed, there is an FDA mandate for LABA safety studies by manufacturers of these agents involving approximately 50 000 patients of all ages with asthma. The results of these studies will not be available for approximately 5 years. In the interim, the consistent trends that we identified for many asthma impairment factors, some of which were statistically significant, favor the continued use of LABAs (Table 2). Thus, in contrast to FDA recommendations of stepping off LABA therapy when asthma is controlled, our analysis supports the continued use of LABAs to maintain asthma control. Although our results indicate that LABA step-off will, on average, lead to worsening of asthma control, it is possible that step-off can be safe in a subset of patients. Individual-level data from the studies conducted to date could help identify characteristics of these patients; such data were not available for this meta-analysis. There is clearly a need for more properly designed and executed randomized trials aimed at rectifying differences between asthma guideline recommendations and FDA safety concerns and guidance. These studies should include a comparison between stepping off LABAs and decreasing the dose of ICSs and be conducted in a variety of populations, since patients with differing severities of asthma might respond differently. Until those data are available, physicians need to evaluate the risk-benefit ratio of LABAs for their individual patients. Results of this review and many previous studies support the positive effects of LABAs for achieving and maintaining asthma control.

Correspondence: Thomas B. Casale, MD, Creighton University, 601 N 30th St, Ste 5850, Omaha, NE 68131 (tbcasale@creighton.edu).

Accepted for Publication: May 21, 2012.

Published Online: August 27, 2012. doi:10.1001/archinternmed.2012.3250

Author Contributions: Dr Brozek had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Brozek, Kraft, Krishnan, Lazarus, Li, Strunk, and Casale. Acquisition of data: Brozek, Kraft, Cloutier, Lazarus, Li, Santesso, Strunk, and Casale. Analysis and interpretation of data: Brozek, Kraft, Krishnan, Lazarus, Li, Strunk, and Casale. Drafting of the manuscript: Brozek, Kraft, Krishnan, Cloutier, Li, Strunk, and Casale. Critical revision of the manuscript for important intellectual content: Kraft, Cloutier, Lazarus, Li, Santesso, Strunk, and Casale. Statistical analysis: Brozek, Lazarus, and Li. Administrative, technical, and material support: Krishnan and Casale. Study supervision: Kraft and Casale.

Funding/Support: Dr Brozek holds a McMaster University Department of Medicine Internal Career Research Award. The American Academy of Allergy Asthma and Immunology and the American Thoracic Society provided staff support for conference calls and meetings.

Additional Contributions: We acknowledge the support and contributions of the American Academy of Allergy Asthma and Immunology and the American Thoracic Society.