Safety of Long-Acting β-Agonists: Are New Data Really Required?

Abstract

Despite 20 years of debate, several US Food and Drug Administration (FDA) hearings, black-box warnings, and many descriptive articles and metaanalyses, controversy regarding the safety of long-acting β-agonist (LABA) treatment in asthma patients continues. This has resulted in a recent call for another large and definitive safety study. This commentary focuses first on data provided in the metaanalysis recently undertaken by the FDA of safety outcomes among 60,954 individuals in 110 LABA trials, and second on the sample size that would be required for a new definitive study of LABA safety in the presence of mandatory treatment with an inhaled corticosteroid (ICS). A critical stratified analysis in the FDA report involving 15,192 individuals indicates that a LABA used with mandatory ICS therapy was not associated with an increased risk of asthma-related mortality, intubations, or exacerbations (risk difference [RD], 0.25 per 1,000 individuals; 95% confidence interval [CI], −1.69 to 2.18). Using the same stratified data to calculate the sample size required to prove or disprove an association between the use of LABA with mandatory ICS therapy and adverse outcomes, assuming the RD is exactly 0.25, and ignoring the 95% CI, which includes 0.0 or even a negative risk, such a study is both logistically and scientifically impossible. A new study is not practicable, nor is one needed in the light of current analyses of existing data. It is time to learn from the past, to rigorously avoid LABA monotherapy in asthma, and to use a LABA (when indicated) always in mandatory combination with appropriate doses of an ICS. Despite 20 years of debate, several US Food and Drug Administration (FDA) hearings, black-box warnings, and many descriptive articles and metaanalyses, controversy regarding the safety of long-acting β-agonist (LABA) treatment in asthma patients continues. This has resulted in a recent call for another large and definitive safety study. This commentary focuses first on data provided in the metaanalysis recently undertaken by the FDA of safety outcomes among 60,954 individuals in 110 LABA trials, and second on the sample size that would be required for a new definitive study of LABA safety in the presence of mandatory treatment with an inhaled corticosteroid (ICS). A critical stratified analysis in the FDA report involving 15,192 individuals indicates that a LABA used with mandatory ICS therapy was not associated with an increased risk of asthma-related mortality, intubations, or exacerbations (risk difference [RD], 0.25 per 1,000 individuals; 95% confidence interval [CI], −1.69 to 2.18). Using the same stratified data to calculate the sample size required to prove or disprove an association between the use of LABA with mandatory ICS therapy and adverse outcomes, assuming the RD is exactly 0.25, and ignoring the 95% CI, which includes 0.0 or even a negative risk, such a study is both logistically and scientifically impossible. A new study is not practicable, nor is one needed in the light of current analyses of existing data. It is time to learn from the past, to rigorously avoid LABA monotherapy in asthma, and to use a LABA (when indicated) always in mandatory combination with appropriate doses of an ICS. confidence interval Food and Drug Administration inhaled corticosteroid long-acting β-agonist odds ratio risk difference Controversy regarding the safety of treatment with long-acting β-agonists (LABAs) in patients with asthma has continued for nearly 20 years. Castle et al1Castle W Fuller R Hall J et al.Serevent nationwide surveillance study: comparison of salmeterol with salbutamol in asthmatic patients who require regular bronchodilator treatment.BMJ. 1993; 306: 1034-1037Crossref PubMed Scopus (425) Google Scholar conducted a large randomized, controlled trial in the United Kingdom comparing salmeterol administration twice daily with salbutamol administration four times daily in subjects who were considered to need regular β-agonist therapy. While exacerbations did not differ and study discontinuations were decreased with salmeterol treatment, there was a disconcerting nonsignificant threefold increase in the risk of mortality. Because of that study, the Food and Drug Administration (FDA) mandated a large phase IV safety trial when salmeterol was introduced into the United States.2Nelson HS Weiss ST Bleecker ER et al.The salmeterol multicenter asthma research trial: a comparison of usual pharmacotherapy for asthma or usual pharmacotherapy plus salmeterol.Chest. 2006; 129: 15-26Abstract Full Text Full Text PDF PubMed Scopus (1047) Google Scholar The outcome of the US study, a fourfold increased risk of death from asthma, most apparent in African Americans, added fuel to the debate. Post hoc analyses that suggested the risk was largely related to the lack of use of inhaled corticosteroids (ICSs) were not regarded as convincing, especially as use of ICS was recorded only at baseline and not monitored throughout the study. After review of the available evidence regarding LABA safety, the FDA issued a black-box warning in 2006 based on the reported risks of mortality and morbidity associated with trials of this class of medication in asthma. A review by the FDA Pediatric Advisory Committee in late 2007 further raised concerns regarding the safety of LABA use in children, and a full review of all available data was requested. In early 2008, in preparation for this review, the FDA asked each of the pharmaceutical companies marketing LABA products in the United States (AstraZeneca; Wilmington, DE; GlaxoSmithKline; Research Triangle Park, NC; and Novartis; East Hanover, NJ) to provide patient-based safety data from all blinded, parallel-arm, randomized, controlled trials conducted with LABAs in the treatment of asthma up to January 2008. The scope of the request included trials in which LABA was administered as randomized treatment, either with or without concomitant ICS or other adjunctive therapy, placebo-controlled, and/or active-controlled trials, and trials in which there was a randomized blinded phase followed by an open-label extension phase. For randomized, double-blind crossover design trials, only the first crossover period of the trial was included. Dr. Mark Levenson of the Quantitative Safety and Pharmacoepidemiology Group in the Division of Biostatistics and Office of Translational Science at the Center for Drug Evaluation and Research of the FDA undertook the requested metaanalysis on this large database3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar and presented the data to the joint meeting of three FDA advisory committees at a hearing held December 10 to 11, 2008. Using data from 110 trials involving 60,954 subjects, Levenson3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar calculated an overall statistically significant risk difference (RD) for LABA vs non-LABA therapy for asthma-related death of 0.40 (95% confidence interval [CI], 0.11 to 0.69) per 1,000 subjects (Table 1). For asthma-related death or intubation, the RD was 0.57 (95% CI, 0.01 to 1.12), and for asthma-related hospitalization the RD was 2.57 (95% CI, 0.90 to 4.23). For the composite outcome (asthma-related deaths, intubations, and hospitalizations), the calculated RD for LABA vs non-LABA therapy was 2.80 per 1,000 subjects (95% CI, 1.11 to 4.49).Table 1Risk Differences for LABA vs Non-LABA for Asthma Related-Deaths, Deaths and Intubations, Hospitalizations, and the Composite Outcome, Reported in the FDA MetaanalysisAll TrialsLABANon-LABARD (95% CI)Risk differences for specific and composite outcomes Asthma death16/30,1484/30,8060.40 (0.11–0.69) Death or intubation44/30,14827/30,8060.57 (0.01–1.12) Hospitalization369/30,148299/30,8062.57 (0.90–4.23) Composite outcome (death, intubation, or hospitalization)381/30,148304/30,8062.80 (1.11–4.49)Composite outcome stratified by use or nonuse of randomized (mandatory) ICS therapy Trials of LABA without randomized ICS vs no LABA350/22,286279/24,4743.63 (1.51–5.75) Trials of LABA with randomized ICS vs randomized ICS therapy31/7,86226/7,3300.25 (−1.69–2.18)Values are given as No. of events/No. of patients at risk, unless otherwise indicated. RD = additional risk of outcome per 1000 subjects treated with LABA compared with no LABA; 0.0 indicates no increased risk. The data are from Table 5 and Figure 1 of Levenson.3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar Open table in a new tab Values are given as No. of events/No. of patients at risk, unless otherwise indicated. RD = additional risk of outcome per 1000 subjects treated with LABA compared with no LABA; 0.0 indicates no increased risk. The data are from Table 5 and Figure 1 of Levenson.3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar In the most relevant stratified subanalyses, the RD for patients receiving LABA without mandatory concomitant randomized ICS was 3.63 (95% CI, 1.51 to 5.75) per 1,000 subjects. In sharp contrast, patients receiving LABA therapy with mandatory concomitant randomized ICS showed a nonsignificant RD of 0.25 (95% CI, −1.69 to 2.18) per 1,000 subjects (Table 1). Levenson3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar also examined outcomes stratifying trials by the prescription of ICSs at baseline and by ICS use during the trial (either randomized ICS use or concomitant but not randomized ICS therapy). The results were very different from the stratified analyses with mandatory randomized ICS noted above. The RD for the composite outcome among those with a baseline ICS prescription was 4.10 (95% CI, 1.67 to 6.53) compared with 2.56 (95% CI, 0.16 to 4.97) among those without a baseline ICS prescription. This result is counterintuitive, unless viewed as evidence of confounding by severity, namely that those patients to whom ICSs had been prescribed at baseline had more severe disease, leading their physician to prescribe ICSs. In the second of these analyses, examining concomitant (but not necessarily randomized) ICS use during the trial, a similar result was seen; the RD for those patients using ICSs during the trial was 3.98 (95% CI, 1.42 to 6.55) compared with 2.80 (95% CI, 0.55 to 5.06) for those not using concomitant ICSs. The sharp contrast between these results and the RD for LABA vs non-LABA therapy among patients using mandatory (randomized) ICSs is only explicable if the subjects not mandated by randomization to ICS were highly noncompliant with concomitant ICS therapy. Further details supporting this perspective are provided in the appendix of the FDA report (Table 13, page 43),3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar which indicates that almost the entire risk of the most severe outcomes (deaths and intubations) was associated with use of LABA without mandatory concomitant ICS therapy. Of 44 deaths and intubations in the LABA-exposed population, 43 occurred among 22,286 individuals (0.19%) in trials that did not mandate the use of ICSs compared with 1 among 7,862 individuals (0.01%) in trials with mandatory ICS therapy. No deaths or intubations were reported for patients treated with single-device combinations of ICSs and LABAs, whether salmeterol/fluticasone or formoterol/budesonide. The logical conclusion from these findings is that the safe use of LABAs can only be assured if subjects receiving a LABA are mandated to receive an ICS. Given the substantial evidence that patients with asthma use β-agonists in preference to ICSs, this means in practice a restriction of the availability of LABAs to combination products in which every dose of the LABA is accompanied by an ICS. Although the findings of the stratified metaanalysis by Levenson3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar would be considered by many to provide strong evidence that LABA used with mandatory concomitant ICS is safe, some believe that more data are required to resolve the issue. In particular, Drazen and O'Byrne,4Drazen JM O'Byrne PM Risks of long-acting β-agonists in achieving asthma control.N Engl J Med. 2009; 360: 1671-1672Crossref PubMed Scopus (87) Google Scholar commenting on the summary report by Kramer5Kramer JM Balancing the benefits and risks of inhaled long-acting β-agonists: the influence of values.N Engl J Med. 2009; 360: 1592-1595Crossref PubMed Scopus (97) Google Scholar on the FDA advisory committee meeting, have issued a call for a very large randomized clinical study addressing the safety of LABAs, in the anticipation that such a study would provide a definitive answer to the questions that continue to be asked. Drazen and O'Byrne4Drazen JM O'Byrne PM Risks of long-acting β-agonists in achieving asthma control.N Engl J Med. 2009; 360: 1671-1672Crossref PubMed Scopus (87) Google Scholar rightly state that, given present knowledge and current asthma guidelines, the appropriate comparison would be of outcomes among patients receiving LABAs with ICS therapy vs patients receiving the same dose of an ICS without a LABA. If further assessment of the safety of LABAs is to be ascertained in a large study, it is appropriate to first ask what sample size would be needed? Fortunately, the comprehensive report of Levenson3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar of the analysis of 110 trials provides the information required for such calculations. Based on these data, it is clear that such a study would require impossibly large numbers or be doomed to repeat the past with insufficient data to truly “clear the air.”6Beasley R Martinez FD Hackshaw A et al.Safety of long-acting β-agonists: urgent need to clear the air remains.Eur Respir J. 2009; 33: 3-5Crossref PubMed Scopus (41) Google Scholar Sample-size calculations for a new study with mandatory ICS therapy can be based on the RD of 0.25 per 1,000 for the composite outcome reported by Levenson3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar among patients mandated to receive concomitant ICS therapy (Table 1). This RD of 0.25 per 1,000 means 1 patient in 4,000 using a LABA with an ICS is at excess risk (in comparison with patients using an ICS without a LABA) of one of the composite outcomes (asthma-related death, intubation, or hospitalization). Each outcome or combination of outcomes can be considered as a candidate for the primary outcome of a new large study. Of all composite outcomes (n = 685 with or without ICS), 20 were deaths (Table 1), meaning death as an outcome occurred 2.9 times in every 100 composite outcomes. If all reported deaths occurred among patients receiving mandatory ICS therapy, then 1 excess death would occur in 137,000 such patients (4,000 × 685/20). However, 15 of 16 deaths in the LABA-exposed population in the full FDA data set occurred in patients not randomized to receive ICS therapy; therefore, this calculation underestimates by a factor of 16 the numbers needed in a trial comparing LABA use with ICS therapy vs ICS therapy alone. The number of subjects required to demonstrate 1 excess death is therefore 137,000 × 16 exposed to LABA with mandatory ICS, yielding a total sample size (assuming 1:1 randomization) of 4,384,000. There were 71 asthma-related deaths and intubations among the 685 composite outcomes (Table 1) or 10.3 such events in every 100 composite outcomes. If all patients experiencing these outcomes were receiving mandatory ICS therapy, 1 excess death or intubation would be seen in 38,591 patients (4,000 × 685/71). However, again, the great majority of events (43 of 44 in LABA-exposed patients) occurred in subjects not using mandatory randomized ICSs, although 19 and 23, respectively, are said to have received baseline ICS therapy or received concomitant ICS therapy throughout the trial. Even if we generously assume that 50% of those experiencing death or intubation in the FDA database were prescribed and were actually receiving an ICS, 1 excess death or intubation would be seen among 77,182 subjects receiving ICS therapy. For a clinically relevant trial outcome, for example, 5 excess deaths or intubations, 385,910 subjects would be required, assuming the true RD is exactly 0.25 per 1,000. However, the 95% CI of this estimate was −0.69 to 2.18 per 1,000 patients. If the true RD is only slightly less, for example, 0.15 per 1,000 patients, then > 640,000 subjects would be needed to identify 5 excess deaths and intubations. Given the CI, the true RD could be closer to or even below 0.0, in which case the number required for a study escalates enormously, including to infinity of the RD is in fact 0.0, which is statistically quite possible. So, a study examining deaths and intubations as the primary outcome also appears not to be feasible. Finally, could the FDA composite measure of deaths, intubations, and hospitalizations be used as an appropriate outcome for this proposed large trial? This would reduce the required sample size substantially, but would introduce further difficulties in interpretation. Several published metaanalyses, each using a subset of the same data used by the FDA, have shown that LABAs used with concomitant randomized ICS therapy are either neutral in risk for exacerbations (dominantly hospitalizations) or in fact reduce the risk. Jaeschke et al7Jaeschke R O'Byrne PM Mezja F et al.The safety of long-acting β2-agonists among patients with asthma using inhaled corticosteroids: systematic review and meta-analysis.Am J Respir Crit Care Med. 2008; 178: 1009-1016Crossref PubMed Scopus (99) Google Scholar examined the risks for the use of formoterol and salmeterol with randomized ICS therapy (62 trials; 15,710 patients exposed to one or other LABA) and reported an odds ratio (OR) for hospitalizations of 0.74 (95% CI, 0.53 to 1.03). Bateman et al8Bateman E Nelson H Bousquet J et al.Meta-analysis: effects of adding salmeterol to inhaled corticosteroids on serious asthma-related events.Ann Intern Med. 2008; 149: 33-42Crossref PubMed Scopus (106) Google Scholar examined GlaxoSmithKline data for salmeterol use with ICS therapy (66 trials, 20,966 patients) and reported an OR for hospitalizations of 1.07 (95% CI, 0.66 to 1.73). Jaeschke et al9Jaeschke R O'Byrne PM Nair P et al.The safety of formoterol among patients with asthma using inhaled corticosteroids: systematic review and meta-analysis.Pol Arch Med Wewn. 2008; 118: 627-635PubMed Google Scholar examined formoterol use with ICS in AstraZeneca trials (16 trials; 5,996 formoterol-exposed patients) and found a significantly decreased risk of hospitalizations (OR, 0.59; 95% CI, 0.37 to 0.93). Cates et al10Cates CJ Lasserson TJ Jaeschke R Regular treatment with formoterol and inhaled steroids for chronic asthma: serious adverse events.Cochrane Database Syst Rev (database online). 2009; Google Scholar performed a Cochrane metaanalysis of formoterol with mandatory ICS therapy and reported among adults a reduced risk of serious adverse events (Peto OR, 0.53; 95% CI, 0.28 to 1.00), while among children the Peto OR was 1.49 (95% CI, 0.48 to 4.61) and not statistically significant. Hence, using the FDA composite outcome including hospitalizations in a large trial comparing LABA with mandatory ICS therapy vs ICS therapy alone would almost certainly move the outcome in the direction opposite to that suggested by the worst-case interpretation of the overall FDA risk assessment.3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar The future treatment of asthma in the United States and elsewhere should not be hampered by a focus on unfortunate previous management strategies.1Castle W Fuller R Hall J et al.Serevent nationwide surveillance study: comparison of salmeterol with salbutamol in asthmatic patients who require regular bronchodilator treatment.BMJ. 1993; 306: 1034-1037Crossref PubMed Scopus (425) Google Scholar, 2Nelson HS Weiss ST Bleecker ER et al.The salmeterol multicenter asthma research trial: a comparison of usual pharmacotherapy for asthma or usual pharmacotherapy plus salmeterol.Chest. 2006; 129: 15-26Abstract Full Text Full Text PDF PubMed Scopus (1047) Google Scholar The use of LABAs without mandatory ICS therapy is, or should be, history. Patients should be treated according to current guidelines, as endorsed by the FDA in 2006 and recently by three of its advisory committees, namely by the appropriate use of a LABA added to ICS therapy when control is not achieved with moderate doses of an ICS. It is highly unlikely that any further metaanalyses of existing data will provide substantially different answers to those already reported that consistently confirm the safety of a LABA used with mandatory ICS therapy.3Levenson M Long-acting β-agonists and adverse asthma events meta-analysis: statistical briefing package for joint meeting of the Pulmonary-Allergy Drugs Advisory Committee, Drug Safety and Risk Management Advisory Committee and Pediatric Advisory Committee, December 10–11, 2008.Available at: http://www.fda.gov/ohrms/dockets/ac/cder08.html#PulmonaryAllergyGoogle Scholar, 7Jaeschke R O'Byrne PM Mezja F et al.The safety of long-acting β2-agonists among patients with asthma using inhaled corticosteroids: systematic review and meta-analysis.Am J Respir Crit Care Med. 2008; 178: 1009-1016Crossref PubMed Scopus (99) Google Scholar, 8Bateman E Nelson H Bousquet J et al.Meta-analysis: effects of adding salmeterol to inhaled corticosteroids on serious asthma-related events.Ann Intern Med. 2008; 149: 33-42Crossref PubMed Scopus (106) Google Scholar, 9Jaeschke R O'Byrne PM Nair P et al.The safety of formoterol among patients with asthma using inhaled corticosteroids: systematic review and meta-analysis.Pol Arch Med Wewn. 2008; 118: 627-635PubMed Google Scholar, 10Cates CJ Lasserson TJ Jaeschke R Regular treatment with formoterol and inhaled steroids for chronic asthma: serious adverse events.Cochrane Database Syst Rev (database online). 2009; Google Scholar A large trial such as that recommended by Drazen and O'Byrne4Drazen JM O'Byrne PM Risks of long-acting β-agonists in achieving asthma control.N Engl J Med. 2009; 360: 1671-1672Crossref PubMed Scopus (87) Google Scholar is neither practicable nor necessary and will not provide any more useful data regarding asthma-related adverse events than what is presently available. It is time to learn from the past, to rigorously avoid LABA monotherapy, and to use LABAs (when indicated) always in mandatory combination with appropriate doses of ICSs. This will allow us to safely provide our patients with the extensively documented and substantial clinical benefit of combination therapy in patients with moderate-to-severe asthma. Financial/nonfinancial disclosures: The author has reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article. Erratum: CHEST 2009; 136(2):604–607CHESTVol. 136Issue 6PreviewThe author regrets that in the August 2009 issue, the Commentary “Safety of Long-Acting β-Agonists: Are New Data Really Required?” (CHEST 2009; 136(2):604–607) by Sears data given in the second paragraph in the right column of page 605 were transposed. The paragraph should read as follows: Full-Text PDF