Inhaled steroids with and without regular salmeterol for asthma: serious adverse events
Abstract
Background
Epidemiological evidence has suggested a link between use of beta₂‐agonists and increased asthma mortality. Much debate has surrounded possible causal links for this association, and whether regular (daily) long‐acting beta₂‐agonists (LABAs) are safe, particularly when used in combination with inhaled corticosteroids (ICSs). This is an update of a Cochrane Review that now includes data from two large trials including 11,679 adults and 6208 children; both were mandated by the US Food and Drug Administration (FDA).
Objectives
To assess risks of mortality and non‐fatal serious adverse events (SAEs) in trials that randomised participants with chronic asthma to regular salmeterol and ICS versus the same dose of ICS.
Search methods
We identified randomised trials using the Cochrane Airways Group Specialised Register of trials. We checked websites of clinical trials registers for unpublished trial data. We also checked FDA submissions in relation to salmeterol. The date of the most recent search was 10 October 2018.
Selection criteria
We included parallel‐design randomised trials involving adults, children, or both with asthma of any severity who were randomised to treatment with regular salmeterol and ICS (in separate or combined inhalers) versus the same dose of ICS of at least 12 weeks in duration.
Data collection and analysis
We conducted the review according to standard procedures expected by Cochrane. We obtained unpublished data on mortality and SAEs from the sponsors, from ClinicalTrials.gov, and from FDA submissions. We assessed our confidence in the evidence according to current GRADE recommendations.
Main results
We have included in this review 41 studies (27,951 participants) in adults and adolescents, along with eight studies (8453 participants) in children. We judged that the overall risk of bias was low for all‐cause events, and we obtained data on SAEs from all study authors. All except 542 adults (and none of the children) were given salmeterol and fluticasone in the same (combination) inhaler.
Deaths
Eleven of a total of 14,233 adults taking regular salmeterol and ICS died, as did 13 of 13,718 taking regular ICS at the same dose. The pooled Peto odds ratio (OR) was 0.80 (95% confidence interval (CI) 0.36 to 1.78; participants = 27,951; studies = 41; I² = 0%; moderate‐certainty evidence). In other words, for every 1000 adults treated for 25 weeks, one death occurred among those on ICS alone, and the corresponding risk among those taking salmeterol and ICS was also one death (95% CI 0 to 2 deaths).
No children died, and no adults or children died of asthma, so we remain uncertain about mortality in children and about asthma mortality in any age group.
Non‐fatal serious adverse events
A total of 332 adults receiving regular salmeterol with ICS experienced a non‐fatal SAE of any cause, compared to 282 adults receiving regular ICS. The pooled Peto OR was 1.14 (95% CI 0.97 to 1.33; participants = 27,951; studies = 41; I² = 0%; moderate‐certainty evidence). For every 1000 adults treated for 25 weeks, 21 adults on ICS alone had an SAE, and the corresponding risk for those on salmeterol and ICS was 23 adults (95% CI 20 to 27).
Sixty‐five of 4229 children given regular salmeterol with ICS suffered an SAE of any cause, compared to 62 of 4224 children given regular ICS. The pooled Peto OR was 1.04 (95% CI 0.73 to 1.48; participants = 8453; studies = 8; I² = 0%; moderate‐certainty evidence). For every 1000 children treated for 23 weeks, 15 children on ICS alone had an SAE, and the corresponding risk for those on salmeterol and ICS was 15 children (95% CI 11 to 22).
Asthma‐related serious adverse events
Eighty and 67 adults in each group, respectively, experienced an asthma‐related non‐fatal SAE. The pooled Peto OR was 1.15 (95% CI 0.83 to 1.59; participants = 27,951; studies = 41; I² = 0%; low‐certainty evidence). For every 1000 adults treated for 25 weeks, five receiving ICS alone had an asthma‐related SAE, and the corresponding risk among those on salmeterol and ICS was six adults (95% CI 4 to 8).
Twenty‐nine children taking salmeterol and ICS and 23 children taking ICS alone reported asthma‐related events. The pooled Peto OR was 1.25 (95% CI 0.72 to 2.16; participants = 8453; studies = 8; I² = 0%; moderate‐certainty evidence). For every 1000 children treated for 23 weeks, five receiving an ICS alone had an asthma‐related SAE, and the corresponding risk among those receiving salmeterol and ICS was seven children (95% CI 4 to 12).
Authors' conclusions
We did not find a difference in the risk of death or serious adverse events in either adults or children. However, trial authors reported no asthma deaths among 27,951 adults or 8453 children randomised to regular salmeterol and ICS or ICS alone over an average of six months. Therefore, the risk of dying from asthma on either treatment was very low, but we remain uncertain about whether the risk of dying from asthma is altered by adding salmeterol to ICS.
Inclusion of new trials has increased the precision of the estimates for non‐fatal SAEs of any cause. We can now say that the worst‐case estimate is that at least 152 adults and 139 children must be treated with combination salmeterol and ICS for six months for one additional person to be admitted to the hospital (compared to treatment with ICS alone). These possible risks still have to be weighed against the benefits experienced by people who take combination treatment.
However more than 90% of prescribed treatment was taken in the new trials, so the effects observed may be different from those seen with salmeterol in combination with ICS in daily practice.
PICOs
Plain language summary
Inhaled steroids with and without regular salmeterol for asthma: serious adverse events
Review question
Is it safe to add regular salmeterol to inhaled corticosteroid (ICS) for adults or children with asthma?
Background
Another Cochrane Review found that using regular salmeterol without regular ICS for adults with asthma led to an increase in serious adverse events (death or admission to hospital). We wanted to find out if more adverse (harmful) effects occur when people take regular salmeterol in addition to ICS. We looked only at adverse effects ‐ deaths, being admitted to hospital and life‐threatening effects. We did not look at the benefits of taking salmeterol for other outcomes. We updated this review in 2018 because of new evidence from large randomised trials of salmeterol in combination with ICS, in 11,679 adults and 6208 children with asthma.
Study characteristics
In total, we have included 41 studies in 27,951 adults and eight studies in 8453 children. Almost all studies used a combination inhaler to deliver salmeterol with ICS and compared this with the same dose of ICS for an average of six months.
Key results
We did not find a difference in the risk of death or serious adverse events in either adults or children.
Eleven of a total of 14,233 adults taking regular salmeterol and ICS died, as did 13 of 13,718 adults taking regular ICS at the same dose. For every 1000 adults treated for 25 weeks, researchers reported one death on ICS alone and a corresponding risk on salmeterol and ICS of one death (95% confidence Interval (CI) 0 to 2 deaths). No deaths in any studies were attributed to asthma, and researchers reported no deaths at all among children.
A non‐fatal serious adverse event of any cause occurred in 332 adults on regular salmeterol with ICS compared to 282 adults on regular ICS alone. For every 1000 adults treated for 25 weeks, 21 serious adverse events occurred on ICS alone, and the corresponding risk on salmeterol and ICS was 23 adults (95% CI 20 to 27).
A total of 65 of 4229 children on regular salmeterol with ICS suffered a serious adverse event of any cause compared to 62 of 4224 children on regular ICS alone. For every 1000 children treated for 23 weeks, 15 serious adverse events occurred on ICS alone, and the corresponding risk on salmeterol and ICS was also 15 children (95% CI 11 to 22).
Quality of the evidence
Reviewers assessed the overall risk of bias for all‐cause events as low. The two new large studies performed independent assessment to identify the cause of asthma‐related serious adverse events. This makes current data on asthma events more reliable than previously reported.
Conclusions
Trials reported no asthma deaths among 27,951 adults or 8453 children randomised to regular salmeterol and ICS or ICS alone over an average of six months. The risk of dying from asthma while receiving either treatment was therefore very low, but we remain uncertain about whether the risk of dying from asthma is altered by adding salmeterol to ICS.
We can now say that the worst‐case estimate (safety margin) from this review is that at least 152 adults and 139 children must be treated with combination salmeterol and ICS for six months for one additional person to be admitted to the hospital (compared to ICS alone). These possible risks must be weighed against the benefits experienced by people who take combination treatment.
People monitored in the new trials took over 90% of their prescribed treatment. This is much more than the average amount of medication that people take outside a trial. Therefore the effects shown in trials may be different from the effects experienced by people at home who are not taking their inhalers as prescribed.
Because very few people die of asthma, trials would have to be very large to detect differences in the death rate. Therefore it is probably not feasible to find out if adding salmeterol to ICS causes more deaths among participants in randomised controlled trials ‐ as these trials would be very large, difficult to run, and expensive. It might be better to use case‐control studies or to review asthma deaths (e.g. from medical records).
Authors' conclusions
Summary of findings
| Inhaled steroids with and without regular salmeterol for asthma | ||||||
| Patient or population: adults and children with chronic asthma Settings: community Intervention: regular salmeterol in addition to regular inhaled corticosteroid (ICS) Comparison: regular ICS (at the same dose) | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No. of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Regular ICS | Regular salmeterol in addition to regular ICS | |||||
| Adults and adolescents | ||||||
| All‐cause mortality | 1 per 1000a | 1 per 1000 | OR 0.80 | 27951 | ⊕⊕⊕⊝ | |
| All‐cause non‐fatal SAEc | 21 per 1000a | 23 per 1000 | OR 1.14 | 27951 | ⊕⊕⊕⊝ | |
| Asthma‐related mortality Follow‐up: mean 25 weeks | No deaths due to asthma | No deaths due to asthma | ‐ | 27951 | ⊕⊕⊕⊝ | Pooled risk difference zero (95% CI ‐0.0009 to 0.0009) |
| Asthma‐related non‐fatal SAEc | 5 per 1000a | 6 per 1000 | OR 1.15 | 27951 | ⊕⊕⊝⊝ | |
| Children | ||||||
| All‐cause mortality Follow‐up: mean 23 weeks | No deaths | No deaths | ‐ | 8453 | ⊕⊕⊕⊝ | Pooled risk difference zero (95% CI ‐0.0013 to 0.0013) |
| All‐cause non‐fatal SAEc | 15 per 1000a | 15 per 1000 | OR 1.04 | 8453 | ⊕⊕⊕⊝ | |
| Asthma‐related mortality Follow‐up: mean 23 weeks | No deaths | No deaths | ‐ | 8453 | ⊕⊕⊕⊝ | Pooled risk difference zero (95% CI ‐0.0013 to 0.0013) |
| Asthma‐related non‐fatal SAEc | 5 per 1000a | 7 per 1000 | OR 1.25 | 8453 | ⊕⊕⊕⊝ | |
| *The basis for the assumed risk is the mean control event rate in the included studies. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; ICS: inhaled corticosteroid; OR: odds ratio; SAE: serious adverse event. | ||||||
| GRADE Working Group grades of evidence. High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. | ||||||
| aMean control event rate. bImprecision (‐1). We regard the upper end of the 95% confidence interval as too high to rule out a potentially important increase. See also Table 3 and Appendix 5. cNon‐fatal SAEs were defined as life‐threatening adverse events, inpatient hospitalisations or prolongation of existing hospitalisations, persistent or significant disabilities or incapacities, or congenital anomalies or birth defects. dMore than half of the events came from trials that did not independently assess the causation of SAEs. | ||||||
Background
Description of the condition
For people whose asthma is not controlled by low‐dose inhaled corticosteroids (ICSs) alone, many asthma guidelines recommend adding long‐acting beta₂‐agonists (LABAs). Several Cochrane Reviews have addressed the efficacy of LABA in addition to ICS (Ni Chroinin 2004; Ni Chroinin 2005), in comparison with placebo (Walters 2007), short‐acting beta₂‐agonists (SABAs; Walters 2002), leukotriene receptor antagonists (LTRAs; Ducharme 2006), and increased doses of ICS (Greenstone 2005). The beneficial effects of LABA on lung function, symptoms, quality of life, and exacerbations requiring oral steroids have been demonstrated.
However, long‐standing controversy surrounds the regular use of beta₂‐agonists for people with asthma. Sears 1986 suggested that excessive use of SABAs might have contributed directly or indirectly to an increase in asthma deaths in New Zealand between 1960 and 1980. Study authors comment that "most deaths were associated with poor assessment, underestimation of severity and inappropriate treatment (over‐reliance on bronchodilators and under use of systemic corticosteroids), and delays in obtaining help".
Concern remains that the symptomatic benefit resulting from treatment with LABA might lead to underestimation of attack severity in acute asthma, and could lead to an increase in asthma‐related deaths. Furthermore, regular treatment with beta₂‐agonists can lead to tolerance to their bronchodilator effects, and this phenomenon may become more marked with longer‐acting, as opposed to shorter‐acting, compounds (Lipworth 1997). Several molecular mechanisms have been proposed to explain the possible detrimental effects of long‐term beta₂‐agonist use in asthma, including receptor down‐regulation and desensitisation (Giembycz 2006).
A meta‐analysis of the effects of LABAs on severe asthma exacerbations and asthma‐related deaths concluded that "long‐acting beta‐agonists have been shown to increase severe and life‐threatening asthma exacerbations, as well as asthma‐related deaths" (Salpeter 2006). However, these researchers considered only trials that compared LABA versus placebo, and 28 of the included trials on 6000 participants did not report asthma‐related deaths, reducing certainty in the strength of review author conclusions.
Description of the intervention
Salmeterol and formoterol are examples of LABAs that are available to treat asthma. These two drugs are known to have differences in receptor activity, and they are used in different ways (e.g. salmeterol has a slower onset of action than salbutamol and formoterol, and is therefore unsuitable for use as a reliever) (Beach 1992). "The Fenoterol Story" is a reminder that all beta₂‐agonists may not carry the same risks (Pearce 2007), so in view of potential differences in adverse effects between salmeterol and formoterol, we have considered the two drugs separately.
How the intervention might work
Since the publication of SMART 2006, much debate has surrounded the interaction between ICS and LABA in relation to serious adverse events (SAEs). This study did not randomise participants to ICS but nevertheless conducted a subgroup analysis of results on the basis of ICS use at baseline. It is tempting to be reassured by the fact that researchers did not find a statistically significant increase in asthma‐related mortality in the subgroup using ICS, but this is not the correct way to test for interaction (Altman 2003), and investigators carried out no assessment during the trial in relation to actual use of ICS during the study. There is a need to systematically review all available data from controlled trials that randomised participants to regular salmeterol in combination with ICS, and to consider all SAEs (fatal and non‐fatal), whether or not they are deemed by investigators to be related to trial medication.
Why it is important to do this review
The focus of this review is on regular salmeterol that has been randomised in combination with ICS (in a single inhaler or in separate inhalers). Due to the difficulty involved in deciding whether adverse events are asthma‐related (particularly in the many studies that do not perform independent outcome assessment of adverse events), the authors of this review are concerned with studies that capture mortality and SAEs, and we have recorded both all‐cause outcomes and those considered by trial investigators to be asthma‐related events. This approach differs from that reported in Bateman 2008, in which review authors restricted outcomes to asthma‐related events.
Regular salmeterol alone is the topic of a previous review (Cates 2008a), as is regular formoterol alone (Cates 2012b). In both of these reviews, review authors demonstrated an increase in SAEs with regular LABA. Another review considered formoterol with ICS (Cates 2009a). Review authors have also provided overviews of the safety of combination therapy in children and in adults (Cates 2012a; Cates 2014, respectively).
Due to ongoing concern over the safety of LABAs, the FDA mandated large trials of salmeterol and ICS in both adults and children. We have included results of these studies in this updated review. The two large new studies were designed to test whether salmeterol was safe when added to ICS; to do this, trial authors used a safety margin derived from the worst end of the 95% confidence interval of the comparative risk for a combined endpoint in these trials (death, intubation, or hospital admission). In adults, the safety margin was a hazard ratio of 2.0; in children, the hazard ratio was 2.675.
Objectives
To assess risks of mortality and non‐fatal SAEs in trials that randomised participants with chronic asthma to regular salmeterol and ICS versus the same dose of ICS.
Methods
Criteria for considering studies for this review
Types of studies
We sought randomised controlled trials (RCTs), with or without blinding, in which researchers randomly assigned salmeterol and ICS to participants with chronic asthma.
Types of participants
We included participants of any age with a clinical diagnosis of asthma, unrestricted by disease severity or previous or current treatment. We did not include studies on acute asthma and exercise‐induced bronchospasm.
Types of interventions
We searched for trials in which investigators prescribed ICS and salmeterol regularly for a period of at least 12 weeks, at any daily dose, and delivered by any single or separate devices (i.e. chlorofluorocarbon metered dose inhaler (CFC‐MDI), hydrofluoroalkane metered dose inhaler (HFA‐MDI), or dry powder inhaler (DPI)). We included studies that used comparison groups given the same dose of ICS; co‐interventions with LTRAs, cromones, oral corticosteroids (OCSs), or theophylline were allowed as long as they were not part of the randomised intervention. We excluded studies that compared different doses of salmeterol or different delivery devices or propellants (with no placebo arm), or that compared salmeterol versus formoterol. We excluded from this review studies that randomised salmeterol without an inhaled steroid; a separate Cochrane Review has considered these studies (Cates 2008a).
Types of outcome measures
Primary outcomes
-
All‐cause mortality
-
All‐cause non‐fatal serious adverse events
Secondary outcomes
-
Asthma‐related mortality
-
Asthma‐related non‐fatal serious adverse events
-
Respiratory‐related mortality
-
Respiratory‐related non‐fatal serious adverse events
-
Cardiovascular‐related mortality
-
Cardiovascular‐related non‐fatal serious adverse events
-
Asthma‐related non‐fatal life‐threatening events (intubation or admission to intensive care)
-
Respiratory‐related non‐fatal life‐threatening events (intubation or admission to intensive care)
Search methods for identification of studies
Electronic searches
We searched the Cochrane Airways Trials Register up to 10 October 2018, with no restrictions on language or type of publication. The Cochrane Airways Trials Register is maintained by the information specialist for Cochrane Airways and contains studies identified from the following sources.
-
Monthly searches of the Cochrane Central Register of Controlled Trials (CENTRAL), in the Cochrane Library, through the Cochrane Register of Studies (CRS).
-
Weekly searches of MEDLINE Ovid SP.
-
Weekly searches of Embase Ovid SP.
-
Monthly searches of PsycINFO Ovid SP.
-
Monthly searches of the Cumulative Index to Nursing and Allied Health Literature (CINAHL) EBSCO.
-
Monthly searches of Allied and Complementary Medicine (AMED) EBSCO.
-
Handsearches of the proceedings of major respiratory conferences.
We identified studies contained in the Trials Register using search strategies based on the scope of Cochrane Airways. We have provided details of these strategies, as well as a list of handsearched conference proceedings, in Appendix 3. See Appendix 4 for search terms used to identify studies for this review.
Searching other resources
We checked the reference lists of all primary studies and review articles for additional references. We checked the websites of clinical trial registers for unpublished trial data. We also checked FDA submissions in relation to salmeterol.
Data collection and analysis
Selection of studies
Two review authors independently assessed studies identified via literature searches by examining title, abstract, and keyword fields. We obtained in full text studies that potentially fulfilled the inclusion criteria. CJC and TL independently assessed these for inclusion (CJC, BS, and SW for the 2018 update). We resolved disagreements by consensus.
Data extraction and management
We extracted data using a prepared checklist before one review author (CJC) with assistance from Susan Hansen (information specialist) entered data into RevMan 5.3, and another review author (TL, MF, SS, BS, or SW) checked data on trial characteristics. A third review author (RJ, MF, SS, BS, or SW) independently extracted outcome data and resolved discrepancies by discussion and correspondence with study sponsors. Data included characteristics (methods, participants, interventions, outcomes) and results of included studies. We contacted authors and sponsors of included studies for unpublished adverse event data, and we searched manufacturers' websites for further details of adverse events. We also searched FDA submissions. We collected all‐cause SAEs (fatal and non‐fatal), and in view of the difficulty involved in deciding whether events were asthma‐related, we noted details of the cause of death and SAEs where available. We also recorded the definition of SAEs, and we sought further information if this was not clear (particularly in relation to hospital admissions and SAEs).
Assessment of risk of bias in included studies
Two review authors (of CJC, Susan Hansen, MF, SS, BS, SW) assessed included studies for bias protection (including sequence generation for randomisation, allocation concealment, blinding of participants and assessors, loss to follow‐up, and completeness of outcome assessment).
Measures of treatment effect
The outcomes of this review were dichotomous, and we recorded the number of participants with one or more outcome events by allocated treated group.
Unit of analysis issues
We confined our analysis to participants with one or more SAEs, rather than analysing the number of events that occurred (as the latter are not independent when one participant experiences multiple events and therefore are not suitable for meta‐analysis).
Dealing with missing data
When we did not find full data on mortality or SAEs in published papers, we searched the sponsors' trial results sites and clinicaltrials.gov to obtain the missing data.
Assessment of heterogeneity
We assessed heterogeneity by using I² to indicate how much of the total heterogeneity was evident between studies (rather than within studies).
Assessment of reporting biases
We found full data on all‐cause mortality and SAEs, so we did not assess reporting biases any further for these outcomes. For asthma‐specific SAEs, we looked for evidence of independent outcome assessment in these studies.
Data synthesis
The outcomes of this review are dichotomous, and we recorded the number of participants with at least one outcome event by allocated treatment group. We calculated pooled odds ratios (ORs) and risk differences (RDs). We used the Peto OR for the primary analysis, as no adjustment for zero cells is required. This property was more important than potential problems with unbalanced treatment arms and large effect sizes (in view of the high proportion of zero cells), but we used the Mantel‐Haenszel method for sensitivity analysis. ORs do not include the large body of evidence derived from trials with no event in either arm, but we included such data in the analysis of absolute rates using RDs. We inspected funnel plots to assess publication bias.
For the 2018 update, we assessed the safety of adding salmeterol to ICS from the worst‐case number needed to treat for an additional harmful outcome (NNTH) for one additional SAE or death to occur. We calculated the NNTH using Visual Rx to transform the upper end of the 95% confidence interval (CI) of the pooled Peto OR, by applying it to the mean event rate in trial control arms (Visual Rx).
Subgroup analysis and investigation of heterogeneity
We planned subgroup analyses on the basis of age (adults vs children), severity of asthma, dose of salmeterol, and dose of inhaled corticosteroid in comparison arms. We made subgroup comparisons using tests for interaction (Altman 2003).
Sensitivity analysis
We carried out sensitivity analysis to assess the impact of the method used to combine study events (Peto OR, Mantel‐Haenszel OR, and RD). We included the degree of bias protection in study designs as part of the sensitivity analysis.
'Summary of findings' tables
We assessed the certainty of evidence as high, moderate, low, or very low in accordance with recommendations outlined by the GRADE Working Group for meta‐analyses of randomised trials (GRADE website). We have presented these assessments in summary of findings Table for the main comparison, alongside the results of our analyses for key outcomes in adults and children (all‐cause mortality, all‐cause non‐fatal SAEs, and asthma‐related mortality and SAEs).
Results
Description of studies
Results of the search
We included 40 studies in the previous version of this review (Cates 2013). The literature search for this review update covered the period from 2011 to October 2018 and yielded 436 references after removal of duplicates. We excluded 403 references on the basis of title and abstract alone, and we selected 33 records for full‐text review. We identified nine new studies and two references to previously included studies. Among adults and adolescents, we identified six new studies including 14,504 adult participants (AUSTRI 2016; Bernstein 2017; Mansfield 2017; Raphael 2017; Sher 2017; Slankard 2016), along with three new studies including 6783 child participants (MASCOT 2013; Ploszczuk 2014; VESTRI 2016).
We excluded six full‐text articles (describing four studies ‐ Bateman 2011; Lotvall 2014; NCT01172808; NCT01172821), as participants used background inhaled corticosteroids, but this was not part of the randomised treatment (Bateman 2011; Casale 2013; Casale 2014; Kerstjens 2014; Kerstjens 2015; Lotvall 2014).
The review now includes a total of 49 studies (Figure 1).
Study flow diagram: review update.
Included studies
We included 41 studies in adults and adolescents (27,951 participants) over the age of 12, 16, or 18 years, according to Characteristics of included studies. We included eight studies in children up to the age of 11 years (8453 participants). The weighted mean duration was 25 weeks in adult studies, and 23 weeks in studies in children.
Most studies were sponsored or supported by GlaxoSmithKline and compared combination fluticasone and salmeterol versus fluticasone alone. Teva Pharmaceuticals sponsored Mansfield 2017 and Raphael 2017, and Mundipharma Research Ltd sponsored Ploszczuk 2014. The NIHR Health Technology Assessment programme sponsored MASCOT 2013.
All except 542 adults (and none of the children) randomised to salmeterol were given fluticasone in the same (combination) inhaler (Table 1). The dose of salmeterol used was 50 μg twice daily in most studies (Table 1). The dose of fluticasone varied from 100 to 1000 μg/d (Table 1), and some studies stratified participants to different daily doses of fluticasone but used the same daily dose of fluticasone in each stratum for comparison with additional salmeterol.
| Study ID | Age of participants (Years) | N on FSC | N on ICS | Daily dose of fluticasone (μg) | Daily dose of salmeterol (μg) | Combined inhaler | Separate inhalers | Duration (weeks) |
| 12+ | 338 | 165 | 1000 | 100 | ✓ | ✓ | 28 | |
| 12+ | 5834 | 5845 | 200/500/1000 | 100 | ✓ | 26 | ||
| 12+ | 239 | 236 | 200 | 100 | ✓ | 52 | ||
| 12+ | 333 | 165 | 200 | 100 | ✓ | 52 | ||
| 12+ | 501 | 499 | 500 | 100 | ✓ | 24 | ||
| 12+ | 1709 | 1707 | 200/500/1000 | 100 | ✓ | 12 | ||
| 18+ | 159 | 159 | 500 | 100 | ✓ | 24 | ||
| 16+ | 171 | 160 | 500 | 100 | ✓ | 28 | ||
| 12+ | 306 | 315 | 500 | 100 | ✓ | 52 | ||
| 12+ | 92 | 90 | 200 | 100 | ✓ | 52 | ||
| 12+ | 310 | 318 | 500 | 100 | ✓ | 12 | ||
| 12+ | 156 | 156 | 200/500/1000 | 100 | ✓ | 40 | ||
| 18+ | 27 | 27 | 500 | 100 | ✓ | 12 | ||
| 4 to 11 | 173 | 177 | 200 | 100 | ✓ | 12 | ||
| 18+ | 95 | 92 | 500 | 100 | ✓ | 12 | ||
| 4 to 11 | 101 | 102 | 200 | 100 | ✓ | 12 | ||
| 12+ | 338 | 335 | 200/400/500/1000 | 25/100 | ✓ | 26 | ||
| 4 to 11 | 23 | 19 | 200 | 100 | ✓ | 48 | ||
| 12+ | 88 | 89 | 200 | 100 | ✓ | 12 | ||
| 12+ | 94 | 91 | 220 | 100 | ✓ | 16 | ||
| 4 to 11 | 171 | 168 | 200 | 100 | ✓ | 16 | ||
| 12+ | 95 | 97 | 200 | 100 | ✓ | 12 | ||
| 12+ | 92 | 89 | 200 | 100 | ✓ | 12 | ||
| 2 to 11 | 169 | 172 | 200 | 100 | ✓ | 12 | ||
| 12+ | 254 | 258 | 100/200 | 25 | ✓ | 12 | ||
| 12+ | 262 | 270 | 200 | 100 | ✓ | 24 | ||
| 12+ | 180 | 182 | 500 | 100 | ✓ | 12 | ||
| 18+ | 29 | 32 | 200/500/1000 | 100 | ✓ | 30 | ||
| 18+ | 42 | 21 | 200 | 100 | ✓ | 52 | ||
| 18+ | 93 | 93 | 1000 | 100 | ✓ | 26 | ||
| 4 to 11 | 181 | 181 | 200 | 100 | ✓ | 24 | ||
| 18+ | 41 | 41 | 200/500/1000 | 100 | ✓ | 52 | ||
| 12+ | 150 | 150 | 200/500/1000 | 100 | ✓ | 40 | ||
| 4 to 11 | 304 | 304 | 100 | 50 | ✓ | 12 | ||
| 12+ | 210 | 212 | 500 | 50 | ✓ | 12 | ||
| 12+ | 151 | 155 | 100 | 50 | ✓ | 12 | ||
| 15+ | 172 | 159 | 200 | 100 | ✓ | 16 | ||
| 15+ | 161 | 161 | 200 | 100 | ✓ | 16 | ||
| 12+ | 262 | 270 | 200 | 100 | ✓ | 24 | ||
| 12+ | 239 | 236 | 200 | 100 | ✓ | 52 | ||
| 18+ | 159 | 159 | 500 | 100 | ✓ | 24 | ||
| 12+ | 84 | 84 | 500 | 100 | ✓ | 12 | ||
| 12+ | 288 | 291 | 200/400 | 25 | ✓ | 12 | ||
| 18+ | 32 | 29 | 88/220/440 | 42 | ✓ | 16 | ||
| 16+ | 14 | 17 | 200 | 100 | ✓ | 12 | ||
| 18+ | 78 | 72 | 200 | 100 | ✓ | 12 | ||
| 12+ | 337 | 172 | 1000 | 100 | ✓ | 12 | ||
| 4 to 11 | 3107 | 3101 | 200/500 | 100 | ✓ | 26 | ||
| 12+ | 18 | 19 | 400 | 100 | ✓ | 12 |
FSC: salmeterol/fluticasone; ICS: inhaled corticosteroid.
Excluded studies
We excluded 81 studies with reasons described in Characteristics of excluded studies.
Risk of bias in included studies
We have provided an overview of the risk of bias in individual studies in Figure 2. We did not downgrade our certainty in the findings for risk of bias in this update, for the reasons outlined below.
Methodological quality summary: review authors' judgements about each methodological quality item for each included study.
Allocation
Studies did not report clear details of sequence generation and allocation concealment, but we found from correspondence with study sponsors that standard methods (as required by regulatory authorities) had been used to protect against selection bias in these studies. We therefore regarded the risk of selection bias as low, even though sequence generation and allocation concealment are marked as unclear in most studies in Figure 2.
Blinding
All studies were reported as double‐blind (with the exception of Mansfield 2017, which was open‐label). Researchers incorporated double‐dummy design when inhaler devices were not the same in each arm. We therefore assessed risks of performance and detection bias as low for all‐cause events.
Indepedent assessment of causation
Previously, no independent assessment had examined the causation of SAEs, so it is possible that asthma‐related events were subject to detection bias. However the two large new trials, which contributed 42% of asthma‐related events in adults (AUSTRI 2016), as well as 92% in children (VESTRI 2016), used independent panels to assess causation for asthma‐related hospitalisation, intubation, or mortality. The risk of bias is therefore reduced for asthma‐related events in this update (compared to the previous version of this review).
Incomplete outcome data
Researchers generally reported safety analyses for all randomised participants who had taken at least one dose of treatment. Some studies reported high proportions of withdrawals; these contributed only a small proportion of participants to the meta‐analyses, so we did not downgrade the overall quality of evidence for attrition bias.
Selective reporting
Sponsors have not found or provided data for all‐cause fatal and non‐fatal SAEs by treatment group for all studies, so we regarded the risk of reporting bias as low for the outcomes considered in this review.
Other potential sources of bias
GlaxoSmithKline sponsored or supported most studies, which compared combination fluticasone and salmeterol to fluticasone alone. Teva Pharmaceuticals sponsored Mansfield 2017 and Raphael 2017, and Mundipharma Research Ltd sponsored Ploszczuk 2014. The NIHR Health Technology Assessment programme sponsored MASCOT 2013. However we do not regard sponsorship as necessarily increasing the risk of bias when studies are well designed.
Effects of interventions
See: Summary of findings for the main comparison Serious adverse events (SAEs) in adults and children
Primary outcomes
All‐cause mortality
Adults and adolescents
Eleven deaths occurred in 14,233 adults on regular salmeterol with inhaled corticosteroid (ICS), and 13 deaths in 13,718 adults on regular ICS at the same dose. The pooled odds ratio was Peto OR 0.80 (95% CI 0.36 to 1.78; participants = 27,951; studies = 41; I² = 0%; moderate‐certainty evidence; Analysis 1.1; Figure 3). In other words, for every 1000 adults treated for 25 weeks, one death occurred on ICS alone, and the corresponding risk for salmeterol and ICS was also one death (95% CI 0 to 2 deaths). We assessed this as moderate‐certainty evidence because only 24 deaths in total occurred across all trials (summary of findings Table for the main comparison). Moreover, even one extra death per 1000 is not a level of risk that we would regard as acceptable (Appendix 5), so we are unable to conclude with a high level of certainty that regular salmeterol with ICS is as safe as regular ICS alone.
Forest plot of comparison: 1 Regular salmeterol in addition to regular inhaled corticosteroids, outcome: 1.1 All‐cause mortality.
We have listed the causes of death in Table 2. Study authors reported none of the deaths as related to asthma.
| Study ID | Treatment arm | Cause of death in adults (n) |
| salmeterol and fluticasone | Heroin overdose (1), stroke (1), and dyspnoea due to metastatic hepatic carcinoma (1) | |
| fluticasone | Aortic dissection (1), sudden cardiac death (1), gastroenteritis (1), sepsis (1), and stroke (2) | |
| salmeterol and fluticasone (separate inhalers) | Bronchial carcinoma (1) | |
| salmeterol and fluticasone | Myocardial infarction (2) and pneumonia (1) | |
| fluticasone | Myocardial infarction (2) | |
| salmeterol and fluticasone (separate inhalers) | Pneumothorax (1) | |
| salmeterol and fluticasone | Cardiac disease (1) | |
| fluticasone | Breast cancer (1) | |
| fluticasone | Cardiac arrest and deep vein thrombosis (1) | |
| fluticasone | Cardiac arrest (1) | |
| fluticasone | Ventricular hypertrophy and aortic hypoplasia (1) | |
| salmeterol and fluticasone | A female patient developed severe jaundice on day 30, which led to withdrawal of the study drug, and she died 6 weeks later. She had begun taking an herbal supplement before the onset of jaundice and then declined recommended interventions and investigations into her condition. The death was considered not related to study treatment | |
| fluticasone | Unknown cause (1) | |
| salmeterol and fluticasone | Leukaemia (1) |
Children
Eight studies on children (8453 participants) reported no deaths. It is not possible to calculate any ORs from these data, but the pooled RD can be assessed with a confidence interval (RD 0.0000, 95% CI ‐0.0013 to 0.0013; Analysis 2.1; Table 3; Figure 4). In other words, for every 1000 children treated with salmeterol and ICS for 23 weeks, the 95% confidence interval is compatible with a possible increase or decrease of one death (summary of findings Table for the main comparison). We also assessed this as moderate‐certainty evidence because no deaths occurred among children across all trials, and even one extra death per 1000 is not a level of risk that we would regard as acceptable (summary of findings Table for the main comparison; Appendix 5).
Forest plot of comparison: 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, outcome: 2.1 All‐cause mortality.
| Peto OR meta‐analysis: regular salmeterol in addition to regular ICS vs ICS alone | ||||||||
| All‐cause mortality | Events Rx | Total Rx | Events control | Total control | Peto OR | CI start | CI end | Worst‐case NNTHa (from OR CI end) |
| Adults and adolescents | 11 | 14,233 | 13 | 13,718 | 0.80 | 0.36 | 1.78 | 1661 |
| Children | 0 | 4229 | 0 | 4224 | ‐ | ‐ | ‐ | ‐ |
| All cause non‐fatal SAE | ||||||||
| Adults and adolescents | 332 | 14,233 | 282 | 13,718 | 1.14 | 0.97 | 1.33 | 152 |
| Children | 65 | 4229 | 62 | 4224 | 1.04 | 0.73 | 1.48 | 139 |
| Asthma‐related mortality | ||||||||
| Adults and adolescents | 0 | 14,233 | 0 | 13,718 | ‐ | ‐ | ‐ | ‐ |
| Children | 0 | 4229 | 0 | 4224 | ‐ | ‐ | ‐ | ‐ |
| Asthma‐related SAE | ||||||||
| Adults and adolescents | 80 | 14,233 | 67 | 13,718 | 1.15 | 0.83 | 1.59 | 351 |
| Children | 29 | 4229 | 23 | 4224 | 1.25 | 0.72 | 2.16 | 128 |
aNNTH (number needed to treat for an additional harmful outcome) calculated using Visual Rx to transform the upper end of the 95% confidence interval of the Peto OR, respectively.
Serious adverse events (non‐fatal all‐cause)
A serious adverse event in general is defined as an event that falls into any of the following categories.
-
Results in death.
-
Is life‐threatening.
-
Requires inpatient hospitalisation or prolongation of existing hospitalisation.
-
Results in persistent or significant disability/incapacity.
-
Is a congenital anomaly/birth defect.
This is the definition from the International Conference on Harmonisation (ICH), and we have assumed that investigators in the included trials used this definition (even though this often was not explicitly reported in the papers, it is the standard definition for regulatory trials (ICHE2a 1995)).
Adults and adolescents
Trials reported non‐fatal SAEs of any cause in 332 of 14,233 adults on regular salmeterol with ICS compared to 282 of 13,718 adults on regular ICS; the pooled Peto OR was 1.14 (95% CI 0.97 to 1.33; participants = 27,951; studies = 41; I² = 0%; moderate‐certainty evidence; Figure 5). For every 1000 adults treated for 25 weeks, 21 adults on ICS alone experienced an SAE, and the corresponding risk for salmeterol and ICS was 23 adults (95% CI 20 to 27; summary of findings Table for the main comparison).
Forest plot of comparison: 1 Regular salmeterol in addition to regular inhaled corticosteroids, outcome: 1.2 All‐cause non‐fatal SAE.
We assessed this as moderate‐certainty evidence because the upper boundary of the confidence interval for the Peto OR translates into an NNTH of 152 for additional adults to suffer a non‐fatal SAE over six months (Table 3). We regarded the upper limit as too high a risk to have a high level of certainty that salmeterol is safe.
Children
Data show 65 of 4229 children with SAEs on regular salmeterol with ICS compared to 62 of 4224 on regular ICS; the pooled odds ratio was Peto OR 1.04 (95% CI 0.73 to 1.48; participants = 8453; studies = 8; I² = 0%; moderate‐certainty evidence; Figure 5).
For every 1000 children treated for 23 weeks, 15 children experienced an SAE on ICS alone, and the corresponding risk on salmeterol and ICS was also 15 children (95% CI 11 to 22).
We assessed this as moderate‐certainty evidence because the upper boundary of the confidence interval for the Peto OR translates into an NNTH of 139 for an additional child to suffer a non‐fatal SAE over six months (Table 3). We regarded the upper limit as too high a risk to have a high level of certainty that salmeterol is safe.
The test for interaction between adults and children did not show a significant impact of age on the treatment effect (Figure 5).
Secondary outcomes
Mortality related to asthma
None of the deaths were reported to be related to asthma. Nevertheless, when all studies were combined in a risk difference meta‐analysis, we were able to use this to estimate the maximum increase in asthma deaths that is compatible with the numbers of adults and children included (using the upper end of the 95% confidence interval). For adults, this yields RD 0.0000 (95% CI ‐0.0009 to 0.0009; participants = 27,951; studies = 41), and for children RD 0.0000 (95% CI ‐0.0013 to 0.0013; participants = 8453; studies = 8; Table 3; Figure 6). In other words, for both adults and children, the pooled risk difference is compatible with one more or one fewer death per 1000 treated for six months.
Forest plot of comparison: 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, outcome: 2.5 Asthma‐related mortality.
We assessed this as moderate‐certainty evidence because no deaths were related to asthma, and even one extra death per 1000 is not a level of risk that we would regard as acceptable (summary of findings Table for the main comparison; Appendix 5).
Non‐fatal serious adverse events related to asthma
Adults and adolescents
Researchers reported asthma‐related SAEs in 80 and 67 adults in each group, respectively, and the pooled odds ratio was Peto OR 1.15 (95% CI 0.83 to 1.59; participants = 27,951; studies = 41; I² = 0%; low‐certainty evidence; Figure 7).
Forest plot of comparison: 1 Regular salmeterol in addition to regular inhaled corticosteroids, outcome: 1.3 Asthma‐related SAE.
For every 1000 adults treated for 25 weeks, five experienced an asthma‐related non‐fatal SAE on ICS alone, and the corresponding risk on salmeterol and ICS was six adults (95% CI 4 to 8).
We assessed this as low‐certainty evidence because the upper boundary of the confidence interval for the Peto OR translates into an NNTH of 351 for an additional adult to suffer a non‐fatal asthma‐related SAE over six months (Table 3). We also downgraded our certainty in this outcome because more than half of the events occurred during trials that provided no independent assessment of the causation of asthma‐related events.
Children
Study results show 29 and 23 children in each treatment group with asthma‐related events (Peto OR 1.25, 95% CI 0.72 to 2.16; participants = 8453; studies = 8; I² = 0%; moderate‐certainty evidence; Figure 7). For every 1000 children treated for 23 weeks, five experienced an asthma‐related SAE on ICS alone, and the corresponding risk on salmeterol and ICS was seven children (95% CI 4 to 12).
We assessed this as moderate‐certainty evidence because the upper boundary of the confidence interval for the Peto OR translates into an NNTH of 128 for an additional child to suffer a non‐fatal asthma‐related SAE over six months (Table 3). We regarded this upper limit as too high a risk to have a high level of certainty that salmeterol is safe, but we did not downgrade our certainty in the evidence for detection bias as almost all events occurred in VESTRI 2016 (in which causation was independently assessed).
The difference between children and adults again was not statistically significant (Figure 7). We did not find sufficient data to assess the other proposed secondary outcomes (such as intensive care unit admission and intubation).
Sensitivity analyses
Risk of bias
We identified one unblinded study that did not report any deaths (Mansfield 2017); removal of this study for non‐fatal SAEs of any cause made very little difference to the pooled Peto OR of 1.13 (95% CI 0.96 to 1.34; participants = 27,278; studies = 41). Removal of the four studies thought to be at high risk of attrition bias also made little difference to the OR of non‐fatal SAEs in adults (Peto OR 1.11, 95% CI 0.94 to 1.31; participants = 26,763; studies = 37) (Figure 2). A funnel plot of non‐fatal SAEs did not suggest obvious publication bias (Figure 8).
Funnel plot of comparison: 1 Regular salmeterol in addition to regular inhaled corticosteroids, outcome: 1.2 All‐cause non‐fatal SAE.
Restricting the results for asthma‐related SAEs to trials with independent outcome assessment made very little difference to the results in children (OR 1.28, 95% CI 0.73 to 2.27; participants = 6208; studies = 1) but decreased the precision of results in adults (OR 1.03, 95% CI 0.64 to 1.67; participants = 11,679; studies = 1).
Methods of analysis
Risk difference
We have summarised the results of using pooled risk differences in Table 4; data show very similar absolute treatment effects to those calculated from the pooled Peto OR in Table 3 and summary of findings Table for the main comparison.
| Risk difference meta‐analysis: regular salmeterol in addition to regular ICS vs ICS alone | ||||||||
| All‐cause mortality | Events Rx | Total Rx | Events control | Total control | Risk difference | CI start | CI end | Worst‐case NNTHa (from RD 95% CI end) |
| Adults and adolescents | 11 | 14,233 | 13 | 13,718 | ‐0.0002 | ‐0.0013 | 0.0009 | 1099 |
| Children | 0 | 4229 | 0 | 4224 | 0.0000 | ‐0.0013 | 0.0013 | 763 |
| All‐cause non‐fatal SAE | ||||||||
| Adults and adolescents | 333 | 14,233 | 283 | 13,718 | 0.0027 | ‐0.0008 | 0.0062 | 161 |
| Children | 65 | 4229 | 62 | 4224 | 0.0006 | ‐0.0047 | 0.0058 | 172 |
| Asthma‐related mortality | ||||||||
| Adults and adolescents | 0 | 14,233 | 0 | 13,718 | 0.0000 | ‐0.0009 | 0.0009 | 1099 |
| Children | 0 | 4229 | 0 | 4224 | 0.0000 | ‐0.0013 | 0.0013 | 763 |
| Asthma‐related non‐fatal SAE | ||||||||
| Adults and adolescents | 82 | 14,233 | 67 | 13,718 | 0.0007 | ‐0.0012 | 0.0026 | 385 |
| Children | 29 | 4229 | 23 | 4224 | 0.0014 | ‐0.0022 | 0.0049 | 205 |
aNNTH (number needed to treat for an additional harmful outcome) calculated from the inverse of the upper end of the 95% confidence interval (CI end) of the risk difference.
For all‐cause mortality in adults and adolescents, the pooled RD was ‐0.0002 (95% CI ‐0.0013 to 0.0009; Table 4; Figure 4). In other words, for every 1000 adults treated with salmeterol and ICS for 25 weeks, the results are compatible with one more or one fewer death (in comparison with one death on regular ICS alone). In children (with no reported deaths and OR that could not be calculated), the pooled RD for all‐cause mortality was 0.0000 (95% CI ‐0.0013 to 0.0013; Analysis 2.1; Table 4; Figure 4). In other words, for every 1000 children treated with salmeterol and ICS for 23 weeks, the 95% confidence interval of the pooled RD is compatible with a possible increase or decrease of one death (summary of findings Table for the main comparison).
For adults and adolescents with a non‐fatal SAE of any cause, the pooled RD was 0.0027 (95% CI ‐0.0008 to 0.0062; Table 4; Figure 9). For children, the pooled RD was 0.0006 (95% CI ‐0.0047 to 0.0058; Table 4; Figure 9).
Forest plot of comparison: 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, outcome: 2.2 All‐cause non‐fatal SAE.
When the upper end of the 95% confidence interval is translated into the worst‐case NNTH, this yields results very similar to those obtained with the Peto OR and the pooled RD (Table 3; Table 4). For adults, the worst‐case NNTH is 152 from the Peto OR and 161 from the RD. Similarly in children, the worst‐case NNTH is 139 from the Peto OR and 172 from the RD.
The absolute difference for asthma‐related non‐fatal SAEs in adults was RD 0.0007 (95% CI ‐0.0012 to 0.0026). Similarly, the absolute difference for children was RD 0.0014 (95% CI ‐0.0022 to 0.0049; Table 3; Figure 10). These results are very similar to the absolute differences derived from the Peto OR in summary of findings Table for the main comparison.
Forest plot of comparison: 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, outcome: 2.4 Asthma‐related SAE.
Mantel‐Haenszel odds ratio
We analysed primary outcomes using Mantel‐Haenszel fixed‐effect and random‐effects models. The result of the fixed‐effect model for mortality in adults was OR 0.80 (95% CI 0.40 to 1.63), and for the random‐effects model OR 0.81 (95% CI 0.38 to 1.72); both were very similar to the Peto OR result of 0.80 (95% CI 0.36 to 1.78). The Mantel‐Haenszel method uses a correction for zero cells that is not required for the Peto OR. With this method, the addition of 0.5 to all cells when the arms include similar randomised numbers will generate an OR of 3 when only one event occurs in the treatment group, and none in the control group. When outcomes are very sparse (as for mortality), the calculated OR is dependent on the size of the zero cell adjustment and on whether treatment arms are balanced.
For non‐fatal serious adverse events in adults, the Peto method (Peto OR 1.14, 95% CI 0.97 to 1.33) yielded almost identical results to the Mantel‐Haenszel fixed‐effect model (OR 1.13, 95% CI 0.97 to 1.33) or the Mantel‐Haenszel random‐effects model (OR 1.12, 95% CI 0.95 to 1.32).
Fatal and non‐fatal serious adverse events combined
When fatal and non‐fatal serious adverse events were combined, the results for adults were almost identical to the pooled result for non‐fatal events in adults (Peto OR 1.12, 95% CI 0.96 to 1.31; Analysis 1.3), and results were unchanged in children, as no deaths occurred in children.
Dose of salmeterol
The dose of salmeterol used in seven studies was less than the usual daily dose of 50 μg twice daily, so we carried out sensitivity analysis out on the primary outcomes while excluding these studies (Mansfield 2017; Raphael 2017; SAS30021; SAS30022; SAS30023; Sher 2017; Slankard 2016). Sher 2017 reported one death, and without this event, the pooled odds ratio was Peto OR 0.72 (95% CI 0.32 to 1.65); among studies in children, no deaths occurred, so mortality results were unaffected. Results for an SAE of any cause without these studies in adults (Peto OR 1.11, 95% CI 0.94 to 1.31; participants = 25,398) and in children (Peto OR 1.08, 95% CI 0.75 to 1.54; participants = 7845) are very similar to the full data set.
Subgroup analyses
Mortality data were too sparse for any subgroup analysis to be performed. For non‐fatal SAEs of any cause, tests for interaction between adults and children did not show a significant difference between results according to age group (test for subgroup differences: Chi² = 0.20, df = 1 (P = 0.65); I² = 0%), nor was there a difference in asthma‐related events between adults and children (test for subgroup differences: Chi² = 0.07, df = 1 (P = 0.79); I² = 0%). Studies that included patients on separate salmeterol and fluticasone inhalers were too few to permit a subgroup comparison between separate and combined inhalers (Table 1).
Discussion
Summary of main results
The number of adults included in this review has doubled and the number of children quadrupled with the addition of new studies to the 2018 update of this review. As a consequence, estimates of risk of non‐fatal serious adverse events have become more precise. The previous upper 95% confidence interval (CI) was Peto odds ratio (OR) 1.44 for adults and 3.91 for children, and is now 1.33 and 1.48, respectively.
However, no deaths in children and no deaths in adults were caused by asthma, so we are not able to draw firm conclusions about the risks of dying from asthma when regular salmeterol is added to inhaled corticosteroid (ICS).
All‐cause mortality in adults
Eleven of a total of 14,233 adults on regular salmeterol and ICS died, as did 13 of 13,718 on regular ICS at the same dose. The pooled Peto OR was 0.80 (95% CI 0.36 to 1.78; participants = 27,951; studies = 41; I² = 0%; moderate‐certainty evidence). In other words, for every 1000 adults treated for 25 weeks, one death occurred on ICS alone, and the corresponding risk on salmeterol and ICS was also one death (95% CI 0 to 2 deaths).
All‐cause non‐fatal serious adverse events
A total of 332 adults on regular salmeterol with ICS suffered a non‐fatal serious adverse event of any cause compared to 282 adults on regular ICS. The pooled Peto OR was 1.14 (95% CI 0.97 to 1.33; participants = 27,951; studies = 41; I² = 0%; moderate‐certainty evidence). For every 1000 adults treated for 25 weeks, 21 adults on ICS alone experienced a serious adverse event, and the corresponding risk on salmeterol and ICS was 23 adults (95% CI 20 to 27).
Sixty‐five of 4229 children on regular salmeterol with ICS experienced a serious adverse event of any cause compared to 62 of 4224 on regular ICS. The pooled Peto OR was 1.04 (95% CI 0.73 to 1.48; participants = 8453; studies = 8; I² = 0%; moderate‐certainty evidence). For every 1000 children treated for 23 weeks, 15 children on ICS alone had a serious adverse event, and the corresponding risk on salmeterol and ICS was also 15 children (95% CI 11 to 22).
Overall completeness and applicability of evidence
Two large surveillance studies examined the use of regular salmeterol without randomised ICS (SMART 2006SNS 1993). As a result of the increase in serious adverse events reported in these studies, the FDA mandated surveillance studies of regular salmeterol in adults and children. These two large surveillance studies ‐ in adults (AUSTRI 2016 ‐ previously ongoing study NCT01475721) ‐ and in children (VESTRI 2016 ‐ previously ongoing study NCT01462344) ‐ assessed the safety of regular salmeterol in combination with an ICS; these studies have now been published, and the results have been incorporated into this updated review. The two large new studies were designed to test whether salmeterol was safe when added to ICS; to do this, investigators used a safety margin that was derived from the worst end of the 95% confidence interval of the comparative risk for a combined endpoint in these trials (death, intubation, or hospital admission). The FDA pre‐defined safety margins for the composite outcomes in these studies of a relative measure of 2.0 in adults and 2.675 in children; confidence intervals of the pooled odds ratios from this review for serious adverse events for adults and children (all‐cause and asthma‐related) now fall within these margins.
Using the same approach ‐ by transforming the worst‐case 95% confidence interval of the pooled odds ratio and the pooled risk difference from all trials in this review ‐ we determined the upper limits of the confidence intervals for both odds ratios and risk differences, and we have presented them in Table 3 and Table 4. We used Visual Rx to convert these upper limits to worst‐case numbers needed to treat for an additional harmful effect (NNTH), for one additional person to suffer an event from these upper limits listed in Table 3 and Table 4.
Although the new FDA studies do not exceed their pre‐defined safety margins for non‐fatal serious adverse events on salmeterol in combination with ICS, the levels of adherence in AUSTRI 2016 (median 95.1%) and VESTRI 2016 (median 94%) were much higher than is usual in day‐to‐day practice. These trials, as well as many recent trials, also excluded participants with life‐threatening asthma. This may limit the applicability of findings from these new studies and from this review.
Almost all studies used combination inhalers; therefore the results of this review do not apply to the use of salmeterol and ICS in separate inhalers.
Quality of the evidence
We assessed risk of bias in the included studies as low for all‐cause outcomes, as the procedures for randomisation and blinding were appropriate, having been designed for regulatory purposes (thereby ensuring common definitions of serious adverse events and minimising the likelihood of selection bias, even though this was not well reported in published papers or trial registers). Although we judged most studies to be at high risk of detection bias, we are not concerned about detection bias for asthma‐related serious adverse events in children, because most events were reported from the new study VESTRI 2016, which provided independent assessment of causation. However we were more concerned about detection bias affecting the treatment effect in adults because the only study with independent assessment of causation was AUSTRI 2016, which yielded a wider confidence interval when data from this study alone were considered.
We previously assessed our confidence in the evidence as moderate due to statistical imprecision. This reflected the low rates of mortality and non‐fatal adverse events across studies included in the review (summary of findings Table for the main comparison).
Our confidence in the findings of this review remains moderate. Included studies reported no deaths from asthma and a total of 24 deaths of any cause among adults. Even with additional data for the 2018 update, the upper confidence intervals for all‐cause non‐fatal serious adverse events remain at a possible NNTH of 152 for adults and 139 for children over six months. We believe that this risk boundary remains too high to grade the evidence with a high level of certainty to show that adding salmeterol to regular ICS was entirely safe.
Potential biases in the review process
Selection of the best method to combine studies with rare events is contentious when event rates are low ‐ not least because of the corrections required to calculate ORs with zero events (Sweeting 2004). Because it became apparent in the course of carrying out our reviews that the pooled ORs were heavily dependent on the zero adjustment used in Mantel‐Haenszel and inverse variance methods, we used the Peto OR and RDs to report results of this review. The likely bias in using the Peto OR is small, as only three trials ‐ Aubier 1999, Bateman 2001, and van Noord 2001 ‐ showed any imbalance in the number of patients in each arm (Sweeting 2004). In these studies, twice the number of patients were randomised to regular salmeterol with ICS in comparison to ICS alone.
Similarly, the included studies were influenced by the decision to restrict the review to trials that randomised participants to salmeterol and ICS, but this decision reduces the risk of bias arising from patients discontinuing their usual inhaled steroid medication if they feel better on the randomised treatment. This pre‐supposes a similar risk of serious adverse events when salmeterol and fluticasone are delivered via one inhaler, and when salmeterol is added to inhaled corticosteroid therapy via a separate inhaler, when both are randomised treatments in a controlled trial.
Agreements and disagreements with other studies or reviews
An overview of Cochrane Reviews on the safety of regular salmeterol or formoterol in children (with and without combination ICS) does not reveal any significant differences between the safety of regular salmeterol and regular formoterol in children, but the number of children studied remains small in comparison to the number of adults (Cates 2012aCates 2014). Moreover, no separate safety results are currently available for adolescents who were recruited in the adult and adolescent trials.
Two systematic reviews on the use of salmeterol with ICS have shown results similar to the findings of this review. Bateman 2008 concentrated on asthma‐related outcomes, and Jaeschke 2008a considered both salmeterol and formoterol in adults in comparison to ICS at the same dose and at higher doses. Neither of these reviews showed a significant increase in the risk of serious adverse events, but results were not precise enough to rule out a clinically important increase or decrease in serious adverse events with regular salmeterol.
Minor discrepancies between results recorded in serious adverse event reports on the GlaxoSmithKline website and in the data provided by Bateman 2008 and Jaeschke 2008a became apparent during preparation of this review. An example of this is the death in Aubier 1999, which is related to the question of whether the adverse event was classified as being "on‐treatment" (see Aubier 1999 notes in Characteristics of included studies). Overall the magnitude of these differences is small, and this relates most often to an external review of company data and inclusion of reviewed data in some analyses; this has not altered the conclusions of this review.
Administration of inhaled corticosteroid (ICS)
No clear difference can be seen between the point estimate and the confidence interval of the OR for non‐fatal serious adverse events in adults (Peto OR 1.14, 95% CI 0.97 to 1.33) or children (Peto OR 1.04, 95% CI 0.73 to 1.48) presented in this review, nor in the Peto OR given in the review comparing salmeterol to placebo (OR 1.14, 95% CI 1.01 to 1.28) (Cates 2008a). However the average non‐fatal serious adverse event rate in control arms of the trials in this review that included randomised ICS in adults was 2.0% over 25 weeks compared to 3.6% over 28 weeks in SMART 2006 (which accounted for most patients in Cates 2008a).
This may reflect greater asthma severity among patients who had been started by their own physician on background ICS (as shown by Sears 2009 in the RELIEF study), but it could also be compounded by known poor adherence to treatment with ICS in routine practice. This raises uncertainty about application of results reported for patients in clinical trials, which usually include much more intensive monitoring of adherence to therapy. For example, the two new large trials added to this review in 2018 show very high adherence rates; the median level of adherence in AUSTRI 2016 was 95.1%, and in VESTRI 2016 94%. Given that we cannot assume that adherence to treatment in trials will be matched in routine practice, care must be exercised in both interpretation and application of trial results (Weiss 2008).
We were not able to investigate possible differences between combined and separate inhalers in trial findings due to the paucity of patients on separate inhalers included in the trials in this review (fewer than 300 patients were randomised to separate fluticasone and salmeterol inhalers; see Table 1).
Study flow diagram: review update.
Methodological quality summary: review authors' judgements about each methodological quality item for each included study.
Forest plot of comparison: 1 Regular salmeterol in addition to regular inhaled corticosteroids, outcome: 1.1 All‐cause mortality.
Forest plot of comparison: 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, outcome: 2.1 All‐cause mortality.
Forest plot of comparison: 1 Regular salmeterol in addition to regular inhaled corticosteroids, outcome: 1.2 All‐cause non‐fatal SAE.
Forest plot of comparison: 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, outcome: 2.5 Asthma‐related mortality.
Forest plot of comparison: 1 Regular salmeterol in addition to regular inhaled corticosteroids, outcome: 1.3 Asthma‐related SAE.
Funnel plot of comparison: 1 Regular salmeterol in addition to regular inhaled corticosteroids, outcome: 1.2 All‐cause non‐fatal SAE.
Forest plot of comparison: 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, outcome: 2.2 All‐cause non‐fatal SAE.
Forest plot of comparison: 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, outcome: 2.4 Asthma‐related SAE.
Comparison 1 Regular salmeterol in addition to regular inhaled corticosteroids, Outcome 1 All‐cause mortality.
Comparison 1 Regular salmeterol in addition to regular inhaled corticosteroids, Outcome 2 All‐cause non‐fatal SAE.
Comparison 1 Regular salmeterol in addition to regular inhaled corticosteroids, Outcome 3 All‐cause SAE (fatal and non‐fatal).
Comparison 1 Regular salmeterol in addition to regular inhaled corticosteroids, Outcome 4 Asthma‐related SAE.
Comparison 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, Outcome 1 All‐cause mortality.
Comparison 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, Outcome 2 All‐cause non‐fatal SAE.
Comparison 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, Outcome 3 All‐cause SAE (fatal and non‐fatal).
Comparison 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, Outcome 4 Asthma‐related SAE.
Comparison 2 Risk difference meta‐analysis: regular salmeterol in addition to regular ICS, Outcome 5 Asthma‐related mortality.
| Inhaled steroids with and without regular salmeterol for asthma | ||||||
| Patient or population: adults and children with chronic asthma Settings: community Intervention: regular salmeterol in addition to regular inhaled corticosteroid (ICS) Comparison: regular ICS (at the same dose) | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No. of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Regular ICS | Regular salmeterol in addition to regular ICS | |||||
| Adults and adolescents | ||||||
| All‐cause mortality | 1 per 1000a | 1 per 1000 | OR 0.80 | 27951 | ⊕⊕⊕⊝ | |
| All‐cause non‐fatal SAEc | 21 per 1000a | 23 per 1000 | OR 1.14 | 27951 | ⊕⊕⊕⊝ | |
| Asthma‐related mortality Follow‐up: mean 25 weeks | No deaths due to asthma | No deaths due to asthma | ‐ | 27951 | ⊕⊕⊕⊝ | Pooled risk difference zero (95% CI ‐0.0009 to 0.0009) |
| Asthma‐related non‐fatal SAEc | 5 per 1000a | 6 per 1000 | OR 1.15 | 27951 | ⊕⊕⊝⊝ | |
| Children | ||||||
| All‐cause mortality Follow‐up: mean 23 weeks | No deaths | No deaths | ‐ | 8453 | ⊕⊕⊕⊝ | Pooled risk difference zero (95% CI ‐0.0013 to 0.0013) |
| All‐cause non‐fatal SAEc | 15 per 1000a | 15 per 1000 | OR 1.04 | 8453 | ⊕⊕⊕⊝ | |
| Asthma‐related mortality Follow‐up: mean 23 weeks | No deaths | No deaths | ‐ | 8453 | ⊕⊕⊕⊝ | Pooled risk difference zero (95% CI ‐0.0013 to 0.0013) |
| Asthma‐related non‐fatal SAEc | 5 per 1000a | 7 per 1000 | OR 1.25 | 8453 | ⊕⊕⊕⊝ | |
| *The basis for the assumed risk is the mean control event rate in the included studies. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; ICS: inhaled corticosteroid; OR: odds ratio; SAE: serious adverse event. | ||||||
| GRADE Working Group grades of evidence. High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. | ||||||
| aMean control event rate. bImprecision (‐1). We regard the upper end of the 95% confidence interval as too high to rule out a potentially important increase. See also Table 3 and Appendix 5. cNon‐fatal SAEs were defined as life‐threatening adverse events, inpatient hospitalisations or prolongation of existing hospitalisations, persistent or significant disabilities or incapacities, or congenital anomalies or birth defects. dMore than half of the events came from trials that did not independently assess the causation of SAEs. | ||||||
| Study ID | Age of participants (Years) | N on FSC | N on ICS | Daily dose of fluticasone (μg) | Daily dose of salmeterol (μg) | Combined inhaler | Separate inhalers | Duration (weeks) |
| 12+ | 338 | 165 | 1000 | 100 | ✓ | ✓ | 28 | |
| 12+ | 5834 | 5845 | 200/500/1000 | 100 | ✓ | 26 | ||
| 12+ | 239 | 236 | 200 | 100 | ✓ | 52 | ||
| 12+ | 333 | 165 | 200 | 100 | ✓ | 52 | ||
| 12+ | 501 | 499 | 500 | 100 | ✓ | 24 | ||
| 12+ | 1709 | 1707 | 200/500/1000 | 100 | ✓ | 12 | ||
| 18+ | 159 | 159 | 500 | 100 | ✓ | 24 | ||
| 16+ | 171 | 160 | 500 | 100 | ✓ | 28 | ||
| 12+ | 306 | 315 | 500 | 100 | ✓ | 52 | ||
| 12+ | 92 | 90 | 200 | 100 | ✓ | 52 | ||
| 12+ | 310 | 318 | 500 | 100 | ✓ | 12 | ||
| 12+ | 156 | 156 | 200/500/1000 | 100 | ✓ | 40 | ||
| 18+ | 27 | 27 | 500 | 100 | ✓ | 12 | ||
| 4 to 11 | 173 | 177 | 200 | 100 | ✓ | 12 | ||
| 18+ | 95 | 92 | 500 | 100 | ✓ | 12 | ||
| 4 to 11 | 101 | 102 | 200 | 100 | ✓ | 12 | ||
| 12+ | 338 | 335 | 200/400/500/1000 | 25/100 | ✓ | 26 | ||
| 4 to 11 | 23 | 19 | 200 | 100 | ✓ | 48 | ||
| 12+ | 88 | 89 | 200 | 100 | ✓ | 12 | ||
| 12+ | 94 | 91 | 220 | 100 | ✓ | 16 | ||
| 4 to 11 | 171 | 168 | 200 | 100 | ✓ | 16 | ||
| 12+ | 95 | 97 | 200 | 100 | ✓ | 12 | ||
| 12+ | 92 | 89 | 200 | 100 | ✓ | 12 | ||
| 2 to 11 | 169 | 172 | 200 | 100 | ✓ | 12 | ||
| 12+ | 254 | 258 | 100/200 | 25 | ✓ | 12 | ||
| 12+ | 262 | 270 | 200 | 100 | ✓ | 24 | ||
| 12+ | 180 | 182 | 500 | 100 | ✓ | 12 | ||
| 18+ | 29 | 32 | 200/500/1000 | 100 | ✓ | 30 | ||
| 18+ | 42 | 21 | 200 | 100 | ✓ | 52 | ||
| 18+ | 93 | 93 | 1000 | 100 | ✓ | 26 | ||
| 4 to 11 | 181 | 181 | 200 | 100 | ✓ | 24 | ||
| 18+ | 41 | 41 | 200/500/1000 | 100 | ✓ | 52 | ||
| 12+ | 150 | 150 | 200/500/1000 | 100 | ✓ | 40 | ||
| 4 to 11 | 304 | 304 | 100 | 50 | ✓ | 12 | ||
| 12+ | 210 | 212 | 500 | 50 | ✓ | 12 | ||
| 12+ | 151 | 155 | 100 | 50 | ✓ | 12 | ||
| 15+ | 172 | 159 | 200 | 100 | ✓ | 16 | ||
| 15+ | 161 | 161 | 200 | 100 | ✓ | 16 | ||
| 12+ | 262 | 270 | 200 | 100 | ✓ | 24 | ||
| 12+ | 239 | 236 | 200 | 100 | ✓ | 52 | ||
| 18+ | 159 | 159 | 500 | 100 | ✓ | 24 | ||
| 12+ | 84 | 84 | 500 | 100 | ✓ | 12 | ||
| 12+ | 288 | 291 | 200/400 | 25 | ✓ | 12 | ||
| 18+ | 32 | 29 | 88/220/440 | 42 | ✓ | 16 | ||
| 16+ | 14 | 17 | 200 | 100 | ✓ | 12 | ||
| 18+ | 78 | 72 | 200 | 100 | ✓ | 12 | ||
| 12+ | 337 | 172 | 1000 | 100 | ✓ | 12 | ||
| 4 to 11 | 3107 | 3101 | 200/500 | 100 | ✓ | 26 | ||
| 12+ | 18 | 19 | 400 | 100 | ✓ | 12 | ||
| FSC: salmeterol/fluticasone; ICS: inhaled corticosteroid. | ||||||||
| Study ID | Treatment arm | Cause of death in adults (n) |
| salmeterol and fluticasone | Heroin overdose (1), stroke (1), and dyspnoea due to metastatic hepatic carcinoma (1) | |
| fluticasone | Aortic dissection (1), sudden cardiac death (1), gastroenteritis (1), sepsis (1), and stroke (2) | |
| salmeterol and fluticasone (separate inhalers) | Bronchial carcinoma (1) | |
| salmeterol and fluticasone | Myocardial infarction (2) and pneumonia (1) | |
| fluticasone | Myocardial infarction (2) | |
| salmeterol and fluticasone (separate inhalers) | Pneumothorax (1) | |
| salmeterol and fluticasone | Cardiac disease (1) | |
| fluticasone | Breast cancer (1) | |
| fluticasone | Cardiac arrest and deep vein thrombosis (1) | |
| fluticasone | Cardiac arrest (1) | |
| fluticasone | Ventricular hypertrophy and aortic hypoplasia (1) | |
| salmeterol and fluticasone | A female patient developed severe jaundice on day 30, which led to withdrawal of the study drug, and she died 6 weeks later. She had begun taking an herbal supplement before the onset of jaundice and then declined recommended interventions and investigations into her condition. The death was considered not related to study treatment | |
| fluticasone | Unknown cause (1) | |
| salmeterol and fluticasone | Leukaemia (1) |
| Peto OR meta‐analysis: regular salmeterol in addition to regular ICS vs ICS alone | ||||||||
| All‐cause mortality | Events Rx | Total Rx | Events control | Total control | Peto OR | CI start | CI end | Worst‐case NNTHa (from OR CI end) |
| Adults and adolescents | 11 | 14,233 | 13 | 13,718 | 0.80 | 0.36 | 1.78 | 1661 |
| Children | 0 | 4229 | 0 | 4224 | ‐ | ‐ | ‐ | ‐ |
| All cause non‐fatal SAE | ||||||||
| Adults and adolescents | 332 | 14,233 | 282 | 13,718 | 1.14 | 0.97 | 1.33 | 152 |
| Children | 65 | 4229 | 62 | 4224 | 1.04 | 0.73 | 1.48 | 139 |
| Asthma‐related mortality | ||||||||
| Adults and adolescents | 0 | 14,233 | 0 | 13,718 | ‐ | ‐ | ‐ | ‐ |
| Children | 0 | 4229 | 0 | 4224 | ‐ | ‐ | ‐ | ‐ |
| Asthma‐related SAE | ||||||||
| Adults and adolescents | 80 | 14,233 | 67 | 13,718 | 1.15 | 0.83 | 1.59 | 351 |
| Children | 29 | 4229 | 23 | 4224 | 1.25 | 0.72 | 2.16 | 128 |
| aNNTH (number needed to treat for an additional harmful outcome) calculated using Visual Rx to transform the upper end of the 95% confidence interval of the Peto OR, respectively. | ||||||||
| Risk difference meta‐analysis: regular salmeterol in addition to regular ICS vs ICS alone | ||||||||
| All‐cause mortality | Events Rx | Total Rx | Events control | Total control | Risk difference | CI start | CI end | Worst‐case NNTHa (from RD 95% CI end) |
| Adults and adolescents | 11 | 14,233 | 13 | 13,718 | ‐0.0002 | ‐0.0013 | 0.0009 | 1099 |
| Children | 0 | 4229 | 0 | 4224 | 0.0000 | ‐0.0013 | 0.0013 | 763 |
| All‐cause non‐fatal SAE | ||||||||
| Adults and adolescents | 333 | 14,233 | 283 | 13,718 | 0.0027 | ‐0.0008 | 0.0062 | 161 |
| Children | 65 | 4229 | 62 | 4224 | 0.0006 | ‐0.0047 | 0.0058 | 172 |
| Asthma‐related mortality | ||||||||
| Adults and adolescents | 0 | 14,233 | 0 | 13,718 | 0.0000 | ‐0.0009 | 0.0009 | 1099 |
| Children | 0 | 4229 | 0 | 4224 | 0.0000 | ‐0.0013 | 0.0013 | 763 |
| Asthma‐related non‐fatal SAE | ||||||||
| Adults and adolescents | 82 | 14,233 | 67 | 13,718 | 0.0007 | ‐0.0012 | 0.0026 | 385 |
| Children | 29 | 4229 | 23 | 4224 | 0.0014 | ‐0.0022 | 0.0049 | 205 |
| aNNTH (number needed to treat for an additional harmful outcome) calculated from the inverse of the upper end of the 95% confidence interval (CI end) of the risk difference. | ||||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 All‐cause mortality Show forest plot | 49 | Peto Odds Ratio (Peto, Fixed, 95% CI) | Subtotals only | |
| 1.1 Adults and adolescents | 41 | 27951 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 0.80 [0.36, 1.78] |
| 1.2 Children | 8 | 8453 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| 2 All‐cause non‐fatal SAE Show forest plot | 49 | Peto Odds Ratio (Peto, Fixed, 95% CI) | Subtotals only | |
| 2.1 Adults and adolescents | 41 | 27951 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 1.14 [0.97, 1.33] |
| 2.2 Children | 8 | 8453 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 1.04 [0.73, 1.48] |
| 3 All‐cause SAE (fatal and non‐fatal) Show forest plot | 49 | Peto Odds Ratio (Peto, Fixed, 95% CI) | Subtotals only | |
| 3.1 Adults and adolescents | 41 | 27951 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 1.12 [0.96, 1.31] |
| 3.2 Children | 8 | 8453 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 1.04 [0.73, 1.48] |
| 4 Asthma‐related SAE Show forest plot | 49 | Peto Odds Ratio (Peto, Fixed, 95% CI) | Subtotals only | |
| 4.1 Adults and adolescents | 41 | 27951 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 1.15 [0.83, 1.59] |
| 4.2 Children | 8 | 8453 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 1.25 [0.72, 2.16] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 All‐cause mortality Show forest plot | 49 | Risk Difference (M‐H, Fixed, 95% CI) | Subtotals only | |
| 1.1 Adults and adolescents | 41 | 27951 | Risk Difference (M‐H, Fixed, 95% CI) | ‐0.00 [‐0.00, 0.00] |
| 1.2 Children | 8 | 8453 | Risk Difference (M‐H, Fixed, 95% CI) | 0.0 [‐0.00, 0.00] |
| 2 All‐cause non‐fatal SAE Show forest plot | 49 | Risk Difference (M‐H, Fixed, 95% CI) | Subtotals only | |
| 2.1 Adults and adolescents | 41 | 27951 | Risk Difference (M‐H, Fixed, 95% CI) | 0.00 [‐0.00, 0.01] |
| 2.2 Children | 8 | 8453 | Risk Difference (M‐H, Fixed, 95% CI) | 0.00 [‐0.00, 0.01] |
| 3 All‐cause SAE (fatal and non‐fatal) Show forest plot | 49 | Risk Difference (M‐H, Fixed, 95% CI) | Subtotals only | |
| 3.1 Adults and adolescents | 41 | 27951 | Risk Difference (M‐H, Fixed, 95% CI) | 0.00 [‐0.00, 0.01] |
| 3.2 Children | 8 | 8453 | Risk Difference (M‐H, Fixed, 95% CI) | 0.00 [‐0.00, 0.01] |
| 4 Asthma‐related SAE Show forest plot | 49 | Risk Difference (M‐H, Fixed, 95% CI) | Subtotals only | |
| 4.1 Adults and adolescents | 41 | 27951 | Risk Difference (M‐H, Fixed, 95% CI) | 0.00 [‐0.00, 0.00] |
| 4.2 Children | 8 | 8453 | Risk Difference (M‐H, Fixed, 95% CI) | 0.00 [‐0.00, 0.00] |
| 5 Asthma‐related mortality Show forest plot | 49 | Risk Difference (M‐H, Fixed, 95% CI) | Subtotals only | |
| 5.1 Adults and adolescents | 41 | 27951 | Risk Difference (M‐H, Fixed, 95% CI) | 0.0 [‐0.00, 0.00] |
| 5.2 Children | 8 | 8453 | Risk Difference (M‐H, Fixed, 95% CI) | 0.0 [‐0.00, 0.00] |
