The Effectiveness of IV β-Agonists in Treating Patients With Acute Asthma in the Emergency Department^*

Materials and Methods

Inclusion Criteria

To be eligible for inclusion in this review, a study had to meet all of the following criteria: (1) design, randomized controlled trials (RCTs) or quasi RCTs (allocation on days of the week, or some other method); (2) population, studies recruiting adult or pediatric patients with severe acute asthma from the ED (or its equivalent); (3) interventions, administration of IV (selective or nonselective) β-agonists vs the administration of placebo, other IV bronchodilators (ie, methylxanthines), or other inhaled selective or nonselective β-agonists; included studies could also use other recognized standard treatment (ie, corticosteroids); and (4) outcomes, pulmonary functions, vital signs, adverse effects, and clinical scores. Agreement for relevance for review was measured using κ statistics.

Study Identification

Electronic databases were searched from 1966 to 2000, exploding β-agonist treatment (all routes, formulations, and brand names) and asthma; the search was restricted to RCTs using standardized and validated strategies.⁶ This search was performed on MEDLINE, EMBASE, CINAHL, the Controlled Trials Register of the Cochrane Library, and the Cochrane Airway Review Group “Asthma and Wheeze” Registry. These Cochrane registries include studies identified by the Cochrane Collaboration through standardized searching and hand searching of journals for controlled clinical trials. Reference lists of all available primary studies and review articles were reviewed to identify potential relevant citations. Trials were not excluded on the basis of language. Included authors, major pharmaceutical producers of asthma medications (AstraZeneca, Boehringer Ingelheim, Glaxo, 3M Pharmaceuticals), and other asthma researchers were contacted regarding the existence of other published, unpublished, or interim results on β-agonist research. The reference lists from the search strategy was independently reviewed, and clearly irrelevant articles were discarded. If the title, abstract, or descriptors suggested any potential relevance, the full text article was retrieved. Agreement for “relevance for review” was measured using statistics. Each relevant article was then assessed by two independent, nonblinded reviewers for inclusion in this review. Agreement for relevance for inclusion was measured using κ statistics. Disagreement was resolved by consensus or third-party adjudication.

Quality Assessment of Trials

The previously validated Jadad 5-point scale (score 0 to 5) was used to assess randomization (0 to 2 points), double blinding (0 to 2 points), and withdrawals and dropouts (0 to 1 point).⁷ For allocation assessment, concealment was described as either adequate, inadequate, or unclear using Cochrane methodology.⁸ Two reviewers independently assessed quality, and interrater reliability was measured by using simple agreement, κ, and weighted κ statistics, with disagreement resolved by third-party adjudication.

Data Extraction

Data for the trials were independently extracted with a structured form by two blinded reviewers and entered into the software program (Review Manager Version 4.0.4; Cochrane Collaboration; Oxford, UK). Primary study authors were requested to confirm data extraction and provide additional clarification or information for the review. In cases where tables were unavailable, graphs were enlarged and values were approximated.

Data Analysis

All similar studies were pooled using random effects weighted mean differences (WMDs) for continuous variables, and random effects odds ratios (ORs) for dichotomous variables, with 95% confidence interval (CIs) where appropriate. With pooled effects, heterogeneity was tested using the Breslow-Day test; p < 0.05 was considered statistically significant. For those summary effect outcome measures with statistical heterogeneity, a priori subgroup analyses were classified on the following basis: (1) population, adult vs pediatric and severity of illness based on pulmonary function test (PFT) results; and (2) intervention, selective vs nonselective β-agonists, IV vs inhaled β-agonists, IV with inhaled vs inhaled β-agonists, IV β-agonists vs IV methylxanthines, and infusion vs bolus β-agonists. Sensitivity analyses were completed on the strength of methodologic quality (high vs low) and statistical method of analysis (random vs fixed effects).

Results

Systematic Review

The Airways Review Group database search revealed 976 references that represented 740 original publications (76%): 258 articles (35%) in EMBASE, 250 articles (34%) in MEDLINE, 2 articles (0.3%) from CINAHL, 224 articles (30%) from both MEDLINE and EMBASE, and 6 articles (0.7%) cited in all three. An independent review of the abstracts and titles of these publications identified 31 potentially relevant studies. The agreement for relevance was high (κ = 0.83). Twenty-four additional references were added from bibliographic searching of relevant articles and overviews; a total of 55 full-text articles were reviewed for inclusion. Unpublished literature was requested from pharmaceutical companies and the authors of all included studies, but none were identified. Of these 55 articles, a total of 15 studies (27%) were included in the overview (κ = 0.87). Of the 40 studies that were excluded, 30 studies (55%) were nonrandomized, 7 studies (13%) included treatment of patients with nonacute asthma or nonasthmatics, and 3 studies (5%) examined non-IV routes of administration. Table 1 illustrates that the evidence for intervention with IV βagonists spans a period of 25 years: 7 articles (47%) published in the 1970s, 5 articles (33%) from the 1980s, and 3 articles (20%) from the 1990s. Twelve of the studies (80%) were conducted in Europe, 1 study (7%) was conducted Asia, and 2 studies (13%) were conducted in Australia. No trials meeting our inclusion criteria were conducted in North America.

Methodologic Quality

Many of the included articles were double-blind, controlled trials; however, the methodologic quality varied across studies. Using the method of Jadad et al,⁷ seven studies (47%) were rated as “strong” (Jadad score 3 to 5) and eight studies (53%) were rated as “weak” (Jadad score 0 to 2). Agreement between the two independent assessments of study quality was high (κ ranged from 0.59 to 1.0 for each domain). There was no significant correlation between quality scores and the year of publication of the trial (Pearson r = 0.38, p = 0.17). Regarding concealment of allocation methodology, 5 studies (33%) were rated as having clearly blinded allocation and 10 studies (67%) were rated as having unclear allocation blinding (κ = 1.0). There was no statistically significant association between those studies that were rated as strong methodologically and those that had blinded allocation (χ² = 2.04, degrees of freedom [df] = 1, p > 0.05).

Study Design

Thirteen of the studies (87%) followed a parallel protocol, whereas 2 of the studies (13%) followed a crossover model.⁹¹⁰ Eleven studies (73%) introduced IV β-agonists immediately on entry. The remaining four studies introduced IV β-agonists within 30 to 75 min of study entry, during which time the patients received either inhaled β-agonists¹¹¹¹² or IV aminophylline.¹² There were three main treatment strategies utilized in the studies under review. Three studies¹¹¹¹² utilized strategy 1 where IV β-agonists were compared to inhaled β-agonist, where both groups of patients received a “run-in” phase of inhaled β-agonist therapy. Six studies^91314151617 utilized strategy 2, where IV β-agonists were compared with inhaled agents, with no inhalational therapy in the IV β-agonist arm. The remaining six studies^101819202122 utilized strategy 3, where IV β-agonists were compared with IV methylxanthines, where neither group received inhaled β-agonist therapy.

Populations

Participants were selected from a sample of patients who presented to the ED or its equivalent with severe acute asthma. All patients were admitted to the hospital. The majority of studies focused on adult patients only (age range, 15 to 65 years), with only three studies¹¹⁷²¹ evaluating children (age range, 0.8 to 14.7 years). The median sample size across the studies was 23, with a range of 13 to 176 patients. All studies enrolled patients with “severe” asthma, though there was variability in the parameters and definitions used for inclusion criteria. Nine studies^{91112131415161920} used vital signs (heart rate > 100 beats/min) and PFT results (< 20% expected) as primary inclusion criteria. Five studies^1214161921 required abnormalities in arterial blood gas (ABG) measurements. Four articles^10182122 listed simple clinical symptoms and signs of “severe shortness of breath or wheezing” as inclusion criteria. The inclusion criteria in two articles¹¹⁴ described standardized clinical assessment scales for severe asthma. These definitions of severity are described elsewhere.²³²⁴

Interventions

All patients received supplemental oxygen by face mask and systemic corticosteroids. No patients received inhaled steroids or inhaled anticholinergic agents in any of the studies. All studies used selective IV β₂-agonists. Table 1 demonstrates that nine studies^{1910121516202122} administered IV β₂-agonists as a bolus (range, 100 to 500 μg, or 4 to 15 μg/kg), whereas six studies^111314181921 administered the IV β₂-agonist as an infusion (range, 8.3 to 20 μg/min to total doses of 500 to 3,000 μg). Most studies (73%) used salbutamol; three studies evaluated terbutaline, and one study evaluated reproterol. One study²² ran a triple-parallel protocol comparing IV salbutamol vs IV terbutaline vs IV aminophylline. Consequently, this study was treated as two studies: IV salbutamol vs aminophylline, and IV terbutaline vs aminophylline.

Outcomes

Each article evaluated primary outcomes within a 2-h period. Nevertheless, six studies extended the observation interval longer: 3 h,²² 5 h,¹¹ 6 h,¹⁴ 24 h,¹²¹ and 36 h.¹⁷¹⁸ Multiple statistical tests were performed in each study, with a mean of 24 (varying from 0 to 80). No mention of adjustments for multiple testing were identified in these articles, and 11 articles (73%) made no mention of possible type I errors.

More than 240 individual outcome measurements were abstracted from the studies. Scores from a variety of symptom scales were occasionally used to describe outcomes; however, due to the different scores used, no pooled analyses were conducted. In addition, a number of PFT results were employed (including peak expiratory flow rate [PEFR], FEV₁, FVC, percentage of predicted PEFR, and percentage of predicted FEV₁); nonetheless, variability in the type of PFT employed limited comparisons between studies. There were no descriptions of any patients who were intubated or died during any of the study observation periods.

Five trials^911161819 used improvements in PFTs (namely PEFR) as the primary outcome. Five articles described a primary outcome variable of “clinical improvement,” but the definition varied widely between studies. Three of these studies¹⁰¹³¹⁵ relied on the “impression by the patient or physician of improvement in symptoms.” The remaining two studies used predefined clinical determinants of success. The first article defined three unique primary clinical measures of success: earlier ED discharge time (defined as the start of hourly inhaled salbutamol therapy), faster recovery time (to cessation of nebulae β₂-agonists every 30 min, and 60 min), and less oxygen dependence (defined at the 2-h window as the requirement for medical oxygen to maintain oxygen saturations > 93%).¹ The second article¹⁴ defined “clinical success” as the presence of at least two of the following points at 60 min: (1) a decrease in a “clinical index rating” of at least three points, (2) a decrease in Paco₂ of at least 3 mm Hg, and (3) an increase in PEFR of at least 50 L/min. Consequent to the variety of outcomes, only seven domains were analyzed in which sufficient data were available and similarly derived: serial PEFR, serial percentage of predicted PEFR, serial FEV₁, serial heart rate, serial ABG values, autonomic side effects, and clinical improvement.

Results

Pulmonary Function

Table 2 demonstrates that across the 6-h observation in the seven articles reporting PEFR, no statistical differences in PEFR were identified between those patients who received IV β₂-agonists vs inhaled β₂-agonists or IV methylxanthines. Moreover, differences between the summary outcome measures in each stratum were of questionable clinical significance with pooled estimates of treatment effect ranging from a 0.4 to 19.4 L/min. Over the course of 6 h, there were no statistically or clinically significant differences with respect to FEV₁ or percentage of predicted PEFRs (not shown). There was intermittent heterogeneity present in these analyses, and this is addressed further in the discussion.

HRs

Nine articles described heart rate results over a 6-h period recorded in Table 2. Over this time, there were higher heart rates in those patients who received IV β₂-agonists (range, 4.0 to 12.3 beats/min). These differences were statistically significant in the 15-min and 45-min periods, and the 2-h to 6-h strata, each of which provided homogeneous pooled estimates. However, the differences in heart rates are of questionable clinical significance.

ABG Measurements

Six articles described ABG measurements, and five articles described carbon dioxide tensions. There was no statistical difference in either the Pao₂ or Paco₂ between IV β₂-agonists and comparison treatments. In addition, there was no heterogeneity across any stratum.

Autonomic Side Effects

Despite concern regarding the potential side effects of IV β₂-agonists, only 10 studies (67%) reported this information. Autonomic effects included cardiovascular (palpitations, tachycardia, hypertension), neurologic (tremor, headache), and/or GI effects (nausea, vomiting). The pooled OR suggests that adverse effects were experienced approximately twice as frequently when receiving IV treatment as with the comparison treatment. Nevertheless, this result was not statistically significant and significant heterogeneity was present in the pooled estimate.

Clinical Improvement

Five articles reported a primary outcome variable of clinical improvement, but there was variability in the measurement. The pooled OR suggests that the proportion of patients who did not improve with IV therapy was the same as the proportion of patients who received the comparison treatment. Significant heterogeneity was present in this pooled estimate.

Subgroup Analysis

An insufficient number of pediatric articles with similar outcome measures were identified, and this precluded any subgroup comparison on the basis of age. An insufficient number of similar outcomes prevented any formal comparison of results based on type or dose of β₂-agonists. There was no statistical difference in any of the outcome domains when comparing β₂-agonists administered as an IV bolus vs infusion. There was no change in the trends of the summary statistics for any of the outcome domains when strategy 2 was compared to strategy 3. Too few studies with sufficient similar outcomes limited any meaningful comparison of strategy 1 vs strategy 2 or strategy 3.

Sensitivity Analysis: Methodologic Quality

Using the methods of Jadad et al⁷, a “strong” methodologic study was defined as having a Jadad score of 3 to 5, and a “weak” study as having a Jadad score of 0 to 2. From Table 3 it is evident that the stronger methodologic studies fail to demonstrate a clinical or statistical difference between IV β₂agonists or the comparison treatment in terms of PEFR and clinical success. Moreover, although not statistically significant, IV β₂-agonists appear to have an increased risk of adverse effects and increased heart rate compared to the control treatment. By comparing the two groups, it is clear that the weak methodologic studies had larger effect sizes, favoring the control treatment. Although these were statistically nonsignificant, the treatment effects from the weak methodologic were orders of magnitude larger or even discordant from the results of the methodologically strong studies. Subgroup analysis by fixed-effects modeling demonstrated no differences in results except for more strata with statistically significant lower serial heart rates for the non-IV groups (range, 0.1 to 14.1 beats/min).

Discussion

The literature has been conflicting regarding the use of IV β₂-agonists in patients with acute asthma, and this systematic review is the first to examine the available evidence of the effect of treating severe acute asthmatics with IV β₂-agonists. The subsequent meta-analysis included 15 randomized trials over 25 years that included 584 adults and children across nine countries. IV β₂-agonists administered either by bolus or infusion compared to inhaled β₂-agonists or IV methylxanthines did not lead to any statistically significant differences in pulmonary functions, laboratory measures of ventilation and oxygenation, or clinical descriptions of improvement. Subgroup and sensitivity analysis consistently demonstrated that the use of IV β₂-agonists was associated with an increased risk of autonomic side effects and higher heart rates; however, they were never shown to be statistically significant in this regard.

When examining the quality of articles involving IV agents in acute asthmatic presentations, it is obvious that greater care must be incorporated into further work if clarity is to emerge. There were broad discrepancies among outcomes from studies⁷⁸ where methodologic quality was scored using two accepted methods. Moreover, statistical planning and sample size calculations were not carefully considered in most studies. No studies were large enough to protect against type II error, and sample size calculations were rarely reported. Furthermore, multiple statistical testing was performed in many studies, increasing the risk of type I error.

The literature has examined three treatment strategies involving IV β₂-agonists. Originally, IV β₂-agonists were compared to IV aminophylline in most clinical trials in the 1970s and early 1980s (40% of the included articles). As the standard of care for asthma has been refined, the routine use of aminophylline has diminished, and inhaled β₂-agonists have been increasingly used.²³⁴⁵ Consequently, there was a shift in focus to compare IV vs nebulae β₂-agonists (40% of the included articles). Whether IV β₂-agonists improve bronchodilator response when administered in addition to nebulae bronchodilators was only addressed in three studies (20%) under review. Although the evidence suggests that IV β₂-agonists alone are no better than inhaled β₂-agonists, the role of IV β₂-agonists in addition to inhaled β₂-agonists remains unclear.

There are several potential limitations of this study. First, this review analyzed only the IV route of administration, and did not evaluate the literature on subcutaneous routes of administration. Based on the consistent lack of benefit shown with the IV β₂-agonists, it is unlikely that subcutaneous agents would differ in effect when compared to nebulized agents.

Second, there was significant heterogeneity in pooled estimates for many of the summary outcome measures despite the demonstration of similarities in design, populations, interventions, and outcome measurements between the 15 studies. Nevertheless, on further sensitivity analysis it appeared that studies of weak methodologic quality account for the majority of this heterogeneity. In particular, one article (Swedish Society, 1990)¹⁵ was responsible for the majority of the heterogeneity based on the following points: (1) differential methodologic quality, the Swedish Society article was rated as having low methodologic quality; (2) different populations, all articles studied patients with extremely severe asthma; however, the majority of studies enrolled patients with mean PEFRs in the range of 50 to 100 L/min, whereas the Swedish study evaluated patients with mean PEFRs in the range of 160 to 170 L/min (still defined as “severe < 200 L/min” by international guidelines); and (3) different cointerventions, the Swedish study did not administer any corticosteroid therapy until 2 h into the study protocol, whereas all other studies introduced corticosteroid therapy at the time of enrollment into the study. The effects of each of these factors on the homogeneity of the outcome domains were confounding in isolation and in whole by the very large sample size of the Swedish study (n = 176) in relation to the relatively smaller studies (sample range, 14 to 71). When analyzed by strong vs weak methodologic quality, the treatment effects favoring control were less pronounced in the methodologically stronger studies. This would suggest that the methodologically stronger studies fail to demonstrate a clear difference between the two comparison treatments.

Third, despite the intensive search strategy employed, there still exists a possibility of study selection bias or publication bias in this meta-analysis. For example, we may be erroneously estimating the nonsignificant effects of IV β₂-agonists due to missing unpublished negative or positive trial results. Although a comprehensive search of the published English and non-English literature for potentially relevant studies was conducted, using a systematic strategy to avoid bias, some studies may have still been missed. In addition, attempts were made to contact first and corresponding authors. Despite these endeavors, no unpublished or non-English articles were uncovered; still, we recognize that they may exist.

Fourth, the best outcome measure for “success” in treating acute asthma was measured variably between studies. Better standardization of this outcome would improve study comparability. Most studies included pulmonary functions as their primary outcome measures. The inherent variability of these PFTs in acute asthma, emphasizes the need for further research into alternative and valid measures. In addition, the evaluation of adverse side effects was complicated by a lack of standardized reporting.

Conclusion

Despite the methodologic limitations, the results of this work clarify the use of IV β₂-agonists in the treatment of severe acute asthma. The use of IV β₂-agonists compared to inhaled β₂-agonists or IV methylxanthines did not lead to any significant differences in pulmonary functions, laboratory measures of ventilation and oxygenation, or clinical failure/success. Although statistically nonsignificant, IV β₂-agonists produced slightly more autonomic side effects and tachycardia than do comparison treatments. Consequently, the clinical benefit appears questionable, while the potential clinical risks are obvious. The only recommendations for IV β₂agonist use should be in those patients in whom inhaled therapy is not feasible, or in the context of a controlled clinical trial comparing IV β₂-agonists with standard care vs standard care alone.