Objective. To evaluate the efficacy of β2-agonists in bronchiolitis.
Design. Critical review and meta-analysis of randomized controlled trials of inhaled β2-agonists.
Results. Three inpatient and five outpatient studies were identified. Inpatient studies (82 patients) were characterized by wide variability in therapeutic regimens and measurement of outcomes. Several problems were identified in the selection and specification of patients, such as failure to assess the prestudy duration of illness or to exclude patients already taking bronchodilators. Meta-analysis was not possible for inpatient trials due to the great variability in study outcomes, timing of outcome assessment, and drug regimens. Results of inpatient trials were contradictory: one found significant reductions in the clinical score and a shorter hospital stay with treatment, whereas two others found only significant oxygen desaturations.
Outpatient studies (251 patients) examined immediate changes in physiologic measures and clinical scores after two treatments, but have not examined the effects of the longer-term regimen customarily used by clinicians. Meta-analysis revealed that short-term β2-agonist therapy had no impact on the hospitalization rate or respiratory rate, and had a statistically significant but clinically insignificant impact on oxygen saturation and heart rate. The diversity of scoring systems precluded pooling of clinical score data.
Conclusions. Despite eight clinical trials, conclusive evidence for the efficacy of β2-agonist therapy for bronchiolitis remains unavailable. Well-designed inpatient trials are needed. Meta-analysis of outpatient studies does not support the use of β2-agonist therapy for bronchiolitis, but investigators have not studied the outcomes and the long-term outpatient regimen customarily used by clinicians.
Bronchiolitis is the most common lower respiratory tract infection in infants,1,2 afflicting 11% to 12%1,3 and hospitalizing 1% to 2%2,4 of all children in their first year. Infants less than 6 months old in day care were found to have an attack rate of 115 episodes of bronchiolitis per 100 children per year in one study.5 The leading cause of bronchiolitis, respiratory syncytial virus (RSV), results in over 54 000 hospitalizations and 2700 deaths annually in children less than 1 year old.6 According to serologic data,7 RSV infects half of infants by the end of their first year and nearly all by the end of their second year.
Given the prevalence and morbidity of bronchiolitis, an effective therapy is needed. Bronchodilators have been used in the treatment of bronchiolitis since the late 1950s, and continue to be the recommended therapy in the most recent editions of American pediatric textbooks.8-10 Bronchodilators are by far pediatricians' therapy of choice for bronchiolitis: in one series of bronchiolitic infants,11 99% were treated with bronchodilators.
Despite this common clinical practice, the available evidence on the efficacy of bronchodilator therapy for bronchiolitis is conflicting. The results of several randomized controlled trials (RCTs) indicated that bronchodilators are effective therapy for bronchiolitis,12-15 whereas others16-19 found that bronchodilators are no more effective than placebo. Furthermore, a Medline search and hand searches of the last 8 years of four major pediatric journals revealed that a formal meta-analysis of this bronchiolitis therapy has never been published. The goal of this study, therefore, was to evaluate the efficacy of bronchodilator therapy in bronchiolitis by critical review of published RCTs and by meta-analysis.
Identification of Studies
To identify studies of bronchodilator treatment of bronchiolitis, a Medline search was conducted using bronchiolitis and drug therapy both as medical subject headings and as text words for the published medical literature up to January 1, 1995. Additional studies were identified by review of the references from studies found by Medline. To assure that the process of identifying studies was comprehensive, we also conducted hand searches of the 1987 to 1994 editions of four major pediatric journals (American Journal of the Diseases of Children, Archives of the Diseases of Children, Journal of Pediatrics, andPediatrics). Authors of studies were contacted when additional data were needed. To identify unpublished RCTs, we contacted the Acute Respiratory Infection Review Group of the Cochrane collaboration.
Only RCTs of inhaled β2-agonists were included. Trials were excluded if all subjects received concomitant therapy (such as other bronchodilators or corticosteroids) in addition to β2-agonists. We focused on clinical response outcomes of greatest interest to clinicians, and thus excluded trials that used outcomes evaluable only in a pulmonary laboratory. We also excluded RCTs that enrolled all wheezing infants regardless of the diagnosis, instead of limiting eligibility only to infants with bronchiolitis. Although there is some disagreement about the precise definition of bronchiolitis, we accepted the following criteria used in most studies and texts: an acute communicable disease presenting predominantly in infancy that is characterized by wheezing (with or without cough, tachypnea, and increased respiratory effort) accompanied by evidence of a viral illness (such as coryza and fever).
Studies published in languages other than English were considered, because they have been shown to be of comparable methodological quality to those published in English.20 Although we had also intended to review published RCTs of oral β2-agonist therapy for bronchiolitis, the small numbers of studies18,19 and total patients precluded meta-analysis.
To ensure the quality of the critical review of the literature and the meta-analysis, the criteria of Sacks et al21were used to evaluate study design and combinability, to avoid potential bias, to perform statistical analysis, and to consider application of results. We also focused on several additional aspects of research design that were considered to be particularly important in the evaluation of therapeutic interventions for bronchiolitis.
Critical Appraisal of Research Design Components
In the critical appraisal of studies, three components of research design were evaluated. First, studies were checked for the presence of problems in selection and specification of patients. Exclusion of those already taking bronchodilators would seem to be essential, because these children may have already achieved maximal bronchodilation. We also considered chest radiographs at study entry to be important for research purposes, to avoid enrollment of subjects with conditions that cause wheezing other than bronchiolitis, such as pneumonia. Exclusion of patients with asthma was viewed as necessary because their inclusion may distort therapeutic response. Statement of the explicit criteria used to make the diagnosis of bronchiolitis is needed to ensure consistent enrollment of suitable patients.
In the specification of study patients, documentation of prestudy duration of illness is needed because early bronchiolitis may be more responsive to therapy than late disease, when alveolar debris and strands of fibrin often form dense plugs that may obstruct small airways.22 In addition, it is reasonable to evaluate baseline attributes that may affect disease severity and therapeutic response, such as recurrent wheezing, a history of prematurity and associated respiratory complications (such as bronchopulmonary dysplasia), and a family history of atopy. Younger infants tend to get more severe bronchiolitis,23,24 so it is also helpful to specify the proportion of study patients less than 6 months old, or to distinguish treatment efficacy by age of the subjects.
The second component was variation in therapeutic maneuvers, including the type of β2-agonist administered, the dose, the dose interval, the total number of doses, the vehicle for nebulization (oxygen or compressed air), and the type of nebulizer used (mechanical or foot pump).
The third component evaluated was variability in study outcomes. We compared the specific types, total number, and timing of the outcomes.
A meta-analytic overview was performed for each outcome that was judged to be similar among trials. For dichotomous outcomes, such as the hospital admission rate, the overall event rate difference Mantel-Haenszel estimator [ERDMH] and its 95% confidence interval (CI) were calculated, using a pooled point estimate.25,26 The ERDMH is analogous to other Mantel-Haenszel estimators of effect and is weighted similarly.25,26 For continuous variables, such as respiratory rate, the overall weighted mean difference and its 95% confidence interval were calculated.26 Statistical heterogeneity in effects among trials was determined for respiratory rate, which was considered the main outcome in the meta-analysis. Heterogeneity was calculated using Cochran's Q.27
Identification of Studies
The initial Medline search generated 161 articles on the use of bronchodilators in bronchiolitis. No unpublished RCTs of β2-agonists have been located by the Acute Respiratory Infection Review Group of the Cochrane collaboration. Limitation of articles identified by the Medline search to RCTs yielded 20 studies, all of which were published in English. Further restriction to trials of β2-agonists that focused on infants with bronchiolitis (rather than all wheezing diagnoses) and used routine clinical rather than pulmonary laboratory outcomes yielded eight studies, three inpatient and five outpatient.
The three inpatient RCTs appeared from 1987 to 1992. In only one12 did the investigators conclude that β2-agonists are an effective therapy for bronchiolitis, finding statistically significant improved clinical scores and shorter mean hospital stays with treatment. In another study,17there were no differences in clinical scores and hospital stays between the treatment and control groups, and consistent with the third study,16 actually found that treatment led to a significant reduction in oxygen saturation. All three studies adequately randomized study subjects. One study12 did not use double-blinding in the research design; it was not stated in the methods section whether single blinding to treatment allocation was performed for only the researchers or only the subjects (and their parents).
Selection and Specification of Patients
Table 1 summarizes the problems and variation in selection and specification of patients that were identified in the inpatient trials. In the selection of patients, none of the studies performed baseline chest radiographs or excluded patients who were already taking bronchodilators at the time of randomization, and one of the studies did not exclude patients with asthma. One study limited enrollment exclusively to infants whose nasopharyngeal aspirate yielded RSV by immunofluorescence. All trials, however, explicitly stated the criteria used to make the diagnosis of bronchiolitis.
In the specification of patients, none of the trials determined the prestudy duration of illness, and only one fully evaluated important baseline patient attributes. The age of study subjects varied greatly: one trial enrolled only infants less than 6 months old (range, 3 weeks to 6 months), another infants less than 1 year old (range, 1 month to 11 months old), and the third only infants 2 months to 2 years old. One study did not specify the proportion of infants less than 6 months old. Viral etiology testing was highly variable: four specific viruses were tested for in one study, RSV only in the second, and no testing was performed in the third. Only one trial explicity stated the emergency department (ED) criteria used to admit patients.
Variation in Therapeutic Maneuvers
Therapeutic maneuvers and co-maneuvers varied among the inpatient trials (Table 2). Albuterol was the β2-agonist used in two trials, whereas the third used fenoterol. One study gave the same β2-agonist dose to all patients regardless of their weight, but the other two dosed by weight, either by milligrams or by milliliters of solution. The dose interval differed in each study, ranging from as little as 30 minutes to as long as 6 hours. The total number of doses given to subjects also differed, ranging from a single dose to as many as 18. Only one study used oxygen for the delivery of nebulizations to all patients; the others used primarily compressed air, with use of oxygen occasionally or only in those already receiving it.
Variability in Outcomes
Table 3 summarizes the variability in the type, total numbers, and timing of outcomes in the inpatient trials. One study used oxygen saturation as the only outcome, another used a clinical scoring system and the duration of the hospital stay, and the third used all three. The timing of outcome measurement also varied greatly. One trial assessed outcomes solely during the nebulization treatment and for 30 minutes after, another trial checked outcomes only once (either on the day of discharge or hospital day 3), and the third evaluated outcomes three times in the first 24 hours, then daily until discharge.
This substantial variability in patients, treatments, and outcomes precluded meta-analytic summary of outcome data for inpatient trials.
The five outpatient RCTs that were evaluated appeared from 1990 to 1994. Three studies concluded that β2-agonists are an effective therapy for bronchiolitis. One study13 found that treatment led to a statistically significant reduction in the respiratory rate, improvement in the accessory muscle score, and increases in the oxygen saturation and heart rate. Another study15 reported significant improvements in the wheezing score and total clinical score with therapy, but a decrease in the oxygen saturation after the first treatment, which improved after the second treatment. The third study14 found only significant improvement in the clinical score, along with an increased heart rate. The remaining two studies18-19 did not find statistically significant improvements in any of the outcomes evaluated, and one19 reported an increase in the heart rate in the treatment group compared with the oral placebo group. All five studies used appropriate randomization techniques and double blinding.
Selection and Specification of Patients
Table 4 summarizes the problems and variation in selection and specification of patients that were identified in the outpatient trials. In the selection of patients, only two of the five trials explicitly stated the criteria used to make the diagnosis of bronchiolitis. Baseline chest radiographs were also performed in only two studies, and three studies excluded patients already taking bronchodilators. All outpatient trials excluded patients with asthma, and none limited enrollment to only patients with documented RSV infection.
In the specification of patients, three trials assessed prestudy duration of illness, and four fully evaluated important baseline attributes. The age of patients was approximately the same in all five studies, but two did not specify the proportion of children less than 6 months old. Viral etiology testing was limited to only RSV in two trials, but included up to four agents in the other trials. All five trials tested for RSV, whereas two tested for influenza, two for parainfluenza, two for adenovirus, and one for cytomegalovirus.
Variation in Therapeutic Maneuvers
Table 5 summarizes the variation in therapeutic maneuvers and co-maneuvers in the outpatient trials. The β2-agonist used in all studies was albuterol. Three trials administered an albuterol nebulization dose of .15 mg/kg, one used a lower dose, and the other used .15 mg/kg for patients weighing >7 kg, but a uniform dose of 1 mg for all patients ≤7 kg. Four trials gave two doses of albuterol 30 minutes apart, but the fifth trial gave three doses at 60-minute intervals. Nebulizations were delivered with oxygen in three trials, and compressed air in two. Mechanical nebulizers were used in all but one study; the remaining study used a foot pump.
None of the outpatient trials examined the albuterol regimen customarily used by many pediatricians for the treatment of bronchiolitis: nebulizations in the office or ED, followed by several days of home oral therapy (or occasionally home nebulizations).
Variability in Outcomes
Four different clinical scores were used in the five outpatient trials. One study13 used two simple scores, a 3-point accessory muscle score based on severity of retractions and nasal flaring, and a second 3-point wheezing score based on the audibility of wheezes and their timing in the respiratory cycle. Two studies14,15 used the Respiratory Distress Assessment Instrument (RDAI), a 17-point score composed of six patient attributes that include the location and respiratory cycle timing of wheezes, and the severity of accessory muscle retractions in each of three locations. The two remaining studies18,19 used a 27-point Severity of Illness Score comprised of nine independently scored patient attributes: grunting, nasal flaring, supraclavicular retractions, intercostal retractions, air entry, air hunger, duration of wheezing during the respiratory cycle, location of wheezes, and general appearance. The clinical scoring systems were considered too diverse for meta-analysis, because of the substantial differences in the variety and number of patient attributes scored and in the maximum total achievable score.
All outpatient trials assessed the following short-term effects after two to three albuterol treatments: changes in physiologic variables (respiratory rate, oxygen saturation, and heart rate), and the hospital admission rate. Because the outcomes and their measurement were the same in all trials, meta-analysis was possible.
Although the short-term effects of therapy were assessed in outpatient RCTs, none of the trials evaluated the longer term effects of extended therapy sought by many pediatricians, namely, rates of return visits to the ED, and long-term rates of hospitalization (instead of the admission rate only immediately after therapy).
Tables 6 and 7 summarize the meta-analysis of similar short-term outcome variables in the five outpatient RCTs, containing a total of 129 patients in the pooled β2-agonist group and 122 patients in the pooled placebo group.
Compared with placebo, β2-agonist therapy had no impact on the hospitalization rate (Table 6). None of the event rate differences for individual trials were statistically significant. Indeed, overall there were 2% more hospitalizations in patients treated with β2-agonists, but this finding also was not statistically significant.
β2-agonist therapy had little or no impact on physiologic status (Table 7). Although the mean difference in the respiratory rate between the treatment and placebo groups after two treatments was as great as −6.3 in one trial, the 95% CIs reveal that neither the overall mean difference nor the mean differences of most of the individual trials were statistically significant. There was no evidence of statistical heterogeneity across trials in the respiratory rate results (χ2 = 6, df = 4,P < .5). β2-agonist therapy resulted in statistically significant increases in oxygen saturation and heart rate, but these findings are clinically insignificant. Although two RCTs found a statistically significant improvement of up to 2.8% in oxygen saturation with β2-agonist therapy, pooling of all trials yielded a clinically insignificant increase of 1.2%. Similarly, the mean differences in two trials indicated a statistically significant increase in heart rate of as much as 15 beats per minute (bpm), but pooling revealed a clinically insignificant overall increase of only 1.4 bpm.
Despite eight RCTs, conclusive evidence for the efficacy of β2-agonist therapy for bronchiolitis remains unavailable. In the case of inpatient studies, the absence of evidence can be attributed to the lack of well-designed trials. We found that inpatient studies were characterized by diverse problems in the selection and specification of patients, most notably inclusion of those already taking bronchodilators, not assessing prestudy duration of illness, and not always explicitly stating the criteria used to admit patients. Inpatient study drug regimens varied considerably, particularly in the dose, dose intervals, and total number of doses administered. The type and total number of study outcomes and their timing were also highly variable. We suggest that future inpatient studies pay careful attention to clinically-relevant attributes when enrolling and specifying study patients. Consistent, well-accepted criteria for study enrollment and admission to the inpatient ward must be adhered to and explicitly stated. Drug regimens should resemble those customarily used on most inpatient wards. Study outcomes should include those of greatest utility and relevance to the clinician, including duration of hospital stay, physiologic measures, and clinical scores.
Data from RCTs have not demonstrated that β2-agonists are effective in the outpatient treatment of bronchiolitis. Meta-analysis revealed that β2-agonist therapy did not have a clinically significant impact on physiologic measures or hospital admission rates. However, lack of evidence for the effectiveness of outpatient β2-agonist therapy may be attributable to limitations in the therapeutic regimens and the outcomes studied to date. Outpatient clinical trials have examined only very short-term responses (30 minutes to 1 hour after therapy) to two or three treatments. Pediatricians customarily treat outpatient bronchiolitis with initial nebulizations, followed by several days of home oral therapy, but this regimen has never been studied. Furthermore, particular outcomes of clinical interest in this extended customary regimen would be the rate of return visits to the office and/or ED, and the overall rate of hospital admission during the course of the illness (rather than the admission rate only immediately after therapy), but these outcomes also have never been studied. When used in the extended customary regimen, β2-agonists may prove to be extremely effective, or, alternatively, useless, but appropriate studies to answer this question have not yet been performed.
Our critical analysis also revealed that studies have not adequately addressed the possibility that β2-agonists are particularly effective therapy for certain subgroups of bronchiolitic patients. For example, therapeutic response may be especially high early in the course of bronchiolitis. Although three of the five outpatient trials assessed the prestudy duration of illness at baseline, none of the three inpatient trials did so, and not one of the eight outpatient or inpatient trials performed a stratified analysis by duration of illness to determine whether patients with early disease are more responsive to therapy. Other factors that have not been adequately investigated but may show important differences in subgroup analysis include age, specific viral etiologies, and severity of illness.
Clinical scoring systems are important in the evaluation of therapeutic response in bronchiolitis because they take into account key measures of severity of illness, such as intensity of wheezing and use of accessory muscles, that are not necessarily reflected in simple physiologic variables like respiratory rate. However, crucial issues in the use of clinical scores were noted as a result of critical reappraisal of the RCTs. The diversity of clinical score components and total scores presents substantial problems in consistency, interstudy comparison, and utilization by clinicians. It is difficult, if not impossible, to compare a simple 3-point score that only evaluates severity of wheezing to a 17-point, 6-component score, and to a 27-point score assessing nine different patient attributes. Because of this substantial diversity, combining separate scores from each trial by using such pooling techniques as z-transformed values was not considered feasible or useful.
Depending on which particular score was used, studies reaching the identical conclusion that therapy results in a significant improvement in clinical score can have substantially different clinical implications. The diversity of clinical scores also makes it difficult for the clinician to choose a reliable instrument for assessing severity of illness and for monitoring response to therapy. Until head-to-head comparisons are made of the individual scoring systems, the RDAI would seem to be the clinical score of choice for future studies, since it has been shown to have high inter-rater reliability, and its two subscores for wheezing and retraction were consistently as good in evaluating change as the multiple respiratory variables used in other studies (such as general appearance, grunting, and so forth).28 It is possible that better studies of β2-agonist therapy using a single, reliable clinical score may find significant improvements in the clinical score without concomitant improvements in physiological variables.
Evidence that conclusively supports or refutes the efficacy of β2-agonist therapy for bronchiolitis is unavailable at present. Critical appraisal of important research components of published clinical trials resulted in the identification of problems and sources of variation, which can provide future researchers with a set of informal guidelines and a checklist of avoidable pitfalls. Conclusive evidence on the efficacy of β2-agonist therapy for bronchiolitis will require a well-designed, multicenter RCT that examines clinically relevant outcomes and the regimens customarily used by pediatricians.
Note Added in Proof.
While this article was in press, a meta-analysis of bronchodilator therapy of bronchiolitis was published (Arch Pediatr Adolesc Med.1996;150:1166–1172). Although the authors found mild improvement in clinical scores with treatment, the study did not specifically focus on the efficacy of β2-agonists, nor address methodological issues.
This work was supported in part by a Robert Wood Johnson Clinical Scholars Program Fellowship to the first author.
We thank Drs Howard Bauchner, David Felson, Robert Merrill, and Joel Alpert for critically reviewing early manuscript drafts. We are grateful to Drs Anne Gadomski, Terry P. Klassen, Suzanne Schuh, and Paula Schweich for providing additional unpublished data from their studies. Dr Elaine Wang kindly provided information from the Acute Respiratory Infection Group of the Cochrane Collaboration.
- Received January 31, 1996.
- Accepted July 26, 1996.
Reprint requests to (G.F.) Division of General Pediatrics, Boston Medical Center, 818 Harrison Ave, Maternity 419, Boston, MA, 02118.
↵‡ Dr Flores conducted a substantial portion of this research while he was a Robert Wood Johnson Clinical Scholar at the Yale University School of Medicine, New Haven, CT.
Presented in part at the joint Ambulatory Pediatric Association/Society for Pediatric Research/American Pediatric Society annual meeting, San Diego, CA, May 9, 1995, and at the American Federation for Clinical Research Eastern Regional Meeting, Boston MA, October 28, 1995.
- RSV =
- respiratory syncytial virus •
- RCT =
- randomized controlled trial •
- ERDMH =
- event rate difference •
- CI =
- confidence interval •
- ED =
- emergency department •
- RDAI =
- Respiratory Distress Assessment Instrument •
- bpm =
- beats per minute
- ↵Phelan PD, Olinsky A, Robertson CF. Respiratory Illness in Children. Oxford, England: Blackwell Scientific Publications; 1994:71–77
- ↵Institute of Medicine. Appendix N: prospects for immunizing against respiratory syncytial virus. In: New Vaccine Development: Establishing Priorities—I: Diseases of Importance in the United States. Washington, DC: National Academy Press; 1985:397–409
- ↵Hanson IC, Shearer WT. Bronchiolitis. In: Oski FA, ed. Principles and Practice of Pediatrics. 2nd ed. Philadelphia, PA: JB Lippincott Co; 1994:1457–1459
- ↵Stern RC. Acute bronchiolitis. In: Behrman RE, ed. Nelson Textbook of Pediatrics. 14th ed. Philadelphia, PA: WB Saunders Co; 1992:1075–1076
- ↵Tooley WH. Bronchiolitis. In: Rudolph AM, ed. Rudolph's Pediatrics. 19th ed. Norwalk, CT: Appleton & Lange; 1991:1520–1521
- Ho L,
- Landau LI,
- Le Souef PN
- Wang EEL,
- Milner R,
- Allen U,
- Maj H
- Gadomski AM,
- Lichenstein R,
- Horton L,
- King J,
- Keane V,
- Permutt T
- Simpson H,
- Matthew DJ,
- Inglis JM,
- George EL
- McMillan JA,
- Tristram DA,
- Weiner LB,
- Higgins AP,
- Sandstrom C,
- Brandon R
- ↵Rothman KJ. Modern Epidemiology. Boston, MA: Little, Brown, & Co; 1986
- ↵Bracken MB. Statistical methods for analysis of effects of treatment in overviews of randomized trials. In: Sinclair JC, Bracken MB, eds. Effective Care of the Newborn. Oxford, UK: Oxford University Press; 1992:13–18
- Cochran W
- Lowell DI,
- Lister G,
- Von Koss H,
- McCarthy P
- Copyright © 1997 American Academy of Pediatrics