Steroids for Otitis Media With Effusion in Children
Systemic
Richard M.
Rosenfeld, MD, MPH; Ellen M. Mandel, MD; Charles D. Bluestone, MD
\s=b\ The role of oral steroids in the treatment of otitis media with effusion in children is the subject of ongoing debate. To critically evaluate the existing evidence, we performed a formal meta-analysis of six randomized clinical trials (n=264 children) selected from 14 articles and abstracts. Children receiving steroids for 7 to 14 days were 3.6 times more likely than placebotreated control subjects to have both ears free of effusion at the end of therapy (95% confidence interval, 2.2 to 4.1). This finding was essentially unchanged when the studies were weighted by a quality score, or stratified by use of concurrent antibiotic. The presence of significant heterogeneity among these studies suggests that additional trials will be needed to identify the specific subset(s) of children most likely to benefit from steroid therapy. (Arch Otolaryngol Head Neck Surg.
1991;117:984-989)
published articles,1"12 and two published abstracts,1314 the efficacy of oral steroid therapy for
Despite
12
chronic otitis media with effusion (OME) in children remains controver¬ sial. Considering that nearly half of these investigations were published subsequent to 1985, the ongoing na¬ ture of the debate is apparent. This renewed interest most likely reflects a dissatisfaction with the cure rates achieved by antimicrobial therapy alone, and a desire to develop effective Accepted for publication February 14,1991. From the Pittsburgh Otitis Media Research Center, Children's Hospital of Pittsburgh (Pa). Dr Rosenfeld is presently with the Department of Pediatric Otolaryngology, Children's National Medical Center, Washington, DC. Reprint requests to the Department of Pediatric Otolaryngology, Children's Hospital of Pittsburgh, 3705 Fifth Ave, Pittsburgh, PA 15213-2583 (Dr Rosenfeld).
medical alternatives for what is at present primarily a surgical disorder. In theory, corticosteroids would have much to offer in the treatment of middle-ear effusion (MEE). Various mechanisms have been proposed: di¬ rect anti-inflammatory action in the middle ear and the eustachian tube by reducing arachidonic acid, thereby in¬ hibiting the cyclo-oxygenase and lipoxygenase pathways for synthesis of in¬ flammatory mediators12,15; increase in eustachian tube surfactant, thus allow¬ ing for better tubai function3; shrink¬ age of peritubal lymphoid tissue, again allowing for better tubai function416; and reduction of middle-ear fluid vis¬ cosity by its action on mucoproteins.4 Since 30% to 50% of chronic effusions also contain bacteria,17 combined ste¬ roid and antibiotic therapy appears to be a logical approach. Treatment, however, must be based on more than logic; it must also be safe and of proven efficacy. Disseminated varicella with occasionally fatal out¬ come has been reported as an unusual complication of steroid therapy,18"20 even after a single short-term course a nonimmunocomprised patient.21 Although extremely rare, the potential for such an outcome highlights the need for proven efficacy of steroids prior to their use in the clinical setting. At present, this efficacy has not been convincingly demonstrated regarding
in
OME. As
an intermediate step between past and future research, we used
meta-analysis to provide a qualitative and quantitative assessment of the ex¬ isting evidence. The specific research hypothesis was that a short course of oral steroids, with
or
without
concur-
rent antibiotic, would be associated with increased odds of resolving MEE compared with untreated or placebotreated control subjects. Besides test¬ ing this hypothesis, we sought to as¬ sess consistency among trials, and to determine which subgroups of patients would be most likely to respond favor¬ ably to steroid therapy. A final objec¬ tive was to identify, if possible, the most promising directions for contin¬ ued research. Meta-analysis, like any other scien¬ tific endeavor, is subject to a number of unconscious biases that may lead to incorrect conclusions. To minimize this possibility, our protocol was based on
published guidelines.22,23 Emphasis was placed on the qualitative aspects of meta-analysis, and the implications of heterogeneity among studies. METHODS Literature Search and Selection of Trials
Computerized literature search combined with a manual check of all source references identified 14 published articles and abstracts through September 1990. Consulting with Current Contents, review articles, text¬ books, and several experts in the field failed to uncover any additional articles. No at¬ tempt was made to identify or obtain data from unpublished trials; instead, we chose to estimate the potential effect of publication bias (the tendency toward preferential publi¬ cation of research with significant results), using statistical techniques.24 Specific inclusion and exclusion criteria (Table 1) were applied to the initial data set to minimize selection bias. Judgments were based solely on information in the methods sections of the various articles; there were no disagreements among two of us (R.M.R. and E.M.M.) regarding which articles should be excluded (Table 2) and included (Table 3) in
Downloaded From: http://archotol.jamanetwork.com/ by a University of Arizona Health Sciences Library User on 06/05/2015
Table 1.—Inclusion and Exclusion Criteria for Meta-Analysis Inclusion criteria Randomized clinical trial Short-term course of peroral steroid (7 to 14
days)
Dichotomous outcome
completion
of
reported therapy
after
Exclusion criteria Published in abstract form only uncontrolled study Nonrandomized allocation of treatment groups Outcome reported only by ears, and not
Retrospective
or
by patients
Control and treatment groups differ by more than just steroid
pooled analysis. Abstracts were exclud¬ ed, since they could not be assigned a quality index (see below), and contained insufficient the
outcome data for pooling. The authors of both published abstracts were contacted to con¬ firm that the work was not subsequently published, or in press. Last, the selected articles were checked to insure that the
study populations represented independent samples.
Quality of Reporting quality index was calculated for each eligible trial using criteria adapted from Chalmer's method for assessing randomized A
'
"' An overall score was calcu¬ control trials. lated based approximately two thirds on measures of internal validity (scientific valid¬ ity of the research), and one third on external validity (generalizability of the results). Two of us (R.M.R. and E.M.M.) surveyed the articles independently to determine the de¬ gree to which each quality measure was sat¬ isfied (Table 4). Disagreements occurred on less than 10% of items, and were resolved after joint discussion and réévaluation of the article. A quality index from 0 to 1 was calcu¬ lated for each study based on the percentage of total points earned. This index was used for assigning relative weights to the studies in the pooled analysis, and should not be construed to represent an absolute measure of the inherent value of a particular article.
Outcome Measures The primary outcome measure in this study was status of MEE at the completion of 7 to 14 days of therapy as determined by the diagnostic method of the author's choice. No attempt was made to assess the efficacy of the intervention beyond this short period. Most studies reported the percentage of sub¬ jects with positive outcomes in both ears (complete cure), and in only one ear (partial
cure). Data from the studies were tabulated in contingency (2 2) table format and veri¬ fied for accuracy prior to statistical analysis.
Studies Excluded From
Table 2.
Meta-analysis
—
Source,
No. of
y
Reason for Exclusion
Subjects
Odds
Ratio* Pt
Steroid alone
Heisse,1 1963
Oppenheimer,2
27 1968
Woodhead et al,13 1986 Steroid + antibiotic Persico et al,3 1978 Puhakka et al,' 1985 Berman et al,'0 1987
162 18
uncontrolled Not randomized Abstract only
Retrospective,
3.5
NS* .
276
ratios
are
.
Not randomized
Outcome reported by ears§ Groups differ by more than steroid|| Daly et al,6 1987_42_Outcome reported at 6 weeksTI 102 Abstract only Heary et al,14 1990
*Odds
.
75
28
relative to the control group. two tailed. NS indicates that
tFisher's Exact Test,
3.8 5.7 5.5
Systemic
Richard M.
Rosenfeld, MD, MPH; Ellen M. Mandel, MD; Charles D. Bluestone, MD
\s=b\ The role of oral steroids in the treatment of otitis media with effusion in children is the subject of ongoing debate. To critically evaluate the existing evidence, we performed a formal meta-analysis of six randomized clinical trials (n=264 children) selected from 14 articles and abstracts. Children receiving steroids for 7 to 14 days were 3.6 times more likely than placebotreated control subjects to have both ears free of effusion at the end of therapy (95% confidence interval, 2.2 to 4.1). This finding was essentially unchanged when the studies were weighted by a quality score, or stratified by use of concurrent antibiotic. The presence of significant heterogeneity among these studies suggests that additional trials will be needed to identify the specific subset(s) of children most likely to benefit from steroid therapy. (Arch Otolaryngol Head Neck Surg.
1991;117:984-989)
published articles,1"12 and two published abstracts,1314 the efficacy of oral steroid therapy for
Despite
12
chronic otitis media with effusion (OME) in children remains controver¬ sial. Considering that nearly half of these investigations were published subsequent to 1985, the ongoing na¬ ture of the debate is apparent. This renewed interest most likely reflects a dissatisfaction with the cure rates achieved by antimicrobial therapy alone, and a desire to develop effective Accepted for publication February 14,1991. From the Pittsburgh Otitis Media Research Center, Children's Hospital of Pittsburgh (Pa). Dr Rosenfeld is presently with the Department of Pediatric Otolaryngology, Children's National Medical Center, Washington, DC. Reprint requests to the Department of Pediatric Otolaryngology, Children's Hospital of Pittsburgh, 3705 Fifth Ave, Pittsburgh, PA 15213-2583 (Dr Rosenfeld).
medical alternatives for what is at present primarily a surgical disorder. In theory, corticosteroids would have much to offer in the treatment of middle-ear effusion (MEE). Various mechanisms have been proposed: di¬ rect anti-inflammatory action in the middle ear and the eustachian tube by reducing arachidonic acid, thereby in¬ hibiting the cyclo-oxygenase and lipoxygenase pathways for synthesis of in¬ flammatory mediators12,15; increase in eustachian tube surfactant, thus allow¬ ing for better tubai function3; shrink¬ age of peritubal lymphoid tissue, again allowing for better tubai function416; and reduction of middle-ear fluid vis¬ cosity by its action on mucoproteins.4 Since 30% to 50% of chronic effusions also contain bacteria,17 combined ste¬ roid and antibiotic therapy appears to be a logical approach. Treatment, however, must be based on more than logic; it must also be safe and of proven efficacy. Disseminated varicella with occasionally fatal out¬ come has been reported as an unusual complication of steroid therapy,18"20 even after a single short-term course a nonimmunocomprised patient.21 Although extremely rare, the potential for such an outcome highlights the need for proven efficacy of steroids prior to their use in the clinical setting. At present, this efficacy has not been convincingly demonstrated regarding
in
OME. As
an intermediate step between past and future research, we used
meta-analysis to provide a qualitative and quantitative assessment of the ex¬ isting evidence. The specific research hypothesis was that a short course of oral steroids, with
or
without
concur-
rent antibiotic, would be associated with increased odds of resolving MEE compared with untreated or placebotreated control subjects. Besides test¬ ing this hypothesis, we sought to as¬ sess consistency among trials, and to determine which subgroups of patients would be most likely to respond favor¬ ably to steroid therapy. A final objec¬ tive was to identify, if possible, the most promising directions for contin¬ ued research. Meta-analysis, like any other scien¬ tific endeavor, is subject to a number of unconscious biases that may lead to incorrect conclusions. To minimize this possibility, our protocol was based on
published guidelines.22,23 Emphasis was placed on the qualitative aspects of meta-analysis, and the implications of heterogeneity among studies. METHODS Literature Search and Selection of Trials
Computerized literature search combined with a manual check of all source references identified 14 published articles and abstracts through September 1990. Consulting with Current Contents, review articles, text¬ books, and several experts in the field failed to uncover any additional articles. No at¬ tempt was made to identify or obtain data from unpublished trials; instead, we chose to estimate the potential effect of publication bias (the tendency toward preferential publi¬ cation of research with significant results), using statistical techniques.24 Specific inclusion and exclusion criteria (Table 1) were applied to the initial data set to minimize selection bias. Judgments were based solely on information in the methods sections of the various articles; there were no disagreements among two of us (R.M.R. and E.M.M.) regarding which articles should be excluded (Table 2) and included (Table 3) in
Downloaded From: http://archotol.jamanetwork.com/ by a University of Arizona Health Sciences Library User on 06/05/2015
Table 1.—Inclusion and Exclusion Criteria for Meta-Analysis Inclusion criteria Randomized clinical trial Short-term course of peroral steroid (7 to 14
days)
Dichotomous outcome
completion
of
reported therapy
after
Exclusion criteria Published in abstract form only uncontrolled study Nonrandomized allocation of treatment groups Outcome reported only by ears, and not
Retrospective
or
by patients
Control and treatment groups differ by more than just steroid
pooled analysis. Abstracts were exclud¬ ed, since they could not be assigned a quality index (see below), and contained insufficient the
outcome data for pooling. The authors of both published abstracts were contacted to con¬ firm that the work was not subsequently published, or in press. Last, the selected articles were checked to insure that the
study populations represented independent samples.
Quality of Reporting quality index was calculated for each eligible trial using criteria adapted from Chalmer's method for assessing randomized A
'
"' An overall score was calcu¬ control trials. lated based approximately two thirds on measures of internal validity (scientific valid¬ ity of the research), and one third on external validity (generalizability of the results). Two of us (R.M.R. and E.M.M.) surveyed the articles independently to determine the de¬ gree to which each quality measure was sat¬ isfied (Table 4). Disagreements occurred on less than 10% of items, and were resolved after joint discussion and réévaluation of the article. A quality index from 0 to 1 was calcu¬ lated for each study based on the percentage of total points earned. This index was used for assigning relative weights to the studies in the pooled analysis, and should not be construed to represent an absolute measure of the inherent value of a particular article.
Outcome Measures The primary outcome measure in this study was status of MEE at the completion of 7 to 14 days of therapy as determined by the diagnostic method of the author's choice. No attempt was made to assess the efficacy of the intervention beyond this short period. Most studies reported the percentage of sub¬ jects with positive outcomes in both ears (complete cure), and in only one ear (partial
cure). Data from the studies were tabulated in contingency (2 2) table format and veri¬ fied for accuracy prior to statistical analysis.
Studies Excluded From
Table 2.
Meta-analysis
—
Source,
No. of
y
Reason for Exclusion
Subjects
Odds
Ratio* Pt
Steroid alone
Heisse,1 1963
Oppenheimer,2
27 1968
Woodhead et al,13 1986 Steroid + antibiotic Persico et al,3 1978 Puhakka et al,' 1985 Berman et al,'0 1987
162 18
uncontrolled Not randomized Abstract only
Retrospective,
3.5
NS* .
276
ratios
are
.
Not randomized
Outcome reported by ears§ Groups differ by more than steroid|| Daly et al,6 1987_42_Outcome reported at 6 weeksTI 102 Abstract only Heary et al,14 1990
*Odds
.
75
28
relative to the control group. two tailed. NS indicates that
tFisher's Exact Test,
3.8 5.7 5.5
