PEDIATRICS Vol. 108 No. 5 November 2001, pp. 1234-1235

Effect of Inhaled Corticosteroids on Growth

To the Editor.

I read with interest the meta-analysis of Sharek and Bergman¹ regarding the effect of inhaled corticosteroids (ICS) on growth, and felt it was important to address 2 issues in response. First, the authors' analysis of the growth effect seen during treatment with fluticasone propionate (FP) 200 µg/day indicated a small but statistically significant reduction in growth rate compared with placebo. This differs from the nonsignificant P value reported in our original paper,² and an explanation for this discrepancy is required. Second, the conclusions arrived at by the authors are nevertheless weakened by a failure to review other growth studies in which FP was the active comparator. The result is a message that may be confusing to practitioners caring for children with asthma.

I would like to begin by addressing the different conclusion arrived at by Drs Sharek and Bergman regarding the effect on growth with FP 200 µg/day. As indicated in our study, we compared the effect of FP 100 µg/day and FP 200 µg/day with placebo in prepubescent children.² The data for the children who remained prepubertal throughout the study were analyzed by analysis of variance (ANOVA) controlling for investigator. We reported a nonsignificant P value of .313, which supported our hypothesis that growth was not significantly impaired after 1-year treatment with FP 100 µg/day or 200 µg/day. I submit that Drs Sharek and Bergman most likely did not have the complete FP data available to them for their meta-analysis. Hence, it appears that the 95% confidence interval reported in their paper was calculated from the raw mean data that we reported along with the sample sizes obtained from Table 1 in our paper, which described the clinical characteristics of the prepubertal children at screening. The number of prepubertal children who actually completed the trial was less than that indicated by Table 1. The numbers of prepubertal children treated with placebo or FP 200 µg/day who completed the study were 57 and 79, respectively. The analysis in our paper used these smaller sample sizes and controlled for investigator interaction effects. Using this same basis for analysis, one would calculate a 95% confidence interval of (0.86, 0.1). This confidence interval includes the zero value and supports the conclusion of our original paper. To not use the smaller sample sizes increases the probability of committing a type 1 error. In addition, including a parameter (used in the model to calculate P values and confidence intervals) for "investigator interaction effects" controls for the potential of asthmatic children with a specific disease severity being recruited at some, but not all, sites. Likewise, as height is measured at each research site, with the data pooled among all sites, the investigator interaction parameter controls for potential inconsistency in stadiometric height measurements by the different study-site coordinators. In their analysis of the data, I do not believe that Drs Sharek and Bergman took this parameter into consideration. Furthermore, as indicated in our paper, we believe that mean change from the baseline growth velocity more accurately assesses the effects of inhaled steroids on growth. As such, we reported no effect of FP 200 µg/day on this parameter, with an overall P value of .380 by ANOVA; a pairwise comparison of the prepubertal children who completed the trial and received either placebo or FP 200 µg/day resulted in a P value of .223 with a 95% confidence interval of (0.83, 0.25).

The robustness of the conclusions of Drs Sharek and Bergman with respect to FP is undermined by the paucity of the data presented. In their search strategy, the authors excluded trials with nonsteroid control arms. This eliminates head-to-head comparisons of ICS that provide the practitioner with relevant information regarding potential for adverse growth effects. Although active control studies could not be included in the meta-analysis based on the authors' selection criteria, they could have been included in the discussion for comparative purposes. The studies of de Benedictis et al³ (FP vs beclomethasone), Ferguson et al⁴ (FP vs budesonide), and Price and colleagues⁵ (FP vs cromolyn) demonstrate that FP has significantly less effect on growth than beclomethasone³ or budesonide⁴ at clinically equivalent doses, and a similar effect when compared to cromolyn.⁵ This reduced effect of FP on growth could have been clearly illustrated in Figure 1 of the article, but the weighted mean difference (WMD) for FP 200 µg/day was inexplicably excluded from this figure.

When considering potential systemic effects of ICS, it is important to keep in perspective the relative benefits and risks of ICS therapy for asthma. The recently published prospective study by Agertoft and Pedersen⁶ demonstrated that the administration of inhaled budesonide to asthmatic children had no effect on these children attaining final adult height, which was similar to asthmatic children who did not receive inhaled steroids, as well as healthy children. Furthermore, Suissa et al⁷ recently showed that the regular use of low-dose ICS is associated with a decreased risk of death from asthma. In their study, the rate of death from asthma among users of ICS decreased by 21% for every additional canister used during the previous year and by 54% for every canister used in the previous 6 months. In both children and adults, the risk of systemic effects of ICS, already markedly reduced compared with oral corticosteroids, can be minimized by titrating to the lowest effective dose.

Although the methodology is admirable, the authors' emphasis on meta-analytical technique obscures the central message of their manuscript. Many paragraphs describing statistical tests assure the reader that the proper route to a meta-analysis has been followed. However, the results do not allow for any generalization nor do they provide the medical professional with any clear sense of differences among ICS or differences among doses. Consequently, this meta-analysis does not fulfill its potential to enhance the full picture of ICS and their use in pediatric asthma.

David B. Allen, MD
Department of Pediatrics
Pediatric Endocrinology and Pediatric Residency Training
University of Wisconsin Children's Hospital
Madison, WI 53792-4108

REFERENCES

1. Sharek PJ, Bergman DA. The effect of inhaled steroids on the linear growth of children with asthma: a meta-analysis. Pediatrics. 2000;106(1). Available at: http://www.pediatrics.org/cgi/content/full/106/1/e8
2. Allen DB, Bronsky EA, LaForce CF, Nathan RA Growth in asthmatic children treated with fluticasone propionate. J Pediatr 1998; 132:472-477 [CrossRef][Medline]
3. de Benedictis FM, Medley HV, Williams L. Long-term study to compare safety and efficacy of fluticasone propionate (FP) with beclomethasone dipropionate (BDP) in asthmatic children. Eur Respir J. 1998;12(suppl);142S
4. Ferguson AC, Spier S, Manja A, Versteegh GA, Mark S, Zhang P Efficacy and safety of high-dose inhaled steroids in children with asthma: a comparison of fluticasone propionate with budesonide. J Pediatr. 1999; 134:422-427 [CrossRef][Medline]
5. Price JF, Russell G, Hindmarsh PC, Weller P, Heaf DP, Williams J Growth during one year treatment with fluticasone propionate or sodium cromoglycate in children with asthma. Pediatr Pulmonol. 1997; 24:178-186 [CrossRef][Medline]
6. Agertoft L, Pedersen S Effect of long-term treatment with inhaled budesonide on adult height in children with asthma. N Engl J Med. 2000; 343:1064-1069 [Abstract/Free Full Text]
7. Suissa S, Ernst P, Benayoun S, Baltzan M, Cai B Low-dose inhaled corticosteroids and the prevention of death from asthma. N Engl J Med. 2000; 343:332-336 [Abstract/Free Full Text]

In Reply.

Thank you for the opportunity to reply to the letter to the editor by Dr Allen. Below, we address these concerns in the order presented by Dr Allen.

The first major concern expressed by Dr Allen is the difference in statistical significance we obtained for the randomized, controlled clinical trial of Allen et al¹ included in our meta-analysis.² As described in our work, we only included studies that provided the direct outcome of linear growth velocity or data convertible to linear growth velocity as an outcome. One purpose, and a major strength of the meta-analysis technique, is to combine similar data to provide a larger number of study participants yielding a more accurate estimate of effect size. This approach combines linear growth velocity (cm per year) data from each study included in the analysis. In Dr Allen's study, the control and intervention groups were effectively randomized. Participants from each group were similar in demographics, clinical characteristics, concurrent asthma medications, oral steroid usage during the study, and compliance rates. The assumption that occurs in a randomized trial is that control and intervention subjects are similar with respect to unknown confounders as well. This assumption is valid if randomization was performed effectively. Given the method of randomization used, use of an off-site envelope, we believed this to be true. Data from Dr Allen's study was therefore directly abstracted and entered into the Review Manager Version 3.1 statistical package (Cochrane Collaboration, Oxford, England) used by the Cochrane Collaboration,³ revealing a significant difference between the participants using fluticasone propionate 200 µg/day and placebo. Efforts to contact the statistician involved in the Allen et al study to discuss some of these issues were unsuccessful.

Dr Allen is also concerned that we "most likely did not have the complete fluticasone propionate data available." This is true, as many important study details were not provided in his article. We were able to estimate the number of participants remaining in the placebo control and fluticasone propionate 200 µg/day groups using the numbers presented and the text. The correct numbers of patients completed in the control group (57) and the fluticasone propionate 200 µg/day group (79) were estimated correctly and used in our calculations. We were not able to control for "investigator interaction effects" as described. The technique of meta-analysis, if viewed as a trial whose subjects are the articles included, should balance out such parameters when the studies included are of large enough numbers. Clearly, because Dr Allen's study was the only one evaluating fluticasone propionate that passed our strict inclusion/exclusion criteria, this assumption is not necessarily a safe one.

Dr Allen states "the robustness of the conclusions of Dr Sharek and Bergman with respect to [fluticasone propionate] is undermined by the paucity of the data presented." We disagree. Regarding robustness, we clearly stated in the first paragraph of the discussion section: "Caution must be used when generalizing about fluticasone, however, because only one study was incorporated and the magnitude of effect was smaller that that of beclomethasone." In addition, we stated in the first paragraph of the conclusion section: "It would be inappropriate to judge the effect of moderate doses of inhaled fluticasone based on the 1 included study." Regarding the paucity of data, our search strategy revealed only one study that had the degree of scientific rigor we felt necessary to draw appropriate conclusions. Researchers who conduct meta-analysis believe one well-done study provides more valid conclusions than do many poorly done studies.

Dr Allen conveys disappointment in our disinterest in comparing 1 inhaled steroid to another to provide relevant information for potential for adverse growth effects. This, however, was beyond the scope of our meta-analysis. As described in the objective section of our abstract, we wished "to determine whether inhaled steroid therapy causes delayed linear growth in children with asthma." Given the persistent uncertainty in the literature and in practice as to whether inhaled steroids do decrease linear growth, we felt it important to first attempt to answer this more basic question comparing an inhaled steroid to a nonsteroidal control. One study cited by Dr Allen, by Price and colleagues,⁴ was reviewed for potential inclusion in our meta-analysis but did not meet the inclusion criteria (many of the patients randomized to the cromolyn arm were switched to the fluticasone propionate arm midway through the study). The other 2 studies cited^5,6 compared 2 different inhaled steroids without a nonsteroidal control group and therefore were removed from consideration based on our inclusion/exclusion criteria. Because we believe we conclusively showed moderate doses of beclomethasone and fluticasone do decrease linear growth velocity in the first year of use, it would therefore appear that direct comparisons are important to determine which inhaled steroid has the smallest growth effect. In our opinion, this question requires a different study design and wouldn't lend itself well to a meta-analysis design. Finally, we chose not to include the fluticasone propionate study in Figure 1 of the article (a weighted mean difference graph) because we suspected, a priori, that each inhaled steroid affects linear growth differently. We therefore subgrouped our results based on the inhaled steroid used and presented in Figure 1 ("inhaled beclomethasone in children with asthma") the data from the beclomethasone subgroup.

Dr Allen suggests that "it is important to keep in perspective the relative benefits and risks of inhaled corticosteroid therapy for asthma." We fully agree. In our conclusion section we state: "The negative effect on linear growth velocity needs to be weighed against the known positive effects of inhaled steroids on such outcomes as quality of life, symptom days, severity of exacerbations, decreased lung architectural changes, and health care utilization before clinical significance is clear." We did not suggest that inhaled steroids should be eliminated from the list of useful medications for childhood asthma but rather that caution should be exercised, doses should be minimized to the lowest effective dose, and height should be followed when they are used. We stated that the effects of inhaled steroids when given for more than 54 weeks, or the effects on adult height, remain unknown. Data published since our meta-analysis was published have shed additional light on the effect of inhaled steroids on adult height.⁷ Interestingly, this same data showed decreases in linear growth velocity in the first few years of the trial of inhaled steroids before catchup growth was displayed. This is consistent with our findings, and, if valid, this adult data is reassuring.

Dr Allen believes our publication focused on showing "the proper route to a meta-analysis was followed," which resulted in "an obscuring of the central message." We did discuss in necessary detail the methods of the meta-analysis, as we wished to provide strong evidence that the results obtained were valid. Regarding Dr Allen's concerns that our "results do not allow for any generalization nor do they provide the medical professional with any clear sense of differences among inhaled corticosteroids or differences among doses," we suggest that we have clearly shown beclomethasone in moderate doses decreases linear growth velocity for at least 54 weeks in children with asthma. We sacrificed generalizability to ensure internal validity. Before our meta-analysis was published, we believe this question was not answered conclusively.

Paul J. Sharek, MD, MPH
David A. Bergman, MD
Department of Pediatrics
Stanford University School of Medicine
Palo Alto, CA 94304

REFERENCES

1. Allen DB, Bronsky EA, LaForce CF, Growth in asthmatic children treated with fluticasone propionate. J Pediatr. 1998; 132:472-477
2. Sharek PJ, Bergman DA. Inhaled steroids and growth in children with asthma: a meta-analysis. Pediatrics. 2000;106(1). Available at: http://www.pediatrics.org/cgi/content/full/106/1/e8
3. Mulrow CD, Oxman AD, ed. Cochrane Collaboration Handbook [updated March 1, 1997]. In: The Cochrane Library [database on disk and CDROM]. The Cochrane Collaboration. Oxford: Update Software; 1996. Updated quarterly
4. Price JF, Russell G, Hindmarsh PC, Weller P, Heaf DP, Williams J Growth during one year treatment with fluticasone propionate or sodium cromoglycate in children with asthma. Pediatr Pulmonol. 1997; 24:178-186
5. De Benedictis FM, Medley HV, Williams L Long term study to compare safety and efficacy of fluticasone propionate with beclomethasone dipropionate in asthmatic children. Eur Respir J. 1998; 12:142S
6. Ferguson AC, Spier S, Manja A, Versteegh GA, Mark S, Zhang P Efficacy and safety of high-dose inhaled steroids in children with asthma: a comparison of fluticasone propionate with budesonide. J Pediatr. 1999; 134:422-427
7. Agertoft L, Pedersen S Effect of long-term treatment with inhaled budesonide on adult height in children with asthma. N Engl J Med 2000; 343:1064-1069