Evaluating the effects of asthma therapy on childhood growth: what can be learnt from the published literature?

Inhaled corticosteroids are widely recommended as first-line treatment for children with persistent asthma, and there is a wealth of evidence indicating their long-term efficacy in controlling this disease 1. While it is known that treatment with oral corticosteroids attenuates childhood growth 2, 3, the effect of inhaled corticosteroid therapy, which results in reduced systemic exposure, is less clear. Debate regarding their potential impairment of growth continues. In 1998, the food and drug administration (FDA) reviewed most of the studies that had attempted to address this issue, although no systematic analysis was performed. As a result, they recommended that all inhaled corticosteroids should carry a warning regarding potential effects on childhood growth 4.

Assessing the effects of inhaled corticosteroids on growth in children with asthma is fraught with difficulties and requires carefully designed and controlled clinical trials. Reduction in final adult height is the principal concern of patients and their parents, but difficulties in obtaining high-quality final height data have led to the adoption of surrogate markers. Growth velocity is the most commonly used surrogate, as it is a sensitive measure of impaired growth, although there is no correlation between growth velocity and final height 5. Most studies attempting to evaluate the effects of asthma therapy on growth in children with asthma are further complicated by the fact that asthma itself impairs growth 6–8. Comparisons of inhaled corticosteroids with placebo or nonsteroidal asthma therapy are medically and ethically justified only in patients with mild-to-moderate asthma, making it difficult to measure directly the absolute effect of high-dose inhaled corticosteroid therapy in children with severe disease.

In the first part of this review 9, recommendations were developed regarding study duration, age/sexual maturity of patients, exclusion criteria for height and growth velocity, permitted therapy during the study, protocol for height measurement, number of patients for adequate statistical power and methods for statistical analysis. In addition, a simple classification system for growth studies with inhaled corticosteroids was developed, separating medium-term randomized controlled trials (classified according to the comparator treatment: Type 1: placebo; Type 2: nonsteroidal asthma therapy; Type 3: inhaled corticosteroid) from “real-life” final height studies where the treatment in both study groups is more flexible (these studies may not be randomized or prospective). To ensure quality, only growth studies with a minimum duration of 1 yr and involving measurement of height with a stadiometer qualify for classification by this system. These criteria were recommended by the FDA to avoid the effect of seasonal variation and because stadiometry is widely acknowledged as being one of the most reliable means of measuring height 4.

The aim of this paper is to apply this classification system to all published studies assessing the effects of inhaled corticosteroids on growth in children with asthma. Subsequently, the design of these studies will be compared with the more detailed trial design recommendations, as a means of assessing the validity of the results and the effect of treatment on growth in those studies meeting the recommendations. As part of this assessment, patient numbers will be compared with the estimates of sample sizes required to establish superiority or noninferiority. However, it is important to note that the estimated patient numbers cannot be used to judge definitively whether a study was adequately powered. Three major caveats to be considered are: changes in variability of the data (the calculations were based on a specific estimate of variability from two good quality studies and increased variability would imply an increase in the required sample size); changes in required power (e.g. 80% instead of 90%); and changes in the minimum detectable difference (e.g. the selection of 0.7 cm·yr⁻¹ rather than 0.4 cm·yr⁻¹ as the equivalence limit in a noninferiority study). Overall, the assessment process should facilitate informed interpretation and appraisal of the literature and offer some explanation for apparently contradictory data.

Literature search

Relevant randomized controlled trials published in all languages were identified by systematically searching four databases for studies on growth/height with inhaled corticosteroids in children with asthma. The databases were Medline (1966–March 2001), Embase (1974–March 2001), Derwent Drug File (1983–March 2001) and Biosis Previews (1970–March 2001). Both free-text and indexing strategies were used to attain maximum recall of relevant references. Four keywords were used in the search: “asthma”, “inhaled”, “steroids” and “growth/height”. “Steroids” encompassed the following: glucocorticoids, corticosteroids, fluticasone, mometasone, budesonide, beclomet(h)asone and triamcinolone. “Growth/height” encompassed the following: growth, leg, height, knemometry, collagen, bone development. Articles were limited to human studies only, while editorial and review articles were excluded. Duplicates between the databases were removed using the DataStar “Drop Duplicate” feature, which identifies exact title matches.

Each title and abstract was reviewed by one reviewer and annotated as either potentially meeting the minimum entry criteria 9 for the classification system or clearly not meeting the criteria. Full papers were obtained for all publications potentially meeting the minimum criteria and classified as appropriate. Randomized prospective studies meeting the minimum entry criteria were classified as type 1, 2 or 3 if the comparator treatment was placebo, nonsteroidal asthma therapy or another inhaled corticosteroid, respectively. “Real life” studies, typically longer than 12 months in duration and where the doses in both study groups were flexible, were classified as type 4 studies.

A total of 362 publications were identified by the literature search, of which 344 were rejected as not meeting the minimum criteria (i.e. ≥12 months duration, used stadiometry, included a control group) for the classification system, or as abstracts from studies subsequently published as full papers. Three studies were classified as type 1, five as type 2, three as type 3 and nine as type 4 (two studies met the criteria for classification as two different study types).

Type 1 growth studies (control: placebo)

Three studies meeting the initial selection criteria of the study classification system compared growth in children with asthma treated with either an inhaled corticosteroid or placebo. Each study was randomized and used stadiometry to measure patient height (table 1⇓). Two were of 12 months duration 10, 11, whereas the third was performed over 27 months 12.

Only one study analysed patients who were prepubertal 11. The inclusion of pubertal patients in the studies of Jonasson et al. 12 and Simons et al. 10 is likely to have increased the variability of growth data and more importantly could have confounded the results if the distribution of pubertal patients between treatment groups had been unequal. Only one of the three studies excluded patients at the extremes of the normal ranges for height and growth velocity 11. Apparently, therefore, this was the only study to eliminate the potential confounding effect of abnormally high or low growth velocities.

No more than four short courses of oral corticosteroids per year were permissible in any of the studies, so oral therapy was unlikely to have substantially confounded the results 13. Nevertheless, at least 50% more courses were administered to patients receiving placebo compared with the inhaled corticosteroid group 10–12. Thus even in well-designed studies, it is questionable whether the absolute effect of inhaled corticosteroids on growth is measurable.

By definition, stadiometry was used to measure height in all of the studies (although in the study of Simons et al. 10 stadiometry was only used at “most” study sites). Height was measured at the recommended frequency (i.e. every 3 months) in two studies; in the third study, height was only measured every 6 months during the second year 12. From the information available in the publications, it is not clear whether height measurements in any of the type 1 studies were made in triplicate or at a consistent time of day, or whether different observers measured height.

In the first part of this review, it was suggested that type 1 studies should be designed to establish noninferiority 9, but this was not the case for any of the three studies listed. The results of the study that found no difference between corticosteroid treatment and placebo 11 should be interpreted with the focus on the confidence intervals (CI) for estimated treatment difference rather than p-values, but these are not available from the publication. Based on the sample size calculations presented in the first part of the review (>165 patients per group for 90% power) 9, this study may have been underpowered to conclude noninferiority within the range 0.3–0.5 cm·yr⁻¹. Nevertheless, the study does provide some useful information about the potential absolute effect of inhaled corticosteroids on growth. The remaining type 1 studies detected considerably slower growth with the inhaled corticosteroid (i.e. beclomethasone dipropionate and budesonide) than with placebo, by >1 cm·y⁻¹ 10, 12. In this context, the discussion of patient numbers for noninferiority becomes trivial, and these studies can be viewed as reasonably reliable.

Only two of the studies provide information on the statistical method used to analyse the growth data 10, 11. In each case, only a limited number of covariates were included in the analysis, reducing the opportunity to correctly allocate differences due to either the drug or other factors.

One of the type 1 studies indicated that growth was similar in children with mild asthma treated with either placebo or an inhaled corticosteroid 11. Conversely, the other two studies reported significant differences between study groups, patients receiving placebo growing faster than children treated with the inhaled corticosteroid 10, 12. The drug used in the study showing no significant difference from placebo was fluticasone propionate at doses of 100 µg and 200 µg·day⁻¹ 11, which is of interest because this inhaled corticosteroid has been shown to have a higher efficacy-to-risk ratio than other inhaled corticosteroids 14, 15. It has been reported in a meta-analysis that growth in the fluticasone propionate 200 µg·day⁻¹ group was slower than with placebo (mean difference of −0.43 cm·yr⁻¹ (95% CI: −0.01–−0.85)) 16, although the validity of such a post hoc analysis is questionable, since this group only contained data from one study, that of Allen et al. 11. In summary, none of these studies met all of the design recommendations detailed in the first part of this review 9, although the study by Allen et al. came closest to fulfilling the most important criteria.

Type 2 growth studies (control: nonsteroidal therapy)

Of the five studies meeting the criteria for classification as type 2 (i.e. inhaled corticosteroid versus nonsteroidal asthma therapy), one was also a type 1 study, while another was also classed as type 3. The design aspects and main outcomes of all published type 2 studies are summarized in table 2⇓ 10, 17–20.

Only one of the type 2 studies included only prepubertal children 18, thus maintaining a predictable level of variability in the data. The remaining four studies all included pubertal patients. Another potential source of variability in all these studies stems from the apparent lack of exclusion criteria for subjects with very high or low centiles for height or growth velocity.

Annual use of oral corticosteroids was confined to ≤4 courses in two of the five studies, minimizing the risk that oral corticosteroid treatment might confound their findings 10, 18. However, in one study 17 oral corticosteroids were permitted for up to 10 days in any 90-day period and in another 20 up to 12 courses were allowed during the 1-yr study period. In the remaining study, permitted oral corticosteroid use was not specified 19. In three of the studies, therefore, administration of more than four courses of oral corticosteroids in 1 yr could have confounded the results. In four of the five studies, more courses of oral corticosteroids were administered to patients receiving nonsteroidal asthma therapy than to those receiving inhaled corticosteroids, although the intergroup difference in this regard varied between studies. Although this imbalance would have favoured corticosteroid treatment, a significantly larger reduction in growth rate was still seen in corticosteroid-treated patients. As with the type 1 studies, therefore, some doubt is cast over the ability of these type 2 studies to estimate the absolute effects of inhaled corticosteroids. In the other study 18, five courses of oral corticosteroids were administered to both the nonsteroidal and the inhaled corticosteroid group.

Height was measured at least every 3 months in four of the five studies 10, 17, 18, 20. In the other study, however, height was not measured between months 6–12 in the inhaled corticosteroid group, or between months 4–12 in the cromone group, reducing the accuracy of annual growth velocity calculations performed by linear regression 19. Interobserver variability in height measurement was eliminated in three of the five studies by ensuring that the same observer measured patients' height at each visit 10, 18, 19. In one study, the data were strengthened further by specifying that readings were made in triplicate, and always at the same time of day 18.

Applying the principles discussed in the first part of this review, type 2 studies should be designed to demonstrate noninferiority. However, none of the five studies appeared to be designed in this manner and the same points made previously for type 1 studies are also relevant here. From assessment of the patient numbers, none of these studies were likely to be sufficiently powered to establish noninferiority using a noninferiority margin of 0.3–0.5 cm·yr⁻¹. One must therefore exercise a degree of caution when interpreting the two studies reporting no difference between an inhaled corticosteroid and a nonsteroidal comparator 9, 19. Therefore it is recommended that the data is assessed, principally by examining the confidence intervals of the estimate of treatment difference, rather than focusing on the p-values. The studies provide valuable information, but cannot be viewed as conclusive in proving noninferiority to the nonsteroid comparator on the basis of the margin mentioned earlier.

Four of the five studies 17–20 specified a per-protocol population for analysis, although none clarified whether pubertal patients were excluded from this population. All of the studies provide some information on the statistical method used to analyse the growth data. Only one study included age and sex as covariates in the analysis, which are arguably the most important factors in predicting growth 18. Nevertheless, as for the type 1 studies, in each of these five studies only a limited number of covariates from the recommended list defined in first part of this review were included in the analysis, thereby reducing the opportunity to correctly allocate differences due to drug and those due to other factors.

The outcomes of three of the five type 2 studies indicated that children treated with inhaled corticosteroids grew a little more slowly than those treated with nonsteroidal therapy (treatment difference 1.3–1.4 cm·yr⁻¹). In the other two studies, no significant differences in growth velocity were reported between children treated with fluticasone propionate and those treated with sodium cromoglycate 18, 19, although growth was slower in children treated with budesonide than in those treated with cromones 19. Reduced growth velocity was reported in all studies of beclomethasone dipropionate 10, 17, 20. However, none of the type 2 studies met all of the principal design criteria and further well-designed studies to compare inhaled corticosteroids with nonsteroidal therapy would be useful.

Type 3 growth studies (control: comparator inhaled corticosteroid)

Table 3⇓ presents the main design aspects and outcomes of the three growth studies comparing different inhaled corticosteroids that were classified as type 3 19, 21, 22.

Two of the type 3 studies included only prepubertal children 21, 22, and thereby avoided possible confounding by the unpredictability of pubertal growth. The inclusion of patients in the pubertal age range in the other study increases the difficulty of establishing a genuine treatment effect 19. No exclusion criteria according to height or growth velocity were specified for two of the three studies 19, 22, introducing a potential source of variability in the results. In the remaining study 21, children with a disorder likely to affect growth were excluded.

In one study, patients who received treatment with systemic corticosteroids were excluded from the growth analysis, removing any potential for additional steroid use to confound the results 21. In the other two studies, use of oral corticosteroids was not specified. In one, no permissible therapy other than study medication was mentioned 19, and in the other, sodium cromoglycate was permitted 22.

All the recommendations made to optimize the accuracy of growth measurement were implemented in one study: measurements were made every 3 months, using the same observer throughout to eliminate interobserver variability, and all measurements were made at a similar time of day to avoid inaccuracies due to circadian variability in height 21. For the remaining two studies, height was measured less often than recommended during the second half of the treatment period 19, 22. Nevertheless, the number of observers was limited to one or two in both studies, and in one of them 22 all measurements were taken at the same time of day.

As discussed in the first part of the review, type 3 studies may be designed to establish either noninferiority or superiority 9. The aim of each of the three studies listed was to evaluate whether one inhaled corticosteroid, fluticasone propionate, was superior to another inhaled corticosteroid. The sample size tables provided in the first part of the review 9 indicate that even to detect a difference as large as 1 cm·yr⁻¹ between treatment groups, with 90% power, approximately 42 patients per treatment group would be required. It is likely that only one of the three type 3 studies listed was adequately powered to establish superiority with a minimum detectable between-group difference of this order 21. The remaining two studies were probably not adequately powered to detect between-group differences in growth velocity of ≤1 cm·yr⁻¹, since none of the treatment groups contained more than 15 patients 19, 22.

All three studies specified a per-protocol population for analysis although only one study 21 stated that pubertal patients were excluded from this population. Only one study included covariates in the analysis for key factors such as age, sex, race, country and baseline height 21. Thus, for the other two studies, it is impossible to correctly allocate differences due to drug and those due to other factors, such as demographic, baseline, study and environmental characteristics. A notable weakness of the study by Rao et al. 22 is the lack of information upon which statistical analysis was based. Fluticasone propionate was compared with either budesonide or beclomethasone dipropionate in all three type 3 studies, and both these comparators had a significantly greater detrimental effect on growth velocity in children. Although two of these studies 19, 22 were apparently underpowered to detect a difference in growth of 1 cm·yr⁻¹ 19, 22, judged according to the recommendations on minimum patient numbers in Part I of this review 9, it is likely that the variability (sd) of the growth measurements in these studies was much smaller than the value on which the current author's recommendations were based. Only one study of the three 21 appears to have been well designed, with no major weaknesses.

Type 4 growth studies (long-term, “real-life”)

Growth velocity studies

As well as final height, four type 4 (“real life”) studies have compared the effects of treatment strategies on growth velocity (table 4⇑) 27–30. Note that one of these publications reported pooled results from three distinct clinical trials 30.

Although none of the four growth velocity type 4 studies specifically excluded children entering puberty, the age range for one of the studies implied that all patients would have been prepubertal 30. For the other three studies 27–29, the possibility of the older patients entering puberty during the study increases the likely variability of the data, and the potential for confounding interpretation. This is most relevant for studies by the Childhood Asthma Management Programme Reasearch Group 29 and by Agertoft and Pederson 27: although the mean ages of subjects at baseline were 9 and 6.2 yrs, they were treated for mean durations of 4.3 and 5.3 yrs, implying that many patients entered puberty during the studies.

Oral corticosteroid use was permissible in three of the four growth velocity studies 27–29, although in one this was limited to 2 weeks·yr⁻¹ and to the budesonide group only 27. In the Skoner et al. 30 publication, oral corticosteroid use was generally balanced between patient groups, with 53–63% of patients in the conventional asthma treatment groups receiving this therapy compared with 46–56% of those in the budesonide group. This publication presented pooled data from three studies, and the control group (“conventional asthma therapy”) was allowed treatment with inhaled corticosteroid in two of the studies. Since the extent of this treatment was unspecified, it could have confounded growth outcomes. In the Child Asthma Management Program Research Group study, use of oral corticosteroids was significantly lower in the budesonide group (0.70 courses per patient per year) and nedocromil group (1.02 courses per patient per year) compared with the placebo group (1.22 courses per patient per year) 29. In this study, it is notable that additional beclomethasone dipropionate therapy was administered to children receiving placebo on 18.7% of days, compared with 17.1% for nedocromil and 6.6% for budesonide. Thus, the use of additional corticosteroid therapy probably affected the results of the Childhood Asthma Management Program Research Group 29 and Skoner et al. 30 studies; the comparison actually achieved was one of treatment strategies rather than fixed-dose therapy with specific drugs. In the remaining study, there was no information on the use of oral corticosteroids 28.

For two of the type 4 growth velocity studies, the interval between each height measurement was more than every 3 months 27, 29. However, both studies were long-term, and the number of measurements for each patient was adequate to prevent introduction of substantial inaccuracy. Height was actually measured more often than recommended in one of the studies 30, but the frequency was not stated in the remaining study 28. Unfortunately, only one study 28 described evidence of measures to optimize the quality of height measurements (the same two observers measured height throughout).

As for the type 4 final height studies, it was suggested in the first part of this review that these studies should be designed to establish noninferiority of the inhaled corticosteroid treatment strategy to nonsteroidal comparator treatment strategy 9. This was not the case for any of the four studies listed.

Only one of the four studies provided information on the populations analysed 30; the other three studies provided no detail in this regard. One of the studies provided no information on the statistical method used to analyse the data 28 but the remainder all analysed the data using a model-based approach, which allows for the adjustment of important covariates. One study specified the covariates adjusted for, and the list appears to cover many of the important factors 29.

All four growth velocity studies classed as type 4 compared a specific inhaled corticosteroid with nonsteroidal asthma therapy, although inhaled corticosteroids were administered to the control groups in two of the studies 29, 30. In one study, there were two control groups, treated with nedocromil and placebo, respectively 29. The findings of the four studies were somewhat contradictory, in that the study by Agertoft and Pederson 27 detected no significant difference between budesonide and nonsteroidal therapy, but growth velocity was reduced with budesonide compared with placebo and nedocromil in the Child Asthma Management Program Research Group study. In the third publication comparing budesonide with a control (“conventional asthma therapy”), there was no significant between-group difference for the pooled data from three studies, but there was a difference in one of the individual studies, where inhaled corticosteroid therapy was not used in the control group. In the remaining study, triamcinolone acetonide was found to attenuate growth in comparison with nonsteroidal therapy 28. The different outcomes could have arisen from design heterogeneity, inadequate statistical power, inadequate allowance in the statistical analysis for factors affecting growth, or differences in the systemic activity of the inhaled corticosteroids 15, 32. Of the four studies of this type, the Child Asthma Management Program Research Group Study appeared to have the fewest design weaknesses, although an assessment of the impact of puberty on the results of this study would be helpful.

Discussion

With the exception of one type 3 study 21, none of the growth studies identified in the comprehensive literature search performed in the present study fulfilled all of the trial design recommendations defined in the first part of this review. The most common shortcoming, for all but the type 3 studies, was the fact that they were not designed to assess noninferiority of the inhaled corticosteroid therapy compared with either placebo or nonsteroidal asthma therapy, and without more complete information on the confidence interval of treatment difference, it is difficult to infer noninferiority conclusions. Many studies also failed to fulfil several other recommendations. Overall, the level of quality was similar across the different growth study types, with one or two studies in each classification approaching fulfilment of the key design criteria.

Despite the flaws identified in the full list of studies, there were few direct contradictions in the outcomes of these trials. In summary, the evidence indicates impaired growth velocity with budesonide and beclomethasone dipropionate in comparison with placebo, nonsteroidal therapy and fluticasone propionate during 1–2 yrs of treatment, but no impact on final height. Children treated with low-dose fluticasone propionate for 1 yr grow similarly to those receiving placebo or nonsteroidal therapy, but there are no data on this drug's effect on medium-term growth, particularly on the timing of puberty and final height. One study with triamcinolone acetonide treatment for 1 yr indicated that this drug impaired growth in comparison with nonsteroidal therapy.

There is evidence of a dose-response relationship for growth impairment with fluticasone propionate 11, but none of the studies which met the criteria for the classification system compared different doses of budesonide or beclomethasone dipropionate. In type 1 and type 2 studies, the most commonly investigated dose of both budesonide and beclomethasone dipropionate was 400 µg·day⁻¹. There was a consistent reduction in growth with this dose of either drug compared with placebo or nonsteroidal therapy, while similar reductions in growth were reported with lower doses of budesonide.

These are the results from those studies deemed to meet the criteria for the classification system outlined in the first part of this review. Some of the other publications identified by this report's literature search were also of interest. Evidence of potential growth impairment with inhaled beclomethasone dipropionate is provided by a randomized, double-blind study of 177 children 33. Those treated with beclomethasone dipropionate 800 µg·day⁻¹ for 1 yr grew significantly more slowly (3.6 cm increase; 95% CI: 3.0–4.2) than children treated with half this dose (5.1 cm increase; 95% CI: 4.5–5.7); height standard deviation score (SDS) decreased by 0.27 cm (95% CI: −0.34–−0.19) and 0.16 cm (95% CI: −0.24–−0.07), respectively during the study. This study was excluded from the classification system because there was no control group receiving only placebo, nonsteroidal therapy or a different inhaled corticosteroid. Another randomized double-blind study compared beclomethasone dipropionate 400 µg·day⁻¹ with placebo for 7 months in 94 children, and those in the beclomethasone dipropionate group grew significantly less than those in the placebo group (2.66 versus 3.66 cm, p<0.0001) 34. This study was excluded because it was <12 months in duration and growth was measured using a Raven Minimeter rather than by stadiometry. A 12-month study in 56 children with asthma compared the effects on growth of budesonide, beclomethasone dipropionate, nonsteroidal therapy, and inhaled plus oral corticosteroids (all doses administered according to clinical requirements) 35. Growth velocity SDS was significantly reduced among children receiving beclomethasone dipropionate (−1.04) and oral corticosteroids (−1.58) compared with nonsteroidal therapy (0.03) and budesonide (−0.20; p<0.001). A 2-yr study indicated that growth among children treated with budesonide 600 µg·day⁻¹ plus salbutamol was no different from that in children treated with salbutamol alone 36. Neither of the previous two studies specified stadiometry as the method for measuring height and hence they were excluded from the list of studies.

A large study of >3,000 children with asthma indicated that children receiving high dosages of inhaled corticosteroids (≥400 µg·day⁻¹), and who were attending hospital for asthma care, had reduced height SDS, and those receiving high-dose inhaled corticosteroids also had low growth velocity (SDS of −0.19) 37. This effect was, however, smaller than that of social deprivation. The “normal” population of children (i.e. those not currently treated with inhaled corticosteroids) with asthma demonstrated no evidence of impaired growth. Unfortunately this 4-yr study did not specify the use of stadiometry to measure height, and no results for specific inhaled corticosteroids were presented. In a primary care-based study, children receiving inhaled budesonide or beclomethasone dipropionate for a year were found to have low growth velocity (SDS scores of −1.0 and −1.7 respectively) 38. However, switching the 20 children with slowest growth to fluticasone propionate treatment for 1 yr resulted in an increase in growth velocity SDS to +1.6. The method for measuring height was again not described for this study.

A final height study was rejected because height was not measured by stadiometry 39. Nevertheless, the results were in agreement with the suggestion in the other beclomethasone dipropionate final height study that the drug may delay pubertal growth 24. Also, in both studies, there were no significant differences in final height between children treated with beclomethasone dipropionate and those treated with nonsteroidal therapy.

A slightly different but highly relevant study confirmed the growth-suppressive effects of oral corticosteroids in 163 patients with severe asthma 40. Height standard deviations of patients receiving intermittent, alternate-day or daily oral corticosteroids were −0.44, −1.22 and −1.95, respectively (the mean dosage of inhaled corticosteroids was 1,675 µg·day⁻¹). Notably, this study indicated that the introduction of inhaled corticosteroids attenuates asthma-associated growth suppression by allowing reduced use of oral corticosteroids and by improving the control of asthma.

Thus, growth studies that did not meet the criteria for inclusion in the classification system did not contradict the findings of those that did and the evidence overall suggested that budesonide and beclomethasone dipropionate may have a similar effect on growth velocity at dosages of ≥400 µg·day⁻¹.

Two meta-analyses of growth studies have been published. The first of these, published in 1994, only included data on beclomethasone dipropionate (200–875 µg·day⁻¹) 3. In light of the results of the present review, it is perhaps surprising that no significant association was found between beclomethasone dipropionate (100–400 µg·day⁻¹) treatment and impaired growth 3. However, closer inspection of the results indicates that there was evidence of a short-term reduction in growth velocity, although final height, the principal outcome, remained unaffected. This is compatible with the present findings. Also, this meta-analysis 3 failed to show any association between dose and growth-suppressive effects of beclomethasone dipropionate, again similar to the present authors' findings. One study, however, has demonstrated a dose-related effect of beclomethasone dipropionate; children treated for 1 yr at a dosage of 800 µg·day⁻¹ grew more slowly than those receiving 400 µg·day⁻¹ 33. The more recently published meta-analysis exercised several quality criteria for inclusion: randomized controlled trials only; inhaled corticosteroid compared with nonsteroidal therapy; children aged <18 yrs not using oral corticosteroids at the outset; clinical diagnosis of asthma; and treatment for a minimum of 3 months 16. Five studies met these criteria: four with beclomethasone dipropionate (328–400 µg·day⁻¹) and one with fluticasone propionate (200 µg·day⁻¹). Decreases in growth velocity of 1.51 cm·yr⁻¹ were reported for beclomethasone dipropionate (95% CI: 1.15–1.87), and 0.43 cm·yr⁻¹ for fluticasone propionate (95% CI: 0.01–0.85). The latter is based on data from the 200 µg·day⁻¹ group in the Allen et al. 11 study, but in the original publication the overall treatment effect was calculated and no significant difference was reported (p≥0.05). Overall, the findings from the two published meta-analyses are consistent with the present authors' results.

Conclusions

This review has examined studies assessing the effects of inhaled corticosteroids and other asthma therapies on childhood growth, taking into account key trial design recommendations derived from important physiological, statistical and trial design principles defined in the first part of this review 9. Many factors potentially confound studies attempting to establish any treatment effects on growth, and almost all of the published trials are susceptible to one or more of these factors. Nevertheless, the outcomes of studies meeting the minimum selection criteria for design quality were largely consistent. Thus, despite the need for caution when interpreting these results, it is probable that the broad findings are correct. All published studies comparing beclomethasone dipropionate or budesonide with placebo or nonsteroidal therapy (i.e. study types 1 and 2) indicate that treatment with either drug at recommended dosages appears to carry a risk of a relatively small degree of growth impairment over 1–2 yrs. In contrast, the original publications from study types 1 and 2 with fluticasone propionate ≤200 µg·day⁻¹ concluded that this drug does not impair growth, although a recent limited meta-analysis suggested that growth was slightly slower with fluticasone propionate 200 µg·day⁻¹ than with placebo. Studies comparing different inhaled corticosteroids also indicate greater growth velocity with fluticasone propionate than with beclomethasone dipropionate or budesonide over 1–2 yrs. None of the inhaled corticosteroids appear to have a significant effect on final height.

Important aims for the future include further growth studies designed to avoid all of the main confounding factors, and long-term prospective ‘real life’ studies to obtain more precise and conclusive data comparing different inhaled corticosteroids. Both short- and long-term studies need to incorporate the recommendations described in the first part of this review in order to obtain more accurate and consistent estimates of the growth effects due to treatment with inhaled corticosteroids. It is also clear that different steroids should be distinguished as they do not all affect growth in the same way, additional well-designed type 3 studies will aid this assessment.