Corticosteroid responsiveness and clinical characteristics in childhood difficult asthma

Difficult asthma (DA) in children can be challenging to treat, and leads to high rates of healthcare utilisation. These patients have continued symptoms despite high-dose inhaled corticosteroids (ICS) and other therapies 1. The reasons for their asthma being difficult to treat require detailed investigation 1. Confirmation of asthma diagnosis, identification and treatment of associated diagnoses is essential. Potential reversible factors, such as poor treatment adherence and continued allergen exposure, should be addressed.

Phenotyping childhood DA according to airway inflammation and corticosteroid responsiveness is challenging 2. However, identification of corticosteroid resistance is important, allowing the clinician to deploy alternative therapies; conversely, in the corticosteroid sensitive patient, the dose of therapy should be minimised to avoid unwanted side-effects.

A protocol driven approach for the investigation of childhood DA helps identify the reasons why asthma is difficult to treat and classifies the underlying inflammatory profile. As in other studies 2–4, we used a protocol involving the administration of a 2-week course of high-dose systemic corticosteroids, with investigation of noninvasive markers of inflammation before and after corticosteroids, followed by bronchoscopy at the end of the corticosteroid trial (fig. 1⇓). Patients were characterised according to their corticosteroid response and appropriate treatment alterations were made.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1—

Difficult asthma investigation protocol. Ig: immunoglobulin; RAST: radioallergosorbent assay; SPT: skin prick testing; BDR: bronchodilator reversibility; F_eNO: exhaled nitric oxide fraction; FOB: fibreoptic bronchoscopy; EB: endobronchial biopsy; BAL: bronchoalveolar lavage; CT: computed tomography. ^#: prednisolone n = 65, triamcinolone n = 16.

Adult DA patients are predominantly female 5, with increased atopy in mild rather than severe asthma 6, and infection rather than allergy is implicated in the pathophysiology 7. Many cases present in early adulthood, not childhood, which may be associated with a poor prognosis 8. Severe adult asthmatics appear to have a different disease compared with mild asthmatics with greater airway obstruction, a lower diffusing ability and more air trapping. Sputum neutrophilia is more common 6. We hypothesised that characteristics of childhood DA are different to those reported in adults and that complete corticosteroid responsiveness is uncommon in paediatric DA 5, 6, 9, 10. We report a consecutive observational study of children, aged >5 yrs, with DA who were referred to the Royal Brompton Hospital, London, UK (mainly tertiary and even quaternary referrals). We aimed to describe their clinical characteristics and their response to a corticosteroid trial in terms of symptoms, spirometry and inflammation, and predictors of corticosteroid response.

METHODS

RESULTS

Cross-sectional study

Tables 1⇑ and 2⇑ summarise the baseline demographics. For patients prescribed oral corticosteroids, the median daily dose was 10 mg (5 mg on alternate days ranging to 40 mg·day⁻¹). The reasons for the variations in denominators are explained in figure 1⇑. Of the 102 children studied, 60 (59%) were male. 88 (86%) children were atopic, of which 32 (36%) owned a pet and of these, 21 (65%) were sensitised to their pet. Of the 102 children studied, 99 had data on history of food allergies/triggers available, of which 24 (24%) had a history of food allergy or triggers to asthma. Of these 24 children, 11 had positive RAST to the food trigger. The mean±sd FEV₁ was 67.1±19.2%. A positive BDR test was observed in 53 (75%) out of 71 atopic children compared to three (30%) out of 10 nonatopic children (p = 0.004). F_eNO was normal in 10 (91%) out of 11 nonatopic patients compared with 36 (51%) out of 70 atopic patients (p = 0.02). The mean±sd F_eNO z score was 1.78±2.1. The distribution of IgE levels is illustrated in figure 2⇓. Of the 102 children studied, 74 underwent serum immune testing and immunological abnormalities were found in 10 (14%) of the patients (table a in the supplementary data). Nonatopic children were significantly more likely to have an abnormality compared to atopic children (p = 0.036). 47 (46%) patients had an HRCT performed (table b in the supplementary data) and of these three (6%) had bronchiectasis. A pH study was completed in 55 (54%) out of 102 children, of which 41 (75%) were abnormal.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2—

Distribution of immunoglobulin (Ig)E in children with difficult asthma and eligibility for anti-IgE therapy.

Corticosteroid response study

89 (87%) subjects underwent a corticosteroid trial as part of the clinical protocol. Data to assess corticosteroid response was available in 81 subjects. There was no difference between the responses of those who were and were not prescribed systemic corticosteroids (refer to supplementary data). Nine (11%) out of the 81 subjects were complete responders, 65 (80%) were partial responders and seven (9%) were nonresponders (fig. 3⇓). Of the 57 (83%) children with evidence of adequate adherence, nine (16%) were complete responders, 47 (82%) were partial responders and one (2%) was a nonresponder. All children that demonstrated complete corticosteroid response had evidence of treatment adherence compared to 48 (84%) out of 57 children who were partial or nonresponders (p = 0.14). None of the steroid nonresponders (none out of seven) were receiving maintenance systemic steroids compared with 30 (41%) out of the 74 subjects with partial or complete responders (p = 0.03). In a post hoc analysis, there was evidence that the 16 subjects given triamcinolone may have had slightly better responses compared with the 65 subjects given prednisolone (refer to supplementary data). Whether this reflected sub-optimal compliance with prednisolone, a dose effect, or intrinsically better anti-inflammatory properties of triamcinolone cannot be determined. In a further post hoc analysis we compared steroid responsiveness in seven subjects with single and 65 subjects with multiple allergies. Although the numbers are small, children with a single allergy had a poorer BDR response after the steroid trial (refer to supplementary data). The significant predictors of response in individual parameters are shown in table 3⇓.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3—

The number of parameters with corticosteroid response for each patient divided according to sex. The tested parameters included symptom improvement, absent bronchodilator reversibility, normal forced expiratory volume in 1 s and normal exhaled nitric oxide fraction at the end of corticosteroid trial. □: female; ▓: male.

View this table:

• View inline
• View popup

Table 3—

Significant predictors of corticosteroid responsiveness from multivariate logistic regression

FEV₁ response

The FEV₁ response of the subjects before and after the 2 week corticosteroid trial is shown in figure 4⇓ (table g in the supplementary data).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4—

Forced expiratory volume in 1 s (FEV₁) pre and post a 2-week trial of systemic corticosteroids. % pred: % predicted. p<0.001.

Baseline mean FEV₁ was 66.3±19.1%, increasing to 77.5±18% after the corticosteroid trial (p<0.001). Spirometry data was available for 75 patients who underwent the trial, of which 35 (47%) reported normal pre-bronchodilator FEV₁ after the corticosteroid trial. Higher baseline FEV₁ (OR 1.08, 95% CI 1.02–1.12) and BAL eosinophilia (OR 5.43, 95% CI 1.13–26.07) were associated with FEV₁ response. Change in FEV₁ was associated with baseline low BDR (OR 0.42, 95% CI 0.05–0.80). PAL was present in 17 (23%) out of the 75 children who underwent the trial and who had spirometry data available. PAL was associated with a low FEV₁ before the corticosteroid trial (OR 0.93, 95% CI 0.90–0.97).

F_eNO response

The F_eNO response of the subjects before and after the 2 week corticosteroid trial is shown in figure 5⇓ (table i in the supplementary data).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5—

Exhaled nitric oxide fraction (F_eNO) pre and post a 2-week systemic cortiosteroid trial. p<0.01.

The mean z scores for F_eNO before and after the corticosteroid trial were 1.78±2.07 and 0.40±1.98, respectively (p<0.01); low baseline F_eNO (OR 0.36, 95% CI 0.18–0.72) was a predictor of F_eNO response.

DISCUSSION

In this large observational study, which is the most detailed characterisation of DA in children in the literature to date, and which includes analyses of clinical corticosteroid responsiveness, we have shown that: 1) in contrast to a study of adults with DA, in which females predominate and atopy is less prominent in severe asthma compared to mild-to-moderate asthma 6, in children with DA, males predominate and atopy is very common; 2) many children are exposed to environmental influences (cigarette smoke) known to cause corticosteroid resistance; 3) associated diagnoses are common especially in nonatopic subjects; 4) response to a corticosteroid trial is difficult to predict, with the majority of subjects exhibiting only a partial response; and 5) BDR is common even after a corticosteroid trial even though most subjects were taking LABA. The latter may reflect noncompliance, tachyphylaxis or a genuine bronchodilator-resistant state.

The study does have a number of limitations. As this was a large observational study, some data is missing, and not all patients underwent the corticosteroid trial because, in some cases, this was not thought to be clinically appropriate. The data is based on a highly selected cohort and may not be applicable to all populations. As with many studies, assessment of adherence to the corticosteroid trial was suboptimal and the invasive assessments of inflammation were performed after the corticosteroid trial. This was done to optimise the safety of the procedure. Later on in the trial, when its safety was established, intramuscular triamcinolone was introduced ensuring adherence to therapy. Due to the severity of asthma in the children evaluated, over a third of the children were on maintenance systemic steroids at baseline, this may have influenced results and we found significantly more children who exhibited a partial or complete response were on maintenance steroids compared with nonresponders. This is an unavoidable disadvantage of an observational clinical study. Finally, a home visit was not performed during the study period, and may have been beneficial to explore the role of psychosocial issues, allergen exposure and compliance in symptom control 4. An adapted version of the protocol is now used which addresses these issues and includes a home visit before proceeding to the steroid trial.

We have shown that atopic patients are more likely to exhibit BDR and have higher F_eNO levels than nonatopic patients, although the number of nonatopic patients was small. Atopy itself causes an elevation in F_eNO 22. Nonatopic patients were also more likely to have immune abnormalities. In keeping with a recent study 23, there was little difference in BAL and EB inflammation in atopic compared with nonatopic children. A large proportion of all patients (atopic and nonatopic) had gastro-oesophageal reflux. We were unable to formally assess the contribution of reflux to the severity of asthma, but in three patients treatment for severe gastro-oesophageal reflux was associated with a significant reduction in inhaled steroid dose with improvement in their symptoms. Importantly, there was a high incidence of alternative or additional diagnoses in the whole cohort (51%), and this was more common in the nonatopic children. A study of adult DA patients found an incidence of alternative or additional diagnoses of 32% 24. One lesson from our study is that at least potential comorbidities are common and ideally should be sought before evaluating steroid sensitivity; our current practice has been modified in light of this.

A complete response to the corticosteroid trial was rare in our study. Our criteria for complete corticosteroid responsiveness were stringent, and required the four measurements (symptoms, FEV₁, BDR and F_eNO) to become normal. In other studies, steroid sensitivity was defined by >15% improvement in pre-bronchodilator FEV₁, after a short course of oral corticosteroids 4, 25. The definition of corticosteroid responsiveness has been debated and it may be that our definition is too rigorous. Although symptoms may be of most importance to patients, symptom perception is often poor in asthmatic children 26, making this less than ideal as a measure of response. Spirometry is often normal in children even with quite severe asthma, and has been criticised as an end-point 27. Conversely, inflammation may predict future outcome, thus F_eNO or sputum eosinophils may be a more objective measure of steroid response 28. We opted for a spectrum of end-points, acknowledging that this is, to some extent arbitrary, and that a true definition of corticosteroid responsiveness in children has not yet been achieved. As this study was based on our clinical protocol and clinicians are more likely to use a combination of end-points, this determined our choice. Furthermore, corticosteroid responsiveness is likely to be a spectrum, rather than an all or nothing phenomenon, and it may be that the use of a higher dose or increased duration of prednisolone therapy might have increased the number of responders. Indeed, we have shown that spirometry may increase still further in the year after a corticosteroid trial 2. However, prolonged high-dose therapy is likely to cause side-effects, so we have kept to traditional doses in order to assess steroid responsiveness. In this study, we found heterogeneity in response in symptoms and inflammation. We found few robust determinants of corticosteroid response and due to regression about the mean, baseline investigations as prediction of outcome measures should be interpreted cautiously.

Partial corticosteroid response was common, and the incomplete response may be due to ongoing exposure to environmental triggers. In keeping with previous studies 4, we have demonstrated that environmental factors are likely to play an important role in DA. We have shown that 65% of atopic children who had pets were allergic to them, and other studies have shown that persistent allergen exposure in those sensitised can lead to an interleukin (IL)-2 and IL-4 mediated corticosteroid resistance 29. Thus, it is likely that the ongoing allergen exposure contributed to their poor symptom control, despite high-dose treatment. We have also shown that those exposed to tobacco smoke at home were less likely to have a symptom response following a corticosteroid trial. In adult DA, it has been shown that active smoking can have detrimental effects on symptoms, and corticosteroid responsiveness 30 and our study suggests the same may be true with passive smoke exposure. Oxidative stress caused by smoke exposure has been shown to alter histone deacetylase activity which has detrimental effects on steroid response 31. The incomplete responsiveness to LABA and systemic corticosteroids suggests the need for other therapies in paediatric DA. It is noteworthy that of the 90 children in whom IgE testing was performed, only 40 (44%) would have been eligible for anti-IgE therapy under the current license (fig. 2⇑) 32. This leaves a large number of children with partially steroid insensitive DA with few, if any, evidence based options for add on therapy. New medications for this group are urgently needed.

A cross sectional study of 125 children and 150 adults with DA, in keeping with our findings, showed that children with DA had a high prevalence of atopy and a male preponderance compared with adults 9. Corticosteroid response was only determined in vitro, in contrast with our report. They found that children displayed greater in vitro responsiveness to glucocorticoids than adults. Ranganathan et al. 4 studied a smaller cohort of children with DA and showed 60% were corticosteroid resistant although their criteria for response was defined by a >15% increase in FEV₁, thus less stringent than ours. In contrast, in a larger cohort of adolescents with DA 25, the same spirometric criteria was used for response, but found only 25% of subjects to be steroid insensitive. Steroid insensitivity was also shown to be associated with the need for oral corticosteroids at a younger age and African-American ethnicity. We had insufficient numbers of non-white children in our cohort to determine any effect of ethnicity. However, lung function did not predict lack of response to corticosteroids. A small paediatric study of difficult asthma has shown that a single dose of intramuscular triamcinolone reduced exacerbations and reduced F_eNO 33, 34, which is in keeping with our findings.

In conclusion, in this large observational study, children with DA differed from reports in adults with respect to their atopic status and sex distribution. Very few children with DA were complete corticosteroid responders, and partial response was more common. There were no predictors of nonresponse, partial response or complete response to corticosteroids, although the statistical power was limited due to the low numbers in the complete and nonresponders groups. The only predictors of response in individual parameters that could be found was that of symptomatic response being less likely if there was a history of smoke exposure in the home, BAL eosinophilia and high FEV₁ prior to steroids predicting lung function response. The reasons for the rarity of complete response are unknown, but it may partly be explained by continued adverse environmental exposures, including smoke, pets and house dust mite in the environment. There are very few and generally weak predictors of corticosteroid response. However, the criteria for steroid response that we used were perhaps too stringent. Certainly, criteria need to be agreed in order to categorise and target therapy in this refractory group of children. As a result of this study, we recommend a more detailed assessment of children with difficult asthma in terms of adherence, environment and psychosocial issues prior to the performance of more invasive investigations such as bronchoscopy, and that the steroid trial should be with triamcinolone, not prednisolone to obviate the issues of adherence. Thorough investigation of children with DA is important to allow exclusion of other diagnoses and comorbidities which when treated may obviate the need to escalate doses of corticosteroids. Confirmation of the diagnosis of asthma is especially important in nonatopic children. Prospective trials are required in carefully phenotyped children to develop an evidence base for DA treatment.

Statement of interest

A statement of interest for D.N.R. Payne can be found at www.erj.ersjournals.com/misc/statements.dtl