Design and duration

The five included studies randomised 14,252 children to receive either an intervention or usual care. All studies were conducted in the Northern Hemisphere. The largest study randomised 12,179 children in a cluster‐randomised trial of a primary care‐based public health intervention (Julious 2016). Data were collected from August 2013 until the end of September 2014 from the Clinical Practice Research Datalink (CPRD), a computerised database of anonymised longitudinal medical records for primary care. The remaining studies were randomised controlled clinical trials employing a pharmacological intervention; three were blinded and one was an open study. Johnston 2007, Weiss 2010, and Morita 2017 studied interventions based upon LTRA administration from school return, whilst Teach 2015a compared administration of the anti‐IgE monoclonal omalizumab to a doubling of ICS dose or placebo initiated four to six weeks before school return. Three pharmacological studies recruited across multiple sites (Morita 2017; Teach 2015a; Weiss 2010), whilst in Johnston 2007 participation was at a single site. In Johnston 2007, 194 children were followed up for 45 days with no run‐in period. In Weiss 2010, 1162 children were followed up for 10 weeks after a 2‐ to 12‐week screening period. The 513 children in Teach 2015a received guideline‐based treatment to gain asthma control during a 4‐ to 9‐month run‐in period and were followed up until 90 days after school return. In Morita 2017, 204 children were enrolled and randomised two to six weeks before entering a 60‐day study period.

Inclusion and exclusion criteria

All studies referenced age and asthma diagnosis in their inclusion criteria. Julious 2016 included children aged 4 to 16 years with a coded diagnosis of asthma within their primary care record and a prescription for asthma medication within the last year. Four‐year‐old children were analysed separately since, whilst they are of school age in the United Kingdom, a diagnosis of asthma in this age group was judged to be more controversial than in older children. Children with neoplastic disease and those judged unsuitable for the intervention by their general practitioner were excluded. Johnston 2007 included children aged 2 to 14 years with doctor‐diagnosed asthma. Additional inclusion criteria were use of a reliever inhaler in the last year, school absence due to asthma in the last year or significant activity limitation, history of asthma exacerbations associated with respiratory viral infections, and ability to communicate in English. Children with significant cardiorespiratory comorbidity were excluded, as were those with an asthma exacerbation during the month before study inception and those using regular OCS or an LTRA. Weiss 2010 included children aged 6 to 14 years with a history of chronic asthma needing asthma medication in the six months preceding screening, at least one asthma exacerbation in the previous year in conjunction with a cold, and an alteration in environment differing from their typical school or education environment throughout August/September. Morita 2017 recruited 1‐ to 14‐year‐old children with physician‐diagnosed asthma, needing a rescue inhaler in the last year, and with a history of asthma exacerbations associated with apparent respiratory viral infections. Exclusion criteria were significant cardiorespiratory comorbidity, regular OCS use, or an asthma exacerbation in the month before the treatment period. Teach 2015a recruited children aged 6 to 17 years with an asthma diagnosis or symptoms for more than a year and at least one asthma exacerbation (requiring systemic corticosteroids or hospitalisation) within the prior 19 months. Additional inclusion criteria were positive perennial allergen skin test response, body weight and total serum IgE levels suitable for omalizumab, school attendance the following August or September, residence in a low‐income census tract, and insurance covering standard medications. There were no exclusions for this study beyond not meeting these inclusion criteria.

Baseline characteristics of participants

All five studies recruited more male than female participants: in each study 60% to 65% of participants were male. No study reported smoking status or exposure to environmental tobacco smoke. Only two studies reported baseline lung function: in Weiss 2010 mean forced expiratory volume in the first second of expiration (FEV₁) was 89.8% predicted in the intervention group and 90.1% in the usual care group, and in Teach 2015a mean FEV₁ across both groups at randomisation was 90.2% predicted. Only Teach 2015a systematically reported asthma severity: 195 randomised children were classified as step 5 according to a severity scale based on the National Heart, Lung and Blood Institute Expert Panel Report‐3 (severe persistent symptoms requiring high‐dose ICS and one adjunctive therapy), and 318 met the criteria for asthma severity steps 2 to 4 (mild‐moderate persistent symptoms requiring preventer medication but no more than medium‐dose ICS and one adjunctive therapy). Johnston 2007 reported that 90% of children were routinely receiving ICS, suggesting moderate severity, whilst only 30% of participants in Weiss 2010 and 50% in Morita 2017 routinely received ICS at randomisation. It is likely that the general practice‐based population in Julious 2016 included more people with mild asthma than the studies recruiting from secondary care.

Description of the intervention

The behavioural public health intervention was a letter sent to parents/carers of school‐aged children with asthma from the child's general practitioner reminding them to maintain their child’s medication and to collect a prescription if they were running low on medication. The letter was sent out during the week commencing 29 July and highlighted that school return is a time when asthma can worsen. The comparison group did not receive a letter (Julious 2016). In the pharmacological studies the interventions were added to usual care and compared with a placebo in addition to usual care. In Johnston 2007, an age‐specific dose of montelukast was given from 1 September to 15 October, whilst participants in Weiss 2010 received 5 mg montelukast from the night before the first day of school for eight weeks. Children in the intervention group in Morita 2017 received pranlukast 7 mg/kg twice daily between 15 September and 14 November. In Teach 2015a, children were randomised 3:3:1 to a standard dose of omalizumab based on serum IgE levels and weight, a doubling of their ICS dose, or placebo from 4 to 6 weeks before the start of the autumn term, continuing for 90 days after school return. Only children at steps 2 to 4 were entered into the ICS boost arm because of concerns that very high‐dose ICS provides limited additional efficacy and increases the risk of side effects.

Outcomes and analysis

Julious 2016 studied a number of outcomes, but the primary outcome was the proportion of children aged 5 to 16 years with unscheduled contacts during September 2013. Secondary outcomes measured in September included number of unscheduled contacts and proportion and total number of contacts (scheduled and unscheduled) and unscheduled contacts for a respiratory diagnosis. These outcomes were also measured throughout September to December 2013, September 2013 to August 2014, and in September 2014 in an 'echo study' to see if there was a maintained effect in the year following the main study and in which there was no study intervention. Between September 2013 and August 2014 time to first contact, first unscheduled contact, and first unscheduled contact for a respiratory diagnosis were also measured. The proportion of children with scheduled contacts was measured in August 2013, August 2014, and between August 2013 and July 2014. The number of participants collecting prescriptions was measured in August for both years. Quality‐adjusted life years (QALYs) gained and NHS health costs were measured between August 2013 and July 2014. Primary analyses were conducted on an intention‐to‐treat basis.

The primary outcome in Johnston 2007 was percentage of days during the intervention period with worsening asthma symptoms. Data were inputted daily by parents/carers into a prospectively completed sticker chart, and further data were collected by questionnaire two weeks after the end of the intervention period. The secondary outcome was number of unscheduled care visits. Analysis was intention‐to‐treat. In Weiss 2010, the primary outcome was percentage of days with worsening asthma symptoms, defined as one or more of increased beta‐agonist use, increased daytime symptoms score, night wakening, increased ICS use, OCS rescue or unanticipated visits to a doctor, emergency department, or hospital for asthma. Secondary outcomes were individual components of the composite primary endpoint and adverse events. Data were collected at 4, 8, and 10 weeks of the study and analysed in the full‐analysis population (all children who received at least one dose of study medication and had a valid measurement of the percentage of days with worsening asthma during the study period, derived from at least seven days of diary data). In Morita 2017, the primary outcome was total asthma score during the 60 study days, calculated based on asthma symptoms, need for medication, and need for an unscheduled physician visit or OCS. The secondary outcomes were days with worse asthma symptoms, number of colds, and days with fever. Data were analysed per protocol only from those compliant with treatment and returning adequate outcome data via a daily sticker chart. Teach 2015a conducted a modified intention‐to‐treat analysis, analysing data from children who were randomised, began study treatment, and had at least one study contact during the 90‐day outcome period. The primary outcome was asthma exacerbation, defined as worsening of asthma control requiring systemic corticosteroids or hospitalisation, during the 90‐day period from the first day of each child's school year. Secondary analyses considered exacerbations during the 90‐day intervention period according to subgroups based upon: exacerbation during run‐in, eosinophil count, total IgE, roach IgE, age, fractional exhaled nitric oxide (FeNO), FEV₁, body mass index, ethnicity and gender. Interferon alpha responses to rhinovirus were measured in peripheral blood mononuclear cells from a subset of children.

Proportion of children experiencing one or more asthma exacerbations during the autumn period

Only Teach 2015a compared the number of children experiencing asthma exacerbations exactly as defined in the primary outcome of this review. Evidence relevant to this outcome is summarised in summary of findings Table for the main comparison and summary of findings Table 2. During the 90‐day period from the first day of each child’s school year, the omalizumab intervention was associated with exacerbation (worsening of asthma control requiring systemic corticosteroids or hospitalisation) in 11.3% of children compared with 21.0% in the placebo arm, odds ratio (OR) 0.48, 95% confidence interval (CI) 0.25 to 0.92 (adjusted for study centre, dosing schedule, and asthma severity step) (Analysis 1.1, Figure 3). Considering those with stage 5 asthma, omalizumab was associated with a reduced odds of exacerbation (OR 0.37, 95% CI 0.17 to 0.81). In contrast, considering only steps 2 to 4 where children were allocated to omalizumab, placebo, or a third arm of a doubling of ICS, exacerbation rates were experienced by 8.4%, 12.7%, and 11.1% of children, respectively. The odds of exacerbation did not differ significantly between any pair of groups (omalizumab versus placebo OR 0.63, 95% CI 0.22 to 1.78; omalizumab versus inhaled steroid boost OR 0.73, 95% CI 0.33 to 1.64; inhaled steroid boost versus placebo OR 0.86, 95% CI 0.32 to 2.30). However, when those experiencing a recent exacerbation (during the four‐ to nine‐month run‐in ending four to six weeks before school return) were considered separately from those without a recent exacerbation, reduced odds of exacerbation were seen across all severity steps 2 to 5 in the omalizumab group compared to placebo (OR 0.12, 95% CI 0.02 to 0.64) and compared to ICS boost across steps 2 to 4 (OR 0.05, 95% CI 0.003 to 0.98). For those without an exacerbation during run‐in, the odds of exacerbation were OR 0.88, 95% CI 0.35 to 2.18 compared to placebo across steps 2 to 5 and OR 1.34, 95% CI 0.56 to 3.25 compared to ICS boost across steps 2 to 4.

Figure 3

---

---

Forest plot of comparison: 1 Interventions for autumn exacerbations of asthma versus usual care, outcome: 1.1 Exacerbations defined according to the review's primary outcome.

Exacerbations reported according to alternative definitions

Johnston 2007 and Weiss 2010 considered the percentage of days with worsening asthma symptoms as the primary outcome, whilst Morita 2017 used a total asthma score based upon symptoms, medication need, and healthcare utilisation. These outcomes were not suitable for expression as an odds ratio. Johnston 2007 defined worsening asthma symptoms as symptoms that were worse than usual or needed extra asthma medication, or required an unscheduled visit to a doctor or treatment with oral corticosteroids; a 53% reduction in days with worsening asthma symptoms was reported compared with placebo during the 45‐day intervention (3.9% versus 8.3%, P = 0.02). Boys aged 2 to 5 years showed greater benefit from montelukast than did older boys, whereas among girls the treatment effect was most evident in 10‐ to 14‐year‐olds. The proportion of participants reporting one or more unscheduled visits to a doctor for asthma symptoms was markedly reduced in the montelukast group compared to the placebo group (4.1% versus 14.6%; OR 0.25, 95% CI 0.08 to 0.79), and it was reported that all prescriptions of OCS for asthma exacerbation occurred as a consequence of an unscheduled visit to a doctor. Weiss 2010 defined worsening asthma symptoms as one or more of the following actions: increased beta‐agonist use; increased daytime symptoms score; being awake 'all night' due to asthma; increased ICS use; OCS rescue; or unanticipated visits to a doctor, emergency department, or hospital for asthma. Analyses were adjusted for treatment, school start date, investigator site type, ICS use at entry, age, and sex. We found no significant difference in worsening symptoms between groups or for any component of this outcome, including OCS use (26.0% versus 30.3%; OR 0.79, 95% CI 0.59 to 1.06) and unanticipated medical contacts (11.8% versus 14.7%; OR 0.77, 95% CI 0.53 to 1.13). There was a consistent direction of effect favouring the intervention for five of the six outcomes, but none reached significance. Prespecified subgroup analyses found significantly fewer days of worsening symptoms in boys and in children aged 10 to 14 years, although interaction terms for age and gender were non‐significant. Morita 2017 based total asthma score on asthma symptoms, need for increased asthma medication, unscheduled physician visit or OCS; an adjustment was made in multivariable analysis for ICS use. There were no significant differences between pranlukast and control group in total asthma score (5.5 versus 7.8, P = 0.35) or days of worsening asthma symptoms (1.5 versus 1.8, P = 0.67). Significantly lower asthma scores and number of colds were seen for boys age one to five years. A higher number of colds and days of fever were seen in the control group compared to the pranlukast group, but only the latter reached significance (P = 0.06 and P = 0.04, respectively). Unscheduled visits to a doctor or OCS did not differ between groups (5.9% versus 8.6%; OR 0.67, 95% CI 0.16 to 2.80, Analysis 1.2, Figure 4). Evidence relevant to LTRA‐based interventions is summarised in summary of findings Table 3.

Figure 4

---

---

Forest plot of comparison: 1 Interventions for autumn exacerbations of asthma versus usual care, outcome: 1.2 Exacerbations defined according to study‐specific definitions.

In Julious 2016, exacerbations were not directly measured, but numbers of unscheduled contacts were reported as a proxy for this, and the study also reported unscheduled contacts coded with a respiratory diagnosis. Neither outcome significantly favoured the intervention. Data were reported between September and December rather than for the three months following school return; during September to December unscheduled contacts for a respiratory diagnosis were recorded for a greater proportion of children receiving the intervention letter than for those in the control group (18.4% versus 16.7%; OR 1.13, 95% CI 0.95 to 1.33), but this difference was not significant (summary of findings Table 4). Unscheduled contacts for any diagnosis also did not differ significantly between the intervention and usual care groups (80.1% versus 79.1%; OR 1.10, 95% CI 0.96 to 1.26). The primary outcome period for this study was September; during this period no significant between‐group differences were reported for the proportion of children for whom any medical contact or any unscheduled contact was recorded. Unscheduled contacts for a respiratory diagnosis were recorded in significantly higher numbers in the intervention than in the usual care arm (5.3% versus 4.2%; OR 1.30, 95% CI 1.03 to 1.66). Analyses were modelled using age, sex, number of contacts the previous September and the trial arm as fixed‐effect, and the design/cluster effect of general practice as random‐effects. The study authors suggested that contacts following the intervention might have occurred as a result of appointments needed to assess children's need for preventer medication.

Asthma exacerbations and the period during which children were considered at risk of exacerbation after school return were defined differently in each trial. Moreover, the interventions trialled rarely used the same approach or medication. For these reasons, we limited meta‐analysis to studies with comparable interventions based upon seasonal administration of montelukast. Even amongst these three trials, participant populations and outcomes varied slightly: in Johnston 2007 the intervention period was fixed for 45 days from 1 September, whilst in Weiss 2010 the intervention period was for eight weeks from the night before each child's school return, and in Morita 2017 children were randomised two to six weeks before a fixed 60‐day study period starting from 15 September. The participant populations in these trials differed according to both age and asthma severity: participants were both younger and more likely to be receiving ICS in Johnston 2007 and Morita 2017 than in Weiss 2010. Despite the higher proportion of children receiving ICS at trial outset in Johnston 2007 and Morita 2017, higher rates of oral steroid prescription occurred in Weiss 2010. It was not possible to assess the review's primary outcome in these studies since, although each separately reported OCS prescription and unscheduled medical contact, the proportion of children with an exacerbation needing hospitalisation or OCS was not reported. Where evidence was based on single studies, the quality was moderate, downgraded due to small numbers of participants randomised or use of an indirect outcome (unscheduled respiratory contacts in Julious 2016 rather than hospitalisation or oral steroid requirement). When results from Johnston 2007 and Weiss 2010 were included in a random‐effects model, the odds ratio for unscheduled medical contacts was 0.50, 95% CI 0.17 to 1.46. We judged the evidence to be of low quality due to poor consistency between studies and concerns about the indirect outcome of unscheduled medical contact, the threshold for which appeared to differ between studies. The I² statistic was 70%, so we deemed a fixed‐effect model to be inappropriate.

Hospital and paediatric intensive care unit admissions and asthma‐related deaths

Although Johnston 2007, Weiss 2010, Teach 2015a, and Morita 2017 reported medical contact data, including hospital admission, no study presented data on hospitalisation or paediatric intensive care unit admission separately from total unscheduled contacts or OCS use. No study reported any asthma‐related deaths.

Asthma control, quality of life, and impact on schooling

No study reported asthma control measured by a standardised tool (e.g. Childhood Asthma Control Test (cACT) or Asthma Control Test (ACT)). Weiss 2010 mentioned increased symptom score, but the scale on which this was measured was not defined. We found no significant between‐group differences for this outcome. No study measured asthma‐related quality of life or absence from school (or employment for those beyond school age).

Adverse events

In the pharmacological studies, there was no evidence that either adverse events or serious adverse event rate differed between the intervention and the usual care group. Adverse events were not formally reported in the primary care intervention study (Julious 2016). Morita 2017 reported that no children discontinued study medication due to an adverse event, and the authors of this study confirmed that no adverse events occurred in either group. In Johnston 2007, minor adverse events occurred in 20.4% of children in the montelukast group and in 28.1% of children in the placebo group (OR 0.66, 95% CI 0.34 to 1.27) (summary of findings Table 3). Adverse events caused two children to discontinue the placebo: one child experienced behavioural change and the other tiredness and appetite changes. A significant behavioural disorder requiring emergency treatment was identified in a participant from the montelukast group at the follow‐up interview. No adverse events were described as serious. Teach 2015a reported adverse events during the period between randomisation and 30 days after the end of the intervention period. Of those children receiving at least one dose of the study drug, 54.5% in the omalizumab arm and 54.8% in the placebo arm experienced an adverse event (OR 0.99, 95% CI 0.61 to 1.58, Analysis 1.3, Figure 5) (summary of findings Table for the main comparison). Considering only children eligible for the ICS boost (steps 2 to 4), one or more adverse events were reported by 43.5% of children in the ICS boost arm and 53.3% of children in the placebo arm (OR 0.67, 95% CI 0.34 to 1.33). Two serious adverse events occurred: a seventh nerve palsy in the placebo group and an episode of anaphylaxis in the ICS boost arm (summary of findings Table 2). In Weiss 2010, 33.9% of children in the montelukast group and 33.6% of those in the placebo group reported at least one adverse event (OR 1.02, 95% CI 0.79 to 1.30) (summary of findings Table 3); the most common adverse events were upper respiratory tract disorders and infections. Four serious adverse events occurred in the intervention group and one in the placebo group (0.7% versus 0.2%). Consequently, there was no evidence in any study that total adverse events occurred more frequently in the intervention than in the usual care arm. Moreover, we found no evidence of a significant difference between these groups when data from the montelukast studies were pooled in a random‐effects model (OR 0.91, 95% CI 0.63 to 1.32; I² = 32%). However, significantly more children experienced local administration site reactions in the intervention group in Teach 2015a compared to the usual care group (15.3% versus 6.5%, P = 0.03). We graded the quality of the evidence for this outcome as high for the pooled montelukast data and moderate for omalizumab or steroid boost intervention, downgrading the evidence due to the imprecision inherent to low participant numbers.

Figure 5

---

---

Forest plot of comparison: 1 Interventions for autumn exacerbations of asthma versus usual care, outcome: 1.3 Adverse effects.

Subgroup analyses

We had planned analyses separating studies of pharmacological interventions from studies of non‐pharmacological interventions, and to consider separately those with mild‐to‐moderate disease and those with severe asthma. Due to the low numbers of studies identified and the likely heterogeneity introduced by combining different pharmacological interventions, the planned subgroup analyses were not justified. We identified only one non‐pharmacological study, and baseline medication use and asthma severity were not always well described.

Sensitivity analyses

We had planned an analysis including only studies without missing data and an analysis excluding cluster‐randomised trials. Due to the low numbers of studies identified, these subgroup analyses were not possible.