Probability of successful inhaled corticosteroids cessation in preschool wheezers: a predictive score

Nearly all asthma predictive tools estimate the future risk of asthma development. However, there is no tool to predict the probability of successful ICS cessation at an early age. Therefore, we aimed to determine the predictors of successful ICS cessation in preschool wheezers, and developed a simple predictive tool for clinical practice. This was a retrospective cohort study involving preschool wheezers who had undergone an ICS therapeutic trial during 2015–2020 at the University Hospital, Southern, Thailand. A predictive scoring system was developed using a nomogram to estimate the probability of successful ICS cessation. We calculated area under ROC curve and used a calibration plot for assessing the tool’s performance. A total of 131 medical records were eligible for analysis. Most of the participants were male (68.9%). More than half of the preschool wheezers had successful ICS cessation after an initial therapeutic trial regimen. The predictors of less successful ICS cessation were perinatal oxygen use [OR 0.10 (0.01, 0.70), P = 0.02], allergic rhinitis [OR 0.20 (0.08, 0.56), P = 0.002], blood eosinophil count > 500 cell/mm3 [OR 0.20 (0.06, 0.67), P = 0.008], and previous ICS use > 6 months [OR 0.30 (0.09, 0.72), P = 0.009]. Conclusions: Predictors of less successful ICS cessation were the following: perinatal oxygen use, allergic rhinitis, blood eosinophil count > 500 cell/mm3, and previous ICS use > 6 months. A simple predictive score developed in this study may help general practitioners to be more confident in making a decision regarding the discontinuation of ICS after initial therapeutic trials. What is Known: • Early allergic sensitization is associated with reduced chances of inhaled corticosteroid cessation at school age. • Prolonged ICS is associated with the emergence of adverse effect and discontinuing too early can result in recurrence symptoms. What is New: • Requirement of oxygen support within 7 days after birth in term neonate is a postnatal factor associated with less successful ICS cessation. • We propose a simple predictive tool with easily available clinical parameters (perinatal oxygen use, allergic rhinitis, blood eosinophil count, parental asthma history, and duration of previous ICS use) to determine the timing of inhalational corticosteroid cessation in preschool wheezers. What is Known: • Early allergic sensitization is associated with reduced chances of inhaled corticosteroid cessation at school age. • Prolonged ICS is associated with the emergence of adverse effect and discontinuing too early can result in recurrence symptoms. What is New: • Requirement of oxygen support within 7 days after birth in term neonate is a postnatal factor associated with less successful ICS cessation. • We propose a simple predictive tool with easily available clinical parameters (perinatal oxygen use, allergic rhinitis, blood eosinophil count, parental asthma history, and duration of previous ICS use) to determine the timing of inhalational corticosteroid cessation in preschool wheezers.


Introduction
Recurrent wheezing is common among children younger than 5 years of age, with its prevalence varying from 30 to 50% worldwide [1]. The rates of related emergency department visits and hospitalizations have been reported to be 71% and 27%, respectively, causing considerable health economic burden and adversely impacting the quality of life [2]. Inhaled corticosteroids (ICS) are the treatment of choice in young children with recurrent wheezing. According to the Global Initiative for Asthma (GINA) guidelines, children under 5 years of age with recurrent wheezing or interval asthma-like symptoms (preschool wheezers) should be given a trial of regular daily low-dose ICS therapy [3]. This initial treatment should be given for at least 3 months to establish its effectiveness in achieving asthma control. If the therapy is discontinued, then symptoms need to be monitored continuously for a minimum of 1 year, assessing for recurrence. Before discontinuing controller medication, determinants such as the wheezing phenotypes (i.e., transient early wheezing, late-onset wheezing, and persistent wheezing) and the future risk for asthma development (e.g., allergic sensitization, atopic disease, and family history of asthma) are considered by the physician [4]. If a preschool wheezer's symptoms disappear without controller medication before the age of 6 years, the child is classified as a transient wheezer [5]. If a preschooler's wheeze needs ICS therapy for symptomatic control until the age of 6 years, the child is classified as a persistent wheezer.
Not all preschool wheezers have asthma when they grow up, but approximately 70% of such children are prescribed intermittent ICS during a bout of respiratory illness [6]. Given that long-term use of ICS has the potential to cause systemic side effects, including adrenal suppression, reduced growth velocity, and bone metabolism disturbance [7], ICS should be prescribed with prudence to prevent an unnecessarily long duration of ICS treatment [8].
A simple tool may be helpful in increasing the general practitioner's confidence in ICS management in preschool wheezers. To determine the timing of ICS cessation in preschool wheezers is crucial because prolonged ICS is associated with the emergence of adverse effects and discontinuing ICS too early can result in recurrence of symptoms. Therefore, this study aimed to develop a scoring system to predict the probability of successful ICS cessation after its first therapeutic trial in preschool wheezers.

Study design and population
This retrospective observational study was conducted at the Pediatric Outpatient Continuity Respiratory Clinic of Songklanagarind Hospital, the university hospital in Southern Thailand. This research was approved by Human Research Ethic Committee (REC. 65-059-1-1), Faculty of Medicine, Prince of Songkla University, Thailand. The written informed consent was waived because of the restrospective nature of the data. Patient information was collected from the hospital records using the International Classification of Diseases, 10 th revision codes. Medical records of children that attended the clinic between 2015 and 2020 with one of the following diagnoses were recruited for initial review: R06.2 (wheezing) and J45 (asthma). Based on the initial review, data of children who met all the following criteria were enrolled as study participants: (1) ≤ 5 years old with recurrent wheezing; (2) initial treatment with regular daily low dose ICS for at least 3 months (therapeutic trial) with good compliance (ICS used > 80% of treatment period); (3) symptoms disappeared without controller medication before the age of six; and (4) follow-up duration at least 1 year after discontinuation of ICS therapy. Children were excluded from the study if they met any of the following criteria: (1) pre-existing congenital airway anomalies such as vascular rings or, bronchial atresia (2) bronchopulmonary dysplasia or chronic lung disease; or (3) incomplete data on important variables (i.e., perinatal history, comorbidities, compliance, and medication) in the medical records.

Data collection
Patient information was collected from medical records at the initial and follow-up visits. The data collected included sex, age at onset of the first wheeze, perinatal use of oxygen, type of feeding (bottle or breast) at the age of six months, family history of asthma, environmental exposure before the age of six (for example, living in an industrial area, smoking exposure, pet ownership, and daycare attendance), wheezing exacerbation related to seasonal changes (seasonal wheezing), comorbidities (allergic rhinitis or, atopic dermatitis), blood eosinophil counts before the initiation of ICS therapy, results of the skin prick test performed at any age, history of respiratory syncytial virus (RSV) infection during the study period, and duration of ICS use.

Outcome measurement
After initial treatment with an ICS therapeutic trial, symptom responsiveness was evaluated and classified into two categories taking ICS prescription into account: (1) symptoms control and lack of wheezing exacerbations without ICS medication until the age of six (successful ICS cessation), or (2) presence of symptoms or wheezing exacerbations and the need for ICS to maintain symptomatic control (recommencement of ICS required). The primary outcomes were the potential factors associated with less successful ICS cessation after the initial therapeutic trial of ICS in preschool wheezers.

Sample size calculation
Sample size calculation was based on a previous study by Sitthisarunkul et al. [9], which reported the skin prick test outcome was predicted ICS cessation in preschool wheezers. The calculated sample size for a two-tailed alpha of 0.05 and predictive power of 80% was 81 subjects. Estimating that 20% of the data would be missing, a final sample size of 98 participants was determined.

Operational definitions
Recurrent wheezing was defined as wheezing episodes occurring more than three times per year. Allergic sensitization to aeroallergens was said to have occurred if the skin prick test was positive for least one aeroallergen. Perinatal oxygen use was defined as oxygen therapy within seven days after birth. Symptoms control was as per the information provided by their caregiver over the previous 4 weeks. Satisfactory symptom control was defined as the absence of use a short acting beta-2-agonist (SABA) reliever more than once a week. Symptoms of an exacerbation presented with an acute or subacute increase in wheeze, shortness of breath and coughing, especially while asleep resulted in reduced exercise tolerance and impairment of daily activities.

Statistical analysis
All statistical analyses were performed with R software, version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria). For continuous variables, normality was tested using the Shapiro-Wilk test. The Wilcoxon ranksum test with continuity correction was used if continuous variables were not normally distributed (the age at onset of the first wheeze, duration of ICS use, blood eosinophil counts). Categorical variables were presented as frequency and percentage (sex, breast feeding, perinatal oxygen use, parental asthma, environmental exposures, allergic rhinitis, and atopic dermatitis). We used the Fisher exact test or a Chi-square test to compare characteristics of the preschool wheezers with and without the outcome. Multiple logistic regression was performed to identify the independent predictors of successful ICS cessation. The covariates with P < 0.2 on univariate analysis were included in the multivariate model. The variable remaining in each final multivariate model was chosen based on backward elimination. P < 0.05 was considered statistically significant.

Prediction model development
We chose the relevant factors associated with successful ICS cessation in transient wheezer using logistic regression. We estimated the impact of predictors using a nomogram [library(nomogramFormula)]. We recorded all potential predictors with more than two response categories into multiple binary variables. The final prediction model allowed the calculation of a predictive score and the probability of successful ICS cessation. To assess performance and determine the discriminative ability of the model, we plotted the receiver operating characteristic curve (ROC) and calculated the area under the curve (AUC) [library(pROC)]. The AUC can take on values from 0 to 1. Discrimination is considered not better than chance if the AUC is 0.5 (moderate discrimination if AUC 0.6-0.8, and good discrimination if AUC > 0.8). For internal validation of our model, we used bootstrap validation and constructed a calibration plot for visualization [library(rms)]. On the calibration plot, a perfect calibration curve would lie exactly on the diagonal line.

Results
Of the 230 preschool wheezers that received an ICS therapeutic trial, 131 transient wheezers were eligible for analysis. The majority of transient wheezers were male (68.9%), and all were born at term by cesarean section (58.9%). The median age was 7.6 (6.7, 8.5) years at the time of data analysis. Approximately 60% of them were breastfed during infancy and 2% had a history of perinatal oxygen use. Parental asthma was present in approximately 15%. Half of the transient wheezers had their first wheeze in their first year of life. Environmental factors inducing wheezing included smoking exposure, pet ownership, and living in an industrial area, which accounted for 41.1%, 16.7%, and 8.9% of transient wheezers, respectively. Seasonal wheezing was observed in 68.9% of transient wheezers, whereas allergic rhinitis and atopic dermatitis were present in 23.3% and 8.9% of transient wheezers, respectively.
Among 131 transient wheezers, successful ICS cessation was observed in 91 children (70%), whilst the other 40 required recommencement of ICS when followed up (57% required ICS recommencement within 6 months of ICS discontinuation) (Fig. 1). The median duration of therapeutic ICS trial among the "successful ICS cessation group" and "ICS recommencement required group" was 7 (IQR 4, 14) and 6 (IQR 4, 11) months, respectively (rank sum test, P = 0.6). For the latter group, the median time to restart ICS was 4 (IQR 3, 9) months after first ICS cessation. The most common reason for recommencement of ICS was episodic respiratory illness. Transient wheezers with seasonal wheezing exacerbation had a sixfold higher chance for early ICS recommencement (within 6 months) compared with those without seasonal wheezing [OR 6.1 (1.03, 36.47), P = 0.04]. We explored the relationship between exacerbation and seasonal changes (Fig. 2) and found a bimodal pattern of exacerbation in February and August-November. The pattern we observed is similar to the pattern of non-RSV triggering wheezing exacerbation. Table 1 shows results from the multivariate analysis. Predictors that were associated with a significantly lower chance of ICS cessation included perinatal oxygen use Although paternal asthma was found to reduce the likelihood of ICS cessation by tenfold compared with maternal asthma, the difference was not statistically significant. Based on multivariate logistic regression, a nomogram for predicting the probability of successful ICS cessation was developed by incorporating five variables, including perinatal oxygen use, allergic rhinitis, parental asthma, blood eosinophil count, and duration of ICS therapeutic trial. The impact of each predictor on the probability of successful ICS cessation was presented by a nomogram (Fig. 3a). The final scoring system allowed the calculation of a predictive score (range from 0 to 450 points) to predict the probability of successful ICS cessation. We stipulated that an area under the ROC curve of 0.75 would guarantee the discriminative ability of our scoring system (Fig. 3b). In a calibration plot, the predicted probability is plotted against the actual probability of successful ICS cessation. The calibration plot showed the agreement of predicted probabilities in our model with a mean absolute error of 0.05, which represents perfect calibration (Fig. 3c).

Discussion
Approximately 70% of preschool wheezers in our study were transient wheezers, defined by successful cessation after first therapeutic ICS trial without recurrence of symptoms after the age of 6 years. We found the predictor of less successful ICS cessation were perinatal oxygen use, allergic rhinitis, ICS use > 6 months, and blood eosinophil > 500 cell/mm 3 . We selected the potential factor and developed a scoring system based on a nomogram to predict successful ICS cessation after the therapeutic ICS trial. General practitioners can use this nomogram in daily practice since the parameters required for scoring are available in most healthcare settings. Having a simple scoring system may help physicians stop ICS treatment with more confidence if they know the chance of exacerbation is low.
The prevalence of transient wheezers in our study was similar to previous reports. In western countries, wheezing resolves at school age in 60-70% of children [10]. A study in Thailand also reported a prevalence of transient wheezers in 65% of preschool wheezers [11]. Our findings suggest that asthma is not specific to any ethnicity. Because transient wheezers account for a large percentage of preschool wheezers, the impact of unnecessary, prolonged use of ICS treatment on children's health is high. Thus, early recognition of transient wheezers would be helpful in preventing this. Predictors for successful ICS cessation vary among studies since asthma management is age specific. Schoolage children can perform a lung function test, which is not applicable to preschool children due to inferior performance. In school-age children, a good lung function test predicts a good prognosis for remission in adult hood [12]. Although predicting asthma remission from a very young age is important, it is limited due to a lack of objective lung function measurement and definitive biomarkers. Previous studies in preschool wheezers showed that atopy and parental asthma are associated with unsuccessful ICS cessation by the age of 6 years [13], while a negative skin prick test is associated with remission [14]. Likewise, in our study, parental asthma and allergic rhinitis were associated with lower chances of successful ICS cessation. We also found that a blood eosinophil count of more than 500 cell/mm 3 was linked to persistent wheezing as previously reported [15]. Allergies to inhaled aeroallergens are known to play a crucial role in mediating chronic airway inflammation and bronchial hyperresponsiveness. However, it is unclear whether intervention against allergen exposure during preschool age will improve the success rate of ICS cessation or not [16]. Another strong predictor for persistent wheezing in this study was a history of oxygen therapy in the perinatal period (about 10 times higher risk). Our study excluded children who were born prematurely since there has been evidence of airway hyperresponsiveness in premature lungs [17]. Our findings suggest that although term newborns do not develop alveolar deficits, they are still vulnerable to airway dysfunction. Perinatal oxygen therapy can increase airway reactivity and airway smooth muscle remodeling, depending on the intensity and duration of oxygenation [18]. We also observed the association between the duration of the therapeutic ICS trial and study outcome. Although a longer duration of ICS use may reflect that patient had poorer symptom control than those who used ICS for a shorter duration, we still included this parameter in our scoring system because it aids in decisionmaking for the practitioner.
The simplicity of the predictive scoring system is the first strength of this study. Users can predict the chance of successful ICS cessation by using one laboratory and four clinical parameters, which are available in a primary health care setting. The scoring system can be used soon after an initial therapeutic ICS trial, making it possible to predict asthma outcome at early age compared to previous scoring systems [19]. Second, the model performance of our scoring system was validated by statistical methods, i.e., calculating the area under the ROC curve and visualizing the calibration plot. There are some limitations related to the retrospective nature of the study. First, the pulmonary function test was not performed due to the limited performance of preschoolers in our study. Second, only approximately 30% of participants had skin prick test results, which was previously reported to be an important predictor [9,14,20]. In our setting, the skin prick test was selectively performed in patients with moderate to severe atopic disease. As we known, food allergy played a role in a part of mechanism of allergic respiratory disease in young children. It was another predictor but, the data of food-induced wheezing episodes did not clearly identify in our medical records. Third, although the tool's performance was assessed, the calibration plot also showed weightage close to the line of equality, and we observed the data set with a probability higher than 80%, quite askew from the ideal line, which could be a result of overfitting of the data. Further studies should externally validate this predictive scoring tool before it can be applied in clinical practice. After external validation, general physicians may use this nomogram when considering discontinuation of ICS after an initial therapeutic trial regimen in preschool wheezers. The compatible features of preschool wheezers' who were born at term without a diagnosis of bronchopulmonary dysplasia or chronic lung disease can be properly examined using this predictive score. For example, if a doctor sees a 3-year-old boy with recurrent wheezing who has received daily low-dose ICS for 3 months (48 points) without wheezing exacerbation during the ICS therapeutic period, without allergic rhinitis (84 points), no perinatal  ), and no eosinophils detected on complete blood count (62 points), applying this information into the predictive scoring system, the total score calculated is 355 points. By this score, one can ascertain that the probability of successful ICS cessation is higher than 90%. Thus, the physician can confirm the decision to discontinue ICS therapy in agreement with the child's parents. A simple predictive scoring tool with easily available clinical parameters for the purpose of determining the timing of ICS cessation in preschool wheezers by general practitioners in clinical practice would be invaluable, thus enabling them to make confident decisions regarding ICS management; a tool that could identify transient wheezers at an earlier stage in which unnecessary ICS use is warranted is thus essential. This decision making is crucial because prolonged ICS use is associated with the emergence of adverse effects; however, discontinuing ICS too early can result in recurrence of symptoms.

Conclusion
Predictors of less successful ICS cessation after an initial therapeutic ICS trial in preschool wheezers were perinatal oxygen use, allergic rhinitis, blood eosinophil count > 500 cell/mm 3 , and previous ICS use > 6 months. We proposed a simple predictive scoring tool to help practitioners make informed decisions on ICS management in preschool wheezers and avoid ICS overuse in early childhood.