Pragmatic RCT of High-dose Oral Montelukast for Moderate and Severe Pediatric Acute Asthma Exacerbations

Study overview: The approach to testing the central hypothesis will be to conduct a two-arm, parallel randomized controlled trial of high-dose (30mg) montelukast versus identical placebo, as add-on to standard treatment of SCS and inhaled SABA, in children aged 5 to 17 years with moderate and severe acute asthma exacerbations. Although this aim will enable us to test each hypothesis, the ability to test each hypothesis is independent of the others.

Randomization: The investigators will use randomly-permuted blocks of 4 to 8 to minimize seasonal bias of exacerbation precipitants that may have independent associations with treatment-response.

Masking of treatment-allocation and outcome-ascertainment: Investigator or clinical team knowledge of treatment allocation may influence assessment of outcomes. Allocation concealment will minimize this bias. The investigators will adhere to established procedures to maintain masking of participants, CTAs, and data analysts.

Aim 1: Testing the primary hypothesis The investigators will measure lung function before and after 2-hours of treatment using airway resistance by impulse oscillometry at 5Hz (%R5).

Expected outcomes of the primary aim: The investigators expect that high-dose montelukast will result in a minimum 15% greater improvement in %R5 and a 10% improvement of %FEV1 between pre-treatment and 2-hours after dosing in comparison with standard treatment. The investigators base this expectation on (1) The known contribution of leukotrienes to airway inflammation during acute asthma exacerbations;8-11 (2) Knowledge that corticosteroids do not inhibit leukotriene synthesis in vivo;7 (3) High-quality trials of IV montelukast in patients with moderate and severe acute asthma exacerbations by Camargo and colleagues that demonstrated rapid and sustained improvement of lung function and decreased need for SCS;25,26 and (4) The pharmacokinetics of IV and oral montelukast indicating that serum levels after high-dose oral montelukast are comparable to the IV doses used in the Camargo trials.29,69-72 The investigators expect that high-dose montelukast will result in meaningful improvement of clinical severity, hospitalization rate, and 72-hour symptom burden.

Secondary Aim and testing the secondary hypothesis: Respiratory viral detection Specimen processing. The investigators will obtain a nasal swab from each participant before treatment. PCR testing for respiratory viruses will be conducted in the laboratory of Dr. Natasha Halasa (Co-I). Nasal and throat swabs combined in 3 ml sterile M4RT transport medium (Remel) will undergo immediate refrigeration at 2-8oC, followed by transportation within 24 hours on ice to the Halasa Lab for processing. Five aliquots in 2-ml screw-cap cyrovials will be prepared three ~0.85-ml volumes in original transport medium and two ~0.2-ml volumes in commercial lysis buffer compatible with extraction methods employed in the Halasa Lab. Aliquots will be flash-frozen prior to storage at -80oC to maximize viral stability and specimen quality.

PCR detection of respiratory viruses: Testing for influenza A, B, and C; RSV; adenovirus; enterovirus; human metapneumovirus; human rhinovirus; parainfluenza 1-4; coronavirus 229E, NL63, OC43, and HKU1; and human bocavirus can be performed according to established protocols using optimized target-specific primers and FAM/BHQ1-congugated hydrolysis probes, AgPath-ID One Step RT-PCR chemistry (Applied Biosystems), and StepOnePlus Real Time PCR System. Total nucleic acid extraction will be performed using the Roche MagNA Pure LC automated extraction system capable of high-throughput specimen processing to yield exceedingly pure RNA. Based on an assay cutoff of Ct equal to 40, specimens demonstrating Ct values less than or equal to 40 for viral sequence signatures will be considered positive for the targets in question. Specimens negative (Ct greater than 40) for RNAseP will be retested using the same preparation of nucleic acid and further tested using a fresh extract if the original extract repeatedly produces a negative result. Specimens persistently demonstrating RNAseP Ct values greater than 40 will be deemed indeterminate for negative viral targets. Viral target detection in RNAseP-negative specimens will be considered true-positive assuming processing and plate controls produce expected patterns of results.

Urinary LTE4 measurement will be by the Vanderbilt Eicosanoid Core Laboratory. Expected outcomes of the secondary aim: First, the investigators expect that there will be a high incidence of viral respiratory detection in the cohort, based on reports from Johnston, Khetsuriani and others. Respiratory viral detection does not imply viral respiratory infection (VRI). Nonetheless, the investigators anticipate that there will be (1) Greater effects of montelukast on lung function in participants with respiratory virus detected in comparison with those who do not have virus detected; (2) Greater effects of montelukast on lung function in proportion to urinary LTE4 levels; and (3) An interaction between viral detection and urinary LTE4 level on treatment-response.

Power Calculation and Statistical Approach: Power calculation for this research is based on the primary outcome measure, %R5 by IOS.

The primary outcome measure will be %R5 because this IOS parameter measures total airway resistance. The data from 192 children aged 5 to 17 years with acute asthma exacerbations include an SD for pre-treatment %R5 of 71.7% and a correlation coefficient (r) of 0.52 between pre-treatment and 2-hour %R5. For residual variance the investigators considered a linear regression model with the 2-hour value as the outcome and the pre-treatment value as an adjustment variable. With 125 participants having complete IOS data in each of the two treatment-allocations arms, the investigators will have 90% power to detect a minimal difference of %R5 of 14% between montelukast and placebo with one interim data analysis conducted when 50% of the subjects have accrued. In order to account for missing IOS data in up to 15% of participants, dropouts and missing data the investigators propose to enroll 330 participants with 165 randomized to each RCT arm.

The investigators will also examine additional IOS parameters as outcomes, including those representing large (R20) and small airway function (R5 to R20, X5 and XA), as well as change of %-predicted FEV1 (%FEV1) and of the AAIRS bedside severity score between pre-treatment and 2-hours. Because the investigators anticipate that approximately 50% of the cohort with moderate and severe exacerbations will be able to provide spirometry meeting ATS quality and reproducibility criteria, there may be insufficient power to detect meaningful differences for this outcome. However, the investigators will be able to score the AAIRS in all participants and thus anticipate sufficient power to detect a minimum 2 point difference of this 17 point severity score between montelukast and placebo arms.

Statistical approach Primary statistical inferential test: Outcomes are measured on a continuous scale and will be analyzed using linear regression and include treatment indicator and baseline value as covariates. The investigators will estimate the bias corrected mean effect of treatment with corresponding 95% confidence intervals that taken into account one planned interim analyses. The investigators will ascertain that the assumptions of inferential tests are satisfied for all analyses. If assumptions are violated, alternatively the investigators will use the proportional odds ordinal logistic regression model which generalizes the non-parametric Wilcoxon rank sum test to a regression setting.

Secondary tests for effect modification by viral detection and urinary LTE4 level: As noted above, the investigators anticipate an interaction of viral detection and urinary LTE4 level on montelukast treatment-response. Additional subgroup analyses by age groups, pretreatment severity (moderate versus severe) and exclusion of participants with rapid response (10 minute) to albuterol are also of interest. To test secondary hypotheses that the treatment effect is modified by covariates, the investigators will fit separate models that also include the interaction of the covariate with the treatment effect. Using the sample size assumptions above and additionally anticipating that 60% to 80% of subjects will have viral detection at baseline, the investigators will have 80% power to detect a 25% to 31% modification of the treatment effect. The investigators will report the results of all subgroup analyses conducted regardless of statistical significance.

Intention-to-treat-analysis and inferential test assumptions. Analyses will conform to the principle of intention-to-treat. All randomized participants will be included in the primary and secondary analyses in their assigned treatment allocation groups.

Stopping rules for use by the Data Safety and Monitoring Committee Stopping boundaries and design operating characteristics were investigated using the RCT-design R package. This package provides a comprehensive suite of functions for evaluating, monitoring, analyzing, and reporting clinical trial designs. While finalized stopping rules will be developed and agreed on by investigators during the R61 startup phase, the investigators summarize some current results. Using the Emerson and Flemming (1989) symmetric test125 with an assumed treatment effect of 14% and standard deviation of 34%, the investigators would stop the trial early at the interim analysis for futility if estimated treatment effect is 0.0% or lower and stop early for efficacy if the treatment effect is 17.0% or greater. If the true treatment effect is 14%, there is an estimated 32% chance of stopping early. Should the trial continue to full enrollment, the bias adjusted treatment effect will be significant for efficacy if it is estimated to be 8.5% or larger.

Missing data will be handled by joint modeling of the treatment effect and informative missing data.

The investigators expect that outcome data on some subjects will be missing, and this data will not be missing (completely) at random. In particular, FEV1 measurements will be more difficult to obtain on subjects with more severe exacerbations. Failing to account for the informative missingness could bias the estimate of the treatment effect, so the investigators will consider joint models for the missing data and estimated treatment effect.126,127 Sensitivity analyses in which the treatment is estimated using different models for the missing data will be conducted to determine the robustness of the treatment effect estimate. Detailed analyses will specify that the missing data models will be created before data collection, during the R61 startup phase.