Flow of participants into study enrollment was slower than expected, with a rate of 22/month compared to the expected 56/month. The 292 enrolled participants had historical characteristics consistent with moderate to severe persistent asthma, with 82% taking medium to high dose ICS (800 mcg budesonide or equivalent or higher): almost one-half of these were inadequately controlled prior to study entry by symptom report. Eighteen percent of enrolled participants had inadequately controlled asthma while taking a lower dose of ICS, 400 mcg budesonide equivalent; these participants were given a higher dose of ICS + salmeterol during the run-in study phase to determine if they would qualify. Of the 292 total enrolled, 55 (19%) were randomized. Length of run-in was median 9 weeks (range 6–21 weeks). Reasons for not achieving randomization are shown in . The two most common reasons were: 1) controlled asthma during the run-in phase (no symptoms or no airway lability demonstrated) (n=59, 25%) on low dose ICS (400 mcg budesonide BID) + salmeterol, and 2) adherence below the standard of 80% (n=89, 37%) (e.g., adherence with Diskus® in these children was a median of 70%, range 17–77%). There were no differences in ages of the children who did not reach randomization criteria (mean ages in years, too severe 12.3, too mild 11.9, non-adherent 11.8) compared to those who were successfully randomized (11.4 years). Elevated liver enzymes (n=6) and a prolonged QTc interval (n =3) were infrequent reasons for ineligibility.
The three treatment groups were well balanced, with characteristics at the time of randomization consistent with guideline definitions of moderate to severe persistent asthma controlled with use of moderate to high dose ICS and salmeterol (). After randomization, 100% of expected visits and 89% of expected phone contacts were completed, along with 90% adherence to diary entries. Adherence to medications estimated from salmeterol Diskus® counter and from Electronic Drug Exposure Monitor records for use of study capsules were 85 and 89%, respectively. There was no difference in adherence to use of study procedures or medications between treatment groups. Analysis of adherence within study groups for capsule (azithromycin or placebo) and tablet (montelukast or placebo) use monitored by eDEM found differences only by minority (Caucasian vs. Minority), with Caucasians doing better than minorities for both capsules and tablets (capsules: 92.2% in Caucasians compared to 81.5% in minorities, with a standard error of the difference = 4.4%, p=0.02; tablets: 91.6% in Caucasians compared to 80.7%, with a standard error of the difference = 4.7%, p=0.02). There was insufficient power to detect a difference when analyzed by treatment group. Rates of adherence for use of Diskus did not differ by gender, ethnicity, SES, or age group overall. There were differences in rates of adherence of use of Diskus among age groups in the placebo and montelukast arms: within the placebo group, adherence was 93.7% for ages 10–14 years compared to 80.7% for those 15–18 years, with a standard error of the difference = 4.7, p=0.0495; for the Montelukast group, adherence was 96.0% for ages 6–9 years compared to 83.4% for those 15–18 years, with a standard error of the difference = 5.6, p=0.0417). For those who had treatment failure, adherence was < 80% for use of Diskus for only 1 patient who was in the placebo group and only at the treatment failure visit. Adherence for oral medications in those achieving treatment failure criteria was < 80% for 1 in the azithromycin group, 1 in the montelukast group, and 3 in the placebo group. Three subjects who completed the full study had adherence < 80 % throughout the treatment phase, 2 in azithromycin and 1 in montelukast.
Characteristics at time of randomization
Early Discontinuation of the Study
The a priori study monitoring plan prescribed that the Data Safety Monitoring Board (DSMB) examine study progress at 6 and 13 months after initiation of enrollment. At the second planned review, enrollment and efficiency of randomization were lower than expected, yielding a much lower than projected number of study participants. As such, the DSMB requested a futility analysis.
The observed data indicated that a reasonable assumption for the true failure rate (that is the proportion of participants who become uncontrolled) in the placebo group was approximately 0.5. Conditional power was calculated to determine what would be the probability of rejecting the null hypothesis of no active treatment effect at the end of the trial (210 randomized children) given that all the groups, active treatments and placebo, displayed approximately a 0.5 failure rate at the interim stage (55 randomized children). illustrates the conditional power curve for a range of true failure rates for either of the active treatments. On the right of the figure are power calculations for higher failure rates in the treatment groups than in the placebo, and on the left, power calculations for higher failure rates in the placebo group than in the treatment groups. shows that there was only 50% power to detect a true failure rate of 0.25 or 0.75 in the treatment groups as compared with a failure rate of 0.50 in the placebo group. Using the lower 95% confidence limit of the failure rates observed among the first 36 randomized children to the active treatments (e.g, 0.28) as the best-case scenario, the maximum conditional power was 40% to detect a significant active treatment effect if the study had continued to completion using the originally planned sample size (210 randomized children). For these reasons, the DSMB judged it to be futile for the MARS trial to continue, and recommended to the NHLBI that the trial be terminated.
Outcomes of those randomized
Of the 55 randomized children, 10 (18.2%) remained controlled throughout ICS reduction until the last planned study visit at 30 weeks after randomization. Seven (12.7%) were still controlled when the study was stopped by the DSMB at a median of 15.9 weeks (range 2.4–23.4 weeks) after randomization. Thirty-five (63.6%) reached inadequately controlled status following randomization at a median 5.1 weeks (range 0.4–29.0) after randomization. There were no differences between those who remained controlled compared to those who reached inadequately controlled status by age (p=0.10), gender (p=0.13), minority status (p=0.19), race (p=0.13), ethnicity (p=0.96), or income (p=0.73). Three subjects were withdrawn due to reasons other than inadequate control: abdominal pain within 24 hours of drug administration (azithromycin, n=1) and withdrawal of assent (azithromycin and placebo, n=1 each).
Of the 35 who were discontinued for attaining inadequately controlled status reasons were an exacerbation requiring oral corticosteroid in 4 participants, symptoms or albuterol use >6 days/week on average over 2 weeks in 16 (2 of these also had nocturnal awakenings >=2 days/week), nocturnal awakening with minimal accompanying daytime symptoms in 3, PEF readings on the diary card <80% expected >6 days/week on average over 2 weeks in 9, and physician discretion in 3. Reasons for reaching inadequately controlled status before the first reduction of ICS at 6 weeks after randomization (N=21, 60% of those reaching inadequately controlled status) were symptoms in 10 (48%), need for OCS in 3 (13%), PEF values <80% of the pre-bronchodilator value at the randomization visit without accompanying symptoms in 6 (29%), and physician discretion 2 (10%).
Among the 55 randomized children, time to inadequate asthma control after randomization is shown in . summarizes the proportion of children within each of the three randomized groups who met the criteria for inadequate asthma control. There is a great deal of overlap among the three confidence intervals, which is mostly attributable to the small sample size within each randomized group. The logrank test was applied to compare the three randomized groups with respect to the time until inadequate asthma control (failure) and it yielded a non-significant result (p = 0.49).
Product-limit survival function estimates for time to return of inadequate control for children assigned to azithromycin (n=17), montelukast (n=19), and placebo (n=19).
Proportion of participants reaching inadequately controlled status after randomization
PCR assays to detect Mycoplasma pneumoniae and Clamydia pneumoniae were done on the 55 participants as designated in the protocol ( in Online Repository), at randomization, visit 5 and completion of the study at visit 7, or at a treatment failure visit. None of the 140 samples assayed had evidence of DNA from Mycoplasma pneumoniae and Clamydia pneumoniae.