| Study summary: |
|
STUDY RATIONALE The obstructive sleep apnea syndrome (OSAS) affects approximately 2% of
children and results in significant morbidity. Continuous positive airway pressure (CPAP)
is the standard therapy for children who fail tonsillectomy and adenoidectomy, or for those
in whom surgery is not indicated. However, CPAP is uncomfortable and is often not tolerated
by children. Thus, a more comfortable technology that would enhance adherence would be
highly desirable. This proposal aims to compare a new technology, Bi-Flex, to standard CPAP
therapy.
Bilevel positive airway pressure (BiPAP): Whereas conventional CPAP therapy delivers a
constant, steady pressure during inspiration and expiration, BiPAP therapy is intended to
respond to both inspiration and expiration by the patient and to deliver a set amount of
pressure when the patient begins spontaneous inhalation, and decreasing pressure when
exhalation begins. Exhaling against lower pressure is thought to be more comfortable for
most patients than the continuous pressure delivered by conventional CPAP therapy, although
it has not been shown objectively to improve adherence. The BiPAP waveform is fixed in that
it provides for set inspiratory and expiratory pressures. For example, the patient's
clinically prescribed BiPAP pressures may be set at an Inspiratory Positive Airway Pressure
(IPAP) =12 cm H2O, and Expiratory Positive Airway Pressure (EPAP) = 8 cm H2O.
The Bi-Flex® feature provides a variation of conventional BiPAP that provides a reduction in
the inspiratory positive airway pressure (IPAP) provided toward the end of the inspiratory
portion of the breathing cycle when the patient's inspiratory airflow normally diminishes,
compared with the level of pressure provided during the majority of inspiration, and also
allows reduction in the expiratory positive airway pressure (EPAP) during the initial
portion of exhalation compared with that provided during the latter portion of expiration.
The purpose of this modification is to provide pressure relief for the patient to smooth the
transition period between the end of IPAP and the beginning of EPAP to allow for a more
comfortable delivery of PAP therapy. Because Bi-Flex provides both a lower expiratory
pressure and pressure relief, it has the potential to be the most comfortable mode of
positive pressure therapy delivery available, and therefore to improve adherence. However,
there have been no studies of adherence to therapy using Bi-Flex in either children or
adults.
STUDY OBJECTIVES
1. To determine whether BiPAP with Bi-Flex as compared to CPAP results in improvement in
adherence of > 30 minutes a night averaged over 3 months of home use.
2. To determine whether BiPAP with Bi-Flex as compared to CPAP results in a lower study
dropout rate.
3. To determine whether Bi-Flex has similar therapeutic efficacy compared to CPAP, as
determined by polysomnographic measurements of OSAS.
4. To determine whether Bi-Flex, compared to CPAP, results in improved subjective comfort
and improved quality of life.
5. To determine whether objective parameters (including demographic and polysomnographic
variables) or subjective parameters predict adherence.
STUDY DESIGN This will be a 3-month, double-blinded, randomized trial of BiPAP with Bi-Flex
vs CPAP in children with OSAS in whom CPAP is medically indicated This will be a 3-month,
double-blinded, randomized trial of BiPAP with Bi-Flex vs CPAP in children with OSAS in whom
CPAP is medically indicated. Subjects will undergo a baseline diagnostic sleep study as
part of their routine clinical care, prior to study enrolment. Those who agree to
participate will then be consented and assented to participate in this study. They will be
randomized to either a Bi-Flex or CPAP treatment arm, in a three to one fashion. In each
arm, each subject will undergo a clinical Bi-Flex or CPAP titration sleep study.
Following these studies, the subjects in both arms will be asked to use their respective
machines, in either the BiPAP with Bi-Flex or CPAP mode, at home for 3 months. During that
time, they will have monthly visits to assess adherence as well as subjective measures of
comfort and quality of life. A repeat sleep study will be performed at the end of 3 months
on the mode assigned (Bi-Flex or CPAP). Subjects will complete standardized questionnaires
regarding symptoms and quality of life throughout the study.
EFFICACY EVALUATIONS Adherence will be assessed by the mean nightly usage as determined by
the equipment compliance recording. Drop-out will be assessed by determining which subjects
are using the device < 2 hours a night. Efficacy of treatment will be assessed using
polysomnographic parameters such as the apnea hypopnea index.
SAFETY EVALUATIONS Polysomnographic parameters such as the apnea hypopnea index, arterial
oxygen saturation and end-tidal PCO2 will be monitored.
STATISTICAL AND ANALYTIC PLAN The primary effectiveness endpoint is the hours of device use
per night over a three month period, as determined from the equipment software.
Analysis Populations
Two analysis populations will be evaluated. An intent-to-treat (ITT) analysis will include
all patients in the trial by their assigned treatment and will be used for effectiveness and
safety. For patients with missing data, values will be imputed as described below. A
second analysis will be done on evaluable patients, i.e., those who have complete data and
these patients will be evaluated for effectiveness.
Primary Safety Analysis
The rate of adverse events will be compared between the two groups by assessing the
proportion of subjects who experience at least one adverse event by Fisher's exact test. If
the test is statistically significant, then this univariate analysis will be followed by a
multivariate logistic regression analysis.
Screening of possible covariates will be done by a method similar to that described in
Hosmer and Lemeshow 25 for logistic regression models. A complete list of clinically
relevant variables to be screened will be provided in a detailed statistical analysis plan
formulated prior to database lock but will include age, gender, height, weight, and others.
The variables on this list will be augmented with variables found to be out of balance
between the treatment groups from the comparability analysis described above. Models will
be fit for each possible covariate to include treatment, the covariate, and the interaction
between the covariate and treatment will be used. Any covariate or interaction significant
at a P-value of 0.15 or lower from the screening analysis will allow the variable to enter
the backward elimination procedure for the final model. If the interaction is significant,
then both the covariate and its interaction need to be included in the model.
The backward elimination will be done manually starting with the interaction with the
highest P-value. Interaction terms will be removed from the model first and then
non-significant main effects. Variables will be retained in the model if the P-value for
the variable or its interaction is 0.05 or less. This analysis is intended to demonstrate
that the treatment affect adjusted for possible covariates is still statistically
significant.
Secondary Safety Analyses The secondary safety analysis will include a descriptive
presentation of individual adverse events each tested with Fisher's exact test. Serious
adverse events will be presented separately with narratives of the events.
Primary Effectiveness Analysis
The primary effectiveness variable is the hours of use per night over three months. This
variable will be summarized with mean, median, standard deviation, minimum and maximum. The
one-sided lower 95% confidence interval will be computed for the Bi-Flex treated subjects
and the upper one-sided 95% confidence interval will be computed for the CPAP treated
subjects. These two intervals will be visually compared to determine if either interval
excludes the mean of the other treatment. Such exclusion will provide suggestive evidence
that the Bi-Flex treatment leads to higher compliance than the CPAP treatment.
Further, a mixed models analysis of hours of use will be done through three months. This
multivariate analysis incorporating clinically important covariates will screen variables
for inclusion in a manner similar to that described above. Screening of possible covariates
will be done by a method similar to that described in Hosmer and Lemeshow 25 for logistic
regression models as discussed above. The same set of potential covariates as described in
the safety analysis above will be screened. Models will be fit for each possible covariate
to include treatment, the covariate, and the interaction between the covariate and treatment
will be used. Any covariate or interaction significant at a P-value of 0.15 or lower from
the screening analysis will allow the variable to enter the backward elimination procedure
for the final model. If the interaction is significant, then both the covariate and its
interaction need to be included in the model.
The backward elimination will be done manually starting with the interaction with the
highest P-value. Interaction terms will be removed from the model first and then
non-significant main effects. Variables will be retained in the model if the P-value for
the variable or its interaction is 0.05 or less. This analysis is intended to demonstrate
that the treatment affect adjusted for possible covariates is still statistically
significant.
Secondary Effectiveness Analyses The drop out rate will be compared between the two groups
will be estimated. The rate for each group will be tabulated and computed with 95% exact
binomial confidence intervals.
The objective sleep parameters of AHI, arterial oxygen saturation nadir, time with elevated
end-tidal PCO2 and the arousal index will be presented descriptively by treatment group.
The tables will include the mean, standard deviation, median, minimum and maximum.
The subjective sleep measures of OSA 18 score, modified Epworth Sleepiness Scale, the NOSE,
and PedQL will be presented descriptively by treatment group. The tables will include the
mean, standard deviation, median, minimum and maximum.
Additional Analysis Exploratory analyses of the relationships of study variables to outcomes
will be done to determine if there is the treatment difference is particularly strong is
specific sub-groups.
Patient Accountability and Missing Data The number and proportion of patients eligible for
and compliant with each follow-up examination will be presented. Patients who withdraw from
the study will be tabulated with the reasons for the withdrawal. If the proportion of
patients withdrawn is larger than the 15% from either arm, an analysis of the demographic
and prognostic characteristics will be made between patients who withdraw and those who
remain in the study. For continuous variables, parametric or non-parametric analysis of
variance will be used. For categorical variables, Chi-square or Fisher's exact test will be
applied.
The evaluation of withdrawn patients presents a special concern. All clinical studies
analyze the results based on the evaluable patients, i.e., those who complete the study.
Because withdrawn patients do not have final data, they present a problem. The statistical
community 26-29 recommends that multiple sensitivity analyses be conducted to determine the
robustness of the result in patients who complete the study. The intention of these
analyses is to demonstrate that the results obtained from the evaluable patients are not
biased.
As a result, sensitivity analyses using multiple imputation analyses will be conducted to
evaluate the robustness of the study result accounting for missing observations. The
imputation will be a non-parametric multiple imputation in which patients withdrawn from the
study will be randomly assigned outcomes by grouping on demographic and prognostic
characteristics including treatment assignment maintaining masking, matching the
characteristics to the withdrawn patients, and randomly selecting the result for the missed
observation from the results for patients with similar characteristics by method such as
"hot deck" imputation or imputation by regression 30. All imputations will be stochastic
imputations to preserve the variability of the imputed value. Also, imputations will be
done in a manner that is consistent with the assumptions of multiple imputation theory
including missing at random to the extent possible. If the missing at random assumption is
clearly violated by the data, other procedures including selection modeling and pattern
mixture modeling will be considered.
Detailed Analysis Plan Prior to database lock a detailed statistical analysis plan will be
written to provide a detailed description of the statistical analysis to be used in the
final analysis. This plan will incorporate lists of relevant clinical variables to be
tested as possible covariates and will incorporate any protocol changes that would affect
the analysis.
Statistical Software
The statistical analyses will be done using SAS version 9.1 or later, StatXact Version 7 or
later, and Systat 10 or later. Each of these software packages provides special features
that will be exploited to provide a comprehensive analysis with excellent graphics support. |