Nasal EPAP for Stroke Patients With Sleep Apnea: a Pilot Trial.: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Nasal EPAP for Stroke Patients With Sleep Apnea: a Pilot Trial. (NCT01703663), a randomized na study. The primary outcome is Apnea-hypopnea Index.

2. Clinical Question

3. Trial Design

This was a randomized, unmasked crossover pilot trial sponsored by the University of Michigan. The registry record lists the study as completed, with enrollment of 40. The recorded start was 2012-11 and the primary completion was 2015-04. The phase is listed as NA.

A crossover design compares conditions within a study in which participants receive more than one condition. This can account for some between-person differences because each participant may contribute observations under multiple conditions. The record identifies an EPAP night and a control night, and gives the primary outcome timeframe as night 1 and night 2. It does not report the treatment sequence, how participants were assigned to sequences, whether there was a washout period, or how any possible period or carryover effects were handled.

The study was unmasked. The registry lists 19 participants for the posted EPAP-night summary and 19 for the control-night summary. These arm-level counts are not the same as the enrollment count, and the registry record does not explain the difference or state the number included in each analysis population. Because this is a crossover study, those arm-level summaries should not automatically be read as independent groups of participants.

The primary outcome was AHI, summarized by median and interquartile range. The trial description says that the prespecified primary absolute treatment effect was based on a linear repeated measures model. The posted analysis is described as ANOVA. Additional details about sequence allocation, model specification, covariate adjustment, and missing-observation handling are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

First Treatment Period

Second Treatment Period

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: All subjects who completed both nights

6. Primary Endpoint

The registry lists the primary outcome as Apnea-hypopnea Index, measured over night 1 and night 2.

The primary endpoint was the apnea-hypopnea index (AHI), expressed as apneas plus hypopneas per hour. The registry description defines AHI as the sum of apneas and hypopneas divided by hours of sleep, with sleep time based on actigraphy. It is intended to quantify the frequency of these breathing events during sleep, making it a direct measure of the sleep apnea severity targeted by the trial.

The registry record gives typical categories of 5-15/hr for mild, 15-30/hr for moderate, and >=30/h for severe AHI. Such categories can make a continuous measure easier to describe, but categorization discards information about the degree of difference between values and can leave people near category boundaries difficult to distinguish. The posted arm summaries use medians and interquartile ranges rather than category counts.

The record identifies night 1 and night 2 as the outcome timeframe. It reports an EPAP-night median of 17.1, with an interquartile range of 13.2 to 26.5, and a control-night median of 25.2, with an interquartile range of 14.2 to 49.6. The trial summary also describes a hypothesis about oxygen saturation, but no oxygen-saturation result is posted as a secondary outcome.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a crossover design mean?

In a crossover design, participants receive or are observed under more than one study condition. The comparison can use each participant as their own reference, which may reduce the influence of stable differences between people. However, observations from the same participant are related, not independent. The analysis should account for that pairing or repeated measurement structure. The registry identifies this trial as crossover and describes an EPAP night and a control night. It does not report the assigned sequences, the sequence-allocation procedure, or how the analysis handled period or carryover effects. Those details matter because outcomes may differ by night or may be influenced by the condition experienced earlier.

Why report medians and interquartile ranges?

The posted AHI summaries use medians and interquartile ranges. A median describes the central point of an ordered set of observations: half the values are at or below it and half are at or above it. The interquartile range describes the middle half of the observations, from the lower quartile to the upper quartile. These summaries are often useful when measurements are skewed or include unusually high values, because they are less sensitive to extremes than a mean and standard deviation. The EPAP-night summary is a median of 17.1 with an interquartile range from 13.2 to 26.5; the control-night summary is a median of 25.2 with an interquartile range from 14.2 to 49.6. Overlap between interquartile ranges is not itself a statistical test, and these summaries do not supply the paired participant-level differences.

What does the posted ANOVA result say?

The registry reports an ANOVA result for an absolute difference. The estimate is -5.73, with a 95% confidence interval from -14.4 to 2.97 and p = 0.183. In general, an estimated difference expresses a contrast between conditions, while its confidence interval describes uncertainty about the contrast under the analysis model. A confidence interval that includes zero is compatible with no difference as well as with possible differences in either direction. It should not be interpreted as proof that the conditions have the same effect.

A p-value is a measure of how unusual data at least as incompatible with a specified null hypothesis would be if that hypothesis and the statistical model were true. It is not the probability that the null hypothesis is true, and it is not a measure of the size or clinical importance of an effect. Here, the reported p-value is 0.183. The registry does not report a prespecified significance threshold, so the interpretation should rest on the estimate, its interval, and the study context rather than an unstated decision rule.

Why does the model description matter?

The endpoint description says the prespecified primary absolute treatment effect was based on a linear repeated measures model, whereas the reported analysis method is listed as ANOVA. Repeated measures methods account for correlations among observations from the same participant. ANOVA is a broad label for methods that compare variability across conditions; by itself, it does not establish which model terms or correlation structure were used. The registry record does not report a full statistical analysis plan or enough model details to reconcile the two descriptions. It is therefore appropriate to describe the posted result as reported, without inferring adjustments or assumptions that are not documented.

How should precision be judged?

Precision concerns how narrowly the data and model locate the effect. The confidence interval communicates this uncertainty more directly than the point estimate alone. In general, a wider interval leaves a broader range of effects compatible with the analysis. The interval in this trial extends from -14.4 to 2.97 and includes zero, so the result remains uncertain about the direction and size of the absolute difference. A pilot trial can provide an estimate and inform future research, but its designation as a pilot does not itself establish the precision or validity of the estimate.

What can be said about sample counts and missing observations?

The registry lists enrollment of 40 and arm-summary counts of 19 for each night. It does not report why these counts differ, whether every participant contributed data under both conditions, how missing outcome observations were handled, or which participants were included in the model. In crossover studies, those details affect how much paired information supports the comparison. Without them, the analysis population and the consequences of missing observations cannot be assessed from the registry record.

10. Analyses Reported by the Sponsor

The registry reports an ANOVA analysis with an absolute difference of -5.73, a 95% confidence interval from -14.4 to 2.97, and p = 0.183. The primary outcome description separately states that the prespecified primary absolute treatment effect was based on a linear repeated measures model. The record does not provide enough detail to determine whether the reported ANOVA corresponds to that prespecified model or to clarify their relationship.

No separate recalculated estimate is reported alongside these results in the registry record, and no discrepancy is flagged. The sponsor's posted analysis is therefore the inferential result to use when describing the trial. The arm medians and interquartile ranges provide descriptive information, but they do not reproduce the model-based contrast, especially in a crossover study where within-participant pairing matters.

The confidence interval includes zero and spans values on both sides of it. This indicates uncertainty about the direction of the estimated difference; it does not demonstrate equivalence or establish that no difference exists. The reported p-value should likewise be interpreted as part of the model-based evidence, not as a probability that the intervention has or lacks an effect.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists one outcome measure and posts one analysis, for AHI. The study summary also states a hypothesis about mean oxygen saturation, but no secondary outcome result is posted. The record does not explain whether oxygen saturation was analyzed, whether other outcomes were considered, or how the primary and additional hypotheses were handled statistically. A multiplicity adjustment or testing hierarchy is not reported in the registry record.

When several outcomes or comparisons are tested, the chance of obtaining at least one apparently notable result can increase. Whether and how that issue applies depends on the planned analyses and their relationship; those details are not reported. It would not be appropriate to assume an adjustment, or to infer that no adjustment was needed, from the single posted analysis alone.

The enrollment count is 40, while the posted summaries list 19 for the EPAP night and 19 for the control night. The registry record does not explain the counts, identify the analysis population, or state whether all participants contributed paired observations. It also does not describe how missing outcome data were handled, whether any observations were excluded, or whether an intention-to-treat approach was used.

These omissions matter for a crossover comparison. The number of complete pairs, the handling of partial records, and the assumptions used for missingness can affect the estimate and its uncertainty. The record does not report sensitivity analyses or a detailed missing-data plan. The trial's summary states the primary question and the outcome measure, but a fuller statistical analysis plan, including multiplicity and analysis-population definitions, is not reported in the registry record.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

The posted adverse-event figures list 19 participants at risk for serious adverse events on the EPAP night, with 0 affected, and 19 at risk on the control night, also with 0 affected. The listed counts for other adverse events are likewise 0 among 19 at risk for each night. These figures describe the posted registry counts; they do not establish that no events occurred outside the recorded categories or observation period.

The death counts are not reported in the registry record. A missing death count should not be read as zero. The record also gives no serious adverse event terms, so there are no specific serious-event descriptions to interpret from the posted results.

These safety counts should be read in light of the pilot trial size and the registry's limited detail about event collection and follow-up. The record does not report definitions, ascertainment procedures, exposure duration beyond the night labels, or a broader safety analysis. The zero counts are therefore a description of what was posted, not evidence that a risk is absent.

13. What the Result Does, and Does Not, Mean

The posted primary result estimates an absolute AHI difference of -5.73, with a 95% confidence interval from -14.4 to 2.97 and p = 0.183. The interval includes zero and allows values on either side of it, so the estimate is not precise enough to establish the direction of the difference. The arm-level medians are 17.1 on the EPAP night and 25.2 on the control night, but those descriptive summaries do not replace the crossover analysis.

The result does not establish that the conditions are equivalent, that the true difference is zero, or that either condition is superior. A non-significant result is not proof of no effect. Nor does the posted analysis establish whether a difference would be clinically important. The record does not provide a threshold for a clinically meaningful AHI difference or a detailed basis for interpreting the observed estimate in clinical terms.

The trial focused on AHI across night 1 and night 2 in patients with recent ischemic stroke and sleep apnea. The record also describes an oxygen-saturation hypothesis, but posts no secondary result for that outcome. It does not provide the analysis population, sequence details, or missing-data approach needed to fully assess the scope of the primary comparison.

As a pilot trial, the study contributes an estimate and a description of the comparison, rather than a conclusive answer about effects beyond the observed study nights. The enrollment count was 40, while the posted summaries list 19 observations for each night; the reason is not reported in the registry record. Conclusions should remain tied to the posted AHI analysis and its uncertainty, without extending them to unreported outcomes or populations.

14. Why This Trial Matters Statistically

This trial raises statistical issues that are especially relevant to crossover studies. The same participant may contribute observations under both conditions, which means the comparison depends on the relationship between those observations. A suitable analysis should account for that dependence. The record identifies a crossover design and states that the prespecified primary effect was based on a linear repeated measures model, but the reported method is ANOVA. Without fuller model details, readers cannot determine exactly how the repeated observations were handled.

The contrast between descriptive summaries and an inferential estimate is also important. The posted medians and interquartile ranges characterize AHI values on each night, while the reported absolute difference and confidence interval express the model-based comparison. In a crossover trial, the distribution of paired differences can matter; separate summaries alone do not recover that distribution.

The confidence interval conveys uncertainty that a single estimate or p-value cannot. Here it includes zero and extends to both negative and positive values, leaving the direction unresolved by the posted analysis. This is a useful reminder that the absence of clear evidence for a difference is not evidence of equivalence.

Finally, the difference between the enrollment count and the counts listed for each night illustrates why analysis populations and missing-data handling matter. The registry does not explain the counts, sequence details, or exclusions. Those omissions limit assessment of how the planned crossover comparison relates to the posted result. The trial is statistically informative as an example of why model specification, paired data, and uncertainty need to be read together.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The registry record gives limited detail about the design implementation. Although the study is identified as randomized and crossover, it does not describe the treatment sequences, the randomization procedure, any washout period, or how period and carryover effects were considered. Masking is listed as none. The control night is not described beyond its label.

The record lists enrollment of 40, but the arm summaries include 19 participants for each night. It does not explain these counts, report how many participants completed both nights, identify the population analyzed, or describe missing-data handling. These gaps limit evaluation of the completeness of the crossover comparison and the possible impact of missing observations.

The endpoint summary provides medians and interquartile ranges, while the inferential analysis is described as ANOVA. The primary outcome description says the prespecified absolute effect was based on a linear repeated measures model. The registry does not provide enough detail to reconcile those statements or assess the model assumptions, terms, and treatment of within-participant correlation.

The registry posts one primary outcome and one analysis, but does not report a full statistical plan, multiplicity strategy, or secondary results. Although oxygen saturation appears in the study hypothesis, no secondary result is posted. The safety record gives zero serious and other adverse events in the listed groups, but death counts are not reported and event-collection details are absent. These are limits of what can be inferred from the registry record, not evidence that unreported methods or outcomes were absent.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Nasal EPAP for Stroke Patients With Sleep Apnea: a Pilot Trial. — NCT01703663.
• Registry record last updated 2016-10-13; status last verified 2016-08.
• Results first posted 2016-10-13.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

NCT01703663 is a completed randomized, unmasked crossover pilot trial sponsored by the University of Michigan. It enrolled 40 and studied nasal EPAP, identified as Provent Sleep Apnea Therapy, against a control night among patients with recent ischemic stroke and sleep apnea. The primary endpoint was AHI, defined using apneas and hypopneas per hour of sleep based on actigraphy.

The posted summaries list 19 participants for each night. The reported ANOVA estimate is an absolute difference of -5.73, with a 95% confidence interval from -14.4 to 2.97 and p = 0.183. The interval includes zero, leaving uncertainty about the direction of the difference. No secondary result is posted; details about sequence, analysis populations, and missing-data handling are not reported in the registry record.