Safety/Efficacy Study With Armodafinil (CEP-10953) in Treatment of Excessive Sleepiness Associated With Chronic SWSD: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Safety/Efficacy Study With Armodafinil (CEP-10953) in Treatment of Excessive Sleepiness Associated With Chronic SWSD (NCT00080288), a randomized phase 3 study. The primary outcome is Multiple Sleep Latency Test (MSLT).

2. Clinical Question

3. Trial Design

The study was randomized, double-blind, placebo-controlled, and parallel-group, and was identified as a phase 3 trial. It compared armodafinil 150 mg/day with placebo in adults with excessive sleepiness associated with chronic shift work sleep disorder. The registry lists enrollment as 254 and the study status as completed. The stated study duration was 12 weeks.

The primary measure was the Multiple Sleep Latency Test (MSLT), an objective assessment of sleepiness. The registry describes five scheduled MSLT naps, each with a maximum duration of 20 minutes, conducted at midnight, 0200, 0400, 0600, and 0800. Assessments took place at weeks 4, 8, and 12. The primary efficacy variable was described as mean change from baseline in MSLT sleep latency at week 12, or at the last postbaseline visit.

The registry separately describes a safety analysis set of 245 participants and a full analysis set of 216 participants. It notes that 9 participants withdrew after randomization but before receiving study drug, and that 29 participants who had withdrawn were non-evaluable for efficacy. The posted primary outcome groups contain 112 participants receiving armodafinil and 104 receiving placebo. The registry record does not report the original sample-size justification or the detailed procedures used to address withdrawals and missing outcome measurements.

Randomization and masking are design features intended to support a fair comparison and reduce the influence of expectations or other systematic differences. In general, random assignment helps balance measured and unmeasured characteristics on average, while double masking can limit behavior or assessment influenced by knowledge of treatment assignment. These design features do not, by themselves, resolve missing-data concerns or establish how every analysis was conducted.

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.

Before assignment: 9 participants (3 female ; 6 male) withdrew after randomization but prior to receiving study drug. These 9 patients are included in the Participant Flow table below.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Safety Analysis set of 245 total patients: 9 participants withdrew after randomization but prior to receiving study drug Full Analysis set of 216 total patients: 29 patients that had withdrawn from the study were non-evaluable for efficacy.

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.

6. Primary Endpoint

The registry lists the primary outcome as Multiple Sleep Latency Test (MSLT), measured over up to 12 weeks.

The primary endpoint was the Multiple Sleep Latency Test (MSLT) sleep latency, measured in minutes. The MSLT assesses how long a participant can remain awake during scheduled nap opportunities; the registry states that longer latencies indicate a greater ability to remain awake. The assessment used scheduled naps of up to 20 minutes and was conducted at weeks 4, 8, and 12.

The primary efficacy variable is described as mean change from baseline in sleep latency at week 12, or at the last postbaseline visit. A change-from-baseline endpoint compares a participant's measurement after treatment with that participant's baseline measurement, then summarizes those changes within treatment groups. The posted outcome gives a mean of 3.1 minutes for the armodafinil group and 0.4 minutes for the placebo group. The registry's reported analysis labels its result as a mean difference in final values, so the precise relationship between that label and the outcome description should be kept in view when interpreting the estimate.

MSLT is used because it provides an objective measure related to the ability to remain awake, rather than relying only on a participant's account of sleepiness. Like any measurement, it captures a particular aspect of the study question and does not represent every dimension of symptoms or daily functioning. The registry describes CGI-C ratings in the study objective, but no secondary outcomes are posted in the record.

7. Results Posted in the Registry

• Armodafinil 150 mg/Day: mean 3.10 versus 0.40 on control; difference 2.70 (95% CI 1.71 to 3.69), standardized effect d = 0.71; z-test p = <0.001.

8. Statistical Interpretation

9. Statistical Methods Explained

The principal reported analysis is an analysis of covariance (ANCOVA), with a mean difference of 2.7 minutes, a 95% confidence interval from 1.67 to 3.69, and p < 0.001. The registry labels this as a mean difference in final values. ANCOVA is a regression-based method that compares groups while accounting for one or more covariates. In general, adjustment for a baseline measurement can improve precision when the baseline value is related to the follow-up outcome. The registry record does not report the covariates included in this ANCOVA or the model specification, so the details of adjustment cannot be reconstructed from the reported result.

A separate simple re-calculation uses the posted group means and standard deviations. It gives a mean difference of 2.7 minutes, with a 95% confidence interval from 1.707 to 3.693 and p < 0.001. This is an unadjusted comparison of summary statistics, not a reproduction of the sponsor's ANCOVA. It uses the posted group sizes of 112 and 104, group means of 3.1 and 0.4, and standard deviations of 4.46 and 2.87. These statistics describe variation among participants within each group; they are not measures of uncertainty in the group difference itself.

What does a mean difference describe?

A mean difference expresses the separation between two group averages in the endpoint's original units. Here, the unit is minutes. It can be easier to interpret in context than a standardized measure because it stays on the MSLT scale. In general, a mean difference does not show the range of individual participant responses, and the group average may not describe any one participant's experience. The registry record does not report the distribution of individual changes or a clinically meaningful threshold for interpreting a difference in minutes.

Why are confidence intervals useful?

A confidence interval communicates uncertainty around an estimate. In general, a narrower interval indicates greater precision under the analysis assumptions, while a wider interval indicates less precision. The sponsor-reported interval and the interval from the unadjusted re-calculation are both centered around an estimated difference of 2.7 minutes, but they are based on different analyses. Their slight differences are expected to reflect that distinction; they should not be combined or treated as independent confirmations.

What does a p-value say?

A p-value describes how compatible the observed data are with a specified null hypothesis under the statistical model. In general, it is not the probability that the null hypothesis is true, and it does not measure the size or importance of an effect. The sponsor-reported p-value is < 0.001, and the unadjusted re-calculation reports p < 0.001. The reported thresholds are not identical, so they should be quoted with their respective analyses rather than presented as if they were one result.

Why does analysis population matter?

The registry describes a safety analysis set of 245 participants and a full analysis set of 216. The record says that 9 participants withdrew after randomization before receiving study drug and that 29 withdrawn participants were non-evaluable for efficacy. An analysis population determines who contributes to an estimate. In general, excluding participants with unavailable outcomes can affect the interpretation if missingness is related to treatment or outcome. The registry record does not report the reasons for all withdrawals, the missing-data method, or sensitivity analyses, so the implications of the exclusions cannot be quantified from the posted information.

Why can variation matter?

The posted standard deviations are 4.46 minutes for armodafinil and 2.87 minutes for placebo. In general, greater variability makes it harder to estimate a group average precisely, all else equal. The unadjusted re-calculation uses a pooled standard deviation of 3.779 minutes to express a standardized difference of 0.714. Standardization expresses a difference relative to a measure of within-group variation, but it is less directly interpretable in minutes. It does not replace the original-scale estimate or its confidence interval.

In general, over-dispersion matters most for count outcomes when variability exceeds what a simple model assumes; that issue is not the central one for this continuous MSLT outcome. The main interpretive concerns in this record are the distinction between the adjusted sponsor analysis and the unadjusted re-calculation, the endpoint description versus analysis label, and the incomplete reporting of missing-data procedures and model details.

10. Confidence Intervals and Precision

The sponsor-reported ANCOVA estimate is a mean difference of 2.7 minutes, with a 95% confidence interval from 1.67 to 3.69. The simple unadjusted re-calculation gives the same stated mean difference of 2.7 minutes, with a 95% confidence interval from 1.707 to 3.693. These intervals describe uncertainty under their respective analyses. They are not identical because one is the sponsor-reported ANCOVA result and the other is a re-calculation from posted group summary statistics.

In general, a 95% confidence interval is produced by a procedure that, under its assumptions and repeated use in comparable settings, would cover the underlying parameter in 95% of such intervals. For a single study, it is best read as a range of effect estimates compatible with the data and method, rather than as a probability statement about the fixed true effect. The interval also gives more information than a significance threshold alone: it conveys both the direction and the range of plausible magnitudes under the model.

Both intervals are above zero, consistent with the reported test results against a null difference of zero. That observation describes statistical evidence in this analysis; it does not establish clinical importance. The registry record does not report a threshold for a meaningful change in MSLT latency, so the interval cannot be classified against such a threshold. Precision is also conditional on the analysis population and assumptions. The record does not describe the ANCOVA covariates or missing-data procedures, and the posted analysis population excludes participants considered non-evaluable for efficacy.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures give participants per arm for 80% power and participants per arm for 90% power. They are not total events: this is a continuous outcome, and the figures are participant counts for each of the two arms. For 100% of the observed effect, the illustration gives 31 participants per arm for 80% power and 42 participants per arm for 90% power. For 75% of the observed effect, it gives 55 and 74 participants per arm, respectively. For 50% of the observed effect, it gives 124 and 165 participants per arm, respectively.

The calculations are a planning illustration based on the observed effect, not the trial's original design justification. The stated assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; each planning scenario uses the stated fraction of the observed effect. In general, the 80% and 90% power columns represent the planned chance of detecting the specified effect at the stated significance level if the assumptions hold and the effect is real. The higher power target requires more participants per arm at each listed effect size.

Smaller effects need more participants to distinguish them from random variation, which is why the planning figures increase as the assumed effect decreases. These values are not observed power. The registry record does not report the original sample-size justification or the assumptions used to plan the trial, so the illustration should not be read as evidence that the study was designed around these particular scenarios.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = 2.700; pooled standard deviation σ = 3.779.
2. n = 2 × (1.96 + 0.842)² × 3.779² / 2.700² = 30.8, rounded up to 31 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
delta <- 2.7                      # observed difference in means
sigma <- 3.77916589471977                     # pooled standard deviation
n_per_arm <- function(power, frac = 1) {
  ceiling(round(2 * (z_a + qnorm(power))^2 * sigma^2 / (delta * frac)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (t-based, so very slightly larger)
cat("power.t.test:", round(power.t.test(delta = delta, sd = sigma, power = 0.80)$n, 1), "\n")

# Simulation: exact two-sample t-test using simulated means and variances
set.seed(80288)
n    <- n_per_arm(0.80)
reps <- 20000
m_diff <- rnorm(reps, delta, sigma * sqrt(2 / n))             # difference in sample means
s2     <- sigma^2 * rchisq(reps, 2 * (n - 1)) / (2 * (n - 1))   # pooled sample variance
tstat  <- m_diff / sqrt(s2 * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(tstat) > qt(0.975, 2 * (n - 1))), reps))

Output from running the code

100% of observed effect: n per arm = 31 (80% power), 42 (90% power)
 75% of observed effect: n per arm = 55 (80% power), 74 (90% power)
 50% of observed effect: n per arm = 124 (80% power), 165 (90% power)
power.t.test: 31.7 
Simulated power at n = 31 per arm: 0.790 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Analyses Reported by the Sponsor

The sponsor's reported analysis is an ANCOVA with a mean difference of 2.7 minutes, a 95% confidence interval from 1.67 to 3.69, and p < 0.001. It is labeled as a mean difference in final values. The separate re-calculation from the posted means, standard deviations, and group sizes is unadjusted; it gives a mean difference of 2.7 minutes, a 95% confidence interval from 1.707 to 3.693, and p < 0.001.

The numerical estimates are close, but the methods differ. ANCOVA can incorporate covariates, while the simple re-calculation compares the posted summaries without that adjustment. The registry record does not identify the ANCOVA covariates or provide enough model detail to recreate the sponsor's analysis. The re-calculation therefore serves as a transparent check based on the posted summary statistics, not as a replacement for the sponsor's analysis.

The registry describes the primary efficacy variable as mean change from baseline at week 12 or the last postbaseline visit, whereas the sponsor analysis is labeled as using final values. The record does not explain this wording difference. Both labels should be retained when describing the reported and re-calculated results, rather than assuming the analyses use identical outcome definitions.

14. Multiplicity, Missing Data and Analysis Populations

The registry's study objective mentions both MSLT sleep latency and Clinical Global Impression of Change (CGI-C) ratings. The primary outcome entry describes MSLT as the primary efficacy variable, and the record lists two outcome measures and two analyses. However, no secondary outcomes are posted. The record does not state how many hypotheses were formally tested, which outcomes were designated as primary or secondary beyond the MSLT description, or whether any adjustment for multiple comparisons was planned or applied. Multiplicity procedures are therefore not reported in the registry record.

In general, testing multiple outcomes or time points can increase the chance of at least one apparently unusual result even when none of the underlying effects is real. A prespecified testing hierarchy or adjustment can address that issue, but the record does not provide such a plan. The reported p-values should consequently be read as the values attached to the posted analyses, without assuming an unreported correction or claiming that no correction was used.

There are also two distinct analysis populations in the record. The safety analysis set contains 245 participants; 9 withdrew after randomization but before receiving study drug. The full analysis set contains 216 participants; 29 who had withdrawn were non-evaluable for efficacy. The record does not report exactly how missing measurements were handled, whether participants were analyzed according to randomized assignment, or how the population used for the ANCOVA maps to these definitions. The outcome groups list 112 participants for armodafinil and 104 for placebo, but the registry does not provide further detail about the connection between those counts and the analysis-set definitions.

In general, missing-data methods can affect estimates when participants with missing outcomes differ systematically from those with observed outcomes. The reasons for withdrawal, timing and amount of missing outcome data, imputation procedures, and sensitivity analyses are not reported in the registry record. Without those details, it is not possible to assess how robust the efficacy result is to alternative assumptions about missing outcomes. This limitation does not establish that the reported analysis is incorrect; it narrows what can be concluded about its handling of unavailable data.

15. 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.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: depression suicidal (0.8%), meningitis viral (0.8%).

The serious adverse event figures list 1 affected participant among 123 at risk in the armodafinil group and 1 among 122 at risk in the placebo group. The registry also lists depression suicidal and meningitis viral among the top serious event terms, each at 0.8%. Those terms are listed without an arm-specific breakdown in the record, so they should not be attributed to a particular group based on this information alone.

These are counts and percentages recorded for the safety population, not estimates of comparative risk adjusted for follow-up or other factors. The number of participants at risk indicates the group denominator used in the safety listing. The record does not provide a statistical comparison of serious adverse events, event timing, or enough detail to determine whether multiple events occurred in the same participant.

Other adverse events are listed for 28 participants in the armodafinil group and 18 in the placebo group, but those figures are not the serious-event counts. Death counts are not reported in the registry record. Safety results should be interpreted alongside the limited detail available: the listing does not establish that all possible events are represented equally or that differences between groups are statistically distinguishable.

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

The posted MSLT results show a mean of 3.1 minutes in the armodafinil group and 0.4 minutes in the placebo group. The sponsor-reported ANCOVA gives a mean difference of 2.7 minutes, with a 95% confidence interval from 1.67 to 3.69 and p < 0.001. The simple unadjusted re-calculation gives a 95% confidence interval from 1.707 to 3.693 and p < 0.001. These are statistical summaries of the comparison for the reported endpoint and analysis populations.

The registry describes longer MSLT latency as indicating a greater ability to remain awake. That makes the endpoint relevant to the study question about excessive sleepiness, but the endpoint is still a measurement of a specific aspect of sleepiness. The record does not provide a threshold for a clinically meaningful difference, so the size of the estimated difference cannot be classified as clinically important from the registry information alone.

The result does not establish what participants experienced in every aspect of daily life, how effects differed across participant subgroups, or whether other outcomes changed. Although CGI-C ratings appear in the study objective, no secondary outcomes are posted. The result also does not show how the estimate would change under alternative methods for handling missing data; those methods and related sensitivity analyses are not reported in the registry record.

In general, a statistically clear difference and a clinically meaningful difference are related but distinct concepts. Statistical inference describes how the data compare with a model of no difference and how uncertain the estimated effect is. Clinical importance requires context, including a meaningful-change threshold and information about outcomes that matter to patients. The registry record does not report such a threshold. These findings should therefore be read as a comparison on the posted MSLT measure, not as a complete account of benefits, risks, or broader outcomes.

17. Why This Trial Matters Statistically

This trial illustrates why the method attached to an effect estimate matters. The sponsor-reported result is an ANCOVA, while the simple re-calculation uses group-level summary statistics without adjustment. Their mean differences are both 2.7 minutes, but their confidence limits and p-value thresholds differ slightly. Reporting the estimates with their methods avoids treating different analyses as though they were identical.

The record also illustrates the importance of distinguishing endpoint descriptions from analysis labels. The primary outcome description refers to change from baseline at week 12 or the last postbaseline visit, while the reported analysis is labeled as a difference in final values. The registry does not explain this wording difference. Clear reporting of the analyzed outcome and model would help readers understand exactly what the estimate represents.

Analysis populations are another central statistical issue. The record describes a safety set of 245 and a full analysis set of 216, and notes withdrawals and non-evaluable efficacy participants. Such distinctions affect the scope of an estimate and raise questions about missing outcomes. The registry does not report the missing-data method or sensitivity analyses, which limits assessment of how the analysis handles those participants.

Finally, the study objective includes CGI-C, but no secondary outcomes are posted, and the record does not describe multiplicity procedures. Statistical interpretation is strongest when outcome hierarchy, analysis populations, model details, and missing-data methods are transparent. This record supports a focused reading of the MSLT comparison, while leaving important details of the broader analysis plan unreported.

18. 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 leaves several trial-specific details unclear. The primary outcome description defines the efficacy variable as change from baseline at week 12 or the last postbaseline visit, while the sponsor analysis is labeled as a mean difference in final values. The record does not explain how those descriptions relate or specify which outcome measurements entered the ANCOVA.

The ANCOVA covariates and model specification are not reported. The analysis populations are described, but the record does not state precisely which population was used for the sponsor's primary analysis or how the listed outcome group sizes relate to the full analysis set. The handling of missing outcomes, reasons for withdrawal, and sensitivity analyses are also not reported in the registry record.

Although CGI-C ratings are named in the study objective, no secondary outcomes are posted. The record does not report a multiplicity plan, the original sample-size justification, or a clinically meaningful threshold for the MSLT endpoint. These omissions limit interpretation of the broader outcome plan, the design assumptions, and the clinical meaning of the estimated difference.

The safety listing is limited: it gives participant counts for serious adverse events and names two top serious event terms, but does not provide arm-specific detail for those terms, event timing, or a statistical comparison. Death counts are not reported in the registry record. These reporting limits do not determine the direction of any unreported result; they identify what cannot be assessed from the posted information.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Safety/Efficacy Study With Armodafinil (CEP-10953) in Treatment of Excessive Sleepiness Associated With Chronic SWSD — NCT00080288.
• Registry record last updated 2013-07-19; status last verified 2013-07.
• Results first posted 2009-09-23.
• 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.

22. Record Summary

The registry describes a completed, randomized, double-blind, placebo-controlled, parallel-group phase 3 study of armodafinil 150 mg/day for adults with excessive sleepiness associated with chronic shift work sleep disorder. Enrollment was 254, and the stated study duration was up to 12 weeks.

The primary outcome was MSLT sleep latency in minutes. Posted means were 3.1 for armodafinil and 0.4 for placebo. The sponsor-reported ANCOVA estimated a mean difference of 2.7 minutes, with a 95% confidence interval from 1.67 to 3.69 and p < 0.001. A simple unadjusted re-calculation gave a 95% confidence interval from 1.707 to 3.693 and p < 0.001.

The registry describes separate safety and efficacy analysis populations, but does not report the missing-data procedures, detailed ANCOVA specification, or original sample-size justification. No secondary outcomes are posted.