Clinical Outcomes in Narcolepsy and Cataplexy: An Evaluation of Reboxetine Treatment (CONCERT): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Clinical Outcomes in Narcolepsy and Cataplexy: An Evaluation of Reboxetine Treatment (CONCERT) (NCT03881852), a randomized phase 2 study. The primary outcome is Change From Baseline in the Weekly Average Total Number of Cataplexy Attacks.

2. Clinical Question

The primary endpoint focuses on the frequency of cataplexy attacks, a symptom specifically named in the trial’s title and eligibility description. The registry describes the endpoint as a change from baseline and states that a positive change indicates improvement. The posted treatment estimates are least-squares means, with standard errors reported for each treatment. A separate sponsor analysis using a mixed models analysis reports a superiority result, but the registry record does not provide its estimate or confidence interval.

The clinical question is limited to the measured endpoint and the comparison of treatment periods in this crossover study. It does not, by itself, establish what happened to other narcolepsy symptoms, longer-term outcomes, or outcomes outside the trial population. No secondary outcome results are posted in the registry record.

3. Trial Design

CONCERT is a Phase 2, randomized, triple-masked, placebo-controlled, multicenter crossover trial. The registry lists enrollment of 21 and a completed status. The recorded start date is 2019-04-05, and the primary completion date is 2019-11-27. The trial’s official title describes a three-week crossover study; its brief description specifies two successive three-week treatment periods.

Participants were assigned in a 1:1 ratio to one of two sequences: placebo for three weeks followed by AXS-12 for three weeks, or AXS-12 for three weeks followed by placebo for three weeks. AXS-12 was described as up to 10 mg daily. The record identifies the study as randomized and triple-masked; it does not give further details about who was masked or how masking was maintained. Specific information about the randomization procedure, any washout period, and the handling of period or sequence effects is not reported in the registry record.

A crossover design means that participants receive both treatments in different periods, allowing treatment comparisons to draw on observations from the same participants. General statistical interpretation of such a design depends on accounting for the pairing of observations and potential period or carryover effects. The registry reports a mixed models analysis for the sponsor’s treatment comparison, but does not describe the model specification or how these design features were handled.

The outcome summaries list 21 observations for AXS-12 and 20 for placebo. The registry does not explain why these treatment-specific counts differ. The counts should not be read as evidence that separate parallel groups were randomized, since the trial is identified as a crossover study.

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.

Overall Study

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: Safety Population

6. Primary Endpoint

The registry lists the primary outcome as Change From Baseline in the Weekly Average Total Number of Cataplexy Attacks, measured over 2 weeks.

The primary endpoint is the change from baseline in the weekly average total number of cataplexy attacks, measured over a two-week timeframe. It is a continuous measure: each participant’s result represents a quantity that can take a range of values, rather than simply an event/no-event outcome. The registry reports the treatment estimates as least-squares means and gives the unit as weekly cataplexy attacks. It explicitly states that a positive change is indicative of improvement.

Counting attacks provides a way to quantify the frequency of a symptom that is central to the population described in the trial. Expressing the measure as a weekly average allows the result to represent attack frequency over a specified observation period, while change from baseline describes how the outcome shifted relative to each participant’s starting point. These are general reasons this kind of endpoint may be used; the registry record does not give a separate rationale for selecting it.

The two-week timeframe is the stated endpoint assessment period, not the duration of each treatment period. The treatment periods are described as three weeks. The record does not provide further detail about the diary or measurement procedure, how baseline was defined, or how observations within the assessment period were combined beyond calling the measure a weekly average.

The endpoint is specific to cataplexy frequency. Although the trial description also mentions excessive daytime sleepiness and other narcolepsy symptoms, no secondary outcome results are posted in the registry record. The primary result therefore should not be interpreted as a summary of every symptom or aspect of the condition.

7. Results Posted in the Registry

• AXS-12 (Reboxetine): mean 13.00 versus -0.28 on control; difference 13.28 (95% CI 3.58 to 22.98), standardized effect d = 0.84; z-test p = 0.007.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous, and the registry reports least-squares means for the treatment periods. A least-squares mean is a model-based estimate of an average. Unlike a simple arithmetic mean, it can reflect adjustment for terms included in the fitted model. The registry does not state which terms were included in the sponsor’s model, so the meaning of any particular adjustment cannot be determined from the record. The posted treatment estimates are 13.0 for AXS-12 and -0.28 for placebo, each with a standard error of 3.5.

A standard error describes how precisely an estimated mean is determined under the analysis used. It is not the same as the standard deviation, which describes variation among observed outcomes. A smaller standard error generally indicates a more precisely estimated mean, while a larger standard error indicates greater uncertainty. The registry gives standard errors for the two treatment estimates but does not provide the underlying participant-level outcome distributions.

The posted mean difference is 13.28. Its confidence interval, from 3.579 to 22.98, and the unadjusted p-value of 0.007 are recalculated from the posted summary statistics. The calculation uses an approximate pooled standard deviation; the record marks this standard deviation as approximate. A pooled standard deviation combines information about variability across treatment groups for a simple comparison. In this case, the reconstruction uses standard errors and counts to approximate that variability. It does not use the actual within-participant covariance, which is important in crossover data.

In a crossover design, each participant receives both treatments in a sequence. General statistical reasoning recognizes that measurements from the same participant are often related. A model that accounts for this relationship can use the paired structure and may estimate treatment differences more efficiently than an analysis that treats treatment-period observations as independent. The sponsor reports a mixed models analysis, a general class of models that can represent relationships among observations and incorporate fixed or random effects. However, the model specification, covariance structure, period terms, sequence terms, and any carryover assessment are not reported in the registry record. It is not possible to verify from the record which features were included.

The sponsor’s mixed models analysis reports p < 0.001 for superiority. The recalculated p-value of 0.007 is unadjusted and based on aggregate summary statistics, so the two p-values are not necessarily expected to match. They arise from analyses with different documented levels of detail and potentially different assumptions. The registry does not provide enough information to reproduce the sponsor’s model. The sponsor’s analysis is the reported trial analysis; the recalculation is a supplementary, simple comparison from the posted summaries.

In general, a confidence interval represents a range of effect values compatible with the observed data under the analysis assumptions and confidence procedure. A narrower interval indicates greater precision than a wider interval, all else equal. A confidence interval is not a guarantee that the true effect lies within its endpoints, nor does it describe the range of individual participant responses. The interval from the recalculation reflects that calculation’s assumptions, not every uncertainty related to crossover design or model specification.

Over-dispersion is general statistical language for variability greater than a model’s assumed distribution allows. When it occurs, a model can understate uncertainty if the extra variability is ignored, making standard errors and tests less reliable. The registry does not report an over-dispersion assessment, and this concept should not be taken as evidence that over-dispersion occurred in CONCERT. More generally, assumptions about outcome distributions and dependence matter because they influence standard errors, confidence intervals, and p-values.

The reported superiority classification indicates that the sponsor framed the comparison as a test for whether AXS-12 differed in the favorable direction from placebo, but the registry record does not state the exact statistical hypotheses or decision rule. The primary endpoint description supplies the direction of improvement: positive change indicates improvement. The registry does not report whether the sponsor used a one-sided or two-sided test for its own model-based result.

The registry lists one outcome measure and one analysis, and no secondary outcomes are posted. That does not establish that no other assessments were collected; it means their results are not reported in the registry record. Similarly, detailed information on missing observations, analysis populations, multiplicity procedures, interim analyses, and the original sample-size justification is not reported in the registry record. These omissions constrain what can be independently assessed about the planned analysis and the completeness of outcome reporting.

10. Confidence Intervals and Precision

The recalculated mean difference is 13.28 weekly cataplexy attacks, with a confidence interval from 3.579 to 22.98. The interval describes uncertainty around the estimated difference under the assumptions of the unadjusted recalculation. Its endpoints are separated by a substantial range of possible values, so the estimate should not be treated as exact. The whole interval is on the positive side of zero, consistent with a positive difference in the endpoint’s stated direction.

Confidence intervals are useful because they show both the direction and the precision of an estimate. A point estimate alone does not indicate how much uncertainty surrounds it. A narrower interval would generally indicate greater precision, while a wider interval would indicate less precision, assuming the same method and conditions. The interval does not capture every possible source of uncertainty: it is derived from aggregate summaries and an approximate pooled standard deviation, and it does not reproduce the sponsor’s mixed models analysis.

The standard error is 3.5 for each treatment estimate. Standard errors describe uncertainty in estimated means, not the range of values among individual participants. The sponsor’s mixed models result has no estimate or confidence interval reported in the registry record, so its precision cannot be assessed directly from that result. Its reported p-value, p < 0.001, is not a substitute for an effect estimate and interval.

Because the study is a crossover trial, precision can depend on how strongly each participant’s outcomes under the two treatments are related. The aggregate summaries do not report that within-participant relationship. The recalculated interval therefore gives a limited view of precision and should be considered alongside the sponsor’s model-based result and the registry’s reporting gaps.

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 are participants per arm, not total event counts. They are illustrations based on the observed effect, not the trial’s original sample-size design. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. Each scenario sets the effect to detect as a stated fraction of the observed effect. The trial’s own original design assumptions and sample-size justification are not reported in the registry record.

For planning at 100% of the observed effect, the illustration gives 23 participants per arm for 80% power and 30 participants per arm for 90% power. For planning at 75% of the observed effect, it gives 40 participants per arm for 80% power and 54 participants per arm for 90% power. For planning at 50% of the observed effect, it gives 90 participants per arm for 80% power and 120 participants per arm for 90% power.

In plain language, the 80% power column represents the illustrative number of participants per arm under which the assumed effect would be detected in 80% of repeated studies if the planning assumptions held. The 90% column uses a higher target probability of detecting that effect and consequently gives larger participant counts in these scenarios. The curve should be read as the general relationship between the effect size assumed for planning and the required number of participants: smaller effects need more participants to distinguish them from random variation.

These values are not observed power. They are planning illustrations calculated from the observed effect and stated assumptions. They do not establish that the original trial was designed with these targets or that its actual analysis had a particular probability of detecting an effect.

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 = 13.280; pooled standard deviation σ = 15.852.
2. n = 2 × (1.96 + 0.842)² × 15.852² / 13.280² = 22.4, rounded up to 23 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 <- 13.28                      # observed difference in means
sigma <- 15.8518784780243                     # 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(3881852)
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 = 23 (80% power), 30 (90% power)
 75% of observed effect: n per arm = 40 (80% power), 54 (90% power)
 50% of observed effect: n per arm = 90 (80% power), 120 (90% power)
power.t.test: 23.4 
Simulated power at n = 23 per arm: 0.797 (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 own reported result comes from a mixed models analysis and is classified as a superiority analysis. The registry reports p < 0.001 but does not report the model-based estimate, its confidence interval, model specification, or additional details that would allow independent reproduction. That result is distinct from the unadjusted recalculation based on the posted treatment means, standard errors, and counts.

The recalculation gives a mean difference of 13.28, a confidence interval from 3.579 to 22.98, and an unadjusted p-value of 0.007. It uses an approximate pooled standard deviation and does not recover the within-participant relationship that is relevant to a crossover comparison. It therefore serves as a simple check from aggregate statistics, not a reconstruction of the sponsor’s mixed model.

The registry does not flag a discrepancy between the sponsor analysis and the recalculation. The two results should nevertheless not be treated as identical analyses: the sponsor’s result is the trial’s reported model-based analysis, while the recalculation rests on simplified assumptions and incomplete aggregate information. The registry record does not contain enough detail to explain precisely why their p-values differ.

14. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when several outcomes, comparisons, time points, or analyses are tested. Repeated testing can increase the chance of obtaining a statistically notable result by chance if no adjustment or prespecified testing strategy is used. General statistical practice addresses this through such approaches as prioritizing a primary endpoint or adjusting tests, but the appropriate method depends on the trial plan.

The registry lists one outcome measure and one analysis, and no secondary outcome results are posted. This does not establish that only one outcome or analysis was planned or conducted. The statistical analysis plan, multiplicity strategy, and any adjustment for multiple testing are not reported in the registry record. It is therefore not possible to determine whether the reported p-value was adjusted or how the primary outcome fit within any broader testing strategy.

The registry record also does not report a missing-data method, participant-level flow through the analysis, or the analysis population definition. It lists 21 observations for AXS-12 and 20 for placebo, but does not explain the difference. The record does not state how incomplete treatment periods, missing endpoint observations, or withdrawals were handled, or whether participants were included according to a particular intention-to-treat or other population definition.

For a crossover study, missing information from either treatment period can affect the paired comparison. General methods may use models that accommodate incomplete observations under explicit assumptions, but the model and assumptions used by the sponsor are not described. The registry identifies a mixed models analysis without reporting its missing-data assumptions or sensitivity analyses.

Interim analyses, stopping rules, and the detailed statistical analysis plan are not reported in the registry record. Their absence from the record should not be interpreted as proof that such procedures were or were not used. These documentation gaps limit independent assessment of how multiplicity, incomplete data, and the crossover analysis were handled.

15. Safety Results

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

Deaths

Other (non-serious) adverse events

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

Most frequent non-serious adverse event terms

The serious adverse event counts are zero among 21 participants in the AXS-12 group and zero among 20 participants in the placebo group. The deaths count is zero for each group. These are counts, not rates, and no serious adverse event terms are listed. The registry also reports other, non-serious adverse events affecting 9 participants in the AXS-12 group and 8 in the placebo group; these are separate from the serious-event counts.

These figures describe what is recorded for the trial and should not be interpreted as proof that serious events cannot occur. A zero count in a study of this size does not establish absence of risk. The registry record does not provide event-level descriptions, exposure-adjusted rates, or details about severity, timing, relatedness, or how safety events were collected. The treatment-specific counts also arise in a crossover study, and the record does not explain the exposure or attribution rules used for adverse events.

The non-serious adverse event counts should not be combined with the serious-event figures. They represent a different category, and the registry gives no further breakdown of those events. These descriptive counts alone do not establish a causal comparison between treatments.

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

The posted results describe a positive estimated change in weekly average cataplexy attacks for AXS-12 and a value close to zero but negative for placebo, using the endpoint’s stated direction in which positive change indicates improvement. The unadjusted recalculation yields a positive mean difference of 13.28, with a confidence interval from 3.579 to 22.98. The sponsor’s mixed models analysis reports p < 0.001 for superiority. These results characterize the trial’s posted statistical findings; they do not establish an individual participant’s outcome.

The result is specific to the primary cataplexy-frequency endpoint and the two-week assessment timeframe. It does not establish changes in every symptom of narcolepsy, longer-term outcomes, or effects beyond the trial population. Although the trial description mentions excessive daytime sleepiness and other narcolepsy symptoms, no secondary outcome results are posted in the registry record.

The confidence interval indicates uncertainty around the recalculated treatment difference, while the sponsor’s reported p-value does not give an effect magnitude or interval. The two analyses also differ in their documented methods: the sponsor reports a mixed models analysis, whereas the recalculation uses aggregate summaries and an approximate pooled standard deviation. Because the model details and within-participant outcome relationship are not reported, the recalculation cannot reproduce the full crossover analysis.

Statistical evidence and clinical importance are different questions. Statistical results address compatibility with a statistical model and uncertainty in an estimated contrast; clinical importance requires context about meaningful changes and patient experience. The registry record does not state a threshold for a clinically important difference. The results therefore should be understood as estimates for the recorded endpoint, with limits in precision and reporting, rather than as a complete account of treatment effects or individual benefit.

17. Why This Trial Matters Statistically

This trial illustrates why design details matter when interpreting treatment comparisons. In a crossover study, participants receive both treatments in sequence, and their results across treatment periods are related. The sponsor reports a mixed models analysis, which is a model-based approach, while the simple recalculation from the posted summaries cannot account for the unreported within-participant relationship. The distinction affects how the estimated difference and its uncertainty should be understood.

The record also illustrates the value of reporting estimates and confidence intervals alongside p-values. The sponsor’s result is reported as p < 0.001, but its effect estimate and interval are not posted. The aggregate summaries permit a separate unadjusted estimate and interval, but that calculation has limitations and is not a substitute for the sponsor’s model. Readers can see both the reported statistical conclusion and the bounds on what can be independently reconstructed.

Planning illustrations are another distinct statistical question. The participant counts for several assumed fractions of the observed effect show that sample-size requirements depend on the effect chosen for planning and the desired power. They are not evidence about the trial’s original design or its observed power. The original design assumptions and sample-size justification are not reported.

Finally, the record highlights the importance of documenting analysis populations, missing-data handling, multiplicity, and crossover model details. These elements help readers assess how the reported result was obtained and how robustly it represents the trial data. Their absence does not invalidate the posted findings, but it limits independent evaluation.

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 important details of the crossover design and analysis unspecified. It does not describe the randomization procedure, treatment sequence implementation, any washout period, or the sponsor model’s treatment of period, sequence, or carryover effects. These details can matter because observations from the two treatment periods are related and may be influenced by timing or residual treatment effects.

The sponsor reports a mixed models analysis but does not provide its estimate, confidence interval, model specification, covariance structure, or missing-data assumptions. The unadjusted recalculation uses an approximate pooled standard deviation derived from aggregate summaries and does not account for the participant-level pairing that a crossover analysis can use. The record does not explain why treatment-specific outcome counts are 21 for AXS-12 and 20 for placebo.

The analysis population definition, participant flow, missing-data strategy, multiplicity plan, interim analyses, and detailed statistical analysis plan are not reported in the registry record. The original sample-size justification and design assumptions are also not reported. No secondary outcome results are posted, despite the trial description referring to assessments of other narcolepsy symptoms.

Safety reporting is limited to summary counts. The record does not give event-level details, timing, severity, relatedness, exposure-adjusted results, or adverse-event attribution rules. The serious-event counts and non-serious counts are descriptive and do not resolve these limitations.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Clinical Outcomes in Narcolepsy and Cataplexy: An Evaluation of Reboxetine Treatment (CONCERT) — NCT03881852.
• Registry record last updated 2026-06-15; status last verified 2026-05.
• Results first posted 2023-08-24.
• 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

CONCERT (NCT03881852) is a completed Phase 2, randomized, triple-masked, placebo-controlled, multicenter crossover trial of AXS-12 (reboxetine) in people with narcolepsy, cataplexy, and excessive daytime sleepiness. Enrollment was 21. The described sequence assigned participants to three weeks of placebo followed by three weeks of AXS-12, or the reverse, with AXS-12 given at up to 10 mg daily.

The primary endpoint was change from baseline in the weekly average total number of cataplexy attacks over a two-week timeframe. The posted least-squares mean changes were 13.0 for AXS-12 and -0.28 for placebo. The sponsor reports a mixed models superiority analysis with p < 0.001. An unadjusted recalculation from the aggregate summaries gives a mean difference of 13.28, with a confidence interval from 3.579 to 22.98 and p = 0.007.

The registry does not report detailed model specifications, missing-data handling, multiplicity procedures, or the original sample-size justification. No secondary outcome results are posted.