Assessing Symptomatic Clinical Episodes in Depression: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Assessing Symptomatic Clinical Episodes in Depression (NCT03595579), a randomized phase 2 study. The primary outcome is MADRS Score - Overall Change From Baseline.

2. Clinical Question

This was a randomized, double-blind, active-controlled comparison. The primary outcome is described as overall change from baseline in MADRS score, with assessments at week 1, 2, 3, 4, 5 and 6 averaged over the entire 6-week treatment period. The posted arm values are means, and the accompanying spread values are standard errors. The primary analysis population is identified as the mITT population, but the record does not explain which participants were included in that population or how missing measurements were handled.

The registry reports 97 enrolled participants overall. The arm-level outcome results report 43 for AXS-05 and 37 for bupropion. These counts need not represent the same population or analysis set as the safety counts, which list 48 participants at risk in each arm. The record does not explain the difference in these counts.

The central statistical question is whether the posted change scores differ between randomized groups and how precisely the comparison is estimated. This is distinct from asking whether either treatment has a clinically important effect, whether the results apply to people outside the trial, or whether the comparison establishes effects on outcomes not reported in the registry.

3. Trial Design

The registry describes a randomized, quadruple-masked, parallel-group, active-controlled phase 2 trial of AXS-05 and bupropion in major depressive disorder. Its official title identifies AXS-05 as administered orally. The registry status is completed. The listed start date is 2018-05-30, and the primary completion date is 2019-01-07. The total enrollment is 97.

Randomization is intended to create treatment groups that are comparable on average at the start of a trial. An active-control design compares the intervention with another treatment rather than with no treatment. A parallel-group design assigns participants to a group and compares outcomes across groups. These general features describe the design, but the registry record does not provide the randomization method, allocation ratio, concealment procedure, or details of the masking arrangements. It also does not explain which parties were included in the quadruple masking.

The two listed interventions are AXS-05 and bupropion. Arm descriptions are blank, so treatment doses, schedules, adherence procedures, and any permitted accompanying care are not reported in the registry record. The study is identified as phase 2; this label alone does not establish how the trial’s sample size or analysis plan was chosen.

The primary outcome is MADRS score change from baseline, assessed repeatedly during treatment and averaged over the stated 6-week period. The results list 43 participants in the AXS-05 arm and 37 in the bupropion arm for this outcome. The sponsor’s reported analysis is identified as a mixed models analysis. The registry record does not provide the model specification, covariates, assumptions, or details about how repeated observations and missing outcomes entered that analysis.

Although the total enrollment is reported, the registry record does not state the trial’s original sample-size justification, planned power, or target treatment difference. The planning illustrations associated with the observed result are not evidence of the original design rationale.

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

Analysis population for the primary outcome, as defined in the registry: modified Intent-to-Treat (mITT) Population

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: Participants randomized

6. Primary Endpoint

The registry lists the primary outcome as MADRS Score - Overall Change From Baseline, measured over Assessed at week 1, 2, 3, 4, 5 and 6 (averaged over the entire 6-week treatment period).

The primary endpoint is overall change from baseline in the Montgomery-Åsberg Depression Rating Scale (MADRS), a measure of depressive-symptom severity. The registry describes a 10-item scale in which each item is scored between 0-6 points. A score of 0 on an item indicates absence of symptoms, while a score of 6 indicates symptoms of maximum severity; the maximum total score is 60 points.

The outcome was assessed at week 1, 2, 3, 4, 5 and 6, with the primary measure described as averaged over the entire 6-week treatment period. Repeated assessment can characterize symptom changes across treatment rather than relying on a single measurement occasion. Averaging across the stated period defines the reported endpoint as an overall measure across those assessments, though the registry record does not explain the exact averaging procedure.

Change from baseline is used to describe how scores during follow-up differ from an initial measurement. The scale’s stated scoring direction means higher scores correspond to more severe symptoms at the item level. However, the registry record does not provide baseline arm means or the precise coding convention for the posted change values. Interpretation should therefore stay close to the registry’s description of the endpoint and the reported group comparison, rather than assuming unreported details about score construction.

The record reports means and standard errors for the endpoint, and identifies the analysis population as the modified Intent-to-Treat (mITT) Population. It does not report a clinically important difference threshold for MADRS.

7. Results Posted in the Registry

• Bupropion: mean 8.80 versus 13.70 on control; difference -4.90 (95% CI -6.72 to -3.08), standardized effect d = -1.19; z-test p = <0.001.

8. Statistical Interpretation

9. Statistical Methods Explained

What does a mean difference describe?

A mean summarizes the average outcome in a group. A mean difference compares the group averages using a specified subtraction order. In the registry’s comparison, the estimate is -4.9 for bupropion relative to AXS-05. The arm means are 8.8 for bupropion and 13.7 for AXS-05. The sign depends on which group is subtracted from which; it should not be interpreted without the contrast definition. Here, the posted contrast makes clear that the bupropion mean is lower for this outcome.

The outcome is a mean of a continuous measure, rather than a count or a binary event. The meaning of the group average depends on the measure and its coding. The MADRS is described as a symptom-severity scale, but the record does not provide a clinical-importance threshold or detailed conventions for constructing the change score.

What do standard errors convey?

The arm results report standard errors of 0.63 for AXS-05 and 0.68 for bupropion. A standard error describes the estimated sampling variability of a group mean: in general, it indicates how much the estimate would vary across repeated samples under a suitable statistical model. It is not the spread of individual participant scores. Individual outcomes can vary substantially even when the mean’s standard error is comparatively small.

The standard errors help convey the precision of each arm mean. They do not, on their own, fully describe the distribution of outcomes or establish that model assumptions are satisfied. The registry record does not report standard deviations for the arm outcomes, nor does it provide the individual-level data needed to examine distributions or outliers.

How should the confidence interval be read?

The recalculated mean difference is -4.9, with a confidence interval from -6.717 to -3.083. A confidence interval is a range of values compatible with the data and the specified method at its stated confidence level. In repeated use of the same procedure under its assumptions, such intervals have a long-run coverage interpretation. The registry comparison reports the interval limits but does not state a confidence level in the comparison itself, so the confidence level should not be inferred from the record.

The interval describes uncertainty in the estimated mean difference, not a range containing a specified proportion of individual participants’ outcomes. Its width reflects sampling precision under the calculation’s assumptions. It cannot capture every source of uncertainty, including possible bias, model misspecification, missing-data effects, or limitations in how the endpoint was defined.

What is the difference between the sponsor analysis and recalculation?

The sponsor identifies a mixed models analysis and reports p <.001. Mixed models are a general class of methods that can represent relationships among repeated measurements or other grouped data. Their exact behavior depends on the model terms, covariance structure, estimation method, and assumptions. Those details are not reported in the registry record, so the sponsor’s analysis cannot be reproduced from the description alone.

The unadjusted recalculation uses the posted group means and standard errors to estimate a simple mean difference and its interval and p-value. It does not reproduce a mixed model or account for unreported model features. It is useful as a transparent comparison based on summary values, but it is not a substitute for the sponsor’s trial analysis. The recalculation’s p-value is p < 0.001; the sponsor’s is p <.001. The distinction in notation should be preserved when describing each result.

Why can over-dispersion matter?

In general, over-dispersion means that observed variability is greater than a statistical model expects. It matters especially for outcomes such as counts when a model assumes a particular level of variability: failing to allow for extra variation can make uncertainty appear too small. The primary endpoint in this trial is continuous, and the registry does not report an over-dispersion assessment. It would therefore be inappropriate to claim that over-dispersion was present or affected this result. More generally, checking whether a model describes the data adequately is part of evaluating statistical uncertainty, but the relevant diagnostic details are not reported in the registry record.

What assumptions should be kept in view?

The recalculation is unadjusted and uses the posted summary statistics. It does not account for covariates, repeated-measures structure, or any features of the sponsor’s model that are not reported. Interpretation also depends on how the mITT population was formed, how measurements were collected, and how missing values were treated. Those details are not reported in the registry record. The sponsor’s own analysis and the recalculation answer related but not identical statistical questions.

10. Confidence Intervals and Precision

The recalculated mean difference is -4.9, with a confidence interval from -6.717 to -3.083. This interval gives a range of plausible values for the contrast under the recalculation’s method and assumptions. Every value in the reported interval is negative, so the interval is consistent with a difference in the direction represented by the estimate, rather than a zero difference, under that calculation. This does not establish that the difference meets a clinical-importance threshold; no such threshold is reported.

Precision is about how narrowly an effect is estimated, not whether the effect is important or free of bias. The posted arm means have standard errors of 0.63 for AXS-05 and 0.68 for bupropion. These quantify uncertainty in the group means, not the full range of participant experiences. The comparison’s interval is more directly relevant to uncertainty in the between-group contrast.

The registry comparison does not state the confidence level associated with this recalculated interval. The sponsor’s mixed-model analysis reports a p-value of p <.001 but no effect estimate or confidence interval. As a result, the sponsor analysis’s precision cannot be assessed from the posted registry result in the same way as the recalculated contrast.

Intervals and standard errors reflect only the uncertainty represented by their calculation. They do not resolve uncertainty from unreported analysis choices, missing outcomes, possible measurement limitations, or the trial’s generalizability. The registry record does not provide enough detail to assess those sources fully.

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 events. They are illustrations based on the observed effect, rather than a description of the trial’s original design. The stated planning assumptions are a two-sided significance level 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set at the stated fraction of the observed effect. The trial’s original sample-size assumptions and justification are not reported in the registry record.

The 80% power column gives the illustrative number of participants per arm under which the specified planning method would have an 80% chance of detecting the selected effect, if its assumptions hold. The 90% power column uses a higher probability of detecting that effect and therefore gives a different planning target. These are prospective design concepts, not probabilities that the observed result is correct.

The rows vary the assumed effect: planning for 100% of the observed effect, planning for 75% of the observed effect, and planning for 50% of the observed effect. Smaller effects need more participants to detect, all else equal, which is why the planning figures increase as the assumed effect becomes a smaller fraction of the observed one. The curve should be read as illustrating this relationship under the stated assumptions, not as a guarantee of the sample size required in another study.

This is not observed power. The calculations use the observed effect as an input to planning illustrations; they do not measure the probability that the completed trial detected a true effect. Because observed effects are uncertain, these figures should be treated as conditional planning aids rather than as evidence about the adequacy of the completed trial’s design.

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 = -4.900; pooled standard deviation σ = 4.134.
2. n = 2 × (1.96 + 0.842)² × 4.134² / -4.900² = 11.2, rounded up to 12 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 <- -4.9                      # observed difference in means
sigma <- 4.13353731920141                     # 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(3595579)
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 = 12 (80% power), 15 (90% power)
 75% of observed effect: n per arm = 20 (80% power), 27 (90% power)
 50% of observed effect: n per arm = 45 (80% power), 60 (90% power)
power.t.test: 12.2 
Simulated power at n = 12 per arm: 0.793 (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 analysis is reported as a mixed models analysis, classified as superiority, with p <.001. The registry does not report its effect estimate, confidence interval, model specification, or detailed assumptions. This is the trial’s reported inferential result, but the sparse summary limits independent assessment of how the model was constructed and what adjustments it included.

The unadjusted recalculation from the posted arm means and standard errors gives a mean difference of -4.9 for bupropion relative to AXS-05, with a confidence interval from -6.717 to -3.083 and p < 0.001. It is a simple comparison derived from summary statistics, not a reconstruction of the mixed-model analysis. The two analyses use different methods, so their numerical results should not be presented as if they were outputs of the same model.

The recalculation is not flagged as discrepant with the sponsor result. Even so, the sponsor’s own analysis takes precedence as the trial’s reported analysis. The recalculated estimate and interval provide an additional, transparent view of the posted summary data, but they do not override the sponsor’s model or fill in details absent from the registry record.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists two outcome measures and two analyses, but reports no posted secondary outcomes. The primary outcome is identified as overall change from baseline in MADRS score. The sponsor reports a mixed models analysis for the comparison and classifies it as superiority. The registry record does not describe a multiplicity strategy, such as whether multiple endpoints, time points, or analyses were treated as a family of tests or whether any adjustment was planned.

Multiplicity is a general concern when several statistical tests are conducted: the chance of at least one apparently unusual result can rise as more tests are considered. Whether and how that issue applies depends on the full analysis plan, which is not reported in the registry record. The listed number of outcomes and analyses alone is not enough to determine which tests were prespecified, how they were related, or whether the reported primary analysis was adjusted.

The endpoint combines assessments at week 1, 2, 3, 4, 5 and 6 into an average over the stated treatment period. The record does not say whether those individual assessment times were also tested separately, whether they were part of a repeated-measures model, or how any such analyses were handled. Do not infer that they were, or were not, included in additional statistical testing.

The primary outcome analysis population is the modified Intent-to-Treat (mITT) Population. The registry record does not define the mITT population, explain participant exclusions, report the number with missing endpoint measurements, or describe how missing values were handled. It also does not state whether all randomized participants were included in the sponsor’s model. Missing-data assumptions can affect interpretation because analyses may rely on assumptions about the unobserved outcomes; the available record does not allow those assumptions to be assessed.

The outcome results list 43 participants for AXS-05 and 37 for bupropion, whereas the safety results list 48 at risk in each arm. The registry record does not explain this difference. These counts should be understood in the context of their respective reported analyses and not assumed to represent the same analysis population. The absence of a detailed participant flow or analysis plan limits assessment of exclusions, missingness, and how the endpoint results relate to all randomized participants.

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 results list 0 affected participants among 48 at risk in the AXS-05 group and 0 among 48 at risk in the bupropion group. The registry also lists 0 deaths in each group. No serious adverse event terms are listed in the record. These are the posted counts for the reported safety categories; they do not establish that no adverse experiences of any kind occurred.

The registry separately reports other adverse events in 35 of 48 participants for AXS-05 and 30 of 48 participants for bupropion. These are a distinct category from serious adverse events. The registry record does not provide further event descriptions in the serious-event terms field, and the figures should not be combined into an unreported summary.

Safety results describe the recorded events in the listed population and observation period. The record does not provide event definitions, severity detail beyond the serious-event category, exposure duration by participant, or a fuller account of how events were collected. The outcome-analysis counts differ from the safety counts, and the registry record does not explain why. These data support a description of the posted counts, not a broader conclusion about comparative safety.

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

The registry reports a difference in the primary MADRS outcome between the trial arms. The posted means are 13.7 for AXS-05 and 8.8 for bupropion, and the recalculated contrast is -4.9 for bupropion relative to AXS-05. Its confidence interval runs from -6.717 to -3.083. The sponsor’s mixed-model analysis is classified as superiority and reports p <.001. These results describe the registered comparison and its statistical evidence, not a universal treatment effect.

The endpoint is a change-from-baseline measure averaged across assessments during the stated 6-week treatment period. The registry record does not provide baseline arm means or detailed score coding. The numerical direction can be described according to the posted contrast, but conclusions about how much participants’ symptoms changed from baseline require care because the underlying baseline values and construction details are not available in the record.

A statistically notable group difference is not by itself proof of clinical importance. The registry does not report a threshold for a clinically meaningful MADRS change. Nor does a p-value indicate the probability that the result is true. The confidence interval conveys uncertainty under the recalculation’s method, but it does not address unreported sources of bias or model uncertainty.

The record does not establish what would happen in a different population, under a different treatment schedule, over a longer period, or for outcomes beyond those reported. It also does not provide posted secondary outcome results. The safety counts are limited to the categories and populations recorded and cannot support conclusions beyond those data.

Accordingly, the defensible interpretation is narrow: the posted primary-outcome summaries and the sponsor’s reported analysis indicate a between-group difference under this trial’s design and analysis. The magnitude, clinical importance, generalizability, and robustness to unreported analytic choices cannot be determined fully from the registry record alone.

17. Why This Trial Matters Statistically

This trial is statistically informative because it reports both a primary continuous outcome and a sponsor analysis of the randomized comparison. The outcome is measured repeatedly during a defined treatment period, while the registry describes an overall average across those assessments. That structure makes it important to distinguish the published summary contrast from the mixed-model result, which may represent a different analysis of the repeated measurements.

The registry gives arm means, standard errors, analysis counts, and a recalculated interval, alongside the sponsor’s p-value and method label. These details allow readers to distinguish a point estimate from its uncertainty and to see that a small p-value is not a complete account of a result. The sponsor’s reported analysis and the unadjusted recalculation are related evidence, but they do not have identical methods or necessarily identical estimands.

The record also highlights the importance of transparent reporting. The mITT population is named but not defined, the outcome-analysis counts differ from the safety counts, and the original sample-size rationale, missing-data handling, model details, and multiplicity approach are not reported. These omissions do not negate the posted result, but they restrict how fully its analysis can be evaluated.

Finally, the planning illustrations demonstrate how sample-size requirements depend on the effect a future study aims to detect. They are conditional planning calculations based on the observed effect, not evidence that the completed trial was adequately powered or that its result will recur. Keeping these distinctions clear is central to reading clinical trial statistics responsibly.

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 does not report the original sample-size justification, planned power, or target difference used when designing the trial. The planning illustrations are based on the observed effect and stated generic assumptions; they cannot be used to infer the trial’s original design rationale.

The sponsor analysis is described only as a mixed models analysis. Model terms, covariance structure, estimation choices, covariate adjustments, and diagnostic checks are not reported in the registry record. The mixed-model p-value therefore cannot be independently reproduced from the posted description. The unadjusted recalculation is based on summary statistics and does not resolve these gaps.

The mITT population is not defined, and the registry record does not explain why the primary outcome counts differ from the safety counts. It also does not report missing outcome counts, participant flow, or missing-data methods. This limits assessment of who contributed to the primary result and whether missing measurements could affect interpretation.

The MADRS endpoint is described as an average over assessments during the 6-week treatment period, but the record does not detail the exact averaging or change-score construction. Baseline arm means and a clinical-importance threshold are not reported. This limits interpretation of both the magnitude and the clinical meaning of the numerical contrast.

The record does not describe the treatment doses or schedules, eligibility criteria, allocation procedures, masking details, multiplicity plan, or secondary outcome results. These trial-specific reporting gaps constrain evaluation of implementation, analysis choices, and generalizability. They should be treated as uncertainties in the available record, not as evidence that a particular procedure was or was not used.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Assessing Symptomatic Clinical Episodes in Depression — NCT03595579.
• Registry record last updated 2021-09-24; status last verified 2021-09.
• Results first posted 2021-09-24.
• 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

NCT03595579 is a completed, randomized, quadruple-masked, parallel-group, active-controlled phase 2 trial in major depressive disorder. The registry lists total enrollment of 97, with AXS-05 compared with bupropion.

The primary outcome is overall change from baseline in MADRS score, averaged over assessments during the 6-week treatment period, in the modified Intent-to-Treat Population. Posted mean values are 13.7 for AXS-05 and 8.8 for bupropion. The sponsor reports a mixed models superiority analysis with p <.001. The unadjusted recalculation gives a mean difference of -4.9 for bupropion relative to AXS-05, with a confidence interval from -6.717 to -3.083 and p < 0.001.

The record reports 0 serious adverse events and 0 deaths in each safety group, with 48 at risk per group. Important analysis details, including the original sample-size rationale, missing-data handling, and full model specification, are not reported in the registry record.