EDG004 Treatment of Adult Patients With Generalized Anxiety Disorder (GAD): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for EDG004 Treatment of Adult Patients With Generalized Anxiety Disorder (GAD) (NCT02305797), a randomized phase 3 study. The primary outcome is Change From Baseline in Hamilton Anxiety Rating Scale (HAM-A) Total Score at Visit 7 (Day 42).

2. Clinical Question

The endpoint measures symptom-score change over the study period, rather than a diagnosis, a long-term outcome, or a direct measure of daily functioning. Negative change represents a lower score at endpoint than at baseline; the registry states that lower HAM-A scores are favorable. The statistical comparison concerns the difference between the groups, not simply whether either group changed from its own baseline.

3. Trial Design

The registry describes a phase 3, randomized, double-blind, placebo-controlled, flexible-dose, parallel-group study of extended-release lorazepam (EDG004) in adults with generalized anxiety disorder. Participants were allocated to EDG004 or placebo. The masking classification is quadruple, although the registry record does not identify the four parties or roles that were masked. The sponsor was Edgemont Pharmaceuticals, LLC.

The study began in 2015-01, had primary completion in 2016-05, and is listed as completed. Enrollment was 495. The primary endpoint was change from baseline in HAM-A total score at Day 42 (Visit 7). The registry states that all 495 patients were randomly assigned and received at least one dose of study treatment. The mITT Set comprised 239 patients assigned to EDG004 and 240 assigned to placebo. The posted arm summaries use those mITT counts.

The record names extended-release lorazepam and placebo as the study interventions and describes the active arm as EDG004 extended-release lorazepam capsules. It does not report the flexible-dose schedule, dose levels, treatment duration beyond the endpoint timeframe, or the detailed visit schedule. Those details should not be inferred from the endpoint visit alone.

Randomization and masking are design features intended, in general, to support a fair comparison and reduce the influence of expectations or treatment assignment on assessment. The registry record does not report the randomization procedure, allocation concealment details, or the specific implementation of masking. It also does not report the original sample-size justification or the trial’s prespecified statistical analysis plan. The planning illustrations derived from the observed effect are separate from those original design decisions.

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: Patients were evaluated for eligibility during a screening period and underwent a washout from prior and concomitant medications using protocol-specified guidelines, where applicable.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: A total of 495 patients were randomly assigned to a treatment group and received at least one dose of study treatment. The mITT Set included 239 patients randomly assigned to EDG004 and 240 patients randomly assigned to placebo.

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: A total of 495 patients were randomly assigned to a treatment group and received at least one dose of study treatment. The mITT Set included 239 patients randomly assigned to EDG004 and 240 patients randomly assigned to placebo.

6. Primary Endpoint

The registry lists the primary outcome as Change From Baseline in Hamilton Anxiety Rating Scale (HAM-A) Total Score at Visit 7 (Day 42), measured over Day 42 (Visit 7).

The primary endpoint is change from baseline in the clinician-rated HAM-A total score at Day 42 (Visit 7). HAM-A is a symptom rating scale for anxiety. The registry states that the total score ranges from 0 to 56 and that a lower score is favorable. Baseline was defined as the last non-missing value before the participant received double-blind study drug.

A change score compares an endpoint measurement with a participant’s baseline measurement. The registry’s primary summary is a least-squares mean, a model-based group mean rather than a simple arithmetic mean. The statistical question is whether average change differed between the randomized groups. That between-group contrast is more informative for the treatment comparison than either group’s change considered in isolation, because both groups may change during a trial for reasons unrelated to their assignment.

HAM-A offers a structured way to quantify clinician-rated anxiety symptoms, making symptom change suitable for a group-level comparison. It does not, by itself, describe every aspect of a participant’s experience or establish the cause of a change. The registry’s rationale was to evaluate whether the mean reduction from baseline to endpoint was greater with EDG004 than with placebo. The endpoint is limited to the specified Day 42 visit; longer-term outcomes and other dimensions of health are not represented by this measure.

7. Results Posted in the Registry

• EDG004: mean -9.01 versus -8.93 on control; difference -0.08 (95% CI -1.40 to 1.24), standardized effect d = -0.01; z-test p = 0.905.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Change From Baseline in the Clinical Global Impression-Severity (CGI-S) Score at Visit 7 (Day 42)

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard errors.
• Result:
  • EDG004: mean -1.14 versus -1.10 on control; difference -0.04 (95% CI -0.24 to 0.16), standardized effect d = -0.04; z-test p = 0.692.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between EDG004 and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists one secondary endpoint: change from baseline in the Clinical Global Impression-Severity (CGI-S) score at Day 42 (Visit 7). The posted arm estimates both move downward, and the recalculated between-group contrast is small relative to its interval. This broad pattern does not provide a clear separation between the groups on the secondary measure.

The secondary outcome is related to the primary question because both assess change over the same timeframe, but they are different scales and represent different assessments. Agreement in the general direction of group summaries does not independently establish a treatment effect. Nor should the secondary outcome be treated as a substitute for the primary HAM-A comparison.

The registry record lists no reported analyses. Its secondary comparison is an unadjusted recalculation, and no multiplicity plan is reported. In general, examining multiple outcomes increases the chance that some result may appear notable by chance, especially when each is interpreted without adjustment. A secondary result therefore carries less weight than a prespecified primary comparison when the adjustment strategy and analysis hierarchy are unknown.

The record does not state whether the secondary endpoint was intended to support a confirmatory claim, how its missing observations were handled, or whether other outcomes were analyzed but not posted. Those details are not reported in the registry record. The secondary evidence is best viewed as descriptive context for the primary result, rather than as an independent confirmation.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Change From Baseline in the Clinical Global Impression-Severity (CGI-S) Score at Visit 7 (Day 42)

9. Statistical Interpretation

10. Statistical Methods Explained

The primary measure is continuous: each participant has a numerical HAM-A score, and the analysis summarizes average change. A least-squares mean is generally a mean estimated by a statistical model. It can account for factors included in that model, but the registry record does not report the model specification or which factors, if any, were included. The posted arm values should therefore be understood as least-squares means without assuming a particular adjustment strategy.

What does the group contrast represent?

The mean difference compares the EDG004 change estimate with the placebo change estimate. The posted contrast is -0.08. Its negative direction means that the EDG004 estimate is slightly more negative than the placebo estimate. Since lower HAM-A scores are favorable, the direction of the point estimate is consistent with a slightly greater decrease in the EDG004 group, but its size and uncertainty matter. The interval from -1.397 to 1.237 spans both negative and positive contrasts, so the data are compatible with a range of possible differences, including no difference.

The comparison is a recalculation from posted arm summaries. It is unadjusted and should not be mistaken for a sponsor’s fitted model result. The registry record lists no reported analyses. It does not provide enough information to recreate a model-based sponsor analysis or to determine whether the registry’s least-squares means came from a model with covariate adjustment.

What do standard errors and confidence intervals tell us?

A standard error describes the uncertainty in an estimated mean across repeated samples under a statistical model. It is not the same as the spread of individual participants’ scores. The primary arm standard errors are 0.47 and 0.48. They indicate uncertainty around the estimated group means, not that individual changes were clustered within those distances.

A confidence interval gives a range of estimates compatible with the observed data under the method used and its assumptions. The contrast interval is more informative than the point estimate alone: it communicates that the estimate has uncertainty in both directions. A confidence interval is not a guarantee that the true value lies in the interval, nor does it address possible bias from missing data, measurement, or model choice. The registry record does not report the method used for a sponsor-level interval because no reported analyses are listed.

Why is the p-value unadjusted?

The posted comparison has an unadjusted p-value of 0.905. In general, a p-value describes how compatible data at least as extreme as those observed would be with a specified null hypothesis, under the analysis assumptions. It is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect. “Unadjusted” means the value is not described as accounting for a family of multiple comparisons. The registry record does not report a multiplicity procedure or confirm whether any other adjustment was prespecified.

Clinical trial interpretation should consider the estimate, its interval, the endpoint, and the analysis population together. A threshold-based reading of a p-value alone would discard information about the direction and precision of the estimate. Here, the contrast is close to zero and the interval is broad enough to include differences in both directions. That is a description of uncertainty, not a finding that the groups are exactly alike.

Why can over-dispersion matter?

In general, over-dispersion means that observed variability is greater than a statistical model expects. It is particularly relevant for count outcomes, where a simple model may understate uncertainty if counts vary more than assumed. The primary endpoint in this study is a continuous score, and the registry supplies standard errors for the least-squares means. It does not report a dispersion assessment or the details needed to evaluate model fit. Over-dispersion should not be assumed to have occurred in this trial.

Other general issues for continuous outcomes include the distribution of change scores, the handling of repeated measurements, and the influence of missing endpoint data. The registry record does not state the diagnostic checks or methods used for these issues. They are methodological considerations, not additional findings about this study.

What is the role of the analysis population?

The registry reports 495 patients randomly assigned and receiving at least one dose, and defines the mITT Set as 239 assigned to EDG004 and 240 assigned to placebo. The primary arm summaries use the mITT counts. The record does not provide a fuller definition of mITT, such as any additional eligibility or outcome requirements, and does not report a separate analysis population or a participant flow accounting for exclusions. Accordingly, conclusions from these summaries pertain to the posted analyzed population rather than automatically to every person enrolled or to a broader population.

Randomization supports comparison of assigned groups, but an analysis can be affected if participants with missing outcomes are excluded or handled differently. Without the missing-data strategy and counts of available Day 42 measurements, the impact of missingness cannot be assessed from the registry record. These gaps limit independent evaluation of the statistical methods, even though the posted summary values can still be interpreted as reported.

11. Confidence Intervals and Precision

The primary between-group estimate is -0.08, with a 95% confidence interval from -1.397 to 1.237. The interval crosses zero, so it includes no difference as well as possible contrasts in either direction. Its width relative to the point estimate shows that the estimate is not precise enough to pin down a narrow range around the observed contrast.

In general, a confidence interval expresses statistical uncertainty around an estimate under the calculation’s assumptions. It does not account automatically for every source of uncertainty: for example, it cannot repair bias from missing outcomes or unreported analytic choices. The registry record does not report missing-data methods, the sponsor’s analysis, or a detailed model specification. The interval is an unadjusted recalculation from the posted summary statistics.

The endpoints of the interval should not be treated as equally likely values or as a prediction of an individual participant’s response. They describe uncertainty about the group contrast. The interval also does not establish equivalence between EDG004 and placebo; an equivalence conclusion requires an appropriate margin and design, neither of which is reported in the registry record.

The arm-level standard errors, 0.47 for EDG004 and 0.48 for placebo, describe uncertainty around the respective least-squares means. They do not replace the interval for the difference between groups, which is the quantity relevant to the comparative question. Reading the contrast, its interval, and the unadjusted p-value together avoids overstating what the point estimate alone can support.

12. 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 illustrations express required sample size as participants per arm, not as a total across both arms and not as total events. The assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. Each illustration uses a stated fraction of the observed effect as the effect to detect. The original trial’s own design assumptions and sample-size justification are not reported in the registry record.

The 80% power column gives the illustrative number of participants per arm for an 80% chance of detecting the assumed effect under those planning assumptions, if that effect and the statistical assumptions were correct. The 90% column gives the corresponding illustrative number for a 90% chance. These are planning quantities, not a description of what the completed trial achieved.

The rows consider planning for 100%, 75%, and 50% of the observed effect. Smaller assumed effects require more participants because a smaller group difference is harder to distinguish from random variation. The very large estimates reflect the small observed primary contrast used as the basis for these illustrations. They should not be read as the intended target enrollment for the trial.

The associated curve illustrates how the planning sample size changes with the assumed effect under the stated assumptions. These figures are not observed power: they are prospective planning illustrations calculated from the observed effect, not a retrospective measure of the probability that this completed study would detect a true effect.

13. 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 = -0.080; pooled standard deviation σ = 7.352.
2. n = 2 × (1.96 + 0.842)² × 7.352² / -0.080² = 132,568.0, rounded up to 132,568 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 <- -0.0800000000000001                      # observed difference in means
sigma <- 7.35174615234161                     # 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(2305797)
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 = 132,568 (80% power), 177,471 (90% power)
 75% of observed effect: n per arm = 235,677 (80% power), 315,504 (90% power)
 50% of observed effect: n per arm = 530,272 (80% power), 709,884 (90% power)
power.t.test: 132569 
Simulated power at n = 132,568 per arm: 0.800 (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.

14. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when a study evaluates several outcomes, time points, subgroups, or contrasts. In general, the more statistical comparisons made, the greater the chance of at least one apparently unusual result under a setting where none of the underlying effects is real. Common approaches include prespecified testing sequences or adjustments to significance thresholds, but no such approach is reported in the registry record.

The registry lists a primary HAM-A endpoint and one secondary CGI-S endpoint. It records no reported analyses. The available material does not specify whether the primary comparison was protected from other testing, whether the secondary outcome was tested formally, or whether any multiplicity correction was used. The unadjusted p-values should therefore be read as unadjusted comparisons, not as results known to account for all study testing.

Missing-data handling is also not reported in the registry record. It does not describe how missing Day 42 scores were treated, whether participants with incomplete measurements were included in the analysis, or what assumptions were made about missingness. The record defines baseline as the last non-missing value before double-blind study drug, but that definition does not establish how missing endpoint values were handled.

The analysis population is described as the mITT Set, with 239 assigned to EDG004 and 240 assigned to placebo. The record also states that 495 patients were randomly assigned and received at least one dose. It does not give a full participant flow or a more detailed definition of mITT. As a result, the extent and reasons for any difference between the overall treated group and the posted analysis population are not reported in the registry record.

These omissions do not erase the posted estimates, but they constrain assessment of how robustly the estimates answer the planned question. In particular, missing-data assumptions and multiplicity procedures can affect uncertainty and interpretation. Neither should be posted by assumption when the registry record does not report them.

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: intentional self-injury (0.4%), acute respiratory failure (0.4%).

The adverse-event summaries use safety groups of 247 participants at risk in the EDG004 group and 248 in the placebo group. Serious adverse events affected 2 participants in the EDG004 group and 0 in the placebo group. The registry lists Intentional Self-Injury and Acute Respiratory Failure among the serious terms, each at 0.4%.

The other-event category affected 131 participants in the EDG004 group and 64 in the placebo group. These counts describe participants affected, not the total number of event episodes. The registry record does not establish from these figures whether an event was caused by the assigned study treatment. Differences in event counts should be interpreted descriptively and in the context of event definitions, duration of observation, and the population at risk, details that are not fully.

Deaths are not reported in the registry record. The adverse-event figures alone do not provide a complete safety assessment: they do not describe event timing, severity beyond the serious classification, duration, treatment discontinuation, or statistical comparisons. They should not be used as a stand-alone basis for an individual treatment decision.

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

The primary posted data show similar average decreases in HAM-A score in the two groups at Day 42, with an unadjusted recalculated mean difference of -0.08. The 95% confidence interval ranges from -1.397 to 1.237, and the unadjusted p-value is 0.905. Together, the estimate and its uncertainty do not provide a precise demonstration of a between-group difference in either direction.

This does not mean the treatments are proven identical. A study can fail to show a clear difference without establishing that any possible difference is negligible. Equivalence or non-inferiority requires a prespecified margin and an analysis designed for that question; those are not reported in the registry record. The confidence interval includes zero, but it also spans nonzero values.

The result concerns a particular symptom scale, analysis population, and endpoint visit. It does not directly answer questions about outcomes outside HAM-A, later follow-up, all adults with GAD, or different doses and treatment settings. Although the trial is described as flexible dose, dose details are not reported in the registry record. Generalization beyond the analyzed participants therefore requires care.

The evidence also has reporting limits. The registry lists no reported analyses, and it does not describe the missing-data method, the detailed statistical model, or multiplicity handling. The recalculated contrast is a transparent summary-statistic comparison, but it cannot replace an unavailable sponsor analysis or resolve those methodological uncertainties.

Finally, group averages do not describe every participant’s experience. The posted results do not establish why scores changed, whether any individual experienced a particular outcome, or whether adverse events were caused by treatment. The appropriate statistical conclusion is limited to the posted group summaries and their precision, not a general claim that the treatment does or does not work.

17. Why This Trial Matters Statistically

This trial illustrates why comparative estimates matter more than separate within-group changes. The HAM-A mean estimates decreased in both groups, but the randomized question concerns whether those changes differed. The recalculated contrast is close to zero, and its confidence interval includes differences in both directions. Reading only the decreases within each arm could therefore give an incomplete account of the evidence.

It also shows why an estimate should be read alongside its precision. A point estimate gives one summary of the data, while an interval communicates how uncertain that summary is. Here the uncertainty does not support a narrowly defined conclusion about the size of any group difference. The unadjusted p-value alone cannot express that range or establish that the groups are identical.

The planning illustrations provide another statistical lesson: sample-size requirements depend strongly on the effect assumed for planning. When the target effect is smaller, more participants are needed to detect it under otherwise fixed assumptions. Because these illustrations are based on the observed effect, they are not a record of the sponsor’s original planning and do not measure the trial’s achieved power.

Finally, registry interpretation depends on transparent analysis reporting. The population counts and endpoint summaries are useful, but the absence of a reported analysis, missing-data approach, and multiplicity procedure restricts independent evaluation. Separating what the posted numbers show from what remains undocumented is central to a careful reading of clinical trial statistics.

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 a sponsor analysis for the primary comparison, although it provides arm summaries and a recalculated contrast. The recalculation is unadjusted and based on summary statistics; it cannot recover an unreported model specification. The record also does not provide the original sample-size justification, so the planning illustrations based on the observed effect should not be mistaken for the trial’s design target.

Important analytic details are not reported in the registry record, including the method for handling missing endpoint data, a detailed definition of the mITT Set, and the handling of multiple outcomes or tests. No full participant flow is given to explain the difference between the total group described as randomized and dosed and the mITT counts used in the outcome summaries. These omissions limit assessment of possible selection effects and robustness.

The primary inference is confined to clinician-rated HAM-A change at Day 42. The record does not establish longer-term effects or provide evidence across other outcome domains. The flexible-dose label is present, but dose levels and adjustment procedures are not reported. The masking category is given as quadruple without identifying the masked roles.

Safety summaries are descriptive. The record provides counts of participants affected and selected serious-event terms, but not the event-level timing, duration, or causal assessment needed to interpret those counts fully. Deaths are not reported in the registry record. Taken together, the results can be described statistically, but several uncertainties about design implementation, analysis, missingness, and safety context remain unresolved.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: EDG004 Treatment of Adult Patients With Generalized Anxiety Disorder (GAD) — NCT02305797.
• Registry record last updated 2018-09-26; status last verified 2016-07.
• Results first posted 2018-09-26.
• 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

NCT02305797 is a completed phase 3 randomized, double-blind, placebo-controlled, flexible-dose, parallel-group study of extended-release lorazepam (EDG004) in adults with GAD. The registry lists enrollment of 495, a start date of 2015-01, and primary completion in 2016-05.

The primary endpoint was change from baseline in clinician-rated HAM-A total score at Day 42 (Visit 7). The posted least-squares mean changes were -9.01 for EDG004 and -8.93 for placebo. The unadjusted recalculated difference was -0.08, with a 95% confidence interval from -1.397 to 1.237 and an unadjusted p-value of 0.905.

The registry also lists a Day 42 CGI-S secondary endpoint and adverse-event summaries. It does not report analyses, the original sample-size justification, missing-data methods, multiplicity handling, or deaths. Interpretation is therefore limited to the posted summaries and their stated precision.