A 2-Part Study to Investigate the Dose-Ranging Safety and Pharmacokinetics, Followed by the Efficacy and Safety of ZX008 (Fenfluramine Hydrochloride) Oral Solution as an Adjunctive Therapy in Children ≥ 2 Years Old and Young Adults With Dravet Syndrome: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A 2-Part Study to Investigate the Dose-Ranging Safety and Pharmacokinetics, Followed by the Efficacy and Safety of ZX008 (Fenfluramine Hydrochloride) Oral Solution as an Adjunctive Therapy in Children ≥ 2 Years Old and Young Adults With Dravet Syndrome (NCT02926898), a randomized phase 3 study. The primary outcome is Change in Convulsive Seizure Frequency (CSF) From the Baseline Period (Baseline) to the Combined Titration + Maintenance (T+M) Period.

2. Clinical Question

The main clinical question is whether convulsive seizure frequency changed differently during treatment with adjunctive ZX008 or matching placebo among children and young adults with Dravet syndrome receiving stiripentol treatment. The primary outcome uses seizure diary data and compares each participant’s frequency during the baseline period with frequency during the combined titration and maintenance period.

This is a question about a defined outcome, treatment period, and study population. It does not by itself answer whether results apply to people with different characteristics, treatment backgrounds, or follow-up periods. The registry also describes a separate first cohort focused on pharmacokinetics and safety after a single dose; the posted primary efficacy comparison concerns Cohort 2.

3. Trial Design

The registry describes NCT02926898 as a completed Phase 3, randomized, parallel-group study with quadruple masking and total enrollment of 87. It lists Dravet syndrome as the condition and ZX008 (fenfluramine hydrochloride) oral solution and matching placebo as the interventions. The study had two parts: an initial cohort assessing pharmacokinetics and safety after a single dose, followed by a randomized placebo-controlled evaluation of efficacy, safety, and tolerability. The primary outcome presented for interpretation is from the latter part, Cohort 2.

In Cohort 2, the ZX008 group received 0.5 mg/kg/day, up to a maximum of 20 mg/day, divided into two daily doses with food. The comparator received matching placebo on the same twice-daily schedule. The registry’s official description specifies that this cohort evaluated ZX008 as adjunctive antiepileptic therapy to stiripentol treatment. It does not report the allocation ratio in the study-design details, although the observed group counts are 43.0 and 44.0.

The primary outcome was assessed over 15 weeks, covering the combined titration and maintenance period. The outcome record defines a modified intent-to-treat population as randomized participants who received at least 1 dose of study treatment and had at least 1 week of diary data. The registry dates the study start as 2017-01-27 and primary completion as 2018-06-05; the overall status is completed.

The trial record does not report the original sample-size justification, the detailed randomization procedure, the full masking procedures, or the precise roles of people included in the quadruple-masking designation. It also does not report the protocol’s planned interim analyses or a complete statistical analysis plan. These omissions limit how specifically the original design choices can be assessed.

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: A total of 115 subjects were screened for eligibility to participate in Study 1504 Cohort 2. Of these, 87 subjects were enrolled and randomized.

Overall Study

Analysis population for the primary outcome, as defined in the registry: Modified intent-to-treat (mITT) Population, defined as all randomized subjects who received at least 1 dose of ZX008 or placebo and for whom at least 1 week of diary data were available.

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 Change in Convulsive Seizure Frequency (CSF) From the Baseline Period (Baseline) to the Combined Titration + Maintenance (T+M) Period, measured over 15 weeks (combined Titration + Maintenance Period).

The primary endpoint is the change in convulsive seizure frequency from baseline to the combined titration and maintenance period. The registry expresses frequency as convulsive seizures per 28 days and reports the period as 15 weeks. Participants recorded seizures in electronic diaries. The listed seizure categories include hemiclonic, focal seizures with clear observable motor signs, generalized tonic clonic, secondarily generalized tonic clonic, tonic, clonic, and drop seizures, including tonic or atonic seizures.

For each participant, the recorded seizure count over the relevant interval was divided by the number of diary days with data and then scaled to a 28-day frequency. This standardization makes frequencies more comparable when diary observation covers different numbers of days. It also means that diary completeness is part of how the endpoint is constructed: a frequency estimate depends on the recorded seizures and the count of nonmissing diary days.

A change-from-baseline endpoint uses participants’ baseline frequency as a reference and asks how their frequency differed during the specified treatment period. The mean summarizes the group’s average change, while the standard deviation describes variation in individual changes around that group mean. In general, such an endpoint is useful when the clinical question concerns change over time rather than only an end-of-period value. It does not, by itself, describe every participant’s experience or the distribution of seizure patterns.

The registry does not give further details about how missing diary days were handled in the analysis, beyond the minimum diary-data criterion for the modified intent-to-treat population.

7. Results Posted in the Registry

• Cohort 2: ZX008 0.5 mg/kg/Day: mean -3.18 versus -0.65 on control; difference -2.53 (95% CI -15.97 to 10.91), standardized effect d = -0.08; z-test p = 0.712.

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.

Percentage of Participants Who Achieved ≥ a 50% Reduction in Convulsive Seizure Frequency From Baseline to the Combined Titration + Maintenance Period

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Cohort 2: ZX008 0.5 mg/kg/Day: 53.5% (23/43) versus 4.5% (2/44) on control; risk difference 48.9 percentage points (95% CI 32.8 to 65.1), risk ratio 11.77 (2.95–46.89); unadjusted two-proportion z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for Cohort 2: ZX008 0.5 mg/kg/Day than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The listed secondary outcomes cover two different aspects of convulsive seizure experience: meeting a threshold for reduction in frequency and the longest seizure-free interval. The posted numerical results show a favorable direction for ZX008 on the responder endpoint, and the registry reports a logistic-regression superiority analysis with p < 0.001. A numerical result for longest convulsive seizure-free interval is not reported in the registry record. The evidence across secondary outcomes is therefore incomplete rather than a fully reported set of convergent results.

The responder result is related to the primary outcome because both concern convulsive seizure frequency over the combined titration and maintenance period. However, one summarizes the average numerical change, while the other divides participants according to whether they meet a specified reduction threshold. They can provide complementary perspectives, but they need not produce the same statistical summary or have the same sensitivity to the distribution of participant-level changes.

Secondary outcomes generally carry less confirmatory weight when their testing is not part of a prespecified multiplicity strategy. Evaluating several outcomes creates multiple opportunities for a low p-value even when the underlying effects are absent. The registry record does not report whether the sponsor used a hierarchy, adjustment, or other procedure to control multiplicity. Consequently, the reported secondary p-value should be read as an unadjusted result unless the record states otherwise, and not as stand-alone confirmation of the primary result.

There are three posted outcome measures in total, including the primary outcome, while two are listed as secondary. The available secondary reporting does not establish how the unreported interval outcome would compare with the primary result. Nor does it make the primary discrepancy disappear: the primary sponsor ANCOVA and the simple unadjusted recalculation remain distinct analyses. A complete interpretation would require the prespecified analysis plan and the results for all listed endpoints.

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

• Percentage of Participants Who Achieved ≥ a 50% Reduction in Convulsive Seizure Frequency From Baseline to the Combined Titration + Maintenance Period
• Longest Convulsive Seizure-Free Interval (Days)

9. Statistical Interpretation

10. Statistical Methods Explained

The primary endpoint is continuous: each participant’s change is expressed as a numerical frequency, and the registry reports group means and standard deviations. The main statistical question is whether the average change differs between treatment groups. Two kinds of results appear in the record: group-level descriptive summaries used for a simple recalculation, and a sponsor-reported ANCOVA result. They answer related questions but are not necessarily the same estimand or analysis.

What does the mean difference describe?

The unadjusted mean difference compares the two posted group means directly. With ZX008 minus placebo as the comparison direction, the estimated difference is -2.53 convulsive seizures per 28 days. This calculation does not adjust for baseline or other participant characteristics. It is a straightforward contrast of the reported means, not a recreation of the sponsor’s model.

The associated unadjusted 95% confidence interval extends from -15.97 to 10.91. A confidence interval describes the range of values reasonably compatible with the data and the analysis assumptions at the stated confidence level. Because this interval includes zero, the posted summaries do not rule out no mean difference under this unadjusted method. The interval also includes differences in both directions, so its range reflects considerable uncertainty about the magnitude and direction of the unadjusted contrast.

Why can the sponsor result differ?

ANCOVA is a regression-based method that can compare groups while accounting for specified covariates. In a trial, adjustment may include baseline outcome values, but the exact variables and model specification must be known to describe a particular analysis. The registry identifies the sponsor’s method as ANCOVA and reports a percent difference, but the adjustment variables and full model details are not reported in the registry record. It is therefore not possible to reproduce that result from the arm means and standard deviations alone.

The reported ANCOVA result is a percent difference of -54.0, with a 95% confidence interval from -67.2 to -35.6 and p < 0.001. The unadjusted recalculation is an absolute mean difference with a different confidence interval and p-value. The different scale and adjustment mean that the two estimates should not be treated as duplicate calculations of the same displayed number. Since the record flags a discrepancy between the sponsor result and the recalculation, the sponsor’s own analysis takes precedence in reporting the trial’s stated analysis.

What does the standard deviation add?

The standard deviation describes how widely participant-level change scores vary around the group mean. It is not the uncertainty in the estimated group mean; that uncertainty is reflected in a standard error or confidence interval. The posted standard deviations are 44.12 in the ZX008 group and 8.767 in the placebo group. This difference in observed spread is important for interpreting the simple mean comparison, because a mean alone can conceal substantial variation among participants.

In general, when variability is high, a group mean may be estimated less precisely unless the study has enough observations to offset that variability. Unequal variability can also make simple assumptions used by some standard comparison methods less suitable. The registry reports a pooled standard deviation of 31.63 for the recalculation, but it does not report a diagnostic assessment of variance assumptions. The sponsor’s ANCOVA model assumptions and diagnostics are not reported in the registry record.

Why does the binary secondary outcome use a different measure?

The responder endpoint classifies participants according to whether they met the stated threshold, rather than retaining each person’s full numerical change. The posted group percentages are 53.5% for ZX008 and 4.5% for placebo. A risk difference describes the absolute difference between the proportions; the recalculated risk difference is 0.489, with a 95% confidence interval from 0.328 to 0.651. The unadjusted p-value is <0.001. The registry also identifies a logistic-regression superiority analysis with p < 0.001, but does not post its model estimate or confidence interval.

These binary and continuous outcomes summarize different features of the data. A threshold result focuses on how many participants crossed a cutoff; a mean change uses the full numerical values. Neither one substitutes for the other. The remaining listed secondary endpoint is longest convulsive seizure-free interval in days, but its numerical result is not reported in the registry record.

How should p-values be understood?

A p-value describes how compatible data are with a specified null hypothesis under a particular statistical model and its 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. The unadjusted primary recalculation has p = 0.712, while the sponsor’s ANCOVA reports p < 0.001. Those values come from different analyses and should be interpreted with their methods and scales in view. The registry record does not report enough detail to independently audit the sponsor model.

All recalculated comparisons are unadjusted. They do not account for multiple outcomes or multiple analyses. The registry record does not report the multiplicity strategy, if any, used for confirmatory testing. That missing information matters when interpreting several outcomes together, because testing multiple hypotheses can increase the chance of at least one apparently notable result under a set of null effects.

11. Confidence Intervals and Precision

The unadjusted primary mean difference is -2.53, with a 95% confidence interval from -15.97 to 10.91. This interval is wide and includes zero. It is compatible with a range of possible group differences, including differences in either direction, so the recalculation is not precise enough to pin down the magnitude of the unadjusted contrast. The interval should be interpreted on the scale of the mean difference in convulsive seizures per 28 days, not as a percent difference.

The sponsor-reported ANCOVA estimate is on a percent-difference scale: -54.0, with a 95% confidence interval from -67.2 to -35.6. The interval does not include zero, and the sponsor reports p < 0.001. This is the sponsor’s adjusted analysis, not the unadjusted contrast. The registry does not provide the details needed to reconcile the intervals or reproduce the ANCOVA result. Given the flagged discrepancy, the sponsor’s own analysis takes precedence when characterizing its reported conclusion.

For the responder outcome, the recalculated risk difference is 0.489, with a 95% confidence interval from 0.328 to 0.651. The interval is on the absolute proportion-difference scale. It does not communicate the same quantity as either the primary mean difference or the sponsor’s percent difference. The sponsor’s logistic-regression result for this outcome is reported as p < 0.001, without a corresponding model estimate or confidence interval in the registry record.

In general, confidence intervals communicate both an estimate and its uncertainty; they do not show the full distribution of participant outcomes or settle whether an effect is important in practice. The contrasting precision of the primary recalculation and the sponsor’s reported analysis should not be attributed to a particular modeling choice without the missing model details. The observed group variability and the limited participant counts are relevant context, but the registry record does not report a formal assessment of precision beyond the posted estimates and intervals.

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

For the continuous primary comparison, the figures are participants per arm. For the binary responder comparison, they are also participants per arm. The general note on the sample-size calculations distinguishes these from time-to-event calculations, which would be expressed as total events; the outcomes with figures in this record are not presented as total-event requirements.

The 80% power column gives the approximate number of participants per arm under the listed assumptions for a study to have an 80% chance of detecting the specified effect if that effect were the true one. The 90% power column gives the corresponding planning number for a 90% chance. These are design probabilities under an assumed effect and statistical model; they are not guarantees about the result of a future study.

The rows for 100%, 75%, and 50% of the observed effect show how the planning requirement changes when the assumed target effect is smaller than the observed one. A smaller effect is harder to distinguish from random variation under otherwise similar assumptions, so it requires more participants per arm. The numerical values are illustrations conditional on the observed effect and the stated planning assumptions, not evidence about the original trial’s intended power or enrollment decision. This is not observed power.

These calculations should not be read as an explanation for the enrollment of 87 participants. The registry record does not report the original target sample size, the assumptions used by the sponsor when designing the trial, or whether any sample-size revisions occurred.

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 = -2.530; pooled standard deviation σ = 31.635.
2. n = 2 × (1.96 + 0.842)² × 31.635² / -2.530² = 2,454.3, rounded up to 2,455 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 <- -2.53                      # observed difference in means
sigma <- 31.6348325052511                     # 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(2926898)
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 = 2,455 (80% power), 3,286 (90% power)
 75% of observed effect: n per arm = 4,364 (80% power), 5,842 (90% power)
 50% of observed effect: n per arm = 9,818 (80% power), 13,143 (90% power)
power.t.test: 2455.3 
Simulated power at n = 2,455 per arm: 0.806 (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. Analyses Reported by the Sponsor

The sponsor’s reported primary analysis is ANCOVA and gives a percent difference of -54.0, a 95% confidence interval from -67.2 to -35.6, and p < 0.001. It is classified in the registry as a superiority analysis. The recalculation from the posted means instead gives an unadjusted mean difference of -2.53, a 95% confidence interval from -15.97 to 10.91, and p = 0.712.

These results are not directly interchangeable. One is a sponsor-reported adjusted analysis on a percent-difference scale; the other is a simple unadjusted difference of group means. The registry does not state the covariates, exact ANCOVA specification, or enough detail to reconstruct the sponsor’s result. It would therefore be inappropriate to claim that the recalculation verifies or disproves the adjusted analysis.

The record flags a discrepancy, so the sponsor’s own analysis takes precedence when describing the reported trial analysis. The recalculation remains useful as a transparent comparison of the posted summary statistics, with its limitations made explicit. Both results should be associated with the method that produced them rather than combined into a single estimate.

The sponsor’s reported logistic-regression analysis of the responder endpoint is also classified as superiority and reports p < 0.001. The model estimate and confidence interval are not reported in the registry record, so that analysis cannot be independently compared with a model-based estimate using the posted group percentages alone.

15. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when a study evaluates more than one outcome, comparison, time point, or statistical hypothesis. If each test is judged separately without an adjustment or testing hierarchy, the chance of obtaining at least one small p-value across the set can be greater than for a single test. Methods such as a prespecified hierarchy or a multiplicity adjustment can address this issue, but the appropriate approach depends on the trial’s aims and planned analyses.

The registry lists a primary continuous outcome and two secondary outcomes. It reports a sponsor ANCOVA result for the primary outcome and a logistic-regression result for the responder outcome. The longest convulsive seizure-free interval is listed but its numerical result is not reported in the registry record. The multiplicity strategy, including whether any adjustment or hierarchy was used, is not reported in the registry record. The reported secondary p-value should therefore not be assumed to have multiplicity control.

The analysis population is described as modified intent-to-treat: randomized participants who received at least 1 dose of ZX008 or placebo and had at least 1 week of diary data. This definition excludes participants without the required treatment exposure or diary availability from the primary analysis population. It is not identical to an analysis of every randomized participant, and the registry record does not report how many randomized participants were excluded under this definition or their group-specific reasons.

The record says the endpoint was based on electronic diary data and required at least 1 week of diary data for inclusion. It does not report the detailed method for handling missing diary days within the endpoint, missing outcome values, or missingness after the minimum diary criterion was met. It also does not report sensitivity analyses examining alternative missing-data assumptions. This matters because seizure frequency is derived from both recorded seizure counts and days with nonmissing diary data.

The original statistical analysis plan, interim-analysis procedures, and any changes to planned analyses are not reported in the registry record. Without those details, it is not possible to determine from the posted results whether the analyses and endpoint hierarchy followed a prespecified confirmatory plan. These omissions do not prove that such procedures were absent; they mean that they cannot be assessed from the registry record.

16. 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 serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: seizure (9.1%), status epilepticus (7.0%), osteochondritis (2.3%), lethargy (2.3%), abdominal pain (2.3%).

The registry reports serious adverse events affecting 6 of 43 participants in the ZX008 group and 7 of 44 participants in the matching-placebo group. It reports zero deaths in each group. These counts describe participants recorded as having serious adverse events; they do not indicate the number of all adverse-event episodes or provide a complete comparison of event severity, timing, or relatedness.

The listed serious adverse-event terms include seizure at 9.1%, status epilepticus at 7.0%, and osteochondritis, lethargy, abdominal pain, and pyrexia at 2.3% each. The record does not establish from these terms alone whether an event was caused by study treatment. Serious events may be recorded under more than one term, and the term percentages should not be added to infer a total participant count.

Other adverse events were recorded for 42 of 43 participants in the ZX008 group and 42 of 44 participants in the placebo group. These figures are separate from the serious-event counts. The registry record does not provide enough detail in these summary details to assess event duration, severity grading, treatment discontinuations, or causal attribution. Safety counts are descriptive and need to be considered alongside the study’s duration, population, and event definitions.

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

The posted results describe a randomized comparison of adjunctive ZX008 and matching placebo in Cohort 2 of a study involving children and young adults with Dravet syndrome receiving stiripentol treatment. The primary outcome measures mean change in convulsive seizure frequency over the combined titration and maintenance period. The sponsor reports an ANCOVA result classified as superiority, with a percent difference of -54.0, a 95% confidence interval from -67.2 to -35.6, and p < 0.001.

At the same time, the unadjusted recalculation from the posted group summaries gives a mean difference of -2.53 and a 95% confidence interval from -15.97 to 10.91, with p = 0.712. The discrepancy is important: the sponsor result and recalculation use different measures and analyses, and the registry does not provide enough ANCOVA detail to reproduce the sponsor’s estimate. The sponsor’s own analysis takes precedence, while the recalculation should be understood only as a simple unadjusted comparison of the posted summaries.

These findings do not, by themselves, settle how much the result matters to participants, how outcomes varied across individuals, or whether the same pattern would apply outside the studied population and period. The confidence intervals describe statistical uncertainty under their respective analyses; they do not capture every source of uncertainty, such as unreported missing-data decisions or analysis-plan details. The registry’s modified intent-to-treat definition also specifies which participants contributed to the primary analysis.

The responder endpoint provides another summary of seizure-frequency change, but it is not the same outcome as the primary mean. The other listed secondary endpoint lacks a numerical result in the registry record. Multiplicity procedures are also not reported, limiting how strongly the secondary finding can be treated as confirmatory. Safety counts provide a descriptive account of serious and other adverse events but do not establish causation.

Accordingly, the record supports a careful description of the sponsor’s reported analysis alongside the distinct unadjusted calculation, not a broader claim about every possible outcome, population, or treatment context. It also does not supply the information needed to resolve the primary-analysis discrepancy independently.

18. Why This Trial Matters Statistically

This trial is statistically informative because its registry record places an adjusted sponsor analysis beside summary statistics that permit a simple unadjusted comparison. The contrast makes clear why an estimate cannot be interpreted without knowing its scale and model. The sponsor reports an ANCOVA percent difference, whereas the recalculation compares group means directly. Those analyses answer related but distinct statistical questions.

The primary endpoint also illustrates the role of outcome construction. Convulsive seizure frequency is derived from electronic diary counts and nonmissing diary days, then standardized to a 28-day measure. The analysis population requires both treatment exposure and a minimum amount of diary data. These choices determine what quantity is summarized and which randomized participants contribute to the result.

Observed variability is another important feature. The standard deviation is 44.12 in the ZX008 group and 8.767 in the placebo group. The large contrast in spread is relevant to the uncertainty of a simple mean comparison and cautions against interpreting the two group means in isolation. It also shows why participant-level distribution and model assumptions matter alongside a point estimate.

Finally, the record emphasizes the difference between a statistical result and a complete evidence account. The primary result is accompanied by a flagged discrepancy, the multiplicity strategy is not reported, and one listed secondary endpoint has no posted numerical result. The example demonstrates the value of reporting analysis details, missing-data methods, and all outcome results so readers can assess how robustly a trial supports its stated statistical conclusions.

19. 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 primary efficacy comparison concerns Cohort 2 and a 15-week combined titration and maintenance period. It does not provide a long-term estimate or a result for every part of the two-cohort study. The record’s title and description include a first cohort focused on single-dose pharmacokinetics and safety, but numerical results for that cohort are not reported in the registry entries used for this analysis.

The primary analysis population requires at least 1 dose of study treatment and at least 1 week of diary data. The record does not report the number of randomized participants excluded from this population or the reasons for exclusion. Nor does it specify how incomplete diary periods and other missing outcomes were handled. Since the endpoint depends on diary information, unreported missing-data procedures limit assessment of how diary completeness may affect estimates.

A central limitation is the discrepancy between the sponsor’s ANCOVA result and the unadjusted recalculation. The ANCOVA’s covariates, model specification, and diagnostics are not reported, so the sponsor analysis cannot be reproduced from the posted group summaries. The sponsor’s own analysis takes precedence, but the available details do not allow an independent assessment of the discrepancy.

The registry record does not report the original sample-size justification, multiplicity strategy, interim-analysis plan, or complete statistical analysis plan. A listed secondary endpoint, longest convulsive seizure-free interval, has no numerical result in the record. These gaps constrain evaluation of confirmatory testing and completeness of outcome reporting.

Finally, the available summary statistics do not show individual participant trajectories or the distribution of changes. The results apply to the defined population, adjunctive treatment comparison, and observation period; they do not establish how the findings would generalize to different populations, treatment contexts, or follow-up durations.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A 2-Part Study to Investigate the Dose-Ranging Safety and Pharmacokinetics, Followed by the Efficacy and Safety of ZX008 (Fenfluramine Hydrochloride) Oral Solution as an Adjunctive Therapy in Children ≥ 2 Years Old and Young Adults With Dravet Syndrome — NCT02926898.
• Registry record last updated 2022-11-02; status last verified 2022-10.
• Results first posted 2022-10-19.
• 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.

23. Record Summary

NCT02926898 is a completed Phase 3, randomized, parallel-group study with quadruple masking and total enrollment of 87. It included an initial pharmacokinetic and safety cohort and a randomized placebo-controlled Cohort 2 evaluation of ZX008 as adjunctive therapy to stiripentol in children and young adults with Dravet syndrome.

The primary Cohort 2 outcome was change in convulsive seizure frequency from baseline to the combined 15-week titration and maintenance period. The sponsor reported an ANCOVA percent difference of -54.0, with a 95% confidence interval from -67.2 to -35.6 and p < 0.001. The unadjusted recalculation from posted means differs; because the discrepancy is flagged, the sponsor’s own analysis takes precedence.

The registry record does not report the full analysis plan, multiplicity strategy, original sample-size justification, or detailed missing-data methods. It lists two secondary outcomes, but a numerical result for longest convulsive seizure-free interval is not reported in the registry record.