Retigabine (Adjunctive Therapy) Efficacy and Safety Study for Partial Onset Refractory Seizures in Epilepsy: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Retigabine (Adjunctive Therapy) Efficacy and Safety Study for Partial Onset Refractory Seizures in Epilepsy (NCT00232596), a randomized phase 3 study. The primary outcome is Percent Change in the 28-day Total Partial Seizure (PS) Frequency From Baseline (BL) to the End of the Double-blind (DB) Phase (Titration and Maintenance Phases).

2. Clinical Question

3. Trial Design

This was a randomized, double-blind, placebo-controlled, multicenter, parallel-group Phase 3 study of retigabine as adjunctive treatment for refractory epilepsy with partial-onset seizures. The registry lists GlaxoSmithKline as sponsor and reports the study as completed. It records a start of 2005-09 and primary completion in 2008-01. Enrollment was 306.

The planned retigabine dose was 1200 mg/day in three equally divided doses. The brief study description says participants were receiving up to three established antiepileptic drugs. Placebo served as the comparator during the double-blind phase. That phase included titration and maintenance, while the primary outcome timeframe spans baseline, identified as Week -7 through Week 0, and the study period identified as Week 1 through Week 18.

The primary outcome analysis describes an Intent-to-Treat population for FDA review as randomized participants who received at least one dose of study drug. It further says that only participants with baseline and post-baseline measures were analyzed. The primary outcome arm counts were 150 for placebo and 151 for retigabine. These are the reported numbers analyzed for that outcome; the registry does not explain all reasons for the difference between enrollment and those outcome counts.

The record lists 20 outcome measures, including the primary seizure-frequency outcome and secondary measures concerning response, seizure-frequency changes, seizure-free status, seizure types, and global impressions. It gives two secondary outcomes with detailed numerical summaries in the data available for this analysis. The protocol’s full allocation procedures, stratification factors, eligibility criteria, and the specific features of the participating centers are not reported in the registry record.

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: Participants who completed the Double-blind Phase (Titration plus Maintenance Phases) who elected to continue in the Open-Label Extension Study (OLE-S) entered the Transition Phase, for titration, if on placebo, or to maintain the blind if on retigabine. Participants who did not enter the Titration Phase entered a Taper Phase.

6-Week Titration Phase

Reasons for not completing this period:

12-Week Maintenance Phase

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Intent-to-Treat (ITT) Population for Food and Drug Administration (FDA) review: all randomized participants (par.) who received at least one dose of the study drug. Only par. with baseline and post-baseline measures were analyzed. Two par. in each treatment arm did not have post-baseline measures and were thus not included in this analysis.

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 Percent Change in the 28-day Total Partial Seizure (PS) Frequency From Baseline (BL) to the End of the Double-blind (DB) Phase (Titration and Maintenance Phases), measured over Baseline (Week -7 through Week 0), Week 1 through Week 18.

The primary endpoint was percent change in 28-day total partial seizure frequency from baseline to the end of the double-blind phase, combining titration and maintenance. Partial seizures are described in the record as seizures limited to a specific area of the brain and also called focal seizures. Frequencies were standardized to a 28-day period by dividing the recorded seizure count by the number of days with available data, then scaling to 28 days. Percent change compares the double-blind-period frequency with the baseline frequency; negative values indicate a reduction.

Using a fixed 28-day scale makes frequency summaries more comparable when the number of days with recorded seizure information varies between periods. The endpoint expresses change relative to each participant’s baseline rather than only reporting a raw count over an observation period. This can make the outcome easier to interpret across participants with different starting frequencies. It remains dependent on baseline values and on the availability and recording of seizure data.

The registry summarizes each arm using a median and full range. A median describes the midpoint of the observed distribution and is less sensitive than a mean to extreme values. The range shows the observed low and high values, but does not indicate how values are distributed between those endpoints. The primary analysis used a non-parametric rank ANCOVA, which compares ranked outcome information rather than relying directly on a normal-distribution model for the raw changes. The record does not report the covariates or other model details.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a median tell the reader? The primary endpoint is reported as a median in each arm, with a full range. The median is the central point of an ordered set of observations: half the observed values lie at or below it and half at or above it, subject to ties. It is useful when outcome values are skewed or include extreme observations, because a few unusually large values have less influence on the median than on the mean. A median is not an average change in the arithmetic sense, and separate arm medians do not themselves constitute an adjusted or unadjusted treatment-effect estimate.

The full range reports the smallest and largest observed values. It conveys the span of observed outcomes, including extremes, but says little about the density or shape of the data between those points. The range can be wide even when many observations cluster near the median. The registry does not provide the primary outcome’s quartiles, standard deviation, or an individual-level distribution, so the median and range are only a partial account of variability.

Why use a rank-based comparison?

The sponsor reports a non-parametric rank ANCOVA. In general, rank-based methods order outcome values and analyze their relative positions rather than treating the raw values as if they followed a specified normal distribution. Such methods can be useful when outcome distributions are skewed or contain extreme observations. ANCOVA denotes a comparison that incorporates covariate information; however, the registry record does not report which covariates were used, how the model was specified, or the assumptions checked. It is therefore not possible to reconstruct the sponsor’s analysis from the posted arm summaries.

A rank-based test result is not automatically an estimate of a difference in medians. Its interpretation depends on the test’s full specification and the data. The reported p-value is p < 0.001, and the record marks the confidence level as 95%, but it does not include the comparison estimate or the confidence-interval limits. General statistical practice treats a confidence interval as a range of effect values compatible with the data and analysis, while communicating uncertainty more directly than a p-value alone. Here, the missing bounds prevent assessment of the interval’s precision.

What does the seizure-frequency scale represent?

The registry converts seizure counts to a 28-day frequency using the count divided by the days with available data, scaled to 28 days. It then expresses change from baseline as a percentage, with negative values indicating a reduction. Standardizing the observation window supports comparisons when data availability differs between periods, but it does not remove the possibility that missing days or differences in recording affect the measure. The record says that only participants with baseline and post-baseline measures were analyzed; it does not describe an imputation method for missing seizure observations.

Percent change also has a baseline value in its denominator. Consequently, the same absolute change in seizure counts can translate into different percentage changes for people with different baseline frequencies. This is a general property of relative-change measures, not a separate result from this trial. The posted endpoint description does not provide an alternative absolute-frequency comparison that would allow readers to examine that scale alongside the percentage change.

How should secondary re-calculations be read?

Some posted secondary comparisons are recalculated from arm-level summary statistics. In general, a simple re-calculation from means and standard deviations is unadjusted: it does not reproduce a sponsor model that may account for covariates, missingness, or other design features. The recalculated comparison for the Clinical Global Impression-Improvement score has a mean difference of -0.5, with a 95% interval from -0.800 to -0.200 and an unadjusted p-value of 0.001. These figures are a summary-statistic calculation, not a sponsor-reported analysis; the registry lists no reported sponsor analysis for that outcome. The record also gives a Patient Global Impression comparison whose estimated mean difference is 0.0 and whose interval runs from -0.318 to 0.318, with an unadjusted p-value of 1.0. That secondary outcome likewise has no reported sponsor analysis in the record.

A confidence interval that includes values on either side of zero does not establish that the true effect is exactly zero. It indicates that the data and calculation do not distinguish zero from the effect values covered by the interval at the stated confidence level. Conversely, a p-value alone does not indicate clinical importance. Interpretation requires attention to the measure, its scale, the estimate, its uncertainty, and the analysis context.

What are the planning illustrations?

The sample-size illustrations use the observed effect for the Clinical Global Impression-Improvement comparison, then consider planning for 100%, 75%, or 50% of that observed effect. They assume a two-sided significance level of 0.05, equal allocation, and a normal approximation. For 80% power, the listed participants per arm are 91, 162, and 363 for those respective effect assumptions; for 90% power, they are 122, 216, and 486. These are planning illustrations, not observed power and not the original trial’s design calculations. The registry explicitly characterizes them as planning based on the observed effect, not the trial’s own design. The assumptions used to plan the actual study are not reported in the registry record.

More generally, a sample-size calculation depends on the target effect, variability assumptions, type I error, power, allocation, and chosen analysis. A calculation based on an effect observed after a trial can be optimistic if that effect is imprecisely estimated. The planning figures should therefore be read as conditional examples, not as evidence that the trial was adequately powered for its primary endpoint or that a future study would necessarily need the listed number of participants.

10. Analyses Reported by the Sponsor

The sponsor’s own primary analysis is the non-parametric rank ANCOVA, which reports p < 0.001 for the comparison of the treatment groups. The registry identifies a 95% confidence level but does not report the sponsor’s effect estimate or confidence-interval bounds. Those missing quantities limit assessment of the comparison’s magnitude and precision, even though the reported test indicates evidence of a group difference under the sponsor’s analysis.

The primary summaries are medians and full ranges. They are descriptive arm-level results and do not reproduce the rank ANCOVA. No recalculated primary comparison is available in the posted results, and the record flags values as missing or not estimable for a two-arm comparison for the other endpoint category. The sponsor’s analysis therefore remains the relevant inferential result for the primary endpoint. The discrepancy indicator is false, so there is no recorded disagreement that would require a recalculated result to be set aside in favor of the sponsor’s analysis.

For the secondary global-impression outcomes, recalculated comparisons are based on the posted group summaries and are unadjusted. The registry lists no sponsor-reported analyses for those outcomes. They should not be presented as if they were the sponsor’s model-based results. Differences in analytic method and adjustment can produce results that are not directly interchangeable, and the record does not provide further detail to reconcile them.

11. Multiplicity, Missing Data and Analysis Populations

The record identifies a primary outcome and lists 20 outcome measures overall, including 18 secondary outcomes posted. It also records two analyses. A planned testing hierarchy, family-wise error procedure, false-discovery approach, or other multiplicity adjustment is not reported in the registry record. It is therefore unclear whether the secondary outcomes were intended as confirmatory tests, exploratory descriptions, or a mixture. When many outcomes are assessed, interpreting each unadjusted p-value as if it were the sole planned test can overstate the strength of evidence. The primary outcome remains the clearest focus because it is explicitly designated as primary, but the record does not provide enough detail to establish how the other analyses were controlled.

The primary population description says that the FDA-review Intent-to-Treat population consisted of all randomized participants who received at least one dose of study drug. It also states that only participants with baseline and post-baseline measures were analyzed. This means the outcome analysis did not include every person merely by virtue of randomization if the required measurements were absent. The registry description notes two participants in each treatment arm without post-baseline measures, but the sentence is incomplete in the record. Further explanation of their handling is not reported in the registry record.

The primary outcome counts are 150 for placebo and 151 for retigabine. These arm counts are the reported analysis counts for this outcome; they do not, by themselves, explain why participants were not included or whether missingness was related to treatment or outcome. No imputation procedure, sensitivity analysis for missing data, or detailed accounting of missing measurements is reported in the registry record. It is therefore not possible to judge from this record whether the analyzed participants represent the randomized groups without bias.

These limitations matter because excluding people with missing measurements can affect estimates if the chance of missingness is associated with seizure outcomes or treatment experience. That is a general statistical concern, not proof that bias occurred in this trial. The registry does not report interim analyses, stopping rules, the analysis-plan version, or a sample-size justification. Those omissions prevent a fuller appraisal of selective testing, missing-data assumptions, and the relationship between planned and posted analyses.

12. 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: psychotic disorder (2.0%), tetany (1.6%), trismus (1.6%), encephalopathy (1.3%), confusional state (1.3%).

The registry reports serious adverse events for the double-blind treatment groups as 8 among 152 participants at risk for placebo and 19 among 153 at risk for retigabine. It also reports other, non-serious adverse events as 91 among 152 for placebo and 121 among 153 for retigabine. These are counts with their reported denominators; they are not the primary efficacy analysis population counts, and the registry’s safety groups should be kept distinct from the outcome groups.

Additional safety groups cover transition-phase treatment: the placebo double-blind group followed by retigabine has 6 serious and 64 other events among 123 at risk, while the retigabine double-blind group followed by retigabine has 5 serious and 35 other events among 92 at risk. These groups represent different treatment periods and follow-up arrangements, so their counts should not be compared as though they arose from the same exposure window. The record does not give deaths; death information is not reported in the registry record.

Serious adverse event terms listed include psychotic disorder at 2.0%, tetany at 1.6%, trismus at 1.6%, encephalopathy at 1.3%, confusional state at 1.3%, and status epilepticus at 1.1%. These percentages are presented in the registry’s serious-event terms listing; the listed figures do not establish causation or a comparative risk. The event counts alone do not describe severity beyond the serious/non-serious classification, duration, resolution, or attribution. Safety interpretation should remain descriptive because no causal assessment or adjusted comparison is reported in the registry record.

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

The posted results support a limited statistical conclusion: the sponsor’s primary rank-based analysis reports a difference between retigabine and placebo in percent change in 28-day partial seizure frequency, with p < 0.001. The arm medians are more negative for retigabine, and negative values indicate reductions. The broad full ranges show that observed outcomes varied and included increases. The registry does not report a primary between-arm estimate or confidence-interval bounds, so the result’s size and precision cannot be fully assessed from the record.

A statistically small p-value is not a measure of treatment benefit, does not describe how large a difference is, and does not say how any individual participant will respond. It addresses how compatible the observed data are with a statistical null under the analysis assumptions. The rank ANCOVA’s exact specification and covariates are not reported, which also limits independent interpretation of the test.

The outcome measures a change in seizure frequency over a defined period, using a 28-day standardization and a percentage scale. It does not, by itself, describe every aspect of epilepsy, participants’ quality of life, or longer-term outcomes. The secondary measures provide other types of information, but the record does not establish a multiplicity strategy, and detailed numerical findings for most listed secondary outcomes are not available in the material.

The safety counts describe events observed in groups with different treatment-period labels and denominators. They should not be read as proof that an event was caused by a study drug, nor do they provide a complete benefit-risk assessment. The record lacks several details needed for that assessment, including death information and a reported causal evaluation.

Overall, these data describe a sponsor-reported primary statistical comparison, variable participant-level outcomes, selected secondary information, and event counts. They do not establish a universal effect, explain individual outcomes, or resolve questions that depend on unreported estimates, analysis details, missing-data handling, and longer-term follow-up. No medical decision should be inferred from the statistical summaries alone.

14. Why This Trial Matters Statistically

This study illustrates why a clinical trial’s reported test result and its effect estimate answer different questions. The primary p-value indicates that the sponsor’s rank-based procedure found evidence of a group difference, but the registry does not give the corresponding between-arm estimate or its confidence-interval limits. Without those, readers cannot judge the precision or practical magnitude of the primary contrast from the record. Reporting an interval alongside an estimate is important because statistical evidence against a null value is not a substitute for quantifying uncertainty.

It also illustrates the distinction between arm summaries and a modeled comparison. Medians and ranges describe the outcome distributions, but they do not reproduce a rank ANCOVA. A comparison of distributions can depend on ranks, covariate adjustment, and model specification. The record names the method but does not provide the covariates or enough detail for independent reconstruction. The finding therefore depends on an analysis whose essential headline result is available but whose full inferential detail is not.

The endpoint itself has statistical features worth noting. It standardizes seizure frequency to 28 days and then expresses change as a percentage of baseline. This supports a common time scale but makes interpretation dependent on baseline frequency and available observation days. The wide ranges signal substantial outcome variation, reinforcing the value of distribution-aware methods and transparent uncertainty reporting.

Finally, the record lists many secondary measures, while the multiplicity plan is not reported. This is a common challenge in clinical research: extensive outcome assessment can enrich description, but unadjusted tests across many measures require cautious interpretation. The trial is statistically informative in showing how a planned primary analysis, descriptive summaries, secondary outcomes, missing-data handling, and safety reporting must be considered together rather than treating a p-value as a complete account.

15. 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 result lacks a reported between-arm effect estimate and confidence-interval bounds. Although the registry gives the confidence level as 95% and reports p < 0.001, readers cannot assess the primary comparison’s interval width or identify the effect values compatible with the analysis. The arm medians and ranges do not fill that gap because they are descriptive and the sponsor’s test is rank based.

The analysis description is incomplete for independent appraisal. The record names non-parametric rank ANCOVA but does not state the covariates, model details, assumptions, or handling of ties and missing observations. The described analysis population requires a dose and baseline and post-baseline measurements, and the record does not report an imputation method or a sensitivity analysis for missing outcomes. The note about participants without post-baseline measurements is itself incomplete.

Multiplicity procedures and a prespecified hierarchy for the secondary outcomes are not reported. Most detailed secondary numerical findings are not available in the registry record for this explanation, so consistency across the full outcome set cannot be assessed. The global-impression comparisons are recalculated and unadjusted, with no corresponding sponsor analyses listed in the record. They are not interchangeable with the sponsor’s primary analysis.

Safety reporting is also limited: the groups span different treatment periods, and the registry does not report deaths. The presented counts do not supply a causal assessment or the detail needed to compare exposure-adjusted risk. The record does not report full eligibility criteria, baseline characteristics, subgroup analyses, the trial’s own sample-size assumptions, interim analyses, or detailed follow-up beyond the specified outcome timeframe. These omissions constrain conclusions about generalizability, robustness, and the planned analysis process.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Retigabine (Adjunctive Therapy) Efficacy and Safety Study for Partial Onset Refractory Seizures in Epilepsy — NCT00232596.
• Registry record last updated 2016-12-08; status last verified 2016-10.
• Results first posted 2011-10-25.
• 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.

19. Record Summary

NCT00232596 is a completed randomized, double-blind, placebo-controlled, multicenter, parallel-group Phase 3 study sponsored by GlaxoSmithKline. It evaluated adjunctive retigabine at 1200 mg/day, given in three equally divided doses, in refractory epilepsy with partial-onset seizures. Enrollment was 306; the recorded study start was 2005-09 and primary completion was 2008-01.

The primary outcome was percent change in 28-day total partial seizure frequency from baseline through the double-blind titration and maintenance phases. The reported medians were -17.5 for placebo and -44.3 for retigabine, with full ranges of -90.0 to 628.0 and -100.0 to 302.0, respectively. The sponsor’s non-parametric rank ANCOVA reports p < 0.001. A primary comparison estimate and confidence-interval bounds are not reported.

The registry lists 20 outcomes, including 18 secondary outcomes posted. Its record does not provide a multiplicity plan, detailed missing-data methods, or the sponsor’s primary effect estimate and interval. Safety counts are reported for double-blind and transition-phase groups, with deaths not reported in the registry record.