Efficacy and Safety of Diazepam in the Management of Refractory Epilepsy in Selected Patients Who Require Intermittent Medical Intervention for Acute Repetitive Seizures.: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Efficacy and Safety of Diazepam in the Management of Refractory Epilepsy in Selected Patients Who Require Intermittent Medical Intervention for Acute Repetitive Seizures. (NCT00319501), a randomized phase 3 study. The primary outcome is Time to Next Seizure or Rescue Medication During the Double-blind Period (Kaplan-Meier 50th Percentile).

2. Clinical Question

The trial also examined rescue medication and medical-care use, global assessments by caregivers and physicians, and safety. Its stated purpose included assessing support provided by caregivers who were not themselves, or were not under the direct supervision of, health care professionals at administration. This makes the setting important to understanding what the study was intended to evaluate: intermittent response to acute repetitive seizures outside direct professional supervision, rather than ongoing management of epilepsy.

The record identifies epilepsy and several epilepsy-related conditions, including partial, complex partial and generalized epilepsy. It does not describe a broader population beyond the selected refractory patients specified in its study description. The findings therefore concern the trial's stated population and double-blind comparison; they should not automatically be generalized to people with other seizure patterns or care circumstances.

3. Trial Design

This was a phase 3, randomized, parallel-group, placebo-controlled study with quadruple masking. The registry lists an enrollment of 234 and describes two double-blind treatment groups: diazepam and placebo. The stated study period also included an optional open-label continuation. The record lists Pfizer as sponsor and gives a start of 2006-01, primary completion of 2014-07, and completed status.

For the double-blind period, the arm descriptions specify a single age- and weight-appropriate injection for an acute repetitive seizure episode. Diazepam was administered at 0.2 to 0.5 mg/kg; the placebo description does not specify an active dose. The descriptions state that a caregiver used a spring-driven, pressure-activated, prefilled autoinjector, with injection into the mid to outer thigh. The diazepam arm description says additional doses were permissible during the open-label period. More detail about the open-label dosing procedures is not reported in the registry record.

The posted primary results use 82 participants in the diazepam group and 81 in the placebo group for the endpoint. These are the group counts attached to that outcome, not a statement that every person in the registry's enrollment contributed to the primary analysis. The primary population was restricted to randomized participants for whom an attempt was made to administer study drug for an acute repetitive seizure episode during the double-blind period.

The record lists 12 outcome measures, six analyses and 10 posted secondary outcomes. Its description of allocation and masking establishes the broad trial structure, but it does not report the sequence-generation procedure, allocation concealment details, or exactly which roles were included in quadruple masking. The registry record also does not report the planned sample-size justification or a complete statistical analysis plan. Those omissions limit reconstruction of the trial's design rationale and analysis decisions from the record alone.

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: Of 234 participants randomized, 71 (29 placebo, 42 diazepam) did not receive/attempt to receive treatment. Discontinuation prior to treatment due to: 3,0 protocol violations; 6,5 sponsor's request; 4,8 Principal Investigator request; 5,11 other; 0,2 lost to follow-up, and 9,10 withdrawn consent.

Double-blind Period

Reasons for not completing this period:

Open-label Period

Reasons for not completing this period:

Open-label Extension Period

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All randomized participants for whom an attempt (successful or not) was made to administer study drug for an ARS event during the Double-blind Period.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: Participants who were randomized to receive treatment

6. Primary Endpoint

The registry lists the primary outcome as Time to Next Seizure or Rescue Medication During the Double-blind Period (Kaplan-Meier 50th Percentile), measured over From 15 minutes to 12 hours after study drug administration for an episode of ARS during the Double-blind Period.

The primary endpoint was time to the next seizure or rescue medication during the double-blind period, reported using the Kaplan-Meier 50th percentile. An event was defined as an acute repetitive seizure episode or the need for rescue medication for such an episode within 15 minutes to 12 hours after study drug administration. Participants without an event were censored at 12 hours.

Time-to-event outcomes retain information about when an event occurs, rather than reducing the result to whether an event occurred at any point in a fixed period. The Kaplan-Meier method estimates the event-free experience over follow-up while allowing for censoring: a participant contributes information up to the point their observation ends, even if the event is not observed. The median is the time by which the estimated event-free probability has reached its midpoint. If the curve does not support that point within the observed follow-up, a median may not be estimable.

The endpoint combines recurrent seizure activity and rescue medication into one event definition. It therefore captures either a seizure recurrence or a clinical response indicating inadequate seizure control that led to rescue medication, as described in the record. It does not isolate these components as separate outcomes. Rescue diazepam rectal gel was allowed only in the first 4 hours and only when the investigator or designee directed the caregiver to administer it.

Diaries recorded seizures and rescue medication during follow-up. If a diary was not returned or did not answer the relevant questions, the stated censoring rule was 15 minutes past treatment time unless another contact was documented. These rules affect how much follow-up time contributes to the estimated event-time distribution.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is a time-to-event measure. For each participant, the relevant duration begins with study drug administration and ends when the defined event occurs or the participant is censored. In this trial an event was an acute repetitive seizure or required rescue medication within 15 minutes to 12 hours. The Kaplan-Meier estimator is a general method for describing the probability of remaining event-free over time when follow-up lengths differ. It incorporates participants' observed time up to an event or censoring point, rather than treating every censored participant as if an event had occurred.

The Kaplan-Meier 50th percentile is the median event time on the estimated curve. It is not necessarily the ordinary median calculated from a list of fully observed event times. Censoring means that the exact event time is unknown for some participants, but their known event-free time can still contribute to the estimate. A median can be difficult or impossible to estimate when the curve does not reach the midpoint during the available follow-up. In this record the diazepam median was 5.92 hours, with lower limit 2.28 hours; the upper limit and placebo median are not reported in the registry record.

The registry also reports a Cox proportional-hazards analysis. In general, a Cox model compares event hazards between groups over time while allowing the underlying baseline hazard to remain unspecified. Its hazard ratio is a relative measure of the instantaneous event rate among participants still under observation and event-free. The reported hazard ratio of 0.55 compares diazepam with placebo for the defined event. A hazard ratio is not a ratio of medians, nor does it directly state how many participants experienced an event. The usual interpretation assumes that the relative hazard is reasonably stable over the modeled follow-up; whether that assumption was assessed is not reported in the registry record.

A confidence interval expresses uncertainty around an estimate under the method used to construct it. The reported 95% interval for the hazard ratio, 0.34 to 0.88, gives a range of values compatible with the observed data under the model and its assumptions. It is not a guarantee that the true value lies in that range, and it does not capture every source of uncertainty, such as limitations in study conduct or reporting. The p-value of 0.012 summarizes how unusual a result at least as incompatible with a no-association null hypothesis would be under that null and the analysis assumptions. It is not the probability that the null is true, nor a measure of clinical importance.

The secondary outcomes include counts of participants with binary events, such as requiring rescue medication or emergency department care. A risk difference compares the event proportions between groups by subtraction, expressed as an absolute difference. The recalculated values are simple unadjusted comparisons from the posted counts. A negative risk difference in the displayed diazepam-versus-placebo direction means the observed proportion was lower in the diazepam group. Its confidence interval conveys uncertainty; if the interval includes zero, the data are compatible with no difference as well as with differences in either direction. Such calculations do not adjust for covariates or for multiple outcome testing.

For rare binary events, the choice of statistical test matters. The sponsor's reported secondary analyses use Fisher Exact tests for the outcomes with posted p-values. Fisher's test is often used for categorical data when event counts are sparse. The recalculated unadjusted p-values arise from a different calculation, so small discrepancies between them and Fisher Exact results are possible without either number being a transcription error. The registry record does not report a detailed rationale for test selection or a complete prespecified analysis plan.

General statistical explanation: a confidence interval for a risk difference and a hypothesis test address related but not identical questions. The interval gives a range of plausible absolute differences under its method, while a p-value measures compatibility with a particular null hypothesis. Different methods can yield different endpoints or p-values, especially with sparse counts. For interpretation, the event counts and the width of the uncertainty interval matter in addition to whether a p-value crosses a conventional threshold.

The record also includes planning illustrations based on observed secondary effects. Such illustrations describe hypothetical participants per arm for target power under specified assumptions; they are not observed power and do not show that the original study was designed around those effects. The stated assumptions include a two-sided significance level of 0.05, equal allocation, a normal approximation, and detection of a stated fraction of the observed effect. The registry's own sample-size justification is not reported in the registry record.

Over-dispersion is a general issue when count data vary more than a model assumes, potentially making conventional standard errors too small. It is not possible to conclude that over-dispersion affected this trial's primary analysis from the information reported: the primary outcome is time-to-event, and the registry record does not report a count-model analysis. More broadly, statistical methods must match outcome type, follow-up and data completeness. The posted descriptions identify the principal methods and endpoint rules, but do not include enough detail to reproduce every analysis decision.

10. Analyses Reported by the Sponsor

The sponsor-reported primary analysis is the Cox proportional-hazards comparison: hazard ratio 0.55, 95% confidence interval 0.34 to 0.88, and p = 0.012. This is the reported inferential analysis for the primary outcome and should be distinguished from any simple recalculation based on summary statistics. No recalculated primary comparison is posted in the record. The diazepam median is reported as 5.92 hours with a lower limit of 2.28 hours, but the placebo median and upper confidence limit for diazepam are not reported in the registry record.

For the secondary binary outcomes, the posted sponsor analyses use Fisher Exact tests. Simple unadjusted recalculations from the arm counts may give different p-values because they need not use the same testing method. For example, the registry's Fisher Exact p-values for rescue medication, emergency department visits and other rescue medical care are 0.066, 0.443 and 0.245, respectively; the corresponding recalculated unadjusted p-values are 0.058, 0.397 and 0.152. These are different analyses of the same posted counts, not conflicting trial datasets. The reported sponsor analyses take precedence when describing the trial's own inferential results. The record does not report whether secondary tests were adjusted for multiplicity.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 12 outcome measures, six analyses, and 10 posted secondary outcomes. It does not report a multiplicity adjustment, a hierarchy for interpreting secondary endpoints, or whether any secondary analyses were designated as confirmatory. Without those details, the chance of at least one apparently notable result across multiple tests cannot be assessed from the record. The secondary p-values should therefore be interpreted cautiously and not as if each test were the only analysis conducted.

The primary endpoint description gives explicit censoring rules. Participants without an acute repetitive seizure or rescue medication event were censored at 12 hours. If a diary was not returned, or did not answer questions about seizures and rescue during follow-up, the stated rule was censoring at 15 minutes past treatment time unless another contact was documented. This is important because censoring assumes, in general, that the reason observation ends does not create an unaccounted relationship with the event process. The registry record does not report how many participants were censored under each rule or whether sensitivity analyses examined the effect of incomplete diaries.

The primary population included randomized participants for whom an attempt, successful or not, was made to administer study drug for an acute repetitive seizure during the double-blind period. This is the population specified for the reported endpoint; it should not be assumed to be identical to all randomized participants. Reasons for exclusion from this endpoint population and details of other analysis populations are not reported in the registry record. Nor does the record describe a full approach to missing secondary outcome data.

The record does not report interim analyses, stopping rules, a detailed sample-size justification, or a complete statistical analysis plan. It is therefore not possible to determine from the registry alone how endpoint hierarchy, missing data, or multiplicity were handled beyond the stated endpoint definitions and analyses. These omissions do not alter the posted numerical results, but they limit assessment of how fully the reported analyses reflect the trial's planned decision process.

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 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: epilepsy (15.2%), gait disturbances (3.0%), pneumonia (3.0%), status epilepticus (3.0%), postictal state (3.0%).

The safety results are listed separately for the double-blind and open-label periods. In the double-blind period, serious adverse events were reported for 2 participants at risk in the diazepam group and 3 participants at risk in the placebo group. Other adverse events were reported for 33 participants at risk in the diazepam group and 32 participants at risk in the placebo group. The listed at-risk denominators are 81 for each of those double-blind groups.

For the open-label period, the record lists 20 serious adverse events among 128 participants at risk and 43 other adverse events among 128 at risk. In the open-label extension, it lists 11 serious adverse events among 33 at risk and 20 other adverse events among 33 at risk. These open-label data do not have a concurrent placebo comparison in the posted safety groups and should not be read as randomized comparative estimates.

The most frequently listed serious adverse-event terms include epilepsy at 15.2%, and gait disturbances, pneumonia, status epilepticus, postictal state and corneal abrasion at 3.0% each. The record does not identify these terms as causal effects of study drug. Death counts are not reported in the registry record. Safety counts describe events observed and recorded in the listed periods; they do not, on their own, establish that an event was caused by treatment or quantify comparative risk with precision.

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

The primary analysis reports a lower estimated hazard of the combined event of seizure or rescue medication with diazepam than with placebo during the analyzed follow-up. The reported hazard ratio is 0.55, with a 95% confidence interval from 0.34 to 0.88 and p = 0.012. The diazepam Kaplan-Meier median is 5.92 hours, with a lower confidence limit of 2.28 hours. Because the placebo median and the upper confidence limit for the diazepam median are not reported in the registry record, the posted medians do not provide a complete group-by-group summary.

This result does not mean that every participant had a longer interval without a seizure or rescue medication. A group-level hazard estimate describes an average pattern over time under the analysis model; individual outcomes can vary. Nor does the combined endpoint establish which component—seizure recurrence or rescue medication—accounts for the result. Those components are not separately quantified in the primary result described in the record.

The reported association does not by itself establish clinical importance, long-term effects, or applicability to people outside the selected refractory-epilepsy population. The double-blind follow-up for this primary measure extended from 15 minutes to 12 hours after study drug administration. It therefore addresses a defined acute period, not ongoing seizure control over a longer duration. The study also included an optional open-label continuation, but open-label observations do not provide the same controlled comparison as the double-blind period.

Finally, a statistical result is conditional on how the endpoint was defined, who contributed data, and how censoring and analysis assumptions were handled. The record reports a primary Cox analysis and endpoint-specific censoring rules, but does not report a complete plan for multiplicity, missing-data sensitivity analyses, or every model diagnostic. The evidence should be understood in that context, rather than as a guarantee about outcomes for an individual or a conclusion beyond the recorded study question.

14. Why This Trial Matters Statistically

This trial is statistically notable because it combines a time-to-event primary endpoint with an acute, caregiver-administered intervention. Time-to-event analysis uses information about when a seizure or rescue-medication event occurs, while accommodating participants whose observation ends without a recorded event. That is a different estimand from comparing only the number of participants with an event by a fixed cutoff. Interpreting the hazard ratio alongside the Kaplan-Meier median helps distinguish a relative event-rate comparison from a summary of event timing.

The endpoint also combines two types of events: acute repetitive seizure and rescue medication. This is clinically relevant to the study's stated objective but statistically means the primary estimate cannot show which component drives the association. The follow-up and diary rules matter as well: incomplete diaries could lead to early censoring unless another contact was documented. This makes documentation and the assumptions behind censoring central to understanding the estimate.

The secondary outcomes illustrate another common issue in clinical trials: sparse event counts can lead to uncertain estimates, and the sponsor's Fisher Exact tests need not match simple recalculations from summary counts. Reporting absolute differences and confidence intervals gives readers a sense of scale and precision, while multiplicity cautions against treating every unadjusted secondary p-value as confirmatory.

The record is informative about the broad randomized, masked, parallel design and the main endpoint analysis, but important elements of the analysis plan are not reported. This gap limits a full audit of endpoint hierarchy, missing-data sensitivity and multiplicity control. The trial therefore offers a useful example of how a statistically significant primary comparison can coexist with incomplete reporting of details needed to judge robustness and the wider pattern across outcomes.

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 outcome is a composite time-to-event measure, combining a seizure episode and required rescue medication. The posted hazard ratio cannot separate these components. The placebo median and upper confidence limit for the diazepam median are not reported in the registry record, which limits interpretation of the group medians alongside the hazard-ratio comparison.

The primary endpoint relied on diaries and specified censoring rules, including early censoring when a diary was missing or lacked relevant responses unless another contact was documented. The record does not report how many participants were affected by each missing-diary circumstance, or the results of sensitivity analyses. Consequently, the effect of incomplete endpoint follow-up cannot be evaluated from the posted information.

Although the study is described as randomized and quadruple masked, details about sequence generation, allocation concealment, and who was masked are not reported in the registry record. The record also does not report a complete statistical analysis plan, planned sample-size justification, interim-analysis procedures, or multiplicity adjustment. These omissions make it harder to assess whether all analysis choices were prespecified and how much confidence to place in secondary findings.

The follow-up for the primary endpoint was limited to the stated window after administration, and the population was selected for refractory epilepsy requiring intermittent intervention. The result should not be generalized automatically beyond that population and setting. Open-label continuation data do not offer a concurrent randomized placebo comparison. Finally, several secondary outcomes have sparse events, which limits precision; the posted counts and methods do not establish causal explanations for particular events or differences in care use.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Efficacy and Safety of Diazepam in the Management of Refractory Epilepsy in Selected Patients Who Require Intermittent Medical Intervention for Acute Repetitive Seizures. — NCT00319501.
• Registry record last updated 2016-09-23; status last verified 2016-07.
• Results first posted 2016-08-16.
• 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

The registry describes a completed phase 3, randomized, quadruple-masked, parallel, placebo-controlled study of diazepam injection in selected patients with refractory epilepsy who required intermittent intervention for acute repetitive seizures. It lists an enrollment of 234 and an optional open-label continuation.

The primary outcome was time to seizure or rescue medication during the double-blind period. The reported Cox analysis gave a hazard ratio of 0.55, 95% confidence interval 0.34 to 0.88, and p = 0.012. The diazepam median was 5.92 hours, with a lower limit of 2.28 hours; a placebo median and the diazepam upper limit are not reported in the registry record.

Secondary outcomes and safety results provide additional descriptions, but some event counts are sparse and multiplicity handling is not reported. The registry record also omits several details needed to fully assess the planned analysis and missing-data sensitivity.