Dabigatran Etexilate for Secondary Stroke Prevention in Patients With Embolic Stroke of Undetermined Source (RE-SPECT ESUS): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Dabigatran Etexilate for Secondary Stroke Prevention in Patients With Embolic Stroke of Undetermined Source (RE-SPECT ESUS) (NCT02239120), a randomized phase 3 study. The primary outcome is Adjudicated Recurrent Stroke.

2. Clinical Question

3. Trial Design

RE-SPECT ESUS was a phase 3 randomized, double-blind trial sponsored by Boehringer Ingelheim. Its title describes a comparison of dabigatran etexilate and acetylsalicylic acid (ASA) for secondary stroke prevention in people with embolic stroke of undetermined source. The registry record lists enrollment of 5390 and two treatment arms: dabigatran etexilate 110 or 150 mg and ASA 100 mg. Each arm lists 2695 participants for the primary outcome.

The brief summary describes a planned allocation ratio of 1:1 and visits every three months. It describes enrollment of approximately 6,000 patients, while the registry record reports enrollment of 5390. The summary says the trial would conclude when the required number of stroke events had been positively adjudicated; the estimated duration in that summary was 3 years, including 2.5 years of enrollment. The primary outcome timeframe is from randomisation until full follow-up, approximately 43 months.

The arm descriptions specify twice-daily dabigatran and once-daily ASA. The dabigatran dose was 150 mg twice daily except for participants meeting the listed age, creatinine-clearance, or gastrointestinal-bleeding conditions, for whom the dose was 110 mg twice daily. The ASA arm used blinded ASA 100 mg once daily and a matching dabigatran placebo. Optional blinded concomitant ASA for coronary artery disease is described in the brief summary, but was not mandatory.

The record lists a start date of 2014-11-27, primary completion on 2018-08-14, and completed status. It also lists nine analyses and ten outcome measures. Detailed information about the randomisation procedure, concealment, visit schedule beyond the brief-summary description, or the protocol’s sample-size justification is 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: All participants were screened for eligibility to participate in the trial. Participants attended specialist sites which would then ensured that all participants met all inclusion/exclusion criteria. Participants were not to be randomized to trial treatment if any one of the specific entry criteria were not met.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Randomised set (RS): RS consisted of all participants who were randomised, regardless of whether they took trial medication. The start date of the observation period for this population was the date of randomisation.

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: Randomised set (RS): The RS consisted of all patients who were randomised, regardless of whether they took trial medication. The start date of the observation period for this population was the date of randomisation.

6. Primary Endpoint

The registry lists the primary outcome as Adjudicated Recurrent Stroke, measured over From randomisation until full follow up period, approximately 43 months..

The primary endpoint was adjudicated recurrent stroke, measured as an annualised event rate (%/ year). The registry defines recurrent stroke to include ischemic, hemorrhagic, or unspecified stroke. It describes the annualised rate as the average number of events per patient during a one-year period. The timeframe ran from randomisation until full follow-up, approximately 43 months, and the analysis population was the randomised set, regardless of whether participants took trial medication.

Adjudication means that outcome events are assessed under a specified process rather than being based solely on an investigator’s routine recording. In general, an adjudicated endpoint can help apply consistent event definitions across treatment groups. The registry record does not report the detailed adjudication procedures, the adjudicators’ masking arrangements, or rules for handling recurrent events.

Stroke is the trial’s central outcome because the study concerns secondary stroke prevention. The registry also lists adjudicated ischemic stroke, disabling stroke, all-cause death, intracranial hemorrhage, fatal and life-threatening bleeding, and a composite cardiovascular outcome among secondary outcomes. These other measures describe additional clinical events and safety outcomes, but they do not replace the stated primary endpoint. A rate averaged over follow-up and a hazard ratio answer related but different questions: the rate describes event frequency over time, while the hazard ratio compares the relative event hazard between groups under the Cox model.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is a count-type event measure expressed as an annualised event rate. The registry describes the annualised rate as the average number of events per patient during a one-year period. This provides a way to describe event frequency while follow-up extends over time. It is not the same as the proportion of participants who had at least one stroke, because a participant-level proportion counts people with an event, whereas an event rate can account for event counts over person-time. The registry does not report the person-time calculation or rules for repeat events.

What does the Cox model estimate?

The reported group comparison used Cox regression and is expressed as a hazard ratio. In general, the hazard is the instantaneous event rate among participants who remain under observation and have not yet had the event. A Cox model compares hazards over time without requiring the baseline hazard to follow a particular parametric shape. Its hazard ratio is a relative comparison, not a difference in annualised event rates and not a ratio of the two displayed rates calculated directly from those summaries.

The registry reports a hazard ratio of 0.85, a 95% confidence interval of 0.69 to 1.03, and p = 0.103. A confidence interval describes a range of estimates compatible with the observed data and model at its stated confidence level. In repeated use under the model’s assumptions, intervals constructed in this way have the stated coverage. It is not a probability statement that the true value lies in this particular interval. Because this interval includes 1, the comparison does not exclude equal hazards at the corresponding two-sided confidence level.

The Cox model relies on assumptions, including proportional hazards for the usual single hazard-ratio interpretation. In general, proportional hazards means the relative hazard is stable over the time considered. If relative hazards change materially over time, a single hazard ratio may not describe the pattern well. The registry record does not report diagnostics or other information about this assumption, nor does it describe the censoring method or model covariates for this analysis.

How should the annualised rates be read?

The rates are reported as 4.09%/ year in the dabigatran group and 4.8%/ year in the ASA group. They describe average event frequency per patient during a one-year period according to the registry’s definition. In general, annualisation allows event experience across follow-up periods of different lengths to be represented on a common time scale. It does not mean that every participant was followed for exactly one year, nor does it imply that the event rate was constant during follow-up. The record does not provide rate confidence intervals or further computational details in the primary outcome entry.

Comparing annualised rates directly is not a substitute for the reported Cox analysis. The rate measure describes average event frequency; the Cox model compares the timing-related hazard. Differences in estimand and modeling mean that the values need not line up as if they were two versions of one calculation. The record’s primary result is the Cox comparison, while the group rates give descriptive context.

What can a recalculation add?

For some outcomes, a simple unadjusted recalculation from group summaries can offer a check on a reported comparison. It is not the same as reproducing an analysis that uses individual participant event times, censoring, covariates, or adjudication details. This endpoint is identified as a count-other outcome, and values are missing or not estimable for a two-arm comparison of that type. No recalculated two-arm result is therefore available for this endpoint in the registry information. The reported Cox analysis remains the relevant posted comparison; there is no discrepancy flag indicating a conflict with a recalculated result.

The sample-size planning illustration is conceptually separate from the observed result. In general, such an illustration uses assumptions about significance level, allocation, and an effect to detect; it describes hypothetical planning, not the power observed in the completed trial. The stated illustration assumptions are a two-sided significance level of 0.05, equal allocation, a normal approximation, and an effect defined as a fraction of the observed effect. For time-to-event outcomes, the stated planning approach concerns total events; other outcomes use participants per arm. Specific planning values and the trial’s own sample-size justification are not reported in the registry record. The illustration should not be interpreted as observed power or as evidence that the trial was designed around a particular effect.

Why does over-dispersion matter for count outcomes?

In general, count data can vary more than a simple model expects, a property called over-dispersion. If a count model assumes less variability than is present, uncertainty estimates may be too small and statistical tests may be too optimistic. An appropriate analysis may account for that extra variation, but the choice depends on how events, follow-up, and repeated outcomes are represented. The registry’s primary comparison is reported as Cox regression, not as a count-regression model. The record does not report a separate over-dispersion assessment, and the existence of a count endpoint alone does not show that over-dispersion occurred.

10. Analyses Reported by the Sponsor

The registry reports the trial comparison as a Cox regression hazard ratio of 0.85, with a 95% confidence interval of 0.69 to 1.03 and p = 0.103. This is the sponsor-reported analysis listed for adjudicated recurrent stroke. Its interval includes 1, and its p-value is above the conventional 0.05 threshold. Those results describe the uncertainty in the reported superiority comparison.

The endpoint is classified as a count-other outcome, for which values are missing or not estimable for a two-arm recalculation. Accordingly, there is no separate recalculated comparison to set against the Cox result. A simple recalculation from group summaries would in any event not reproduce a time-to-event analysis using participant-level event timing and censoring. The registry does not report the underlying participant-level data or additional model details needed to reconstruct that analysis independently.

The discrepancy indicator is false, so the registry does not flag a disagreement between the sponsor analysis and a recalculation. This does not mean that an independent recalculation confirms the Cox model; no such two-arm recalculation is available for this endpoint. The sponsor’s reported analysis should therefore be read as the available inferential comparison, alongside the annualised rates and their stated limitations.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists nine analyses and ten outcome measures, including the primary recurrent-stroke endpoint and secondary outcomes such as adjudicated ischemic stroke, disabling stroke, all-cause death, intracranial hemorrhage, bleeding outcomes, and a composite cardiovascular outcome. When a trial examines multiple outcomes or makes multiple comparisons, the chance of at least one apparently unusual result can increase. In general, multiplicity procedures may control that risk, but the appropriate approach depends on which outcomes and comparisons were planned and how they relate.

Whether the trial used a multiplicity adjustment, a hierarchical testing procedure, or another strategy is not reported in the registry record. The p-value for the primary comparison should therefore be understood as the reported result, without assuming that it was adjusted or unadjusted for other analyses. The registry lists the analysis as a superiority comparison but does not provide a detailed statistical-analysis plan in the registry record.

The primary analysis population was the randomised set, defined as all randomised participants regardless of whether they took trial medication. The observation period began at randomisation. This definition is reported, but other analysis populations and their roles are not described for the primary outcome. The registry record does not report how missing outcome information, loss to follow-up, treatment discontinuation, or censoring was handled. It also does not report sensitivity analyses for missing data or alternative analysis populations.

These omissions matter because missing follow-up and censoring can affect estimates in time-to-event analyses, while different analysis populations can answer different questions. The randomised-set definition identifies who was intended to contribute, but it does not by itself explain how incomplete observations were treated. The record also does not report interim analyses, stopping rules, or whether the reported analysis was prespecified in a dated analysis plan. Those details should not be inferred from the trial’s completed status or from the reported p-value.

12. 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: cerebrovascular accident (3.6%), ischaemic stroke (1.8%), transient ischaemic attack (1.7%), cerebral infarction (1.6%), atrial fibrillation or atrial flutter (1.3%).

The registry reports serious adverse events among 2676 participants at risk in the dabigatran arm, with 724 affected, and among 2674 participants at risk in the ASA arm, with 740 affected. For other adverse events, it reports the same at-risk counts, with 306 affected in the dabigatran arm and 295 in the ASA arm. The deaths field lists 56 in the dabigatran group and 58 in the ASA group.

These are posted counts, not adjusted comparisons. The record does not give a statistical comparison, confidence intervals, or a detailed definition of the adverse-event risk period in the registry record describing these figures. The serious adverse-event terms listed include cerebrovascular accident (3.6%), ischaemic stroke (1.8%), transient ischaemic attack (1.7%), cerebral infarction (1.6%), atrial fibrillation or atrial flutter (1.3%), and epilepsy (0.9%). The registry does not specify the denominator or group attribution for those term percentages in the listed information, so they should not be interpreted as arm-specific rates.

In general, adverse-event counts describe events recorded under the trial’s reporting definitions; they do not alone establish that a treatment caused an event or that groups differ. The record says adverse events and serious adverse events would be recorded, but detailed coding, exposure-adjusted analyses, and causal assessments are not reported in the registry record.

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

The posted primary result compares adjudicated recurrent stroke between participants assigned to dabigatran etexilate and those assigned to ASA. The annualised rates are 4.09%/ year and 4.8%/ year, respectively. The Cox model reports a hazard ratio of 0.85, with a 95% confidence interval of 0.69 to 1.03 and p = 0.103. Together, these figures describe a point estimate below 1 with uncertainty that includes 1.

That pattern does not establish that the treatments are equivalent. Equivalence requires a prespecified margin and an analysis designed to assess whether any difference falls within that margin. The registry’s listed analysis is a superiority comparison; an equivalence margin is not reported in the registry record. Nor does a p-value above 0.05 prove that there is no difference. It indicates that the result did not meet the conventional threshold for statistical significance under the reported comparison.

The interval also does not show that a clinically important difference is absent. Interpretation of clinical importance would require a prespecified threshold and relevant clinical context, neither of which is reported for this result. The annualised rates are descriptive; they are not a substitute for the Cox model or a complete account of how risk changes over follow-up.

The result is specific to the trial’s population, endpoint definition, follow-up, and analysis population. It concerns participants with embolic stroke of undetermined source and evaluates adjudicated recurrent stroke, including ischemic, hemorrhagic, or unspecified stroke. The registry does not report enough detail to assess every modeling and missing-data assumption independently. The listed adverse-event counts provide additional descriptive safety information, but they do not resolve the uncertainty around the primary comparison.

In short, the registry reports an estimate favoring a lower relative hazard for dabigatran in the direction of the point estimate, but the confidence interval includes equal hazards and the p-value is 0.103. This statistical summary does not establish that one treatment is preferable, that the treatments have the same effect, or that no clinically important effect exists. It is not medical advice.

14. Why This Trial Matters Statistically

This trial illustrates why an endpoint’s scale and analysis method matter. Recurrent stroke is reported as an annualised event rate, while the primary group comparison is a Cox-model hazard ratio. The rate gives a summary of average event frequency per patient during a one-year period; the hazard ratio compares relative hazards over time. They provide complementary descriptions, but they are not interchangeable measures and should not be compared as though one were a direct recalculation of the other.

The result also demonstrates the importance of reading an estimate together with its uncertainty. The hazard ratio is 0.85, but the 95% confidence interval extends from 0.69 to 1.03, and p = 0.103. The estimate alone might invite an overly simple directional interpretation. The interval and p-value make clear that the data do not exclude equal hazards under the reported model and do not meet a conventional 0.05 significance threshold.

Randomisation and double masking are design features relevant to how a treatment comparison is conducted, while the analysis population defines who is included in the primary outcome analysis. Here, the randomised set includes all randomised participants regardless of whether they took trial medication. These design descriptions help readers understand the comparison, but they do not remove uncertainty or replace information about follow-up, missing data, censoring, and model assumptions.

Finally, this trial is a useful example of the boundary between a reported inferential analysis and a simple summary-based recalculation. The endpoint is a count-other measure for which a two-arm recalculation is not estimable in the registry information. The Cox result is therefore the available reported comparison, and it should not be presented as independently reproduced from the annualised rates. Clear separation of descriptive rates, model-based comparisons, and planning illustrations helps prevent those quantities from being mistaken for one another.

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 registry record provides the primary rates, hazard ratio, confidence interval, p-value, endpoint definition, and randomised-set description, but it does not report the detailed statistical-analysis plan. The model’s covariates, proportional-hazards checks, censoring rules, handling of repeat strokes, and treatment of missing outcome data are not reported in the registry record. These details limit independent assessment of how well the model represents the observed event process and how sensitive the estimate may be to analytic choices.

The annualised rates are listed without uncertainty intervals in the primary outcome details. No participant-level data or event-time information is available in the record, and the endpoint is not estimable for a two-arm summary-based recalculation. The Cox result therefore cannot be independently reproduced from the posted group rates alone. The discrepancy indicator is false, but that does not supply a separate recalculated estimate.

The registry lists multiple outcomes and analyses, but whether the primary p-value was adjusted for multiplicity, or whether testing followed a hierarchy, is not reported in the registry record. The record also does not report a sample-size justification, interim-analysis procedures, stopping boundaries, or a detailed protocol for event adjudication. The brief summary describes concluding the trial after the required number of stroke events had been positively adjudicated, but the required number is not stated in the registry record.

For safety, the registry gives counts of affected participants and selected serious adverse-event terms, but the registry entries do not provide uncertainty estimates or an adjusted comparison. The term percentages are not identified by arm or denominator in the listed information. These limitations make the safety counts descriptive rather than a complete comparative safety analysis. The registry’s completed status and enrollment figure do not resolve these reporting gaps.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Dabigatran Etexilate for Secondary Stroke Prevention in Patients With Embolic Stroke of Undetermined Source (RE-SPECT ESUS) — NCT02239120.
• Registry record last updated 2019-09-06; status last verified 2019-08.
• Results first posted 2019-09-06.
• 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

RE-SPECT ESUS was a phase 3 randomized, double-blind trial sponsored by Boehringer Ingelheim. The registry record lists enrollment of 5390 and compares dabigatran etexilate 110 or 150 mg with ASA 100 mg in patients with embolic stroke of undetermined source. The primary endpoint was adjudicated recurrent stroke, reported as an annualised event rate over approximately 43 months in the randomised set.

The posted annualised rates were 4.09%/ year for dabigatran and 4.8%/ year for ASA. The reported Cox hazard ratio was 0.85, with a 95% confidence interval of 0.69 to 1.03 and p = 0.103. The interval includes 1, and the p-value is above the conventional 0.05 threshold. The record also lists serious adverse-event counts, other adverse-event counts, deaths, and secondary outcomes. Detailed information on multiplicity, missing-data handling, model assumptions, and the trial’s sample-size justification is not reported in the registry record.