This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
PACIFIC-STROKE was a randomized, parallel phase 2 trial evaluating the oral FXIa inhibitor BAY 2433334 in patients following a recent non-cardioembolic ischemic stroke. The study enrolled 1808 participants across four arms and used quadruple masking.
| Feature | PACIFIC-STROKE |
|---|---|
| Trial name | PACIFIC-STROKE |
| ClinicalTrials.gov | NCT04304508 |
| Phase | Phase 2 |
| Condition | Acute Non-cardioembolic Ischemic Stroke |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 1808 |
| Arms | 4 |
| Primary endpoints | 2 |
| Primary endpoint types | Binary; Time-to-event |
| Study start | June 15, 2020 |
| Primary completion | February 18, 2022 |
| Status | Completed |
| Lead sponsor | Bayer |
2. Clinical Question
The clinical question was whether oral BAY 2433334, administered at different dose levels, could be evaluated for efficacy and safety in patients following a recent non-cardioembolic ischemic stroke.
Population
Patients following a recent non-cardioembolic ischemic stroke, described in the registry as acute non-cardioembolic ischemic stroke.
Intervention
BAY 2433334, an oral FXIa inhibitor, evaluated across three dose groups: 10 mg, 20 mg, and 50 mg.
Comparator
BAY 2433334 matching placebo.
Primary question
How did BAY 2433334 dose groups compare with placebo for the composite efficacy endpoint through 26 weeks, and how did bleeding safety compare through 52 weeks?
3. Trial Design
10 mg dose group
The 10 mg BAY 2433334 group was compared with placebo for the primary efficacy and safety endpoints.
20 mg dose group
The 20 mg BAY 2433334 group was compared with placebo for the primary efficacy and safety endpoints.
50 mg dose group
The 50 mg BAY 2433334 group was compared with placebo for the primary efficacy and safety endpoints.
Matching placebo
The placebo group served as the comparator for each individual dose and for the dose-response comparison.
The design is statistically useful because it supports two different questions. The pairwise comparisons ask how each BAY 2433334 dose compares with placebo. The MCP-Mod analysis asks a different question: whether the collection of dose groups provides evidence of a dose-response relationship.
4. Endpoints
| Primary endpoint | Time frame | Endpoint type | Registry definition |
|---|---|---|---|
| Efficacy: Number of Participants With Composite of Symptomatic Ischemic Stroke or Covert Brain Infarcts Detected by MRI | From baseline up to 26 weeks | Time-to-event in the registered endpoint; binary in the posted statistical analyses | Composite of adjudicated symptomatic ischemic stroke and incident covert brain infarcts detected by serial MRI. |
| Safety: Number of Participants With Composite of ISTH Major Bleeding and Clinically Relevant Non-major (CRNM) Bleeding | From baseline up to 52 weeks | Binary in the registered endpoint; time-to-event in the posted statistical analyses | Composite of ISTH major bleeding and ISTH clinically relevant non-major bleeding. |
Efficacy endpoint definition
Ischemic stroke was defined as either rapid onset, or presence on awakening, of a new focal neurological deficit with clinical symptoms or signs lasting more than 24 hours or imaging evidence of infarction that was not attributable to a non-ischemic cause; or acute worsening of an existing focal neurological deficit judged attributable to a new infarction or extension of the previous infarction in the same vascular territory, based on persistent symptoms or signs or imaging evidence and no evidence of a non-ischemic etiology.
When imaging was inconclusive, persistent symptoms or signs had to be significant, defined as worsening of the NIHSS score by 4 or more, and sustained for at least 24 hours or until death. Covert brain infarcts were defined as incident infarcts detected by serial MRI in the absence of an adjudicated stroke consistent with the location of the infarct.
Safety endpoint definition
ISTH major bleeding included fatal bleeding; symptomatic bleeding in a critical area or organ; clinically overt bleeding associated with a recent decrease in hemoglobin of at least 2 g/dL compared with the most recent available value before the event; or clinically overt bleeding leading to transfusion of 2 or more units of packed red blood cells or whole blood.
ISTH clinically relevant non-major bleeding was defined as a sign or symptom of hemorrhage that did not meet the definition of major bleeding but required medical intervention by a healthcare professional, led to hospitalization or increased level of care, or prompted a face-to-face evaluation.
5. Statistical Methodology
Pairwise efficacy comparisons
The posted analyses compare each BAY 2433334 dose with placebo using a crude incidence ratio. The analysis population is the FAS. The registry does not report a named statistical method for these three pairwise efficacy estimates, so the effect measure should be interpreted as reported rather than relabeled as a risk ratio, rate ratio, or hazard ratio.
The registry reports the effect as a crude incidence ratio. A value above 1 indicates a higher observed crude incidence in the BAY 2433334 group relative to placebo; a value below 1 would indicate a lower observed crude incidence. The ratio itself does not establish a causal mechanism or describe the timing of individual events.
Safety time-to-event analysis
The safety analyses use the safety population (SAF), a log-rank test, and Cox proportional-hazards estimates. The posted effect measure is a hazard ratio with a two-sided 90% confidence interval.
The Cox hazard ratio is a relative time-to-event measure. It is not an absolute difference in the number of participants with bleeding, and it does not mean that the same relative difference applies to every individual participant.
MCP-Mod dose-response analysis
The registry reports an MCP-Mod dose-response analysis comparing the three BAY 2433334 doses and placebo. MCP-Mod combines multiple comparison procedures with model-based dose-response evaluation. The purpose is to assess whether the observed pattern across doses provides evidence consistent with a dose-response relationship.
The posted dose-response analysis reports a P-value of 0.7976 and does not report an effect estimate or confidence interval.
Confidence intervals
The posted pairwise analyses use two-sided 90% confidence intervals. The confidence interval describes statistical uncertainty around the reported effect estimate under the corresponding analysis framework. It is not a range containing the effects that individual participants experienced.
Analysis populations
The efficacy analyses use the FAS, while the safety analyses use the SAF. Keeping these populations distinct is important because an efficacy estimate and a safety estimate may be based on different participant sets.
| Analysis | Population | Method | Effect measure | Confidence interval |
|---|---|---|---|---|
| 10 mg vs placebo efficacy | FAS | Not reported | Crude incidence ratio | Two-sided 90% |
| 20 mg vs placebo efficacy | FAS | Not reported | Crude incidence ratio | Two-sided 90% |
| 50 mg vs placebo efficacy | FAS | Not reported | Crude incidence ratio | Two-sided 90% |
| Dose-response efficacy | FAS | MCP-Mod | No estimate reported | No CI reported |
| 10 mg vs placebo safety | SAF | Log-rank; Cox | Hazard ratio | Two-sided 90% |
| 20 mg vs placebo safety | SAF | Log-rank; Cox | Hazard ratio | Two-sided 90% |
| 50 mg vs placebo safety | SAF | Log-rank; Cox | Hazard ratio | Two-sided 90% |
| Total BAY 2433334 vs placebo safety | SAF | Log-rank; Cox | Hazard ratio | Two-sided 90% |
6. Results
The registry contains eight posted statistical analyses: four for the primary safety endpoint, three pairwise efficacy comparisons, and one MCP-Mod dose-response analysis. Seven of the eight analyses include an estimate and confidence interval; the MCP-Mod analysis reports a P-value without an effect estimate or confidence interval.
Primary efficacy: composite symptomatic ischemic stroke or covert brain infarcts
| Comparison | Crude incidence ratio | 90% CI |
|---|---|---|
| Asundexian 10 mg vs placebo | 1.044 | 0.8105–1.3478 |
| Asundexian 20 mg vs placebo | 1.1963 | 0.9281–1.5428 |
| Asundexian 50 mg vs placebo | 1.0485 | 0.8082–1.3619 |
10 mg vs placebo
Two-sided 90% CI: 0.8105–1.3478
Effect measure: crude incidence ratio · Analysis population: FAS
The reported crude incidence ratio of 1.044 means that the observed crude incidence in the Asundexian 10 mg group was approximately 1.044 times that in the placebo group under the registry's reported analysis. It does not mean that participants receiving 10 mg had a 4.4% higher individual risk, and it is not a hazard ratio.
The two-sided 90% confidence interval of 0.8105–1.3478 describes the uncertainty around the estimate. Its width indicates that the estimate is not highly precise. Because the interval spans 1, the reported interval is compatible with both a lower and a higher crude incidence relative to placebo.
No P-value is posted for this pairwise efficacy comparison. A P-value, when available, addresses evidence against a specified null hypothesis; it does not measure the size or clinical importance of an effect. The registry also does not report a named statistical method for this pairwise analysis, so the estimate should be interpreted specifically as the reported crude incidence ratio.
20 mg vs placebo
Two-sided 90% CI: 0.9281–1.5428
Effect measure: crude incidence ratio · Analysis population: FAS
The crude incidence ratio of 1.1963 indicates an observed crude incidence approximately 1.1963 times the placebo incidence in the 20 mg group. This is a relative incidence measure, not a statement that an individual participant's probability increased by exactly 19.63%.
The two-sided 90% confidence interval of 0.9281–1.5428 is fairly broad relative to the point estimate and crosses 1. The interval therefore includes values consistent with a lower crude incidence as well as values consistent with a higher crude incidence relative to placebo.
No P-value is posted for this pairwise efficacy estimate. The absence of a reported P-value means the confidence interval is the primary measure of statistical uncertainty posted on ClinicalTrials.gov for this comparison. The registry does not identify the formal method used to generate the crude incidence ratio.
50 mg vs placebo
Two-sided 90% CI: 0.8082–1.3619
Effect measure: crude incidence ratio · Analysis population: FAS
The crude incidence ratio of 1.0485 indicates an observed crude incidence approximately 1.0485 times the placebo incidence in the 50 mg group. The estimate is close to 1, but the point estimate alone should not be treated as evidence of equivalence or as proof that the treatment groups have identical event rates.
The two-sided 90% confidence interval of 0.8082–1.3619 spans 1 and extends in both directions. It therefore allows for a lower or higher crude incidence relative to placebo under the reported analysis.
No P-value is posted for this pairwise comparison. The analysis also is not reported as a Cox model, so proportional-hazards assumptions do not attach to this particular efficacy estimate.
Primary efficacy: dose-response analysis
MCP-Mod dose-response test
Comparison: Asundexian 10 mg vs 20 mg vs 50 mg vs placebo
Analysis population: FAS · Method: MCP-Mod (dose-response)
The MCP-Mod analysis evaluates the overall dose-response question rather than one dose-versus-placebo comparison. The reported P-value is 0.7976. A P-value is a measure of compatibility with a specified statistical null framework; it is not an estimate of the magnitude of a dose-response relationship.
The registry does not report an effect estimate or confidence interval for this MCP-Mod analysis. Consequently, the P-value should not be translated into a percentage change, dose-response slope, or other quantitative treatment effect.
MCP-Mod is particularly useful when several plausible dose-response shapes are considered simultaneously. Its purpose here is different from simply comparing the three dose groups one at a time, because the analysis addresses evidence for a systematic relationship across the dose range.
Primary safety: ISTH major bleeding or CRNM bleeding
| Comparison | Hazard ratio | Two-sided 90% CI | P-value |
|---|---|---|---|
| Asundexian 10 mg vs placebo | 1.724 | 0.924–3.215 | = 0.1507 |
| Asundexian 20 mg vs placebo | 1.285 | 0.662–2.494 | = 0.5339 |
| Asundexian 50 mg vs placebo | 1.749 | 0.938–3.262 | = 0.1401 |
| Total Asundexian vs placebo | 1.585 | 0.918–2.736 | = 0.1653 |
10 mg vs placebo
Two-sided 90% CI: 0.924–3.215 · P = 0.1507
Log-rank test with Cox proportional-hazards effect estimate · Analysis population: SAF
A hazard ratio of 1.724 means that the estimated instantaneous hazard of the composite ISTH major bleeding or CRNM bleeding endpoint was 1.724 times the placebo hazard under the fitted Cox model. It does not mean that 72.4% of participants experienced bleeding, nor does it mean that each participant had a 72.4% increase in risk.
The two-sided 90% confidence interval of 0.924–3.215 is wide and includes 1. The interval therefore contains values compatible with a lower hazard as well as substantially higher hazards relative to placebo.
The reported P-value of 0.1507 is not a measure of effect size. It addresses the statistical evidence under the specified testing framework. The confidence interval is important because it shows that the point estimate alone does not provide a precise description of the underlying hazard ratio.
Because the estimate is from a Cox proportional-hazards analysis, interpretation of a single hazard ratio is most straightforward when the proportional-hazards assumption is reasonable. The registry also states that cause-specific HRs were calculated only if at least 3 events occurred in one of the compared groups and at least 1 event occurred in each compared treatment group.
20 mg vs placebo
Two-sided 90% CI: 0.662–2.494 · P = 0.5339
Log-rank test with Cox proportional-hazards effect estimate · Analysis population: SAF
The hazard ratio of 1.285 corresponds to an estimated instantaneous hazard 1.285 times that of placebo under the Cox model. It is not an absolute risk difference and should not be read as a 28.5% increase in the probability of bleeding for every participant.
The two-sided 90% confidence interval of 0.662–2.494 is wide and spans 1. This indicates substantial statistical uncertainty around the point estimate and allows for both lower and higher hazards relative to placebo.
The P-value of 0.5339 does not quantify the size of the hazard ratio. It should be considered together with the confidence interval and the event-time analysis rather than used as a substitute for an effect estimate.
The Cox interpretation also depends on the proportional-hazards framework. The registry's event-count condition for cause-specific hazard ratios indicates that the availability of certain hazard-ratio calculations depended on the observed number of events.
50 mg vs placebo
Two-sided 90% CI: 0.938–3.262 · P = 0.1401
Log-rank test with Cox proportional-hazards effect estimate · Analysis population: SAF
The hazard ratio of 1.749 indicates an estimated instantaneous hazard 1.749 times the placebo hazard under the reported Cox model. This is a relative time-to-event estimate and does not imply that the absolute probability of bleeding was 74.9% higher.
The two-sided 90% confidence interval of 0.938–3.262 spans 1 and is wide. The data summarized by this interval therefore allow for a range of possible relative hazards, including values close to the null and values considerably above it.
The P-value of 0.1401 is a measure of statistical evidence under the testing framework, not a measure of clinical importance or effect magnitude. The confidence interval supplies additional information about the precision of the estimate.
As with the other Cox analyses, a single hazard ratio summarizes a time-to-event comparison under the model and should not be interpreted as an individual-level probability or absolute risk difference.
Total Asundexian vs placebo
Two-sided 90% CI: 0.918–2.736 · P = 0.1653
Log-rank test with Cox proportional-hazards effect estimate · Analysis population: SAF
The total Asundexian hazard ratio of 1.585 corresponds to an estimated instantaneous hazard 1.585 times the placebo hazard under the Cox model. This pooled comparison combines the BAY 2433334 dose groups for the reported safety analysis; it does not provide a dose-specific estimate.
The two-sided 90% confidence interval of 0.918–2.736 crosses 1 and remains relatively wide. Thus, the point estimate should be interpreted with attention to the uncertainty represented by the interval.
The reported P-value is 0.1653. As with the dose-specific analyses, the P-value does not describe the size of the observed effect. The hazard ratio and its confidence interval provide the quantitative description of the treatment comparison.
The registry also notes an event-count requirement for cause-specific hazard-ratio calculations. That condition is relevant when interpreting Cox estimates for relatively infrequent safety outcomes.
7. Safety Results by Arm
The registry reports serious adverse events by treatment arm. These figures describe the number affected and the number at risk for each group and are separate from the primary composite bleeding endpoint analyses.
| Treatment group | Serious adverse events affected | At risk |
|---|---|---|
| Asundexian 10 mg | 90 | 445 |
| Asundexian 20 mg | 83 | 446 |
| Asundexian 50 mg | 93 | 443 |
| Placebo | 98 | 452 |
The affected and at-risk counts should not be substituted for the primary bleeding hazard ratios. Serious adverse events are a broader safety category, whereas the primary safety endpoint specifically combines ISTH major bleeding and CRNM bleeding.
8. Statistical Methods Explained
Why are there both pairwise dose comparisons and an MCP-Mod analysis?
The pairwise analyses address three specific comparisons: 10 mg versus placebo, 20 mg versus placebo, and 50 mg versus placebo. MCP-Mod addresses a broader dose-response question across the dose range. These are complementary statistical questions rather than interchangeable analyses.
What does a crude incidence ratio of 1.044 mean?
A crude incidence ratio of 1.044 means that the observed crude incidence in the 10 mg group was 1.044 times the observed crude incidence in placebo under the reported analysis. It does not mean that every patient had a 4.4% higher risk, and it is not equivalent to a Cox hazard ratio.
Why does the efficacy analysis not use a hazard ratio?
The registry's posted efficacy analyses report a crude incidence ratio and identify the endpoint as binary for those analyses. Although the registered endpoint is characterized as time-to-event, the posted statistical analysis does not report a Cox model or other named time-to-event method for the pairwise efficacy comparisons. The effect measure should therefore be interpreted as reported.
What does a hazard ratio of 1.585 mean?
For the total Asundexian versus placebo safety analysis, a hazard ratio of 1.585 means that the fitted Cox model estimates the instantaneous hazard of the composite bleeding endpoint to be 1.585 times the placebo hazard. It does not describe the absolute percentage of participants who experienced the event and does not imply a uniform individual-level increase.
Why does the confidence interval matter?
A point estimate is only one estimate from the observed data. The two-sided 90% confidence interval describes statistical uncertainty around that estimate. For example, the total Asundexian safety HR of 1.585 has a 90% CI of 0.918–2.736, showing substantially more uncertainty than the point estimate alone would suggest.
Why does the P-value not measure effect size?
A P-value summarizes the evidence against a specified null hypothesis under a statistical testing framework. It is affected by the amount of information and variability in the data. Effect magnitude is communicated by the effect estimate itself, while confidence intervals describe uncertainty around that estimate.
What does MCP-Mod contribute to the analysis?
MCP-Mod is designed for dose-response assessment. Multiple comparison procedures can address candidate dose-response patterns while controlling the relevant statistical testing structure, and model-based components then describe the dose-response relationship. Here, the registry reports the MCP-Mod P-value as 0.7976 but does not report a corresponding dose-response estimate or confidence interval.
9. Interpreting the Efficacy Endpoint
The primary efficacy endpoint combines two clinically and statistically distinct types of events: symptomatic ischemic stroke and covert brain infarcts detected by serial MRI. A participant can therefore contribute to the composite through either an adjudicated symptomatic event or an incident MRI-detected infarct meeting the registry definition.
Composite endpoint
Combining multiple event types can increase the number of observed endpoint events, but interpretation depends on the clinical meaning and relative contribution of each component.
MRI component
Covert brain infarcts are defined using serial MRI in the absence of an adjudicated stroke consistent with the infarct location.
Symptomatic stroke component
The registry definition incorporates clinical findings, imaging, duration, and exclusion of non-ischemic etiologies.
Analysis population
The posted pairwise efficacy analyses use the FAS, making the population definition part of the interpretation of each crude incidence ratio.
The three point estimates are 1.044 for 10 mg, 1.1963 for 20 mg, and 1.0485 for 50 mg. The confidence intervals for all three pairwise comparisons span 1. The separate MCP-Mod analysis reports a P-value of 0.7976 and no effect estimate.
10. Interpreting the Safety Endpoint
The primary safety endpoint is a composite of ISTH major bleeding and clinically relevant non-major bleeding. The posted analyses use time-to-event methodology, making both the timing of events and censoring relevant to the statistical comparison.
| Safety comparison | HR | 90% CI | P-value |
|---|---|---|---|
| 10 mg vs placebo | 1.724 | 0.924–3.215 | = 0.1507 |
| 20 mg vs placebo | 1.285 | 0.662–2.494 | = 0.5339 |
| 50 mg vs placebo | 1.749 | 0.938–3.262 | = 0.1401 |
| Total Asundexian vs placebo | 1.585 | 0.918–2.736 | = 0.1653 |
All four reported confidence intervals include 1. The intervals are also relatively broad, emphasizing that the point estimates alone do not provide a precise description of the underlying safety hazard. The P-values likewise should not be treated as measures of the magnitude of the observed hazard ratios.
The registry specifies that cause-specific hazard ratios were calculated only if at least 3 events occurred in one of the compared groups and at least 1 event occurred in each of the compared treatment groups. This event-count condition is particularly relevant for safety endpoints because individual adverse-event categories can be relatively infrequent.
11. Confidence Intervals and P-values
The efficacy pairwise estimates and safety hazard ratios use two-sided 90% confidence intervals. The distinction between a confidence interval and a P-value is important when reading the posted analyses.
Confidence interval
Describes uncertainty around the reported effect estimate. A wider interval indicates less precision than a narrower interval.
P-value
Addresses evidence under a specified null hypothesis. It does not measure how large or clinically important an effect is.
Ratio near 1
For the crude incidence ratio and hazard ratio, 1 represents the no-relative-difference value for the corresponding measure.
Interval crossing 1
When a two-sided ratio confidence interval crosses 1, the interval includes the null value for that relative measure.
These concepts should not be conflated. For example, the 10 mg safety HR is 1.724 with a 90% CI of 0.924–3.215 and P = 0.1507. The HR describes the estimated relative hazard, the CI describes uncertainty around that estimate, and the P-value describes statistical evidence under the testing framework.
12. Randomization and Masking
Randomization is a central design feature because treatment assignment is determined independently of participants' subsequent outcomes. In a randomized parallel-group trial, this provides the foundation for comparing outcomes between groups while reducing systematic allocation differences.
PACIFIC-STROKE used quadruple masking. Masking can reduce the potential for knowledge of treatment assignment to influence participant behavior, clinical management, outcome assessment, or other aspects of trial conduct. The degree to which masking matters depends on the endpoint; MRI-detected outcomes and adjudicated clinical events can involve different pathways of assessment.
13. Time-to-Event Analysis and Censoring
The primary safety analyses are explicitly time-to-event analyses. In a time-to-event framework, participants contribute information over the period they are observed, and participants who do not experience the event during their observed follow-up may be censored.
The survival function represents the probability that the event time T exceeds time t. For this trial's safety endpoint, the event is the composite of ISTH major bleeding and CRNM bleeding.
The log-rank test compares the event-time experience of the treatment groups over follow-up. The Cox model then expresses the relative difference as a hazard ratio.
The hazard is an instantaneous event rate conditional on remaining event-free immediately before a given time. Consequently, a hazard ratio is fundamentally different from a cumulative incidence ratio. This distinction is why the efficacy crude incidence ratios should not be interpreted using the same language as the safety hazard ratios.
14. Cox Proportional-Hazards Interpretation
HR > 1 → higher modeled hazard in the first-listed group
HR < 1 → lower modeled hazard in the first-listed group
For PACIFIC-STROKE, the first-listed group in each safety comparison is the BAY 2433334 group and the reference group is placebo.
The Cox model produces a single relative hazard summary over the analyzed period. Its most direct interpretation assumes that the proportional-hazards structure is an adequate representation of the relationship between treatment and the event hazard over time.
A hazard ratio should therefore not be converted directly into an absolute percentage of participants with bleeding. Absolute event probabilities require a different presentation, such as cumulative incidence or survival estimates at specified times.
15. MCP-Mod and Dose-Response Reasoning
The presence of three BAY 2433334 dose groups makes dose-response a central statistical question. Simply looking at the three crude incidence ratios does not formally test whether the treatment effect changes systematically with dose.
| Analysis component | Question answered |
|---|---|
| 10 mg vs placebo | How does the 10 mg group compare with placebo? |
| 20 mg vs placebo | How does the 20 mg group compare with placebo? |
| 50 mg vs placebo | How does the 50 mg group compare with placebo? |
| MCP-Mod | Is there statistical evidence of a dose-response relationship across the evaluated dose levels? |
The reported MCP-Mod P-value is 0.7976. No corresponding dose-response estimate or confidence interval is posted in the statistical analysis record. Therefore, the dose-response result is appropriately described through the reported testing result rather than converted into an estimated slope or other quantitative measure.
16. Multiplicity and Multiple Dose Comparisons
Three dose-versus-placebo comparisons create a multiple-comparison problem if each is interpreted as an independent confirmatory hypothesis test. MCP-Mod provides a framework specifically designed to address dose-response evaluation in the presence of multiple candidate comparisons and dose-response models.
The registry identifies the dose-response analysis as "MCP Mod" and describes it as a dose-response test using multiple comparison procedures. It does not provide a detailed alpha-allocation scheme in the information summarized here.
17. Primary Endpoint Results at a Glance
| Endpoint | Comparison / analysis | Estimate | 90% CI | P-value |
|---|---|---|---|---|
| Efficacy composite | 10 mg vs placebo | Crude incidence ratio 1.044 | 0.8105–1.3478 | Not reported |
| Efficacy composite | 20 mg vs placebo | Crude incidence ratio 1.1963 | 0.9281–1.5428 | Not reported |
| Efficacy composite | 50 mg vs placebo | Crude incidence ratio 1.0485 | 0.8082–1.3619 | Not reported |
| Efficacy composite | MCP-Mod dose-response | Not reported | Not reported | 0.7976 |
| Bleeding composite | 10 mg vs placebo | HR 1.724 | 0.924–3.215 | 0.1507 |
| Bleeding composite | 20 mg vs placebo | HR 1.285 | 0.662–2.494 | 0.5339 |
| Bleeding composite | 50 mg vs placebo | HR 1.749 | 0.938–3.262 | 0.1401 |
| Bleeding composite | Total Asundexian vs placebo | HR 1.585 | 0.918–2.736 | 0.1653 |
18. How to Read the Two Primary Endpoints Together
The efficacy and safety endpoints use different effect measures and different follow-up periods. The efficacy endpoint is assessed from baseline through 26 weeks, whereas the safety endpoint is assessed from baseline through 52 weeks.
Efficacy question
What was the crude incidence of the composite symptomatic ischemic stroke or covert brain infarct endpoint through 26 weeks in each dose group relative to placebo?
Safety question
How did the time-to-event hazard of ISTH major bleeding or CRNM bleeding through 52 weeks compare between BAY 2433334 and placebo?
Different estimands
A crude incidence ratio and a hazard ratio summarize different aspects of the observed data and should not be compared numerically as if they were the same measure.
Dose-response question
The MCP-Mod analysis evaluates the overall relationship across the four groups rather than providing another single dose-versus-placebo effect estimate.
This separation is important for statistical interpretation. A trial can have one endpoint analyzed through crude incidence and another through survival methods without those analyses being interchangeable. The endpoint definition, follow-up window, analysis population, and effect measure all contribute to what the numerical result means.
19. Limitations
- Efficacy method not reported: the registry provides crude incidence ratios for the three pairwise efficacy analyses but does not identify the statistical method used to calculate those estimates.
- Different endpoint characterizations: the registered efficacy endpoint is characterized as time-to-event, while the posted pairwise efficacy analyses are labeled binary. The registered safety endpoint is characterized as binary, while the posted analyses use time-to-event methods.
- Limited effect-measure information for MCP-Mod: the dose-response analysis reports a P-value of 0.7976 but no effect estimate or confidence interval.
- Wide safety confidence intervals: the posted safety hazard-ratio intervals are broad, particularly relative to their point estimates, indicating substantial uncertainty.
- Composite endpoints: the efficacy endpoint combines symptomatic ischemic stroke with covert MRI-detected brain infarcts, while the safety endpoint combines ISTH major bleeding with CRNM bleeding. Composite interpretation depends on the components being clinically and statistically coherent for the question being asked.
- Analysis-population distinction: efficacy analyses use the FAS, whereas safety analyses use the SAF. Estimates from these populations should not be assumed to represent exactly the same participant set.
- Event-count restrictions: the registry states that cause-specific hazard ratios were calculated only when at least 3 events occurred in one compared group and at least 1 event occurred in each compared treatment group.
- Cox model assumptions: the interpretation of a single hazard ratio depends on the proportional-hazards framework. The registry does not provide additional proportional-hazards diagnostics in the posted analyses summarized here.
- Multiplicity: three pairwise dose comparisons and a dose-response analysis raise different multiple-testing considerations. The registry identifies MCP-Mod as the dose-response method but does not provide the complete alpha-allocation details in the posted statistical-analysis information.
20. Why This Trial Matters Statistically
PACIFIC-STROKE is a useful teaching case because it combines randomized dose-ranging design with two different statistical endpoint structures. The efficacy endpoint is a composite involving clinical and MRI-detected events, while the safety endpoint is analyzed using time-to-event methodology. The trial therefore illustrates why endpoint definition and statistical method must be read together.
| Statistical concept | How it appears in PACIFIC-STROKE |
|---|---|
| Randomization | Four parallel randomized groups |
| Blinding | Quadruple masking |
| Multiple doses | 10 mg, 20 mg, and 50 mg BAY 2433334 plus placebo |
| Composite endpoint | Symptomatic ischemic stroke or covert brain infarcts for efficacy |
| MRI endpoint assessment | Covert brain infarcts detected by serial MRI |
| Crude incidence ratio | Posted pairwise efficacy effect measure |
| Confidence intervals | Two-sided 90% intervals for reported pairwise estimates |
| Log-rank test | Posted primary safety analysis method |
| Cox proportional hazards | Posted safety effect measure |
| Hazard ratio | Safety comparison of BAY 2433334 with placebo |
| MCP-Mod | Dose-response analysis across the four groups |
| Analysis populations | FAS for efficacy; SAF for safety |
21. Overall Statistical Interpretation
The three reported crude incidence ratios for the composite symptomatic ischemic stroke or covert brain infarct endpoint were 1.044, 1.1963, and 1.0485 for 10 mg, 20 mg, and 50 mg versus placebo, respectively. Their two-sided 90% confidence intervals were 0.8105–1.3478, 0.9281–1.5428, and 0.8082–1.3619. The dedicated MCP-Mod dose-response analysis reported P = 0.7976.
The posted Cox hazard ratios for ISTH major bleeding or CRNM bleeding were 1.724 for 10 mg, 1.285 for 20 mg, 1.749 for 50 mg, and 1.585 for total Asundexian versus placebo. The corresponding two-sided 90% confidence intervals were 0.924–3.215, 0.662–2.494, 0.938–3.262, and 0.918–2.736. The corresponding P-values were 0.1507, 0.5339, 0.1401, and 0.1653.
The reported estimates describe the statistical comparisons recorded in the registry. The efficacy results use crude incidence ratios, while the safety results use hazard ratios from Cox proportional-hazards analyses. These measures answer different questions and should not be directly compared numerically. Confidence intervals show the uncertainty around the estimates, while P-values address statistical evidence under the associated testing framework rather than measuring effect magnitude.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: NCT04304508 — PACIFIC-STROKE.
- PubMed: PMID 40515384.
- PubMed: PMID 38174569.
- PubMed: PMID 38038437.
- PubMed: PMID 37950392.
- PubMed: PMID 36063821.
Continue through the Clinical Biostats statistical library
Explore deeper tutorials on survival analysis, confidence intervals, dose-response methodology, randomization, and other methods used in clinical research.
25. Record Summary
PACIFIC-STROKE provides a useful example of a randomized phase 2 dose-ranging trial in which efficacy and safety are evaluated through different statistical frameworks. The primary efficacy endpoint combines symptomatic ischemic stroke with covert brain infarcts detected by MRI and is reported through crude incidence ratios for the three dose-versus-placebo comparisons, together with an MCP-Mod dose-response analysis. The primary safety endpoint combines ISTH major bleeding and CRNM bleeding and is analyzed using log-rank testing and Cox proportional-hazards estimates.
The statistical interpretation therefore depends on keeping several distinctions clear: crude incidence ratios are not hazard ratios; hazard ratios are not absolute risks; P-values do not measure effect magnitude; confidence intervals describe uncertainty rather than individual treatment responses; and a dose-response analysis addresses a different question from separate pairwise comparisons.