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
AXIOMATIC-SSP was a randomized, quadruple-masked, parallel phase 2 trial evaluating BMS-986177 in patients with acute ischemic stroke or transient ischemic attack receiving aspirin and clopidogrel. The registry reports eight arms and 2366 enrolled participants, with a primary binary endpoint assessed through a model-based dose-response analysis.
| Feature | AXIOMATIC-SSP |
|---|---|
| Trial name | AXIOMATIC-SSP |
| NCT identifier | NCT03766581 |
| Phase | Phase 2 |
| Status | Completed |
| Therapeutic area | Cardiovascular |
| Conditions | Acute Ischemic Stroke; Transient Ischemic Attack (TIA) |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Prevention |
| Enrollment | 2366 |
| Primary endpoint type | Binary |
| Outcome measures posted | 23 |
| Statistical analyses posted | 12 |
2. Clinical Question
The central statistical question was whether increasing doses of BMS-986177 were associated with a dose-response relationship for a composite of new ischemic stroke during treatment and new covert brain infarction detected by MRI by day 90 in patients receiving aspirin and clopidogrel.
Population
Patients with acute ischemic stroke or transient ischemic attack enrolled in a randomized phase 2 prevention study.
Intervention
BMS-986177, evaluated across multiple dose regimens including 25 mg QD, 50 mg QD, 100 mg QD, 25 mg BID, 50 mg BID, 100 mg BID, and 200 mg BID.
Background therapy
Aspirin and clopidogrel were part of the study's intervention framework.
Primary question
Does the model-based dose-response analysis provide evidence of a relationship between BMS-986177 dose and the day 90 composite event?
3. Trial Design
Study arms represented in the safety data
Placebo
Placebo was the reference group for the reported relative-risk comparisons and safety summary.
Milvexian 25 mg QD
The 25 mg once-daily regimen was included in the model-based primary endpoint analysis and in the prespecified ischemic-stroke analysis.
Milvexian 50 mg QD
The 50 mg once-daily regimen was included in the prespecified ischemic-stroke analysis.
Milvexian 100 mg QD
The 100 mg once-daily regimen was included in the prespecified ischemic-stroke analysis.
Milvexian 25 mg BID
The 25 mg twice-daily regimen was included in the model-based primary endpoint analysis and in the prespecified ischemic-stroke analysis.
Milvexian 50 mg BID
The 50 mg twice-daily regimen was included in the model-based primary endpoint analysis and in the prespecified ischemic-stroke analysis.
Milvexian 100 mg BID
The 100 mg twice-daily regimen was included in the model-based primary endpoint analysis and in the prespecified ischemic-stroke analysis.
Milvexian 200 mg BID
The 200 mg twice-daily regimen was included in the model-based primary endpoint analysis and in the prespecified ischemic-stroke analysis.
4. Primary Endpoint
The registered primary endpoint was the percent of participants with model-based assessment of a composite of new ischemic stroke during treatment and new covert brain infarction (FLAIR + DWI) detected by MRI by day 90.
| Endpoint feature | Registry definition |
|---|---|
| Endpoint | Percent of Participants With Model Based Assessment of Composite of New Ischemic Stroke During Treatment and New Covert Brain Infarction (FLAIR + DWI) Detected by MRI by Day 90 |
| Time frame | From randomization to up to 90 days after randomization |
| Endpoint type | Binary |
| Effect measure | Relative Risk (RR) |
| Primary statistical method | MCP-MOD |
| Confidence interval | 95%, two-sided |
| Confidence interval construction | Bootstrap with 10000 samples |
Primary analysis population
The registry defines the analysis population as all randomized participants who experienced a new ischemic stroke by day 90, or an evaluable day 90 MRI regardless of when the MRI was collected. The dose-response model-based endpoint was prespecified for data collected only in the placebo, 25 mg QD, and BID dose regimens.
5. Statistical Methodology
MCP-Mod for dose-response analysis
The primary endpoint used MCP-Mod, or Multiple Comparison Procedures and Modeling, to evaluate the dose-response relationship. In this framework, candidate dose-response patterns can first be evaluated through multiple comparison procedures, followed by model-based estimation of the relationship when supported by the data.
For AXIOMATIC-SSP, the registry describes the primary endpoint as a model-based assessment and specifically identifies MCP-Mod as the customized statistical analysis used to check for a dose-response relationship. The reported confidence intervals for the composite event were based on bootstrap resampling with 10000 samples.
The statistical objective is broader than asking whether one selected dose differs from placebo. The dose-response framework asks whether the pattern across prespecified doses is consistent with an underlying dose-response relationship.
Relative risk
The registry reports relative risk (RR) for the primary composite endpoint. For a binary outcome, an RR compares the probability of the event in one group with the probability in the reference group.
An RR of 1 represents equal estimated risk between the compared groups. An RR below 1 represents lower estimated risk in the BMS-986177 group; an RR above 1 represents higher estimated risk in that comparison.
Bootstrap confidence intervals
The primary endpoint confidence intervals were constructed using bootstrap resampling with 10000 samples. Bootstrap methods repeatedly resample the observed data to characterize the sampling variability of an estimator. Here, the registry specifically identifies bootstrap as the basis for the 95% confidence interval around the MCP-Mod composite-event estimates.
Wald confidence limits for ischemic stroke
The prespecified ischemic-stroke analyses used Wald Confidence Limits. These analyses reported relative risks and two-sided 95% confidence intervals for each specified dose compared with placebo.
For a relative-risk analysis, Wald-based intervals are typically constructed on an appropriate transformed scale and then returned to the RR scale. The resulting interval quantifies uncertainty around the estimated relative risk.
6. Primary Results
The registry contains five posted primary endpoint analyses. Each compares a specified BMS-986177 dose with placebo using MCP-Mod and reports a relative risk with a two-sided 95% bootstrap confidence interval. The posted analyses do not list p-values.
Milvexian 25 mg QD vs Placebo
Primary composite endpoint
95% CI: 0.87–1.10 · MCP-Mod
95% confidence interval based on bootstrap resampling with 10000 samples.
An RR of 0.99 means that the model-based estimated relative risk of the composite event for Milvexian 25 mg QD compared with placebo was approximately the same on the relative-risk scale. More specifically, an RR of 0.99 corresponds to an estimated relative risk that is 1% lower than the placebo reference, but the point estimate alone should not be treated as evidence that the true effect is exactly 1% lower.
The 95% CI of 0.87–1.10 indicates that the estimate has uncertainty extending below and above 1. The interval therefore includes values compatible with lower relative risk and values compatible with higher relative risk. The width of the interval also indicates that the estimate is not highly precise.
The registry does not provide a p-value for this posted MCP-Mod analysis. A p-value, when reported, would address evidence against a specified null hypothesis; it would not measure the magnitude or clinical importance of the effect. The RR and its confidence interval are therefore the key quantitative measures reported here.
Because the endpoint is model-based and dose-response oriented, this comparison should also be distinguished from a simple standalone pairwise test. The MCP-Mod framework incorporates the dose-response structure specified for the analysis.
Milvexian 25 mg BID vs Placebo
Primary composite endpoint
95% CI: 0.83–1.15 · MCP-Mod
95% confidence interval based on bootstrap resampling with 10000 samples.
The estimated RR of 0.99 again places the point estimate very close to the placebo reference value of 1. On the relative-risk scale, the point estimate corresponds to approximately 1% lower estimated risk in the 25 mg BID comparison.
The 95% CI of 0.83–1.15 is wider than the corresponding point estimate itself and crosses 1. This means the interval includes both relative-risk values below 1 and values above 1. The confidence interval therefore provides important information that cannot be obtained from the point estimate alone.
No p-value is listed for this posted analysis. The absence of a listed p-value means that the registry's quantitative result here is best read through the RR, the two-sided 95% CI, and the MCP-Mod analysis framework rather than through an unreported significance threshold.
Milvexian 50 mg BID vs Placebo
Primary composite endpoint
95% CI: 0.76–1.16 · MCP-Mod
95% confidence interval based on bootstrap resampling with 10000 samples.
An RR of 0.93 corresponds to an estimated relative risk approximately 7% lower than the placebo reference. That is a relative statement about the modeled comparison; it is not a 7-percentage-point reduction in event probability and does not establish how many individual participants benefited.
The 95% CI of 0.76–1.16 spans 1. The interval is therefore compatible with a lower relative risk as well as a higher relative risk under the statistical uncertainty represented by the interval.
The confidence interval is also relatively broad compared with the point estimate. That matters because an estimate close to 1 can still be associated with substantial uncertainty. The registry does not list a p-value for this primary analysis.
Milvexian 100 mg BID vs Placebo
Primary composite endpoint
95% CI: 0.73–1.18 · MCP-Mod
95% confidence interval based on bootstrap resampling with 10000 samples.
The estimated RR of 0.92 corresponds to approximately 8% lower estimated relative risk compared with placebo. This does not mean that the absolute event probability was reduced by 8 percentage points, nor does it mean that every participant experienced an 8% reduction in risk.
The 95% CI of 0.73–1.18 crosses 1 and extends from a potentially lower relative risk to a potentially higher relative risk. The interval is therefore essential to understanding the uncertainty around the point estimate.
The posted analysis does not include a p-value. In addition, the result should be interpreted as one component of a dose-response analysis rather than as an isolated dose comparison detached from the overall MCP-Mod framework.
Milvexian 200 mg BID vs Placebo
Primary composite endpoint
95% CI: 0.69–1.31 · MCP-Mod
95% confidence interval based on bootstrap resampling with 10000 samples.
An RR of 0.91 corresponds to approximately 9% lower estimated relative risk than placebo at the point estimate. Again, this is a relative measure and is not equivalent to an absolute risk difference.
The 95% CI of 0.69–1.31 is the widest of the five posted primary composite comparisons. It crosses 1 and includes values representing both lower and higher relative risk. The width of the interval indicates considerable statistical uncertainty around the point estimate.
No p-value is reported for the posted MCP-Mod analysis. The appropriate reading is therefore to consider the point estimate together with its confidence interval and the prespecified dose-response modeling framework rather than treating the point estimate alone as the result.
7. Primary Endpoint Results Across Dose Levels
Viewing the five primary analyses together is more informative than reading any one relative risk in isolation. The registry reports point estimates below 1 for all five displayed comparisons, with confidence intervals that cross 1 in every case.
| Comparison | Method | RR | 95% CI | CI construction |
|---|---|---|---|---|
| Milvexian 25 mg QD vs Placebo | MCP-Mod | 0.99 | 0.87–1.10 | Bootstrap, 10000 samples |
| Milvexian 25 mg BID vs Placebo | MCP-Mod | 0.99 | 0.83–1.15 | Bootstrap, 10000 samples |
| Milvexian 50 mg BID vs Placebo | MCP-Mod | 0.93 | 0.76–1.16 | Bootstrap, 10000 samples |
| Milvexian 100 mg BID vs Placebo | MCP-Mod | 0.92 | 0.73–1.18 | Bootstrap, 10000 samples |
| Milvexian 200 mg BID vs Placebo | MCP-Mod | 0.91 | 0.69–1.31 | Bootstrap, 10000 samples |
8. Prespecified Ischemic Stroke Results
The registry also posts seven prespecified analyses of percent of participants with ischemic stroke events from randomization to up to 90 days after randomization. These analyses use all randomized participants and compare individual BMS-986177 regimens with placebo using Wald Confidence Limits.
| Comparison | Method | RR | 95% CI |
|---|---|---|---|
| Milvexian 25 mg QD vs Placebo | Wald Confidence Limits | 0.83 | 0.46–1.49 |
| Milvexian 25 mg BID vs Placebo | Wald Confidence Limits | 0.69 | 0.36–1.30 |
| Milvexian 50 mg BID vs Placebo | Wald Confidence Limits | 0.72 | 0.39–1.33 |
| Milvexian 100 mg BID vs Placebo | Wald Confidence Limits | 0.65 | 0.33–1.25 |
| Milvexian 200 mg BID vs Placebo | Wald Confidence Limits | 1.40 | 0.87–2.25 |
| Milvexian 50 mg QD vs Placebo | Wald Confidence Limits | 2.48 | 0.83–7.42 |
| Milvexian 100 mg QD vs Placebo | Wald Confidence Limits | 1.01 | 0.15–6.96 |
These ischemic-stroke estimates illustrate why relative risks should always be presented with their confidence intervals. Several point estimates are below 1, while two are above 1, but the confidence intervals are wide and cross 1 for every listed comparison.
The Milvexian 100 mg BID estimate is 0.65 with a 95% CI of 0.33–1.25. The point estimate corresponds to approximately 35% lower estimated relative risk, but the interval extends above 1, so the estimate is substantially uncertain.
The Milvexian 200 mg BID estimate is 1.40 with a 95% CI of 0.87–2.25. The point estimate is above 1, but the confidence interval includes 1 and therefore does not support reading the point estimate as a precise measure of increased relative risk.
The especially wide intervals for 50 mg QD and 100 mg QD indicate limited precision in those reported comparisons. For 50 mg QD, the RR is 2.48 with a 95% CI of 0.83–7.42; for 100 mg QD, the RR is 1.01 with a 95% CI of 0.15–6.96.
9. Primary Composite Endpoint vs Ischemic Stroke Endpoint
The two sets of analyses answer related but different statistical questions. The primary endpoint combines new ischemic stroke during treatment with new covert brain infarction detected by MRI. The other prespecified endpoint considers ischemic stroke events alone.
| Feature | Primary composite | Ischemic stroke |
|---|---|---|
| Endpoint | New ischemic stroke during treatment plus new covert brain infarction detected by MRI | Ischemic stroke events |
| Time frame | From randomization to up to 90 days after randomization | From randomization to up to 90 days after randomization |
| Endpoint type | Binary | Count / rate |
| Analysis method | MCP-Mod | Wald Confidence Limits |
| Effect measure | Relative Risk | Relative Risk |
| CI construction | Bootstrap, 10000 samples | Wald Confidence Limits |
The distinction is important because a composite endpoint can capture more events than one of its individual components, potentially affecting precision and the way the overall treatment comparison is interpreted. At the same time, the clinical meaning of the composite depends on how its components are combined and measured.
10. Safety Results
The registry reports serious adverse events by arm as affected participants divided by participants at risk. These figures provide an arm-level safety summary across the placebo group and seven BMS-986177 dose regimens.
| Arm | Serious adverse events | At risk | Affected / at risk |
|---|---|---|---|
| Placebo | 94 | 682 | 94/682 |
| Milvexian 25 mg QD | 37 | 325 | 37/325 |
| Milvexian 50 mg QD | 5 | 22 | 5/22 |
| Milvexian 100 mg QD | 3 | 17 | 3/17 |
| Milvexian 25 mg BID | 39 | 313 | 39/313 |
| Milvexian 50 mg BID | 41 | 325 | 41/325 |
| Milvexian 100 mg BID | 42 | 306 | 42/306 |
| Milvexian 200 mg BID | 54 | 344 | 54/344 |
The affected/at-risk values are descriptive counts rather than adjusted treatment-effect estimates. In particular, the arm sizes differ substantially, including 22 participants at risk for the 50 mg QD regimen and 17 for the 100 mg QD regimen. A small denominator can produce substantial statistical uncertainty even when the number of affected participants is small.
11. Statistical Methods Explained
Why was MCP-Mod used?
MCP-Mod is designed for dose-response studies in which the scientific question concerns the relationship between dose and response rather than only one treatment-versus-control comparison. It combines multiple comparison procedures with model-based dose-response estimation. In AXIOMATIC-SSP, the registry specifically identifies MCP-Mod as the customized statistical analysis used to check for a dose-response relationship.
What does an RR of 0.91 mean?
An RR of 0.91 means that the estimated relative risk in the BMS-986177 group was 0.91 times the placebo reference risk. Expressed as a simple relative comparison, that is approximately 9% lower estimated risk. It does not mean a 9-percentage-point reduction in absolute risk and does not describe the outcome of every individual participant.
Why is the confidence interval more informative than the point estimate alone?
The point estimate is only one estimate of the treatment effect. The confidence interval communicates statistical uncertainty around that estimate. For example, the 200 mg BID primary comparison has an RR of 0.91 but a 95% CI of 0.69–1.31. The interval demonstrates that the estimate is considerably less precise than the single number 0.91 might suggest.
Why were bootstrap confidence intervals used for the primary endpoint?
The registry states that the 95% confidence intervals for the composite event were based on bootstrap resampling with 10000 samples. Bootstrap procedures are useful when the sampling distribution of a complex estimator is not conveniently represented by a simple closed-form formula. In this study, the approach was applied to the MCP-Mod model-based endpoint.
What does a Wald confidence limit mean for the ischemic-stroke endpoint?
The ischemic-stroke analyses used Wald Confidence Limits to construct two-sided 95% confidence intervals around the reported relative risks. A Wald approach uses an estimate and its estimated standard error to characterize uncertainty. Its interpretation is still centered on the same basic principle: the interval describes uncertainty around the estimated relative risk rather than the range of outcomes for individual participants.
Why should the five primary dose comparisons not simply be ranked by their RRs?
The primary analysis was explicitly a dose-response analysis using MCP-Mod. The point estimates were 0.99, 0.99, 0.93, 0.92, and 0.91 for the five reported comparisons, but those estimates have different confidence intervals and arise within a dose-response modeling framework. A lower point estimate at one dose does not by itself establish that dose as statistically or clinically superior to another dose.
What does it mean that the confidence intervals cross 1?
For a relative risk, 1 is the value representing equal estimated risk between the compared groups. When a two-sided 95% confidence interval crosses 1, the interval includes values below and above the reference value. That means the interval alone does not exclude either a lower or higher relative risk at that confidence level. It does not mean that the treatment effects at all possible values within the interval are equally plausible or that the treatment has been shown to have exactly no effect.
12. Interpreting Relative Risk Correctly
The reported RR for Milvexian 100 mg BID versus placebo was 0.92, with a 95% CI of 0.73–1.18.
The point estimate corresponds to approximately 8% lower estimated relative risk. However, this is not an absolute risk reduction, and it does not mean that 8% of participants avoided an event because of treatment.
The confidence interval extends from 0.73 to 1.18, so the statistical uncertainty includes relative risks below and above 1. The interval therefore supplies substantially more information than the point estimate by itself.
The reported RR for Milvexian 200 mg BID versus placebo was 1.40, with a 95% CI of 0.87–2.25.
The point estimate is above 1, corresponding to approximately 40% higher estimated relative risk at the point estimate. But the confidence interval includes 1 and extends over a broad range. The point estimate should therefore not be treated as a precise estimate of increased risk.
The reported RR for Milvexian 50 mg QD versus placebo was 2.48, with a 95% CI of 0.83–7.42.
The point estimate is substantially above 1, but the interval is very wide and includes 1. The result illustrates how a point estimate can appear large while remaining statistically imprecise.
13. Understanding the Dose-Response Structure
The seven BMS-986177 dose regimens provide two different dimensions of dosing: once-daily regimens and twice-daily regimens. The primary MCP-Mod analysis was prespecified for the placebo, 25 mg QD, and BID dose regimens, while the posted primary comparisons shown by the registry include five dose-versus-placebo estimates.
| Regimen | Included in posted primary composite analysis | Posted ischemic-stroke analysis |
|---|---|---|
| 25 mg QD | Yes | Yes |
| 50 mg QD | No | Yes |
| 100 mg QD | No | Yes |
| 25 mg BID | Yes | Yes |
| 50 mg BID | Yes | Yes |
| 100 mg BID | Yes | Yes |
| 200 mg BID | Yes | Yes |
This structure is one reason the MCP-Mod framework is important. A dose-response question cannot be reduced to the statement that the smallest or largest observed RR is the preferred dose. The model considers the prespecified dose-response structure, while the individual relative risks provide dose-specific descriptive estimates.
14. Precision and Sample Size Across Arms
The safety denominators demonstrate substantial differences in the number of participants contributing to individual arms. The placebo group has 682 participants at risk, while the 50 mg QD and 100 mg QD groups have 22 and 17 participants at risk, respectively.
| Arm | Participants at risk | Serious adverse events |
|---|---|---|
| Placebo | 682 | 94 |
| Milvexian 25 mg QD | 325 | 37 |
| Milvexian 50 mg QD | 22 | 5 |
| Milvexian 100 mg QD | 17 | 3 |
| Milvexian 25 mg BID | 313 | 39 |
| Milvexian 50 mg BID | 325 | 41 |
| Milvexian 100 mg BID | 306 | 42 |
| Milvexian 200 mg BID | 344 | 54 |
These denominators help explain why some arm-specific estimates are much less precise than others. The registry's ischemic-stroke confidence intervals for 50 mg QD and 100 mg QD are particularly wide, consistent with the small numbers of participants contributing to those arm-level comparisons.
15. What the Statistical Results Do — and Do Not — Establish
What the RR provides
A relative comparison between the event risk in a specified BMS-986177 regimen and the placebo reference group.
What the CI provides
A quantitative description of uncertainty around the reported relative-risk estimate under the specified analysis framework.
What MCP-Mod adds
A prespecified model-based approach for evaluating dose-response rather than treating dose comparisons as unrelated isolated tests.
What is not contained in an RR
An RR alone does not provide the absolute event probabilities, number needed to treat, or individual-level probability of benefit.
16. Limitations
- Primary endpoint complexity: the primary endpoint is a composite involving new ischemic stroke during treatment and new covert brain infarction detected by MRI, so the result reflects the combined endpoint rather than either component alone.
- Model-based analysis: the primary endpoint uses MCP-Mod, so interpretation should consider the prespecified dose-response modeling framework rather than relying only on individual pairwise relative risks.
- Confidence-interval width: several posted estimates have broad confidence intervals, particularly some of the prespecified ischemic-stroke comparisons.
- Small QD denominators: the safety data show only 22 participants at risk in the 50 mg QD arm and 17 in the 100 mg QD arm, which limits precision for arm-specific descriptions.
- No posted p-values: the posted statistical analyses provide effect estimates and confidence intervals, but the registry entries shown here do not provide p-values for those analyses.
- Different statistical methods: the primary composite endpoint uses MCP-Mod with bootstrap confidence intervals, whereas the ischemic-stroke analyses use Wald Confidence Limits. These are related but distinct statistical procedures.
- Composite interpretation: a composite endpoint can contain clinically and statistically different components. The overall composite estimate does not show whether each component behaves identically.
- Multiple dose levels: interpretation of a dose-response study requires attention to the prespecified modeling strategy and multiple-comparison framework rather than ranking observed point estimates alone.
17. Why This Trial Matters Statistically
AXIOMATIC-SSP is a useful teaching example because it illustrates a clinical-trial setting in which the principal statistical question is explicitly about dose-response, rather than simply whether one experimental treatment differs from placebo.
| Concept | How it appears in AXIOMATIC-SSP |
|---|---|
| Randomization | The trial used randomized allocation with a parallel design. |
| Blinding | The trial used quadruple masking. |
| Dose-response analysis | MCP-Mod was used for the model-based primary endpoint assessment. |
| Relative risk | The registry reports RR for the primary composite endpoint and ischemic-stroke analyses. |
| Confidence intervals | Two-sided 95% confidence intervals accompany the posted effect estimates. |
| Bootstrap inference | The primary composite confidence intervals were based on bootstrap resampling with 10000 samples. |
| Wald inference | Prespecified ischemic-stroke analyses used Wald Confidence Limits. |
| Binary outcomes | The primary endpoint was registered as binary. |
| Composite endpoint | The primary endpoint combined new ischemic stroke during treatment with new covert brain infarction detected by MRI. |
18. Statistical Methods Explained in Practical Terms
Why use a model rather than one pairwise test?
Because the study evaluates multiple doses and the scientific question concerns whether the response changes systematically with dose.
Why use a bootstrap?
The registry specifies bootstrap resampling with 10000 samples for the primary composite confidence intervals, providing an empirical way to characterize uncertainty around the model-based estimate.
Why report RR instead of only a p-value?
The RR describes the estimated magnitude and direction of the relative effect, while the confidence interval describes its precision.
Why distinguish the composite from stroke alone?
The composite contains two event types, whereas the ischemic-stroke endpoint contains only stroke events. They therefore represent different outcome definitions.
19. Related Tutorials
Learn more about the methods used in this trial:
20. Related Calculators
21. Sources
- ClinicalTrials.gov: NCT03766581 — AXIOMATIC-SSP.
- PubMed: Publication indexed under PMID 38101902.
- PubMed: Publication indexed under PMID 38038437.
Continue through Clinical Biostats
Connect the trial's endpoints and statistical methods with deeper tutorials and practical statistical tools.
22. Record Summary
AXIOMATIC-SSP provides a focused example of dose-response methodology in a randomized phase 2 clinical trial. The primary endpoint was a binary composite assessed from randomization through up to 90 days, and the registry reports five MCP-Mod relative-risk estimates with two-sided 95% bootstrap confidence intervals. The additional ischemic-stroke analyses use Wald Confidence Limits and provide a complementary set of dose-specific relative-risk estimates.
The most important statistical lesson is that a dose-response trial should be interpreted at several levels: the effect estimate, the confidence interval, the endpoint definition, and the model used to relate dose to outcome. AXIOMATIC-SSP illustrates why those elements need to be considered together rather than reducing a multidose trial to a single point estimate.