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
RIVER-PCI was a randomized, parallel-group, triple-masked, phase 3 trial evaluating ranolazine versus placebo in participants with coronary artery disease and angina pectoris after percutaneous coronary intervention. The primary endpoint was a time-to-event composite of ischemia-driven revascularization or ischemia-driven hospitalization without revascularization.
| Feature | RIVER-PCI |
|---|---|
| Trial name | RIVER-PCI |
| ClinicalTrials.gov identifier | NCT01442038 |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Cardiovascular |
| Conditions | Coronary Artery Disease; Angina Pectoris |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 2651 |
| Interventions | Ranolazine; Placebo |
| Lead sponsor | Gilead Sciences |
| Sponsor type | Industry |
2. Clinical Question
The central question was whether treatment with ranolazine, compared with placebo, altered the time from randomization to the first occurrence of ischemia-driven revascularization or ischemia-driven hospitalization without revascularization in participants with incomplete vessel revascularization after PCI.
Population
Participants with coronary artery disease and angina pectoris enrolled in a phase 3 trial following percutaneous coronary intervention, with the primary efficacy analysis excluding participants with no qualifying PCI.
Intervention
Ranolazine.
Comparator
Placebo.
Primary question
Does ranolazine change the time to first ischemia-driven revascularization or ischemia-driven hospitalization without revascularization compared with placebo?
3. Trial Design
Ranolazine
- Ranolazine drug intervention.
- Primary comparison against placebo.
- Primary analysis based on the Full Analysis Set defined in the registry.
Placebo
- Placebo drug intervention.
- Comparator for all posted efficacy analyses.
- Primary analysis based on the same Full Analysis Set framework.
The study began in October 2011 and reached primary completion in February 2015. The primary endpoint used follow-up from baseline through the end of the study, with an average follow-up of 90 weeks.
4. Endpoints
The registry identifies 1 primary endpoint and reports 4 posted statistical analyses. All four analyses use time-to-event methodology and compare ranolazine with placebo.
| Endpoint role | Endpoint | Time frame | Analysis |
|---|---|---|---|
| Primary | Kaplan-Meier Estimates for Time From Randomization to First Occurrence of Ischemia-driven Revascularization or Ischemia-driven Hospitalization Without Revascularization | Baseline through end of study (average 90 weeks) | Kaplan-Meier estimation; Cox proportional-hazards model |
| Secondary | Kaplan-Meier Estimates for Time From Randomization to Sudden Cardiac Death | Baseline through end of study (average 90 weeks) | Cox proportional-hazards model |
| Secondary | Kaplan-Meier Estimates for Time From Randomization to Cardiovascular Death | Baseline through end of study (average 90 weeks) | Cox proportional-hazards model |
| Secondary | Kaplan-Meier Estimates for Time From Randomization to Myocardial Infarction | Baseline through end of study (average 90 weeks) | Cox proportional-hazards model |
5. Analysis Populations
The registry distinguishes the population used for the primary efficacy analysis from the broader Full Analysis Set used for the secondary analyses.
| Population | Registry definition / role |
|---|---|
| Full Analysis Set for primary endpoint | All participants in the Safety Analysis Set who were randomized and received at least one dose of study drug, except participants with no qualifying percutaneous coronary intervention (PCI; formerly known as angioplasty with stent). |
| Full Analysis Set for secondary endpoints | Full Analysis Set. |
| Safety Analysis Set | Participants who were randomized and received at least one dose of study drug, as reflected in the primary endpoint population definition. |
This distinction is statistically important. The primary analysis is described in the registry as incorporating an intention-to-treat concept, but the exact registry definition is narrower than the simple phrase "everyone randomized": participants had to have received at least one dose of study drug and, for the primary endpoint, had to have a qualifying PCI.
6. Statistical Methodology
Kaplan-Meier estimation
The primary endpoint and all three posted secondary endpoints are time-to-event outcomes. The registry states that time-to-event distributions were estimated using the Kaplan-Meier method. This approach estimates the probability of remaining event-free over time while allowing participants to be censored when their event status is not observed through the end of follow-up.
Here, di represents the number of events at a particular event time and ni represents the number of participants at risk immediately before that time.
Cox proportional-hazards model
The registry reports a Cox proportional-hazards model for the primary endpoint and each posted secondary endpoint. The model expresses the treatment comparison through a hazard ratio.
An HR below 1 corresponds to a lower estimated instantaneous event rate in the ranolazine group under the fitted model; an HR above 1 corresponds to a higher estimated instantaneous event rate.
Superiority framework
Each posted analysis is identified as a superiority hypothesis. That means the statistical question is whether the time-to-event experience differs between ranolazine and placebo, rather than whether ranolazine is merely no worse than placebo by a prespecified non-inferiority margin.
Intention-to-treat concept
The registry identifies intention-to-treat analysis as a concept associated with the posted analyses. At the same time, the primary endpoint population is explicitly defined as a Full Analysis Set based on participants who were randomized and received at least one dose of study drug, with an additional exclusion for participants without a qualifying PCI. The exact analysis population therefore matters when interpreting the reported hazard ratio.
Two-sided confidence intervals
The primary and secondary analyses report two-sided 95% confidence intervals for the hazard ratios. A confidence interval communicates statistical precision around the estimated treatment effect. It should be read together with the point estimate rather than treated as a substitute for the estimate itself.
7. Primary Result
The primary analysis compared ranolazine with placebo for time from randomization to the first occurrence of ischemia-driven revascularization or ischemia-driven hospitalization without revascularization, from baseline through the end of the study, with an average follow-up of 90 weeks.
Primary hazard ratio
95% CI: 0.818–1.099 · P = 0.48
Cox proportional-hazards model; Full Analysis Set; superiority hypothesis.
| Primary endpoint component | Reported analysis |
|---|---|
| Endpoint | Time from randomization to first ischemia-driven revascularization or ischemia-driven hospitalization without revascularization |
| Time frame | Baseline through end of study (average 90 weeks) |
| Groups | Ranolazine vs Placebo |
| Analysis population | Full Analysis Set as defined by the registry |
| Method | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| Estimate | 0.948 |
| 95% CI | 0.818–1.099 |
| P-value | 0.48 |
| Hypothesis | Superiority |
The estimated hazard ratio of 0.948 is below 1. Under the Cox model, this corresponds to an estimated instantaneous event rate that is approximately 94.8% of the comparator rate, or, equivalently, an estimated hazard about 5.2% lower in the ranolazine group relative to placebo. That is a model-based relative interpretation, not a statement that 5.2% fewer participants experienced the endpoint.
The 95% confidence interval of 0.818–1.099 spans 1.0. Thus, the interval includes values corresponding to lower and higher hazards for ranolazine relative to placebo. The interval also shows that the point estimate should not be interpreted as a highly precise estimate of the treatment effect.
The P-value of 0.48 addresses the statistical evidence against the superiority null hypothesis under the specified analysis. It does not measure the magnitude of the treatment effect, the probability that ranolazine works, or the probability that the null hypothesis is true. The hazard ratio and confidence interval provide the more direct description of effect size and precision.
Because this is a time-to-event analysis, interpretation also depends on censoring and on the assumptions underlying the Cox model, including the proportional-hazards assumption. The registry reports a Cox model but does not provide a proportional-hazards diagnostic in the information summarized here.
8. Secondary Endpoint Results
The registry reports three secondary time-to-event analyses. Each compares ranolazine with placebo using a Cox proportional-hazards model and a two-sided 95% confidence interval.
Sudden Cardiac Death
Hazard ratio for sudden cardiac death
95% CI: 0.244–1.691 · P = 0.40
Full Analysis Set; baseline through end of study (average 90 weeks).
The point estimate of 0.665 is below 1, corresponding to an estimated instantaneous hazard approximately 33.5% lower in the ranolazine group under the Cox model. The confidence interval is wide, extending from 0.244 to 1.691. It therefore includes both a materially lower and a materially higher hazard relative to placebo.
The P-value of 0.40 does not quantify the size of the apparent treatment difference. The wide confidence interval is particularly important here because the estimate is substantially less precise than the primary endpoint estimate.
Cardiovascular Death
Hazard ratio for cardiovascular death
95% CI: 0.579–1.994 · P = 0.82
Full Analysis Set; baseline through end of study (average 90 weeks).
The point estimate of 1.073 is above 1, corresponding to an estimated instantaneous cardiovascular-death hazard approximately 7.3% higher in the ranolazine group under the fitted model. The estimate should not be interpreted in isolation: the 95% confidence interval of 0.579–1.994 is broad and includes 1.0.
The P-value of 0.82 indicates little statistical evidence against the superiority null hypothesis for this endpoint under the posted analysis. It does not establish equivalence between treatments, nor does it prove that the true hazard ratio is exactly 1.
Myocardial Infarction
Hazard ratio for myocardial infarction
95% CI: 0.745–1.256 · P = 0.81
Full Analysis Set; baseline through end of study (average 90 weeks).
The point estimate of 0.968 corresponds to an estimated instantaneous myocardial-infarction hazard approximately 3.2% lower in the ranolazine group under the Cox model. The 95% confidence interval of 0.745–1.256 includes 1.0 and permits a range of treatment effects on either side of the null value.
The P-value of 0.81 is not a measure of the clinical importance of the observed point estimate. It reflects the evidence against the superiority null hypothesis within this statistical framework.
9. Summary of Posted Statistical Results
| Endpoint | Role | HR | 95% CI | P-value |
|---|---|---|---|---|
| Ischemia-driven revascularization or ischemia-driven hospitalization without revascularization | Primary | 0.948 | 0.818–1.099 | 0.48 |
| Sudden cardiac death | Secondary | 0.665 | 0.244–1.691 | 0.40 |
| Cardiovascular death | Secondary | 1.073 | 0.579–1.994 | 0.82 |
| Myocardial infarction | Secondary | 0.968 | 0.745–1.256 | 0.81 |
A useful way to read this table is to separate three quantities that answer different questions. The hazard ratio describes the relative treatment effect estimated by the Cox model. The 95% confidence interval describes the precision and range of values compatible with the statistical model and data. The P-value describes evidence against the specified null hypothesis; it does not describe the magnitude or clinical importance of the effect.
10. Serious Adverse Events
The registry reports serious adverse events by treatment arm using the number of affected participants over the corresponding number at risk.
| Treatment arm | Participants with serious adverse events | Participants at risk |
|---|---|---|
| Ranolazine | 516 | 1322 |
| Placebo | 485 | 1297 |
These safety figures are descriptive counts by treatment arm. They should be interpreted separately from the time-to-event efficacy analyses because a serious adverse event is a safety outcome, whereas the primary efficacy endpoint is a composite time-to-event outcome involving ischemia-driven revascularization or ischemia-driven hospitalization without revascularization.
The denominators also matter. The reported serious-adverse-event counts are presented with the corresponding numbers at risk rather than being treated as if they represented the same analysis population used for every efficacy endpoint.
11. Statistical Methods Explained
Why was Kaplan-Meier estimation used?
The primary and secondary endpoints are defined by the time from randomization until an event occurs. Kaplan-Meier estimation is designed for this structure because participants may have different follow-up times and some may not experience the event during observation. Instead of reducing the outcome to a simple yes/no event count, the method uses the timing of events and censoring to estimate the event-free distribution over follow-up.
What does a hazard ratio of 0.948 mean?
A hazard ratio of 0.948 means that the estimated instantaneous event rate in the ranolazine group is 0.948 times the corresponding rate in the placebo group under the fitted Cox model. This can be expressed as an estimated 5.2% relative reduction in hazard, calculated as 1 minus 0.948. It does not mean a 5.2-percentage-point reduction in the proportion of participants experiencing the endpoint.
Why does the confidence interval matter?
The primary 95% confidence interval is 0.818–1.099. The point estimate alone would give only one estimate of the treatment effect. The interval adds information about statistical precision and shows that the observed estimate is compatible with a range of possible underlying hazard ratios. Because the interval crosses 1.0, the data do not exclude either a lower or higher hazard relative to placebo under the specified model.
Why doesn't the P-value measure effect size?
The primary P-value is 0.48. A P-value is calculated relative to a null hypothesis and a specified statistical model; it does not describe how large or clinically important the treatment effect is. Effect size is communicated primarily by the hazard ratio, while the confidence interval communicates uncertainty around that estimate.
What does "superiority" mean here?
The registry identifies the hypothesis type as superiority. The analysis therefore tests whether the treatment groups differ in the specified time-to-event outcome rather than asking whether ranolazine is within a prespecified non-inferiority margin of placebo. The interpretation of the primary result should therefore be tied to the reported superiority framework.
Why does the analysis population matter?
The primary analysis uses a Full Analysis Set consisting of participants in the Safety Analysis Set who were randomized and received at least one dose of study drug, with participants lacking a qualifying PCI excluded. This is important because the phrase "intention-to-treat" can otherwise be interpreted too broadly. The registry's actual population definition determines which participants contribute to the reported primary hazard ratio.
What assumption is important for the Cox model?
The Cox proportional-hazards model assumes that the relative hazard represented by the hazard ratio is sufficiently stable over time for the model to provide an appropriate summary. If proportional hazards do not hold, a single HR may obscure changes in the treatment contrast during follow-up. The registry reports the Cox model but does not provide a proportional-hazards diagnostic in the summarized information.
12. How to Read the Primary Result Statistically
The primary HR of 0.948 lies close to 1.0. The treatment contrast estimated by the Cox model is therefore relatively small on the hazard-ratio scale.
The 95% CI of 0.818–1.099 extends on both sides of 1.0. The interval therefore indicates uncertainty about whether the underlying relative hazard is below or above the null value.
The P-value of 0.48 does not provide evidence against the superiority null hypothesis at conventional significance thresholds. Importantly, this does not establish that the two treatments are identical or clinically equivalent.
The posted primary analysis does not provide a time-specific absolute event probability in the information summarized here. Consequently, the HR should not be converted into an absolute risk difference without the necessary event-time information.
13. Understanding the Secondary Results
The three secondary analyses illustrate why a treatment-effect estimate should always be presented together with its uncertainty.
| Endpoint | Point estimate | 95% CI interpretation | Key statistical feature |
|---|---|---|---|
| Sudden cardiac death | HR 0.665 | 0.244–1.691 | Wide interval spanning 1.0 |
| Cardiovascular death | HR 1.073 | 0.579–1.994 | Interval spans 1.0 and is broad |
| Myocardial infarction | HR 0.968 | 0.745–1.256 | Interval spans 1.0 |
The sudden-cardiac-death estimate is below 1, while the cardiovascular-death estimate is above 1 and the myocardial-infarction estimate is close to 1. These point estimates should not be interpreted as evidence that the treatment effects are definitively different from one another. Their confidence intervals overlap the null value, and the registry does not provide an interaction analysis establishing heterogeneity among these endpoints.
There is also an important distinction between a non-significant P-value and evidence of no effect. For example, an HR of 0.665 with a wide confidence interval is statistically different from an HR known to be exactly 1.0 in terms of the information it conveys: the former suggests a point estimate below 1 but with substantial uncertainty. The confidence interval is therefore essential to the interpretation.
14. Composite Endpoint Considerations
The primary endpoint combines two clinically distinct events: ischemia-driven revascularization and ischemia-driven hospitalization without revascularization. A composite endpoint can increase the number of observed events and therefore improve statistical efficiency, but interpretation depends on understanding the components.
What the composite measures
The primary endpoint captures time to the first occurrence of either ischemia-driven revascularization or ischemia-driven hospitalization without revascularization.
Why first occurrence matters
Once a participant experiences the first qualifying component event, that participant has reached the composite endpoint for the primary time-to-event analysis.
Why components matter
A composite treatment effect does not automatically imply that every component is affected to the same degree. Component-specific results are needed to understand the clinical composition of the overall endpoint.
What the HR does not show
The primary HR alone does not reveal the separate absolute frequencies or hazard ratios for the two components.
This is one reason a composite endpoint should be read at the endpoint-definition level rather than reduced to a single label such as "ischemic events." The registry's exact definition specifies which events constitute the outcome.
15. Censoring and Follow-Up
All posted efficacy endpoints use the time from randomization through the end of the study, with an average follow-up of 90 weeks. Time-to-event analysis is particularly useful in this setting because participants can contribute different amounts of observed follow-up.
A participant who reaches a qualifying event contributes an event time. A participant who remains without the event through the available observation period contributes information up to the censoring time.
Kaplan-Meier estimation and the Cox model both depend on the handling of censoring. Standard survival-analysis interpretation generally relies on assumptions about the relationship between censoring and the event process. The registry identifies the survival methods but does not provide a detailed missing-data or censoring-assumption analysis in the information summarized here.
16. Randomization and Masking as Statistical Design Features
RIVER-PCI used randomized allocation in a parallel-group design and was triple-masked. Randomization is central to the causal comparison because it is intended to distribute prognostic factors between treatment groups without systematically assigning participants according to their baseline risk.
Masking addresses a different source of bias. By keeping treatment assignment masked, the design can reduce the influence of knowledge of treatment on participant behavior, clinical management, outcome assessment, and other aspects of trial conduct. The registry identifies the study as triple-masked but does not specify in the summarized information which three groups of individuals were masked.
| Design feature | Statistical role |
|---|---|
| Randomization | Creates the treatment comparison used for causal efficacy inference. |
| Parallel design | Participants remain assigned to their randomized treatment groups for the treatment comparison rather than being assigned sequentially to both interventions. |
| Triple masking | Reduces the potential influence of treatment knowledge during the conduct of the trial. |
| Superiority hypothesis | Frames the primary comparison as a test for a treatment difference rather than a non-inferiority margin. |
17. Safety and Efficacy Use Different Statistical Perspectives
The RIVER-PCI record illustrates a fundamental distinction in clinical-trial statistics: efficacy and safety answer different questions and can use different analysis populations or summaries.
Efficacy
The primary efficacy outcome is a time-to-event endpoint analyzed with Kaplan-Meier estimation and a Cox proportional-hazards model.
Safety
Serious adverse events are reported as affected participants and participants at risk in each treatment arm.
A hazard ratio for the primary endpoint should therefore not be combined mathematically with serious-adverse-event counts to produce a single "net benefit" statistic. Such a conclusion would require an explicitly defined benefit-risk framework and appropriate statistical methodology.
18. Limitations
- Analysis-population definition: the primary endpoint analysis excludes participants with no qualifying PCI and requires participants to have been randomized and to have received at least one dose of study drug. This should be considered when interpreting the phrase intention-to-treat.
- Composite endpoint: the primary outcome contains two different clinical events. The overall hazard ratio does not establish that both components have identical treatment effects.
- Time-to-event assumptions: Cox modeling relies on assumptions about the hazard structure, including proportional hazards. The registry identifies the Cox model but does not report a diagnostic assessment in the summarized record.
- Confidence-interval width: the secondary sudden-cardiac-death estimate has a broad 95% confidence interval, illustrating substantial statistical uncertainty around the point estimate.
- Secondary endpoints: the three secondary analyses are separate statistical comparisons. Their P-values should not automatically be interpreted as though they constitute a single independently powered confirmatory testing family unless the multiplicity framework is specified.
- Absolute risks: the posted analyses provide hazard ratios and confidence intervals but do not provide time-specific absolute event estimates in the information summarized here.
- Generalizability: the population represented by the trial is defined by the study's eligibility and PCI-related requirements, so results should be interpreted within that clinical context.
19. Why This Trial Matters Statistically
RIVER-PCI is a useful teaching example because its registry record brings together several core concepts in survival analysis without relying on a large collection of different endpoint types. The primary outcome is explicitly time-to-event, the registry provides a Kaplan-Meier definition and a Cox proportional-hazards model, and the results include both point estimates and two-sided 95% confidence intervals.
| Concept | How it appears in RIVER-PCI |
|---|---|
| Randomization | Participants were randomized to ranolazine or placebo in a parallel design. |
| Blinding | The study was triple-masked. |
| Kaplan-Meier estimation | Time-to-event distributions were estimated by the Kaplan-Meier method. |
| Hazard ratio | The treatment effect was expressed as a hazard ratio for each posted statistical analysis. |
| Cox proportional-hazards model | The registry reports a Cox model for the primary and secondary endpoints. |
| Confidence intervals | Each posted hazard ratio includes a two-sided 95% confidence interval. |
| P-values | Each posted analysis includes a P-value for the superiority comparison. |
| Intention-to-treat concept | The analysis text identifies intention-to-treat analysis, while the primary population has a specific Full Analysis Set definition. |
| Composite endpoint | The primary endpoint combines ischemia-driven revascularization and ischemia-driven hospitalization without revascularization. |
| Safety analysis | Serious adverse events are reported separately by treatment arm. |
20. A Practical Framework for Reading the RIVER-PCI Results
A disciplined interpretation can proceed in five steps.
What event is being measured?
The primary estimand concerns time from randomization to the first occurrence of either ischemia-driven revascularization or ischemia-driven hospitalization without revascularization.
Who contributes to the analysis?
The primary analysis uses the registry-defined Full Analysis Set, including randomized participants who received at least one dose of study drug, with participants lacking a qualifying PCI excluded.
How is the treatment effect estimated?
The registry reports Kaplan-Meier estimation for the time-to-event distribution and a Cox proportional-hazards model for the hazard ratio.
What is the estimate and how uncertain is it?
The primary HR is 0.948 with a 95% CI of 0.818–1.099. The interval is as important as the point estimate because it describes the uncertainty surrounding the estimated treatment effect.
What does the result establish?
The P-value of 0.48 describes the evidence against the superiority null hypothesis under the specified model. It does not establish equivalence, eliminate uncertainty, or describe an absolute risk difference.
21. Primary Result in Context
The primary hazard ratio is close to the null value of 1.0, while the confidence interval extends on both sides of 1.0. This combination is more informative than the P-value alone: it shows both the direction of the point estimate and the range of treatment effects compatible with the statistical analysis.
Primary statistical summary
Hazard ratio for the first occurrence of ischemia-driven revascularization or ischemia-driven hospitalization without revascularization.
95% CI 0.818–1.099 · P = 0.48
The result should not be translated into a claim that ranolazine and placebo have identical effects. A non-significant superiority test is not an equivalence test. Conversely, the point estimate alone should not be treated as proof of a treatment benefit. The appropriate statistical description incorporates all three reported quantities: the HR, its confidence interval, and the P-value.
22. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary Cox model produced an HR of 0.948 with a two-sided 95% CI of 0.818–1.099 and P = 0.48. The confidence interval includes the null hazard ratio of 1.0.
Clinical interpretation
The primary endpoint concerns ischemia-driven revascularization or ischemia-driven hospitalization without revascularization. Clinical interpretation requires attention to the definition of the composite endpoint, follow-up period, analysis population, and safety findings rather than the HR alone.
Keeping these two levels separate is important. Statistical significance is a property of a statistical comparison under specified assumptions; clinical importance concerns the magnitude, consequences, and context of the outcomes being studied.
23. Sources
- ClinicalTrials.gov: RIVER-PCI, NCT01442038.
- PubMed: PMID 26474810.
- PubMed: PMID 26555329.
- PubMed: PMID 30771861.
Continue through the Clinical Biostats statistical methods library
Explore the methods that appear in RIVER-PCI, including survival analysis, hazard ratios, confidence intervals, randomization, and time-to-event endpoints.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Calculators
26. Record Summary
RIVER-PCI provides a focused example of randomized time-to-event analysis. The trial used randomization, a parallel design, triple masking, Kaplan-Meier estimation, and Cox proportional-hazards modeling to evaluate a primary composite endpoint and three secondary time-to-event outcomes. The primary analysis reported an HR of 0.948, with a two-sided 95% CI of 0.818–1.099 and P = 0.48.
The most useful statistical interpretation combines the hazard ratio, confidence interval, and P-value rather than relying on any single quantity. It also requires attention to the exact analysis population, the composite nature of the primary endpoint, censoring, the proportional-hazards framework, and the distinction between efficacy and safety outcomes.