This page separates reported trial results from statistical interpretation. Numerical results and trial characteristics are restricted to the ClinicalTrials.gov data posted on ClinicalTrials.gov for VICTORIA. The registry record is the official trial record.
1. Trial at a Glance
VICTORIA was a randomized, double-blind, parallel-group phase 3 trial evaluating vericiguat versus placebo in participants with heart failure and chronic heart failure with reduced ejection fraction. The registry reports 5050 enrolled participants, two study arms, one registered primary time-to-event endpoint, and eight posted statistical analyses.
| Feature | VICTORIA |
|---|---|
| Trial name | VICTORIA |
| Brief title | A Study of Vericiguat in Participants With Heart Failure With Reduced Ejection Fraction (HFrEF) (MK-1242-001) |
| Phase | Phase 3 |
| Status | COMPLETED |
| Therapeutic area | Cardiology |
| Conditions | Heart Failure; Chronic Heart Failure With Reduced Ejection Fraction |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 5050 |
| Interventions | Vericiguat; placebo for vericiguat |
| Lead sponsor | Merck Sharp & Dohme LLC |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT02861534 |
2. Clinical Question
The primary statistical question was whether the time to first occurrence of the composite endpoint of cardiovascular death or heart failure hospitalization differed between participants randomized to vericiguat and those randomized to placebo.
Population
Participants with heart failure and chronic heart failure with reduced ejection fraction, as described in the registered trial condition and brief title.
Intervention
Vericiguat.
Comparator
Placebo for vericiguat.
Primary question
Does vericiguat change the time to first cardiovascular death or heart failure hospitalization relative to placebo?
3. Trial Design
Vericiguat
- Intervention: vericiguat
- Drug intervention
- Compared with placebo for vericiguat
Placebo
- Intervention: placebo for vericiguat
- Drug intervention
- Comparator for the vericiguat group
4. Trial Timing and Analysis Cutoff
Trial start
The registry lists 20-September-2016 as the trial start date.
Primary completion
The registry lists 18-June-2019 as the primary completion date and uses the same date as the primary analysis database cutoff.
Approximately 33 months
The registered primary endpoint and the posted statistical analyses use a time frame of up to approximately 33 months through the 18-June-2019 primary analysis database cutoff.
5. Primary Endpoint
| Endpoint | Registered definition / time frame | Posted analysis |
|---|---|---|
| Time to First Occurrence of Composite Endpoint of Cardiovascular (CV) Death or Heart Failure (HF) Hospitalization | Time to first occurrence of the composite endpoint of CV death or HF hospitalization; up to approximately 33 months, through the primary analysis database cutoff date of 18-June-2019. | All randomized participants; vericiguat vs placebo; Cox proportional hazard model; hazard ratio. |
The registry's primary-endpoint definition states that the endpoint was analyzed using a one-sided stratified log-rank test. It also states that randomized participants without an HF hospitalization or CV death event at the time of analysis were censored according to the available follow-up information or the primary analysis database cutoff. The posted statistical-analysis record separately reports the Cox proportional-hazards model used for the hazard-ratio estimate.
6. Primary Result
Time to First Cardiovascular Death or Heart Failure Hospitalization
Hazard ratio: vericiguat vs placebo
95% CI: 0.82–0.98 · P = 0.019
Analysis population: all randomized participants · Hypothesis type: superiority
| Endpoint | Vericiguat vs placebo | 95% CI | P-value | Analysis |
|---|---|---|---|---|
| Time to first CV death or HF hospitalization | HR 0.90 | 0.82–0.98 | 0.019 | Cox proportional-hazards model |
What the estimate means: an HR of 0.90 means that, under the fitted Cox model, the estimated instantaneous rate of experiencing the composite event was approximately 10% lower with vericiguat than with placebo over the analyzed follow-up.
What it does not mean: it does not mean that exactly 10% fewer participants experienced the endpoint, that every participant had a 10% reduction in risk, or that the absolute probability of an event was reduced by 10 percentage points. A hazard ratio is a relative, model-based time-to-event measure.
Precision: the 95% CI of 0.82–0.98 describes uncertainty around the estimated hazard ratio under the analysis framework. The interval is relatively narrow compared with the estimate itself, but it still represents statistical uncertainty rather than a range of effects that must occur in individual patients.
The p-value: P = 0.019 is evidence against the null hypothesis under the specified testing framework. It does not measure the size of the treatment effect. Effect size is communicated by the HR, while precision is communicated by the confidence interval.
Important time-to-event cautions: the Cox interpretation depends on the proportional-hazards framework. The registry also describes censoring of participants who had not experienced the composite event at the time of analysis. Censoring assumptions therefore matter to interpretation. In addition, the registry identifies the formal primary endpoint analysis as a one-sided stratified log-rank test, whereas the posted estimate is from a Cox proportional-hazards model; these should not be treated as interchangeable statistical procedures.
7. Secondary Endpoint Results
The registry posts seven additional statistical analyses. Six are time-to-event analyses and one is a binary safety-related endpoint. The time-to-event analyses generally use hazard ratios from Cox proportional-hazards models, while the two binary safety endpoints use risk differences estimated with the Miettinen & Nurminen method.
| Secondary endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Time to the First Occurrence of CV Death | Hazard ratio | 0.93 | 0.81–1.06 | 0.269 |
| Time to the First Occurrence of HF Hospitalization | Hazard ratio | 0.90 | 0.81–1.00 | 0.048 |
| Time to Total HF Hospitalizations (Including First and Recurrent Events) | Hazard ratio | 0.91 | 0.84–0.99 | 0.023 |
| Time to First Occurrence of Composite Endpoint of All-Cause Mortality or HF Hospitalization | Hazard ratio | 0.90 | 0.83–0.98 | 0.021 |
| Time to All-Cause Mortality | Hazard ratio | 0.95 | 0.84–1.07 | 0.377 |
| Percentage of Participants Who Experienced Symptomatic Hypotension | Risk difference | 1.2 | -0.3–2.8 | 0.121 |
| Percentage of Participants Who Experienced Syncope | Risk difference | 0.6 | -0.5–1.6 | 0.303 |
Cardiovascular Death
The estimate is below 1, but the reported 95% confidence interval extends from below 1 to above 1. The registry identifies this as a superiority analysis using a Cox proportional-hazards model in all randomized participants.
Heart Failure Hospitalization
The hazard-ratio estimate corresponds to an approximately 10% lower estimated instantaneous rate of first HF hospitalization under the Cox model. The upper confidence limit reaches 1.00, illustrating why the estimate should be read together with its interval rather than from the point estimate alone.
Total Heart Failure Hospitalizations
Unlike a simple time-to-first-event endpoint, this analysis concerns total HF hospitalizations, including first and recurrent events. The registry reports an Andersen-Gill model rather than a standard Cox proportional-hazards model for this endpoint.
All-Cause Mortality or Heart Failure Hospitalization
This composite uses all-cause mortality rather than cardiovascular mortality in combination with HF hospitalization. The registry reports a Cox proportional-hazards model in all randomized participants.
All-Cause Mortality
The point estimate is below 1, but the confidence interval spans 1.00. The reported p-value is 0.377. This illustrates why a point estimate below 1 should not by itself be described as evidence of a statistically established treatment effect.
Symptomatic Hypotension
The analysis population was all randomized participants who received at least one dose of study treatment. The registry reports the Miettinen & Nurminen method for this percentage comparison.
Syncope
The analysis population was all randomized participants who received at least one dose of study treatment. The reported method was the Miettinen & Nurminen method.
8. Understanding the Primary Hazard Ratio
The primary HR of 0.90 is a useful example of why clinical-trial interpretation should separate relative effect, precision, and statistical evidence.
Relative effect
HR 0.90 corresponds to a 10% lower estimated instantaneous event rate under the fitted Cox model, because 1 − 0.90 = 0.10.
Precision
The 95% CI of 0.82–0.98 shows the uncertainty surrounding the estimated hazard ratio under the reported analysis.
Statistical evidence
The reported P = 0.019 quantifies evidence under the specified hypothesis-testing framework; it is not a measure of clinical magnitude.
Absolute effects
The hazard ratio does not provide an absolute event probability or an absolute risk difference. Those require the underlying event-time distribution or corresponding absolute-risk estimates.
A hazard ratio describes the relative instantaneous event rate within a time-to-event model. It is therefore different from comparing the proportion of participants who eventually experience an event. The distinction becomes particularly important when follow-up times differ, censoring occurs, or the event rate changes over time.
The interval 0.82–0.98 provides information about statistical precision that the point estimate alone cannot provide. A point estimate of 0.90 could be interpreted very differently if its confidence interval were extremely wide. Here, the ClinicalTrials.gov record gives a specific two-sided 95% interval that remains below 1.00.
P = 0.019 does not say that the probability of benefit is 98.1%, nor does it say that the treatment effect is 1.9%. The p-value is a statement about the compatibility of the observed result with a null hypothesis under the specified statistical framework.
9. Statistical Methodology
Cox Proportional-Hazards Model
The registry reports a Cox proportional-hazards model for the primary endpoint and for several secondary time-to-event endpoints. The model estimates a relative hazard associated with treatment while using event times and censoring information rather than reducing follow-up to a simple yes/no event indicator.
For a binary treatment indicator, exp(β) is interpreted as the hazard ratio comparing the treatment groups under the proportional-hazards model.
Stratified Log-Rank Test
The registered primary endpoint definition states that the composite endpoint was analyzed using a one-sided stratified log-rank test. A log-rank test compares observed and expected event patterns between treatment groups across follow-up, while stratification allows the comparison to account for prespecified strata.
The ClinicalTrials.gov record does not identify the variables used for stratification. They therefore should not be inferred from external sources or from the mere fact that the registry describes the test as stratified.
Kaplan-Meier Estimation
The VICTORIA primary endpoint is a time-to-event outcome. Kaplan-Meier estimation is the standard nonparametric framework for describing the event-time distribution in the presence of right censoring.
Here, di is the number of events at time ti, and ni is the number at risk immediately before that time.
Andersen-Gill Model for Recurrent Hospitalizations
Total HF hospitalizations, including first and recurrent events, were analyzed using an Andersen-Gill model. This is an important methodological distinction: a recurrent-event endpoint cannot always be adequately represented by simply analyzing the time to the first hospitalization.
By retaining recurrent events, the analysis addresses the occurrence of multiple hospitalization events rather than discarding later events after the first one. The registry reports the effect measure as a hazard ratio with estimate 0.91 and a 95% CI of 0.84–0.99.
Miettinen & Nurminen Method
The symptomatic-hypotension and syncope endpoints are binary outcomes expressed as percentages of participants. The registry reports the Miettinen & Nurminen method for the difference in percentages.
A positive risk difference means the percentage experiencing the event was higher in the vericiguat group; a negative value means it was lower. The reported confidence intervals quantify uncertainty around that difference.
Analysis Populations
The primary and time-to-event efficacy analyses in the ClinicalTrials.gov record uses all randomized participants. The symptomatic-hypotension and syncope analyses use all randomized participants who received at least 1 dose of study treatment.
| Endpoint category | Analysis population | Statistical method |
|---|---|---|
| Primary composite time-to-event endpoint | All randomized participants | One-sided stratified log-rank test for the registered endpoint; Cox proportional-hazards model for the posted HR |
| Secondary time-to-event endpoints | All randomized participants | Cox proportional-hazards model, except total HF hospitalizations, which used an Andersen-Gill model |
| Symptomatic hypotension | All randomized participants who received at least 1 dose | Miettinen & Nurminen method |
| Syncope | All randomized participants who received at least 1 dose | Miettinen & Nurminen method |
10. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
The primary endpoint is defined by time to first occurrence of cardiovascular death or HF hospitalization. A Cox model is designed for this setting because it uses both event timing and censoring information and expresses the treatment comparison as a hazard ratio.
What does an HR of 0.90 mean?
It means that the estimated instantaneous event rate under the fitted model was 0.90 times that of the comparator, or approximately 10% lower. It does not mean that 10% of participants avoided the endpoint or that the absolute probability was reduced by 10 percentage points.
Why is the confidence interval important?
The estimate 0.90 is only one point on the range of values compatible with the statistical uncertainty represented by the analysis. The 95% CI of 0.82–0.98 shows the precision of the reported HR and should be interpreted together with the point estimate.
Why is the primary endpoint described with both a log-rank test and a Cox model?
The registry identifies a one-sided stratified log-rank test as the analysis of the registered primary endpoint, while the posted statistical analysis reports the Cox proportional-hazards model as the method used to generate the HR. The test and the model therefore play related but distinct roles in the statistical presentation.
Why is total HF hospitalization analyzed differently from first HF hospitalization?
A time-to-first-event analysis stops counting once the first hospitalization occurs. The total-hospitalization endpoint explicitly includes first and recurrent events, so the registry reports an Andersen-Gill model designed for recurrent-event data.
What does a risk difference of 1.2 mean?
For symptomatic hypotension, the reported effect measure is a difference in percentages. An estimate of 1.2 means a 1.2-percentage-point difference between the treatment groups under the registry's direction of comparison. It is not a hazard ratio and should not be interpreted as a relative 1.2% increase.
Why should the secondary p-values be interpreted cautiously?
The ClinicalTrials.gov record reports several secondary hypothesis tests but do not specify an overall multiplicity-adjustment scheme. Without that information, a reader should distinguish the reported nominal p-values from a claim that the entire set of secondary comparisons is familywise-error controlled.
11. Primary Endpoint and Censoring
The registry's definition specifies that randomized participants without an HF hospitalization or CV death event at the time of analysis were censored at their last available information, the date of their non-CV death, or the primary analysis database cutoff date of 18-June-2019, whichever occurred according to the registry's stated censoring rule.
Event
The first occurrence of either cardiovascular death or HF hospitalization constitutes the composite endpoint event.
Censoring
Participants without a qualifying event at analysis contribute follow-up until their applicable censoring time.
Composite endpoint
The first qualifying component event determines the time to the composite endpoint.
Analysis cutoff
The primary analysis database cutoff was 18-June-2019, with the registered time frame extending up to approximately 33 months.
12. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm. The registry summary gives the number affected and the number at risk for each arm.
| Safety measure | Vericiguat | Placebo |
|---|---|---|
| Serious adverse events | 852 / 2519 | 897 / 2515 |
The denominators shown above are the registry-reported numbers at risk for this safety summary. The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or p-value for serious adverse events, so this page does not create one.
13. Statistical Interpretation of the Secondary Results
| Finding | Statistical reading |
|---|---|
| CV death: HR 0.93 | The estimated hazard is below 1, but the 95% CI of 0.81–1.06 spans 1.00 and the reported P-value is 0.269. |
| First HF hospitalization: HR 0.90 | The estimated hazard is approximately 10% lower under the model; the 95% CI is 0.81–1.00 and P = 0.048. |
| Total HF hospitalizations: HR 0.91 | The recurrent-event analysis gives an estimated HR of 0.91 with 95% CI 0.84–0.99 and P = 0.023. |
| All-cause mortality or HF hospitalization: HR 0.90 | The estimated hazard is approximately 10% lower under the model, with 95% CI 0.83–0.98 and P = 0.021. |
| All-cause mortality: HR 0.95 | The point estimate is below 1, while the 95% CI of 0.84–1.07 spans 1.00 and P = 0.377. |
| Symptomatic hypotension: RD 1.2 | The estimated percentage difference is 1.2 percentage points, with 95% CI -0.3 to 2.8 and P = 0.121. |
| Syncope: RD 0.6 | The estimated percentage difference is 0.6 percentage points, with 95% CI -0.5 to 1.6 and P = 0.303. |
A useful statistical distinction emerges from these results: point estimates below 1 are not equivalent to statistically established differences. For example, both CV death and all-cause mortality have HR estimates below 1, but their confidence intervals include 1.00 and their reported p-values are 0.269 and 0.377, respectively.
14. What the Primary Result Does — and Does Not — Establish
The registry reports a superiority analysis for the primary composite time-to-event endpoint, with a Cox-model HR of 0.90, a two-sided 95% CI of 0.82–0.98, and P = 0.019. The registered endpoint definition identifies a one-sided stratified log-rank test as the formal endpoint analysis.
The HR does not specify the absolute number of cardiovascular deaths or HF hospitalizations prevented, the absolute probability of experiencing the composite event, or the treatment effect for every individual participant.
The primary endpoint combines cardiovascular death and HF hospitalization. A composite endpoint can increase the number of observed events and improve statistical efficiency, but its interpretation depends on understanding that its components are being treated as a single time-to-first-event outcome.
The registry identifies one primary endpoint and multiple secondary endpoints. Secondary results provide additional information about specific components and related outcomes, but their statistical role is different from that of the prespecified primary endpoint.
15. Hypothesis Testing and P-values
The ClinicalTrials.gov record identifies the primary hypothesis type as superiority. This means the statistical objective was to test whether the treatment groups differed in the specified primary time-to-event endpoint rather than to establish non-inferiority against a prespecified margin.
The p-value and confidence interval answer related but different questions. The p-value concerns evidence against the null under the testing framework; the confidence interval describes the uncertainty around the estimated effect.
The registry describes the primary endpoint as using a one-sided stratified log-rank test, while the posted Cox-model confidence interval is explicitly two-sided 95%. These conventions should be retained as reported rather than silently converting one testing framework into another.
16. Missing Data, Imputation, and Other Design Features
The ClinicalTrials.gov record provides specific information about censoring for the primary time-to-event endpoint, but they do not report a separate missing-data or imputation strategy for the posted statistical analyses.
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Non-inferiority margin | Not reported in the ClinicalTrials.gov record; the primary hypothesis is identified as superiority. |
| Crossover | Not reported in the ClinicalTrials.gov record. |
| Factorial design | Not reported; the design model is parallel with two arms. |
| Multiplicity adjustment | No alpha-allocation or multiplicity-adjustment scheme is reported in the ClinicalTrials.gov record. |
| Interim analysis | Not reported in the ClinicalTrials.gov record. |
| Missing-data imputation | No imputation method is reported in the ClinicalTrials.gov record. |
| Stratification variables | The primary endpoint is described as using a stratified log-rank test, but the ClinicalTrials.gov record does not identify the stratification variables. |
| Bayesian methods | No Bayesian method is reported. |
17. Why the Analysis Population Matters
The registry specifies all randomized participants for the primary and most secondary time-to-event analyses. This aligns the main efficacy comparison with treatment assignment rather than restricting the analysis to participants who remained on treatment.
Randomized analysis
All randomized participants form the analysis population for the primary endpoint and the reported time-to-event secondary endpoints.
Safety-related binary endpoints
Symptomatic hypotension and syncope use randomized participants who received at least one dose of study treatment.
Why the distinction matters
Changing the analysis population can change the estimand. Results from treated participants should not automatically be described as if they were identical to results based on randomization.
Causal interpretation
Randomization provides the foundation for the treatment-group comparison, while subsequent follow-up and censoring determine how much time-to-event information is available for analysis.
18. Time-to-Event Endpoints: A Statistical Walkthrough
VICTORIA is particularly useful for teaching survival analysis because its primary endpoint is not simply whether an event occurred. The analysis incorporates when the event occurred and how long participants remained under observation without experiencing the event.
| Step | Statistical question | VICTORIA application |
|---|---|---|
| 1. Define time origin | When does follow-up begin? | The endpoint is described as time to first occurrence following randomization. |
| 2. Define event | What counts as the endpoint? | First CV death or HF hospitalization. |
| 3. Follow participants | How much event-time information is observed? | Up to approximately 33 months through the 18-June-2019 cutoff. |
| 4. Handle censoring | What happens when no qualifying event is observed? | Participants are censored according to the registry's stated follow-up/cutoff rules. |
| 5. Compare groups | Do event-time distributions differ? | The registered primary analysis uses a one-sided stratified log-rank test. |
| 6. Quantify relative effect | How large is the modeled treatment difference? | The posted Cox analysis reports HR 0.90 with 95% CI 0.82–0.98. |
19. Limitations
- Registry-level detail: the ClinicalTrials.gov record does not include all protocol or statistical-analysis-plan details, so some design features cannot be characterized beyond what is explicitly reported.
- Stratification details: the primary endpoint is described as using a stratified log-rank test, but the ClinicalTrials.gov record does not identify the stratification factors.
- Proportional-hazards assumption: the Cox hazard ratio is model-based and its conventional interpretation relies on the proportional-hazards framework.
- Composite endpoint: cardiovascular death and HF hospitalization are combined into one primary endpoint, so the primary HR is not an estimate for either component alone.
- Recurrent-event endpoint: total HF hospitalizations require a different analytical framework from time to first HF hospitalization, and the registry reports an Andersen-Gill model for this purpose.
- Multiplicity: multiple secondary endpoints are reported, but the ClinicalTrials.gov record does not specify an overall multiplicity-adjustment strategy.
- Censoring: time-to-event inference depends on the handling and assumptions associated with censoring and follow-up.
- Safety comparisons: the registry-reported serious-adverse-event data are descriptive counts and denominators without a formal comparison, confidence interval, or p-value.
- Incomplete numerical detail: the ClinicalTrials.gov record does not provide absolute event rates, Kaplan-Meier survival probabilities, median event times, baseline characteristics, subgroup analyses, or graphical survival estimates.
20. Why This Trial Matters Statistically
VICTORIA provides a compact teaching example of several core clinical-trial statistical concepts. The same trial record moves from a randomized treatment comparison to a composite time-to-event endpoint, a formal stratified log-rank test, a Cox hazard ratio, recurrent-event modeling, and score-based confidence intervals for binary outcomes.
| Concept | How it appears in VICTORIA |
|---|---|
| Randomization | The trial uses randomized allocation with two parallel arms. |
| Blinding | The registry identifies the study as double-masked. |
| Time-to-event endpoint | The primary endpoint measures time to first CV death or HF hospitalization. |
| Stratified log-rank test | The registered primary endpoint definition specifies a one-sided stratified log-rank test. |
| Cox model | The posted primary analysis reports a Cox proportional-hazards model and HR 0.90. |
| Hazard ratio | The primary HR is 0.90 with a two-sided 95% CI of 0.82–0.98. |
| Confidence interval | The 95% CI communicates uncertainty around the estimated HR. |
| P-value | The primary reported p-value is 0.019. |
| Composite endpoint | CV death and HF hospitalization form the registered primary composite endpoint. |
| Recurrent-event analysis | Total HF hospitalizations are analyzed using an Andersen-Gill model. |
| Risk difference | Symptomatic hypotension and syncope use differences in percentages. |
| Miettinen & Nurminen method | The binary endpoints use this score-based method for the reported percentage differences. |
| Safety analysis | Serious adverse events are reported by treatment arm with affected and at-risk counts. |
21. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
22. Related Statistical Calculators
23. Sources
- ClinicalTrials.gov: VICTORIA, NCT02861534.
- PubMed: PMID 31820546.
- PubMed: PMID 41058565.
- PubMed: PMID 38934967.
- PubMed: PMID 38363272.
- PubMed: PMID 37671551.
Continue through the Clinical Biostats statistical library
Explore the statistical methods behind randomized trials, survival analysis, confidence intervals, and clinical-trial endpoints.
24. Record Summary
VICTORIA is a useful statistical case study because the registry combines a randomized, double-masked, parallel-group phase 3 design with a primary composite time-to-event endpoint. The primary result is reported as a Cox-model hazard ratio of 0.90 with a two-sided 95% CI of 0.82–0.98 and P = 0.019, while the registered endpoint definition identifies a one-sided stratified log-rank test as the primary endpoint analysis. The same record illustrates why different endpoint structures require different statistical tools: Cox models for several time-to-event outcomes, an Andersen-Gill model for recurrent HF hospitalizations, and the Miettinen & Nurminen method for binary percentage differences.