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
ACCELERATE was a randomized, parallel-group, quadruple-masked phase 3 study of evacetrapib versus placebo in cardiovascular diseases. The registry reports 12,092 enrolled participants, two arms, a binary registered primary endpoint, and statistical analyses based on both Cox proportional-hazards models and ANOVA.
| Feature | ACCELERATE |
|---|---|
| Trial name | ACCELERATE |
| Brief title | A Study of Evacetrapib in High-Risk Vascular Disease |
| Phase | Phase 3 |
| Status | Terminated |
| Therapeutic area | Cardiovascular |
| Condition | Cardiovascular Diseases |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 12,092 |
| Interventions | Evacetrapib (drug); Placebo (drug) |
| Lead sponsor | Eli Lilly and Company |
| Sponsor type | Industry |
| Start | 2012-10 |
| Primary completion | 2016-03 |
2. Clinical Question
The central statistical question was whether participants randomized to evacetrapib differed from those randomized to placebo in the time to first occurrence of the registered composite primary endpoint of cardiovascular death, myocardial infarction, stroke, coronary revascularization, or hospitalization for unstable angina.
Population
Participants enrolled in the phase 3 study of evacetrapib in high-risk vascular disease, with the registry condition listed as cardiovascular diseases.
Intervention
Evacetrapib, classified in the registry as a drug intervention.
Comparator
Placebo, classified in the registry as a drug intervention.
Primary question
Does evacetrapib change the time to first occurrence of the composite cardiovascular endpoint compared with placebo?
3. Trial Design
Evacetrapib
- Drug intervention
- Compared with placebo
- Evaluated within a randomized parallel-group design
Placebo
- Drug intervention classified as placebo
- Comparator for evacetrapib
- Evaluated within the same randomized parallel-group design
4. Endpoints
| Endpoint | Time frame | Registry / analysis type |
|---|---|---|
| Number of Participants With Composite Primary Endpoint of Cardiovascular (CV) Death, Myocardial Infarction (MI), Stroke, Coronary Revascularization, or Hospitalization for Unstable Angina (UA) | Baseline to Study Completion (Up to 4 years) | Registered primary endpoint; formal time-to-event analysis posted |
| Mean Percent Change From Baseline to 3 Months in Low-Density (LDL-C) and High-Density Lipoprotein Cholesterol (HDL-C) Levels | Baseline, 3 Months | Secondary; ANOVA |
| Number of Participants With Composite Endpoint of All-Cause Mortality, MI, Stroke, Coronary Revascularization, or Hospitalization for UA | Baseline through End of Study (Up to 4 years) | Secondary; Cox proportional-hazards model |
| Number of Participants With Composite Endpoint of CV Death, MI, or Coronary Revascularization | Baseline through End of Study (Up to 4 years) | Secondary; Cox proportional-hazards model |
| Number of Participants With Composite Endpoint of CV Death, MI, Stroke, or Hospitalization for UA | Baseline through End of Study (Up to 4 years) | Secondary; Cox proportional-hazards model |
| Number of Participants With Triple Composite Endpoint of CV Death, MI, or Stroke | Baseline through End of Study (Up to 4 years) | Secondary; Cox proportional-hazards model |
5. Statistical Methodology
Primary time-to-event analysis
The registry reports a Cox proportional-hazards model for the primary endpoint, using all randomized participants and comparing evacetrapib with placebo. The effect measure is the hazard ratio, with a two-sided 95% confidence interval and a superiority hypothesis.
In a Cox model, the treatment coefficient is transformed into a hazard ratio. The model compares the instantaneous event rates associated with the treatment groups over follow-up rather than simply comparing the final proportions of participants who experienced an event.
ANOVA for lipid outcomes
The registry reports ANOVA for the secondary endpoint measuring mean percent change from baseline to 3 months in LDL-C and HDL-C levels. The reported effect measure is the mean difference in final values between evacetrapib and placebo.
For a continuous endpoint, a mean difference directly describes the separation between group means on the analyzed scale. Unlike a hazard ratio, it is not a relative time-to-event measure. The registry's lipid analyses therefore illustrate why statistical method selection follows the structure of the endpoint.
Analysis populations
The primary and composite cardiovascular endpoint analyses use all randomized participants. The LDL-C and HDL-C analysis uses all randomized participants with evaluable LDL-C and HDL-C levels. This distinction matters because the analysis population defines which observations contribute to each estimate.
Superiority hypothesis
The registry classifies the primary and posted secondary analyses under a superiority hypothesis. In a superiority analysis, the treatment effect is tested against the null hypothesis of no difference rather than against a prespecified non-inferiority margin.
What the registry reports about other design features
The ClinicalTrials.gov record identifies randomization, parallel allocation, quadruple masking, Cox proportional-hazards modeling, ANOVA, hazard ratios, mean differences, and superiority hypotheses. The ClinicalTrials.gov record does not report a non-inferiority margin, crossover scheme, factorial structure, Bayesian method, interim-analysis boundary, missing-data imputation method, or stratification factors. Those features are therefore not used here to characterize the trial.
6. Primary Results
Composite cardiovascular primary endpoint
The registered primary endpoint was the number of participants with a composite of cardiovascular death, myocardial infarction, stroke, coronary revascularization, or hospitalization for unstable angina, measured from baseline to study completion, up to 4 years. The formal analysis was a Cox proportional-hazards model in all randomized participants.
Hazard ratio: evacetrapib vs placebo
95% CI: 0.911–1.111 · P = 0.9054
Two-sided 95% confidence interval · Superiority hypothesis
| Primary endpoint | Analysis population | Method | Effect estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Time to first occurrence of CV death, MI, stroke, coronary revascularization, or hospitalization for UA | All randomized participants | Cox proportional-hazards model | HR 1.006 | 0.911–1.111 | 0.9054 |
The estimated hazard ratio of 1.006 means that the fitted model estimated the instantaneous rate of the composite event in the evacetrapib group to be approximately 1.006 times that in the placebo group over the analyzed follow-up. Because the estimate is very close to 1, the estimated relative difference between the groups is correspondingly small.
The hazard ratio does not mean that 1.006% of participants experienced the endpoint, nor does it represent a difference in cumulative incidence. It is a model-based relative measure of event rates over time.
The 95% confidence interval, 0.911–1.111, describes statistical uncertainty around the estimated hazard ratio under the analysis model. It spans 1, so the data are compatible with a range of relative hazard differences in either direction rather than establishing a precise directional effect.
The p-value of 0.9054 addresses the statistical evidence against the relevant null hypothesis; it does not measure the size, clinical importance, or probability of the treatment effect. A p-value should therefore be interpreted alongside the hazard ratio and its confidence interval.
Because the analysis is based on a Cox proportional-hazards model, interpretation of a single hazard ratio also depends on the model's proportional-hazards framework. The registry result does not provide enough information here to independently assess that assumption from event-time data.
7. Secondary Endpoint Results
LDL-C and HDL-C: 3-month mean percent change
The registry reports an ANOVA analysis of mean percent change from baseline to 3 months in LDL-C and HDL-C levels. The analysis population consisted of all randomized participants with evaluable LDL-C and HDL-C levels.
| Measure | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| LDL-C | ANOVA | Mean Difference (Final Values) | -37.11 | -38.15 to -36.08 | <0.0001 |
| HDL-C | ANOVA | Mean Difference (Final Values) | 131.55 | 130.01 to 133.09 | <0.0001 |
The LDL-C estimate of -37.11 is a mean difference on the registry's reported analysis scale, with the evacetrapib-versus-placebo direction producing a negative estimate. The corresponding 95% confidence interval is -38.15 to -36.08.
The HDL-C estimate of 131.55 is the corresponding mean difference for HDL-C, with a 95% confidence interval of 130.01 to 133.09.
Both p-values are reported as <0.0001. These p-values provide evidence against the relevant null hypotheses for the analyzed lipid comparisons, but they do not quantify the magnitude of the changes. The estimates and confidence intervals are needed to understand the size and precision of the observed group differences.
These laboratory outcomes also answer a different statistical question from the primary cardiovascular endpoint. A treatment can produce a large difference in a biomarker while the relationship between that biomarker change and a clinical time-to-event endpoint remains a separate question. The registry results should therefore not be treated as interchangeable evidence.
Composite of all-cause mortality, MI, stroke, coronary revascularization, or hospitalization for UA
| Endpoint | Method | HR | 95% CI | P-value |
|---|---|---|---|---|
| Time to first occurrence of composite endpoint of all-cause mortality, MI, stroke, coronary revascularization, or hospitalization for UA | Cox proportional-hazards model | 0.991 | 0.901–1.089 | 0.8463 |
| Time to first occurrence of composite endpoint of CV death, MI, or coronary revascularization | Cox proportional-hazards model | 1.001 | 0.901–1.112 | 0.9874 |
| Time to first occurrence of composite endpoint of CV death, MI, stroke, or hospitalization for UA | Cox proportional-hazards model | 1.003 | 0.893–1.127 | 0.9574 |
| Time to first occurrence of triple composite endpoint of CV death, MI, or stroke | Cox proportional-hazards model | 0.965 | 0.846–1.100 | 0.5917 |
These secondary time-to-event estimates are all close to 1, with confidence intervals that include 1. The registry reports the same Cox proportional-hazards framework for each comparison, allowing the estimates to be interpreted on a common hazard-ratio scale while recognizing that each endpoint represents a different composite definition.
8. Secondary Results: Statistical Reading of the Composite Endpoints
HR 0.991
The estimate is slightly below 1, corresponding to an estimated hazard approximately 0.991 times that of placebo for this secondary composite. Its 95% CI spans both sides of 1.
HR 1.001
The estimate is essentially centered on 1. The 95% CI of 0.901–1.112 indicates uncertainty extending in both directions around the null value.
HR 1.003
The estimate is very close to 1. The 95% CI of 0.893–1.127 includes values below and above the null value.
HR 0.965
The estimate is below 1, but its 95% CI of 0.846–1.100 includes 1. The reported p-value is 0.5917.
The important statistical point is that the direction of a point estimate alone is not sufficient to establish evidence of a treatment difference. Confidence intervals show how precisely the treatment effect has been estimated, while the p-value summarizes evidence against the relevant null hypothesis. The three quantities should be considered together.
9. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
The primary and several secondary cardiovascular endpoints are time-to-event outcomes. Participants can experience the event at different times, and some observations can be censored before the event occurs. A Cox model is designed for this structure because it models relative event rates over follow-up rather than collapsing all follow-up information into a single end-of-study proportion.
What does a hazard ratio of 1.006 mean?
A hazard ratio of 1.006 means the estimated instantaneous event rate in the evacetrapib group was approximately 1.006 times that in the placebo group under the fitted model. It does not mean that 1.006% of participants experienced the event, and it is not an absolute risk difference.
Why is 1 the important reference value for a hazard ratio?
For a ratio measure, 1 represents equality between the two groups. Values below 1 indicate a lower estimated hazard in the numerator group, while values above 1 indicate a higher estimated hazard. The primary estimate of 1.006 is therefore very close to the null value.
What does the confidence interval add?
The confidence interval communicates precision. A narrow interval indicates a more precisely estimated hazard ratio than a wide interval, all else equal. For the primary endpoint, the interval from 0.911 to 1.111 indicates that the point estimate should not be interpreted in isolation.
Why does the p-value not measure effect size?
A p-value summarizes the evidence against a null hypothesis under the statistical model. It is not a measure of how large or clinically important an effect is. Effect size is described by the hazard ratio or mean difference, while uncertainty is described by the confidence interval.
Why was ANOVA used for the lipid endpoint?
The lipid endpoint is a continuous measurement summarized as mean percent change, rather than a time-to-event outcome. ANOVA provides a linear-model framework for comparing mean outcomes between groups. This differs fundamentally from the Cox model used for cardiovascular event times.
Why should the composite endpoints be interpreted separately?
Each composite combines a different set of clinical events. Even when two composites share several components, changing the definition can alter which events contribute to the endpoint. A hazard ratio for one composite therefore should not automatically be treated as the hazard ratio for another.
10. Confidence Intervals and Effect Size
The point estimate is close to the null value of 1, and the interval extends below and above 1. The interval therefore provides important information that is not contained in the point estimate alone.
A useful discipline in clinical-trial interpretation is to separate three questions:
- What is the estimated effect? For the primary endpoint, the reported hazard ratio is 1.006.
- How precise is that estimate? The reported two-sided 95% confidence interval is 0.911–1.111.
- How strong is the statistical evidence against the null? The reported p-value is 0.9054.
None of these quantities by itself answers every clinical question. In particular, a hazard ratio does not provide the absolute probability of experiencing an event, and a p-value does not tell the reader how large an effect would be important in practice.
11. Time-to-Event Endpoints and Censoring
The primary endpoint is defined over baseline to study completion, up to 4 years, and the posted statistical analysis is explicitly classified as time-to-event. That means the analysis incorporates when the first event occurred, rather than only whether an event was eventually observed.
Kaplan-Meier estimation uses the number of participants at risk and the number of events at each event time to estimate the probability of remaining event-free over time.
The ClinicalTrials.gov record does not post a Kaplan-Meier estimate, median event time, event counts by treatment arm for the primary endpoint, or the individual event and censoring times. Those quantities are therefore not added here.
The distinction between a binary endpoint label in the registered endpoint description and a time-to-event analysis is also instructive. The registry identifies the primary endpoint as binary in its endpoint classification, while the posted formal analysis evaluates time to first occurrence using a Cox model. The analysis method therefore contains additional timing information beyond a simple yes/no comparison at one fixed time.
12. Analysis Populations
| Analysis | Population | Why it matters |
|---|---|---|
| Primary composite cardiovascular endpoint | All randomized participants | Anchors the efficacy comparison to randomized treatment assignment. |
| Secondary cardiovascular composites | All randomized participants | Maintains the randomized comparison for the posted time-to-event analyses. |
| LDL-C / HDL-C analysis | All randomized participants with evaluable LDL-C and HDL-C levels | Limits the continuous-outcome analysis to participants with evaluable measurements. |
Analysis-population definitions are not a minor technical detail. A treatment effect can differ depending on which participants contribute data. For this reason, an estimate should always be read together with the population to which it applies.
13. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected participants divided by participants at risk.
| Safety measure | Evacetrapib | Placebo |
|---|---|---|
| Serious adverse events | 2,306 / 6,036 | 2,341 / 6,052 |
These figures describe the reported number of participants affected by serious adverse events and the corresponding number at risk in each arm. They should not be substituted for an efficacy endpoint, because adverse-event outcomes and cardiovascular efficacy outcomes address different questions.
The reported serious-adverse-event data should be interpreted using both the affected count and the corresponding at-risk denominator. Comparing counts alone can be misleading when denominators differ. The ClinicalTrials.gov record supports reporting the arm-specific affected/at-risk values but do not provide a formal statistical comparison of serious adverse events.
14. What the Primary Hazard Ratio Does — and Does Not — Mean
The primary HR of 1.006 is the estimated relative hazard for the composite endpoint comparing evacetrapib with placebo under the Cox proportional-hazards model.
It is not a percentage of participants, not an absolute event probability, not a mean difference, and not the probability that evacetrapib is effective or ineffective. It also does not imply that every participant experiences the same relative change in event risk.
The 95% CI of 0.911–1.111 expresses uncertainty around the estimated hazard ratio. Because it includes 1, the reported interval is compatible with both a lower and a higher hazard relative to placebo.
The p-value of 0.9054 summarizes evidence against the null hypothesis used for the superiority analysis. It should not be interpreted as an effect-size metric or as the probability that the null hypothesis is true.
15. Secondary Endpoint Interpretation
The secondary results illustrate two different statistical scales. The lipid outcomes use mean differences, while the cardiovascular composite outcomes use hazard ratios.
Mean difference
A mean difference describes separation between two group means on the analyzed outcome scale. The LDL-C and HDL-C results are expressed this way.
Hazard ratio
A hazard ratio compares modeled event rates over time. The cardiovascular composite outcomes are expressed this way.
Confidence interval
The interval supplies information about precision around the estimated effect and should accompany the point estimate.
P-value
The p-value addresses statistical evidence against a null hypothesis. It does not replace the effect estimate or confidence interval.
This distinction is especially important when several endpoints are presented on the same page. A value of -37.11 for a mean difference and a value of 0.991 for a hazard ratio are not directly comparable numerical quantities. Their interpretation depends on their respective statistical scales.
16. Multiple Endpoints and Statistical Interpretation
The registry reports one primary endpoint and five secondary outcome analyses, with multiple posted statistical comparisons. The ClinicalTrials.gov record does not specify an adjustment procedure for the secondary endpoints.
| Endpoint role | Statistical method | Effect measure | Interpretive focus |
|---|---|---|---|
| Primary cardiovascular composite | Cox proportional-hazards model | Hazard ratio | Primary superiority comparison |
| LDL-C | ANOVA | Mean difference | Continuous lipid outcome |
| HDL-C | ANOVA | Mean difference | Continuous lipid outcome |
| All-cause mortality / MI / stroke / revascularization / UA | Cox proportional-hazards model | Hazard ratio | Secondary time-to-event composite |
| CV death / MI / revascularization | Cox proportional-hazards model | Hazard ratio | Secondary time-to-event composite |
| CV death / MI / stroke / UA | Cox proportional-hazards model | Hazard ratio | Secondary time-to-event composite |
| CV death / MI / stroke | Cox proportional-hazards model | Hazard ratio | Secondary triple composite |
When multiple endpoints are analyzed, the chance of observing at least one small p-value can increase if each test is considered independently. The correct interpretation therefore depends on the prespecified testing hierarchy and multiplicity strategy. Those details are not provided in the ClinicalTrials.gov record, so no additional multiplicity conclusion is assigned here.
17. Trial Timeline
Study start
The registry lists October 2012 as the study start.
Randomized cardiovascular study
The study is classified as phase 3, randomized, parallel, quadruple-masked, and treatment-focused, with evacetrapib and placebo as the two drug interventions.
Primary completion
The registry lists March 2016 as the primary completion date.
Terminated
The ClinicalTrials.gov record classifies the study status as terminated.
18. Statistical Questions Raised by the Design
Randomization
Randomization provides the structural basis for comparing evacetrapib with placebo while preserving treatment assignment as the defining exposure for the primary efficacy analysis.
Masking
Quadruple masking is a design feature intended to reduce the influence of treatment knowledge on trial conduct and assessment.
Time-to-event structure
The cardiovascular endpoints incorporate event timing, so methods designed for censored survival data are more appropriate than a simple comparison of final event proportions.
Endpoint construction
Composite endpoints combine several event types. Interpretation should therefore remain tied to the exact registered definition.
19. Limitations
- Registry-level information: the ClinicalTrials.gov record contains the posted statistical analyses but not the full statistical analysis plan, protocol, or individual participant data.
- No median event times: the ClinicalTrials.gov record does not report median time to the primary or secondary cardiovascular endpoints.
- No event counts by arm for efficacy: the statistical analyses posted on ClinicalTrials.gov provide hazard ratios, confidence intervals, and p-values but do not provide the corresponding cardiovascular event counts by treatment arm.
- No Kaplan-Meier estimates: the ClinicalTrials.gov record does not provide the survival-function estimates or underlying event/censoring times needed to construct a trial-specific Kaplan-Meier curve.
- No stratification information: the ClinicalTrials.gov record does not identify stratification factors used in the Cox analyses.
- No interim-analysis details: the ClinicalTrials.gov record does not identify an interim-analysis schedule, stopping boundary, or alpha-spending method.
- No missing-data method: the ClinicalTrials.gov record does not specify an imputation strategy for missing lipid measurements or other outcomes.
- No non-inferiority margin: the hypothesis type is superiority, and the ClinicalTrials.gov record does not report a non-inferiority framework.
- No crossover information: the ClinicalTrials.gov record does not report crossover or treatment-switching procedures.
- Multiple secondary analyses: several secondary p-values are posted, but the ClinicalTrials.gov record does not specify a multiplicity-adjustment strategy for interpreting them jointly.
- Safety comparison: serious adverse events are reported as affected/at-risk counts by arm, but a formal statistical comparison is not provided in the ClinicalTrials.gov record.
- Generalizability: the registry summary identifies cardiovascular diseases and high-risk vascular disease but does not provide the baseline clinical characteristics needed to evaluate applicability to particular patient subgroups.
20. Why This Trial Matters Statistically
ACCELERATE is a useful teaching case because it places several fundamental statistical ideas in the same randomized trial. The primary endpoint is a composite time-to-event outcome, while an important secondary endpoint uses continuous laboratory measurements. Consequently, the registry demonstrates why endpoint structure determines the statistical method.
| Concept | How it appears in ACCELERATE |
|---|---|
| Randomization | The study is randomized with two parallel intervention groups. |
| Blinding | The registry classifies masking as quadruple. |
| Time-to-event analysis | The primary and cardiovascular secondary composites are analyzed as time-to-event outcomes. |
| Cox proportional-hazards model | Used for the primary and cardiovascular secondary analyses. |
| Hazard ratio | Used as the effect measure for cardiovascular time-to-event endpoints. |
| Confidence interval | Two-sided 95% confidence intervals accompany the reported effect estimates. |
| P-value | Reported for the primary and secondary analyses. |
| ANOVA | Used for the LDL-C and HDL-C continuous outcomes. |
| Mean difference | Used as the effect measure for the 3-month lipid analysis. |
| Composite endpoint | The primary endpoint combines CV death, MI, stroke, coronary revascularization, and hospitalization for UA. |
| Analysis population | Primary efficacy analysis uses all randomized participants. |
The most instructive contrast is between the lipid and cardiovascular findings. The lipid analyses report substantial mean differences with p-values below 0.0001, whereas the primary cardiovascular analysis reports a hazard ratio of 1.006 with a p-value of 0.9054. Statistically, these are not contradictory quantities: they answer different questions on different outcome scales.
This is a core principle of clinical-trial interpretation. A statistically strong effect on a surrogate or laboratory outcome should not automatically be translated into a clinical-event effect. The clinical endpoint must be analyzed and interpreted on its own terms.
21. A Practical Framework for Reading the ACCELERATE Results
The primary endpoint is a composite cardiovascular event outcome assessed over time.
The registry reports a Cox proportional-hazards model in all randomized participants.
The treatment effect is summarized as a hazard ratio.
The primary HR is 1.006, with a two-sided 95% CI of 0.911–1.111.
The reported p-value is 0.9054. It addresses evidence against the superiority null hypothesis rather than the magnitude of the effect.
The primary analysis uses all randomized participants.
This framework prevents several common interpretation errors: treating a hazard ratio as a risk percentage, treating a p-value as an effect size, ignoring the confidence interval, or comparing a cardiovascular event endpoint directly with a continuous biomarker endpoint.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Statistical Calculators
24. Sources
- ClinicalTrials.gov: NCT01687998 — ACCELERATE.
- Linked PubMed record: PMID 32639518 — PubMed.
- Linked PubMed record: PMID 31852422 — PubMed.
- Linked PubMed record: PMID 31752637 — PubMed.
- Linked PubMed record: PMID 28514624 — PubMed.
Continue through the Clinical Biostats statistical pathway
Explore the statistical concepts behind randomized trials, time-to-event endpoints, confidence intervals, hypothesis testing, and continuous-outcome analysis.
25. Record Summary
ACCELERATE provides a compact example of how a randomized clinical trial can combine several statistical frameworks. The primary cardiovascular endpoint is a time-to-event composite analyzed with a Cox proportional-hazards model, producing a hazard ratio of 1.006 with a two-sided 95% CI of 0.911–1.111 and a p-value of 0.9054. Secondary lipid outcomes use ANOVA and mean differences, while additional cardiovascular composites use Cox models and hazard ratios.
The central statistical lesson is that an effect estimate must always be interpreted in the context of its endpoint, analysis method, population, confidence interval, and hypothesis framework. The primary hazard ratio is not a percentage of participants and the p-value is not an effect-size measure. Likewise, the strong statistical evidence reported for the LDL-C and HDL-C analyses describes those continuous outcomes rather than automatically establishing a corresponding effect on the clinical cardiovascular composites.