This page separates reported trial results from statistical interpretation. Numerical trial results on this page are restricted to the ClinicalTrials.gov record.
Trial at a Glance
AIM-HIGH was a randomized, parallel-group, quadruple-masked phase 3 prevention trial comparing extended-release niacin plus simvastatin with placebo plus simvastatin. The primary registered endpoint was a composite time-to-event outcome measured from randomization through the last follow-up visit.
| Feature | AIM-HIGH |
|---|---|
| Phase | Phase 3 |
| Therapeutic area | Cardiology |
| Conditions | Cardiovascular Diseases; Heart Diseases; Cerebrovascular Accident; Coronary Disease; Atherosclerosis; Myocardial Infarction |
| Design | Randomized, parallel-group |
| Masking | Quadruple |
| Primary purpose | Prevention |
| Enrollment | 3414 |
| Interventions | Extended release niacin; Simvastatin |
| Primary endpoint | Composite End Point of CHD Death, Nonfatal MI, Ischemic Stroke, Hospitalization for Non-ST Segment Elevation Acute Coronary Syndrome (ACS), or Symptom-driven Coronary or Cerebral Revascularization |
| Primary analysis | Intention-to-treat Cox proportional-hazards model |
| Lead sponsor | Axio Research. LLC |
| Sponsor type | Industry |
| Status | Terminated |
Clinical Question
The central statistical question was whether participants assigned to extended-release niacin plus simvastatin had a different time to first occurrence of the registered composite vascular endpoint than participants assigned to placebo plus simvastatin.
Population
The registered conditions were cardiovascular diseases, heart diseases, cerebrovascular accident, coronary disease, atherosclerosis, and myocardial infarction.
Intervention
Extended-release niacin plus simvastatin.
Comparator
Placebo plus simvastatin.
Primary question
Does assignment to extended-release niacin plus simvastatin alter time to the first event in the registered composite endpoint compared with placebo plus simvastatin?
Trial Design
ERN + Simvastatin
- Extended-release niacin
- Simvastatin
Placebo + Simvastatin
- Placebo
- Simvastatin
Endpoints
The registry contains one registered primary endpoint and three additional posted statistical analyses for secondary endpoints. All four are time-to-event outcomes analyzed using Cox regression, with hazard ratio as the reported effect measure.
| Role | Endpoint | Time frame | Analysis |
|---|---|---|---|
| Primary | Composite End Point of CHD Death, Nonfatal MI, Ischemic Stroke, Hospitalization for Non-ST Segment Elevation Acute Coronary Syndrome (ACS), or Symptom-driven Coronary or Cerebral Revascularization | Time to first event measured from date of randomization through last follow-up visit (common termination) for an average of 36 months follow-up, maximum 66 months | Intention-to-treat Cox proportional-hazards model |
| Secondary | Composite Endpoint of CHD Death, Non-fatal MI, High-risk ACS or Ischemic Stroke | Time to first event measured from date of randomization through last follow-up visit (common termination) for an average of 36 months follow-up, maximum 66 months | Cox proportional-hazards model |
| Secondary | Composite Endpoint of CHD Death, Non-fatal MI, or Ischemic Stroke | Time to first event measured from date of randomization through last follow-up visit (common termination) for an average of 36 months follow-up, maximum 66 months | Intention-to-treat Cox proportional-hazards model |
| Secondary | Cardiovascular Mortality | Time to first event measured from date of randomization through last follow-up visit (common termination), for an average of 36 months follow-up, maximum 66 months. | Intention-to-treat Cox proportional-hazards model |
Statistical Methodology
Primary analysis population
The primary endpoint was analyzed using the intention-to-treat population. In an ITT analysis, randomized participants remain associated with their assigned treatment group for the principal comparison. This preserves the treatment contrast created by randomization and avoids redefining the primary comparison around treatment exposure after randomization.
Cox proportional-hazards model
The registry reports a Cox regression analysis for the primary endpoint. The model estimates a hazard ratio comparing the instantaneous event rate between the randomized groups over the observed time-to-event follow-up.
The coefficient associated with treatment is transformed into a hazard ratio. A hazard ratio below one indicates a lower estimated instantaneous event rate in the treatment group; a hazard ratio above one indicates a higher estimated instantaneous event rate.
Covariate adjustment
The primary analysis notes that the Cox model was adjusted for gender and history of diabetes, which were the randomization stratification factors identified in the registry analysis.
This is an important distinction from an unadjusted comparison. The reported hazard ratio is not simply a ratio obtained by counting events in the two groups. It is the estimate from a regression model that incorporates the specified stratification factors.
Stratified analysis
The registry identifies stratified analysis among the additional concepts associated with the primary statistical analysis. Stratification allows the treatment comparison to account for the prespecified randomization factors without requiring the analysis to assume that the baseline hazard is identical across those strata.
Hazard ratio and confidence interval
The primary effect measure was the hazard ratio, reported with a two-sided confidence interval. The point estimate and confidence interval should be interpreted together rather than treating the point estimate as an exact description of the treatment effect.
The point estimate is close to one. The confidence interval spans one, indicating that the registry's reported estimate is compatible with both a lower and a higher hazard under the model.
Statistical Methods Explained
Why was a Cox proportional-hazards model used?
The registered endpoints are time-to-event outcomes. A Cox model is designed for this setting because it uses both the occurrence of events and the timing of those events while accommodating right censoring. It produces a hazard ratio rather than requiring every participant to have experienced an event by the end of follow-up.
What does a hazard ratio of 1.02 mean?
A hazard ratio of 1.02 means that the estimated instantaneous event rate for the ERN plus simvastatin group was 1.02 times the corresponding rate in the placebo plus simvastatin group under the fitted Cox model. Because the estimate is close to one, the reported point estimate represents little difference in the modeled hazard between groups.
This does not mean that every participant had a 1.02-fold individual risk, nor does it mean that the cumulative probability of an event was exactly 1.02 times higher. A hazard ratio is a relative time-to-event measure, not an individual-level probability.
Why does the confidence interval matter?
The 95% confidence interval of 0.87–1.21 communicates uncertainty around the estimated hazard ratio. It gives a range of parameter values compatible with the statistical model and data under the stated confidence framework. It is not a range containing the treatment effect for individual participants.
Why is the p-value not an effect size?
The primary p-value was 0.80. A p-value evaluates the compatibility of the observed data with the null hypothesis used for the test; it does not quantify how large or clinically important an effect is. Effect size is conveyed here by the hazard ratio, while precision is conveyed by its confidence interval.
Why use intention-to-treat analysis?
The primary analysis population was intention-to-treat. An ITT comparison keeps participants grouped according to randomized assignment. This maintains the causal comparison generated by randomization and prevents the primary efficacy analysis from being defined by post-randomization treatment behavior.
What does censoring mean in these analyses?
In a time-to-event analysis, censoring occurs when the exact event time is not observed during the period for which a participant contributes follow-up. The participant can still contribute information up to the censoring time. Cox regression is designed to incorporate this type of incomplete event-time information under its statistical assumptions.
What does the proportional-hazards assumption mean?
The Cox model is commonly interpreted through a hazard ratio that represents a relative comparison of hazards over time. The proportional-hazards assumption concerns whether that relative hazard is adequately represented as stable over the modeled follow-up. If the underlying hazards change substantially relative to one another, a single hazard ratio may provide an incomplete summary of the time-varying treatment difference.
Primary Result
Composite vascular endpoint
The primary endpoint was the time to first event of CHD death, nonfatal MI, ischemic stroke, hospitalization for non-ST segment elevation acute coronary syndrome (ACS), or symptom-driven coronary or cerebral revascularization.
Hazard ratio for the primary endpoint
95% CI: 0.87–1.21 · P = 0.80
ERN + Simvastatin vs Placebo + Simvastatin
| Feature | Reported result |
|---|---|
| Analysis population | Intention-to-treat |
| Model | Regression, Cox |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 1.02 |
| 95% CI | 0.87–1.21 |
| CI sides | Two-sided |
| P-value | 0.80 |
| Hypothesis type | Superiority |
| Adjustment | Gender and history of diabetes, the randomization stratification factors |
The primary hazard ratio of 1.02 indicates that the fitted Cox model estimated an instantaneous event rate very close to that of the placebo plus simvastatin group. Because the hazard ratio is above one, the point estimate itself is slightly above the null value, but the magnitude is small.
The estimate does not mean that participants receiving ERN plus simvastatin had exactly 2% greater individual risk. It is a model-based relative comparison of event hazards over the analyzed follow-up.
The 95% CI of 0.87–1.21 is important because it includes the null value of one. The interval indicates that the observed estimate is not precise enough to exclude a range extending below and above the null under the reported statistical framework.
The p-value of 0.80 is a hypothesis-test result, not a measure of the size of the treatment effect. It should not be interpreted as an 80% probability that the null hypothesis is true, nor as a probability that the treatment has no effect.
The analysis is also subject to the usual interpretive considerations for a Cox model, including censoring and the proportional-hazards assumption. The result is an ITT analysis adjusted for the registered randomization stratification factors of gender and history of diabetes.
Secondary Endpoint Results
Composite Endpoint of CHD Death, Non-fatal MI, High-risk ACS or Ischemic Stroke
Hazard ratio
95% CI: 0.87–1.34 · P = .49
ERN + Simvastatin vs Placebo + Simvastatin
This secondary time-to-event analysis used Cox regression and reported a hazard ratio of 1.08. The 95% confidence interval of 0.87–1.34 spans the null value, and the reported p-value was .49.
Composite Endpoint of CHD Death, Non-fatal MI, or Ischemic Stroke
Hazard ratio
95% CI: 0.90–1.42 · P = .30
ERN + Simvastatin vs Placebo + Simvastatin
This secondary endpoint was analyzed in the intention-to-treat population using Cox regression. The reported hazard ratio was 1.13, with a 95% confidence interval of 0.90–1.42 and a p-value of .30.
Cardiovascular Mortality
Hazard ratio
95% CI: 0.76–1.80 · P = 0.47
ERN + Simvastatin vs Placebo + Simvastatin
Cardiovascular mortality was also analyzed as a time-to-event endpoint using intention-to-treat Cox regression. The reported hazard ratio was 1.17, with a 95% confidence interval of 0.76–1.80 and a p-value of 0.47.
| Secondary endpoint | HR | 95% CI | P-value |
|---|---|---|---|
| CHD death, non-fatal MI, high-risk ACS or ischemic stroke | 1.08 | 0.87–1.34 | .49 |
| CHD death, non-fatal MI, or ischemic stroke | 1.13 | 0.90–1.42 | .30 |
| Cardiovascular mortality | 1.17 | 0.76–1.80 | 0.47 |
Interpreting the Pattern of Hazard Ratios
The four posted analyses produce a coherent statistical pattern in the sense that the point estimates are 1.02 for the primary endpoint and 1.08, 1.13, and 1.17 for the secondary endpoints. However, the point estimates alone are insufficient for judging the precision of the comparisons.
| Endpoint | Point estimate | Confidence interval | What the interval communicates |
|---|---|---|---|
| Primary composite | 1.02 | 0.87–1.21 | Compatible with values below and above the null |
| CHD death, non-fatal MI, high-risk ACS or ischemic stroke | 1.08 | 0.87–1.34 | Compatible with values below and above the null |
| CHD death, non-fatal MI, or ischemic stroke | 1.13 | 0.90–1.42 | Compatible with values below and above the null |
| Cardiovascular mortality | 1.17 | 0.76–1.80 | Compatible with a comparatively broad range around the null |
The cardiovascular mortality estimate illustrates why precision matters. A point estimate of 1.17 is not sufficient by itself to characterize the evidence because its confidence interval extends from 0.76 to 1.80. The interval is substantially broader than the point estimate and includes both values below and above one.
Safety
The ClinicalTrials.gov record reports serious adverse events by treatment arm. The affected and at-risk counts were 587/1718 for ERN plus simvastatin and 551/1696 for placebo plus simvastatin.
| Safety measure | ERN + Simvastatin | Placebo + Simvastatin |
|---|---|---|
| Serious adverse events, affected / at risk | 587/1718 | 551/1696 |
What is reported
The ClinicalTrials.gov record identifies the number affected and the number at risk in each randomized arm for serious adverse events.
What is not inferred
No additional adverse-event categories, event percentages, relative risks, or statistical tests are added here because they are not contained in the ClinicalTrials.gov record.
Safety and efficacy answer different questions. The serious-adverse-event counts describe an exposure-related safety outcome, whereas the primary statistical analysis evaluates time to the first event in a specific vascular composite endpoint.
Analysis of the Primary Endpoint
Why the primary endpoint is a composite
The registered primary endpoint combines several clinically distinct events: CHD death, nonfatal MI, ischemic stroke, hospitalization for non-ST segment elevation ACS, and symptom-driven coronary or cerebral revascularization. The statistical analysis treats the first occurrence of any component as the event of interest.
A composite endpoint can increase the number of observed events and therefore provide more information for a time-to-event comparison. But its interpretation depends on the component events being clinically meaningful and on understanding which components contribute to the overall result. The ClinicalTrials.gov record does not provide component-specific estimates, so no component-level conclusion is made on this page.
Why time to first event matters
The endpoint is not simply whether a participant ever experienced an event. It is explicitly defined as time to first event from randomization through the last follow-up visit. The timing of an event therefore contributes to the statistical analysis.
This also means that participants who remain event-free for different lengths of follow-up can contribute different amounts of information. Cox regression is appropriate for this structure because it can use observed follow-up while accounting for right censoring.
Primary Hypothesis and Statistical Evidence
The registry identifies the hypothesis type as superiority and the reported confidence interval as two-sided. The exact hypothesis notation above is a conceptual representation of the null comparison underlying a two-sided hazard-ratio analysis; the registry's reported hypothesis type remains the authoritative description.
The primary estimate was 1.02, the 95% confidence interval was 0.87–1.21, and the p-value was 0.80. The confidence interval includes the null hazard ratio of one, and the p-value does not provide evidence against the null at conventional significance thresholds.
That statement should not be translated into a claim that the treatment has been proven to have exactly no effect. Statistical non-significance means that the observed data do not provide sufficient evidence for the specified superiority hypothesis under the reported analysis. The confidence interval is essential for understanding what effect sizes remain compatible with the analysis.
Covariate Adjustment and Stratification
The primary analysis adjusted for gender and history of diabetes, identified in the registry as randomization stratification factors. This means the reported hazard ratio incorporates these variables rather than representing a purely unadjusted treatment-group comparison.
Why adjust?
Adjustment can account for prespecified design factors and can make the treatment estimate reflect the randomized comparison conditional on those factors.
What adjustment does not do
Covariate adjustment does not transform a non-significant result into a positive one, and it does not eliminate uncertainty represented by the confidence interval.
Because gender and history of diabetes were used as randomization stratification factors, their presence in the analysis is connected directly to the trial design rather than being an arbitrary post hoc selection of predictors.
Blinding and Randomization
The trial was randomized and quadruple-masked. Randomization is central to the causal interpretation of the comparison because treatment assignment is established before outcomes are observed. Masking can reduce the influence of knowledge of treatment assignment on participant, investigator, care, or assessment behavior.
The statistical model does not create randomization; it analyzes the comparison generated by the randomized design. That distinction is important: statistical adjustment is not a substitute for randomized treatment allocation.
Limitations
- Registry-level scope: the ClinicalTrials.gov record contains the primary and three secondary statistical analyses but do not provide baseline characteristics, component-specific event counts, subgroup estimates, median event times, or Kaplan-Meier estimates. Those quantities are therefore not added.
- Composite endpoint: the primary endpoint contains several different clinical events. The registry analysis reported here reports the composite result, not separate treatment-effect estimates for each component.
- Hazard-ratio interpretation: a single Cox hazard ratio is a model-based summary and should not be treated as an individual patient's risk ratio or as an absolute risk difference.
- Proportional hazards: interpretation of a Cox hazard ratio depends on the suitability of the model for the observed time-to-event process, including the proportional-hazards assumption.
- Censoring: time-to-event analysis relies on information contributed before censoring and therefore depends on assumptions concerning the censoring process.
- Secondary endpoints: multiple secondary analyses create a multiplicity consideration. The ClinicalTrials.gov record identifies the endpoints and their p-values but do not provide an alpha-allocation or multiplicity-adjustment strategy, so no such procedure is inferred.
- Missing data and imputation: the ClinicalTrials.gov record does not specify a missing-data or imputation procedure. None is assumed.
- Interim analysis: the ClinicalTrials.gov record does not report an interim-analysis strategy or alpha-spending procedure. No such design feature is inferred.
- Crossover: the ClinicalTrials.gov record does not report a crossover strategy. No crossover effect is therefore discussed.
- Bayesian methods: the analyses posted on ClinicalTrials.gov are frequentist Cox regression analyses. No Bayesian method is reported.
Why This Trial Matters Statistically
AIM-HIGH is a useful teaching case because its registry record illustrates how a randomized prevention trial can be represented almost entirely through time-to-event methodology. The primary question is expressed through a composite first-event endpoint, analyzed in an intention-to-treat population using a Cox proportional-hazards model with adjustment for randomization stratification factors.
| Statistical concept | How it appears in AIM-HIGH |
|---|---|
| Randomization | The allocation is randomized. |
| Parallel-group design | The trial has two parallel treatment arms. |
| Blinding | The registry describes the trial as quadruple-masked. |
| Intention-to-treat | The primary analysis uses the intention-to-treat population. |
| Time-to-event endpoint | The primary endpoint measures time to first event from randomization through follow-up. |
| Composite endpoint | The primary outcome combines CHD death, nonfatal MI, ischemic stroke, hospitalization for non-ST segment elevation ACS, or symptom-driven coronary or cerebral revascularization. |
| Cox regression | The primary and secondary statistical analyses use Cox regression. |
| Hazard ratio | Hazard ratio is the reported effect measure for all posted analyses. |
| Covariate adjustment | The primary model adjusts for gender and history of diabetes. |
| Stratified analysis | Stratified analysis is identified among the concepts in the primary analysis record. |
| Confidence intervals | The primary estimate has a two-sided 95% confidence interval. |
| Superiority testing | The registry identifies the hypothesis type as superiority. |
What the Primary Hazard Ratio Does — and Does Not — Mean
The primary HR of 1.02 means that the fitted Cox model estimated the instantaneous event rate in the ERN plus simvastatin group to be 1.02 times that in the placebo plus simvastatin group. It is a relative model-based measure, not an absolute probability.
The 95% CI of 0.87–1.21 expresses uncertainty around the estimated hazard ratio. Because the interval includes one, the data are compatible with both a hazard below the null value and a hazard above it under the reported model.
The p-value of 0.80 measures the statistical evidence against the specified null hypothesis under the testing framework. It does not measure treatment efficacy, clinical importance, or the probability that one treatment is better than the other.
The registry data posted on ClinicalTrials.gov for AIM-HIGH provide hazard ratios and confidence intervals but do not provide absolute cumulative event probabilities or median time-to-event estimates. The hazard ratio should therefore not be converted into an absolute risk difference.
Understanding the Secondary Analyses
The secondary endpoints progressively define narrower clinical composites. The first includes CHD death, non-fatal MI, high-risk ACS or ischemic stroke. The second includes CHD death, non-fatal MI, or ischemic stroke. Cardiovascular mortality isolates a mortality endpoint.
These analyses can be useful for understanding whether the statistical pattern is confined to the broad primary composite or is also reflected in related outcomes. However, they should remain identified as secondary analyses. The ClinicalTrials.gov record does not specify a multiplicity adjustment, so the reported p-values should not be interpreted as though a particular familywise error-control strategy had been established in the ClinicalTrials.gov record.
Sources
- ClinicalTrials.gov: AIM-HIGH, NCT00120289.
- Linked publication record: PubMed record for PMID 22085343.
- Linked publication record: PubMed record for PMID 23881958.
- Linked publication record: PubMed record for PMID 23973688.
- Linked publication record: PubMed record for PMID 23916935.
- Linked publication record: PubMed record for PMID 34134520.
Continue with the statistical methods used in this trial
Explore the underlying concepts behind time-to-event endpoints, Cox regression, hazard ratios, confidence intervals, randomization, and intention-to-treat analysis.
Related Tutorials
Learn more about the methods used in this trial:
Related Calculators
Record Summary
AIM-HIGH provides a focused example of randomized time-to-event analysis. The primary registered endpoint was a composite first-event outcome analyzed in the intention-to-treat population using Cox regression, with adjustment for gender and history of diabetes. The resulting hazard ratio was 1.02, with a two-sided 95% confidence interval of 0.87–1.21 and a p-value of 0.80.
The three posted secondary analyses used the same broad time-to-event framework and produced hazard ratios of 1.08, 1.13, and 1.17, respectively. Their confidence intervals were 0.87–1.34, 0.90–1.42, and 0.76–1.80. The registry also reports serious adverse-event counts of 587/1718 for ERN plus simvastatin and 551/1696 for placebo plus simvastatin.
The principal statistical lesson is that a clinical-trial result should be read through several linked quantities: the endpoint definition, analysis population, model, effect measure, confidence interval, and hypothesis-test result. In AIM-HIGH, the hazard ratio describes a relative time-to-event comparison, while the confidence interval describes its statistical precision and the p-value addresses the specified superiority test. None of these quantities, considered alone, describes individual patient benefit or absolute event probability.