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
ADVANCE was a randomized, double-blind, phase 3 prevention trial evaluating blood pressure lowering with perindopril-indapamide and glucose lowering with a gliclazide MR-based regimen in high-risk patients with type 2 diabetes. The registry identifies the overall design as factorial.
| Feature | ADVANCE |
|---|---|
| Trial acronym | ADVANCE |
| Phase | Phase 3 |
| Condition | Diabetes Mellitus, Type 2 |
| Primary purpose | Prevention |
| Design | Randomized, double-blind, factorial |
| Enrollment | 11,140 |
| Number of arms | 2 |
| Interventions | Perindopril-indapamide; gliclazide MR-based glucose lowering |
| Trial status | Completed |
| Start date | 2001-06 |
| Primary completion date | 2008-03 |
| Lead sponsor | The George Institute |
2. Clinical Question
The central clinical question was whether pharmacologic lowering of blood pressure and glucose could reduce vascular disease outcomes in patients with type 2 diabetes who were considered at high risk.
Population
High-risk patients with type 2 diabetes mellitus.
Interventions
Perindopril-indapamide for blood pressure lowering and gliclazide MR-based glucose lowering.
Comparator
The registry record identifies a factorial design but does not specify the comparator regimens in the record summarized here.
Primary questions
Whether the randomized intervention strategy affected the composite of major cardiovascular events and the composite of new or substantially worsening nephropathy or microvascular eye disease.
3. Trial Design
The registry classifies ADVANCE as a randomized, double-blind factorial trial. Factorial designs are useful when investigators want to study more than one intervention within the same randomized trial structure.
Why a factorial design?
A factorial design can evaluate two intervention dimensions within one randomized experiment. In a conventional two-factor factorial structure, the statistical analysis can estimate the effect associated with each intervention while also providing a framework for assessing whether the effects interact. That interaction question is important: if the effect of one intervention changes depending on the presence of the other, a simple additive interpretation may be inappropriate.
4. Endpoints
The registry lists two primary endpoints. Both are composite endpoints, meaning that each combines multiple clinically distinct outcomes into a single prespecified outcome measure.
| Primary endpoint | Registry time frame | Statistical structure |
|---|---|---|
| Composite of non-fatal stroke, non-fatal myocardial infarction or death from any cardiovascular cause | July 2001 - December 2007 | Time-to-event composite endpoint |
| Composite of new or substantially worsening nephropathy or microvascular eye disease. | July 2001 - December 2007 | Composite microvascular endpoint |
Why composite endpoints require careful interpretation
A composite endpoint counts an event when any qualifying component occurs. This can improve statistical efficiency when the components are clinically related and occur with sufficient frequency. But the interpretation of the composite is not automatically equivalent to the interpretation of every component.
For the cardiovascular composite, for example, non-fatal stroke, non-fatal myocardial infarction, and death from any cardiovascular cause are different clinical outcomes. A statistically meaningful effect on the composite would therefore need to be understood alongside the component-specific event patterns when those results are available.
The same principle applies to the microvascular composite. New or substantially worsening nephropathy and microvascular eye disease represent different manifestations of diabetic microvascular disease. A composite result should not be interpreted as though it were a single homogeneous clinical event.
5. Statistical Methodology
Factorial treatment comparison
The central statistical feature of ADVANCE is its factorial structure. In a factorial trial with two intervention factors, randomization creates a framework in which the effect of one intervention can be compared while accounting for the other intervention factor. The primary statistical question for each factor is its average effect across the levels of the other factor, provided the interaction is sufficiently small to support that interpretation.
The interaction term represents whether the effect associated with one intervention depends on the level of the other intervention.
Time-to-event analysis
The primary cardiovascular endpoint is naturally structured as a time-to-event outcome because patients can experience the first qualifying event at different times and some patients may remain event-free through the end of observation. A standard analysis for this type of endpoint would use survival methods such as Kaplan-Meier estimation and a Cox proportional-hazards model, with an appropriate comparison between randomized treatment groups.
The registry information summarized here does not specify the exact survival model, test statistic, stratification factors, censoring rules, or effect measure used for the posted analysis. The statistical methods described below therefore explain the conventional approach to these endpoint types rather than attributing an unreported procedure to ADVANCE.
Composite-event analysis
For a composite endpoint, the usual time-to-event analysis considers the time until the first qualifying component. Thus, if a participant experiences one component first, the composite endpoint is considered to have occurred at that time.
A hazard ratio is a relative time-to-event measure. It is not an absolute risk reduction, a probability of benefit, or the percentage of patients who avoid the event.
Microvascular composite analysis
The second primary endpoint combines nephropathy and microvascular eye disease. If analyzed as a time-to-first-event composite, the same general survival framework can be applied: participants are followed until the first qualifying microvascular event, censoring, or the end of the specified observation period.
6. Statistical Methods Explained
Why is factorial randomization useful?
A factorial design permits two intervention questions to be studied within a common randomized population. Instead of treating the two interventions as unrelated experiments, the design creates a formal structure for estimating each factor's effect and, when prespecified, examining whether the effects interact.
What does the interaction term mean?
An interaction asks whether the effect of intervention A is different depending on whether intervention B is present. If there is little evidence of interaction, investigators may summarize the average effect of A across the levels of B. If interaction is substantial, the marginal effect of A can conceal clinically important differences between the relevant factorial combinations.
Why use time-to-event methods for the cardiovascular endpoint?
The endpoint includes events that occur over follow-up rather than at a single fixed time. Survival analysis uses the timing of events and can incorporate participants who have not experienced the endpoint by the time observation ends. This makes it more informative than simply classifying every participant as an event or non-event without considering follow-up time.
Why combine stroke, myocardial infarction, and cardiovascular death?
A composite can increase the number of observed events and thereby improve statistical efficiency. The tradeoff is interpretability: the component events may differ in clinical importance, frequency, and responsiveness to an intervention. A composite result is therefore most informative when the components are also examined individually.
What does censoring mean?
Censoring occurs when the exact event time is not observed during the available follow-up. For example, a participant who remains event-free until the end of observation contributes information up to that point. Standard survival methods require assumptions about the relationship between censoring and the event process.
Why should the two primary endpoints be considered separately?
ADVANCE has two distinct primary endpoints: one cardiovascular composite and one microvascular composite. They represent different clinical domains. Statistical interpretation should therefore preserve that distinction rather than treating the two endpoints as interchangeable measures of a single outcome.
7. Planned Analysis
The registry identifies two primary outcomes to be evaluated during the period from July 2001 - December 2007. The first is the composite of non-fatal stroke, non-fatal myocardial infarction or death from any cardiovascular cause. The second is the composite of new or substantially worsening nephropathy or microvascular eye disease.
Cardiovascular endpoint
A time-to-event analysis would typically estimate the distribution of time to the first qualifying cardiovascular event and compare the randomized intervention groups using a survival-analysis framework.
Microvascular endpoint
A corresponding analysis would typically examine time to the first qualifying nephropathy or microvascular eye-disease event, with appropriate censoring of participants who remain event-free.
Factorial structure
The statistical analysis would ordinarily account for the two intervention factors and assess whether an interaction between them is relevant to interpretation.
Results status
No formal statistical analyses are posted to ClinicalTrials.gov for these endpoints. Results have not been posted on ClinicalTrials.gov.
8. Interpreting a Composite Time-to-Event Endpoint
The cardiovascular primary endpoint counts the first occurrence of any of three specified events: non-fatal stroke, non-fatal myocardial infarction, or death from any cardiovascular cause. The resulting analysis describes the time until the first qualifying component.
A composite cardiovascular endpoint is not equivalent to three separate analyses. A treatment effect on the composite cannot by itself establish that every individual component is affected to the same extent.
If one component occurs much more frequently than the others, that component can contribute disproportionately to the composite result. Conversely, rare components may have limited statistical precision even when they are clinically important.
9. Factorial Design: The Statistical Logic
The factorial design is the feature that most clearly distinguishes ADVANCE statistically from a simple two-group randomized trial. With two intervention dimensions, the design can answer two broad questions within the same trial population.
| Question | Statistical interpretation |
|---|---|
| Effect of blood pressure lowering | Compare outcomes across the levels of the blood-pressure intervention while accounting for the glucose-lowering factor. |
| Effect of glucose lowering | Compare outcomes across the levels of the glucose-lowering intervention while accounting for the blood-pressure factor. |
| Interaction between interventions | Determine whether the effect associated with one intervention differs according to the level of the other intervention. |
This structure can be statistically efficient because one participant population contributes information to more than one intervention question. The efficiency depends on the assumptions underlying the factorial analysis, particularly the interpretation of the intervention effects when interaction is present.
Main effects versus interaction
A main effect is the average effect associated with one intervention factor across the levels of the other factor. An interaction is different: it asks whether that effect changes according to the second factor.
The exact statistical scale for an interaction depends on the outcome model. For a time-to-event endpoint, the interaction may be evaluated on the model's coefficient or hazard scale rather than by subtracting hazard ratios directly.
10. Analysis of the Cardiovascular Composite
The first primary endpoint combines non-fatal stroke, non-fatal myocardial infarction, and death from any cardiovascular cause. Because the registry supplies a defined follow-up window from July 2001 - December 2007, the endpoint is naturally suited to a time-to-event framework.
For a Cox-model analysis, the hazard ratio would compare the estimated instantaneous rate of the first cardiovascular composite event between randomized intervention groups. A Kaplan-Meier curve would provide an estimate of the proportion remaining free of the composite event over time.
The confidence interval would quantify uncertainty around the estimated relative effect. A narrow interval would indicate greater statistical precision than a wide interval, but neither the point estimate nor its confidence interval alone describes the absolute number of events prevented.
11. Analysis of the Microvascular Composite
The second primary endpoint is a composite of new or substantially worsening nephropathy or microvascular eye disease. This endpoint addresses a different clinical domain from the cardiovascular composite.
As with the cardiovascular endpoint, a time-to-first-event analysis can accommodate different event times and right censoring. The interpretation should retain the distinction between the two components because nephropathy and microvascular eye disease are not identical outcomes.
12. Confidence Intervals, P-Values, and Effect Size
When results from a randomized trial are reported, three related but distinct quantities are commonly considered: the effect estimate, its confidence interval, and the hypothesis-test p-value.
Effect estimate
Describes the magnitude and direction of the observed association between randomized treatment assignment and the endpoint.
Confidence interval
Describes statistical uncertainty around the estimated treatment effect under the specified model and sampling framework.
P-value
Measures the compatibility of the observed data with a specified null hypothesis under the statistical testing framework. It is not a measure of clinical importance.
Absolute effect
Communicates the difference in event probability or survival at a specified time and can complement a relative effect measure such as a hazard ratio.
For ADVANCE, the registry information summarized here contains no posted estimates, confidence intervals, or p-values for either primary endpoint. Consequently, there is no observed effect size to interpret from the registry record itself.
13. Statistical Assumptions
Proportional hazards
If a Cox proportional-hazards model is used, the hazard ratio has its clearest interpretation when the relative hazard is reasonably stable over time. Strong departures from proportional hazards can make a single hazard ratio an incomplete summary of the treatment effect.
Independent censoring
Standard Kaplan-Meier and Cox analyses rely on assumptions concerning censoring. Informally, participants who are censored should not have an event risk that is systematically different from what the analysis assumes after accounting for observed information.
Factorial interaction
The interpretation of a factorial main effect depends on the relationship between the two intervention factors. If a meaningful interaction exists, reporting only marginal treatment effects can conceal an important difference between factorial combinations.
Composite endpoint coherence
A composite endpoint is most straightforward to interpret when its components are sufficiently related in clinical importance, occurrence, and expected treatment response. If those properties differ substantially, the overall composite can be driven by only one component.
14. Multiplicity and Two Primary Endpoints
ADVANCE has two registered primary endpoints. Multiple primary endpoints create a statistical multiplicity issue because testing more than one primary hypothesis can increase the overall probability of a false-positive conclusion unless the testing strategy accounts for the number and relationship of hypotheses.
| Primary endpoint | Clinical domain | Multiplicity question |
|---|---|---|
| Non-fatal stroke, non-fatal myocardial infarction or cardiovascular death | Cardiovascular | How should the two primary hypotheses be controlled within the confirmatory testing framework? |
| New or substantially worsening nephropathy or microvascular eye disease | Microvascular | How should evidence for the second primary endpoint be interpreted alongside the cardiovascular endpoint? |
The registry record summarized here does not specify the multiplicity-adjustment procedure, alpha allocation, hierarchical testing strategy, or gatekeeping rule.
15. Analysis Populations and Randomization
Randomization is the foundation for causal comparison in a randomized clinical trial. An intention-to-treat analysis generally preserves that randomized comparison by analyzing participants according to their assigned intervention factor, regardless of later adherence or treatment changes.
By contrast, an analysis restricted to participants who fully adhered to treatment can introduce post-randomization selection. For that reason, the analysis population and treatment-assignment rule are important components of the statistical interpretation.
16. Missing Data and Censoring
Longitudinal prevention trials can encounter incomplete follow-up, withdrawal, or other forms of missing outcome information. For time-to-event endpoints, the statistical analysis typically handles incomplete observation through censoring rather than treating every participant without an observed event as having the same follow-up duration.
The key distinction is between a participant who remains event-free through a known observation period and a participant whose subsequent outcome becomes unknown after loss to follow-up. The assumptions used to handle censoring can affect the validity of survival estimates.
17. Safety and Adverse Events
The registry information summarized here identifies the two interventions but does not provide serious adverse-event counts by arm. A treatment-effect analysis therefore cannot be paired with an arm-specific serious-adverse-event table from the registry record.
For a randomized trial of preventive therapy, safety interpretation is especially important because the statistical question is not only whether an intervention changes a disease endpoint but also how frequently adverse outcomes occur under each randomized strategy.
| Safety information | Registry record |
|---|---|
| Serious adverse events by arm | Not reported in the record summarized here. |
| Arm-specific adverse-event counts | Not reported in the record summarized here. |
| Safety analysis population | Not specified in the record summarized here. |
18. Trial Timeline
Trial start
The ADVANCE trial began in June 2001.
Primary endpoint time frame
The registry specifies this period for both registered primary endpoints.
Primary completion
The registry lists March 2008 as the primary completion date.
Trial status
The registry lists the study status as completed.
19. What a Hazard Ratio Would Mean in This Trial
If a hazard ratio below 1 were reported for an intervention factor, it would indicate a lower estimated instantaneous rate of first cardiovascular composite events in that randomized intervention group relative to its comparator under the fitted survival model.
If a hazard ratio below 1 were reported for the microvascular composite, it would indicate a lower estimated instantaneous rate of first nephropathy or microvascular eye-disease events under the fitted model.
A hazard ratio would not mean that the same percentage of every participant's personal risk was reduced, nor would it describe the absolute number of events prevented. Absolute risks at clinically relevant time points would provide complementary information.
20. Why This Trial Matters Statistically
ADVANCE is an instructive statistical design because it combines randomized prevention research with a factorial intervention structure and two clinically distinct composite primary endpoints. It illustrates why trial interpretation depends not only on the final effect estimate but also on how interventions, endpoints, follow-up, and statistical hypotheses are structured.
| Concept | How it appears in ADVANCE |
|---|---|
| Randomization | The registry classifies allocation as randomized. |
| Factorial design | The registry identifies a factorial design involving blood pressure and glucose-lowering interventions. |
| Double blinding | The registry identifies the study as double-blind. |
| Composite endpoint | Both registered primary endpoints combine multiple clinical outcomes. |
| Time-to-event analysis | The cardiovascular composite has a longitudinal July 2001 - December 2007 time frame and is naturally suited to survival methods. |
| Interaction | A factorial design creates a specific question about whether one intervention's effect depends on the other intervention factor. |
| Multiplicity | Two registered primary endpoints create a need to distinguish separate hypotheses and their error-control framework. |
| Censoring | Longitudinal event analysis requires appropriate treatment of participants whose event status is not observed for the entire follow-up period. |
| Component interpretation | Each composite endpoint requires attention to the individual events contributing to the composite. |
21. Important Limitations and Interpretation Issues
- No posted statistical analyses: ClinicalTrials.gov does not provide estimates, confidence intervals, or p-values for the two primary endpoints in the record summarized here.
- Comparator details: The registry information summarized here identifies the interventions and factorial design but does not specify the comparator regimens or allocation ratios.
- Factorial interaction: A factorial design requires attention to possible interaction between intervention factors before marginal effects are interpreted as broadly applicable across the other factor.
- Composite endpoints: The components of a composite can differ in frequency, clinical importance, and response to treatment, so the composite should not automatically be treated as equivalent to each component.
- Two primary endpoints: Separate primary hypotheses require an explicitly defined multiplicity strategy for confirmatory interpretation.
- Censoring and follow-up: Time-to-event estimates depend on the handling and assumptions surrounding censoring and incomplete follow-up.
- Unreported statistical details: The registry record summarized here does not specify the exact survival model, stratification variables, missing-data strategy, or multiplicity procedure.
- Safety information: Serious adverse events by randomized arm are not reported in the record summarized here.
22. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial uses a randomized, double-blind factorial structure with two registered primary composite endpoints. The appropriate analysis must account for time-to-event follow-up, factorial intervention factors, composite-event definitions, and the presence of two primary hypotheses.
Clinical interpretation
The clinical questions concern prevention of cardiovascular and microvascular disease in high-risk patients with type 2 diabetes. The cardiovascular and microvascular endpoints represent distinct domains and should therefore remain clinically distinguishable when interpreting results.
23. Related Statistical Tutorials
Learn more about the methods used in this trial:
24. Related Statistical Calculators
25. Sources
- ClinicalTrials.gov: NCT00145925 — ADVANCE.
Continue through Clinical Biostats
Explore statistical tutorials, calculators, and clinical trial analyses for deeper study of trial design and statistical methodology.
26. Record Summary
ADVANCE is a phase 3 randomized, double-blind factorial trial enrolling 11,140 high-risk patients with type 2 diabetes. The registry identifies two intervention strategies—perindopril-indapamide for blood pressure lowering and gliclazide MR-based glucose lowering—and two primary composite endpoints assessed during July 2001 - December 2007: non-fatal stroke, non-fatal myocardial infarction or death from any cardiovascular cause; and new or substantially worsening nephropathy or microvascular eye disease.
From a statistical perspective, the trial provides a useful framework for understanding factorial randomization, main effects and interaction, composite endpoints, time-to-event analysis, censoring, and multiplicity across multiple primary hypotheses. The registry record does not post formal statistical analyses or outcome estimates for these endpoints, so the appropriate statistical focus is the structure of the planned comparisons and the assumptions required to interpret them.