This page separates reported trial results from statistical interpretation. Numerical results are restricted to the information contained in the ClinicalTrials.gov record.
1. Trial at a Glance
CREDENCE was a randomized, double-blind, parallel phase 3 trial evaluating canagliflozin versus placebo in participants with type 2 diabetes and diabetic nephropathy. The registered primary endpoint was a time-to-event composite of doubling of serum creatinine, end-stage kidney disease, and renal or cardiovascular death.
| Feature | CREDENCE |
|---|---|
| Trial name | CREDENCE |
| Phase | Phase 3 |
| Status | Completed |
| Population | Participants with type 2 diabetes and diabetic nephropathy |
| Design | Randomized, double-blind, parallel |
| Primary purpose | Treatment |
| Enrollment | 4401 |
| Interventions | Canagliflozin and placebo |
| Primary endpoint type | Time-to-event |
| Primary hypothesis | Superiority |
| Primary analysis | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| ClinicalTrials.gov | NCT02065791 |
| Lead sponsor | Janssen Research & Development, LLC |
2. Clinical Question
The central statistical question was whether participants randomized to canagliflozin experienced a different time-to-event rate for the registered renal and cardiovascular composite endpoint than participants randomized to placebo.
Population
Participants with Type 2 Diabetes Mellitus and Diabetic Nephropathy.
Intervention
Canagliflozin, recorded in the trial data as the active drug intervention.
Comparator
Placebo, recorded in the trial data as the comparator drug intervention.
Primary question
Does canagliflozin reduce the hazard of the registered primary composite endpoint relative to placebo?
3. Trial Design
Canagliflozin 100 mg
- Canagliflozin
- 2200 participants at risk for the reported serious-adverse-event summary
Placebo
- Placebo
- 2197 participants at risk for the reported serious-adverse-event summary
The ClinicalTrials.gov record identifies the trial as randomized, parallel, double-blind, and designed for treatment. They do not provide additional design features such as a crossover scheme, factorial structure, stratification factors, interim-monitoring rules, missing-data imputation strategy, or Bayesian methods. Those features are therefore not inferred here.
4. Endpoints
| Endpoint | Registry definition / description | Time frame | Type |
|---|---|---|---|
| Primary Composite Endpoint of Doubling of Serum Creatinine (DoSC), End-stage Kidney Disease (ESKD), and Renal or Cardiovascular (CV) Death | Primary composite endpoint is the composite of DoSC, ESKD, and renal or CV death. DoSC is based on a baseline average determination that is sustained and confirmed by repeat central laboratory measurement after at least 30 days and preferably within 60 days. ESKD includes initiation of maintenance dialysis for at least 30 days, renal transplantation, or an estimated glomerular filtration rate (eGFR) value of less than the threshold specified in the registry definition. | Up to 4.6 years | Time-to-event |
The registry data identify 1 primary endpoint and 8 posted outcome measures. The formal statistical analyses in the ClinicalTrials.gov record include the primary endpoint and seven secondary endpoints.
5. Statistical Methodology
Intention-to-treat analysis
The primary analysis population was the intent-to-treat (ITT) population, defined in the registry analysis as all randomized participants. This is an important feature of the causal comparison because treatment groups are defined by randomization rather than by what treatment participants ultimately received.
Cox proportional-hazards model
The registered primary analysis used a Cox proportional-hazards model. This is a survival-analysis model for comparing the instantaneous event rate between groups while accounting for differing follow-up times and right-censoring.
For the reported analyses, an HR below 1 indicates a lower estimated instantaneous event rate in the canagliflozin group than in the placebo group, under the Cox-model framework.
Hazard ratio as the effect measure
The trial data report treatment effects as hazard ratios, accompanied by two-sided 95% confidence intervals. A hazard ratio is a relative time-to-event measure. It should not be interpreted as an absolute risk difference, a probability of an individual experiencing an event, or a percentage of participants who benefit.
Superiority hypothesis
The reported hypothesis type is superiority. Thus, the inferential question is whether the canagliflozin group has a different event hazard from the placebo group, with the reported direction of the estimates favoring a hazard ratio below 1.
Time-to-event structure
The primary and all seven registry-reported secondary analyses are time-to-event analyses with a time frame of up to 4.6 years. This means that follow-up duration is part of the statistical structure rather than treating every participant as if they had identical observation time.
6. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
The primary endpoint records time until a composite clinical event, rather than simply whether an event ever occurred. Participants can also have different amounts of observed follow-up. A Cox model is designed for this setting because it uses event timing and accommodates right-censored observations.
What does an HR of 0.70 mean?
An HR of 0.70 means that the estimated hazard in the canagliflozin group was 0.70 times the estimated hazard in the placebo group under the fitted Cox model. Expressed as a relative comparison, this corresponds to a 30% lower estimated hazard because \(1-0.70=0.30\).
This does not mean that exactly 30% fewer participants experienced the endpoint, nor does it mean that each individual participant had exactly 30% lower risk.
Why is the ITT population important?
All randomized participants are retained in the primary analysis population according to their randomized assignment. This preserves the comparison created by randomization and reduces the potential for treatment-group comparisons to be distorted by post-randomization treatment decisions.
What does the 95% confidence interval tell us?
The 95% confidence interval describes statistical uncertainty around the estimated hazard ratio under the analysis model and sampling framework. For the primary endpoint, the interval is 0.59 to 0.82. It therefore communicates considerably more than the point estimate alone: it shows the range of model-based effect estimates compatible with the data at the stated confidence level.
The confidence interval does not describe the range of effects that individual patients experienced.
Why does the p-value not measure effect size?
The p-value addresses the strength of evidence against the null hypothesis within the specified testing framework. It is not a measure of how large or clinically important the treatment effect is. Effect size is better represented here by the hazard ratio and its confidence interval, while absolute event measures would provide a different perspective.
What assumption accompanies a Cox hazard ratio?
The conventional Cox proportional-hazards interpretation assumes proportionality of the hazards over time. If the hazard ratio changes substantially over follow-up, a single HR can become a less complete description of the treatment-time relationship. The ClinicalTrials.gov record does not report a formal proportional-hazards diagnostic, so no such diagnostic is inferred here.
7. Primary Result
Primary Composite Endpoint
The primary endpoint was the composite of doubling of serum creatinine, end-stage kidney disease, and renal or cardiovascular death, assessed for up to 4.6 years. The analysis used the ITT population and compared canagliflozin 100 mg with placebo using a Cox proportional-hazards model.
Hazard ratio for the primary composite endpoint
95% CI: 0.59–0.82 · P < 0.0001
Two-sided 95% confidence interval · Superiority hypothesis
The estimated hazard ratio of 0.70 indicates that the fitted Cox model estimated the instantaneous rate of experiencing the primary composite endpoint in the canagliflozin group at 70% of the corresponding estimated hazard in the placebo group. Equivalently, the estimate corresponds to a 30% lower estimated hazard under the model.
The HR does not mean that 30% of participants avoided the endpoint, that each participant's probability was reduced by exactly 30%, or that the absolute event risk differed by 30 percentage points. Hazard ratios are relative time-to-event measures.
The 95% CI of 0.59–0.82 expresses uncertainty around the estimated HR. It does not represent the range of individual treatment responses. Because the interval remains below 1.00, the reported estimate is directionally consistent with a lower hazard in the canagliflozin group throughout the interval.
The P < 0.0001 result provides evidence against the null hypothesis in the reported superiority analysis. It does not quantify the magnitude of the treatment effect; that information comes from the HR and confidence interval.
The analysis should also be interpreted within the Cox-model framework, including its proportional-hazards assumption and the handling of censoring. The ClinicalTrials.gov record does not report a formal proportional-hazards diagnostic or a reconstructed Kaplan-Meier curve.
8. Secondary Endpoint Results
The registry data contain seven secondary statistical analyses. All were performed in the ITT population, compared placebo with canagliflozin 100 mg, used Cox proportional-hazards methodology, and reported two-sided 95% confidence intervals.
| Secondary endpoint | HR | 95% CI | P-value |
|---|---|---|---|
| Composite Endpoint of CV Death and Hospitalized Heart Failure (HHF) | 0.69 | 0.57–0.83 | =0.0001 |
| Major Adverse Cardiac Event (MACE) | 0.80 | 0.67–0.95 | =0.0121 |
| Hospitalized Heart Failure (HHF) | 0.61 | 0.47–0.80 | =0.0003 |
| Renal Composite Endpoint | 0.66 | 0.53–0.81 | <0.0001 |
| Cardiovascular (CV) Death | 0.78 | 0.61–1.00 | =0.0502 |
| All-cause Mortality | 0.83 | 0.68–1.02 | = 0.0727 |
| CV Composite Endpoint | 0.74 | 0.63–0.86 | 0.0001 |
Composite Endpoint of CV Death and Hospitalized Heart Failure
Hazard ratio
95% CI: 0.57–0.83 · P = 0.0001
The estimate corresponds to a 31% lower estimated hazard for the composite endpoint under the reported Cox model. The confidence interval is entirely below 1.00, while the p-value provides evidence against the null hypothesis under the reported superiority analysis.
Major Adverse Cardiac Event
Hazard ratio
95% CI: 0.67–0.95 · P = 0.0121
The HR of 0.80 corresponds to a 20% lower estimated hazard for MACE under the Cox model. The 95% CI remains below 1.00, although its upper limit is closer to the null value than several of the other secondary estimates.
Hospitalized Heart Failure
Hazard ratio
95% CI: 0.47–0.80 · P = 0.0003
The HR of 0.61 corresponds to a 39% lower estimated hazard for hospitalized heart failure under the reported Cox model. The confidence interval indicates uncertainty around that estimate while remaining below 1.00.
Renal Composite Endpoint
Hazard ratio
95% CI: 0.53–0.81 · P < 0.0001
The HR of 0.66 corresponds to a 34% lower estimated hazard for the renal composite endpoint. The interval from 0.53 to 0.81 describes uncertainty around the model-based estimate.
Cardiovascular Death
Hazard ratio
95% CI: 0.61–1.00 · P = 0.0502
The point estimate corresponds to a 22% lower estimated hazard. Unlike several other reported secondary endpoints, the 95% CI reaches 1.00, and the reported p-value is = 0.0502. The appropriate interpretation is therefore based on the full estimate and interval rather than treating the p-value as a measure of effect size.
All-cause Mortality
Hazard ratio
95% CI: 0.68–1.02 · P = 0.0727
The point estimate corresponds to a 17% lower estimated hazard, but the 95% CI extends above 1.00. The reported p-value is = 0.0727. The point estimate should therefore not be presented as establishing a definitive reduction in all-cause mortality.
CV Composite Endpoint
Hazard ratio
95% CI: 0.63–0.86 · P = 0.0001
The HR of 0.74 corresponds to a 26% lower estimated hazard under the Cox model. The confidence interval is below 1.00 across its reported range.
9. Interpreting the Pattern of Hazard Ratios
| Endpoint | HR | Approximate relative hazard interpretation |
|---|---|---|
| Primary composite | 0.70 | 30% lower estimated hazard |
| CV death + HHF | 0.69 | 31% lower estimated hazard |
| MACE | 0.80 | 20% lower estimated hazard |
| HHF | 0.61 | 39% lower estimated hazard |
| Renal composite | 0.66 | 34% lower estimated hazard |
| CV death | 0.78 | 22% lower estimated hazard |
| All-cause mortality | 0.83 | 17% lower estimated hazard |
| CV composite | 0.74 | 26% lower estimated hazard |
The reported estimates are all below 1.00, but their confidence intervals and p-values differ. This distinction is important: the point estimate alone does not establish the precision or inferential strength of an analysis.
The smallest reported HR among these endpoints is 0.61 for hospitalized heart failure, while the largest is 0.83 for all-cause mortality. These are separate endpoint-specific estimates, not evidence that one endpoint is intrinsically more important than another.
10. Primary Endpoint: What the Confidence Interval Adds
The primary HR of 0.70 is the single best estimate reported by the fitted Cox model for the relative treatment effect.
The 0.59–0.82 interval communicates the statistical precision of the estimate. It is narrower than an interval extending over a very large range and remains below 1.00 throughout.
The reported P < 0.0001 describes evidence against the null hypothesis within the stated statistical framework. It should not be read as a 0.01% probability that the null hypothesis is true, nor as a measure of clinical magnitude.
These three quantities answer different questions. The HR describes the estimated relative effect, the confidence interval describes uncertainty around that estimate, and the p-value addresses evidence against the null hypothesis.
11. Safety
The ClinicalTrials.gov record reports serious adverse events by randomized arm using affected participants over the stated at-risk populations.
| Safety measure | Placebo | Canagliflozin 100 mg |
|---|---|---|
| Serious adverse events | 806/2197 | 737/2200 |
Safety and efficacy answer different questions and should not be collapsed into a single numerical "benefit-risk" statistic. The ClinicalTrials.gov record provides the serious-adverse-event counts and denominators but do not provide a formal comparative safety analysis, confidence interval, or p-value for this safety measure.
12. Statistical Interpretation of the Primary Analysis
Relative effect versus absolute effect
The primary HR of 0.70 is a relative measure. It describes the relationship between the estimated hazards in the two randomized groups. It does not tell us the absolute number of events prevented, the absolute risk difference, or the number needed to treat.
Those measures require absolute event probabilities or event counts over a defined time horizon. The ClinicalTrials.gov record does not provide those quantities for the primary endpoint, so they are not calculated here.
Why the time dimension matters
A time-to-event endpoint contains information about when events occur. Two studies could have the same eventual proportion of participants experiencing an event but different timing of those events. Cox regression uses the event-time structure rather than reducing follow-up to a simple binary outcome.
Censoring
Time-to-event analyses commonly encounter participants who have not experienced the endpoint by the time their available follow-up ends. Such observations are censored rather than treated as if they experienced the event. The ClinicalTrials.gov record identifies the endpoint as time-to-event and the analysis as Cox proportional hazards, but do not provide detailed censoring rules. No additional censoring assumptions are therefore attributed to CREDENCE here.
Composite endpoints
The primary endpoint combines several clinically distinct event types: doubling of serum creatinine, end-stage kidney disease, and renal or cardiovascular death. A composite endpoint can increase the number of observed events and provide an overall time-to-first-composite-event comparison, but the interpretation of the composite is not automatically identical to the interpretation of each individual component.
13. What the Primary HR Does — and Does Not — Mean
What it means
Under the reported Cox model, the estimated hazard for the primary composite endpoint was 0.70 times the placebo-group hazard.
What it does not mean
It does not mean that 30% of participants were protected, that absolute risk fell by 30 percentage points, or that every participant experienced the same relative reduction.
What the CI means
The 95% CI of 0.59–0.82 quantifies uncertainty around the estimated HR under the specified statistical framework.
What the p-value means
P < 0.0001 indicates strong evidence against the null hypothesis in the reported analysis; it does not measure effect magnitude.
14. Analysis Population and Causal Interpretation
The primary efficacy analysis used the ITT population consisting of all randomized participants. This is especially important in a randomized trial because the treatment comparison is defined at the moment of randomization.
The strength of the ITT framework is that it preserves the original randomized comparison rather than redefining groups according to subsequent treatment exposure.
The registry data do not report a separate per-protocol or as-treated efficacy analysis. Accordingly, no such analysis is presented here.
15. Limitations
- Registry-level information: this analysis is constrained to the ClinicalTrials.gov record. Additional protocol or statistical-analysis-plan details are not inferred.
- No absolute event counts for efficacy: the statistical analyses posted on ClinicalTrials.gov report event rates per 1000 participant-years as the outcome unit, but do not provide the underlying event counts in the ClinicalTrials.gov record.
- No Kaplan-Meier estimates reported: the data identify time-to-event methodology but do not provide Kaplan-Meier curves, medians, or time-specific survival estimates.
- Proportional-hazards assumption: the Cox HR is model-based. The ClinicalTrials.gov record does not report a formal diagnostic for proportional hazards.
- Composite endpoint: the primary outcome combines several distinct renal and cardiovascular events, so the overall HR should not automatically be interpreted as the effect on every component separately.
- Multiplicity information is incomplete: seven secondary analyses are reported, but the ClinicalTrials.gov record does not specify a multiplicity-adjustment procedure or hierarchical testing strategy.
- Secondary endpoints: the individual secondary p-values should be interpreted in the context of their role as secondary analyses rather than automatically treated as separate confirmatory hypotheses.
- Safety comparison: serious adverse-event counts are available, but the ClinicalTrials.gov record does not include a formal comparative safety model or confidence interval.
- Missing-data and censoring details: the ClinicalTrials.gov record does not specify detailed rules for handling missing observations or censoring beyond identifying the endpoint as time-to-event.
16. Design Features Not Reported in the Supplied Data
Some statistical features commonly encountered in large randomized trials are not present in the registry-reported CREDENCE data. They are deliberately not reconstructed from external knowledge.
| Design topic | What can be stated from the ClinicalTrials.gov record |
|---|---|
| Non-inferiority margin | Not applicable to the reported hypothesis type; the analyses posted on ClinicalTrials.gov specify superiority. |
| Crossover | Not reported in the ClinicalTrials.gov record. |
| Factorial design | The design model is parallel; no factorial structure is reported. |
| Interim analysis | Not reported in the ClinicalTrials.gov record. |
| Multiplicity adjustment | Not reported in the ClinicalTrials.gov record. |
| Missing-data / imputation method | Not reported in the ClinicalTrials.gov record. |
| Stratification factors | Not reported in the ClinicalTrials.gov record. |
| Bayesian methods | Not reported; the registry-reported method is Cox proportional hazards. |
This distinction is important for a statistical analysis page: absence of a registry-reported methodological detail is not evidence that the underlying trial did not have such a procedure. It simply means the procedure cannot be responsibly described from the ClinicalTrials.gov record.
17. Why This Trial Matters Statistically
CREDENCE provides a compact teaching example of how randomized clinical-trial evidence is constructed around a time-to-event endpoint. The statistical story is not simply the reported p-value. It combines the randomized design, ITT population, composite endpoint, Cox model, hazard ratio, confidence interval, and the distinction between primary and secondary analyses.
| Concept | How it appears in CREDENCE |
|---|---|
| Randomization | The trial is randomized with 4401 participants enrolled. |
| Blinding | The design is double-blind. |
| ITT analysis | The primary analysis population consists of all randomized participants. |
| Time-to-event endpoint | The primary endpoint is assessed over up to 4.6 years. |
| Cox proportional hazards | The primary and registry-reported secondary analyses use Cox proportional-hazards methodology. |
| Hazard ratio | The primary estimate is HR 0.70, with secondary HRs ranging from 0.61 to 0.83. |
| Confidence interval | The primary 95% CI is 0.59–0.82. |
| P-value | The primary analysis reports P < 0.0001. |
| Composite endpoint | The primary endpoint combines doubling of serum creatinine, ESKD, and renal or CV death. |
| Secondary analyses | Seven secondary time-to-event analyses are reported. |
| Safety | Serious adverse-event counts are reported by randomized arm. |
18. A Practical Framework for Reading the CREDENCE Results
Step 1 · Identify the estimand
The primary question concerns the relative time-to-event rate for the registered composite endpoint comparing randomized canagliflozin and placebo groups.
Step 2 · Identify the population
The primary efficacy analysis uses all randomized participants in the ITT population.
Step 3 · Identify the model
The registry-reported analysis uses a Cox proportional-hazards model and reports a hazard ratio.
Step 4 · Read the full estimate
Read HR 0.70 together with the 95% CI 0.59–0.82 rather than relying on the p-value alone.
Step 5 · Separate primary from secondary
The primary endpoint has a distinct role from the seven secondary endpoints, even when their estimates point in the same direction.
Step 6 · Keep safety separate
Serious adverse-event counts provide safety information but are not directly interchangeable with the efficacy hazard ratios.
19. Trial Timeline
Trial start
The ClinicalTrials.gov record lists February 17, 2014 as the study start date.
Primary completion
The ClinicalTrials.gov record lists October 30, 2018 as the primary completion date.
Registry status
The trial is recorded as completed, with results posted for eight outcome measures and eight statistical analyses.
20. Reported Results at a Glance
| Endpoint role | Endpoint | HR | 95% CI | P-value |
|---|---|---|---|---|
| Primary | Primary composite of DoSC, ESKD, and renal or CV death | 0.70 | 0.59–0.82 | <0.0001 |
| Secondary | CV death + HHF | 0.69 | 0.57–0.83 | =0.0001 |
| Secondary | MACE | 0.80 | 0.67–0.95 | =0.0121 |
| Secondary | HHF | 0.61 | 0.47–0.80 | =0.0003 |
| Secondary | Renal composite endpoint | 0.66 | 0.53–0.81 | <0.0001 |
| Secondary | CV death | 0.78 | 0.61–1.00 | =0.0502 |
| Secondary | All-cause mortality | 0.83 | 0.68–1.02 | = 0.0727 |
| Secondary | CV composite endpoint | 0.74 | 0.63–0.86 | 0.0001 |
All eight registry-reported analyses use the same broad inferential structure: ITT population, placebo versus canagliflozin 100 mg, Cox proportional-hazards analysis, hazard ratio as the effect measure, and a time frame of up to 4.6 years.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: CREDENCE, NCT02065791.
- Linked publication record: PubMed record — PMID 42475421.
- Linked publication record: PubMed record — PMID 41926219.
- Linked publication record: PubMed record — PMID 41817689.
- Linked publication record: PubMed record — PMID 41198855.
- Linked publication record: PubMed record — PMID 41196655.
The numerical results and trial characteristics presented throughout this page are restricted to the registry-reported CREDENCE trial data. The linked PubMed records are provided as publication-navigation sources; no additional numerical trial results from those records are incorporated into this analysis.
Continue through the Clinical Biostats statistical pathway
Use the trial as a starting point for deeper study of survival analysis, hazard ratios, confidence intervals, randomization, and intention-to-treat analysis.
24. Record Summary
CREDENCE is a randomized, double-blind, parallel phase 3 trial with 4401 enrolled participants and a primary time-to-event endpoint assessed for up to 4.6 years. The primary ITT analysis used a Cox proportional-hazards model and reported a hazard ratio of 0.70 with a two-sided 95% CI of 0.59–0.82 and P < 0.0001. Seven secondary time-to-event analyses were also posted, with hazard ratios ranging from 0.61 to 0.83.
The most important statistical lesson is that these results should be read as a coordinated set of design and inference elements: randomization establishes the comparison, the ITT population preserves that randomized assignment, the time-to-event framework accounts for follow-up and event timing, the Cox model produces the hazard ratio, and the confidence interval communicates uncertainty around that estimate. Secondary endpoints and safety results add information but should retain their distinct statistical roles.