← Clinical Trials
Chronic Kidney Disease Phase 3 Time-to-Event NCT03036150

DAPA-CKD: Complete Statistical Analysis of Dapagliflozin in Chronic Kidney Disease

An independent statistical review of the randomized phase 3 DAPA-CKD trial evaluating dapagliflozin versus placebo for a composite renal and cardiovascular mortality endpoint and related time-to-event outcomes in patients with chronic kidney disease.

Study start: 2017-02-02  ·  Primary completion: 2020-06-12  ·  Enrollment: 4304
Scope of this record

This page separates reported trial results from statistical interpretation. The numerical results presented here are taken from the ClinicalTrials.gov record. The registry provides the official trial record.

Registry note: 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

DAPA-CKD was a randomized, parallel-group, quadruple-masked phase 3 treatment trial evaluating dapagliflozin versus placebo in patients with chronic kidney disease. The primary endpoint was a time-to-event composite incorporating sustained decline in estimated glomerular filtration rate, end-stage renal disease, cardiovascular death, or renal death.

4304
Enrolled
2 treatment arms
2
Arms
Dapagliflozin vs placebo
0.61
Primary HR
95% CI 0.51–0.72
< 0.0001
Primary P-value
Superiority analysis
FeatureDAPA-CKD
Trial nameDAPA-CKD
PhasePhase 3
Therapeutic areaNephrology
ConditionChronic Kidney Disease
DesignRandomized, parallel-group
MaskingQuadruple
AllocationRandomized
Primary purposeTreatment
Enrollment4304
InterventionsDapagliflozin; placebo
Primary endpoint typeTime-to-event
Primary hypothesisSuperiority
Primary analysisCox proportional-hazards model
Lead sponsorAstraZeneca
Study statusCompleted
ClinicalTrials.govNCT03036150

2. Clinical Question

The central statistical question was whether treatment with dapagliflozin, compared with placebo, changed the time until the first occurrence of the registered composite primary endpoint in patients with chronic kidney disease.

Population

Patients with chronic kidney disease enrolled in the phase 3 DAPA-CKD trial.

Intervention

Dapagliflozin.

Comparator

Placebo.

Primary question

Does dapagliflozin produce a different time-to-event experience from placebo for the registered composite endpoint?

The trial's statistical hypothesis was explicitly one of superiority. That distinction matters: the objective was to estimate and test whether the randomized treatment groups differed, rather than to demonstrate that dapagliflozin was merely no worse than placebo within a prespecified non-inferiority margin.

3. Trial Design

01
Randomize4304 enrolled
02
AssignDapagliflozin or placebo
03
FollowTime-to-event outcomes
04
AnalyzeCox regression
05
EstimateHazard ratios and 95% CIs
ARM A

Dapagliflozin

  • Dapagliflozin was the active intervention.
  • Patients were analyzed according to randomized treatment assignment for the reported efficacy analyses.
ARM B

Placebo

  • Placebo was the comparator intervention.
  • Patients were analyzed according to randomized treatment assignment for the reported efficacy analyses.
Allocation
Randomized allocation in a parallel-group design.
Masking
Quadruple masking.
Primary purpose
Treatment.
Study period
Started 2017-02-02 and reached primary completion on 2020-06-12.

4. Endpoints

EndpointRegistry definition / time frameType
Primary composite endpoint Time to the First Occurrence of Any of the Components of the Composite: ≥50% Sustained Decline in eGFR or Reaching ESRD or CV Death or Renal Death.

Up to 38.2 months.
Time-to-event
Secondary renal composite Time to the First Occurrence of Any of the Components of the Composite: ≥50% Sustained Decline in eGFR or Reaching ESRD or Renal Death.

Up to 38.2 months.
Time-to-event
Secondary cardiovascular composite Time to the First Occurrence of Either of the Components of the Composite: CV Death or Hospitalization for Heart Failure.

Up to 38.2 months.
Time-to-event
Secondary mortality endpoint Time to Death From Any Cause.

Up to 38.2 months.
Time-to-event

The primary endpoint is a composite time-to-event endpoint. The registry definition combines several clinically distinct events into one time-to-first-event outcome. The statistical analysis therefore asks whether the time until the first qualifying component differs between treatment groups; it does not, by itself, establish that every component has the same treatment effect.

ESRD definition: the registry defines end-stage renal disease as sustained eGFR <15 mL/min/1.73m2, chronic dialysis treatment, or receiving a renal transplant.

5. Statistical Methodology

Cox proportional-hazards model

The registry reports a Cox proportional-hazards regression model for the primary endpoint and each of the three reported secondary analyses. The effect measure was the hazard ratio.

For the primary endpoint, the Full Analysis Set was used. This included all patients who were randomized to study treatment, with patients analyzed according to their randomized treatment assignment. This analysis structure preserves the treatment contrast created by randomization.

Conceptual Cox model
h(t|X) = h0(t) exp(βX)

The coefficient associated with treatment is transformed into a hazard ratio. In this setting, the reported HR compares the estimated hazard of experiencing the endpoint under dapagliflozin with that under placebo, conditional on the model structure.

Stratification

The primary and secondary Cox analyses were stratified by the randomization stratification of Type 2 Diabetes and urine albumin creatinine ratio. Stratification allows the baseline hazard structure to differ across these strata while retaining a common treatment-effect parameter in the reported model.

Covariate adjustment

The analyses also adjusted for baseline eGFR. This means the reported treatment comparison is not simply an unadjusted comparison of event times; baseline eGFR was incorporated into the regression analysis as a covariate.

Randomization

Randomization establishes the treatment comparison before outcomes occur. An analysis based on randomized assignment retains that central design feature.

Stratification

The Cox model was stratified using the reported randomization strata of Type 2 Diabetes and urine albumin creatinine ratio.

Covariate adjustment

Baseline eGFR was included as an adjustment variable in the reported Cox analyses.

Effect measure

The treatment effect was expressed as a hazard ratio rather than as a risk ratio or absolute risk difference.

Kaplan-Meier estimation and the Cox model

Kaplan-Meier estimation is a standard descriptive method for displaying time-to-event outcomes, while the Cox model provides a regression-based estimate of the relative hazard. The registry analysis specifically reports Cox regression; it does not identify Kaplan-Meier estimation as the formal inferential method in the posted statistical-analysis fields.

Hazard-ratio interpretation
HR = hDapagliflozin(t) / hPlacebo(t)

An HR below 1 indicates a lower estimated instantaneous event rate under dapagliflozin than under placebo within the fitted model. The HR is a relative time-to-event measure, not a direct measure of absolute risk reduction.

6. Analysis Population and Estimand Structure

The registry gives an unusually important detail about the analysis population: the Full Analysis Set was used for the primary endpoint, and it included all patients who were randomized to study treatment. Patients were analyzed according to their randomized treatment assignment.

ElementReported approach
Primary analysis populationFull Analysis Set
Population definitionAll patients randomized to study treatment
Treatment assignment.Analyzed according to randomized treatment assignment.
Primary comparisonDapagliflozin vs placebo
Regression modelCox proportional-hazards model
StratificationType 2 Diabetes and urine albumin creatinine ratio
Adjustment covariateBaseline eGFR

This is closely related to the intention-to-treat principle. The key statistical idea is that treatment assignment remains the defining exposure for the efficacy comparison. That reduces the risk that the estimated treatment effect becomes primarily a comparison of patients who remained on treatment versus patients who did not.

7. Primary Endpoint Result

The primary endpoint was analyzed through the Full Analysis Set using a stratified Cox proportional-hazards model adjusted for baseline eGFR. The comparison was dapagliflozin versus placebo, with a superiority hypothesis.

Hazard ratio for the primary composite endpoint

0.61

95% CI: 0.51–0.72   ·   P < 0.0001

Time frame: up to 38.2 months

Primary endpointDapagliflozin vs placeboAnalysis
Time to first occurrence of ≥50% sustained decline in eGFR, reaching ESRD, CV death, or renal death HR 0.61
95% CI 0.51–0.72
P < 0.0001
Stratified Cox regression, adjusted for baseline eGFR
Clinical Biostats interpretation

An HR of 0.61 means that, under the reported Cox model, the estimated instantaneous rate of experiencing the first qualifying component of the composite endpoint was 0.61 times that under placebo. Expressed as a relative model-based interpretation, this corresponds to an estimated 39% lower hazard for the dapagliflozin group.

The HR does not mean that 39% of patients avoided the endpoint, that an individual patient's absolute risk was reduced by 39%, or that every component of the composite was reduced by exactly 39%. The endpoint is defined by the first occurrence of any qualifying component.

The 95% CI of 0.51–0.72 describes uncertainty around the estimated hazard ratio under the model and sampling framework. It is not an interval containing the treatment effect experienced by individual patients.

The very small P-value, P < 0.0001, addresses evidence against the null hypothesis in the reported superiority analysis. A P-value is not a measure of effect size and does not indicate the probability that the treatment hypothesis is true.

Because the estimate comes from a Cox proportional-hazards model, interpretation of a single HR should also be understood in the context of the model's proportional-hazards assumption. The registry result itself is a model-based summary of the time-to-event comparison rather than a direct statement about absolute event probabilities at every point in follow-up.

8. Secondary Endpoint Results

The registry reports three secondary time-to-event analyses. Each used the Full Analysis Set, randomized treatment assignment, Cox regression, the same reported stratification variables, and adjustment for baseline eGFR.

Renal composite

Hazard ratio for renal composite

0.56

95% CI: 0.45–0.68   ·   P < 0.0001

Time frame: up to 38.2 months

The endpoint was time to the first occurrence of any component of the composite of ≥50% sustained decline in eGFR, reaching ESRD, or renal death. An HR of 0.56 corresponds to an estimated 44% lower hazard under the fitted Cox model for dapagliflozin relative to placebo.

Cardiovascular death or hospitalization for heart failure

Hazard ratio for cardiovascular composite

0.71

95% CI: 0.55–0.92   ·   P = 0.0089

Time frame: up to 38.2 months

The endpoint was time to the first occurrence of either cardiovascular death or hospitalization for heart failure. The HR of 0.71 represents an estimated 29% lower hazard under the fitted model for dapagliflozin relative to placebo.

Death from any cause

Hazard ratio for all-cause mortality

0.69

95% CI: 0.53–0.88   ·   P = 0.0035

Time frame: up to 38.2 months

The endpoint was time to death from any cause. An HR of 0.69 corresponds to an estimated 31% lower hazard of death under the fitted model for dapagliflozin relative to placebo.

EndpointHR95% CIP-value
Primary composite: ≥50% sustained decline in eGFR, ESRD, CV death, or renal death0.610.51–0.72< 0.0001
Renal composite: ≥50% sustained decline in eGFR, ESRD, or renal death0.560.45–0.68< 0.0001
CV death or hospitalization for heart failure0.710.55–0.920.0089
Death from any cause0.690.53–0.880.0035
Educational note: the reported registry fields provide hazard ratios, confidence intervals, P-values, analysis populations, and model specifications. They do not provide the underlying event-time and censoring records needed to reconstruct an exact Kaplan-Meier curve from first principles.

9. How the Primary and Secondary Results Relate

The four reported analyses form a coherent time-to-event analysis set, but they do not answer identical questions. The primary composite combines renal and mortality events. The renal secondary endpoint removes cardiovascular death from that composite, the cardiovascular secondary endpoint focuses on cardiovascular death or hospitalization for heart failure, and the all-cause mortality endpoint considers death from any cause.

Endpoint familyWhat it asksHR
Primary composite When does a patient first experience any of the four qualifying renal or mortality events? 0.61
Renal composite When does a patient first experience a qualifying renal event? 0.56
CV death / heart failure hospitalization When does a patient first experience either cardiovascular death or hospitalization for heart failure? 0.71
All-cause mortality When does the patient die from any cause? 0.69

The HRs should not be interpreted as interchangeable estimates of one underlying quantity. Each corresponds to a different endpoint definition. In particular, the primary composite HR should not be described as though it were simply an all-cause mortality HR.

10. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

The primary endpoint is explicitly a time-to-event outcome. A Cox model is designed to compare event hazards while accounting for differing follow-up times and right censoring. Rather than reducing each participant to a simple yes/no event indicator, the analysis uses the timing of the event or censoring information.

What does a hazard ratio of 0.61 mean?

An HR of 0.61 means the estimated hazard under dapagliflozin was 0.61 times the hazard under placebo in the reported Cox model. The complementary interpretation is a 39% lower estimated hazard. This is not equivalent to saying that the probability of an event was 39% lower at every time point.

Why adjust for baseline eGFR?

Baseline eGFR was specified as a covariate in the reported analysis. Covariate adjustment incorporates information about baseline renal function into the regression model. It can improve statistical efficiency and account for baseline variation relevant to the outcome, while the treatment comparison remains anchored to randomized assignment.

Why stratify the Cox model?

The analysis was stratified by the randomization stratification of Type 2 Diabetes and urine albumin creatinine ratio. A stratified Cox model permits different baseline hazard functions across the strata without requiring separate treatment-effect estimates for every stratum.

Why is the Full Analysis Set important?

The registry states that the Full Analysis Set included all randomized patients and that patients were analyzed according to randomized treatment assignment. This is important because analyzing patients according to the treatment they were assigned to preserves the comparison generated by randomization more directly than restricting analysis to patients who remained exposed to treatment.

What does the 95% confidence interval tell us?

For the primary HR, the 95% CI is 0.51–0.72. It quantifies uncertainty around the estimated HR under the statistical model and sampling framework. A narrower interval generally indicates greater precision than a wider interval, but the interval should not be interpreted as a range of individual patient outcomes.

Why does the P-value not tell us the size of the effect?

The P-value addresses the compatibility of the observed data with the null hypothesis under the specified testing framework. It does not measure how clinically large the effect is. Effect size is communicated by the HR, while precision is communicated by the confidence interval. Both should be read alongside the endpoint definition and analysis population.

11. Composite Endpoints: An Important Statistical Distinction

The primary endpoint combines four components: ≥50% sustained decline in eGFR, reaching ESRD, cardiovascular death, and renal death. The statistical analysis records the time to the first occurrence of any component.

One analysis, multiple events

The primary HR summarizes time to the first qualifying component rather than estimating four independent treatment effects.

Component interpretation

A composite result should not automatically be interpreted as though every component contributes equally to the observed treatment effect.

Different clinical events

The components span renal deterioration and cardiovascular or renal mortality, so the endpoint represents a broad outcome construct.

Secondary endpoint context

The separately reported renal, cardiovascular, and all-cause mortality endpoints help distinguish the questions addressed by each analysis.

This distinction is especially important when communicating the primary result. Saying that the HR was 0.61 for the primary endpoint is precise. Saying that dapagliflozin "reduced kidney failure by 39%" would be a different claim and is not an accurate restatement of the registered primary endpoint.

12. Confidence Intervals and Precision

The confidence intervals posted on ClinicalTrials.gov for the four reported analyses are:

Reported 95% confidence intervals for hazard ratios
Primary composite
0.51–0.72
Renal composite
0.45–0.68
CV / HF
0.55–0.92
All-cause death
0.53–0.88

The intervals provide a useful reminder that an observed HR is an estimate, not a fixed biological constant. The primary estimate of 0.61 is accompanied by a 95% CI extending from 0.51 to 0.72. The corresponding secondary estimates also have uncertainty, with different interval widths reflecting the information available for each endpoint.

A practical reading rule

Read the result in three steps: endpoint first, then effect estimate, then uncertainty. For DAPA-CKD, "primary composite, HR 0.61, 95% CI 0.51–0.72" is substantially more informative than reporting the HR alone.

13. Statistical Interpretation of the P-values

EndpointP-valueWhat it contributes
Primary composite< 0.0001Evidence against the null hypothesis under the reported superiority analysis.
Renal composite< 0.0001Evidence against the null hypothesis for the reported secondary analysis.
CV death or hospitalization for heart failure0.0089Evidence against the null hypothesis for the reported secondary analysis.
Death from any cause0.0035Evidence against the null hypothesis for the reported secondary analysis.

These P-values should be kept distinct from the corresponding effect estimates. For example, the primary result contains both an HR of 0.61 and a P-value of < 0.0001. The former describes the estimated relative hazard; the latter describes the strength of evidence against the null hypothesis under the specified statistical test.

The ClinicalTrials.gov record identifies the hypothesis type as superiority, but do not provide enough information here to reconstruct a complete multiplicity framework for the family of primary and secondary tests. The P-values should therefore be reported exactly as posted rather than assigning them an additional hierarchy or adjusted interpretation that is not contained in the ClinicalTrials.gov record.

14. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm using affected participants divided by participants at risk.

Safety measureDapagliflozinPlacebo
Serious adverse events, affected / at risk633 / 2149729 / 2149

The reported figures indicate the number of participants affected by serious adverse events and the corresponding number at risk in each arm. These are counts by treatment arm, not a hazard ratio and not an efficacy endpoint.

Dapagliflozin

633 participants with serious adverse events among 2149 at risk.

Placebo

729 participants with serious adverse events among 2149 at risk.

Because the ClinicalTrials.gov record provides the affected and at-risk counts rather than a formal statistical comparison, the safety results should not be converted into a treatment-effect estimate or described as statistically different without an appropriate analysis.

15. Trial Timeline

2017-02-02

Study start

The DAPA-CKD study began on February 2, 2017.

Phase 3

Randomized treatment comparison

The completed phase 3 trial used a randomized, parallel-group design with quadruple masking and two interventions: dapagliflozin and placebo.

Up to 38.2 months

Registered outcome window

The primary endpoint and the three reported secondary endpoints were evaluated over a time frame of up to 38.2 months.

2020-06-12

Primary completion

The trial reached primary completion on June 12, 2020.

16. Why This Trial Matters Statistically

DAPA-CKD is a useful statistical teaching case because the registry record brings together several core concepts in survival analysis: randomized treatment assignment, a composite time-to-event endpoint, a Full Analysis Set, Cox proportional-hazards regression, stratification, covariate adjustment, hazard ratios, confidence intervals, and superiority testing.

ConceptHow it appears in DAPA-CKD
RandomizationPatients were randomized to dapagliflozin or placebo.
Parallel-group designThe trial used two parallel treatment arms.
BlindingThe registry reports quadruple masking.
Time-to-event endpointThe primary endpoint was time to first occurrence of a composite event.
Composite endpointThe primary endpoint combined four qualifying components.
Cox regressionThe primary and secondary analyses used Cox regression.
Hazard ratioTreatment effects were reported as HRs.
Confidence intervalAll four statistical analyses registry-reported 95% CIs.
Covariate adjustmentBaseline eGFR was included in the reported models.
Stratified analysisModels were stratified by Type 2 Diabetes and urine albumin creatinine ratio.
Full Analysis SetAll randomized patients were included in the reported efficacy analysis population.
Superiority testingThe registered hypothesis type was superiority.

17. Important Limitations and Interpretation Issues

18. What the Primary Hazard Ratio Does — and Does Not — Mean

Effect estimate

The primary HR of 0.61 means that the estimated hazard of the first qualifying primary composite event was 0.61 times the placebo hazard under the reported Cox model. The corresponding relative interpretation is an estimated 39% lower hazard.

What it does not mean

It does not mean that exactly 39% fewer patients experienced an event, that each individual patient had a 39% lower probability of an event, or that every component of the composite was reduced by 39%.

Precision

The 95% CI of 0.51–0.72 expresses uncertainty around the estimated HR. It should be read together with the endpoint definition, analysis population, stratification, and baseline-eGFR adjustment.

P-value

The P-value of < 0.0001 addresses statistical evidence against the null hypothesis in the reported superiority analysis. It is not a measure of the magnitude or clinical importance of the HR.

19. Clinical Interpretation vs Statistical Interpretation

Statistical interpretation

The primary Cox model estimated an HR of 0.61 with a 95% CI of 0.51–0.72 and P < 0.0001. The model was stratified by Type 2 Diabetes and urine albumin creatinine ratio and adjusted for baseline eGFR.

Clinical interpretation

The primary endpoint represents a broad composite of renal and mortality outcomes. Its HR describes the time-to-first-event comparison, while the separate secondary analyses provide distinct perspectives on renal, cardiovascular/heart-failure, and all-cause mortality outcomes.

Keeping these two levels separate is important. Statistical significance does not by itself determine clinical importance, and a hazard ratio cannot be interpreted properly without knowing precisely which event definition generated it.

20. Comparing Relative Effects Across the Reported Endpoints

The reported HRs all fall below 1, but each belongs to a different endpoint. A useful way to read them is descriptively rather than treating their numerical ordering as a ranking of clinical importance.

EndpointHR95% CIModel adjustment
Primary composite0.610.51–0.72Stratified by Type 2 Diabetes and urine albumin creatinine ratio; adjusted for baseline eGFR
Renal composite0.560.45–0.68Stratified by Type 2 Diabetes and urine albumin creatinine ratio; adjusted for baseline eGFR
CV death or hospitalization for heart failure0.710.55–0.92Stratified by Type 2 Diabetes and urine albumin creatinine ratio; adjusted for baseline eGFR
Death from any cause0.690.53–0.88Stratified by Type 2 Diabetes and urine albumin creatinine ratio; adjusted for baseline eGFR

Notice that the confidence intervals overlap substantially across the reported estimates. Numerical differences between HRs should not be interpreted as evidence that one endpoint has a statistically different treatment effect from another. Formal comparisons would require an analysis specifically designed to test such differences.

21. A Worked Reading of the Primary Result

Step 1 · Identify the endpoint

The outcome is time to the first occurrence of ≥50% sustained decline in eGFR, reaching ESRD, CV death, or renal death, over up to 38.2 months.

Step 2 · Identify the analysis

The Full Analysis Set was used. Patients were analyzed according to randomized treatment assignment. A stratified Cox proportional-hazards model was used, with stratification by Type 2 Diabetes and urine albumin creatinine ratio and adjustment for baseline eGFR.

Step 3 · Read the effect estimate

The HR was 0.61. Under the model, the estimated hazard under dapagliflozin was 0.61 times the hazard under placebo.

Step 4 · Read the uncertainty

The 95% CI was 0.51–0.72. This communicates uncertainty around the estimated HR.

Step 5 · Read the P-value

The P-value was < 0.0001. This provides evidence against the null hypothesis under the reported superiority analysis, but it does not quantify the size of the treatment effect.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Statistical Calculators

24. Sources

Continue through the Clinical Biostats statistical library

Explore the statistical concepts behind randomized trials, time-to-event endpoints, Cox regression, hazard ratios, confidence intervals, and clinical-trial analysis.

25. Record Summary

DAPA-CKD provides a clear example of how a randomized clinical trial can evaluate a composite renal and cardiovascular time-to-event outcome using Cox proportional-hazards regression. The registry analysis used the Full Analysis Set, analyzed patients according to randomized treatment assignment, stratified the Cox model by Type 2 Diabetes and urine albumin creatinine ratio, and adjusted for baseline eGFR.

The primary analysis reported an HR of 0.61 with a 95% CI of 0.51–0.72 and P < 0.0001 for time to the first occurrence of ≥50% sustained decline in eGFR, reaching ESRD, CV death, or renal death over up to 38.2 months. Three secondary analyses also reported HRs below 1: 0.56 for the renal composite, 0.71 for CV death or hospitalization for heart failure, and 0.69 for death from any cause.

The statistical lesson is broader than any individual number. Correct interpretation requires identifying the exact endpoint, preserving the randomized analysis population, understanding the Cox model and its assumptions, distinguishing hazard ratios from absolute risks, reading confidence intervals as measures of precision, and keeping P-values separate from effect size.

Clinical Biostats methodology: A trial-results page should not merely repeat a registry result. The goal is to reconstruct the statistical story of the trial in a standardized format while clearly separating reported evidence from educational interpretation and avoiding conclusions that are not supported by the available data.