A Study to Evaluate the Effect of Dapagliflozin on Renal Outcomes and Cardiovascular Mortality in Patients With Chronic Kidney Disease: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Evaluate the Effect of Dapagliflozin on Renal Outcomes and Cardiovascular Mortality in Patients With Chronic Kidney Disease (NCT03036150), a randomized phase 3 study. The primary outcome is 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..

2. Clinical Question

The question is clinically framed around kidney outcomes and cardiovascular mortality, but the statistical endpoint combines several types of event. A participant could meet the composite through any of its listed components. The overall primary result therefore pertains to the combined endpoint; it does not by itself establish that the association was the same for each component considered separately. The registry record gives no component-specific primary analyses.

The trial was randomized and quadruple masked, with a parallel-group model. Those design features describe how treatment assignment and masking were organized in general terms. Which individuals or roles were included in the quadruple masking is not reported in the registry record. The posted primary result concerns assignment to treatment, not a comparison based on treatment actually received.

3. Trial Design

This was a Phase 3, randomized, quadruple-masked, parallel-group trial sponsored by AstraZeneca. The registry lists enrollment as 4304 and describes allocation in a 1:1 ratio between dapagliflozin and matching placebo. The study is marked completed. Its recorded start date is 2017-02-02, and its primary completion date is 2020-06-12.

The trial’s central comparison was between participants assigned to dapagliflozin and participants assigned to placebo. For the primary endpoint, the registry says that the Full Analysis Set included all patients randomized to study treatment, with patients analyzed according to randomized assignment. That is an intention-to-treat-style assignment-based description; the registry’s specific label is Full Analysis Set. Details on adherence, treatment discontinuation, crossover, and handling of participants after discontinuation are not reported in the registry record.

The primary outcome is a time-to-first-event composite, and the registry describes a proportional hazards Cox regression model that takes time to event into account. The reported primary analysis is a hazard ratio with a 95% confidence interval and a p-value. The endpoint timeframe is “Up to 38.2 months.” This is the listed observation timeframe, not a statement that every participant was followed for the same duration.

Four outcome measures and four analyses are reported in the record. These comprise the primary composite and three secondary outcomes: a kidney composite without cardiovascular death, a composite of cardiovascular death or hospitalization for heart failure, and death from any cause. Their posted analyses also use Cox regression and report hazard ratios. Details of the randomization sequence, allocation concealment, masking roles, trial monitoring, interim analyses, and prespecified statistical plan are not reported in the registry record.

The record includes adverse-event counts for the Dapa 10 mg and placebo groups. The listed safety denominators are 2149 per group, whereas the primary efficacy arms are listed as 2152 per group. The reason for the difference between these recorded counts is not reported in the registry record, so the efficacy and safety denominators should be read as the distinct denominators stated for their respective results.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: For the primary endpoint the Full Analysis Set was used. This included all patients who were randomized to study treatment. Patients were analyzed according to their randomized treatment assignment.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as 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., measured over Up to 38.2 months.

The primary endpoint is the time to the first occurrence of any component in a composite: a sustained eGFR decline of at least 50%, reaching end stage renal disease, cardiovascular death, or renal death. The registry defines end stage renal disease as sustained eGFR below 15 mL/min/1.73m2, chronic dialysis treatment, or receiving a renal transplant. The stated timeframe is up to 38.2 months.

A time-to-first-event endpoint combines information about whether an event occurred with information about when it occurred. The Cox regression analysis is described in the registry as taking time to event into account. In general, this approach is useful when follow-up time can differ among participants or when some participants have not had the event by the end of their observed follow-up. The registry does not give participant-level follow-up distributions or details about censoring.

Composite endpoints can capture several clinically consequential outcomes in a single prespecified outcome measure. In general, this can provide a broader summary than any component alone, but it also means the combined estimate does not reveal which components contribute most to the result. The registry record does not provide component-specific results for the primary composite. Therefore, the primary hazard ratio should be interpreted as applying to the composite event definition, not as a separate estimate for each kidney or cardiovascular component.

The registry reports both numbers of participants with an event and a Cox hazard ratio. The counts offer a straightforward description of observed event occurrence by assigned group; the hazard ratio is the sponsor-reported time-to-event comparison. These measures answer related but not identical statistical questions.

7. Results Posted in the Registry

• Dapagliflozin: 9.2% (197/2152) versus 14.5% (312/2152) on control; risk difference -5.3 percentage points (95% CI -7.3 to -3.4), risk ratio 0.63 (0.53–0.75); unadjusted two-proportion z-test p = <0.001.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Time to the First Occurrence of Any of the Components of the Composite: ≥50% Sustained Decline in eGFR or Reaching ESRD or Renal Death.

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Dapagliflozin: 6.6% (142/2152) versus 11.3% (243/2152) on control; risk difference -4.7 percentage points (95% CI -6.4 to -3.0), risk ratio 0.58 (0.48–0.71); unadjusted two-proportion z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically lower for Dapagliflozin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Time to the First Occurrence of Either of the Components of the Composite: CV Death or Hospitalization for Heart Failure.

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Dapagliflozin: 4.6% (100/2152) versus 6.4% (138/2152) on control; risk difference -1.8 percentage points (95% CI -3.1 to -0.4), risk ratio 0.72 (0.56–0.93); unadjusted two-proportion z-test p = 0.011.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically lower for Dapagliflozin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Time to Death From Any Cause.

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Dapagliflozin: 4.7% (101/2152) versus 6.8% (146/2152) on control; risk difference -2.1 percentage points (95% CI -3.5 to -0.7), risk ratio 0.69 (0.54–0.89); unadjusted two-proportion z-test p = 0.003.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically lower for Dapagliflozin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Across the three posted secondary outcomes, the sponsor-reported hazard ratios are below 1, and their 95% confidence intervals are below 1. The outcomes cover a kidney composite, cardiovascular death or hospitalization for heart failure, and death from any cause. Taken together, this is a consistent directional pattern across the secondary results rather than a pattern in which some reported hazard ratios favor one group and others the opposite group.

The secondary outcomes relate to the primary result because they examine narrower or related clinical outcomes. The kidney composite excludes cardiovascular death from the listed components; the heart-failure composite combines cardiovascular death and hospitalization for heart failure; and the all-cause mortality outcome concerns death from any cause. Their direction is compatible with the primary composite result, but they do not show which primary composite component drove the overall association. The registry does not provide component-level analyses of the primary endpoint.

The secondary p-values are posted as analysis results, but multiplicity handling is not reported in the registry record. When several outcomes are tested, the chance of observing at least one apparently notable result under a collection of null hypotheses can increase. In general, prespecified testing sequences or adjustment procedures can address this issue, but whether either was used is not reported. For that reason, unadjusted secondary results generally carry less confirmatory weight than a prespecified primary outcome analyzed under a clear testing plan.

The registry marks the analyses as superiority and reports Cox models for these time-to-event outcomes. The posted estimates should still be read with their confidence intervals and with attention to each outcome definition. A secondary result does not independently establish a treatment effect across every related endpoint, and a consistent direction does not remove uncertainty about the analysis plan, outcome selection, or multiplicity. Those details are not available in the registry record.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Time to the First Occurrence of Any of the Components of the Composite: ≥50% Sustained Decline in eGFR or Reaching ESRD or Renal Death.
• Time to the First Occurrence of Either of the Components of the Composite: CV Death or Hospitalization for Heart Failure.
• Time to Death From Any Cause.

9. Statistical Interpretation

10. Statistical Methods Explained

What does the event count describe?

The primary outcome lists the number of participants recorded with at least one event in the composite. The dapagliflozin arm has 197 events among 2152 listed participants, and the placebo arm has 312 among 2152. These counts are useful descriptive summaries, but they do not encode when events occurred or how long each participant was observed. The endpoint is explicitly time-to-event, so the count comparison is not the complete analysis reported by the sponsor.

The unadjusted recalculation uses the posted group counts to estimate a risk difference. The reported estimate is -0.053, with interval limits of -0.073 and -0.034. A risk difference is an absolute-scale comparison between the proportions with an event in the two groups. Its negative direction reflects fewer recorded events in the dapagliflozin group than in the placebo group. The accompanying p-value is < 0.001. This recalculation is unadjusted and does not account for event timing through a survival model.

Why report a Cox hazard ratio?

The registry identifies the sponsor’s method as Cox regression, reports a hazard ratio of 0.61, and says the proportional hazards Cox regression model takes time to event into account. In general, a hazard is an event rate among those who have not yet experienced the event, and a hazard ratio compares those rates between groups. A value below 1 indicates a lower estimated hazard in the dapagliflozin group relative to placebo under the fitted model. It is not the same as a ratio of cumulative risks, and it should not be read as a direct percentage change in the probability of an event.

Time-to-event methods can use information about event timing and accommodate participants whose follow-up ends before an event is recorded, through censoring. In general, this relies on assumptions about the model and censoring process. The registry does not report diagnostic checks, the number censored, details of censoring, or further model specification. The proportional hazards assumption is part of the Cox model’s interpretation; whether and how it was evaluated is not reported in the registry record.

How should the confidence interval be read?

The sponsor’s primary hazard ratio is 0.61, with a 95% confidence interval from 0.51 to 0.72. In general, a confidence interval describes the range of values reasonably compatible with the data under the specified statistical procedure and its assumptions. It conveys precision as well as direction: a narrower interval indicates less statistical uncertainty than a wider interval on the same scale. The interval does not mean there is a 95% probability that the fixed underlying effect lies within these particular limits, and it does not account for every possible source of uncertainty.

The reported p-value is < 0.001. In general, a p-value summarizes how incompatible the observed data are with a specified null hypothesis under the analysis assumptions; it is not the probability that the null hypothesis is true. The registry identifies the analysis as superiority. The precise testing hierarchy and any multiplicity adjustment are not reported in the registry record, so the p-value should be read as the posted result rather than as proof that all possible sources of bias or uncertainty have been resolved.

Why can the risk difference and hazard ratio differ in meaning?

The recalculated risk difference compares observed event proportions. The sponsor’s hazard ratio uses event-time information through Cox regression. Follow-up duration, censoring, and the timing of events can make these estimates differ in magnitude and interpretation even when they point in the same direction. The risk difference is expressed as an absolute difference in event proportions; the hazard ratio is a relative comparison of modeled event hazards over time. Neither measure can be converted into the other from the posted aggregate results without additional information.

The risk-difference confidence interval and p-value come from an unadjusted recalculation of posted summary counts. The Cox estimate and its interval and p-value are the sponsor-reported analysis. They are not competing estimates of precisely the same quantity. The registry record does not report an adjusted risk-difference analysis or a model that directly links these measures.

What does randomization contribute?

Random assignment is intended, in general, to balance measured and unmeasured prognostic factors across groups on average. The registry says participants were randomized 1:1 and that the primary Full Analysis Set includes all randomized patients analyzed according to assigned treatment. This assignment-based approach preserves the randomized comparison even if treatment received later differs from assignment, although the registry does not report adherence or switching details. Randomization does not ensure that every baseline characteristic is exactly balanced, and the record does not provide baseline tables within the registry entries described for this analysis.

Masking is listed as quadruple, but the registry does not identify which people or roles were masked. Masking can generally reduce the influence of expectations on care, assessment, or reporting, depending on how it is implemented. Without the roles and procedures, the exact protections in this trial cannot be assessed from the registry record.

What is not described by the reported method?

The record reports the method label, the hazard ratio, interval, p-value, outcome definition, and primary analysis population. It does not report the original statistical analysis plan, sample-size justification, interim-analysis plan, covariate adjustment details, model diagnostics, missing-data strategy, multiplicity approach, or whether the analysis was amended. These details matter when evaluating the full inferential context. Their absence from the registry record should not be filled by assumptions about standard practice.

11. Confidence Intervals and Precision

The primary Cox estimate is a hazard ratio of 0.61, with a 95% confidence interval from 0.51 to 0.72. The interval is entirely below 1, so the posted analysis is consistent with a lower modeled event hazard in the dapagliflozin group than in the placebo group. The interval also communicates uncertainty around the point estimate: the data are compatible with a range of hazard ratios, not only the estimate of 0.61. The reported p-value is < 0.001.

The interval for the unadjusted risk difference runs from -0.073 to -0.034 around an estimate of -0.053. Its limits are both below zero, consistent with fewer recorded events in the dapagliflozin group on this unadjusted absolute scale. The estimate and interval are recalculated from the posted event counts, rather than taken from the sponsor’s time-to-event model. They do not account for event timing or follow-up differences as the Cox model does.

In general, an interval is more informative than a point estimate alone because it shows a range of values compatible with the data under the procedure used. It is not a guarantee that the true effect falls within the interval, nor does it represent uncertainty about every issue in trial conduct, measurement, or generalizability. The interval’s interpretation relies on the analysis assumptions. Details needed to assess all those assumptions are not reported in the registry record.

The number of observed events contributes to precision, but the event counts alone do not determine it: the time-to-event analysis also uses event timing and censoring information. The registry does not post participant-level data or a detailed follow-up distribution, so precision cannot be independently reconstructed for the Cox analysis from the count summaries. The sponsor’s reported interval is the relevant posted interval for the primary hazard ratio.

Precision is also distinct from clinical importance. Statistical intervals describe uncertainty in an estimated quantity; they do not decide whether an effect is important to patients or applicable to every patient with chronic kidney disease. That broader judgment requires context not established by the numerical interval alone.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning illustrations use the observed effect as a starting point and consider planning for 100% of the observed effect, 75% of the observed effect, and 50% of the observed effect. The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The original trial’s own sample-size justification and design assumptions are not reported in the registry record.

The unit needs careful reading. The planning entries are labeled as participants per arm, while the accompanying method note says time-to-event tables give total numbers of events using the Schoenfeld method, and that other tables give participants per arm. Because the outcomes considered are time-to-event outcomes, these descriptions do not unambiguously identify the unit for the displayed planning figures. The registry record does not resolve the conflict; readers should not silently treat the entries as either per-arm participant targets or total event targets without checking the display’s labeling and method note.

The 80% power column represents the planning size intended to give an 80% chance of detecting the specified effect under the listed assumptions, if that effect and the assumptions were true. The 90% power column represents a higher planned chance of detection under those same conditions. These are prospective planning concepts, not probabilities that the trial’s result is correct. The sample-size illustrations are not observed power.

The illustrations also show why smaller effects require larger planned samples or event totals: a weaker signal is harder to distinguish from random variation at the same significance level and target power. The estimates are based on fractions of the observed effect, which may not be the effect a future study should expect. They are planning scenarios, not evidence that the original study was designed around these specific assumptions.

13. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.145; treatment proportion p₁ = 0.092; average p̄ = 0.118.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.4567 = 0.8951.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.4551 = 0.3830.
4. n = (0.8951 + 0.3830)² / (-0.0534)² = 572.0, rounded up to 573 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
p0 <- 0.144981412639405                          # control proportion
p1 <- 0.091542750929368                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(3036150)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 573 (80% power), 766 (90% power)
 75% of observed effect: n per arm = 1,068 (80% power), 1,429 (90% power)
 50% of observed effect: n per arm = 2,512 (80% power), 3,363 (90% power)
power.prop.test: 572 
Simulated power at n = 573 per arm: 0.802 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

14. Analyses Reported by the Sponsor

The sponsor’s primary analysis is a Cox regression with a hazard ratio of 0.61, a 95% confidence interval from 0.51 to 0.72, and p < 0.001. This is the registry’s reported time-to-event result and takes precedence as the sponsor’s own analysis when describing the trial’s primary inferential finding. The registry does not flag a discrepancy between the posted event counts and this analysis.

The recalculated result uses the event counts and arm sizes to estimate an unadjusted risk difference of -0.053, with a 95% confidence interval from -0.073 to -0.034 and p < 0.001. It answers a more limited question: how the observed proportions with an event compare, without modeling time to event. It is useful as a transparent summary of the posted counts, but it is not an independent replication of the Cox analysis and should not replace it.

Agreement in direction between the risk difference and hazard ratio is descriptively reassuring, but the estimates remain different measures. The risk difference is absolute and based on event proportions; the hazard ratio is a relative time-to-event measure. Their numerical values should not be compared as if they shared a scale. The sponsor’s model uses information not recoverable from the aggregate counts alone, including event timing. The registry does not provide enough participant-level detail to reproduce that model.

For secondary endpoints, the same distinction applies: posted counts allow unadjusted proportion comparisons, while reported Cox analyses summarize time-to-event contrasts. The Cox results are the sponsor’s inferential analyses; count-based calculations are supplementary descriptions. Neither eliminates uncertainty about unreported model details, missingness, or multiplicity.

15. Multiplicity, Missing Data and Analysis Populations

The registry lists a primary composite and three secondary outcomes, with four outcome measures and four analyses reported. It does not report whether the outcomes were organized into a formal testing hierarchy, whether any secondary hypotheses were tested only after a primary result, or whether multiplicity adjustments were used. The registry record therefore does not establish that the posted secondary p-values control the overall chance of false-positive findings across outcomes.

In general, testing multiple outcomes creates more opportunities for chance findings than testing a single outcome. Methods such as a prespecified hierarchy or adjusted significance thresholds can manage that issue, but neither should be assumed in the absence of documentation. The p-values remain descriptions of the reported analyses; their interpretation as confirmatory evidence depends on the prespecified statistical plan, which is not reported in the registry record.

Missing-data and censoring details are also not reported. This is important for a time-to-event analysis because the Cox model uses event times and may include participants whose event was not observed before follow-up ended. General interpretation of such analyses depends on how follow-up ends and on assumptions about censoring. The registry does not provide the number or reasons for censoring, a missing-data strategy, or sensitivity analyses addressing incomplete outcome information.

The primary population is described as the Full Analysis Set, including all randomized patients analyzed by assigned treatment. The registry does not give a corresponding population description for each secondary analysis beyond the posted group counts and analyses. It also does not report exclusions from the primary set, the handling of unavailable endpoint data, or whether any analyses differed from the planned population.

Interim analyses, stopping rules, sample-size justification, analysis-plan amendments, and prespecified subgroup analyses are not reported in the registry record. Their absence does not show that such procedures were or were not used; it means the record does not allow them to be assessed. These gaps constrain interpretation of the nominal p-values and the completeness of the analysis history.

16. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 2% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pneumonia (3.3%), cardiac failure (2.7%), acute kidney injury (2.4%), acute myocardial infarction (2.1%), end stage renal disease (1.7%).

The safety record lists serious adverse events for Dapa 10 mg and placebo, with 2149 participants at risk in each group. Serious adverse events affected 633 in the Dapa 10 mg group and 729 in the placebo group. The record also lists other adverse events affecting 479 and 500, respectively, and deaths of 107 and 159. These are posted counts; the registry record does not give a detailed account of event duration, severity grading beyond the serious classification, attribution, or how events were handled in the analysis.

The listed serious adverse-event terms include pneumonia at 3.3%, cardiac failure at 2.7%, acute kidney injury at 2.4%, acute myocardial infarction at 2.1%, end stage renal disease at 1.7%, and chronic kidney disease at 1.4%. These are term-specific percentages as listed in the registry. The record does not state in these results whether participants could contribute to more than one term, so the terms should not be added together or treated as mutually exclusive categories.

Safety counts are descriptive and do not, by themselves, establish that treatment caused or prevented an event. The safety denominator differs from the 2152 participants listed in each efficacy arm; the reason is not reported in the registry record. Comparisons of counts should therefore preserve the denominators attached to each result and should not be interpreted as adjusted estimates or as complete measures of benefit and harm.

17. What the Result Does, and Does Not, Mean

The posted primary result shows a lower estimated hazard of the composite outcome in the dapagliflozin-assigned group than in the placebo group under the sponsor’s Cox model. The reported hazard ratio is 0.61, with a 95% confidence interval from 0.51 to 0.72 and p < 0.001. The unadjusted event-count comparison points in the same direction, with a risk difference of -0.053. These are findings about the trial groups and the defined composite over the recorded follow-up.

The result does not mean that every participant assigned dapagliflozin experienced a lower risk, nor does it predict an individual participant’s outcome. A group-level estimate describes an average contrast under the study’s design and analysis. It also does not show that every component of the composite changed to the same degree. Component-specific primary results are not reported in the registry record.

A hazard ratio is not an absolute probability, and the reported estimate should not be translated into a percentage-point change. The unadjusted risk difference provides an absolute-scale summary of observed event proportions, but it does not incorporate event timing as the Cox analysis does. Neither measure alone describes long-term outcomes beyond the listed timeframe of up to 38.2 months.

The findings do not resolve all questions about safety, generalizability, or mechanism. The record gives safety counts but not full event-level detail, and it does not provide enough information to assess how the study population compares with every clinical population. The trial’s eligibility criteria beyond chronic kidney disease, background care, and relevant subgroup results are not reported in the registry record.

Statistical significance is not equivalent to clinical importance or certainty. The confidence interval quantifies uncertainty under the model and its assumptions, while omitted details such as missing-data handling, multiplicity, and model diagnostics limit what can be concluded from the registry record alone. The evidence should be understood as a randomized group comparison with a favorable direction in the posted primary analysis, not as a guarantee, a complete account of every endpoint, or medical advice.

18. Why This Trial Matters Statistically

This trial is statistically informative because it combines randomized assignment with a time-to-event primary endpoint and reports both event counts and a Cox model estimate. Randomization supports a group comparison based on assigned treatment, while the time-to-event analysis uses more information than a simple end-of-follow-up event proportion. The Full Analysis Set description also makes clear that the primary analysis follows randomized assignment.

The primary endpoint’s composite structure illustrates a common trade-off in clinical-trial design. Combining kidney outcomes and cardiovascular or renal death can capture a broader set of events, but the combined estimate cannot reveal how much each component contributes. The absence of component-specific primary results matters when interpreting what the composite represents.

The contrast between the hazard ratio and the unadjusted risk difference is a practical lesson in estimands. The Cox analysis compares modeled event hazards over time; the risk difference compares observed proportions. Both may be useful, but they answer different questions and rely on different information. Reporting them together encourages interpretation that separates relative time-to-event evidence from absolute event-count summaries.

The secondary outcomes further illustrate why a consistent pattern is not the same as a complete multiplicity assessment. Their directions align with the primary result, but the registry does not report whether testing was adjusted or organized in a hierarchy. Statistical interpretation requires attention not only to estimates and p-values, but also to the number of outcomes and the plan governing their analysis.

Finally, the record distinguishes efficacy and safety populations through different listed denominators and supplies event counts for both. This highlights the importance of checking the denominator attached to each analysis rather than assuming all results use identical populations. Missing documentation about the reasons for the denominator difference and other analysis details remains relevant to a complete appraisal.

19. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The registry record does not report the full eligibility criteria, baseline characteristics, or details of background care. Those omissions limit assessment of how closely the enrolled population resembles other groups of patients with chronic kidney disease. The trial’s named condition alone does not define the full clinical profile represented by its participants.

The primary result is a composite, and component-specific primary estimates are not reported. This limits interpretation of whether the overall result reflects a similar pattern across kidney decline, end stage renal disease, cardiovascular death, and renal death, or is concentrated in particular components. The record also does not provide subgroup results that might describe variation across participant characteristics.

Although the registry identifies Cox regression, it does not report the model’s detailed specification, diagnostics, or assessment of the proportional hazards assumption. Censoring counts and reasons, missing-data procedures, sensitivity analyses, and endpoint adjudication details are not reported in the registry record. These are relevant to evaluating a time-to-event estimate but cannot be inferred from the hazard ratio alone.

Multiplicity methods, the original sample-size justification, interim-analysis procedures, and the prespecified analysis plan are not reported. Accordingly, the secondary p-values cannot be confidently characterized as adjusted or confirmatory based on this record. The sample-size planning illustrations are not the trial’s documented design justification, and the displayed unit descriptions are not fully consistent for time-to-event planning.

The safety results list group counts and selected serious adverse-event terms, but do not provide comprehensive details about event attribution, timing, repeated events, or analysis methods. The safety denominator is 2149 per group, compared with 2152 per arm for efficacy, and the reason for that difference is not reported. These limitations do not negate the posted results, but they constrain a complete assessment of their context and robustness.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A Study to Evaluate the Effect of Dapagliflozin on Renal Outcomes and Cardiovascular Mortality in Patients With Chronic Kidney Disease — NCT03036150.
• Registry record last updated 2021-07-07; status last verified 2021-06.
• Results first posted 2021-07-07.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

23. Record Summary

The record describes a completed Phase 3 randomized, quadruple-masked, parallel-group comparison of dapagliflozin and matching placebo in patients with chronic kidney disease. Enrollment is listed as 4304, with 1:1 allocation. The primary endpoint is a time-to-first-event composite of sustained eGFR decline of at least 50%, end stage renal disease, cardiovascular death, or renal death.

The primary Cox analysis reports a hazard ratio of 0.61, with a 95% confidence interval from 0.51 to 0.72 and p < 0.001. The unadjusted recalculation from event counts gives a risk difference of -0.053, with a 95% confidence interval from -0.073 to -0.034. Three secondary time-to-event outcomes are also posted, with hazard ratios below 1 and confidence intervals below 1.

The record describes the primary Full Analysis Set as all randomized patients analyzed according to assigned treatment. Multiplicity procedures, missing-data handling, detailed model diagnostics, and the original sample-size justification are not reported in the registry record. Safety counts and selected serious adverse-event terms are listed separately.