← Clinical Trials
Type 2 Diabetes Chronic Kidney Disease Phase 3 Completed NCT03819153

FLOW: Complete Statistical Analysis of Semaglutide in Type 2 Diabetes and Chronic Kidney Disease

An independent statistical review of the randomized phase 3 FLOW trial evaluating semaglutide versus placebo in people with type 2 diabetes and chronic kidney disease, with emphasis on the prespecified composite renal endpoint and its Cox proportional-hazards analysis.

Phase 3  ·  Randomized  ·  Parallel  ·  Quadruple-masked  ·  Enrollment 3533
Scope of this record

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

FLOW was a completed phase 3 randomized, parallel-group trial evaluating semaglutide versus placebo in people with type 2 diabetes and chronic kidney disease. The registry reports a total enrollment of 3533 participants and a primary time-to-event analysis based on a Cox proportional-hazards model.

3533
Enrolled
Total trial enrollment
2
Arms
Semaglutide vs placebo
0.76
Primary HR
95% CI 0.66–0.88
0.0001
P-value
Primary analysis
FeatureFLOW
Trial nameFLOW
NCT identifierNCT03819153
PhasePhase 3
StatusCompleted
Therapeutic areaDiabetes
ConditionDiabetes Mellitus, Type 2
AllocationRandomized
Design modelParallel
MaskingQuadruple
Primary purposeTreatment
Enrollment3533
Lead sponsorNovo Nordisk A/S
Sponsor typeIndustry
Trial datesStart: 2019-06-17; primary completion: 2024-01-09
Primary endpointComposite renal and cardiovascular time-to-event endpoint
Primary analysisCox proportional-hazards model

2. Clinical Question

The primary question was whether semaglutide, compared with placebo, affected the time from randomization to the first occurrence of the registered composite endpoint in people with type 2 diabetes and chronic kidney disease.

Population

People with type 2 diabetes and chronic kidney disease enrolled in the FLOW trial.

Intervention

Semaglutide.

Comparator

Placebo (semaglutide).

Primary question

Does semaglutide change the time to first occurrence of the registered composite renal and cardiovascular event compared with placebo?

3. Trial Design

01
Randomize3533 participants
02
Parallel armsSemaglutide vs placebo
03
Quadruple maskMasked trial design
04
Follow-upFrom Week 0 up to Week 234
05
Event analysisCox model
Allocation
Randomized allocation to two parallel treatment groups.
Masking
Quadruple masking was specified for the trial.
Primary purpose
Treatment.
Study status
Completed, with primary completion recorded as 2024-01-09.
ARM 1

Semaglutide

  • Intervention: semaglutide
  • Randomized trial arm
  • Included in the primary efficacy comparison
ARM 2

Placebo

  • Intervention: placebo (semaglutide)
  • Randomized trial arm
  • Comparator for the primary efficacy comparison

4. Primary Endpoint

The registered primary endpoint was a composite time-to-event outcome. The registry defines it as the number of participants from time of randomization to the first occurrence of any of the following components:

Registered componentDescription
Persistent eGFR reductionOnset of persistent ≥50% reduction in eGFR (CKD-EPI).
Very low eGFROnset of persistent eGFR (CKD-EPI) <15 mL/min/1.73m2.
Renal replacement therapyInitiation of chronic renal replacement therapy.
Renal deathRenal death.
Cardiovascular deathCV death.
Registered time frame: From Week 0 up to Week 234.

The registry classifies the primary endpoint as Binary in its registered endpoint information, while the posted statistical analysis treats the endpoint as a time-to-event outcome. That distinction matters statistically: the formal analysis uses not only whether a participant experienced an event, but also the time from randomization to the first qualifying event and the censoring information available during follow-up.

5. Analysis Population and Treatment Comparison

The posted primary analysis used the Full Analysis Set (FAS). The registry defines this population as including all unique randomized participants who were grouped according to the treatment assigned at randomization.

Analysis featurePosted specification
Analysis populationFAS: all unique randomized participants, grouped according to treatment assigned at randomization.
Groups comparedSemaglutide vs Placebo.
Outcome unitParticipants.
Endpoint type for analysisTime-to-event.
Effect measureHazard Ratio (HR).
Hypothesis typeSuperiority.

This analysis population is important because treatment assignment at randomization is retained for the efficacy comparison. That preserves the randomized comparison rather than redefining groups according to treatment received after randomization.

6. Statistical Methodology

Cox proportional-hazards model

The primary endpoint was analyzed using a Cox proportional-hazards model. The registry reports the model as a regression analysis with treatment entered as a categorical fixed factor.

The model estimates a relative hazard between the semaglutide and placebo groups while accounting for the timing of the first qualifying event. Participants who do not experience the primary event during the observed follow-up can contribute information until the point at which their observation is censored.

Conceptual hazard-ratio interpretation
HR = estimated hazard in semaglutide group ÷ estimated hazard in placebo group

An HR below 1 indicates a lower estimated instantaneous event rate in the semaglutide group under the fitted model. The HR is a relative time-to-event measure; it is not an absolute risk difference and does not mean that every participant experiences the same proportional change.

Stratification

The primary Cox analysis was stratified by use of a sodium glucose cotransporter-2 (SGLT-2) inhibitor at baseline, categorized as yes or no.

Stratification factorCategories
SGLT-2 inhibitor use at baselineYes / No

Stratification allows the treatment comparison to account for the prespecified baseline SGLT-2 inhibitor-use categories without requiring the stratification factor to be represented as an ordinary coefficient in the Cox model. It is particularly relevant here because the registry explicitly identifies this factor as part of the primary analysis specification.

Kaplan-Meier estimation and time-to-event analysis

A time-to-event endpoint is naturally represented using survival-analysis methods. Kaplan-Meier estimation describes the event-free probability over time, while the Cox model summarizes the relative event hazard between randomized treatment groups.

Kaplan-Meier concept
S(t) = probability of remaining free of the event through time t

The key advantage of a time-to-event framework is that participants who have not yet experienced the event can remain informative through their observed follow-up, rather than being treated simply as if they had all contributed the same duration of observation.

7. Primary Result

The registry reports one formal statistical analysis for the primary endpoint. The analysis compares semaglutide with placebo using a Cox proportional-hazards model in the FAS, stratified by baseline SGLT-2 inhibitor use.

Primary composite endpoint

HR 0.76

95% CI: 0.66–0.88   ·   P = 0.0001

Two-sided 95% confidence interval  ·  Superiority analysis

Primary endpointSemaglutide vs placebo
Registered time frameFrom Week 0 up to Week 234
Analysis populationFAS
ModelStratified Cox proportional-hazards model
StratificationBaseline SGLT-2 inhibitor use: yes/no
Effect measureHazard ratio
Estimate0.76
95% CI0.66–0.88
P-value0.0001
HypothesisSuperiority
Clinical Biostats interpretation

The estimated hazard ratio of 0.76 means that, under the fitted Cox model, the estimated instantaneous rate of experiencing the first qualifying component of the composite endpoint was approximately 24% lower in the semaglutide group than in the placebo group. That 24% figure is a direct interpretation of the relative hazard, not a statement that 24% of participants avoided an event.

The HR does not mean that every participant had a 24% reduction in personal risk, nor does it provide the absolute probability that an individual participant will experience the composite endpoint. It is a model-based relative comparison over the analyzed time-to-event data.

The 95% CI of 0.66–0.88 describes the statistical uncertainty around the estimated hazard ratio under the analysis framework. It is not the range of individual treatment effects and does not imply that individual participants' hazards must fall within that interval.

The P-value of 0.0001 quantifies evidence against the null hypothesis under the specified statistical testing framework. It does not measure the size, clinical importance, or certainty of the treatment effect. The effect size is described by the hazard ratio and its confidence interval.

Because the analysis uses a Cox proportional-hazards model, the hazard-ratio interpretation relies on the model's proportional-hazards framework. The ClinicalTrials.gov record does not provide a separate assessment of that assumption, so the HR should be understood as the reported model-based summary rather than as a complete description of how hazards behaved at every point in follow-up.

8. Interim Analysis and Alpha Spending

The primary analysis notes that the nominal significance level was updated based on the available number of events for analysis. The registry reports a nominal significance level of 0.01612, using the Lan-DeMets alpha spending function.

Why alpha spending matters

When accumulating trial data are examined during an interim analysis, repeatedly testing for efficacy without adjustment can increase the overall probability of a false-positive conclusion. Alpha spending provides a framework for allocating the permitted type I error across information accumulated over time.

What the 0.01612 means

The registry reports 0.01612 as the updated nominal significance level based on the available number of events. It is therefore part of the trial's error-control framework rather than an alternative effect-size measure.

The reported primary P-value of 0.0001 is below the stated nominal significance level of 0.01612. The statistical comparison therefore provides evidence under the reported superiority-testing framework. The P-value itself should still not be interpreted as the probability that the treatment effect is real, nor as a measure of the magnitude of the effect.

Important distinction: alpha spending controls statistical error across planned looks at accumulating information. It does not change the numerical meaning of the hazard ratio or its confidence interval.

9. Understanding the Composite Endpoint

The primary outcome combines several clinically distinct events into one time-to-first-event endpoint. The first qualifying occurrence determines when a participant is counted as having experienced the composite event.

Composite componentStatistical role
Persistent ≥50% reduction in eGFRRenal event component.
Persistent eGFR <15 mL/min/1.73m2Renal event component.
Initiation of chronic renal replacement therapyRenal event component.
Renal deathDeath component.
CV deathDeath component.

A composite endpoint can increase the number of events available for analysis by combining several clinically relevant outcomes. But interpretation must remain tied to the exact components: a single hazard ratio summarizes time to the first qualifying event across the composite, rather than providing separate effect estimates for each component.

The ClinicalTrials.gov record contains one formal statistical analysis for the composite endpoint. It does not provide separate statistical estimates for each individual component in the ClinicalTrials.gov record, so no component-specific treatment effects are presented here.

10. Why a Hazard Ratio Was Used

The primary endpoint is fundamentally about when the first event occurs, not simply whether an event ever occurred. That makes a survival-analysis framework appropriate.

Binary analysis alone

A simple binary comparison would distinguish participants who experienced the composite event from those who did not, but it would discard much of the information about follow-up time.

Time-to-event analysis

The Cox model uses the timing of the first event and accommodates participants whose observations are censored before an event occurs.

Hazard ratio

The HR summarizes the relative instantaneous event rate between semaglutide and placebo under the fitted model.

Stratification

The primary model was stratified by baseline SGLT-2 inhibitor use, preserving that specified analysis feature.

11. Safety Results

The registry reports serious adverse events by randomized treatment arm. These figures describe the number of affected participants divided by the number at risk in each arm.

Safety measureSemaglutidePlacebo
Serious adverse events877 / 1767950 / 1766
Serious adverse events by randomized arm
Semaglutide
877/1767
Placebo
950/1766

Serious adverse events are a safety outcome and should be interpreted separately from the primary efficacy endpoint. The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or P-value for these serious adverse-event counts, so this page does not construct one.

Interpretation caution: the serious-adverse-event figures are affected/at-risk counts, not hazard ratios. They should not be interpreted using the same framework as the primary time-to-event analysis.

12. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

The primary endpoint is a time-to-event outcome measured from randomization to the first occurrence of a qualifying composite event. A Cox model is designed for this structure because it uses event timing and accommodates right-censored observations. The reported effect measure is a hazard ratio.

What does an HR of 0.76 mean?

An HR of 0.76 means that the estimated instantaneous event rate in the semaglutide group was 0.76 times that in the placebo group under the fitted model. Expressed as a relative reduction, 1 − 0.76 = 0.24, or approximately 24%. This is not an absolute risk reduction and does not mean that 24% of participants avoided an event.

Why is the confidence interval important?

The 95% CI of 0.66–0.88 communicates uncertainty around the estimated HR. A point estimate alone can give a false impression of precision. The interval provides information about the range of parameter values compatible with the statistical model and data under the stated confidence procedure.

Why doesn't the P-value measure effect size?

The P-value of 0.0001 describes the evidence against the null hypothesis under the specified testing framework. It is influenced by both the observed data and the amount of information available. Effect size is communicated by measures such as the HR, while precision is communicated by the confidence interval.

Why was the analysis stratified by SGLT-2 inhibitor use?

The posted analysis explicitly stratified the Cox model by baseline SGLT-2 inhibitor use, yes or no. Stratification allows the model to account for the specified baseline categories while estimating the treatment effect across those strata.

Why does alpha spending matter?

The registry states that the nominal significance level was updated to 0.01612 using the Lan-DeMets alpha spending function based on the available number of events. Alpha spending is designed to preserve the overall type I error when accumulating trial information is examined according to the planned monitoring framework.

13. Confidence Interval and Statistical Precision

Primary effect estimate

0.76

95% CI: 0.66–0.88

The point estimate provides the single best estimate reported by the fitted model, while the confidence interval shows the statistical precision around that estimate. The interval remains below 1.00, the conventional null value for a hazard ratio.

How to read the interval

The lower confidence-limit estimate is 0.66 and the upper confidence-limit estimate is 0.88. Thus, the reported 95% interval spans hazard-ratio values from 0.66 to 0.88 under the specified statistical framework.

This interval should not be interpreted as saying that 95% of future individual participants will have treatment effects somewhere between 0.66 and 0.88. It is an interval for the population-level model parameter under the analysis framework.

14. P-Value, Significance Level, and Hypothesis Testing

The primary analysis is identified as a superiority hypothesis. The registry reports a P-value of 0.0001 and a nominal significance level updated to 0.01612 through the Lan-DeMets alpha-spending approach.

Testing quantityReported valueInterpretation
Hypothesis typeSuperiorityThe analysis tests whether the treatment groups differ in the prespecified direction of the superiority framework.
Nominal significance level0.01612Updated according to the available number of events using Lan-DeMets alpha spending.
P-value0.0001Evidence against the null hypothesis under the reported testing framework.
Effect estimateHR 0.76Magnitude and direction of the modeled relative treatment effect.
95% CI0.66–0.88Statistical uncertainty around the HR estimate.

These quantities answer different questions. The HR describes the estimated relative treatment effect, the confidence interval describes statistical precision, and the P-value describes evidence against the null under the specified testing procedure.

15. What the Analysis Does — and Does Not — Establish

What it establishes statistically

The reported randomized comparison has a Cox-model HR of 0.76 with a two-sided 95% CI of 0.66–0.88 and P = 0.0001 under the stated superiority-testing framework.

What it does not establish

The HR does not provide an individual participant's absolute probability of an event, and it does not show that every participant experiences the same relative change in hazard.

What the composite represents

The estimate concerns time to the first occurrence of any component of the registered composite endpoint.

What the P-value represents

The P-value represents statistical evidence against the null under the specified analysis; it is not a measure of clinical magnitude.

16. Limitations and Interpretation Issues

17. Why This Trial Matters Statistically

FLOW provides a useful teaching example because the primary analysis brings together several fundamental clinical-trial concepts: randomization, masking, a composite time-to-event endpoint, a Full Analysis Set, stratified Cox regression, hazard ratios, confidence intervals, superiority testing, and interim alpha spending.

ConceptHow it appears in FLOW
RandomizationThe trial uses randomized allocation.
Parallel designTwo parallel treatment arms are compared.
Quadruple maskingThe trial is described as quadruple-masked.
Time-to-event endpointThe primary analysis models time from randomization to the first composite event.
Composite endpointFive qualifying renal or cardiovascular components are combined into the registered primary outcome.
FAS analysisAll unique randomized participants are grouped according to treatment assigned at randomization.
Cox regressionThe primary endpoint is analyzed using a Cox proportional-hazards model.
Stratified analysisThe Cox model is stratified by baseline SGLT-2 inhibitor use.
Hazard ratioThe treatment effect is reported as HR 0.76.
Confidence intervalThe primary HR has a two-sided 95% CI of 0.66–0.88.
Interim analysis / alpha spendingThe analysis reports Lan-DeMets alpha spending and an updated nominal significance level of 0.01612.
Superiority testingThe posted hypothesis type is superiority.

18. Trial Timeline

2019-06-17

Trial start

The FLOW trial is recorded as starting on 2019-06-17.

Phase 3

Randomized parallel comparison

The completed trial enrolled 3533 participants and used two randomized treatment arms with quadruple masking.

Primary analysis

Composite renal and cardiovascular endpoint

The primary endpoint was analyzed from Week 0 up to Week 234 using a stratified Cox proportional-hazards model.

2024-01-09

Primary completion

The registry records primary completion on 2024-01-09.

19. Statistical Interpretation of the Primary Result

Effect size

The primary hazard ratio was 0.76. Because the comparison is semaglutide versus placebo, the value below 1 indicates a lower estimated hazard for the semaglutide group under the fitted Cox model. The corresponding relative interpretation is approximately a 24% lower estimated hazard.

Precision

The 95% CI of 0.66–0.88 indicates the statistical uncertainty around the HR estimate. The interval is entirely below 1, while still showing that the precise magnitude of the relative effect is not known exactly.

Statistical evidence

The reported P = 0.0001 is evaluated in the context of the trial's superiority hypothesis and the reported nominal significance level of 0.01612 established using Lan-DeMets alpha spending. It provides statistical evidence against the null under that framework, but it is not itself a measure of effect size.

Endpoint meaning

The result applies to time to the first occurrence of the registered composite endpoint: persistent ≥50% eGFR reduction, persistent eGFR <15 mL/min/1.73m2, initiation of chronic renal replacement therapy, renal death, or CV death.

20. Clinical Interpretation vs Statistical Interpretation

Statistical interpretation

The randomized comparison produced an HR of 0.76 with a two-sided 95% CI of 0.66–0.88 and P = 0.0001 in the reported stratified Cox analysis.

Clinical interpretation

The primary endpoint represents a composite of renal and cardiovascular events. Statistical interpretation should therefore remain tied to the exact composite rather than treating the HR as a separate effect estimate for every component.

21. Related Tutorials

Learn more about the methods used in this trial:

22. Related Calculators

23. Sources

Continue through the Clinical Biostats statistical pathway

Explore the statistical methods behind randomized clinical trials, time-to-event endpoints, regression models, confidence intervals, and hypothesis testing.

24. Record Summary

FLOW is a phase 3 randomized, parallel, quadruple-masked trial with 3533 participants and two treatment arms. Its registered primary endpoint is a composite of persistent ≥50% reduction in eGFR, persistent eGFR <15 mL/min/1.73m2, initiation of chronic renal replacement therapy, renal death, or CV death, assessed from Week 0 up to Week 234.

The posted primary analysis used the Full Analysis Set and compared semaglutide with placebo using a Cox proportional-hazards model stratified by baseline SGLT-2 inhibitor use. The reported hazard ratio was 0.76, with a two-sided 95% CI of 0.66–0.88 and P = 0.0001. The analysis also reports a nominal significance level of 0.01612 based on the available number of events and a Lan-DeMets alpha-spending function. eGFR was calculated using the CKD-EPI formula.

Statistically, the central lesson is the distinction among effect size, precision, and evidence. The HR describes the relative time-to-event effect, the confidence interval describes uncertainty around that estimate, and the P-value addresses the hypothesis-testing question under the specified interim-monitoring framework. The composite nature of the endpoint and the Cox model's time-to-event assumptions are equally important to interpreting what the reported result does and does not mean.

Clinical Biostats methodology: A trial-results page should distinguish the registry's reported statistical evidence from educational interpretation. For FLOW, the ClinicalTrials.gov record supports one formal primary endpoint analysis, so the page focuses on the reported composite-event hazard ratio, its confidence interval and P-value, the stratified Cox methodology, interim alpha spending, and the appropriate interpretation of the result without adding unsupported component-level or secondary efficacy estimates.