Numerical trial results on this page are restricted to the information contained in the ClinicalTrials.gov record. The page separates reported registry results from statistical interpretation. Where the registry supplies an analysis method or effect estimate, that information is reported without reconstructing or recalculating the result.
1. Trial at a Glance
SOLOIST-WHF was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating sotagliflozin versus placebo in participants with type 2 diabetes mellitus and heart failure. The registry reports 1222 enrolled participants and a primary endpoint based on the total number of cardiovascular death, hospitalizations for heart failure, and urgent heart-failure visits through 21.9 months.
| Feature | SOLOIST-WHF |
|---|---|
| Phase | Phase 3 |
| Therapeutic area | Cardiology |
| Conditions | Heart Failure; Type 2 Diabetes Mellitus |
| Design | Randomized, parallel-group |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 1222 |
| Interventions | Sotagliflozin and placebo |
| Primary endpoint type | Count / rate |
| Hypothesis type | Superiority |
| Study status | Terminated |
| Start | 2018-06-15 |
| Primary completion | 2020-06-05 |
| Lead sponsor | Lexicon Pharmaceuticals |
| Sponsor type | Industry |
2. Clinical Question
The registered primary question can be framed statistically as whether sotagliflozin changes the rate of the combined cardiovascular and heart-failure event burden compared with placebo after randomization.
Population
Participants with type 2 diabetes mellitus and heart failure enrolled in the SOLOIST-WHF trial.
Intervention
Sotagliflozin.
Comparator
Placebo.
Primary question
Does sotagliflozin alter the total number of cardiovascular death, heart-failure hospitalization, and urgent heart-failure visit occurrences after randomization?
3. Trial Design
Sotagliflozin
- Drug intervention.
- Compared with placebo in the randomized analysis.
- Included in the ITT population for efficacy analyses.
Placebo
- Placebo intervention.
- Compared with sotagliflozin in the randomized analysis.
- Included in the ITT population for efficacy analyses.
4. Endpoints
The registry reports eight posted outcome measures and eight statistical analyses. The primary endpoint is a recurrent-event measure expressed as events per 100 person-years, while several secondary outcomes use the same Cox modeling framework for time-to-event comparisons.
| Endpoint | Registry time frame | Type / unit | Analysis |
|---|---|---|---|
| Number of Total Occurrences of Cardiovascular (CV) Death, Hospitalizations for Heart Failure (HHF) and Urgent Visits for Heart Failure (HF) | Up to 21.9 months | Count / rate; events per 100 person-years | Cox proportional-hazards model |
| Total Number of Occurrences of HHF and Urgent HF Visits | Up to 21.9 months | Time-to-event; events per 100 person-years | Cox proportional-hazards model |
| Total Number of Deaths From Cardiovascular Causes | Up to 21.9 months | Time-to-event; events per 100 person-years | Cox proportional-hazards model |
| Total Number of Occurrences of CV Death, HHF, Non-fatal Myocardial Infarction and Non-fatal Stroke | Up to 21.9 months | Time-to-event; events per 100 person-years | Cox proportional-hazards model |
| Total Number of Occurrences of HHF, Urgent HF Visit, CV Death, and HF While Hospitalized | Up to 21.9 months | Time-to-event; events per 100 person-years | Cox proportional-hazards model |
| Total Number of Deaths From Any Cause | Up to 21.9 months | Time-to-event; events per 100 person-years | Cox proportional-hazards model |
| Change From Baseline in Kansas City Cardiomyopathy Questionnaire-12 (KCCQ-12) Scores at Month 4 | Baseline to Month 4 | Score on a scale | ANCOVA |
| Change From Baseline in Estimated Glomerular Filtration Rate (eGFR) | Baseline up to 21.9 months | mL/min/1.73 m2 | MMRM |
Primary endpoint definition
The registered primary endpoint is the total number of occurrences of cardiovascular death, hospitalizations for heart failure, and urgent heart-failure visits after randomization. The registry specifies that both first and potentially subsequent occurrences are included and that events are calculated as the total number of events per 100 person-years of follow-up.
5. Statistical Methodology
Intention-to-treat analysis
The registry defines the efficacy analysis population as the ITT population, including all randomized participants. The treatment comparison is therefore anchored to randomized assignment rather than restricting efficacy analysis to participants who remained on treatment or who completed follow-up.
Marginal Cox proportional-hazards model
The primary analysis used a marginal Cox proportional-hazards model. The registry specifies stratification by region and ejection fraction, with non-cardiovascular death treated as a competing event.
The reported effect measure is a hazard ratio with a two-sided 95% confidence interval.
ANCOVA for KCCQ-12
The change from baseline to Month 4 in KCCQ-12 score was analyzed using an ANCOVA model. Treatment group was included as a factor, while baseline KCCQ-12 score and randomization stratification factors were included as covariates.
MMRM for eGFR
The rate of decline in eGFR over time was analyzed using an MMRM. The registry describes absolute change in eGFR from baseline as the outcome, with a random intercept and fixed effects for treatment, baseline value, and time.
| Method | Registry-supported use in SOLOIST-WHF | Effect measure |
|---|---|---|
| Cox proportional-hazards model | Primary and several secondary time-to-event analyses | Hazard ratio |
| ANCOVA | Change from baseline in KCCQ-12 at Month 4 | Registry reports hazard ratio |
| MMRM | Change from baseline in eGFR over time | Difference in least squares means |
6. Statistical Methods Explained
Why use an intention-to-treat population?
ITT preserves the randomized comparison. Once participants have been randomized, analyzing them according to their assigned group avoids selectively removing participants because of treatment discontinuation or other post-randomization events. In this registry, all randomized participants are included in the ITT definition used for the reported efficacy analyses.
Why use a Cox model for the primary endpoint?
The registry classifies the primary statistical analysis as a Cox proportional-hazards model and reports the endpoint as a time-to-event analysis in the statistical-analysis record. Cox regression provides a model-based comparison of event hazards over follow-up and expresses the treatment contrast as a hazard ratio.
What does a hazard ratio of 0.67 mean?
A hazard ratio of 0.67 means that the estimated hazard under the fitted model was 0.67 times the corresponding hazard in the comparator group. Expressed as a relative model-based quantity, this corresponds to an estimated 33% lower hazard, because 1 − 0.67 = 0.33.
It does not mean that exactly 33% fewer participants experienced an event, nor does it directly give an absolute reduction in the number of events. Those are different quantities.
Why is the confidence interval important?
The primary 95% confidence interval is 0.52 to 0.85. It describes statistical uncertainty around the estimated hazard ratio under the specified model and sampling framework. It does not describe the range of treatment effects that must occur for individual participants.
Why doesn't the p-value measure effect size?
The primary p-value is < 0.001. A p-value addresses evidence against the null hypothesis under the specified testing framework; it is not a measure of how large or clinically important the treatment effect is. The hazard ratio and its confidence interval provide the direct information about the estimated relative effect and its precision.
Why are region and ejection fraction included as strata?
The registry specifies that the marginal Cox model is stratified by region and ejection fraction. Stratification allows the baseline hazard to vary across the specified strata while estimating the treatment comparison within the model framework. The ClinicalTrials.gov record does not provide the numerical distribution of participants across those strata.
Why does competing risk matter?
The primary and several secondary Cox analyses specify non-cardiovascular death as a competing event. This matters because a participant who dies from a non-cardiovascular cause can no longer experience a later cardiovascular or heart-failure event. Treating that competing event explicitly changes how the time-to-event analysis accounts for the available event pathways.
7. Primary Result
Total Occurrences of Cardiovascular Death, HHF and Urgent HF Visits
The primary endpoint was analyzed in the ITT population using a marginal Cox proportional-hazards model stratified by region and ejection fraction, with non-cardiovascular death treated as a competing event.
Primary hazard ratio
95% CI: 0.52–0.85 · P < 0.001
Effect measure: Hazard Ratio · Two-sided 95% CI · Superiority hypothesis
| Primary endpoint | Analysis population | Model | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Total occurrences of CV death, HHF and urgent HF visits, up to 21.9 months | ITT; all randomized participants | Marginal Cox proportional-hazards model, stratified by region and ejection fraction, with non-CV death as competing event | HR 0.67 | 0.52–0.85 | < 0.001 |
What the estimate means: the estimated hazard ratio of 0.67 indicates a lower estimated hazard for the sotagliflozin group relative to placebo under the reported Cox model. Numerically, 0.67 corresponds to an estimated 33% lower hazard relative to the comparator.
What it does not mean: it does not mean that 33% of participants were protected, that the number of events was reduced by exactly 33%, or that each participant experienced the same reduction in risk. Because the registered endpoint includes potentially subsequent events, the hazard ratio should not be casually translated into a simple proportion of participants having a first event.
What the confidence interval says: the two-sided 95% CI of 0.52–0.85 indicates uncertainty around the estimated hazard ratio. The interval remains below 1, but its width shows that the estimate is not known with perfect precision.
What the p-value says: the p-value < 0.001 provides evidence against the null hypothesis in the reported superiority analysis. It does not measure the magnitude of the treatment effect; the HR and CI do that.
Important modeling cautions: the result depends on the specified Cox model, its treatment of follow-up, the stratification variables, and the handling of non-cardiovascular death as a competing event. The ClinicalTrials.gov record does not provide enough information to independently assess the proportional-hazards assumption.
8. Secondary Endpoint Results
The registry reports seven secondary statistical analyses. Five use Cox proportional-hazards models for time-to-event outcomes, one uses ANCOVA for KCCQ-12 change at Month 4, and one uses MMRM for eGFR change over time.
| Secondary endpoint | Estimate | 95% CI | P-value | Method |
|---|---|---|---|---|
| Total occurrences of HHF and urgent HF visits | HR 0.64 | 0.49–0.83 | < 0.001 | Cox proportional-hazards model |
| Total deaths from cardiovascular causes | HR 0.84 | 0.58–1.22 | = 0.36 | Cox proportional-hazards model |
| Total occurrences of CV death, HHF, non-fatal myocardial infarction and non-fatal stroke | HR 0.72 | 0.56–0.92 | Not reported | Cox proportional-hazards model |
| Total occurrences of HHF, urgent HF visit, CV death, and HF while hospitalized | HR 0.68 | 0.54–0.86 | Not reported | Cox proportional-hazards model |
| Total deaths from any cause | HR 0.82 | 0.59–1.14 | Not reported | Cox proportional-hazards model |
| Change from baseline in KCCQ-12 scores at Month 4 | 4.1 | 1.3–7 | Not reported | ANCOVA |
| Change from baseline in eGFR | −0.16 | −1.3–0.98 | Not reported | MMRM |
Total Occurrences of HHF and Urgent HF Visits
Hazard ratio
95% CI: 0.49–0.83 · P < 0.001
This secondary analysis used the ITT population and a Cox proportional-hazards model stratified by region and ejection fraction, with non-cardiovascular death treated as a competing event.
Total Number of Deaths From Cardiovascular Causes
Hazard ratio
95% CI: 0.58–1.22 · P = 0.36
The point estimate is below 1, but the 95% confidence interval extends from below 1 to above 1. The registry reports a two-sided p-value of 0.36 for this superiority analysis.
Composite of CV Death, HHF, Non-fatal MI and Non-fatal Stroke
Hazard ratio
95% CI: 0.56–0.92
The registry reports this endpoint as a time-to-event analysis in the ITT population, using a Cox proportional-hazards model. The reported confidence interval provides the registry-reported measure of precision around the hazard ratio.
Composite of HHF, Urgent HF Visit, CV Death, and HF While Hospitalized
Hazard ratio
95% CI: 0.54–0.86
This endpoint was analyzed with a Cox proportional-hazards model in the ITT population. The registry identifies region and ejection fraction as stratification factors and non-cardiovascular death as a competing event for this analysis.
Total Deaths From Any Cause
Hazard ratio
95% CI: 0.59–1.14
The registry reports a Cox proportional-hazards analysis in the ITT population. Unlike an endpoint restricted to cardiovascular death, all-cause mortality includes deaths from any cause. The ClinicalTrials.gov record does not provide a p-value for this analysis.
KCCQ-12 Change From Baseline at Month 4
Reported estimate
95% CI: 1.3–7
ANCOVA; baseline to Month 4
The registry states that the change from baseline to Month 4 was analyzed with ANCOVA, using treatment group as a factor and baseline KCCQ-12 score plus randomization stratification factors as covariates.
Change From Baseline in eGFR
Difference in Least Squares Means
95% CI: −1.3–0.98
MMRM; baseline up to 21.9 months
The registry describes an MMRM with absolute change in eGFR from baseline as the outcome, a random intercept, and fixed effects for treatment, baseline value, and time. The reported effect measure is the difference in least squares means.
9. How to Read the Secondary Results
Relative event measures
The reported Cox hazard ratios for several cardiovascular and heart-failure endpoints are below 1, indicating lower estimated hazards for sotagliflozin under the corresponding models.
Cardiovascular mortality
The cardiovascular-death HR is 0.84 with a 95% CI of 0.58–1.22 and a p-value of 0.36, illustrating why the point estimate alone is not enough to characterize statistical uncertainty.
Patient-reported outcome
KCCQ-12 was analyzed using covariate-adjusted ANCOVA rather than a survival model, demonstrating that different endpoint types require different analytical frameworks.
Longitudinal renal outcome
eGFR was analyzed with an MMRM, which uses repeated observations over time rather than reducing the longitudinal record to a single time point.
10. Safety Results
The registry reports serious adverse events by randomized treatment arm using affected participants over the corresponding at-risk denominators.
| Safety measure | Sotagliflozin | Placebo |
|---|---|---|
| Serious adverse events | 235/605 | 251/611 |
Sotagliflozin
235 participants with serious adverse events among 605 participants at risk.
Placebo
251 participants with serious adverse events among 611 participants at risk.
The ClinicalTrials.gov record provides affected/at-risk counts rather than a formal between-group safety hypothesis test. Accordingly, this page does not attach a p-value or comparative effect estimate to the serious-adverse-event counts.
11. Stratification and Covariate Adjustment
Stratification and covariate adjustment play different roles in the reported analyses. The Cox analyses use region and ejection fraction as stratification factors. The KCCQ-12 ANCOVA additionally adjusts for baseline KCCQ-12 score and randomization stratification factors.
| Analysis | Adjustment / stratification specified in registry | Purpose in interpretation |
|---|---|---|
| Primary Cox analysis | Region and ejection fraction; non-CV death as competing event | Defines the reported stratified survival-model comparison |
| Secondary Cox analyses | Region and ejection fraction; non-CV death as competing event where specified | Defines the corresponding time-to-event model |
| KCCQ-12 ANCOVA | Baseline KCCQ-12 score and randomization stratification factors as covariates | Accounts for baseline score and specified stratification variables when comparing change |
| eGFR MMRM | Treatment, baseline value and time as fixed effects; random intercept | Models repeated measurements over time |
12. Time-to-Event Analysis and the Hazard Ratio
Several SOLOIST-WHF endpoints are represented in the registry as time-to-event analyses. The hazard ratio is a relative measure of the event hazard under a fitted survival model. It is particularly useful when follow-up times differ among participants and censoring occurs.
An HR below 1 indicates a lower estimated hazard in the treatment group under the model. An HR of 1 would correspond to equal hazards, while an HR above 1 would indicate a higher estimated hazard.
For the primary endpoint, the registry reports HR 0.67. This is a relative model-based measure. It is not an absolute event-rate difference, a risk ratio, a probability of avoiding hospitalization, or a statement about an individual participant's outcome.
Why person-years appear in the endpoint definition
The primary endpoint is described as total events per 100 person-years of follow-up. Person-time incorporates both the number of observed events and the amount of follow-up contributing to the analysis. This is particularly relevant when participants have different lengths of observation.
13. ANCOVA and Covariate Adjustment for KCCQ-12
The KCCQ-12 endpoint is fundamentally different from the recurrent cardiovascular event endpoint. It asks about change in a score from baseline to Month 4, so the registry uses ANCOVA rather than a time-to-event model.
The registry specifies treatment group as a factor and baseline KCCQ-12 score plus randomization stratification factors as covariates.
Covariate adjustment can improve precision when baseline measurements explain part of the variation in the follow-up outcome. The important statistical distinction is that the treatment comparison is still based on randomized groups; the ANCOVA specifies how the continuous outcome is modeled.
14. MMRM for eGFR
The eGFR analysis illustrates a third statistical framework. Rather than analyzing only one follow-up measurement, the registry describes a mixed model for repeated measures using absolute change from baseline as the outcome.
Repeated observations
MMRM is designed for outcomes observed repeatedly over time, allowing the analysis to use the longitudinal structure rather than treating each time point as an unrelated analysis.
Random intercept
The registry specifies a random effect for the intercept, allowing the model to account for participant-level variation in the longitudinal outcome.
Fixed effects
Treatment, baseline value, and time are specified as fixed effects in the registered analysis description.
Reported estimate
The effect measure is the difference in least squares means: −0.16, with a 95% CI of −1.3 to 0.98.
The confidence interval spans zero, which is the reference value for a difference. The ClinicalTrials.gov record does not provide a p-value for this analysis, so no additional significance statement is assigned here.
15. Multiplicity and the Collection of Endpoints
SOLOIST-WHF has one registered primary endpoint and multiple secondary endpoints. The ClinicalTrials.gov record identifies the hypothesis type as superiority, but they do not provide an alpha-allocation strategy, hierarchical testing procedure, multiplicity-adjustment method, or interim-analysis alpha-spending plan.
| Endpoint role | Number reported in the ClinicalTrials.gov record | Interpretive role |
|---|---|---|
| Primary endpoint | 1 | Primary superiority comparison |
| Secondary statistical analyses | 7 | Additional efficacy and clinical outcome assessments |
| Total statistical analyses posted | 8 | Primary plus secondary analyses |
16. Missing Data and Analysis Assumptions
The ClinicalTrials.gov record identifies the ITT population and the statistical methods, but they do not provide a missing-data or imputation strategy for the primary endpoint, KCCQ-12, or eGFR analysis.
That absence matters because different endpoint types create different missing-data problems. A time-to-event analysis can use available follow-up until an event or censoring, whereas a longitudinal score such as KCCQ-12 can be affected by missing Month 4 assessments. An MMRM similarly relies on assumptions about the observed longitudinal data and the missingness mechanism.
The ClinicalTrials.gov record does not provide enough information to determine the precise imputation or sensitivity-analysis framework. No such method is therefore attributed to the trial here.
17. Interim Analysis, Crossover, and Bayesian Methods
The ClinicalTrials.gov record does not report an interim-analysis method, crossover design, factorial structure, or Bayesian statistical method.
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Interim analysis | Not reported in the ClinicalTrials.gov record. |
| Alpha spending | Not reported in the ClinicalTrials.gov record. |
| Crossover | Not reported in the ClinicalTrials.gov record. |
| Factorial design | No; the registered design model is parallel. |
| Bayesian methods | Not reported in the ClinicalTrials.gov record. |
| Non-inferiority margin | Not applicable to the reported superiority hypothesis; no non-inferiority margin is reported. |
18. Study Status and Premature Termination
The registry status is TERMINATED. The registry-reported limitations field states that the study was terminated prematurely due to a business decision.
The registry also identifies limitations of the trial such as small numbers of participants analyzed or technical problems leading to unreliable data. Because the ClinicalTrials.gov record does not quantify these issues beyond that statement, this page does not assign an additional numerical adjustment or sensitivity estimate.
19. Limitations
- Premature termination: the study was terminated prematurely due to a business decision, which can limit the amount of information available for analysis.
- Potentially small analysis numbers: the registry limitation statement specifically notes small numbers of participants analyzed as a possible source of unreliable data.
- Technical problems: the registry also identifies technical problems as a potential source of unreliable data.
- Recurrent-event endpoint: the primary endpoint includes first and potentially subsequent occurrences, so its interpretation is more complex than a simple time-to-first-event endpoint.
- Model dependence: the primary hazard ratio is a model-based estimate from a marginal Cox proportional-hazards model.
- Competing event: non-cardiovascular death is treated as a competing event in the primary and several secondary Cox analyses.
- Proportional-hazards assumption: the ClinicalTrials.gov record does not provide enough information to independently evaluate whether the proportional-hazards assumption is appropriate.
- Multiplicity: the ClinicalTrials.gov record does not state how multiplicity across secondary endpoints was controlled.
- Incomplete statistical detail: the ClinicalTrials.gov record does not report a detailed missing-data strategy, interim-analysis framework, or Bayesian method.
- Registry effect-measure labeling: the KCCQ-12 ANCOVA analysis is labeled with “Hazard Ratio” in the ClinicalTrials.gov record even though ANCOVA is the reported method and the outcome is a score.
20. Why This Trial Matters Statistically
SOLOIST-WHF is a useful statistical teaching case because a single randomized trial contains several distinct outcome structures and therefore several distinct analytical frameworks. The primary endpoint combines recurrent cardiovascular and heart-failure events, while secondary outcomes extend the analysis to mortality, additional composite events, a patient-reported score, and longitudinal renal function.
| Statistical concept | How it appears in SOLOIST-WHF |
|---|---|
| Randomization | Randomized allocation in a two-arm parallel design. |
| Blinding | Quadruple masking. |
| Intention-to-treat analysis | ITT population includes all randomized participants. |
| Time-to-event analysis | Used for the primary and several secondary cardiovascular endpoints. |
| Hazard ratio | Primary effect measure and effect measure for multiple secondary Cox analyses. |
| Stratified analysis | Primary Cox model stratified by region and ejection fraction. |
| Competing risk | Non-cardiovascular death treated as a competing event in the reported Cox analyses. |
| ANCOVA | Used for change from baseline in KCCQ-12 at Month 4. |
| Covariate adjustment | Baseline KCCQ-12 and randomization stratification factors included in the KCCQ-12 ANCOVA. |
| MMRM | Used to analyze change in eGFR over time. |
| Confidence intervals | Reported for all eight statistical analyses in the ClinicalTrials.gov record. |
| Superiority testing | Registered hypothesis type for the primary and reported secondary analyses. |
21. A Statistical Reading of the Primary Result
The primary HR of 0.67 is a relative measure. Under the specified marginal Cox model, the estimated hazard in the sotagliflozin group is 0.67 times that in the placebo group.
The 95% CI of 0.52–0.85 shows the uncertainty around that estimate. It is substantially more informative than the point estimate alone because it indicates how precisely the treatment effect was estimated under the statistical framework.
The reported two-sided P < 0.001 indicates strong evidence against the null hypothesis used for the superiority comparison. It should not be interpreted as a probability that the treatment effect is exactly a particular size.
The registry's primary result is presented as a hazard ratio rather than an absolute risk difference. Without additional event totals, cumulative incidence estimates, or other absolute measures in the ClinicalTrials.gov record, an absolute treatment effect should not be calculated or inferred.
22. Primary and Secondary Results in Context
The reported results show that the statistical evidence is not uniform across every endpoint. Several composite cardiovascular and heart-failure outcomes have hazard ratios below 1 with confidence intervals entirely below 1, whereas the cardiovascular-death analysis has an HR of 0.84 with a confidence interval extending above 1. The eGFR analysis likewise has a confidence interval that crosses zero.
This pattern illustrates why a clinical-trial statistical review should examine the endpoint definition, estimand, analysis method, effect estimate, confidence interval, and testing framework together. A collection of point estimates cannot be interpreted correctly without knowing what each endpoint measures and how it was analyzed.
| Question | Primary endpoint | Secondary endpoint example |
|---|---|---|
| What is being measured? | Total occurrences of CV death, HHF and urgent HF visits | Change in KCCQ-12 at Month 4 |
| Outcome structure | Event-based / time-to-event | Continuous change score |
| Model | Marginal Cox proportional-hazards model | ANCOVA |
| Effect measure | Hazard ratio | Registry labels estimate as HR despite ANCOVA method |
| Primary estimate | 0.67 | 4.1 |
23. Related Tutorials
Learn more about the statistical methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: SOLOIST-WHF, NCT03521934.
- PubMed: PMID 39260929.
- PubMed: PMID 38878007.
- PubMed: PMID 38771012.
- PubMed: PMID 38770818.
- PubMed: PMID 34152828.
Continue with the statistical methods behind the trial
Explore the underlying survival, regression, longitudinal, and clinical-trial methods used to analyze endpoints such as those in SOLOIST-WHF.
26. Record Summary
SOLOIST-WHF provides a useful example of how a randomized clinical trial can require several statistical frameworks within one analysis plan. Its primary endpoint combines total occurrences of cardiovascular death, hospitalization for heart failure, and urgent heart-failure visits, including first and potentially subsequent occurrences. The primary analysis uses a marginal Cox proportional-hazards model stratified by region and ejection fraction, with non-cardiovascular death treated as a competing event.
The reported primary hazard ratio is 0.67 with a two-sided 95% CI of 0.52–0.85 and P < 0.001. Secondary analyses extend the statistical story to additional heart-failure and mortality endpoints, KCCQ-12 change analyzed by ANCOVA, and eGFR change analyzed by MMRM. The reported serious-adverse-event counts were 235/605 for sotagliflozin and 251/611 for placebo.
The most important statistical lesson is that these results should not be reduced to a single number. The primary hazard ratio, its confidence interval, the recurrent-event endpoint definition, competing-risk treatment, ITT population, stratification, secondary endpoints, ANCOVA adjustment, MMRM structure, and premature termination all contribute to the correct interpretation of the evidence.