This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the results posted in the ClinicalTrials.gov record; calculations added by Clinical Biostats are labelled as such.
1. Trial at a Glance
This page covers the registry results for A Study Using Medical Records of Danish People With Type 2 Diabetes Comparing Empagliflozin and Glucagon-Like Peptide-1 Receptor Agonists (GLP1-RA) in the Occurrence of Serious Cardiovascular Outcomes (NCT03993132), a registered phase not stated study. The primary outcome is Incidence Rate of Expanded Major Adverse Cardiovascular Event (MACE) Composite Outcome - OT Analysis.
| Feature | Registry value |
|---|---|
| ClinicalTrials.gov identifier | NCT03993132 |
| Phase | Phase not stated |
| Condition | Diabetes Mellitus, Type 2 |
| Allocation | Not Stated |
| Design model | Not Stated |
| Masking | Not Stated |
| Enrollment | 26774 |
| Arms in the primary outcome | n/a |
| Lead sponsor | Boehringer Ingelheim |
| Results posted | Yes |
2. Clinical Question
Population
Patients with type 2 diabetes mellitus included in the propensity score balanced cohort of empagliflozin or liraglutide users. The registry describes a Danish nationwide study using medical records. It does not report additional eligibility criteria or the characteristics of the balanced cohort.
Intervention
Initiation of empagliflozin. The primary outcome population is described as patients who initiated one of the study drugs and entered the propensity score balanced cohort. The registry does not report treatment-group sizes or the details of how treatment initiation was identified.
Comparator
Initiation of liraglutide, a glucagon-like peptide-1 receptor agonist. Although the research question refers to GLP1-RA treatment, liraglutide is the named comparator in the intervention record and the primary outcome description.
Primary question
Among people with type 2 diabetes, how does initiation of empagliflozin compare with initiation of liraglutide in the adjusted incidence rate of the expanded major adverse cardiovascular event composite during on-treatment follow-up?
This is a comparative effectiveness question about event rates during treatment exposure, not a randomized treatment assignment question. The registry states that the primary research question is whether initiating empagliflozin changes the adjusted incidence of outcomes compared with initiating GLP1-RA therapy. The listed primary analysis is on-treatment and concerns an expanded composite of cardiovascular events and all-cause death.
The study enrolled 26774 participants overall, but the registry does not report how many participants were in either balanced treatment cohort. It also does not report a numerical estimate for the primary comparison. As a result, the precise direction and magnitude of the reported primary contrast cannot be assessed from the registry record.
3. Trial Design
The registry describes a Danish nationwide comparative effectiveness study using medical records of people with type 2 diabetes. The study compares initiators of empagliflozin with initiators of liraglutide. Its status is completed, and total enrollment is 26774. The registry lists a start date of 2018-10-01 and primary completion on 2022-06-16.
This is not a randomized comparison; allocation is not stated in the registry record. The primary outcome population consists of patients included in a propensity score balanced cohort of empagliflozin or liraglutide users. Propensity score balancing is intended to make measured characteristics more comparable between groups, but does not by itself establish that the groups are exchangeable or eliminate differences in unmeasured factors. The registry record does not report which characteristics entered the score, the balance diagnostics, or the matching or weighting procedure.
The primary analysis is described as on-treatment. Follow-up begins with first initiation of empagliflozin or liraglutide and continues until the end of follow-up, up to 6 years. For the on-treatment analysis, the estimated number of days covered by each prescription was based on the number of drug packages multiplied by the numerical volume of a package, with a 30-day grace period added. Participants were censored at treatment cessation, initiation of an alternative drug in the study drug class, or initiation of a drug from the comparator study drug class.
The registry lists 14 outcome measures and 14 analyses. The secondary outcomes include rates involving heart failure hospitalization, all-cause death, loop-diuretic initiation, all-cause hospitalization, and death; several are listed under both on-treatment and intention-to-treat approaches. Details about the intention-to-treat analysis are not reported in the registry record.
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.
Before assignment: All subjects were screened for eligibility to ensure that they (the subjects) strictly met all inclusion and none of the exclusion criteria.
Overall Study
| Stage | Empagliflozin - PS Balanced Cohort | Liraglutide - PS Balanced Cohort |
|---|---|---|
| STARTED | 14148 | 12626 |
| COMPLETED | 14148 | 12626 |
| NOT COMPLETED | 0 | 0 |
Analysis population for the primary outcome, as defined in the registry: All patients included in the propensity score (PS) balanced cohort of either empagliflozin or liraglutide users.
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.
Population: All patients included in the propensity score (PS) balanced cohort of either empagliflozin or liraglutide users.
| Characteristic | Unit | Empagliflozin - PS Balanced Cohort | Liraglutide - PS Balanced Cohort | Total |
|---|---|---|---|---|
| Participants analysed | 14148 | 12626 | 26774 | |
| Age, Continuous | Years | 61.55 (52.73 to 69.83) | 61.19 (52.31 to 69.83) | 61.37 (52.54 to 69.70) |
| Sex: Female, Male: Female | Participants | 5692 | 5162 | 10854 |
| Sex: Female, Male: Male | Participants | 8456 | 7464 | 15920 |
| Race and Ethnicity Not Collected | Participants | — | — | 0 |
6. Primary Endpoint
The registry lists the primary outcome as Incidence Rate of Expanded Major Adverse Cardiovascular Event (MACE) Composite Outcome - OT Analysis, measured over From first initiation of empagliflozin or liraglutide until end of follow-up, up to 6 years..
| Element | Registry entry |
|---|---|
| Result type | Number |
| Measure of spread | not applicable |
| Unit | Events per 1000 person-years |
| Endpoint classification | Event-Rate Endpoint |
The primary endpoint is the incidence rate of an expanded major adverse cardiovascular event composite, reported in events per 1000 person-years. It includes all-cause death; acute admission with non-fatal stroke within 30 days; acute admission with non-fatal myocardial infarction within 30 days; admission with unstable angina; coronary revascularization; or acute admission with non-fatal heart failure. The registry describes the primary analysis as on-treatment and defines follow-up from first initiation of either study drug until the end of follow-up, up to 6 years.
In general, a composite endpoint combines multiple kinds of events into one outcome measure. This can capture a broader range of clinically important events than any single component, but it also means that the overall rate can reflect different component events. The registry does not report how many events occurred for each component, so the contribution of death, stroke, myocardial infarction, unstable angina, revascularization, and heart failure cannot be distinguished from the record.
An incidence rate uses accumulated person-time as its exposure denominator. In general, person-time accounts for differences in how long participants contribute follow-up, which can be useful when follow-up duration varies. The stated unit, events per 1000 person-years, expresses the rate against a common amount of follow-up time. The registry does not report the primary event rates, an adjusted rate contrast, or an uncertainty interval. The endpoint definition therefore identifies what was intended to be measured, but does not establish the observed comparative result.
7. Why No Re-analysis Is Shown
8. Statistical Interpretation
The registry identifies a clinically broad primary outcome and an on-treatment comparison between empagliflozin initiators and liraglutide initiators. However, it does not report the primary outcome estimates. The event rate in either cohort, the adjusted comparison between cohorts, and its statistical uncertainty are therefore not available in the registry record. No conclusion about the direction or size of the observed association can be drawn from the endpoint description alone.
The outcome is an incidence rate rather than a simple count or proportion. In general, an incidence rate relates observed events to time at risk, so its interpretation depends on both event occurrence and the amount of follow-up contributed. The registry gives the unit as events per 1000 person-years but does not report the cohort-specific event counts, person-time totals, or the rate estimates. It also does not report a confidence interval or p-value for the primary comparison. Without these, neither the precision of the estimate nor compatibility with a range of possible associations can be evaluated.
The primary analysis is on-treatment. The registry says participants were censored when treatment stopped, an alternative drug in the study drug class was initiated, or a drug from the comparator study drug class was initiated. This defines the treatment-exposure window for the stated analysis. In general, censoring ends a participant’s contribution to an analysis at a specified point; valid interpretation can depend on how censoring relates to subsequent outcome risk. The registry does not report how such potential informative censoring was addressed, or whether sensitivity analyses tested alternative assumptions.
Propensity score balancing is described for the primary outcome cohort, but the registry does not report the variables used or the balance achieved. In general, adjustment can address differences in measured characteristics, but residual confounding may remain if important predictors were not measured or adequately balanced. This is especially relevant to a nonrandomized comparison, where treatment initiation is not assigned at random.
The registry lists secondary rate outcomes for heart failure hospitalization, all-cause death, hospitalizations, and loop-diuretic initiation, with on-treatment and intention-to-treat labels on several outcomes. It provides no estimates for these outcomes in the record described. The number of listed analyses alone does not indicate which result, if any, was statistically persuasive. Multiplicity procedures and missing-data handling are not reported in the registry record.
Accordingly, the available information supports a description of the study question, endpoint, and on-treatment follow-up rules, but not an estimate of comparative effectiveness. Any planning illustration based on an observed effect is a design calculation under stated assumptions, not observed power and not a substitute for the primary estimate or its uncertainty interval.
9. Statistical Methods Explained
The primary outcome is described as an event rate per 1000 person-years, estimated in a propensity score balanced cohort and analyzed on-treatment. The registry supplies some operational definitions for this analysis, but does not report the numerical outcome results or a full statistical analysis plan.
What does an incidence rate represent?
In general, an incidence rate divides the number of events by the accumulated time participants contribute while under observation. It is useful when participants contribute different durations of follow-up, because a participant observed for longer contributes more time at risk. The registry’s stated unit is events per 1000 person-years. It does not report the event counts, person-time, cohort-specific rates, or adjusted rate comparison, so none of those quantities can be reconstructed from the record.
A rate is not the same as the probability that an individual experiences an event over a fixed period. In general, the rate describes the pace of events relative to observed time, while a fixed-time risk describes the proportion experiencing an event over a specified interval. The endpoint’s follow-up may extend up to 6 years, but no fixed-time risks are reported.
What does the composite endpoint combine?
The primary endpoint groups all-cause death, acute admission with non-fatal stroke within 30 days, acute admission with non-fatal myocardial infarction within 30 days, admission with unstable angina, coronary revascularization, and acute admission with non-fatal heart failure. In general, a composite outcome can capture several event types in one analysis, but it can be harder to interpret than a single outcome if components differ in frequency or importance. The registry does not report component-specific event results, so the overall endpoint cannot be attributed to particular components.
What does the on-treatment approach mean?
The registry states that treatment duration was estimated from prescription packages and package volume, with a 30-day grace period added. Follow-up was censored at treatment cessation, initiation of an alternative drug in the study drug class, or initiation of a drug from the comparator study drug class. In general, an on-treatment analysis focuses on outcomes during the defined treatment period rather than continuing to count follow-up after treatment changes or discontinuation. This can address a question about outcomes while participants remain on the initial treatment, but it may also make interpretation sensitive to the censoring process. The registry does not report additional details on adherence measurement, handling of exposure misclassification, or methods for accounting for informative censoring.
How should propensity score balancing be understood?
In general, a propensity score summarizes the probability of receiving a treatment given measured characteristics. Balancing on such a score is intended to improve comparability on recorded variables between treatment groups. It does not make a nonrandomized study equivalent to a randomized trial: unmeasured or poorly measured characteristics can still differ, and the outcome estimate can remain confounded. The registry identifies a propensity score balanced cohort but does not report the score model, covariates, balance measures, exclusions during cohort construction, or whether balancing used matching, weighting, stratification, or another procedure.
The registry also does not describe the model used to estimate adjusted incidence rates, the adjustment set beyond the propensity score balancing description, or the method for calculating uncertainty. It does not report whether assumptions about the rate model were checked. In general, rate analyses can be affected by over-dispersion, when observed variability exceeds what a simple model expects; the registry does not state whether this issue arose or how it was assessed.
What can planning calculations tell readers?
Planning illustrations based on an observed effect are hypothetical calculations, not study findings and not observed power. The stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, a normal approximation, and an effect defined as a stated fraction of the observed effect. For time-to-event outcomes, the stated planning method uses the total number of events; other outcomes use participants per arm. These assumptions describe an illustration rather than the study’s original design justification. A protocol-based sample-size justification and any observed-effect planning values are not reported in the registry record. Such calculations cannot establish that the study was adequately powered or substitute for reporting the estimated effect and its precision.
10. Multiplicity, Missing Data and Analysis Populations
The registry lists 14 outcome measures and 14 analyses. The primary outcome is the on-treatment rate of the expanded major adverse cardiovascular event composite. Secondary outcomes include heart failure hospitalization or all-cause death; first hospitalized heart failure or initiation of community loop-diuretic therapy; all-cause hospitalization or death; and all-cause hospitalization. Several are separately labeled as on-treatment and intention-to-treat analyses. The registry does not report estimates for these outcomes in the record described.
When a study evaluates multiple outcomes or analyses, the chance of finding at least one apparently notable result can increase. In general, multiplicity procedures may adjust significance thresholds, or investigators may designate a hierarchy of outcomes to guide interpretation. The registry record does not report whether any multiplicity adjustment, testing hierarchy, or prespecified distinction between confirmatory and exploratory analyses was used. It also does not report interim analyses or stopping rules.
Missing-data methods are not reported in the registry record. The record does not describe the extent or handling of missing covariates, outcome data, or prescription information. The primary analysis does specify censoring at treatment cessation and specified treatment changes, but that is an analysis rule rather than a complete account of missing-data handling. Whether censoring was treated as independent of future outcome risk, and whether methods were used to address informative loss of follow-up, are not reported.
The primary population is described as the propensity score balanced cohort, but the registry does not report cohort sizes, participant flow, or how many people were excluded during balancing. It does not report further details about the intention-to-treat population or whether all enrolled participants contributed to each secondary analysis. These omissions limit assessment of analysis-population differences and the impact of missingness or exclusions.
11. What the Result Does, and Does Not, Mean
The record establishes the study’s question and describes how the primary endpoint was defined and followed. It does not report the primary event rates, the adjusted comparison, a confidence interval, or a p-value. Therefore, it cannot show from the registry record whether the observed rate was higher or lower in either treatment cohort, how large any difference was, or how precisely it was estimated.
The stated endpoint combines death and several cardiovascular events. In general, a composite result represents the combined endpoint and should not automatically be read as evidence that every component behaved in the same way. Because component-specific results are not reported, the record does not show which events contributed to the composite or whether the result was driven by any one component.
The comparison uses propensity score balancing in a nonrandomized study. In general, this approach can improve balance on measured characteristics, but it cannot guarantee comparability on unmeasured factors. The available description does not include the score model or balance diagnostics, so the adequacy of measured confounding control cannot be evaluated from the registry record.
The on-treatment definition focuses the analysis on follow-up during estimated treatment exposure and specifies when censoring occurs. This does not answer every possible treatment question. In general, an on-treatment analysis differs from an intention-to-treat analysis, which follows participants according to their initial treatment assignment or group even after treatment changes, depending on the study’s definition. The registry lists intention-to-treat secondary analyses but does not provide their results or operational details.
These limitations do not mean that the study had no result; they mean the registry record described does not report the numerical results needed to assess it. The record supports an account of design and endpoint construction, not a claim that either treatment is more effective or less effective.
12. Why This Trial Matters Statistically
This study addresses a comparative effectiveness question using medical records in a nationwide Danish setting. Statistically, it illustrates the distinction between a prespecified clinical question and the information needed to interpret an estimate. Knowing the endpoint definition and analysis population is important, but the primary event rates, adjusted contrast, and uncertainty are also needed to judge the magnitude and precision of the association.
The use of a propensity score balanced cohort makes confounding control central to interpretation. In general, balancing measured characteristics can improve comparability in nonrandomized data, while the credibility of the comparison also depends on which characteristics were measured, how the score was constructed, and whether balance was achieved. Those details are not reported in the registry record, so readers cannot assess those parts of the design.
The on-treatment approach also makes exposure definition and censoring relevant. Prescription-based treatment duration, a grace period, and specified censoring events determine which follow-up time contributes to the primary analysis. In general, these choices can affect both the estimand and the possibility of bias if treatment discontinuation or switching is related to outcome risk. A clear report of the analytic implementation and sensitivity analyses would help readers evaluate robustness; such details are not reported in the registry record.
Finally, the study lists multiple outcomes and analysis populations. Reporting the primary result alongside component outcomes, uncertainty intervals, missing-data procedures, and multiplicity decisions would help distinguish the principal comparison from other analyses. The registry record’s value is its description of the question and approach; its limits underscore why complete statistical reporting matters for interpreting comparative evidence.
13. 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 primary endpoint estimate, the cohort-specific incidence rates, the adjusted rate contrast, or an uncertainty interval. It therefore does not support a quantitative interpretation of the primary comparison. It also does not provide component-specific results for the expanded cardiovascular composite, so the contribution of each event type cannot be assessed.
Important design details are not reported, including treatment-group sizes, participant flow into the propensity score balanced cohort, eligibility criteria beyond the described population, the covariates used in propensity score estimation, the balancing procedure, and balance diagnostics. Without these details, it is not possible to assess how well the cohorts were made comparable on measured characteristics. Because allocation was not randomized and is not stated, residual confounding remains a concern for interpretation in general.
Exposure measurement is described using prescription package counts and package volume, but the registry does not report validation of that measure, adherence information, or sensitivity analyses for alternative exposure definitions. It states that follow-up was censored at treatment cessation and specified treatment changes, but does not report methods for addressing possible informative censoring or differences in follow-up. Model specification and uncertainty estimation for the rates are also not reported.
Multiplicity procedures, interim analyses, sample-size justification, missing-data methods, and detailed intention-to-treat definitions are not reported in the registry record. The record lists adverse-event group counts with zero serious and other affected participants and lists deaths of 881 in the empagliflozin cohort and 765 in the liraglutide cohort, but does not provide context sufficient to interpret these counts as comparative safety results. No serious adverse terms are listed.
14. Related Tutorials
15. Related Statistical Calculators
16. Sources
- ClinicalTrials.gov: A Study Using Medical Records of Danish People With Type 2 Diabetes Comparing Empagliflozin and Glucagon-Like Peptide-1 Receptor Agonists (GLP1-RA) in the Occurrence of Serious Cardiovascular Outcomes — NCT03993132.
- Registry record last updated 2024-02-12; status last verified 2023-06.
- Results first posted 2024-02-12.
- 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.
Continue through the Clinical Biostats statistical pathway
Use this trial's endpoint as a starting point for deeper study of effect measures, confidence intervals, and sample-size planning.
17. Record Summary
NCT03993132 is a completed Danish nationwide comparative effectiveness study using medical records of patients with type 2 diabetes. Total enrollment is 26774. The study compares initiation of empagliflozin with initiation of liraglutide, using a propensity score balanced cohort for the primary outcome.
The primary endpoint is an on-treatment incidence rate for an expanded major adverse cardiovascular event composite, expressed in events per 1000 person-years. Follow-up is described as beginning at first initiation and continuing up to 6 years, with specified censoring at treatment cessation and treatment changes. The registry lists 14 outcome measures and 14 analyses.
The registry record does not report the primary outcome estimates, uncertainty intervals, detailed propensity score methods, multiplicity procedures, or missing-data methods. It describes the research question and analysis approach, but does not provide enough numerical results to assess the comparative outcome.