1. Introduction: Why the Trial Design Matters

Most cardiovascular outcome trials (CVOTs) in type 2 diabetes conducted after the 2008 FDA guidance used placebo-controlled designs to establish cardiovascular safety, with some subsequently demonstrating superiority. SURPASS-CVOT departed from that model. Because cardiovascular benefit had already been established for the GLP-1 receptor agonist class, including dulaglutide in REWIND, the trial was designed as a head-to-head comparison against an active drug rather than placebo ^[1][2].

This design addressed a clinically relevant question but changed the interpretation of the treatment effect. A non-inferiority trial against an active comparator estimates the difference between two active treatment strategies; it does not directly identify the absolute effect of tirzepatide relative to placebo or untreated standard care. Any inference about absolute cardiovascular protection therefore requires external evidence about the comparator and depends on the assumption that its effect is sufficiently preserved across different populations and background therapies. That inferential structure is essential when interpreting both the primary result and the subsequent secondary analyses.

2. Trial Design and Population

SURPASS-CVOT randomized 13,299 participants at 640 sites in 30 countries between 2020 and 2022 ^[2]. Eligibility required age ≥40 years, HbA1c 7.0–10.5%, BMI ≥25 kg/m², and established atherosclerotic cardiovascular disease. Participants received once-weekly tirzepatide escalated to a maximally tolerated dose of up to 15 mg, or dulaglutide at a fixed dose of 1.5 mg weekly, with sham dose escalation to preserve blinding. The trial was event-driven and planned to continue until at least 1,615 adjudicated MACE events had accrued ^[2].

The population was typical of a secondary-prevention CVOT: mean age 64 years, 29% women, more than 80% White, mean diabetes duration 14.7 years, mean HbA1c 8.4%, and mean BMI 32.6 kg/m². About half of participants were receiving insulin and approximately 30% an SGLT2 inhibitor at baseline ^[1][2]. These characteristics define the population to which the results apply most directly: adults with long-standing type 2 diabetes, overweight or obesity, and established atherosclerotic cardiovascular disease. Extrapolation to primary prevention, leaner phenotypes, older and more multimorbid patients, or people with obesity without diabetes requires additional evidence.

The primary endpoint was time to first occurrence of cardiovascular death, myocardial infarction, or stroke (MACE-3). Non-inferiority was declared if the upper bound of the confidence interval for the hazard ratio was below 1.05; the 95.3% confidence level reflected alpha spent at interim analysis. Superiority required an upper confidence bound below 1.00. Secondary endpoints were placed within a hierarchical testing strategy that was conditional on superiority for the primary endpoint ^[1][2].

3. Principal Findings

The main findings, drawn from the primary publication and the pre-specified exploratory kidney analysis, are summarized in Table 1 ^[1][3]. Estimates marked with an asterisk were outside the confirmatory testing hierarchy and should therefore be interpreted as nominal or exploratory.

Table 1. Key outcomes of SURPASS-CVOT at a median follow-up of 4 years.

* Nominal or exploratory estimates: the hierarchical testing sequence stopped when superiority for the primary endpoint was not achieved. Kidney composite: persistent macroalbuminuria, sustained ≥50% eGFR decline, end-stage kidney disease, or death from kidney disease ^[3].

Tirzepatide produced substantially greater metabolic effects than dulaglutide, with an HbA1c reduction of 1.66% versus 0.88% and a body-weight reduction of 11.6% versus 4.5%, together with greater reductions in systolic blood pressure and triglycerides ^[1]. Gastrointestinal adverse events were more frequent with tirzepatide and contributed to a higher rate of discontinuation because of adverse events (13.2% vs 10.1%). Serious adverse-event rates were similar, and approximately one participant in five in each group permanently discontinued the assigned regimen ^[1].

Subsequent analyses extended the clinical picture. A pre-specified exploratory kidney analysis reported a 23% relative reduction in the composite kidney outcome (6.0% vs 7.6%; HR 0.77, 95% CI 0.68–0.88), with an annual eGFR slope difference of 0.29 mL/min/1.73 m² overall and 0.93 mL/min/1.73 m² in the high-risk CKD subgroup ^[3]. A post hoc analysis using a six-component cardiorenal composite reported an HR of 0.84 (95% CI 0.79–0.90), with the kidney composite at 0.79 (0.68–0.91) and heart-failure hospitalization at 0.96 (0.79–1.17) ^[4]. These analyses are clinically informative but occupy a different inferential level from the primary confirmatory endpoint.

4. Methodological Limits and Interpretive Boundaries

The following considerations do not invalidate SURPASS-CVOT or diminish its importance. They define the limits of what its design can establish. Several have been discussed by the investigators or in subsequent scientific commentaries ^{[1][5][7][8][9][10]}. Statements about cardiovascular protection, mortality reduction, or kidney benefit should therefore distinguish clearly between the confirmatory primary result and nominal, exploratory, indirect, or post hoc findings.

4.1. Non-Inferiority Margin and the Constancy Assumption

The non-inferiority margin of 1.05 was selected to preserve a substantial proportion of the established cardiovascular effect of dulaglutide ^[2][10]. As in every active-comparator non-inferiority trial, this approach depends on a constancy assumption: the comparator is assumed to retain a sufficiently similar effect in the new trial setting. REWIND and SURPASS-CVOT, however, enrolled clinically different populations. REWIND included a majority of participants without established cardiovascular disease, whereas SURPASS-CVOT was entirely a secondary-prevention trial and was conducted in an era of substantially greater use of SGLT2 inhibitors and other cardioprotective therapies ^[2][11]. A subsequent pre-specified indirect comparison combining SURPASS-CVOT and REWIND estimated an HR of 0.72 (95% CI 0.55–0.94) for tirzepatide versus an imputed placebo ^[12]. This analysis is informative and biologically plausible, but it remains an indirect cross-trial estimate rather than a randomized placebo comparison within SURPASS-CVOT.

4.2. Comparator Dose and Treatment-Intensity Asymmetry

Tirzepatide was escalated to 15 mg (or the highest tolerated dose), whereas dulaglutide was maintained at 1.5 mg, the dose used in REWIND and the dose with established cardiovascular-outcome evidence ^[1][2]. Higher dulaglutide doses of 3.0 and 4.5 mg were already available and have been shown to produce greater glycaemic and weight reductions than 1.5 mg ^[6]. The comparator choice is therefore defensible from an outcomes-evidence perspective, but it introduces an asymmetry in treatment intensity for metabolic endpoints. SURPASS-CVOT can establish the cardiovascular comparison between the tested regimens; it cannot determine how much of the observed HbA1c and weight difference would persist in a dose-intensified dulaglutide comparison, nor whether such dose intensification would alter cardiovascular outcomes.

4.3. Hierarchical Testing and Nominal Secondary Endpoints

SURPASS-CVOT was designed and powered primarily for non-inferiority. Because superiority for MACE-3 was not achieved (p = 0.09), the pre-specified hierarchy did not progress to formal confirmatory testing of secondary endpoints ^[1]. Consequently, findings such as the reduction in all-cause mortality (HR 0.84; 95% CI 0.75–0.94) are clinically important signals but are not protected by the planned multiplicity-control procedure. This distinction is substantive rather than semantic: a nominally significant secondary endpoint can be consistent, biologically credible, and worthy of further study while still falling short of confirmatory evidence under the trial’s statistical architecture ^[1][8][10].

4.4. Interpreting the Mortality Signal

The observed difference in all-cause mortality was driven largely by non-cardiovascular deaths, whereas cardiovascular death itself was not significantly reduced (HR 0.89; 95% CI 0.77–1.02) ^[1]. This pattern raises potentially important hypotheses involving metabolic reserve, inflammation, infection, respiratory disease, malignancy, frailty, or other pathways ^[7][8]. However, SURPASS-CVOT was not designed to test a specific mechanism for non-cardiovascular mortality, and the available cause-of-death information is insufficient to establish one. The mortality finding therefore warrants mechanistic and phenotype-specific investigation rather than a single causal interpretation.

4.5. Metabolic Superiority and Atherothrombotic Outcomes

Tirzepatide produced approximately twice the HbA1c reduction and substantially greater weight loss than dulaglutide, yet the MACE-3 hazard ratio was 0.92 ^[1]. Several interpretations remain compatible with this result. Cardiovascular benefit from incretin-based therapy may become progressively less incremental once an effective GLP-1 receptor agonist is already present; alternatively, MACE-3 may be relatively insensitive to benefits expressed through kidney disease, heart failure, liver disease, functional status, or non-cardiovascular mortality ^[5][7][8]. Duration of follow-up may also matter when translating metabolic change into atherosclerotic events. What the trial does not support is a simple linear assumption that a larger reduction in HbA1c or body weight must produce a proportionally larger reduction in MACE-3.

4.6. Post-Randomization Co-Interventions

The investigators reported imbalances in post-randomization treatment, including initiation of SGLT2 inhibitors ^[1]. In a blinded trial, differential co-intervention can arise because clinicians respond to observable differences in glycaemic control, weight, blood pressure, kidney status, or other clinical features. If additional cardioprotective therapy was initiated more often in one treatment group, it could attenuate the between-group cardiovascular contrast. The intention-to-treat analysis remains the appropriate estimate of the randomized treatment strategy under actual trial conduct, but it does not isolate a pure pharmacological comparison between tirzepatide and dulaglutide in the absence of subsequent treatment changes. The magnitude, timing, and clinical consequences of these co-interventions therefore remain relevant to interpretation.

4.7. Kidney Outcomes: Exploratory and Partly Surrogate-Driven

The kidney analysis was pre-specified but exploratory. Its composite endpoint included persistent macroalbuminuria as well as sustained ≥50% eGFR decline, end-stage kidney disease, and death from kidney disease ^[3]. In the low-to-moderate-risk CKD population, the observed benefit was primarily driven by a lower incidence of new persistent macroalbuminuria, whereas the high-risk CKD subgroup showed a more substantial difference in eGFR decline ^[3]. This distinction matters because albuminuria is clinically relevant but remains a surrogate outcome, while hard kidney events were less frequent. The overall eGFR slope difference of 0.29 mL/min/1.73 m² per year was modest; the 0.93 mL/min/1.73 m² difference in the high-risk CKD subgroup is more clinically compelling but derives from a smaller subgroup. The post hoc six-component cardiorenal composite adds supportive information, but by combining mortality, revascularization, heart failure, and kidney outcomes it also combines events with different clinical weight and mechanisms ^[4].

4.8. Absolute Effects, Generalizability, and Sponsorship

Because SURPASS-CVOT did not include a placebo group, it does not directly provide an absolute risk reduction for tirzepatide versus standard care. It provides a 0.9-percentage-point absolute difference in MACE-3 over approximately four years relative to dulaglutide, corresponding to an approximate number needed to treat of 110, and a nominal 1.6-percentage-point absolute difference in all-cause mortality, corresponding to an approximate NNT of 63 ^[1]. These figures should be interpreted in the context of an active comparator already known to reduce cardiovascular risk. Generalizability is also limited by the study population, which was 71% male, more than 80% White, and restricted to participants with BMI ≥25 kg/m² and established atherosclerotic cardiovascular disease ^[1][2]. Finally, the trial was funded by the manufacturer of tirzepatide, as is common for large regulatory outcome trials. Sponsorship does not invalidate randomized evidence, but transparent attention to design choices, endpoint hierarchy, analysis plans, and the framing of secondary results remains appropriate in the appraisal of any industry-funded trial.

5. What SURPASS-CVOT Established

The strongest conclusion is straightforward. In a large, long-duration, double-blind active-comparator trial, tirzepatide was non-inferior to dulaglutide 1.5 mg for MACE-3 in patients with type 2 diabetes and established atherosclerotic cardiovascular disease ^[1]. Tirzepatide also produced substantially greater and durable improvements in HbA1c and body weight, with a safety profile dominated by gastrointestinal intolerance and without a new major safety signal. Across mortality, kidney, revascularization, and broader cardiorenal analyses, the direction of effect was generally favorable to tirzepatide, although the inferential status of these findings ranges from nominal to exploratory or post hoc ^[1][3][4]. SURPASS-CVOT also demonstrated the feasibility of an active-comparator CVOT in a therapeutic field in which placebo-controlled cardiovascular testing is becoming progressively less informative.

6. Stay Hungry, Stay Foolish: Questions That Remain

A positive trial should define what is known while making the remaining uncertainty more precise. In that sense, SURPASS-CVOT is not the end of the cardiovascular question for dual incretin agonism; it identifies the next questions that can now be asked more clearly.

First, the comparison most relevant to contemporary clinical practice is increasingly tirzepatide versus semaglutide rather than tirzepatide versus dulaglutide 1.5 mg. A 2026 target-trial emulation in adults with type 2 diabetes and atherosclerotic cardiovascular disease found similar rates of modified MACE among tirzepatide and semaglutide initiators (HR 1.03; 95% CI 0.90–1.17) ^[9]. This real-world evidence is valuable, but its observational design cannot establish cardiovascular equivalence or substitute for a randomized head-to-head cardiovascular outcome trial. In the absence of such a trial, the relative cardiovascular positioning of tirzepatide and semaglutide continues to rely on indirect comparisons and observational evidence.

Second, the non-cardiovascular mortality signal requires decomposition by cause of death, baseline phenotype, frailty, comorbidity burden, and achieved metabolic response. If the signal is reproducible, understanding its mechanism may be more informative than simply repeating the aggregate all-cause mortality estimate. If it is not reproducible, identifying why it emerged in SURPASS-CVOT will be equally important.

Third, the relationship between weight loss and clinical outcome should increasingly be studied through body composition rather than body weight alone. Percentage weight loss cannot distinguish fat mass from lean mass and cannot determine whether equivalent weight reduction has the same prognostic meaning across different baseline BMI, age, sex, or frailty phenotypes. Observational cohorts with serial body-composition assessment may help address questions that conventional CVOTs were not designed to answer.

Fourth, the endpoint architecture of future incretin trials deserves reconsideration. MACE-3 remains essential, but a time-to-first atherothrombotic composite may not fully capture the clinical effects of drugs whose actions extend across kidney disease, heart failure, liver disease, body composition, functional status, and mortality. Multistate, recurrent-event, or integrated cardiorenal-metabolic approaches have therefore been proposed as complementary strategies ^[7][8]. SURPASS-CVOT illustrates why endpoint selection can shape the apparent magnitude and location of treatment benefit.

Fifth, important population gaps remain. Evidence is still limited in older, leaner, frailer, and more multimorbid patients with type 2 diabetes, as well as in people with obesity without diabetes. These groups may differ in competing risk, body-composition response, tolerability, and the balance between metabolic improvement and treatment burden. The results of SURPASS-CVOT should therefore inform, rather than replace, dedicated evidence in these populations.

The phrase “stay hungry, stay foolish” captures the methodological attitude that these unresolved questions require. “Hungry” means treating a statistically successful trial as a source of further testable questions rather than as the end of inquiry. “Foolish” means remaining willing to question assumptions that have become conventional when the observed data do not fit them neatly. SURPASS-CVOT established that tirzepatide is not inferior to dulaglutide for MACE-3 and strongly favors tirzepatide on multiple metabolic outcomes. It also leaves a productive tension: a large metabolic separation produced only a modest difference in the primary atherothrombotic endpoint, while some of the most interesting signals appeared elsewhere. Explaining that tension, rather than smoothing it away, is likely to be one of the most useful contributions of the next generation of cardiovascular and cardiometabolic studies.

This entry is adapted from: https://doi.org/10.1056/NEJMoa2505928