A post hoc analysis of SURMOUNT-1, published in the European Journal of Preventive Cardiology, reports that tirzepatide was associated with a lower predicted 10-year cardiovascular disease risk than placebo over 176 weeks in adults with obesity or overweight and prediabetes. Model-derived hazard…
# Tirzepatide associated with lower predicted 10-year CVD risk over 176 weeks
A post hoc analysis of the SURMOUNT-1 trial, published in the European Journal of Preventive Cardiology, reports that tirzepatide was associated with a lower predicted 10-year cardiovascular disease risk than placebo over 176 weeks in adults with obesity or overweight and prediabetes.
What is new in this analysis is not the claim that tirzepatide improves individual risk factors. That was already established. The new material is durability and method. The analysis assesses whether the predicted risk reduction persists across roughly three years, and it validates the prediction tool itself, checking the Framingham Heart Study equation's estimate of treatment effect against observed cardiovascular events in the REWIND trial.
That two-part structure carries most of the paper's weight. It claims not only that tirzepatide moved the inputs of a risk equation, but that the equation used to measure the benefit has a track record of tracking real events in the same drug class. Event-driven cardiovascular outcome trials require very large samples and many years of follow-up. In their absence, a validated risk equation applied to data from a completed trial is one of the few ways to project a clinical effect from surrogate changes.
The authors concluded that tirzepatide was associated with a lower predicted 10-year cardiovascular disease risk versus placebo over three years, with confirmation required in an event-based cardiovascular outcomes trial such as SURMOUNT-MMO . Predicted risk is model-derived, not a count of heart attacks and strokes, and that distinction is central to how the findings should be read.
The analysis compared predicted 10-year CVD risk at two time points. At week 72, the first point of comparison, tirzepatide 15 mg was associated with an absolute risk reduction of 1.92% in predicted 10-year risk, while placebo was associated with an absolute risk increase of 1.10%, a difference significant at p < 0.001 with a model-derived hazard ratio HR = 0.73 .
At week 176, tirzepatide-treated participants retained an absolute risk reduction of 1.31%, while placebo participants showed a larger absolute risk increase of 3.84%. The difference was again significant at p < 0.001, and the model-derived hazard ratio improved HR = 0.62 .
The pattern across the two time points rewards careful reading. At week 72 the absolute separation between the 15 mg arm and placebo was 3.02 percentage points 1.92 plus 1.10 . By week 176 the separation between the tirzepatide group and placebo had grown to 5.15 percentage points 1.31 plus 3.84 . Randomization supports the assumption that the arms began from comparable baseline risk, so the widening gap reflects two forces: a maintained reduction in the tirzepatide group and continued accrual in the placebo group. The treatment arm's absolute reduction actually narrowed slightly between the two time points, from 1.92% to 1.31%, while the placebo arm's increase more than tripled, from 1.10% to 3.84%. The growing separation is therefore driven predominantly by what happened to untreated participants, not by deepening benefit in treated ones.
The hazard ratio improved from 0.73 to 0.62 over the same interval, meaning the relative gap widened even as the absolute gap shifted. An HR of 0.62 corresponds to a 38% lower predicted risk on the relative scale. That is the kind of number that can dominate a headline, and it needs the absolute figures beside it. A 1.31 percentage point reduction in predicted 10-year risk is the magnitude of effect associated with established primary prevention therapies, and it is a statement about predicted risk, not about counted events.
The two time points are not reported symmetrically. The week 72 result is specific to the 15 mg dose, while the week 176 result is reported for tirzepatide-treated participants generally. The distinction matters because dose response remains a live question in this dataset. The week 72 finding establishes that the effect exists at the highest dose; the week 176 finding shows persistence across the tirzepatide group as a whole. How the 5 mg and 10 mg doses compare with 15 mg on predicted risk at either time point is not reported, and the analysis leaves that question open.
SURMOUNT-1 was a phase 3, double-blind, randomised controlled trial in which 2,539 participants were randomised to placebo or tirzepatide 5, 10, or 15 mg for 176 weeks. Of the randomised participants, 1,032 had prediabetes and 976 had no pre-existing cardiovascular disease. The post hoc analysis included 962 participants who had a baseline and at least one post-baseline predicted CVD risk score. The analytic subset is therefore a derived population, not the full trial cohort, and it was selected through criteria that make sense for a primary prevention analysis: prediabetes, no established CVD, and sufficient data to compute the score.
The outcome was not observed cardiovascular events. Predicted 10-year cardiovascular disease risk was calculated using the Framingham Heart Study equation , which converts an individual's risk factor profile into a predicted probability of a first cardiovascular event within a decade. The classic risk factors that feed the score are:
In the standard Framingham formulations these inputs are joined by age, sex, and smoking status, and the score is calibrated against the events it is meant to anticipate: heart attacks and strokes.
The risk-equation-predicted hazard ratio for the treatment effect was then compared with observed cardiovascular events in REWIND, the dulaglutide cardiovascular outcomes trial in type 2 diabetes. In REWIND, the Framingham-based predicted hazard ratio HR = 0.83 was consistent with the observed cardiovascular event hazard ratio HR = 0.85 . The agreement is close enough to matter. Fed with dulaglutide-induced changes in blood pressure, body weight, lipids, and diabetes status, the equation projected nearly the entire event reduction that REWIND actually observed. On this evidence, the classic risk factors captured most of dulaglutide's measured cardiovascular benefit.
That agreement is the paper's methodological anchor, and it deserves scrutiny. The risk equation was developed decades before incretin therapies existed, and it was designed to predict events in individuals, not to estimate drug effects. Using it to generate a predicted treatment-effect hazard ratio presupposes that changes in classic risk factors capture most of the cardiovascular benefit of the drug in question. REWIND supplies the test: dulaglutide changed blood pressure, body weight, lipids, and diabetes status, and the equation said those changes should have produced a hazard ratio of 0.83. The observed hazard ratio for cardiovascular events was 0.85. The two numbers sit within 0.02 of each other.
What the REWIND check does not establish is equally important. It is one test, in one molecule, and that molecule is not tirzepatide. The validation is indirect, using dulaglutide data to stand in for a direct tirzepatide benchmark that does not yet exist. It shows that the Framingham approach can track the event effect of an incretin peptide. It does not show that it will track every incretin peptide, and it leaves open the possibility that a dual agonist with a different receptor profile works partly through pathways the equation does not measure.
The mechanics of the analysis explain both its utility and its limits. At baseline, at week 72, and at week 176, each participant's risk factor values were entered into the Framingham equation to produce a predicted 10-year risk score. The between-arm comparison of those scores is the basis of the absolute risk reduction and the model-derived hazard ratio.
The absolute risk reduction is straightforward: the change in predicted risk from baseline within each arm, expressed as the difference between treatment and placebo. The model-derived hazard ratio is a different object. Because no participant was followed for the full 10-year horizon within the trial, the analysis treats the predicted risk score as a pseudo-outcome and fits a proportional hazards model to it. The resulting hazard ratio of 0.73 at week 72 and 0.62 at week 176 is therefore a single-number summary of the between-arm difference in predicted risk, not a measure of observed events. It is not a license to tell a patient that tirzepatide cut their chance of a heart attack by 38%. The events the score points to are real, but the analysis has not counted them.
The distinction between patient-level and drug-level prediction is worth making explicit. A risk equation is calibrated to answer one question: given this patient's risk factors, what is the probability of an event within 10 years? Applying it to trial data asks a second question: how much of this drug's effect on events is explained by the risk factors the equation measures? The REWIND result is evidence, for one drug, that the second question can be answered with reasonable accuracy using the Framingham equation. It is not evidence that the equation captures unmeasured pathways. A therapy that prevents events mainly through effects on inflammation, thrombosis, or plaque stability could produce a flat predicted-risk curve while reducing real events, which is the failure mode the REWIND validation is designed to exclude for the incretin class.
The time scales also deserve attention. The trial ran for 176 weeks, about three years, while the risk score describes a 10-year horizon. The survival model must bridge that gap by assuming the relative treatment effect on the predicted scale is stable over the remaining years. That is an assumption, not a measurement. The authors' own language, "predicted" risk and "model-derived" hazard ratio, keeps the distinction visible.
Tirzepatide is a synthetic peptide that acts as a dual agonist at the glucose-dependent insulinotropic polypeptide GIP receptor and the glucagon-like peptide-1 GLP-1 receptor . GLP-1 receptor agonism enhances glucose-dependent insulin secretion, suppresses glucagon, slows gastric emptying, and reduces appetite through central pathways. GIP, long understood as an incretin hormone, is now thought to amplify and extend these effects, and the combination produces greater weight loss than selective GLP-1 agonists in head-to-head trials.
Weight loss is the most visible driver of the predicted risk reduction, but it is not the only one. Tirzepatide lowers blood pressure, improves lipid profiles, and slows progression from prediabetes to type 2 diabetes, which is itself a major cardiovascular risk factor. Because the Framingham equation treats diabetes as a categorical input, present or absent, the difference between arms can be driven not only by gradations in weight, blood pressure, and lipids, but by whether participants crossed the threshold into diabetes. A cohort enrolled with prediabetes sits close to that threshold, which makes the diabetes input potentially decisive in the score's behavior. Each of these effects maps directly onto an input of the risk equation: blood pressure, diabetes, body weight, and blood lipids, the classic risk factors named in the analysis. That mapping is what allows the Framingham score to register a treatment effect at all.
The time course helps explain the durability finding. Weight loss in incretin trials typically plateaus within the first year, yet the predicted-risk separation here widened from week 72 to week 176. Several mechanisms can produce that pattern.…
Peptides referenced: Tirzepatide, Dulaglutide, Glucagon, GLP-1.
Related reading: Lilly Confirms Q1 2027 Retatrutide BLA After Phase 3 Wins, Patent Argument Offers Rationale, Not Evidence, for Unapproved Peptides, ALXN2420 Phase 1: GH receptor peptide antagonist lowers IGF-1, Peptide Nanostructures Kill Resistant Bacteria and Restore Meropenem.