Cardiovascular safety has shaped development programs for type 2 diabetes and obesity drugs for close to two decades. A cardiovascular outcomes trial's endpoint definition, noninferiority margin, and event-driven sizing set its cost and timeline. They also decide whether the program can support approval, a postmarketing commitment, or a cardiovascular benefit claim on the label. Regulatory and clinical teams planning these programs need to know what FDA and EMA have actually accepted, not only what the guidance documents say.
The analysis below sets the FDA and EMA guidance frameworks side by side with the CVOTs that regulators reviewed in individual diabetes and obesity programs. It covers how MACE was defined and adjudicated, which noninferiority margins were prespecified and accepted, how premarketing and postmarketing evidence requirements were split, and how each agency turned trial results into labeling. Each point cites the regulatory source documents.
Cardiovascular outcomes trials for diabetes and obesity drugs: how FDA and EMA judged MACE endpoints and noninferiority margins
FDA and EMA have both treated cardiovascular (CV) safety as a central question in diabetes and obesity drug development. They have also converged on the same core endpoint: three-point major adverse cardiovascular events (MACE). Where they differ is in how the margins were set and in how a positive CVOT result reaches the label. This article covers the guidance frameworks, the MACE definitions and noninferiority margins regulators accepted in individual programs, and how each agency turned CVOT results into labeling.
Key takeaways
- Standard endpoint. Both agencies accepted time to first 3-point MACE (CV death, nonfatal myocardial infarction, nonfatal stroke) as the primary endpoint in nearly every CVOT reviewed here, with independent adjudication 370172876.
- FDA's 1.8 / 1.3 framework. Under the December 2008 FDA guidance, sponsors had to exclude an upper 95% confidence bound of 1.8 for the CV risk ratio before approval. If the bound fell between 1.3 and 1.8, a postmarketing trial generally had to exclude 1.3 977978. In March 2020, a draft guidance proposed replacing this approach with minimum safety-database requirements 154921922.
- EMA's approach is less prescriptive. The CHMP reflection paper treats an upper bound below 1.8 (when the HR is about 1) as a reasonable planning assumption at initial licensing. It does not present 1.8 as a universal margin 879.
- Margins actually accepted. These ranged from 1.8 (SUSTAIN 6, PIONEER 6), to the common 1.3 (LEADER, EMPA-REG OUTCOME, CANVAS, DECLARE, VERTIS CV, SAVOR, EXAMINE, CARMELINA, DEVOTE, EXSCEL, ELIXA, TECOS), to 2.0/1.4 for an obesity drug (LIGHT), to 1.05 for an active-comparator CVOT (SURPASS-CVOT). Some trials were designed for superiority with no margin at all (SELECT) 54376937170624120023.
- Labeling diverges. FDA has granted standalone indications to reduce MACE risk (for example Victoza, Ozempic, Invokana, Wegovy) 374854422952. CHMP has repeatedly declined separate CV-prevention indications in SmPC section 4.1. Instead it broadens the type 2 diabetes wording and places the CVOT results in section 5.1 341996667827203.
The FDA framework
The 2008 guidance and how reviewers applied it
FDA's December 2008 guidance, Diabetes Mellitus: Evaluating Cardiovascular Risk in New Antidiabetic Therapies to Treat Type 2 Diabetes, recommended CVOTs and prospective meta-analyses for new antidiabetic therapies 152153. FDA reviews describe its decision rules this way 977978979:
- Before approval, sponsors compare CV events against a control group. They must show that the upper bound of the two-sided 95% CI for the estimated risk ratio is below 1.8, with a reassuring point estimate.
- If the upper bound falls between 1.3 and 1.8 and benefit-risk supports approval, a required postmarketing trial generally must establish an upper bound below 1.3.
- If premarketing data already exclude 1.3, a postmarketing trial generally may not be needed.
- Events should be prospectively and blindly adjudicated, ideally within a prespecified phase 2/3 meta-analysis. A large safety trial can also supply the preapproval bound, alone or combined with other trials.
In practice, sponsors used two steps. A pooled, adjudicated phase 2/3 meta-analysis excluded 1.8, and a dedicated CVOT then excluded 1.3 994. Program examples:
- The Farxiga meta-analysis pooled 21 trials 982983.
- The Invokana, Steglatro, Tanzeum and Trulicity reviews describe meta-analyses that often used adjudicated MACE+ (MACE plus hospitalization for unstable angina) to test the 1.8 margin 984985986987.
- For Jardiance, the eight-trial meta-analysis included interim CVOT data. FDA said the dedicated CVOT should provide the 1.3 assessment, rather than drawing conclusions from overlapping data sources without multiplicity control 988989.
- For Mounjaro, a seven-trial MACE-4 meta-analysis was used against the 1.8 margin 791.
Not every product was held to this framework:
- Onglyza and Victoza were caught by the guidance after development or submission was already under way. Reviewers acknowledged that retrospective event analyses could not fully replace prospective adjudication 980981.
- The Tresiba review states that injectable insulins were not generally held to the 1.8/1.3 bounds, although FDA still expected CV events to be collected and analyzed 993. DEVOTE nonetheless prespecified a 1.3 margin 649.
The 2020 draft guidance
In March 2020 FDA issued the draft guidance Type 2 Diabetes Mellitus: Evaluating the Safety of New Drugs for Improving Glycemic Control. It proposes withdrawing and replacing the 2008 guidance 154. FDA gave two reasons. First, none of the CVOTs completed since 2008 had found an increased risk of ischemic CV events, and some showed reduced risk 919. Second, a 2018 advisory committee recommended postmarketing requirements tied to program-specific risk signals rather than a one-size-fits-all approach 916.
In place of a fixed risk margin, the draft asks for a minimum safety database 921922:
| Element | Recommended minimum |
|---|---|
| Total phase 3 exposure | 4,000 patient-years |
| Long-term exposure | 1,500 patients for at least 1 year; 500 patients for at least 2 years |
| High-risk subgroups | 500 with stage 3/4 CKD; 600 with established CV disease; 600 older than 65 |
| Overlap target | At least 1,200 patients with at least one of these characteristics |
The draft keeps rigorous adjudication of CV events. Where a specific premarket concern exists, enough events may still be needed to exclude a meaningful degree of risk 923.
Obesity drugs
The January 2025 FDA draft guidance Obesity and Overweight: Developing Drugs and Biological Products for Weight Reduction recommends:
- monitoring cardiometabolic parameters and running a comprehensive CV assessment, generally including ambulatory blood-pressure monitoring 748749
- a CVOT if a CV risk signal emerges, either before or after approval depending on the signal 750
The guidance passages reviewed do not set a MACE endpoint or a numerical risk margin 751.
The EMA framework
Diabetes guideline
Revision 1 of the CHMP Guideline on clinical investigation of medicinal products in the treatment or prevention of diabetes mellitus (29 June 2012) sets out the following:
- MACE definition. The principal MACE are CV death, non-fatal MI and stroke. Hospitalization for unstable angina may be added when the composite's main purpose is to exclude a safety signal 172.
- Adjudication and pooling. CV outcomes should be predefined consistently, centrally adjudicated with blinding, and pooled under a prospective statistical plan 173.
- Two routes to the evidence. One is a meta-analysis, provided the database can detect uncommon serious events. The other, or the expected route when a CV signal is suspected, is a long-term controlled outcome study with at least 18 to 24 months of follow-up 174.
- Timing. An integrated adjudicated MACE analysis is required at submission. A fully powered CV assessment should be submitted before authorization when the mechanism raises an intrinsic concern or a concern emerges during development 175.
- No numerical CV margin. Revision 1 sets no 1.8 or 1.3 CV margin. Its 0.3% noninferiority margin applies to HbA1c efficacy 177.
Revision 2 (CPMP/EWP/1080/00 Rev.2, 2024) points to the separate CHMP reflection paper for CV safety requirements 801803.
CHMP reflection paper on CV safety
The CHMP reflection paper EMA/CHMP/50549/2015 makes these points:
- Preferred endpoint. MACE (CV death, non-fatal MI, non-fatal stroke) is the preferred endpoint for both pooled analyses and dedicated CVOTs. A broader MACE-plus composite is acceptable only if justified in advance as more sensitive to harm, and each component must be reported 876.
- Adjudication. An independent committee should adjudicate all major predefined events 264.
- Sensitivity of the data. Applicants must justify that exposure and follow-up are sufficient. A claim of similar or lower risk should not simply reflect an inability to detect a difference 878.
- Margin. The general planning assumption is an upper 95% CI bound below 1.8 when the estimated HR is about 1. The paper calls this a reasonable basis at initial licensing, not a universal noninferiority margin, and notes that narrower targets may be appropriate 879.
Weight management guideline
The CHMP Guideline on clinical evaluation of medicinal products used in weight management (EMA/CHMP/311805/2014) expects CV safety to be adequately characterized at MAA. It singles out new mechanisms and drug classes with unestablished or questioned CV safety, and it defers to the reflection paper for quantification 262. The final text does not itself mandate a MACE outcome study 252.
MACE definitions regulators accepted
Most programs used classic 3-point MACE. Several accepted variants:
| Variant | Trials | Notes |
|---|---|---|
| Nonfatal ischemic stroke instead of all stroke | SAVOR (saxagliptin), DECLARE (dapagliflozin) | 286144673 |
| MACE+ with hospitalization for unstable angina | ELIXA (lixisenatide), TECOS (sitagliptin) | 464294 |
| MACE-4 / MACE+ in premarketing meta-analyses | Mounjaro, Invokana, Steglatro, Tanzeum, Trulicity programs | 791984985 |
| CV death including death of undetermined cause | SELECT (semaglutide 2.4 mg), SURPASS-CVOT (tirzepatide), as described by CHMP | 22199 |
Noninferiority margins and results by trial
Some trials reported CIs adjusted for interim analyses rather than an unadjusted 95% CI. The table gives these as reported (for example 95.02%, 95.3%, 95.6%).
| Trial (product) | Comparator | Primary endpoint | Prespecified margin | Primary result | FDA outcome | EMA outcome |
|---|---|---|---|---|---|---|
| LEADER (Victoza) | Placebo | 3-point MACE | 1.3, then superiority 371338 | HR 0.87 (0.78 to 0.97); superiority shown 372339 | CV risk-reduction indication in established CVD 374 | Results in 5.1; 4.1 broadened to "treatment of" T2DM 341342 |
| SUSTAIN 6 (Ozempic) | Placebo | 3-point MACE | 1.8 543618 | HR 0.74 (0.58 to 0.95) 544620 | CV risk-reduction indication approved January 16, 2020 854855 | Superiority tested post hoc only; favourable signal, not formal superiority 621622 |
| PIONEER 6 (Rybelsus) | Placebo | 3-point MACE | 1.8 769 | HR 0.79 (0.57 to 1.11); superiority not shown 770771 | Safety finding, no CV indication from this trial 773 | Not reviewed here (the later SOUL CVOT is outside this table) |
| REWIND (Trulicity) | Placebo | 3-point MACE | FDA review: rule out 1.3 442; CHMP: designed for superiority 94 | HR 0.88 (0.79 to 0.99) 44395 | Labeled as significant MACE reduction 443 | Superiority concluded; 5.1 plus 4.1 cross-reference 9799 |
| EMPA-REG OUTCOME (Jardiance) | Placebo | 3-point MACE | 1.3, then superiority 70662 | HR 0.86 (0.74 to 0.99; 95.02% CI per CHMP) 70864 | Benefit attributed to reduced CV death 711 | Separate CV-death indication declined; results in 5.1 66 |
| CANVAS Program (Invokana) | Placebo | 3-point MACE | 1.3 424 | HR 0.86 (0.75 to 0.97); superiority not in prespecified testing strategy 426427 | MACE risk-reduction indication in established CVD 422 | Not reviewed here |
| DECLARE (Farxiga/Forxiga) | Placebo | CV death, MI, ischemic stroke | 1.3 145674 | HR 0.93 (0.84 to 1.03); superiority not shown 145675 | Noninferiority, not MACE reduction 148 | Noninferiority only; 4.1 reworded; separate CV/HF indications based on DECLARE rejected 676678 |
| VERTIS CV (Steglatro) | Placebo | 3-point MACE | 1.3 313488 | HR 0.97 (95.6% CI 0.85 to 1.11) 314489 | Noninferiority statement in labeling 315 | 4.1 broadened despite no MACE superiority 494496 |
| SAVOR-TIMI 53 (Onglyza) | Placebo | CV death, MI, ischemic stroke | 1.3 286 | HR 1.00 (95% CI 0.89 to 1.12) 287 | Safety labeling update; heart-failure warning 284289 | Noninferiority; HF hospitalization flagged 213215 |
| EXAMINE (Nesina) | Placebo, post-ACS patients | 3-point MACE | 1.3 599 | HR 0.96 (upper bound of the one-sided repeated CI 1.16) 600 | Safety evidence 602 | Not reviewed here |
| TECOS (Januvia) | Placebo | 4-point MACE (incl. unstable angina) | 1.3 295 | HR 0.98 (0.88 to 1.09) 296 | Not located in FDA records searched | Noninferiority; results in 5.1 297298 |
| CARMELINA (Tradjenta/Trajenta) | Placebo | 3-point MACE | 1.3 42731 | HR 1.02 (0.89 to 1.17) 43732 | "Cardiovascular Safety Trials" labeling 44 | 5.1 updated, no 4.1 change 738 |
| DEVOTE (Tresiba) | Insulin glargine | 3-point MACE | 1.3 649 | HR 0.91 (0.78 to 1.06) 649 | CV safety, no CV indication 653650 | Not reviewed here |
| EXSCEL (Bydureon) | Placebo | 3-point MACE | 1.3, then superiority 569 | HR 0.91 (0.832 to 1.004); superiority p=0.06 570 | "Did not increase the risk of MACE" 570 | Not reviewed here |
| ELIXA (Adlyxin) | Placebo, post-ACS patients | MACE+ (incl. unstable angina) | 1.3 465 | HR 1.02 (0.89 to 1.17) 466 | Excess risk ruled out 467 | Not reviewed here |
| SURPASS-CVOT (Mounjaro) | Dulaglutide 1.5 mg | MACE-3 | 1.05 (accepted by CHMP) 200 | HR 0.92 (95.3% CI 0.83 to 1.01); superiority not shown 823201 | Noninferior to dulaglutide in labeling 823 | Separate MACE indication declined; 4.1 cross-references 5.1 203 |
| SELECT (Wegovy) | Placebo, CVD with overweight/obesity, no diabetes | 3-point MACE | None; superiority design 95023 | HR 0.80 (0.72 to 0.90) 39624 | MACE risk-reduction indication 952 | No separate indication; results in 5.1 27 |
| LIGHT / NB-CVOT (Contrave/Mysimba) | Placebo | 3-point MACE | 2.0 interim; 1.4 final 241516 | Interim ITT HR 0.59 (0.39 to 0.90) per CHMP 516 | Interim result accepted as safety reassurance only; new CVOT required 244246 | Reassurance for short- to intermediate-term safety only 516517 |
How the agencies used CVOT results
FDA: standalone risk-reduction indications
FDA has been willing to convert a superior CVOT into a separate indication, usually restricted to the population with the strongest evidence.
- LEADER (Victoza). Reviewers accepted a single trial as substantial evidence. They found it large, well conducted, statistically robust and clinically meaningful, with all three MACE components trending favorably 813814815. The resulting indication was narrower than the sponsor proposed: adults with type 2 diabetes and established CV disease rather than "high CV risk" 818819.
- Jardiance. FDA framed the benefit around CV death. The review did not consider the single trial enough to conclude a general reduction in macrovascular events 711712.
- SELECT (Wegovy). FDA's statistical review found no major issues: components were consistent and missing data minimal (3%). CV death, the key secondary endpoint, did not reach superiority, so lower-ranked secondaries could not be called significant 953. The approved indication is MACE risk reduction in adults with established CV disease and obesity or overweight 952.
Noninferior results go into the label as safety statements, not efficacy claims. Examples are the SAVOR, EXAMINE, CARMELINA, DEVOTE and EXSCEL labeling 28460244653570.
EMA: section 5.1 with a broadened section 4.1
CHMP followed a consistent pattern:
- Section 5.1. CVOT results go into section 5.1.
- Section 4.1. The indication changes from "improvement of glycaemic control" to "treatment of" type 2 diabetes, with a cross-reference to section 5.1. CHMP adopted this approach for Victoza, Trulicity and Forxiga 34199678.
- Separate indications declined. CHMP turned down separate section 4.1 CV indications for Jardiance (CV death), Mounjaro (MACE reduction) and Wegovy (CV prevention). Its reasoning was that the studied populations were already covered by the existing indication 6620327.
- Steglatro exception. For Steglatro, CHMP allowed the broadened 4.1 wording even without MACE superiority. It noted that this wording had previously been reserved for products with demonstrated superiority, and justified it on mechanism and class evidence 496.
Methodological issues reviewers raised
- Post hoc superiority. CHMP recorded that SUSTAIN 6 was a 1.8-margin safety trial with no prespecified 1.3-then-superiority sequence. Its nominal superiority p value came from an unpenalized post hoc test 618621. FDA advisory committee members made a similar point: superiority was not a prespecified test 958.
- What drives the composite. Regulators looked at which components carried each result:
- Generalizability and subgroups. CHMP flagged:
- Active-comparator margins. In SURPASS-CVOT, a placebo-anchored 1.3 margin was not appropriate against an active CV-protective comparator. CHMP accepted a 1.05 margin and considered an indirect placebo comparison via REWIND supportive but not definitive 200206.
- Interim data integrity. For Contrave, FDA's statistical reviewer recommended keeping the interim LIGHT findings out of labeling. Because of trial-integrity concerns, FDA concluded that LIGHT could not satisfy the postmarketing requirement to exclude 1.4 245246. CHMP described the interim data as short- to intermediate-term reassurance only, and some members dissented 516524.
- Heart failure as a separate safety signal. SAVOR showed a MACE-neutral result alongside more heart-failure hospitalizations. This led to an FDA heart-failure warning and a CHMP caution 289215.
Implications for development programs
- Make adjudicated 3-point MACE the default primary safety endpoint for both agencies. Justify any MACE-plus or ischemic-stroke variant prospectively 876172.
- For type 2 diabetes, FDA's current draft position relies on safety-database size and composition rather than a fixed margin 921922. EMA defers to the reflection paper's 1.8 planning assumption and case-by-case justification 879.
- To support a CV benefit claim, prespecify superiority testing after noninferiority. FDA will grant a standalone indication on one robust trial 813. In the EU, expect the result to sit in section 5.1 with a broadened section 4.1 rather than a new CV-prevention indication 34127.
- For obesity drugs, neither agency's guidance sets a numerical MACE margin. A CVOT becomes necessary when a mechanism-based or emergent CV signal exists 750262, and the LIGHT precedent shows how interim-analysis integrity affects approval 246.