Drugs in the Ozempic Family Slowed Heart Plaque More Than Other Diabetes Medicines, Real-World Scan Study Finds

GLP-1 receptor agonists, the class that includes semaglutide (Ozempic, Wegovy) and liraglutide, were the only non-insulin diabetes medicines tied to slower coronary plaque buildup in a new 880-scan comparison from the TOCCATA registry, published in La Radiologia Medica and indexed on PubMed. SGLT2 inhibitors, a class often credited with heart-protective effects, showed no plaque benefit in the same comparison.

The paper is a retrospective look at 880 serial coronary CT angiography (CCTA) scans from people with type 2 diabetes (T2DM) treated at one center. CCTA is the non-invasive X-ray scan that maps the fatty, calcium-rich buildup, called atherosclerotic plaque, inside the arteries that feed the heart. Researchers tracked how those plaques changed scan to scan in patients on five non-insulin drug classes: metformin (357 patients), SGLT2 inhibitors (258), DPP-4 inhibitors (97), GLP-1 receptor agonists (88), and TZDs (80). They used two endpoints: vessel-level stenosis progression (any artery segment narrowing further) and patient-level plaque scores, including MAXS for the worst lesion, SIS for the total integrated plaque burden, and SSS for a summed segment score.

GLP-1 receptor agonists came out ahead on both. The adjusted hazard ratio for vessel-level stenosis progression was 0.68 (95% confidence interval 0.49 to 0.95, p=0.024), meaning a 32% lower relative hazard of a segment getting worse. Patient-level signals pointed the same way: HR 0.34 for MAXS progression (p=0.002), 0.34 for SIS (p<0.001), and 0.41 for SSS (p<0.001). The consistency across four separate plaque metrics is what makes the class-level signal harder to dismiss as a one-score fluke.

SGLT2 inhibitors, the family that includes empagliflozin (Jardiance) and dapagliflozin (Farxiga), told a quieter story. Their vessel-level stenosis HR was 0.99 (95% CI 0.82 to 1.19, p=0.90), essentially null. The patient-level SSS score did trend toward worse plaque (HR 1.23, 95% CI 1.00 to 1.52, p=0.055), and the lower bound of that interval just clips 1.0, so the paper's authors call it borderline rather than clean. Either way, the class did not replicate the GLP-1 result on the same imaging endpoint.

DPP-4 inhibitors, an older and cheaper class that includes sitagliptin (Januvia), trended in the opposite direction: HR 1.31 for vessel-level stenosis progression (95% CI 1.01 to 1.69, p=0.039). Metformin and TZDs (pioglitazone and rosiglitazone) were neutral on every endpoint the team measured.

Semaglutide is one of several GLP-1 receptor agonists in the comparison, and the 88-patient arm is modest. CCTA plaque metrics are also surrogate endpoints: they track what a scan sees, not whether a patient has a heart attack, needs a stent, or dies from cardiovascular disease. The paper does not, and cannot, show that any of these drugs prevent those hard events. Prior SGLT2 inhibitor cardiovascular outcome trials, including EMPA-REG and DAPA-HF, and SELECT-style GLP-1 imaging work sit alongside, not inside, this study, and some of those earlier trials pointed in different directions on different endpoints.

In real-world practice, on the same imaging surrogate, the GLP-1 receptor agonist class is the only one of the five tied to slower coronary plaque progression. For a patient whose dominant risk is coronary artery disease rather than weight, sugar, or heart failure, that is a more usable data point than the broader "good for the heart" framing usually attached to these drugs.

The TOCCATA registry is single-center and observational, so confounding by indication is plausible: the patients who ended up on a GLP-1 receptor agonist may have differed in ways the multivariable adjustment did not fully capture. Replication in a multi-center, ideally randomized cohort is the open question, and the authors flag it as such.