A real-world Athens cohort started 62 adults who had type 2 diabetes and a fatty liver on one of three drugs and held them on it for a year. The GLP-1 receptor agonist dulaglutide moved a fine-grained measure of cardiac muscle function by 22.6 percent. The SGLT-2 inhibitor dapagliflozin, in the same prescription orbit, moved the same measure by 8.5 percent. A DPP-4 inhibitor, an older diabetes class, moved it 5.9 percent. The P value on the three-way split was 0.015.
The measure is left ventricular global longitudinal strain. It tracks how much the long axis of the heart shortens during a beat, and is a more sensitive index of cardiac mechanics than ejection fraction, the number cardiologists usually quote on a discharge summary. A patient with a normal ejection fraction can still have an abnormal strain, and the strain reading will degrade earlier in the natural history of diabetic cardiomyopathy. Twenty-two percent is a real movement in this metric.
The cohort was reported in the Journal of Hypertension ↗ in April, with a print issue in July. The group at Attikon University Hospital in Athens has been running this glycocalyx-and-strain methodology on cardiometabolic cohorts for the better part of a decade. The new analysis pulled consecutive patients with type 2 diabetes plus metabolic-dysfunction-associated steatotic liver disease, the renamed version of what cardiometabolic clinicians used to call NAFLD, and matched them by propensity score across three drug arms. Twenty-one started on dapagliflozin, 21 on dulaglutide ↗, and 20 on a DPP-4 inhibitor. Both baseline and 12-month measurements ran through the same imaging suite.
All three arms got the diabetes basics. HbA1c dropped, body mass dropped, both at P less than 0.001. All three arms also moved the same secondary panel in the same direction. Central systolic blood pressure fell, arterial stiffness (measured as pulse wave velocity) fell, the endothelial glycocalyx thinned by less (a marker called the perfused boundary region), liver fat fell, liver stiffness fell, coronary flow reserve rose, and strain rose. Every shift held at P less than 0.01 across the cohort. The story is not in the direction. It is in the size of each lift in each arm.
Dulaglutide pulled ahead on cardiac strain three-way, 22.6 percent versus 8.5 percent versus 5.9 percent. It also pulled ahead on the glycocalyx marker (P=0.037). On coronary flow reserve, the measure of how much the small heart vessels can dilate under stress, dulaglutide and dapagliflozin tied. Both separated from the DPP-4 inhibitor. Neither separated from the other. Liver fat reduction was similar across all three arms; the place where the GLP-1 receptor ↗ and SGLT-2 mechanisms diverged was the cardiac mechanics readout, not the liver one.
A second result inside the result. The drop in central systolic blood pressure tracked with the improvement in glycocalyx, arterial stiffness, coronary flow, and strain at 12 months, all at P less than 0.01. The drop in liver fat tracked with the glycocalyx and strain numbers separately, at P less than 0.05. The blood-pressure-to-vascular-function axis and the liver-fat-to-vascular-function axis both held inside the cohort, which gives the headline GLS finding some mechanistic ballast: the drugs that lower central pressure more, and the drugs that strip more liver fat, are the same drugs that pull cardiac strain higher.
The numbers are small. Sixty-two patients across three arms puts each arm in the low twenties. The design was unblinded and real-world. Propensity score matching is the tool clinicians reach for when randomization is not possible, and it carries the usual unmeasured-confounder warning. The follow-up is one year, which is short for endpoints that drift over a decade. The Athens group has published similar microvascular-endothelium readouts using the same glycocalyx and PBR methodology on at least three prior cardiometabolic cohorts, so the technique is in their hands. The question is whether 21 patients per arm picks up an effect that a multi-center registry would also see, or whether it picks up the local idiosyncrasies of one clinic's referral pattern.
The clinical context is moving fast. The American Diabetes Association, the European Association for the Study of Diabetes, and the AASLD MASLD guidance all now list GLP-1 receptor agonists and SGLT-2 inhibitors as preferred second-line agents for type 2 diabetes when MASLD or established cardiovascular disease is in the picture. DPP-4 inhibitors remain in the formulary but get fewer label expansions per cycle. The comparative case for picking dulaglutide over dapagliflozin in that population has usually been made on weight loss and HbA1c. This readout pushes the case onto cardiac mechanics for the first time, in a population where structural cardiac change is what the clinician is trying to slow.
The press release on this paper has not circulated. The trial would not have been registered because it was a real-world cohort, not an interventional study; the patients started the drug their clinic recommended, the metrics were measured before and after. The number that should travel is the three-way split on strain, not the conclusion line of the abstract that says both drug classes improve hemodynamics. They both did. One did more.