The study drew on administrative claims data from Optum’s deidentified Clinformatics Data Mart Database, which captures health records from large commercial and Medicare Advantage plans spanning all 50 US states. Researchers identified adults with confirmed diagnoses of both type 2 diabetes and established atherosclerotic cardiovascular disease who initiated oral semaglutide or another noninsulin glucose-lowering therapy between October 2019 and April 2024. After propensity score matching to balance baseline characteristics, the analysis compared 10,878 new users of oral semaglutide against 28,639 users of other noninsulin therapies, and conducted additional matched comparisons against specific drug classes, including dipeptidyl peptidase 4 inhibitors and sodium-glucose cotransporter 2 inhibitors. An as-treated approach was the primary analysis, with intention-to-treat sensitivity analyses confirming robustness.

The investigators assessed a battery of cardiovascular endpoints, ranging from the classic 3-point definition of major adverse cardiovascular events—ischemic stroke, myocardial infarction, and cardiovascular-related death—to broader composites that add hospitalization for unstable angina and heart failure, as well as modified versions that substitute all-cause death. Death was categorized as cardiovascular-related when claims showed a cardiovascular primary diagnosis within 30 days before death, while other events required an inpatient claim with a qualifying ICD-10-CM diagnosis code in the primary position. This multi-layered endpoint strategy matters because composite measures aggregate more events than their individual components, providing greater statistical power to detect treatment differences in a real-world population where follow-up periods are often short.

The headline finding was striking. Compared with users of other noninsulin glucose-lowering therapies, new users of oral semaglutide had a 17 percent lower risk of 3-point major adverse cardiovascular events and a 21 percent lower risk of the modified 3-point version. The benefits extended across the endpoint spectrum: oral semaglutide initiation was associated with a 26 percent lower risk of 5-point major adverse cardiovascular events, a 27 percent lower risk of the modified 5-point measure, and a 16 percent lower risk of 2-point events. Individually, oral semaglutide users experienced 23 percent lower risk of ischemic stroke and 29 percent lower risk of all-cause death, though the differences in myocardial infarction and cardiovascular-related death did not reach statistical significance in this dataset.

The class-specific comparisons sharpened the picture. Against matched users of dipeptidyl peptidase 4 inhibitors, each group comprising 7,218 patients, oral semaglutide users had a 22 percent lower risk of 3-point major adverse cardiovascular events, a 24 percent lower risk of the modified version, a 39 percent lower risk of ischemic stroke, and a 36 percent lower risk of all-cause death. Against sodium-glucose cotransporter 2 inhibitor users, the comparison was somewhat narrower, with 7,491 oral semaglutide users matched against 21,572 SGLT2 inhibitor users, yet oral semaglutide still delivered a 21 percent lower risk of 3-point events, a 22 percent lower risk of modified 3-point events, and significant reductions across the broader 5-point and 2-point composites. These are notable gains considering SGLT2 inhibitors themselves carry established cardiovascular benefits, meaning oral semaglutide outperformed an already cardioprotective comparator.

The economic implications were also meaningful. Oral semaglutide users accumulated fewer hospitalizations and lower overall medical costs than their counterparts on other therapies. Specifically, they had 19 percent fewer all-cause inpatient visits, 19 percent lower all-cause inpatient costs, and roughly $2,927 lower all-cause medical costs per patient per year compared with users of other noninsulin therapies. Against dipeptidyl peptidase 4 inhibitors, the savings were larger: 31 percent fewer all-cause inpatient visits and nearly $4,972 lower annual all-cause medical costs. Emergency room visits and costs were the notable exception, showing no significant differences in most comparisons, suggesting the cost advantages concentrate in inpatient and outpatient care rather than acute unscheduled visits.

The findings dovetail with the SOUL randomized clinical trial, which demonstrated that oral semaglutide was superior to placebo in reducing cardiovascular events among people with type 2 diabetes and established cardiovascular disease or chronic kidney disease, showing a 14 percent reduction in 3-point major adverse cardiovascular events. The real-world cohort had a slightly older average age than the trial population, a higher proportion of women, and broader representation of race, ethnicity, and geography, making it a useful complement to the trial’s controlled conditions. The US Food and Drug Administration has since approved oral semaglutide for cardiovascular risk reduction in adults with type 2 diabetes at high risk, including those without prior cardiovascular events, expanding its potential reach.

What makes oral semaglutide scientifically interesting is the chemistry behind the pill. Peptides like semaglutide are normally destroyed by digestive enzymes and cannot cross the intestinal wall, which is why GLP-1 receptor agonists have historically required injection. Oral semaglutide overcomes this by co-formulation with sodium N-(8-[2-hydroxybenzoyl]amino)caprylate, known as SNAC, an absorption enhancer that enables localized uptake of the peptide through the stomach lining without disrupting absorption of other molecules. This technological achievement matters clinically because injection aversion remains a real barrier for many patients, and an oral option with comparable cardiovascular protection could help combat therapeutic inertia in diabetes management.

The authors caution that the observational design cannot fully establish causation, and that claims data lack clinical granularity on kidney function, blood pressure, lipids, smoking, and lifestyle factors that may influence cardiovascular risk. Negative-control analyses examining outcomes theoretically unrelated to the drugs, such as breast or prostate cancer and arm or shoulder fractures, found no differences between groups, providing some reassurance against unmeasured confounding. Follow-up was relatively short, averaging roughly nine to ten months, and the study population was limited to the US health care setting. Still, the consistency of the results with the SOUL trial, combined with prior real-world evidence showing cardiovascular benefits of once-weekly injectable semaglutide, builds a coherent case that this class of drugs genuinely protects the heart in high-risk patients, whether delivered by needle or by pill.

Subject of Research: Real-world cardiovascular outcomes of oral semaglutide in adults with type 2 diabetes and atherosclerotic cardiovascular disease.

Article Title: Comparing Cardiovascular Outcomes in New Users of Oral Semaglutide Versus Other Noninsulin Glucose-Lowering Therapies Among Adults with Type 2 Diabetes and Atherosclerotic Cardiovascular Disease

Article References: Tan, X., Liang, Y., Zhong, C., Xie, L., Guevarra, M., Swift, C., & de Havenon, A. (2026). Comparing Cardiovascular Outcomes in New Users of Oral Semaglutide Versus Other Noninsulin Glucose-Lowering Therapies Among Adults with Type 2 Diabetes and Atherosclerotic Cardiovascular Disease. Diabetes Therapy. https://doi.org/10.1007/s13300-026-01919-8

Image Credits: AI Generated

DOI: 10.1007/s13300-026-01919-8

Keywords: oral semaglutide, type 2 diabetes, cardiovascular outcomes, major adverse cardiovascular events, GLP-1 receptor agonist, DPP4 inhibitors, SGLT2 inhibitors, ischemic stroke, real-world evidence, SNAC absorption, Diabetes Therapy, comparative effectiveness

The rationale for repurposing antidiabetic drugs against dementia rests on a concept researchers have provocatively dubbed “type 3 diabetes.” Older adults with type 2 diabetes (T2D) face roughly a twofold increased risk of cognitive impairment compared with people without the disease, and patients with T2D develop Alzheimer’s disease (AD) at higher rates than the general population. At the center of this link sits brain insulin resistance. When neurons stop responding properly to insulin, the consequences cascade through exactly the pathways that define AD pathology: increased production and accumulation of beta-amyloid (Aβ), formation of neurofibrillary tangles through hyperphosphorylation of tau protein, oxidative stress, and escalating neuroinflammation. Studies of brain tissue from patients with AD have even demonstrated inactivation of the insulin-like growth factor 1 receptor and insulin receptor substrates 1 and 2—molecular signatures strikingly similar to the peripheral insulin resistance seen in diabetic patients.

The mechanistic detail is intricate. Chronic hyperglycemia and hyperinsulinemia promote oxidative stress, endothelial damage, and the formation of advanced glycation end products (AGEs) that impair neuronal function, while insulin resistance increases blood–brain barrier permeability, allowing peripheral inflammatory signals to flood the brain, where reactive astrocytes and activated microglia amplify the damage in a self-perpetuating cycle. Disrupted insulin signaling also derails mitochondrial structure and function, choking off the energy metabolism neurons need, and impairs the synthesis and release of neurotransmitters and neurotrophic factors in memory-critical regions. Cross-sectional imaging studies point to structural correlates as well: reductions in gray matter volume and pronounced hippocampal and amygdalar atrophy may account for the memory impairment so often observed in patients with T2DM. Elevated serum levels of inflammatory markers such as interleukin-6 and high-sensitivity C-reactive protein have been linked to increased risk of mild cognitive impairment (MCI), further tightening the association between metabolic dysfunction and eroding cognition.

Against this backdrop, GLP-1 receptor agonists have emerged as the most intensively studied candidates. GLP-1 receptors are expressed in brain regions central to memory and cognition, including the hippocampus—the same territory compromised earliest in AD pathogenesis. Crucially, these drugs can cross the blood–brain barrier. Once inside, they appear to act through multiple converging mechanisms: they re-sensitize insulin signaling by raising PI3K levels, which rescues the pathway and inhibits GSK3β activity; they suppress Aβ-induced excitotoxicity; they reduce Aβ production by inhibiting BACE1 while boosting α-secretase; and they elevate brain-derived neurotrophic factor (BDNF), a molecule that promotes neuronal survival, neurogenesis, synaptic plasticity, and remyelination. Because GLP-1 receptors are also expressed on glial cells, the drugs can dampen neuroinflammation directly by activating the PI3K/Akt pathway and inhibiting NF-κB, thereby lowering pro-inflammatory cytokines such as TNFα, IL-1β, and IL-6 and restraining microglial and astrocyte activation.

Preclinical data have been striking, if not universally consistent. In diabetic rats and 5xFAD mouse models of AD, liraglutide treatment reduced amyloid-β plaque deposition, tamed astrocyte reactivity and microglial activation in the cortex and hippocampus, and prevented synaptic loss. Semaglutide, in animal models, appears to shift microglia from the pro-inflammatory M1 state toward the neuroprotective M2 phenotype, a polarization change correlated with rescued cognition and reduced neuroinflammatory markers such as Iba-1 and glial fibrillary acidic protein. Lixisenatide, an exenatide analogue, prevented Aβ-related synaptic plasticity and spatial memory impairment by blocking Aβ-induced hippocampal GSK3β activation, and decreased both amyloid plaques and neurofibrillary tangles while enhancing long-term potentiation. The review’s authors are careful to note, however, that not all animal studies concur—some models failed to show reduced Aβ accumulation or cognitive gains, differences likely attributable to genetic background, dosing regimens, treatment duration, and the stage of disease at which treatment began.

SGLT2 inhibitors, best known for their renal and cardiovascular benefits, are building their own neuroprotective case. These lipid-soluble drugs cross the blood–brain barrier and engage SGLT1 and SGLT2 co-receptors expressed in the human central nervous system, including the hippocampus, where they help maintain glucose homeostasis and support learning. In db/db mice, SGLT2i treatment improved learning and memory by reducing brain inflammation and oxidative stress while ameliorating neuronal plasticity and mitochondrial dysfunction. Empagliflozin and dapagliflozin increase neurotrophic factors such as BDNF, GDNF, and VEGF, enhance synaptophysin expression, and restore the PI3K/Akt/GSK-3β pathway. In AD animal models, SGLT2is reduce tau phosphorylation and senile plaque density, and they appear to protect neurons from apoptosis by reducing Bax and caspase-3 expression while raising Bcl-2 levels. DPP4 inhibitors add a further layer: linagliptin attenuated Aβ-induced cytotoxicity in human neuronal cells by restoring insulin signaling through increased IRS-1 and Akt phosphorylation, while a novel DPP4 inhibitor, gramcyclin A, produced dose-dependent improvements in spatial learning in triple transgenic mice alongside reduced Aβ and p-tau levels and enhanced brain glucose uptake.

The clinical picture is genuinely mixed—and the review does not shy away from that. Early signals were tantalizing: a phase IIb ELAD study of liraglutide in 204 patients with mild AD dementia missed its primary outcome of change in cerebral glucose metabolic rate, yet scores on the ADAS-Exec composite declined more slowly in treated patients, suggesting the drug was safe and possibly active. A real-world target-trial emulation study found that patients with T2DM treated with semaglutide had a 67% lower risk of a first AD diagnosis over three years compared with insulin treatment. In Parkinson’s disease, exenatide-treated patients showed a five-point advantage on the Mattis Dementia Rating Scale-2 that persisted after drug withdrawal, and a recent meta-analysis of five randomized trials confirmed improvements in both motor and nonmotor symptoms. A large TriNetX cohort study reported that semaglutide or tirzepatide use was associated with significantly reduced dementia (HR 0.63) and ischemic stroke (HR 0.81) compared with other antidiabetic drugs.

Then came the disappointments. The phase III EVOKE and EVOKE Plus trials—which enrolled 1,855 and 1,953 participants respectively across 566 sites in 40 countries to test semaglutide in early-stage symptomatic AD—failed to confirm superiority over placebo in slowing disease progression as measured by the Clinical Dementia Rating–Sum of Boxes score. Mean changes in CDR-SB from baseline to week 104 were nearly identical between semaglutide and placebo groups. Encouragingly, semaglutide did improve AD-related biomarkers, including canonical CSF markers such as p-tau181 and p-tau217 and neuroinflammatory markers such as YKL-40, with changes in the 5–10% range—but these biomarker shifts did not translate into delayed cognitive decline. On the SGLT2i front, however, large cohort data remain favorable: in a study of more than 708,000 patients with T2D, SGLT2i use was associated with substantially lower incidence of overall dementia (2.9% versus 6.7%; adjusted HR 0.77) compared with DPP4 inhibitors, across vascular dementia, AD, and other subtypes, alongside markedly lower all-cause mortality. A separate phase II trial found that empagliflozin lowered CSF tau and modulated immune and inflammatory biomarkers in patients with amnestic MCI or AD without diabetes, and a single-arm study detected reduced brain glutamate and upregulated IGF-1 and insulin signaling proteins in neuronal-origin extracellular vesicles after just 14 days of treatment.

The authors argue that the field’s next steps are clear: rigorously designed randomized controlled trials specifically enrolling patients with AD or other neurodegenerative diagnoses, standardized neurocognitive batteries, molecular and imaging biomarkers, and extended follow-up periods. Promising candidate biomarkers—plasma neurofilament light chain, GFAP, and the p-tau/β-amyloid ratio—could help identify which patients stand to benefit most, but require longitudinal validation in diabetic populations. Combination strategies also merit attention; preliminary evidence suggests that dapagliflozin paired with cognitive behavior training improved cognitive function and quality of life in elderly patients with T2D and MCI, outperforming pharmacological treatment alone. The therapeutic landscape may broaden further still: dual GIP/GLP-1 receptor agonists such as tirzepatide may modulate central insulin signaling, mitochondrial bioenergetics, and synaptic plasticity in ways that selective GLP-1 agonism alone cannot, opening a next-generation chapter in metabolic neuroprotection.

For now, the message to clinicians and the millions of patients living with type 2 diabetes is one of measured hope. The convergence of epidemiology, molecular biology, animal data, and large observational cohorts makes a persuasive case that glucose-lowering drugs could become genuine tools against dementia—and the pharmaceutical industry’s willingness to run massive phase III AD trials with diabetes drugs signals how seriously the hypothesis is now taken. But as the EVOKE results demonstrated, improved biomarkers do not guarantee slowed decline, and heterogeneity among study populations, treatment durations, and outcome measures continues to frustrate definitive conclusions. Whether incretin-based therapies, SGLT2 inhibitors, and DPP4 inhibitors can ultimately earn a place in dementia prevention will depend on the biomarker-driven, well-controlled trials now underway—trials that will determine if protecting the brain’s metabolism is truly the next frontier of neurology.