Abstract
Purpose
Type 2 diabetes mellitus (T2DM) and COVID-19 are independently associated with an elevated risk of dementia, and COVID-19 survivors with T2DM may represent a clinically vulnerable population in whom metabolic, inflammatory, and vascular factors coexist. However, whether antidiabetic treatment choice is associated with incident dementia risk in this population remains unknown.
Methods
In this target trial emulation using the TriNetX US Collaborative Network, we compared COVID-19 survivors aged ≥50 years with T2DM initiating glucagon-like peptide-1 receptor agonist (GLP-1RA) therapy with those initiating dipeptidyl peptidase-4 inhibitor (DPP-4i) therapy between January 1, 2020 and December 31, 2024. We applied an active-comparator, new-user design with a 90-day survivor landmark and 1:1 propensity score matching. The primary outcome was incident all-cause dementia (a composite of vascular dementia, Alzheimer’s disease, and other dementias). Secondary outcomes included dementia subtypes, mild cognitive impairment, glycemic outcomes, and all-cause mortality.
Results
After matching, 25,263 patients per group were followed for up to 6 years (median, 919 days [GLP-1RA] vs 955 days [DPP-4i]). GLP-1RA initiation was associated with a lower hazard of incident dementia (hazard ratio [HR], 0.64; 95% confidence interval [CI], 0.58–0.72), vascular dementia (HR, 0.67; 95% CI, 0.52–0.85), and Alzheimer’s disease (HR, 0.74; 95% CI, 0.59–0.93), as well as more favorable glycemic control and lower all-cause mortality (HR, 0.63; 95% CI, 0.59–0.68). The estimate for mild cognitive impairment was borderline (HR, 0.87; 95% CI, 0.75–1.00) and not consistently supported across sensitivity analyses. The association with incident dementia remained evident after the 6-month landmark analysis (HR, 0.66; 95% CI, 0.58–0.75) and was consistent across sensitivity and subgroup analyses.
Conclusion
Among COVID-19 survivors with T2DM, GLP-1RA initiation was associated with a lower hazard of incident dementia than DPP-4i initiation. These hypothesis-generating results warrant confirmation before they can be used to inform clinical decision-making.
Keywords: glucagon-like peptide-1 receptor agonist, DPP-4 inhibitor, dementia, COVID-19 survivors, type 2 diabetes mellitus, target trial emulation
Introduction
Type 2 diabetes mellitus (T2DM) is a well-established risk factor for dementia, with prior meta-analyses showing an approximately 1.5- to 2-fold higher risk of all-cause dementia among individuals with T2DM than among those without diabetes.1–5 This excess risk is thought to arise from chronic hyperglycemia, insulin resistance, microvascular injury, systemic inflammation, and accelerated neurodegenerative processes.6–10 COVID-19 may further exacerbate these pathways, as SARS-CoV-2 infection has been increasingly recognized as a contributor to long-term neurological and cognitive sequelae.11–13 Large cohort studies and meta-analyses have reported elevated risks of incident cognitive impairment, memory disorders, cerebrovascular disease, Alzheimer’s disease, and vascular dementia after COVID-19.14–16 Together, these observations suggest that COVID-19 survivors with T2DM may constitute a particularly vulnerable population in whom diabetes-related neurodegenerative susceptibility and post-acute viral inflammatory or vascular injury may coexist and contribute to cognitive vulnerability. However, whether these conditions interact to confer an additive or synergistic dementia risk remains uncertain.
Glucagon-like peptide-1 receptor agonists (GLP-1RAs) have attracted growing interest for their potential neuroprotective effects beyond glycemic control.17–19 Current evidence suggests that GLP-1RAs may influence dementia-related pathways through several potentially distinct mechanisms, including improved glycemic control and weight reduction, cardiovascular and vascular benefits, direct effects on neuroinflammation, oxidative stress, and amyloid/tau-related pathways, and attenuation of systemic inflammatory processes.20–24 Recent target trial emulation studies in general or older populations with T2DM have reported lower hazards of incident dementia among GLP-1RA initiators than among dipeptidyl peptidase-4 inhibitor (DPP-4i) initiators.25–27 However, these studies did not specifically address COVID-19 survivors with T2DM, in whom persistent inflammatory and vascular sequelae may create a clinically distinct context for dementia development. The principal evidence gap is therefore whether the previously observed association extends to this clinically distinct post-COVID population. Persistent post-COVID inflammation and vascular injury may therefore provide a biologically plausible context in which the anti-inflammatory, vascular, and neuroprotective actions of GLP-1RAs could be relevant to subsequent cognitive outcomes. However, whether COVID-19 modifies the association between GLP-1RA initiation and dementia remains unknown.
To address this evidence gap, we conducted an active-comparator, new-user target trial emulation using electronic health record data to compare incident dementia among COVID-19 survivors with T2DM initiating GLP-1RA versus DPP-4i. DPP-4i was selected as a clinically relevant comparator because both classes are established glucose-lowering therapies, although they differ in glycemic, weight, cardiovascular, and potentially neurocognitive effects and may therefore be prescribed to systematically different patients. Residual treatment-selection differences may remain despite adjustment for measured baseline characteristics. Our objective was to determine whether the association previously reported in broader T2DM populations was also observed in this clinically distinct post-COVID population, rather than to test whether COVID-19 modifies the treatment association. This post-COVID setting is relevant because metabolic, vascular, and inflammatory vulnerabilities may coexist. The findings are intended to inform future comparative evaluations of diabetes therapies, but the observational design does not support treatment selection based on cognitive outcomes alone.
Methods
Study Design, Data Source, and Ethical Approval
This retrospective active-comparator, new-user cohort study emulated a target trial to evaluate the association between GLP-1RA initiation and incident dementia among COVID-19 survivors with T2DM, using DPP-4 inhibitors as an active comparator. Data were obtained from the TriNetX US Collaborative Network, a federated real-world database of de-identified electronic health records from participating healthcare organizations. TriNetX provides standardized information on diagnoses, procedures, medications, laboratory results, and encounters using common clinical terminology, including ICD-10-CM, RxNorm, and LOINC. As only aggregated, de-identified data were available to investigators, informed consent was waived. Access to and use of the de-identified TriNetX data complied with applicable data protection and privacy regulations and with the data-use requirements governing the TriNetX Research Network. The study protocol was approved by the Institutional Review Board of Chi Mei Medical Center, Tainan, Taiwan, and all procedures followed the Declaration of Helsinki and the TARGET reporting guidance for target trial emulation studies.
Study Population and Exposure Definition
We identified adults aged ≥50 years with T2DM and a recorded COVID-19 diagnosis whose first eligible GLP-1RA or DPP-4 inhibitor prescription occurred between January 1, 2020 and December 31, 2024 in the TriNetX US Collaborative Network. The database query was performed on July 3, 2026. TriNetX does not provide a single uniform calendar data-cutoff date across all contributing healthcare organizations. To emulate cohort entry after the acute infection phase, patients were required to survive for at least 90 days after COVID-19 diagnosis. Among these survivors, the index date or time zero was defined as the first eligible prescription of either a GLP-1RA or DPP-4 inhibitor on or after the 90-day post-COVID-19 landmark. Patients were followed for up to 6 years after the index date.
Patients initiating GLP-1RA therapy were assigned to the GLP-1RA cohort, and those initiating DPP-4 inhibitor therapy were assigned to the active-comparator cohort. A new-user design was applied by excluding patients with any recorded use of the assigned drug class within three years before the index date. The 3-year washout was used to operationalize new use of the initiated drug class. To ensure mutual exclusivity, patients were also excluded if they had used the comparator drug class within 1 year before the index date or were using the comparator drug class concurrently at the index date. The 1-year comparator washout was intended to reduce recent cross-class exposure and, together with exclusion of concurrent use, to establish mutually exclusive treatment strategies at cohort entry. Patients were required to have at least one healthcare encounter during the 5 years before the index date to support baseline covariate ascertainment.
Exclusion Criteria
Patients were excluded if they had evidence of preexisting or potentially undiagnosed cognitive impairment, including dementia, mild cognitive impairment, other major neurocognitive disorders, or use of anti-dementia medications, including donepezil, rivastigmine, galantamine, or memantine, before the index date. We also excluded patients with neurological or psychiatric conditions that could independently affect cognitive outcome ascertainment, including intracranial injury, multiple sclerosis, epilepsy or recurrent seizures, schizophrenia, bipolar disorder and Parkinson’s disease. Additional exclusions included type 1 or gestational diabetes, CKD stages 4–5, end-stage renal disease or dialysis dependence, and prior bariatric surgery. All codes used to define the exclusion criteria are listed in Supplemental Table 1.
Target Trial Emulation Framework
We emulated a hypothetical randomized trial in which eligible adults aged ≥50 years with T2DM, who survived at least 90 days after COVID-19 diagnosis and were free of baseline cognitive impairment, initiated either a GLP-1RA or a DPP-4 inhibitor at time zero. The treatment strategies were initiation of GLP-1RA versus initiation of DPP-4i. Time zero was defined as the first eligible prescription of either drug class on or after the 90-day post-COVID-19 landmark. Treatment assignment was based on the drug class initiated at time zero, and the primary outcome was incident all-cause dementia. Follow-up was specified for up to 6 years after the individual index date. Because randomization was not possible, baseline confounding was addressed using propensity score matching. The primary analysis compared the hazard of incident dementia after GLP-1RA versus DPP-4i initiation under an intention-to-treat–like framework. Accordingly, the estimand reflected treatment initiation rather than sustained treatment exposure throughout follow-up. The full target trial specifications and emulation are summarized in Supplemental Table 2.
Propensity Score Matching
To minimize baseline imbalances between patients who initiated GLP-1RA and those who initiated DPP-4i, we applied 1:1 propensity score matching using a greedy nearest-neighbor approach with a caliper set at 0.1 of the pooled standard deviation of the logit of the propensity score. The propensity score was estimated using a logistic regression model incorporating demographic characteristics, comorbidities, concomitant medications, and laboratory values measured in the 5 years before the index date. Because diabetes duration from disease onset could not be reliably ascertained, diabetes-related complications, HbA1c, and glucose-lowering medications were included as available measures of diabetes severity and chronicity. The full code list used for the covariate definitions is provided in Supplemental Table 3. Covariate balance before and after matching was evaluated using standardized mean differences (SMDs); an absolute SMD < 0.10 indicated an acceptable balance.
Outcomes and Healthcare-Utilization Indicators
The primary outcome was incident all-cause dementia after the index, defined as a composite of vascular dementia, dementia in other diseases, unspecified dementia, and Alzheimer’s disease. Secondary outcomes included vascular dementia and Alzheimer’s disease analyzed separately, mild cognitive impairment, glycemic outcomes (hyperglycemia, hypoglycemia, and HbA1c ≥9%), and all-cause mortality. Any ambulatory healthcare visit during follow-up was assessed as a healthcare utilization indicator to evaluate differential post-index surveillance and retention in the electronic health record. Age-related cataracts were included as a negative control outcome because they are not expected to be directly affected by GLP-1RA or DPP-4 inhibitor initiation. The diagnostic, laboratory, and medication codes used to define all outcomes and indicators are provided in Supplemental Table 4.
Sensitivity Analyses
Sensitivity analyses were performed to assess the robustness of primary findings. Model I was restricted to patients who survived throughout the follow-up period to reduce the influence of differential mortality. Model II included only patients with at least one healthcare visit during the follow-up period to address incomplete post-index observation or differential healthcare retention. Model III excluded patients with previous SGLT2 inhibitor use to reduce confounding factors from co-prescribed agents potentially associated with cognitive outcomes. Models IV and V evaluated GLP-1RA subtype-specific associations by restricting the exposure cohort to semaglutide and dulaglutide users, respectively. Model VI was a switching-restricted sensitivity analysis that excluded patients who initiated the comparator drug class during follow-up. For each model, the GLP-1RA and DPP-4 inhibitor cohorts were redefined, rematched, and analyzed using the same time-to-event framework as in the primary analysis.
Landmark and Descriptive Competing-Risk Analysis
Additional landmark analysis was performed 6 months after the index date to assess whether early dementia ascertainment or differential healthcare contact influenced the primary association between GLP-1RA initiation and incident dementia. The 6-month landmark represented the start of this sensitivity analysis rather than the duration of follow-up. In this analysis, patients who developed dementia or died before the landmark time were excluded, and follow-up was restarted from the landmark date using the same matched time-to-event framework. A descriptive competing-risk analysis was also conducted using the Aalen–Johansen estimator, with incident dementia as the event of interest and death as a competing event. This analysis was performed in the unmatched cohort and was intended to visualize cumulative incidence patterns in the presence of differential mortality. It was descriptive and was not intended as a formal between-group competing-risk hypothesis test.
Subgroup Analyses
Prespecified subgroup analyses were conducted to evaluate whether the association between GLP-1RA initiation and incident dementia differed across clinically relevant strata, including age, sex, CKD, cerebrovascular disease, hypertension, ischemic heart disease, glycemic control, obesity, and COVID-19 severity. Age was categorized as 50–65 years or >65 years, and glycemic control was defined as HbA1c ≥9% versus <9%. COVID-19 severity was classified according to hospitalization status on the date of COVID-19 diagnosis, serving as a proxy for acute illness severity. Within each subgroup, the GLP-1RA and DPP-4 inhibitor cohorts were separately matched.
Statistical Analysis
Baseline characteristics were summarized using means with standard deviations or frequencies with percentages, and covariate balance was assessed using SMDs. Time-to-event outcomes were analyzed using Cox proportional hazards models, Kaplan–Meier curves, and Log rank tests. The proportional hazards assumption was assessed using Schoenfeld residuals within the TriNetX platform, and no significant violations were detected. Missing data were not imputed, and analyses were performed using the available structured data in the TriNetX platform. Patients were not excluded solely because individual laboratory measurements were unavailable, and no complete-case restriction based on laboratory availability was applied. All statistical tests were two-sided, and p-values below 0.05 were considered statistically significant for the primary outcome. Secondary outcomes, negative control outcomes, healthcare utilization analyses, sensitivity analyses, subgroup analyses, and landmark and competing risk analyses were considered exploratory for the purpose of statistical inference; therefore, no adjustment for multiple comparisons was applied.
Results
Cohort Assembly and Covariate Balance
Before matching, 89,812 GLP-1RA initiators and 30,100 DPP-4i initiators were identified among COVID-19 survivors with type 2 diabetes mellitus (Table 1). GLP-1RA initiators were younger (mean age, 62.9 vs 68.1 years) and more often female (53.5% vs 49.5%), with a marked imbalance in age (standardized mean difference [SMD] 0.521). The groups also differed in COVID-19 vaccination status (38.6% vs 25.4%; SMD, 0.285) and race distribution (SMD up to 0.312), along with smaller imbalances in comorbidities and laboratory values. After 1:1 propensity score matching, 25,263 patients remained in each group, and balance was achieved across all measured covariates, including age (66.3 vs 66.4 years; SMD, 0.005) and sex (50.3% vs 50.6%; SMD, 0.006). Balance was also maintained for comorbidities, medications, and laboratory values (eg, hypertension, SMD 0.000; HbA1c ≥9%, SMD 0.002), with all SMDs below 0.10. The median follow-up was 919 days in the GLP-1RA group and 955 days in the DPP-4i group, with the longest observed follow-up extending to 2,191 days (approximately 6 years).
Table 1.
Baseline Characteristics Before and After Propensity Score Matching Among COVID-19 Survivors with Type 2 Diabetes Initiating GLP-1RA versus DPP-4i
| Variables | Before Matching | After Matching | ||||
|---|---|---|---|---|---|---|
| GLP-1 Group(n = 89,812) | DPP-4 Group(n = 30,100) | SMD | GLP-1 Group(n = 25,263) | DPP-4 Group(n = 25,263) | SMD | |
| Patient characteristics | ||||||
| Age at index (years) | 62.9±9.2 | 68.1±10.6 | 0.521 | 66.3±9.6 | 66.4±10.1 | 0.005 |
| Female | 48091 (53.5) | 14,889 (49.5) | 0.082 | 12,697 (50.3) | 12,772 (50.6) | 0.006 |
| BMI≥40 kg/m2 | 26648 (29.7) | 3689 (12.3) | 0.438 | 4057 (16.1) | 3659 (14.5) | 0.044 |
| White | 60895 (67.8) | 15,874 (52.7) | 0.312 | 14,862 (58.8) | 15,103 (59.8) | 0.019 |
| Black or African American | 16536 (18.4) | 4218 (14.0) | 0.120 | 3898 (15.4) | 4036 (16.0) | 0.015 |
| Asian | 3111 (3.5) | 2421 (8.0) | 0.198 | 1699 (6.7) | 1630 (6.5) | 0.011 |
| Comorbidities and health services | ||||||
| Essential (primary) hypertension | 75892 (84.5) | 24,308 (80.8) | 0.099 | 20,840 (82.5) | 20,843 (82.5) | 0.000 |
| Disorders of lipoprotein metabolism and other lipidemias | 73962 (82.4) | 22,289 (74.1) | 0.202 | 19,893 (78.7) | 19,986 (79.1) | 0.009 |
| Overweight and obesity | 53192 (59.2) | 10,332 (34.3) | 0.515 | 10,384 (41.1) | 10,145 (40.2) | 0.019 |
| Neoplasms | 36161 (40.3) | 11,506 (38.2) | 0.042 | 9840 (39.0) | 9821 (38.9) | 0.002 |
| Ischemic heart diseases | 27667 (30.8) | 10,455 (34.7) | 0.084 | 8532 (33.8) | 8566 (33.9) | 0.003 |
| Sleep disorders | 39187 (43.6) | 8305 (27.6) | 0.340 | 8047 (31.9) | 7876 (31.2) | 0.015 |
| Chronic pain, not elsewhere classified | 30507 (34.0) | 6895 (22.9) | 0.247 | 6613 (26.2) | 6615 (26.2) | 0.000 |
| Chronic kidney disease (CKD) | 16699 (18.6) | 6963 (23.1) | 0.112 | 5705 (22.6) | 5618 (22.2) | 0.008 |
| Other anxiety disorders | 25801 (28.7) | 5990 (19.9) | 0.207 | 5690 (22.5) | 5619 (22.2) | 0.007 |
| Type 2 diabetes mellitus with kidney complications | 17158 (19.1) | 6607 (22.0) | 0.071 | 5490 (21.7) | 5450 (21.6) | 0.004 |
| Type 2 diabetes mellitus with neurological complications | 18569 (20.7) | 5983 (19.9) | 0.020 | 5457 (21.6) | 5393 (21.3) | 0.006 |
| Heart failure | 14778 (16.5) | 5531 (18.4) | 0.051 | 4496 (17.8) | 4420 (17.5) | 0.008 |
| Diseases of liver | 17957 (20.0) | 5037 (16.7) | 0.084 | 4332 (17.1) | 4441 (17.6) | 0.011 |
| Nicotine dependence | 15208 (16.9) | 4291 (14.3) | 0.074 | 3998 (15.8) | 3989 (15.8) | 0.001 |
| Cerebrovascular diseases | 10412 (11.6) | 4682 (15.6) | 0.116 | 3677 (14.6) | 3665 (14.5) | 0.001 |
| Other chronic obstructive pulmonary disease | 11687 (13.0) | 4342 (14.4) | 0.041 | 3554 (14.1) | 3572 (14.1) | 0.002 |
| Type 2 diabetes mellitus with circulatory complications | 9972 (11.1) | 3331 (11.1) | 0.001 | 2936 (11.6) | 2972 (11.8) | 0.004 |
| Type 2 diabetes mellitus with ophthalmic complications | 8294 (9.2) | 2681 (8.9) | 0.011 | 2343 (9.3) | 2350 (9.3) | 0.001 |
| Major depressive disorder, recurrent | 8131 (9.1) | 1563 (5.2) | 0.151 | 1546 (6.1) | 1524 (6.0) | 0.004 |
| Alcohol related disorders | 3283 (3.7) | 994 (3.3) | 0.019 | 846 (3.3) | 897 (3.6) | 0.011 |
| Malnutrition | 1527 (1.7) | 979 (3.3) | 0.100 | 714 (2.8) | 655 (2.6) | 0.014 |
| Medications | ||||||
| ANTILIPEMIC AGENTS | 67107 (74.7) | 21,201 (70.4) | 0.096 | 18,531 (73.4) | 18,536 (73.4) | 0.000 |
| Glucocorticoids | 61352 (68.3) | 18,208 (60.5) | 0.164 | 15,908 (63.0) | 15,857 (62.8) | 0.004 |
| Metformin | 58493 (65.1) | 17,900 (59.5) | 0.117 | 15,602 (61.8) | 15,561 (61.6) | 10.003 |
| INSULINS AND ANALOGUES | 42824 (47.7) | 15,675 (52.1) | 0.088 | 13,201 (52.3) | 13,039 (51.6) | 0.013 |
| SARS-CoV-2 (COVID-19) Vaccine | 34664 (38.6) | 7657 (25.4) | 0.285 | 7406 (29.3) | 7359 (29.1) | 0.004 |
| Sodium-glucose co-transporter 2 (SGLT2) inhibitors | 22928 (25.5) | 6820 (22.7) | 0.067 | 6218 (24.6) | 6162 (24.4) | 0.005 |
| Laboratory data | ||||||
| eGFR≥60 mL/min/1.73 m2 | 80689 (89.8) | 24,974 (83.0) | 0.202 | 21,527 (85.2) | 21,576 (85.4) | 0.005 |
| Hemoglobin ≥12 g/dL | 78631 (87.6) | 24,955 (82.9) | 0.131 | 21,322 (84.4) | 21,339 (84.5) | 0.002 |
| Albumin ≥3.5 g/dL | 79321 (88.3) | 24,056 (79.9) | 0.231 | 21,072 (83.4) | 21,137 (83.7) | 0.007 |
| HbA1c≥9% | 31,296 (34.8) | 8912 (29.6) | 0.112 | 8142 (32.2) | 8115 (32.1) | 0.002 |
Notes: Data are presented as number (%) or mean ± standard deviation. Standardized mean differences (SMDs) <0.10 indicated an acceptable covariate balance.
Abbreviations: BMI, body mass index; CKD, chronic kidney disease; COVID-19, coronavirus disease 2019; DPP-4i, dipeptidyl peptidase-4 inhibitor; eGFR, estimated glomerular filtration rate; GLP-1RA, glucagon-like peptide-1 receptor agonist; HbA1c, hemoglobin A1c; SGLT2, sodium-glucose cotransporter 2; SMD, standardized mean difference.
Primary and Secondary Outcomes
In the matched cohort, incident all-cause dementia occurred in 520 GLP-1RA initiators (2.06%) and 832 DPP-4i initiators (3.29%) (Table 2). GLP-1RA initiation was associated with a lower hazard of incident dementia (hazard ratio [HR], 0.64; 95% CI, 0.58–0.72), with Kaplan-Meier curves showing progressive separation between the groups over the follow-up period (Figure 1). Assessment based on Schoenfeld residuals showed no significant violation of the proportional hazards assumption for the primary dementia analysis. Inverse associations were observed for vascular dementia (HR, 0.67; 95% CI, 0.52–0.85) and Alzheimer’s disease (HR, 0.74; 95% CI, 0.59–0.93). For mild cognitive impairment, the estimate was borderline (HR, 0.87; 95% CI, 0.75–1.00; p=0.049). GLP-1RA initiation was also associated with lower hazards of hyperglycemia, hypoglycemia, HbA1c ≥9%, and all-cause mortality (HR, 0.63; 95% CI, 0.59–0.68) (all p<0.05). GLP-1RA initiators also had a slightly higher hazard of recording an ambulatory healthcare visit during follow-up (HR, 1.07; 95% CI, 1.05–1.08), suggesting modest differences in post-index healthcare contact that could influence outcome ascertainment. The negative-control outcome, age-related cataract, showed a statistically significant association favoring the DPP-4i group (HR, 1.08; 95% CI, 1.03–1.14). Because a negative-control outcome would ideally show no association, this finding provides empirical evidence that residual confounding or differential ascertainment may remain despite propensity score matching.
Table 2.
Association Between GLP-1RA Initiation and Incident Dementia, Glycemic Outcomes, Mortality, Healthcare Utilization, and Negative-Control Outcome
| Outcome | GLP-1 Group (n=25,263) | DPP-4 Group (n=25,263) | HR (95% CI) | p value |
|---|---|---|---|---|
| Events (%) | Events (%) | |||
| Primary outcome | ||||
| Overall dementia | 520 (2.06) | 832 (3.29) | 0.64 (0.58–0.72) | <0.001 |
| Secondary outcomes | ||||
| Vascular dementia | 110 (0.44) | 173 (0.68) | 0.67 (0.52–0.85) | 0.001 |
| Alzheimer’s disease | 124 (0.49) | 176 (0.70) | 0.74 (0.59–0.93) | 0.010 |
| Mild cognitive impairment | 344 (1.36) | 412 (1.63) | 0.87 (0.75–1.00) | 0.049 |
| Hyperglycemia | 13,332 (52.77) | 15,746 (62.33) | 0.74 (0.73–0.76) | <0.001 |
| Hypoglycemia | 2,493 (9.87) | 2,748 (10.88) | 0.92 (0.87–0.97) | 0.003 |
| HbA1c ≥9% | 5,820 (23.04) | 6,390 (25.29) | 0.90 (0.87–0.93) | <0.001 |
| Mortality | 1,457 (5.77) | 2,359 (9.34) | 0.63 (0.59–0.68) | <0.001 |
| Health utilization indicator | ||||
| Any healthcare visit | 25,045 (99.14) | 24,893 (98.54) | 1.07 (1.05–1.08) | <0.001 |
| Negative control outcome | ||||
| Age-related cataract | 2,833 (11.21) | 2,640 (10.45) | 1.08 (1.03–1.14) | 0.004 |
Notes: Data are presented as number of events (%), hazard ratios, 95% confidence intervals, and p-values. Any ambulatory healthcare visit during follow-up was assessed as a healthcare utilization indicator. Age-related cataract was included as a negative-control outcome.
Abbreviations: CI, confidence interval; COVID-19, coronavirus disease 2019; DPP-4i, dipeptidyl peptidase-4 inhibitor; GLP-1RA, glucagon-like peptide-1 receptor agonist; HbA1c, hemoglobin A1c; HR, hazard ratio.
Figure 1.
Kaplan–Meier curves for dementia-free probability among GLP-1RA and DPP-4i initiators. Kaplan–Meier curves show dementia-free probability after the index date among propensity score–matched COVID-19 survivors with type 2 diabetes who initiated GLP-1RA or DPP-4i therapy. The shaded areas indicate the 95% confidence intervals. Follow-up extended to 2,191 days, with median follow-up durations of 919 days in the GLP-1RA group and 955 days in the DPP-4i group. Numbers at risk at prespecified time points were not available from the TriNetX platform.
Sensitivity Analyses
The association between GLP-1RA initiation and incident dementia was consistent across models restricted to patients who survived follow-up (HR, 0.67; 0.60–0.76), those with at least one healthcare visit (HR, 0.67; 0.60–0.74), and those without prior SGLT2 inhibitor use (HR, 0.67; 0.59–0.76) (Table 3). The association with Alzheimer’s disease attenuated toward null when restricted to survivors (HR, 0.95; 0.75–1.22). The mild cognitive impairment association was also attenuated in several sensitivity analyses and was not statistically significant in Models I–III. An inverse association with incident dementia was also observed within GLP-1RA subtypes, with a somewhat larger magnitude for semaglutide (HR, 0.60; 0.53–0.67) than for dulaglutide (HR, 0.65; 0.58–0.74) versus the DPP-4i comparator, indicating that the overall association was not driven by a single agent (Table 4). In Model VI, after excluding patients who initiated the comparator drug class during follow-up, the association with incident dementia remained similar to the primary analysis (HR, 0.62; 0.56–0.69). Associations were also observed for vascular dementia (HR, 0.62; 0.49–0.78) and Alzheimer’s disease (HR, 0.74; 0.59–0.93), whereas the estimate for mild cognitive impairment was not statistically significant (HR, 0.91; 0.79–1.05).
Table 3.
Sensitivity Analyses for the Association Between GLP-1RA Initiation and Incident Dementia: Models I–III
| Outcomes | Model I | Model II | Model III | |||
|---|---|---|---|---|---|---|
| HR (95% CI) | p value | HR (95% CI) | p value | HR (95% CI) | p value | |
| Overall dementia | 0.67 (0.60–0.76) | <0.001 | 0.67 (0.60–0.74) | <0.001 | 0.67 (0.59–0.76) | <0.001 |
| Vascular dementia | 0.68 (0.52–0.89) | 0.005 | 0.73 (0.59–0.92) | 0.006 | 0.62 (0.47–0.81) | <0.001 |
| Alzheimer’s disease | 0.95 (0.75–1.22) | 0.693 | 0.70 (0.56–0.87) | 0.002 | 0.62 (0.48–0.81) | <0.001 |
| Mild cognitive impairment | 0.87 (0.75–1.01) | 0.072 | 0.91 (0.79–1.04) | 0.177 | 0.90 (0.76–1.06) | 0.204 |
| Hyperglycemia | 0.75 (0.73–0.77) | <0.001 | 0.75 (0.73–0.77) | <0.001 | 0.72 (0.70–0.74) | <0.001 |
| Hypoglycemia | 0.96 (0.91–1.02) | 0.204 | 0.92 (0.87–0.97) | 0.003 | 0.95 (0.89–1.01) | 0.111 |
| HbA1c ≥9% | 0.87 (0.84–0.90) | <0.001 | 0.88 (0.85–0.91) | <0.001 | 0.87 (0.84–0.91) | <0.001 |
| Mortality | NA | NA | 0.63 (0.59–0.67) | <0.001 | 0.61 (0.57–0.66) | <0.001 |
| Any healthcare visit | 1.08 (1.06–1.10) | <0.001 | 1.06 (1.04–1.08) | <0.001 | 1.07 (1.04–1.09) | <0.001 |
| Age-related cataract | 1.06 (1.00–1.12) | 0.037 | 1.07 (1.02–1.13) | 0.008 | 1.06 (0.99–1.12) | 0.078 |
Notes: Model I was restricted to patients who survived throughout the follow-up period (n=23,511 per group). Model II was restricted to patients with at least one healthcare visit during follow-up (n=25,176 per group). Model III excluded patients who had previously used SGLT2 inhibitors (n=18,232 per group). Data are presented as hazard ratios, 95% confidence intervals, and p-values.
Abbreviations: CI, confidence interval; COVID-19, coronavirus disease 2019; DPP-4i, dipeptidyl peptidase-4 inhibitor; GLP-1RA, glucagon-like peptide-1 receptor agonist; HbA1c, hemoglobin A1c; HR, hazard ratio; NA, not applicable; SGLT2, sodium-glucose cotransporter 2.
Table 4.
Sensitivity Analyses for the Association Between GLP-1RA Initiation and Incident Dementia: Models IV–VI
| Outcomes | Model IV | Model V | Model VI | |||
|---|---|---|---|---|---|---|
| HR (95% CI) | p value | HR (95% CI) | p value | HR (95% CI) | p value | |
| Overall dementia | 0.60 (0.53–0.67) | <0.001 | 0.65 (0.58–0.74) | <0.001 | 0.62 (0.56–0.69) | <0.001 |
| Vascular dementia | 0.51 (0.39–0.68) | <0.001 | 0.71 (0.55–0.92) | 0.008 | 0.62 (0.49–0.78) | <0.001 |
| Alzheimer’s disease | 0.72 (0.56–0.92) | 0.009 | 0.66 (0.51–0.86) | 0.002 | 0.74 (0.59–0.93) | 0.010 |
| Mild cognitive impairment | 0.75 (0.64–0.88) | <0.001 | 0.87 (0.75–1.02) | 0.077 | 0.91 (0.79–1.05) | 0.207 |
| Hyperglycemia | 0.66 (0.65–0.68) | <0.001 | 0.83 (0.81–0.85) | <0.001 | 0.74 (0.72–0.76) | <0.001 |
| Hypoglycemia | 0.86 (0.81–0.92) | <0.001 | 0.96 (0.91–1.02) | 0.147 | 0.83 (0.78–0.88) | <0.001 |
| HbA1c ≥9% | 0.82 (0.79–0.85) | <0.001 | 0.97 (0.94–1.01) | 0.139 | 1.04 (1.00–1.09) | 0.046 |
| Mortality | 0.51 (0.48–0.55) | <0.001 | 0.64 (0.59–0.68) | <0.001 | 0.54 (0.51–0.58) | <0.001 |
| Any healthcare visit | 1.05 (1.03–1.07) | <0.001 | 1.07 (1.05–1.09) | <0.001 | 1.05 (1.03–1.07) | <0.001 |
| Age-related cataract | 0.99 (0.94–1.04) | 0.659 | 1.24 (1.18–1.31) | <0.001 | 1.20 (1.14–1.27) | <0.001 |
Notes: Model IV was restricted to semaglutide initiators and their matched DPP-4i comparators (n=23,783 in each group). Model V was restricted to dulaglutide initiators and their matched DPP-4i comparators (n=21,286 per group). Model VI excluded patients who initiated the comparator drug class during follow-up; the remaining cohorts were subsequently rematched using the same propensity score matching approach as in the primary analysis (n=22,575 per group). Data are presented as hazard ratios, 95% confidence intervals, and p values.
Abbreviations: CI, confidence interval; COVID-19, coronavirus disease 2019; DPP-4i, dipeptidyl peptidase-4 inhibitor; GLP-1RA, glucagon-like peptide-1 receptor agonist; HbA1c, hemoglobin A1c; HR, hazard ratio.
Landmark and Competing-Risk Analyses
The association remained evident after applying the 6-month landmark (HR, 0.66; 95% CI, 0.58–0.75), reducing concern that the primary finding was driven mainly by early differential detection (Table 5). In the descriptive competing-risk analysis performed in the unmatched cohort using the Aalen–Johansen estimator, the cumulative incidence of dementia at the end of follow-up was 3.0% among GLP-1RA initiators versus 8.5% among DPP-4i initiators, whereas the cumulative incidence of death, modeled as a competing event, was 11.2% versus 18.6%, respectively (Figure 2). Despite their lower cumulative mortality, GLP-1RA initiators still had a lower cumulative incidence of dementia than DPP-4i initiators. This finding was reassuring, but because the analysis was descriptive and conducted in the unmatched cohort, it does not exclude an influence of differential mortality on the primary association.
Table 5.
Six-month Landmark Analysis of the Association Between GLP-1RA Initiation and Subsequent Incident Dementia
| Outcome | GLP-1 Group (n=24,897) | DPP-4 Group (n=24,897) | HR (95% CI) | p value |
|---|---|---|---|---|
| Events (%) | Events (%) | |||
| Overall dementia | 396 (1.59) | 635 (2.55) | 0.66 (0.58–0.75) | <0.001 |
| Vascular dementia | 80 (0.32) | 131 (0.53) | 0.66 (0.50–0.87) | 0.003 |
| Alzheimer’s disease | 86 (0.35) | 133 (0.53) | 0.70 (0.54–0.92) | 0.010 |
| Mild cognitive impairment | 260 (1.04) | 335 (1.35) | 0.83 (0.71–0.98) | 0.029 |
| Hyperglycemia | 11,467 (46.06) | 13,236 (53.16) | 0.80 (0.78–0.82) | <0.001 |
| Hypoglycemia | 1,987 (7.98) | 2,135 (8.57) | 0.97 (0.92–1.03) | 0.376 |
| HbA1c ≥9% | 4,857 (19.51) | 5,512 (22.14) | 0.88 (0.84–0.91) | <0.001 |
| Mortality | 1,101 (4.42) | 1,695 (6.81) | 0.69 (0.64–0.74) | <0.001 |
| Any healthcare visit | 24,038 (96.55) | 23,518 (94.46) | 1.07 (1.05–1.09) | <0.001 |
| Age-related cataract | 2,582 (10.37) | 2,454 (9.86) | 1.07 (1.01–1.13) | 0.016 |
Notes: The 6-month landmark analysis excluded patients who developed dementia or died within 6 months after the index date, and follow-up was restarted from the landmark date. Data are presented as number of events (%), hazard ratios, 95% confidence intervals, and p-values.
Abbreviations: CI, confidence interval; COVID-19, coronavirus disease 2019; DPP-4i, dipeptidyl peptidase-4 inhibitor; GLP-1RA, glucagon-like peptide-1 receptor agonist; HbA1c, hemoglobin A1c; HR, hazard ratio.
Figure 2.
Descriptive Aalen–Johansen competing-risk analysis for incident dementia, with death as a competing event. Aalen–Johansen cumulative incidence curves show the cumulative incidence of incident dementia and death among (A) GLP-1RA initiators and (B) DPP-4i initiators. Death was considered a competing event for incident dementia. At the end of follow-up, the cumulative incidence of dementia was 3.0% in the GLP-1RA group and 8.5% in the DPP-4i group, whereas the cumulative incidence of death was 11.2% and 18.6%, respectively. This analysis was descriptive, and no formal between-group competing-risk hypothesis test was performed. Therefore, the curves and cumulative incidence estimates should not be interpreted as a formal statistical comparison between treatment groups.
Subgroup Analyses
The association was broadly consistent across most subgroups defined by age, sex, chronic kidney disease, cerebrovascular disease, hypertension, glycemic control, obesity, and COVID-19 severity (Figure 3). Among patients aged 50–65 years, GLP-1RA initiation was associated with a lower hazard of incident dementia (HR, 0.57; 95% CI, 0.33–0.97; p=0.034). A similar association was observed among patients aged >65 years (HR, 0.72; 95% CI, 0.64–0.81; p<0.001), with no significant interaction by age (p for interaction=0.375). Pre- and post-matching baseline characteristics for these age-specific analyses are presented in Supplemental Tables 5 and 6. One exception involved ischemic heart disease. The association was stronger among patients with ischemic heart disease (HR, 0.58; 95% CI, 0.50–0.68) than among those without ischemic heart disease (HR, 0.80; 95% CI, 0.69–0.93; p for interaction=0.004).
Figure 3.
Subgroup analyses of the association between GLP-1RA initiation and incident all-cause dementia. Forest plot showing hazard ratios and 95% confidence intervals for incident all-cause dementia across the prespecified subgroups. Hazard ratios compare GLP-1RA initiators with DPP-4i initiators as the reference group. Interaction p-values assess the heterogeneity of the association across subgroup strata.
Abbreviations: CKD, chronic kidney disease; COVID-19, coronavirus disease 2019; DPP-4i, dipeptidyl peptidase-4 inhibitor; GLP-1RA, glucagon-like peptide-1 receptor agonist; HbA1c, hemoglobin A1c; HR, hazard ratio.
Discussion
In this target trial emulation among COVID-19 survivors with type 2 diabetes mellitus, GLP-1RA initiation was associated with a lower hazard of incident all-cause dementia than DPP-4i initiation. The primary association remained evident across sensitivity, subgroup, landmark, and GLP-1RA subtype analyses. The descriptive competing-risk analysis was directionally consistent but not intended for formal inference. Together, these findings provide comparative real-world evidence for an association rather than a causal effect. However, the significant negative-control association with age-related cataract indicates that residual confounding or differential ascertainment cannot be excluded. Several secondary and subtype-specific cognitive outcomes were attenuated toward null; in particular, the mild cognitive impairment estimate was borderline and was not statistically significant in several sensitivity analyses, and should not be considered robust. To our knowledge, this is the first target trial emulation to evaluate this comparison specifically among COVID-19 survivors with type 2 diabetes.
Compared with previous target trial emulations in the general population with type 2 diabetes,25–27 our study showed both consistent and distinct findings. Similar to the Swedish National Register emulation and a more recent electronic health record–based emulation,25,26 we observed that GLP-1RA initiation was associated with a lower hazard of incident dementia than DPP-4i initiation. Our findings also partly overlap with those of a Medicare-based emulation,27 in which the overall association was not definitive but was more favorable among participants younger than 75 years. The main difference is the study population: prior studies focused on older or general type 2 diabetes cohorts,25–27 whereas our analysis was restricted to COVID-19 survivors with type 2 diabetes, a group potentially characterized by additional post-acute neurological, inflammatory, and vascular vulnerability. Data sources may also contribute, as registers, claims, and electronic health records capture dementia diagnoses with differing sensitivity and timing. Thus, our study is consistent with prior evidence reporting lower hazards of incident dementia with GLP-1RA versus DPP-4i, while extending it to a post-COVID population.
The GLP-1RA subtype analyses helped clarify whether the primary association was driven by a specific agent. Both semaglutide and dulaglutide were associated with a lower hazard of incident all-cause dementia than DPP-4i, suggesting that the observed association was not driven by a single GLP-1RA agent. However, the point estimate for incident dementia was lower in the semaglutide-restricted analysis than in the dulaglutide-restricted analysis, and this pattern broadly paralleled stronger reductions in hyperglycemia, hypoglycemia, and HbA1c ≥9% in the semaglutide-restricted analysis. Although these analyses were not designed to formally compare GLP-1RA molecules or test glycemic mediation, they raise the possibility that better post-index glycemic control may contribute to the observed association with dementia. This interpretation should be approached with caution because differences between semaglutide and dulaglutide users may reflect prescribing indications, obesity severity, adherence, dose escalation, treatment duration, healthcare engagement, or residual confounding rather than pharmacological differences alone. Future studies comparing individual GLP-1RA agents should incorporate longitudinal HbA1c trajectories, glucose variability, weight change, adherence, dose, and formal mediation analyses to clarify whether the cognitive association is mediated by improved glycemic control, weight reduction, cardiovascular benefit, or glycemia-independent neuroprotection.
Several biologically plausible and potentially overlapping mechanisms could hypothetically contribute to the observed association. Of these, only post-index glycemic outcomes were directly evaluated in the present study. First, GLP-1RAs generally provide stronger glucose-lowering effects than DPP-4i,28–30 and the lower hazards of post-index hyperglycemia and HbA1c ≥9% observed among GLP-1RA initiators suggest that improved glycemic control may represent an intermediate pathway contributing to the observed association with dementia rather than a baseline confounder. This pathway is biologically plausible because chronic hyperglycemia, insulin resistance, and glycemic variability have been linked to microvascular injury, oxidative stress, and accelerated neurodegeneration.31–33 Second, GLP-1RAs may have glycemia-independent neuroprotective properties, including attenuation of systemic and neuroinflammation, reduction of oxidative stress, and modulation of amyloid- and tau-related pathways.19,24,34,35 Central GLP-1 receptor signaling may also provide a direct neurobiological pathway through which GLP-1RAs influence neuronal and inflammatory processes. These mechanisms may be particularly relevant after COVID-19, in which persistent neuroinflammatory and vascular injury may contribute to subsequent cognitive vulnerability. Third, cardiovascular mechanisms may be involved. GLP-1RAs reduce atherosclerotic cardiovascular events in high-risk patients with type 2 diabetes,36 and the stronger association observed among patients with ischemic heart disease in our subgroup analysis raises the possibility that vascular pathways contribute to the overall finding. Our observational design cannot determine whether glycemic, neuroinflammatory, cardiovascular, or other pathways mediate the observed association, and formal mediation analyses are required.
From a clinical perspective, the observed HR of 0.64 is notable but is insufficient to support selection of GLP-1RA therapy specifically for dementia prevention. Residual confounding, treatment-selection differences, and differential ascertainment remain possible. Treatment decisions should therefore continue to consider established metabolic and cardiovascular indications, gastrointestinal and other adverse effects, tolerability, patient preferences, and treatment access or cost. The present findings instead support prospective evaluation of cognitive outcomes as an additional component of comparative diabetes treatment research in post-COVID populations.
Several design features support the robustness of the primary findings, although they do not establish causality. The target trial emulation aligned eligibility, treatment assignment, and time zero; the active-comparator new-user design was intended to reduce prevalent-user bias and confounding by indication; and the 90-day post-COVID survivor landmark limited the influence of acute COVID-19 mortality. The primary association remained evident across sensitivity analyses, including restrictions based on survival, healthcare follow-up, prior SGLT2 inhibitor use, delayed landmark entry, GLP-1RA subtype, and post-index cross-class switching. However, greater post-index healthcare utilization and the significant negative-control association with age-related cataract indicate that residual confounding or differential ascertainment may remain despite measured covariate balance. The lower mortality among GLP-1RA initiators (HR, 0.63) may also reflect unmeasured differences in health status, treatment selection, or healthcare engagement and makes death an important competing event. In the descriptive unmatched Aalen–Johansen analysis, cumulative incidences of both dementia and death were lower among GLP-1RA initiators, but this analysis cannot exclude differential mortality, informative censoring, or residual confounding. Treatment discontinuation and incomplete adherence also remain possible. Accordingly, the primary estimate should be interpreted as an intention-to-treat–like association of treatment initiation rather than the effect of sustained GLP-1RA versus DPP-4i exposure.
Several limitations should be considered. First, prescribing indication, socioeconomic status, exact diabetes duration, healthcare-seeking behavior, and granular obesity characteristics were unavailable or incompletely captured. Although propensity score matching balanced measured covariates, residual treatment-selection confounding may remain, and restricting the analysis to patients with sufficient propensity-score overlap may limit generalizability; consequently, the observed HR of 0.64 may partly reflect unmeasured differences between treatment groups. Second, post-index glycemic control and weight change represent potential intermediate pathways rather than baseline confounders. Although GLP-1RA initiators showed more favorable glycemic outcomes, this study cannot determine the relative contributions of glycemic control, weight reduction, neuroprotection, cardiovascular benefit, or residual confounding. Third, EHR-based outcome definitions may misclassify dementia subtypes or miss undiagnosed cognitive decline. Finally, detailed data on medication dose, duration, adherence, discontinuation, and uniformly timed post-index HbA1c and body-weight measurements were unavailable. Cross-class switching could be identified and addressed in Model VI, but this post-index switching-restricted analysis should not be interpreted as a per-protocol estimate. Missing laboratory values were not imputed, and routine testing may be incomplete or non-random. COVID-19 severity was approximated by hospitalization status without more granular measures such as intensive care or oxygen support, and post-COVID inflammatory biomarkers were unavailable. The TriNetX US Collaborative Network may also not represent patients in other healthcare systems or countries, limiting generalizability.
Conclusion
Among COVID-19 survivors with type 2 diabetes, GLP-1RA initiation was associated with a lower hazard of incident all-cause dementia than DPP-4i initiation. The primary dementia association remained evident across several sensitivity, subgroup, landmark, and GLP-1RA subtype analyses, although some subtype-specific cognitive outcomes were attenuated in selected analyses. However, residual confounding, differential mortality, and differential healthcare utilization or ascertainment cannot be excluded, particularly in light of the significant negative-control association. Therefore, these findings do not establish a dementia-preventive effect of GLP-1RA therapy. Confirmation in independent datasets and prospective studies is warranted before cognitive protection can be considered in treatment selection. For clinicians managing post-COVID patients with T2DM, continued attention to cognitive symptoms during follow-up may be warranted.
Funding Statement
This research did not receive any specific grants from funding agencies in the public, commercial, or not-for-profit sectors.
Declaration of AI-Assisted Technologies
During the preparation of this manuscript, ChatGPT (OpenAI; GPT-5.6 Sol) was used to assist with language editing, wording refinement, and improvement of figure formatting and visual presentation. All scientific content, numerical values, effect estimates, confidence intervals, tables, and figures were reviewed and verified by the authors against the original statistical outputs. ChatGPT was not used to generate or modify primary research data. The authors take full responsibility for the accuracy, integrity, and final content of the submitted manuscript.
Data Sharing Statement
The data utilized in this study were obtained from the TriNetX Research Network under a collaborative agreement and are not publicly available. De-identified data may be made available upon reasonable request to the corresponding author, pending approval from TriNetX and compliance with the required data-sharing agreement or network membership.
Ethics Approval Statement
The study protocol was approved by the Institutional Review Board of Chi Mei Medical Center (IRB No. 11406-E01). The requirement for informed consent was waived in accordance with institutional and national regulations governing observational research.
Patient Consent Statement
Informed consent was not required for this retrospective study, as it involved secondary analysis of pre-existing data without any interventions or direct participant interaction.
Author Contributions
All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.
Disclosure
The authors declare that they have no competing financial interests or personal relationships that could have influenced the work reported in this paper.
References
- 1.Bello-Chavolla OY, Antonio-Villa NE, Vargas-Vázquez A, Ávila-Funes JA, Aguilar-Salinas CA. Pathophysiological mechanisms linking type 2 diabetes and dementia: review of evidence from clinical translational and epidemiological research. Curr Diabetes Rev. 2019;15:456–15. doi: 10.2174/1573399815666190129155654 [DOI] [PubMed] [Google Scholar]
- 2.Xue M, Xu W, Ou YN, et al. Diabetes mellitus and risks of cognitive impairment and dementia: a systematic review and meta-analysis of 144 prospective studies. Ageing Res Rev. 2019;55:100944. doi: 10.1016/j.arr.2019.100944 [DOI] [PubMed] [Google Scholar]
- 3.Athanasaki A, Melanis K, Tsantzali I, et al. Type 2 diabetes mellitus as a risk factor for alzheimer’s disease: review and meta-analysis. Biomedicines. 2022;10:778. doi: 10.3390/biomedicines10040778 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Mohamed-Mohamed H, García-Morales V, Sánchez Lara EM, et al. Physiological mechanisms inherent to diabetes involved in the development of dementia: alzheimer’s disease. Neurol Int. 2023;15:1253–1272. doi: 10.3390/neurolint15040079 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kan W, Qu M, Wang Y, Zhang X, Xu L. A review of type 2 diabetes mellitus and cognitive impairment. Front Endocrinol. 2025;16:1624472. doi: 10.3389/fendo.2025.1624472 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Singh DD, Shati AA, Alfaifi MY, et al. Development of dementia in type 2 diabetes patients: mechanisms of insulin resistance and antidiabetic drug development. Cells. 2022;11:3767. doi: 10.3390/cells11233767 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Patel VN, Chorawala MR, Shah MB, et al. Emerging pathophysiological mechanisms linking diabetes mellitus and alzheimer’s disease: an old wine in a new bottle. J Alzheimers Dis Rep. 2022;6:349–357. doi: 10.3233/ADR-220021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Tyagi A, Pugazhenthi S. Targeting Insulin Resistance to Treat Cognitive Dysfunction. Mol Neurobiol. 2021;58:2672–2691. doi: 10.1007/s12035-021-02283-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Takechi R, Sharif A, Brook E, et al. Is type 2 diabetes associated dementia a microvascular early-Alzheimer’s phenotype induced by aberrations in the peripheral metabolism of lipoprotein-amyloid? Front Endocrinol. 2023;14:1127481. doi: 10.3389/fendo.2023.1127481 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Vinuesa A, Pomilio C, Gregosa A, et al. Inflammation and insulin resistance as risk factors and potential therapeutic targets for alzheimer’s disease. Front Neurosci. 2021;15:653651. doi: 10.3389/fnins.2021.653651 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Rahmati M, Yon DK, Lee SW, et al. New-onset neurodegenerative diseases as long-term sequelae of SARS-CoV-2 infection: a systematic review and meta-analysis. J Med Virol. 2023;95:e28909. doi: 10.1002/jmv.28909 [DOI] [PubMed] [Google Scholar]
- 12.Xu E, Xie Y, Al-Aly Z. Long-term neurologic outcomes of COVID-19. Nat Med. 2022;28:2406–2415. doi: 10.1038/s41591-022-02001-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Reiss AB, Greene C, Dayaramani C, et al. Long COVID, the brain, nerves, and cognitive function. Neurol Int. 2023;15:821–841. doi: 10.3390/neurolint15030052 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Park HY, Song IA, Oh TK. Dementia risk among coronavirus disease survivors: a nationwide cohort study in South Korea. J Pers Med. 2021;11:1015. doi: 10.3390/jpm11101015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Wang L, Davis PB, Volkow ND, Berger NA, Kaelber DC, Xu R. Association of COVID-19 with New-Onset Alzheimer’s Disease. J Alzheimers Dis. 2022;89:411–414. doi: 10.3233/JAD-220717 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Shan D, Wang C, Crawford T, Holland C. Association between COVID-19 infection and new-onset dementia in older adults: a systematic review and meta-analysis. BMC Geriatr. 2024;24:940. doi: 10.1186/s12877-024-05538-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Maskery MP, Holscher C, Jones SP, et al. Glucagon-like peptide-1 receptor agonists as neuroprotective agents for ischemic stroke: a systematic scoping review. J Cereb Blood Flow Metab. 2021;41:14–30. doi: 10.1177/0271678X20952011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Cheng D, Yang S, Zhao X, Wang G. The Role of glucagon-like peptide-1 receptor agonists (glp-1 ra) in diabetes-related neurodegenerative diseases. Drug Des Devel Ther. 2022;16:665–684. doi: 10.2147/DDDT.S348055 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Urkon M, Ferencz E, Szász JA, et al. Antidiabetic GLP-1 receptor agonists have neuroprotective properties in experimental animal models of alzheimer’s disease. Pharmaceuticals. 2025;18:614. doi: 10.3390/ph18050614 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Mosenzon O, Capehorn MS, De Remigis A, Rasmussen S, Weimers P, Rosenstock J. Impact of semaglutide on high-sensitivity C-reactive protein: exploratory patient-level analyses of SUSTAIN and PIONEER randomized clinical trials. Cardiovasc Diabetol. 2022;21:172. doi: 10.1186/s12933-022-01585-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Cukierman-Yaffe T, Gerstein HC, Colhoun HM, et al. Effect of dulaglutide on cognitive impairment in type 2 diabetes: an exploratory analysis of the REWIND trial. Lancet Neurol. 2020;19:582–590. doi: 10.1016/S1474-4422(20)30173-3 [DOI] [PubMed] [Google Scholar]
- 22.Moaket OS, Obaid SE, Obaid FE, et al. GLP-1 and the degenerating brain: exploring mechanistic insights and therapeutic potential. Int J Mol Sci. 2025;26:10743. doi: 10.3390/ijms262110743 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Reich N, Hölscher C. The neuroprotective effects of glucagon-like peptide 1 in Alzheimer’s and Parkinson’s disease: an in-depth review. Front Neurosci. 2022;16:970925. doi: 10.3389/fnins.2022.970925 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Spezani R, Mandarim-de-Lacerda CA. Beyond diabetes and obesity: GLP-1 receptor agonists in disrupting the vicious cycle of metabolic dysfunction and neuroinflammation. Diabetes Obes Metab. 2026;28:1622–1637. doi: 10.1111/dom.70400 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Zhou T, Tang H, Zhang B, et al. Association between glucagon-like peptide-1 receptor agonists and risk of dementia in older adults with type 2 diabetes: a target trial emulation. Diabetes Obes Metab. 2026;28:1984–1996. doi: 10.1111/dom.70384 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Tang B, Sjölander A, Wastesson JW, et al. Comparative effectiveness of glucagon-like peptide-1 agonists, dipeptidyl peptidase-4 inhibitors, and sulfonylureas on the risk of dementia in older individuals with type 2 diabetes in Sweden: an emulated trial study. EClinicalMedicine. 2024;73:102689. doi: 10.1016/j.eclinm.2024.102689 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Inoue K, Saliba D, Gotanda H, et al. Glucagon-like peptide-1 receptor agonists and incidence of dementia among older adults with type 2 diabetes: a target trial emulation. Ann Intern Med. 2025;178:1258–1267. doi: 10.7326/ANNALS-24-02648 [DOI] [PubMed] [Google Scholar]
- 28.Gilbert MP, Pratley RE. GLP-1 analogs and dpp-4 inhibitors in type 2 diabetes therapy: review of head-to-head clinical trials. Front Endocrinol. 2020;11:178. doi: 10.3389/fendo.2020.00178 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Campos C, Unger J. Primary care management of type 2 diabetes: a comparison of the efficacy and safety of glucagon-like peptide-1 receptor agonists and dipeptidyl peptidase-4 inhibitors. Postgrad Med. 2021;133:843–853. doi: 10.1080/00325481.2021.1971461 [DOI] [PubMed] [Google Scholar]
- 30.Aroda VR, Henry RR, Han J, et al. Efficacy of GLP-1 receptor agonists and DPP-4 inhibitors: meta-analysis and systematic review. Clin Ther. 2012;34:1247–58.e22. doi: 10.1016/j.clinthera.2012.04.013 [DOI] [PubMed] [Google Scholar]
- 31.Barone E, Di Domenico F, Perluigi M, Butterfield DA. The interplay among oxidative stress, brain insulin resistance and AMPK dysfunction contribute to neurodegeneration in type 2 diabetes and Alzheimer disease. Free Radic Biol Med. 2021;176:16–33. doi: 10.1016/j.freeradbiomed.2021.09.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.watt C, Sanchez-Rangel E, Hwang JJ. Glycemic Variability and CNS Inflammation: reviewing the Connection. Nutrients. 2020;12:3906. doi: 10.3390/nu12123906 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Mule NK, Singh JN. Diabetes Mellitus to neurodegenerative disorders: is oxidative stress fueling the flame? CNS Neurol Disord Drug Targets. 2018;17:644–653. doi: 10.2174/1871527317666180809092359 [DOI] [PubMed] [Google Scholar]
- 34.Grasso M, Fidilio A, Varrasi S, Chiechio S. GLP-1 receptor agonists in neuropathic pain and neurodegenerative diseases: mechanisms, therapeutic potentials, and future perspectives. Curr Neuropharmacol. 2026;24. doi: 10.2174/011570159X450007260305034049 [DOI] [PubMed] [Google Scholar]
- 35.Corcoran E, Kettlety M, Mogul U, Azah JN, Cork SC. The effects of GLP-1 receptor agonists on Alzheimer’s pathophysiology: a systematic review. Mol Cell Neurosci. 2026;137:104091. doi: 10.1016/j.mcn.2026.104091 [DOI] [PubMed] [Google Scholar]
- 36.Chen X, Zhang X, Xiang X, Fang X, Feng S. Effects of glucagon-like peptide-1 receptor agonists on cardiovascular outcomes in high-risk type 2 diabetes: a systematic review and meta-analysis of randomized controlled trials. Diabetol Metab Syndr. 2024;16:251. doi: 10.1186/s13098-024-01497-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data utilized in this study were obtained from the TriNetX Research Network under a collaborative agreement and are not publicly available. De-identified data may be made available upon reasonable request to the corresponding author, pending approval from TriNetX and compliance with the required data-sharing agreement or network membership.