20 Novembre 2026

Psicologia Punto One | Psicologia e Psicologi nel Web

Menu

Menu

SCHEDA SCIENTIFICA2026Altra ricerca

Distinct temporal neuropsychiatric outcomes associated with GLP-1 receptor agonists versus SGLT2 inhibitors after acute kidney disease in type 2 diabetes.

Yu-Chen Cheng, Yuahn-Sieh Huang, Jui-Yi Chen, Vin-Cent Wu, Ruei-Ming Chen, Kuo-Cheng Lu, Ko-Lin Kuo, Chien-Lin Lu, Cai-Mei Zheng, Yi-Chou Hou

Frontiers in endocrinology · 2026

Abstract originale

BACKGROUND: Neurological and neuropsychiatric complications may persist after acute kidney disease (AKD), particularly among patients with type 2 diabetes mellitus (T2DM). Although glucagon-like peptide-1 receptor agonists (GLP-1RAs) and sodium-glucose cotransporter-2 inhibitors (SGLT2is) provide established cardiorenal benefits, their associations with neurological outcomes after AKD remain uncertain. METHODS: We conducted a retrospective cohort study using the TriNetX Global Collaborative Network. Adults with T2DM and dialysis-requiring AKD receiving GLP-1RAs or SGLT2is were identified and balanced using 1:1 propensity score matching, yielding 3,398 patients in each group. Incident delirium and other neuropsychiatric outcomes were evaluated. The 90-365-day period was the primary analysis, with 90-180- and 90-720-day windows used to assess temporal patterns. Kaplan-Meier and multivariable Cox regression analyses were performed. RESULTS: During 90-365 days, GLP-1RA use was associated with better delirium-free survival than SGLT2i use (118/3,032 vs. 155/2,982 events; log-rank P = 0.009). This association remained significant after multivariable adjustment (adjusted hazard ratio [HR], 0.668; 95% CI, 0.488-0.914; P = 0.012). Statin use was independently associated with lower delirium risk (adjusted HR, 0.674; 95% CI, 0.492-0.924; P = 0.014), while exploratory analyses suggested a distinct association between LDL cholesterol of 100-129.9 mg/dL and delirium risk. In contrast, SGLT2i use showed a favorable association with depression during early follow-up, which attenuated over time. The GLP-1RA-delirium association was also attenuated by 90-720 days. Exploratory sensitivity analyses using non-overlapping follow-up intervals showed outcome-specific associations: adjusted differences in delirium were not significant, whereas depression and overall neurological complications were more frequent with GLP-1RA than with SGLT2i. CONCLUSIONS: Among patients with dialysis-requiring AKD and T2DM, GLP-1RA and SGLT2i use showed outcome-specific associations with subsequent neuropsychiatric events. Although the primary analysis suggested a lower delirium risk with GLP-1RA, this association was not consistently observed in adjusted time-stratified analyses. These findings should be considered exploratory and require prospective validation.

Parole chiave

HumansDiabetes Mellitus, Type 2FemaleRetrospective StudiesGlucagon-Like Peptide-1 Receptor AgonistsMaleSodium-Glucose Transporter 2 InhibitorsAgedMiddle AgedAcute Kidney InjuryDeliriumHypoglycemic AgentsGLP-1 RASGLT2 (sodium-glucose cotransporter 2) inhibitoracute kidney diseasedeliriumdepressiontype 2 diabetes
Leggi il testo integrale originale

Testo estratto dalla pubblicazione: sono esclusi immagini, media, supplementi e parti soggette a diritti distinti.

Distinct temporal neuropsychiatric outcomes associated with GLP-1 receptor agonists versus SGLT2 inhibitors after acute kidney disease in type 2 diabetes

Background

Neurological and neuropsychiatric complications may persist after acute kidney disease (AKD), particularly among patients with type 2 diabetes mellitus (T2DM). Although glucagon-like peptide-1 receptor agonists (GLP-1RAs) and sodium-glucose cotransporter-2 inhibitors (SGLT2is) provide established cardiorenal benefits, their associations with neurological outcomes after AKD remain uncertain.

Methods

We conducted a retrospective cohort study using the TriNetX Global Collaborative Network. Adults with T2DM and dialysis-requiring AKD receiving GLP-1RAs or SGLT2is were identified and balanced using 1:1 propensity score matching, yielding 3,398 patients in each group. Incident delirium and other neuropsychiatric outcomes were evaluated. The 90–365-day period was the primary analysis, with 90–180- and 90–720-day windows used to assess temporal patterns. Kaplan–Meier and multivariable Cox regression analyses were performed.

Results

During 90–365 days, GLP-1RA use was associated with better delirium-free survival than SGLT2i use (118/3,032 vs. 155/2,982 events; log-rank P = 0.009). This association remained significant after multivariable adjustment (adjusted hazard ratio [HR], 0.668; 95% CI, 0.488–0.914; P = 0.012). Statin use was independently associated with lower delirium risk (adjusted HR, 0.674; 95% CI, 0.492–0.924; P = 0.014), while exploratory analyses suggested a distinct association between LDL cholesterol of 100–129.9 mg/dL and delirium risk. In contrast, SGLT2i use showed a favorable association with depression during early follow-up, which attenuated over time. The GLP-1RA–delirium association was also attenuated by 90–720 days. Exploratory sensitivity analyses using non-overlapping follow-up intervals showed outcome-specific associations: adjusted differences in delirium were not significant, whereas depression and overall neurological complications were more frequent with GLP-1RA than with SGLT2i.

Conclusions

Among patients with dialysis-requiring AKD and T2DM, GLP-1RA and SGLT2i use showed outcome-specific associations with subsequent neuropsychiatric events. Although the primary analysis suggested a lower delirium risk with GLP-1RA, this association was not consistently observed in adjusted time-stratified analyses. These findings should be considered exploratory and require prospective validation.

Introduction

Acute kidney injury (AKI) is common and clinically important in patients with type 2 diabetes mellitus (T2DM), in whom metabolic, microvascular, cardiovascular, and kidney abnormalities increase susceptibility to acute and persistent kidney dysfunction (1). Although AKI has traditionally been viewed as an acute event, acute kidney disease (AKD) represents persistent kidney dysfunction after AKI and bridges AKI and chronic kidney disease (CKD) (2). Persistent or severe kidney injury is associated with long-term risks of CKD progression, cardiovascular events, and mortality (3). These risks are particularly relevant in diabetes, where recurrent kidney injury and cardiorenal-metabolic abnormalities may compromise recovery. Diabetic AKD therefore represents a distinct therapeutic challenge (4). Emerging real-world evidence suggests that contemporary glucose-lowering therapies may modify post-AKD outcomes. SGLT2 inhibitor use after AKD has been associated with favorable kidney, cardiovascular, and survival outcomes in patients with T2DM (5), highlighting pharmacological strategies beyond glycemic control.

Beyond kidney and cardiovascular outcomes, neurological and neuropsychiatric complications are increasingly recognized as post-AKI and post-critical illness morbidity (6). AKI is linked to acute and long-term neurological consequences through systemic and neuroinflammation, endothelial dysfunction, blood–brain barrier disruption, oxidative stress, uremic toxin accumulation, and altered cerebral metabolism (6). Delirium is especially relevant because systemic inflammation may trigger endothelial and blood–brain barrier injury, microglial activation, oxidative stress, and disruption of neuronal networks (7), while delirium during critical illness is associated with later cognitive impairment (8). Biomarkers of endothelial activation and blood–brain barrier injury have also been associated with delirium in critically ill patients, supporting a kidney–systemic–brain interaction during and after severe illness (9). Depression, cognitive impairment, and dementia may further impair recovery and quality of life. However, most AKI and AKD studies have focused on mortality, kidney progression, and cardiovascular events, leaving post-AKD neuropsychiatric outcomes understudied, particularly in diabetes.

No established pharmacological strategy specifically prevents neuropsychiatric complications after AKD. However, the expanding cardiorenal indications of glucagon-like peptide-1 receptor agonists (GLP-1RAs) and sodium-glucose cotransporter-2 inhibitors (SGLT2is) raise the possibility of benefits beyond metabolic and cardiorenal outcomes. SGLT2 inhibitors have demonstrated kidney and cardiovascular benefits and favorable post-AKD outcomes in patients with T2DM (5), while comparative evidence suggests differential kidney outcomes between SGLT2 inhibitors and GLP-1RAs in patients with T2DM (10). Accumulating experimental and clinical evidence also suggests that GLP-1 receptor activation may affect neuroinflammation, microglial activity, oxidative stress, mitochondrial function, and cerebral glucose metabolism, supporting biological plausibility for neurological effects. Whether these two glucose-lowering drug classes are associated with different post-AKD neuropsychiatric outcomes remains unclear. Therefore, using a large multinational real-world database, we compared incident delirium, depression, and other neurological outcomes between GLP-1RA and SGLT2 inhibitor users with T2DM after dialysis-requiring AKD. We also examined whether these associations differed across early, intermediate, and extended follow-up to explore outcome-specific and temporal patterns of neurological vulnerability during post-AKD recovery.

Materials and Methods

Study design and data source

We conducted a retrospective, multinational cohort study using the TriNetX Global Collaborative Network, a federated electronic health record (EHR) platform with de-identified longitudinal clinical data from participating healthcare organizations (HCOs). The database includes demographics, diagnoses, procedures, medications, laboratory measurements, and healthcare encounters. The query was generated on May 16, 2026; 173 HCOs responded, and 45 contributed eligible patients. Source data spanned September 1, 2012, to September 30, 2023.

All analyses were performed within the secure TriNetX analytics environment using de-identified data. Diagnoses were identified primarily using International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes; medications using Anatomical Therapeutic Chemical (ATC) codes; procedures using Current Procedural Terminology (CPT), Healthcare Common Procedure Coding System (HCPCS), SNOMED CT, and relevant ICD procedure codes; and laboratory measurements using Logical Observation Identifiers Names and Codes (LOINC).

Study population and definition of acute kidney disease

We identified adults aged 18–80 years with type 2 diabetes mellitus (T2DM) who had a qualifying hospitalization with severe kidney dysfunction requiring dialysis or renal replacement therapy. Hospital discharge was identified using CPT codes 99239, 99238, and 99217 with corresponding grouped discharge concepts. To establish temporality, dialysis or renal replacement therapy was required from 3 months before to 1 day after discharge.

In this study, AKD was defined using a prespecified EHR-based hospitalization–dialysis phenotype rather than retrospective reconstruction of Kidney Disease: Improving Global Outcomes (KDIGO) serum creatinine criteria. This definition was selected to identify patients with clinically severe, persistent kidney dysfunction within a multinational federated EHR database.

Patients with documented delirium or altered consciousness during the qualifying hospitalization were excluded to reduce misclassification of prevalent or hospitalization-associated delirium as incident post-AKD events. The pre-index exclusion was defined using ICD-10-CM codes F05, R41.82, R41.0, and R40.0. For each outcome analysis, patients with the corresponding outcome documented before the assessment window were also excluded to focus on incident events.

Exposure definition

Eligible patients were classified into two mutually exclusive cohorts by exposure to contemporary glucose-lowering therapies. The GLP-1RA cohort included patients exposed to a glucagon-like peptide-1 receptor agonist, identified by ATC code A10BJ, without concomitant SGLT2 inhibitor exposure. The SGLT2i cohort included patients exposed to a sodium-glucose cotransporter-2 inhibitor, identified by ATC code A10BK, without GLP-1RA exposure.

The mutually exclusive exposure definition minimized treatment overlap and enabled active-comparator analysis between drug classes. Before propensity score matching, 3,822 patients were identified in the GLP-1RA cohort and 5,937 in the SGLT2i cohort. After 1:1 propensity score matching, 3,398 patients remained in each cohort.

Baseline characteristics and covariates

Baseline characteristics were assessed before matching and included demographic, metabolic, cardiovascular, kidney, laboratory, and medication variables. Demographics included age and sex. Comorbidities included hypertension, ischemic heart disease, heart failure, cerebrovascular disease, chronic obstructive pulmonary disease, and hyperlipidemia.

Kidney and metabolic variables included estimated glomerular filtration rate (eGFR), urinary albumin-to-creatinine ratio (UACR), body mass index (BMI), and glycated hemoglobin (HbA1c). Laboratory variables included hemoglobin, serum albumin, potassium, C-reactive protein, total cholesterol, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol, and triglycerides, where available.

Medication covariates included cardiovascular and metabolic therapies, including statins, beta-blockers, renin–angiotensin system-modifying therapies, and diuretics. Variables were selected for potential associations with treatment selection, cardiorenal status, metabolic severity, or subsequent neurological outcomes.

Detailed diagnostic, medication, laboratory, and procedural codes used to construct cohorts and covariates are provided in the Supplementary Table.

Propensity score matching

To reduce baseline differences between cohorts, 1:1 propensity score matching (PSM) was performed using demographics, major comorbidities, kidney and metabolic parameters, laboratory measurements, and concomitant medications available in TriNetX.

Covariate balance was assessed using standardized mean differences (SMDs) before and after matching, with smaller absolute SMDs indicating better balance. PSM improved overall comparability, although residual imbalance remained for selected clinically relevant characteristics. Because GLP-1RAs and SGLT2 inhibitors have different guideline-supported indications by metabolic, cardiovascular, and kidney phenotypes, residual confounding by indication was considered when interpreting findings.

Outcome assessment

The primary objective was to compare incident neurological and neuropsychiatric outcomes between GLP-1RA and SGLT2i users after dialysis-requiring AKD. The primary assessment period was 90–365 days after the index event. The outcomes of interest included incident delirium, depression, and a composite neurological outcome. Incident delirium was identified using the complete prespecified TriNetX outcome concept set, which included multiple diagnosis, procedure, and medication concepts related to delirium and was therefore not restricted to ICD-10-CM F05 alone. Depression and the composite neurological outcome were similarly identified using their prespecified TriNetX concept sets. The complete coding definitions for all outcomes are provided in Supplementary Table S1. To minimize contamination by prevalent disease, patients with the corresponding outcome documented before each assessment window were excluded from that outcome-specific analysis. Because neurological manifestations after severe systemic illness may evolve over time, the prespecified analyses additionally examined an early post-AKD window (90–180 days) and an extended follow-up window (90–730 days), in addition to the primary 90–365-day assessment. The 90-day landmark was used to focus on post-acute neurological events rather than complications occurring during hospitalization or early recovery. To further evaluate the robustness of the findings and avoid overlap between follow-up windows, exploratory sensitivity analyses were subsequently performed using three mutually exclusive intervals: 90–180, 181–365, and 366–730 days. Crude Kaplan–Meier analyses and multivariable Cox models using a consistent covariate set were performed for delirium, depression, and the composite neurological outcome across these intervals. These interval-specific analyses were considered exploratory and hypothesis-generating rather than formal tests of treatment-by-time interaction. These prespecified windows assessed whether associations between treatment class and neurological outcomes differed by post-AKD recovery stage.

Subgroup and exploratory analyses

To characterize factors associated with incident delirium during the primary 90–365-day period, additional analyses examined demographics, kidney function, metabolic and laboratory parameters, and concomitant medication use. Lipid parameters and statin exposure were emphasized because of their potential links with vascular, inflammatory, and neurocognitive outcomes. LDL-C was evaluated using clinically defined ranges, including <100, 100–129.9, and ≥130 mg/dL. These ranges were evaluated as separate indicator variables (presence versus absence of the specified range) rather than as mutually exclusive categorical levels referenced to a single LDL-C category. These analyses were exploratory and were not intended to establish an optimal LDL-C treatment target.

Statistical analysis

Baseline characteristics were compared between cohorts before and after propensity score matching. Continuous variables were summarized as means with standard deviations, and categorical variables as numbers and percentages. Between-group differences were assessed using appropriate statistical tests within TriNetX, and covariate balance was evaluated using SMDs.

Time-to-event analyses were performed for incident neurological outcomes. Kaplan–Meier methods estimated cumulative event-free probabilities, and log-rank tests compared treatment cohorts. Hazard ratios (HRs) with 95% confidence intervals (CIs) were estimated using Cox proportional hazards models. Death was treated as a censoring event in the primary time-to-event analyses; mortality was additionally compared between treatment groups as a sensitivity assessment for potential differential competing risk.

For incident delirium during the primary 90–365-day period, multivariable Cox proportional hazards regression assessed whether the association between treatment class and delirium persisted after adjustment for clinically relevant demographic, kidney, metabolic, laboratory, and medication covariates. Because residual imbalance remained after propensity score matching for ischemic heart disease, beta-blocker use, and diuretic use, additional multivariable Cox sensitivity analyses were performed. These models consistently adjusted for treatment group, age, sex, ischemic heart disease, beta-blocker use, and diuretic use across all three neuropsychiatric outcomes and the prespecified follow-up intervals.

The 90–365-day interval served as the primary analytical window, whereas 90–180- and 90–730-day analyses characterized early and extended temporal patterns. Because statistical significance across separate windows does not constitute a formal treatment-by-time interaction, differences were interpreted as exploratory temporal patterns rather than definitive evidence of time-varying treatment effects. To facilitate interpretation of temporal patterns, additional exploratory analyses were performed using three non-overlapping follow-up intervals (90–180, 181–365, and 366–730 days). These analyses were conducted separately and were not equivalent to a formal treatment-by-time interaction test; therefore, comparisons of statistical significance across intervals were considered descriptive and hypothesis-generating. Given the multiple secondary outcomes, follow-up intervals, and subgroup analyses, these analyses were considered exploratory and hypothesis-generating. All statistical tests were two-sided, and P<0.05 was considered statistically significant. Analyses were conducted within TriNetX.

Results

Baseline characteristics of the study population

After applying the predefined eligibility criteria, 3,822 patients with acute kidney disease (AKD) receiving GLP-1 receptor agonists (GLP-1RAs) without SGLT2 inhibitors (SGLT2is) and 5,937 patients receiving SGLT2is without GLP-1RAs were included in the final study cohorts. After 1:1 propensity score matching, 3,398 patients remained in each treatment group (Figure 1).

Baseline demographic, clinical, laboratory, and medication characteristics before and after propensity score matching are summarized in Table 1. Before matching, substantial differences were observed between the two treatment groups. Patients receiving GLP-1RAs were younger (55.5 ± 12.2 vs. 59.5 ± 10.7 years, P < 0.001; SMD = 0.354), less frequently male (47.6% vs. 62.7%, P < 0.001; SMD = 0.308), and had a higher BMI (36.5 ± 9.5 vs. 33.3 ± 8.4 kg/m², P < 0.001; SMD = 0.359). Conversely, ischemic heart disease, COPD, and use of several cardiovascular medications were more frequent among SGLT2i users. In particular, ischemic heart disease was present in 30.1% of GLP-1RA users versus 52.2% of SGLT2i users (SMD = 0.461), while beta-blocker use was 50.7% versus 68.6% (SMD = 0.370), respectively.

Table 1Baseline demographic, clinical, laboratory, and medication characteristics of patients receiving GLP-1 receptor agonists or SGLT2 inhibitors before and after propensity score matching.

Characteristics GLP-1RA Before (n=3,822) SGLT2i Before (n=5,937) P value SMD GLP-1RA After (n=3,398) SGLT2i After (n=3,398) P value SMD

Demographics

Age at index, years 55.5 ± 12.2 59.5 ± 10.7 <0.001 0.354 57.2 ± 11.1 57.1 ± 11.2 0.803 0.006

Male sex, n (%) 1,817 (47.6%) 3,701 (62.7%) <0.001 0.308 1,750 (51.5%) 1,750 (51.5%) 1.000 <0.001

Comorbidities

Overweight/obesity, n (%) 2,185 (57.2%) 2,774 (47.0%) <0.001 0.205 1,872 (55.1%) 1,766 (52.0%) 0.010 0.063

Hypertension, n (%) 3,216 (84.2%) 5,270 (89.3%) <0.001 0.152 2,952 (86.9%) 2,965 (87.3%) 0.638 0.011

Ischemic heart disease, n (%) 1,148 (30.1%) 3,078 (52.2%) <0.001 0.461 1,087 (32.0%) 1,574 (46.3%) <0.001 0.297

Cerebrovascular disease, n (%) 388 (10.2%) 775 (13.1%) <0.001 0.093 373 (11.0%) 395 (11.6%) 0.399 0.020

COPD, n (%) 522 (13.7%) 1,135 (19.2%) <0.001 0.151 486 (14.3%) 617 (18.2%) <0.001 0.105

Hyperlipidemia, n (%) 2,265 (59.3%) 3,694 (62.6%) 0.001 0.068 2,075 (61.1%) 2,041 (60.1%) 0.399 0.020

Kidney and metabolic characteristics

BMI, kg/m² 36.5 ± 9.5 33.3 ± 8.4 <0.001 0.359 35.9 ± 9.3 35.0 ± 8.8 0.001 0.095

eGFR, mL/min/1.73 m² 81.2 ± 31.7 78.4 ± 27.8 <0.001 0.094 80.2 ± 30.7 80.4 ± 29.6 0.782 0.007

UACR, mg/g 757.4 ± 2697.2 801.0 ± 3105.4 0.844 0.015 828.7 ± 2972.2 864.8 ± 3552.3 0.901 0.011

Laboratory parameters

Hemoglobin, g/dL 11.7 ± 2.1 12.0 ± 2.4 <0.001 0.127 11.8 ± 2.1 12.0 ± 2.3 <0.001 0.089

Total cholesterol, mg/dL 167.1 ± 68.0 157.7 ± 72.2 <0.001 0.134 164.8 ± 66.0 164.5 ± 77.5 0.925 0.003

LDL cholesterol, mg/dL 87.8 ± 42.6 83.2 ± 41.6 <0.001 0.110 86.3 ± 42.4 86.6 ± 43.5 0.843 0.007

HDL cholesterol, mg/dL 37.4 ± 14.9 35.2 ± 16.6 <0.001 0.139 37.6 ± 15.1 36.5 ± 15.7 0.052 0.069

Triglycerides, mg/dL 222.6 ± 256.0 184.6 ± 244.0 <0.001 0.152 216.6 ± 251.0 204.2 ± 283.6 0.182 0.046

HbA1c, % 9.1 ± 2.7 8.4 ± 2.5 <0.001 0.255 9.0 ± 2.7 8.6 ± 2.5 <0.001 0.139

Potassium, mmol/L 4.0 ± 0.5 4.0 ± 0.5 0.051 0.041 4.0 ± 0.5 4.0 ± 0.5 0.618 0.012

Medications

DPP-4 inhibitors, n (%) 250 (6.5%) 414 (7.0%) 0.368 0.019 229 (6.7%) 252 (7.4%) 0.277 0.026

Biguanides, n (%) 1,305 (34.2%) 1,720 (29.2%) <0.001 0.108 1,133 (33.3%) 1,120 (33.0%) 0.738 0.008

RAAS inhibitors, n (%) 2,065 (54.1%) 3,784 (64.1%) <0.001 0.206 1,958 (57.6%) 1,972 (58.0%) 0.731 0.008

Beta-blockers, n (%) 1,937 (50.7%) 4,045 (68.6%) <0.001 0.370 1,787 (52.6%) 2,160 (63.6%) <0.001 0.224

Diuretics, n (%) 1,916 (50.2%) 3,918 (66.4%) <0.001 0.334 1,748 (51.4%) 2,124 (62.5%) <0.001 0.225

Calcium-channel blockers, n (%) 1,321 (34.6%) 2,404 (40.8%) <0.001 0.127 1,224 (36.0%) 1,276 (37.6%) 0.191 0.032

Statins, n (%) 2,384 (62.4%) 4,280 (72.6%) <0.001 0.218 2,199 (64.7%) 2,350 (69.2%) <0.001 0.095

Benzodiazepine derivatives, n (%) 127 (3.3%) 160 (2.7%) 0.081 0.036 113 (3.3%) 101 (3.0%) 0.405 0.020

Antipsychotics, n (%) 817 (21.4%) 1,129 (19.1%) 0.007 0.056 701 (20.6%) 700 (20.6%) 0.976 0.001

Baseline characteristics are presented for the GLP-1RA and SGLT2i cohorts before and after 1:1 propensity score matching. Continuous variables are reported as mean ± standard deviation and categorical variables as number (%). P values represent between-group comparisons. Standardized mean differences (SMDs) are provided to describe covariate balance.

GLP-1RA, glucagon-like peptide-1 receptor agonist; SGLT2i, sodium-glucose cotransporter-2 inhibitor; SMD, standardized mean difference; COPD, chronic obstructive pulmonary disease; BMI, body mass index; eGFR, estimated glomerular filtration rate; UACR, urine albumin-to-creatinine ratio; LDL, low-density lipoprotein; HDL, high-density lipoprotein; HbA1c, glycated hemoglobin A1c; DPP-4, dipeptidyl peptidase-4; RAAS, renin-angiotensin-aldosterone system.

Propensity score matching substantially improved overall covariate balance. Age, sex, hypertension, cerebrovascular disease, hyperlipidemia, kidney function, UACR, lipid parameters, and several medication exposures were well balanced after matching. Nevertheless, residual differences persisted for selected characteristics, particularly ischemic heart disease (32.0% vs. 46.3%; SMD = 0.297), COPD (14.3% vs. 18.2%; SMD = 0.105), HbA1c (9.0 ± 2.7% vs. 8.6 ± 2.5%; SMD = 0.139), beta-blocker use (52.6% vs. 63.6%; SMD = 0.224), and diuretic use (51.4% vs. 62.5%; SMD = 0.225). Given these residual imbalances, ischemic heart disease, beta-blocker use, and diuretic use were included, together with age and sex, in additional multivariable Cox sensitivity analyses. BMI showed a smaller residual difference after matching (35.9 ± 9.3 vs. 35.0 ± 8.8 kg/m²; SMD = 0.095). These remaining imbalances were considered in subsequent multivariable analyses.

Primary neuropsychiatric outcomes during 90–365 days

The prespecified primary analysis evaluated neurological and neuropsychiatric outcomes between 90 and 365 days after the index hospitalization (Figure 2). The most prominent between-group difference was observed for incident delirium. Patients receiving GLP-1RAs had a significantly lower risk of delirium than those receiving SGLT2is, with an HR of 0.57 (95% CI, 0.39–0.82; P = 0.002). The separation between the two curves became increasingly apparent during follow-up, suggesting that the relative difference in delirium risk emerged during the post-acute recovery period rather than immediately after discharge.

In contrast, no significant difference was observed for depression during the 90–365-day primary window (HR, 1.07; 95% CI, 0.77–1.48; P = 0.673). Similarly, the composite neurological outcome was comparable between the two groups (HR, 0.98; 95% CI, 0.723–1.350; P = 0.940). Thus, the neurological association observed during the primary follow-up period appeared to be outcome-specific, predominantly driven by a lower incidence of delirium in the GLP-1RA group, rather than a generalized reduction in all neurological or neuropsychiatric complications.

In additional analyses using non-overlapping follow-up intervals, crude Kaplan–Meier estimates showed that the between-group difference in depression was significant during 90–180 days (HR, 1.572; 95% CI, 1.003–2.463; P = 0.0465), but not during 181–365 or 366–730 days. A similar early difference was observed for overall neurological complications during 90–180 days (HR, 1.559; 95% CI, 1.001–2.428; P = 0.0478), whereas no significant between-group difference in delirium was observed across any of the three intervals. Event counts and complete crude estimates are provided in Supplementary Table 2. After additional adjustment for age, sex, ischemic heart disease, diuretic use, and beta-blocker use, GLP-1RA treatment was associated with a higher hazard of depression and overall neurological complications across all three non-overlapping intervals. In contrast, the adjusted association with delirium remained nonsignificant during 90–180 days (HR, 0.778; 95% CI, 0.528–1.148; P = 0.2067), 181–365 days (HR, 0.814; 95% CI, 0.609–1.089; P = 0.1660), and 366–730 days (HR, 0.926; 95% CI, 0.735–1.168; P = 0.5180). Complete adjusted estimates are provided in Supplementary Table 3. In an additional agent-specific sensitivity analysis, empagliflozin and semaglutide were the most frequently represented individual agents, accounting for approximately 83.7% (4,971/5,937) of the SGLT2i cohort and 53.3% (2,039/3,822) of the GLP-1RA cohort, respectively. After 1:1 propensity score matching (n=1,951 per group), no significant differences were observed for depression (HR 0.880, 95% CI 0.575–1.347; P = 0.557), delirium (HR 1.556, 95% CI 0.920–2.634; P = 0.097), or overall neurological complications (HR 1.091, 95% CI 0.735–1.620; P = 0.665) (Supplementary Table S4).

Factors associated with incident delirium during the primary follow-up period

To determine whether the observed association with delirium was independent of baseline clinical characteristics, laboratory parameters, and concomitant medications, multivariable Cox proportional hazards analyses were performed for the 90–365-day period (Figure 3).

After multivariable adjustment, GLP-1RA use remained independently associated with a lower risk of incident delirium compared with SGLT2i use (adjusted HR, 0.668; 95% CI, 0.488–0.914; P = 0.012). Importantly, statin use was also independently associated with a lower delirium risk (adjusted HR, 0.674; 95% CI, 0.492–0.924; P = 0.014).

Among laboratory parameters, LDL cholesterol of 100–129.9 mg/dL showed an association with a higher hazard of incident delirium. LDL-C ranges were evaluated as separate indicator variables rather than as mutually exclusive categories referenced to a single LDL-C stratum. In contrast, most other evaluated clinical and biochemical parameters—including age, sex, renal function categories, HbA1c, potassium, CRP, albumin, hemoglobin, and several concomitant medications—were not independently associated with delirium during this interval.

Collectively, these findings indicate that the lower delirium risk associated with GLP-1RA exposure persisted after adjustment for clinically relevant covariates. The concomitant association with statin therapy and the differential signal across LDL cholesterol strata further suggests that lipid-related or vascular-metabolic factors may contribute to post-AKD delirium risk, although these exploratory associations should not be interpreted as defining an optimal LDL target. As a sensitivity assessment for potential competing mortality, death occurred in 113 of 3,275 patients (3.5%) in the GLP-1RA group and 123 of 3,277 patients (3.8%) in the SGLT2i group. Mortality did not differ significantly between treatment groups (risk difference, −0.003; 95% CI, −0.012 to 0.006; P = 0.510; risk ratio, 0.919; 95% CI, 0.715–1.181).

Early neuropsychiatric outcomes during 90–180 days

Analysis of the early post-AKD period revealed a different pattern (Figure 4). During days 90–180, the GLP-1RA group had a higher risk of depression relative to the SGLT2i group (HR, 1.84; 95% CI, 1.07–3.15; P = 0.023), corresponding to a lower early risk of depression among SGLT2i users.

A similar early pattern was observed for the composite neurological outcome, with an HR of 1.71 (95% CI, 1.00–2.92; P = 0.045) for GLP-1RA versus SGLT2i. In contrast, no significant difference in incident delirium was detected during this early interval (HR, 0.85; 95% CI, 0.41–1.74; P = 0.660).

These findings suggest that the relative neurological associations of the two drug classes differed according to both outcome phenotype and follow-up period. Specifically, the early post-AKD period was characterized by a favorable association of SGLT2i exposure with depression and the composite neurological outcome, whereas a significant difference in delirium had not yet emerged.

Extended neuropsychiatric outcomes during 90–730 days

Extended follow-up further demonstrated the temporal heterogeneity of these associations (Figure 5). The early difference in depression was no longer evident over the 90–730-day period (HR, 0.937; 95% CI, 0.748–1.175; P = 0.577). Similarly, there was no significant difference in the composite neurological outcome (HR, 0.91; 95% CI, 0.73–1.13; P = 0.416).

In contrast, the association between GLP-1RA exposure and lower delirium risk persisted during extended follow-up. GLP-1RA users had a significantly lower risk of incident delirium than SGLT2i users during the 90–730-day period (HR, 0.729; 95% CI, 0.574–0.926; P = 0.009).

Taken together, the analyses across the three follow-up windows revealed a distinct temporal and outcome-specific pattern. SGLT2i exposure was associated with a lower risk of depression during the early 90–180-day post-AKD period, but this association attenuated during longer follow-up. Conversely, the difference in delirium was not apparent during the early period but became evident during the 90–365-day primary analysis and remained significant through extended follow-up. Neither treatment demonstrated a consistent advantage for the overall composite neurological outcome across longer observation periods.

Discussion

The first major finding of the present study was that GLP-1 receptor agonist (GLP-1RA) use was associated with a lower risk of incident delirium than SGLT2 inhibitor use during the 90–365-day period following dialysis-requiring acute kidney disease (AKD). This observation is clinically relevant because direct evidence linking GLP-1RA therapy specifically to delirium prevention remains limited, although an expanding body of evidence supports broader neurological effects of GLP-1 receptor activation. Several biologically plausible mechanisms may underlie the potential neuroprotective effects of GLP-1RAs. Experimental studies have shown that GLP-1RA treatment may attenuate neuroinflammation by modulating microglial activation and promoting a shift from pro-inflammatory toward anti-inflammatory microglial phenotypes (1, 2). In patients with type 2 diabetes, GLP-1RA therapy has also been associated with reduced oxidative stress and inflammatory signaling, together with improvements in mitochondrial membrane potential and respiratory function (3). Beyond these anti-inflammatory and mitochondrial effects, liraglutide has been shown to preserve cerebral glucose metabolism and improve blood–brain glucose transfer in patients with Alzheimer’s disease, suggesting a potential role in maintaining cerebral metabolic homeostasis (4, 5).

The apparent specificity of this association for delirium, rather than depression or the broader composite neurological outcome, may be particularly relevant to the underlying pathophysiology of delirium. Delirium represents an acute brain dysfunction that is highly sensitive to systemic inflammatory insults and has been mechanistically linked to neuroinflammation, microglial activation, endothelial dysfunction, blood–brain barrier disruption, oxidative stress, and impaired cerebral energy metabolism (6, 7). Such mechanistic convergence may be especially relevant after severe AKD, in which persistent systemic inflammatory, metabolic, and vascular disturbances may increase susceptibility to subsequent acute brain dysfunction.

In contrast, depression is a more heterogeneous and multifactorial neuropsychiatric disorder involving not only inflammatory pathways but also monoaminergic and glutamatergic neurotransmission, hypothalamic–pituitary–adrenal axis dysregulation, neurotrophic signaling, synaptic plasticity, and psychosocial factors (8). GLP-1RA-mediated inflammatory and metabolic modulation may not translate into a consistent antidepressant effect. Likewise, combining biologically distinct neuropsychiatric outcomes may dilute a delirium-specific association. Importantly, emerging human evidence supports an association between GLP-1RA exposure and reduced neurocognitive vulnerability. A population-based longitudinal study of 109,778 patients with type 2 diabetes reported a lower incidence of dementia among GLP-1RA users than nonusers (adjusted HR 0.90, 95% CI 0.83–0.97 (9)). Similarly, a large target-trial emulation of more than one million patients with type 2 diabetes found that semaglutide use was associated with a lower risk of first-time Alzheimer disease diagnosis than several comparator glucose-lowering therapies (10)]. These findings do not establish GLP-1RA as a treatment for delirium, but they provide clinical plausibility for our observation that GLP-1RA exposure may be associated with reduced neurological vulnerability during recovery from severe AKD. Notably, the treatment difference in delirium was not evident during the early 90–180-day period but emerged during the primary 90–365-day analysis and remained evident during extended follow-up. This temporal pattern raises the possibility that GLP-1RA-associated neurological effects may become more apparent during post-acute recovery rather than immediately after AKD. Nevertheless, given the observational nature of our study, these findings should be interpreted as temporal and outcome-specific associations rather than evidence of a direct causal effect of GLP-1RA therapy on delirium prevention.

The additional time-stratified analyses also highlight the influence of residual covariate imbalance on the observed associations. In the crude analyses, between-group differences in depression and overall neurological complications were primarily observed during the early 90–180-day interval, whereas these associations became more consistent across intervals after adjustment for age, sex, ischemic heart disease, diuretic use, and beta-blocker use. In contrast, delirium remained nonsignificantly different between treatment groups across all three non-overlapping intervals after adjustment. The differences between the crude and adjusted estimates suggest that residual imbalance after propensity score matching may have influenced some of the observed secondary neuropsychiatric associations. Given the multiple outcome- and interval-specific comparisons and the absence of formal multiplicity adjustment or treatment-by-time interaction testing, these findings should be considered exploratory and hypothesis-generating rather than confirmatory evidence of differential or time-dependent treatment effects.

A second finding was that statin use was independently associated with a lower risk of incident delirium, whereas LDL-C levels of 100–129.9 mg/dL showed an exploratory association with a higher delirium hazard. Importantly, the LDL-C ranges in our analysis were evaluated as separate indicator variables rather than as mutually exclusive categories using a single LDL-C stratum as the reference. These observations should not necessarily be interpreted as manifestations of the same mechanism. Rather, they may reflect complementary lipid-related and lipid-independent pathways influencing neurological vulnerability after AKD. The association confined to an intermediate LDL-C range is particularly intriguing because the relationship between cholesterol and neurological outcomes may be nonlinear. In patients with type 2 diabetes, Zhang et al. demonstrated an inverted U-shaped association between LDL-C and cognitive performance, with an estimated inflection point of 2.686 mmol/L, approximately 104 mg/dL (11). Although that study evaluated cognitive function rather than delirium, it supports the concept that neurological outcomes may not follow a simple linear relationship with LDL-C concentration. In contrast, the association between statin exposure and delirium has more direct clinical support. A propensity score-matched study of 110,090 older adults with type 2 diabetes found a lower incidence of delirium among statin users than matched nonusers (12). Earlier evidence in critically ill patients similarly suggested a reduced risk of delirium during statin administration and proposed attenuation of systemic inflammation as a potential mechanism (13). Beyond LDL-C lowering, statins exert several pleiotropic effects, including attenuation of inflammation and oxidative stress and improvement of endothelial function (14). These effects may also extend to the neurovascular unit, where statins have been proposed to preserve blood–brain barrier integrity, reduce endothelial inflammatory responses, and modulate neuroinflammation (15, 16). Such mechanisms are potentially relevant to delirium, a condition strongly linked to systemic inflammation, endothelial dysfunction, and disruption of the blood–brain barrier. Notably, previous clinical evidence in critically ill patients has suggested an association between statin exposure and a reduced risk of delirium (13). Therefore, our findings may indicate that lipid status and statin exposure provide complementary information regarding susceptibility to post-AKD delirium. Nevertheless, the LDL-C 100–129.9 mg/dL finding should be regarded as exploratory and should not be interpreted as establishing this range as an optimal LDL-C target for delirium prevention.

The third major observation was a favorable association between SGLT2 inhibitor use and depression during the earlier post-AKD period, which was no longer evident with extended follow-up. Emerging real-world studies have consistently associated SGLT2 inhibitor use with a lower incidence of depression in patients with type 2 diabetes, although these studies did not specifically characterize an early, time-limited effect (17, 18). SGLT2 inhibitors exert rapid metabolic and hemodynamic effects after treatment initiation, including early improvements in glycemic dynamics (19), providing a potential biological basis for a stronger association during the vulnerable early recovery phase after AKD. Glycemic status may also influence these associations. HbA1c reflects chronic glycemic exposure and has recently been linked to broader metabolic and microvascular vulnerability (20, 21). In our cohort, HbA1c remained slightly higher in GLP-1RA users after matching (9.0% vs. 8.6%; SMD = 0.139). Because longitudinal HbA1c changes were not assessed, whether differential glycemic improvement contributed to the observed neuropsychiatric associations remains uncertain.More broadly, recent real-world analyses of glucose-lowering therapies have demonstrated that neuropsychiatric associations may differ between early and later follow-up periods, supporting the relevance of landmark-based temporal assessment (22). Therefore, the disappearance of the depression signal during extended follow-up in our study should be interpreted as a novel, hypothesis-generating temporal pattern rather than confirmation of an established time-limited antidepressant effect of SGLT2 inhibition. Taken together, our findings suggest outcome-specific and temporally heterogeneous neuropsychiatric associations after AKD, with an early favorable depression profile associated with SGLT2i use and a later, sustained reduction in incident delirium associated with GLP-1RA use.

Limitations

Several limitations should be acknowledged. First, this retrospective EHR-based study remains susceptible to residual and unmeasured confounding despite propensity score matching and multivariable adjustment. Complete balance was not achieved for several clinically relevant characteristics, including BMI and cardiovascular comorbidities. Residual imbalance may reflect confounding by indication, as ADA guidelines recommend GLP-1RAs and SGLT2 inhibitors according to distinct metabolic, cardiovascular, and kidney profiles (23, 24); therefore, causal inference remains limited. Second, death was treated as a censoring event in the conventional Kaplan–Meier and Cox analyses rather than explicitly modeled as a competing event. However, mortality was relatively low and similar between the GLP-1RA and SGLT2i groups (3.5% vs. 3.8%; P = 0.510), reducing concern that differential mortality substantially accounted for the observed between-group associations. Nevertheless, residual competing-risk bias cannot be excluded, and future studies with patient-level data should confirm these findings using competing-risk methods. Third, delirium and depression were identified using diagnostic codes rather than standardized prospective assessments. Claims-based algorithms for delirium have high specificity but relatively low sensitivity, potentially resulting in underrecognition and outcome misclassification (25). Medication exposure recorded in EHRs also does not ensure adherence or continued treatment, and treatment switching or discontinuation during follow-up may introduce exposure misclassification (26, 27). Furthermore, incomplete laboratory data warrant cautious interpretation of the LDL-C and statin findings, particularly the LDL-C 100–129.9 mg/dL signal, which should be considered hypothesis-generating. Finally, the different associations across the 90–180-, 90–365-, and 90–730-day windows do not establish a formal treatment-by-time interaction and require prospective validation. Generalizability beyond patients with T2DM and dialysis-requiring AKD may also be limited.

Conclusion

In this retrospective cohort study of patients with type 2 diabetes following dialysis-requiring acute kidney disease, GLP-1 receptor agonist and SGLT2 inhibitor use were associated with outcome-specific neuropsychiatric profiles. Although the primary analysis suggested a lower risk of incident delirium with GLP-1RA use, this association was not consistently observed in adjusted analyses across non-overlapping follow-up intervals. In contrast, adjusted interval-specific analyses showed a higher risk of depression and overall neurological complications with GLP-1RA compared with SGLT2i use. Statin use was independently associated with a lower risk of delirium, suggesting a potential contribution of vascular-metabolic and inflammatory pathways to post-AKD neuropsychiatric vulnerability. Given the observational design, residual confounding, and multiple interval-specific comparisons, these findings should be considered exploratory and hypothesis-generating. Prospective studies are needed to validate these associations and determine whether treatment effects truly vary over time.

References

1Wang ZJ Han WN Chai SF Li Y Fu CJ Wang CF .
Semaglutide promotes the transition of microglia from M1 to M2 type to reduce brain inflammation in APP/PS1/tau mice . Neuroscience . (2024 ) 563 :222–34. doi: 10.1016/j.neuroscience.2024.11.022 39547338

2Mi R Cheng H Chen R Bai B Li A Gao F .
Effects and mechanisms of long-acting glucagon-like peptide-1 receptor agonist semaglutide on microglia phenotypic transformation and neuroinflammation after cerebral ischemia/reperfusion in rats . Brain Circ . (2024 ) 10 :354–65. doi: 10.4103/bc.bc_38_24 40012598
PMC11850941

3Luna-Marco C de Marañon AM Hermo-Argibay A Rodriguez-Hernandez Y Hermenejildo J Fernandez-Reyes M .
Effects of GLP-1 receptor agonists on mitochondrial function, inflammatory markers and leukocyte-endothelium interactions in type 2 diabetes . Redox Biol . (2023 ) 66 :102849. doi: 10.1016/j.redox.2023.102849 37591012
PMC10457591

4Gejl M Gjedde A Egefjord L Møller A Hansen SB Vang K .
In Alzheimer's disease, 6-month treatment with GLP-1 analog prevents decline of brain glucose metabolism: Randomized, placebo-controlled, double-blind clinical trial . Front Aging Neurosci . (2016 ) 8 :108. doi: 10.3389/fnagi.2016.00108 27252647
PMC4877513

5Gejl M Brock B Egefjord L Vang K Rungby J Gjedde A .
Blood-brain glucose transfer in Alzheimer's disease: Effect of GLP-1 analog treatment . Sci Rep . (2017 ) 7 :17490 . doi: 10.1038/s41598-017-17718-y 29235507 . Research Support. Speaker’s Bureau; Pfizer Inc. L. Egefjord: None. K. Vang: None J. Rungby: Advisory Panel; Merck & Co. Inc. Novo Nordisk A/S, Eli Lilly and Company, Sanofi U.S. Employee; Spouse/Partner; Novo Nordisk A/S. Research Support; Novo Nordisk A/S, Eli Lilly and Company, Amylin Pharmaceuticals, LLC. Speaker’s Bureau; Merck & Co. Inc. Novo Nordisk A/S, Johnson & Johnson, Sanofi U.S. AstraZeneca/Bristol-Myers Squibb. A. Gjedde: Advisory panel; Executive Council, European Dana Alliance for the Brain, European Research Council. Consultant; National Expert, Horizon2020 SC1 PC Health, European Commission, Novo-Nordisk A/S. Research support; Danish Council of Independent Research, National Institutes of Health BRAIN Initiative. Speaker’s bureau; MINDexult. Stocks/shareholder; Danske Bank.
PMC5727512

6Wilson JE Mart MF Cunningham C Shehabi Y Girard TD MacLullich AMJ .
Delirium . Nat Rev Dis Primers . (2020 ) 6 :90 . doi: 10.1038/s41572-020-00223-4 33184265
PMC9012267

7Hughes CG Pandharipande PP Thompson JL Chandrasekhar R Ware LB Ely EW .
Endothelial activation and blood-brain barrier injury as risk factors for delirium in critically ill patients . Crit Care Med . (2016 ) 44 :e809–817. doi: 10.1097/ccm.0000000000001739 27088157
PMC4987204

8Felger JC Lotrich FE .
Inflammatory cytokines in depression: neurobiological mechanisms and therapeutic implications . Neuroscience . (2013 ) 246 :199 –229 . doi: 10.1016/j.neuroscience.2013.04.060 23644052
PMC3741070

9Cheng HW Yang SF Liao PL Lee CH Jong GP .
Impact of glucagon-like peptide-1 receptor agonists on the dementia incidence in patients with type 2 diabetes mellitus: A population-based longitudinal cohort study . Diabetes Metab Res Rev . (2025 ) 41 :e70058. doi: 10.1002/dmrr.70058 40481813
PMC12145076

10Wang W Wang Q Qi X Gurney M Perry G Volkow ND .
Associations of semaglutide with first-time diagnosis of Alzheimer's disease in patients with type 2 diabetes: Target trial emulation using nationwide real-world data in the US . Alzheimer's Dementia: J Alzheimer's Assoc . (2024 ) 20 :8661–72. doi: 10.1002/alz.14313 39445596
PMC11667504

11Zhang H Zhu W Niu T Wang Z An K Cao W .
Inverted U-shaped correlation between serum low-density lipoprotein cholesterol levels and cognitive functions of patients with type 2 diabetes mellitus . Lipids Health Dis . (2021 ) 20 :103 . doi: 10.1186/s12944-021-01534-5 34511118
PMC8436464

12Sun M Wang X Lu Z Yang Y Lv S Miao M .
Statins reduce delirium risk in older adults with type 2 diabetes . J Am Med Dir Assoc . (2025 ) 26 :105490. doi: 10.1016/j.jamda.2025.105490 39952282

13Page VJ Davis D Zhao XB Norton S Casarin A Brown T .
Statin use and risk of delirium in the critically ill . Am J Respir Crit Care Med . (2014 ) 189 :666–73. doi: 10.1164/rccm.201306-1150OC 24417431
PMC3974585

14Liao JK Laufs U .
Pleiotropic effects of statins . Annu Rev Pharmacol Toxicol . (2005 ) 45 :89 –118 . doi: 10.1146/annurev.pharmtox.45.120403.095748 15822172
PMC2694580

15Liu JC Lei SY Zhang DH He QY Sun YY Zhu HJ .
The pleiotropic effects of statins: A comprehensive exploration of neurovascular unit modulation and blood-brain barrier protection . Mol Med . (2024 ) 30 :256 . doi: 10.1186/s10020-024-01025-0 39707228
PMC11660731

16Buttmann M Lorenz A Weishaupt A Rieckmann P .
Atorvastatin partially prevents an inflammatory barrier breakdown of cultured human brain endothelial cells at a pharmacologically relevant concentration . J Neurochem . (2007 ) 102 :1001–8. doi: 10.1111/j.1471-4159.2007.04563.x 17442054

17Mui JV Li L Chou OHI Azfar N Lee A Hui J .
Comparing sodium-glucose cotransporter 2 inhibitors and dipeptidyl peptidase-4 inhibitors on new-onset depression: A propensity score-matched study in Hong Kong . Acta Diabetologica . (2023 ) 60 :917–27. doi: 10.1007/s00592-023-02063-6 37000300
PMC10198893

18Kao MJ Huang YC Huang YC Yang HW To SY Liao CC .
Sodium-glucose cotransporter 2 inhibitors and lower risk of depression in population with type 2 diabetes mellitus: A population-based active comparator, new-user design . Clin Epidemiol . (2025 ) 17 :797 –806 . doi: 10.2147/clep.s536542 41064008
PMC12503185

19Torimoto K Okada Y Koikawa K Tanaka Y .
Early effects of sodium-glucose co-transporter 2 inhibitors in type 2 diabetes: Study based on continuous glucose monitoring . Diabetol Metab Syndr . (2017 ) 9 :60 . doi: 10.1186/s13098-017-0258-5 28785316
PMC5544992

20Popoviciu MS Pop AM Ghitea TC Dorobantu FR Pantis C Pop NO .
HbA1c as a continuous marker of microvascular vulnerability: Development of a non-linear risk framework in a real-world cohort . Metabolites . (2026 ) 16 (3 ):197 . doi: 10.3390/metabo16030197 41893346
PMC13029085

21Popoviciu MS Ghitea TC Pantis C Pop NO Brata RD .
HbA1c-based metabolic stratification: A retrospective study of clinical and biological differences across normoglycemic, prediabetic, and diabetic subjects . Diagnostics (Basel Switzerland) . (2026 ) 16 (4 ):566 . doi: 10.3390/diagnostics16040566 41750711
PMC12940064

22Chen SC Wang TJ Chou CK Lee C .
Neuropsychiatric outcomes with tirzepatide, semaglutide, and other GLP-1 receptor agonists . Diabetes Obes Metab . (2026 ). doi: 10.1111/dom.71088 42420795
PMC13538802

23American Diabetes Association Professional Practice Committee for Diabetes.
10. Cardiovascular disease and risk management: Standards of care in diabetes-2026 . Diabetes Care . (2026 ) 49 :S216–45. doi: 10.2337/dc26-S010 41358899
PMC12690187

24American Diabetes Association Professional Practice Committee for Diabetes.
11. Chronic kidney disease and risk management: Standards of care in diabetes-2026 . Diabetes Care . (2026 ) 49 :S246–60. doi: 10.2337/dc26-S011 41358881
PMC12690176

25Kim DH Lee J Kim CA Huybrechts KF Bateman BT Patorno E .
Evaluation of algorithms to identify delirium in administrative claims and drug utilization database . Pharmacoepidemiol Drug Saf . (2017 ) 26 :945–53. doi: 10.1002/pds.4226 28485014
PMC5583076

26Thai TN Winterstein AG .
Core concepts in pharmacoepidemiology: Measurement of medication exposure in routinely collected healthcare data for causal inference studies in pharmacoepidemiology . Pharmacoepidemiol Drug Saf . (2024 ) 33 :e5683. doi: 10.1002/pds.5683 37752827

27Galozy A Nowaczyk S Sant'Anna A Ohlsson M Lingman M .
Pitfalls of medication adherence approximation through EHR and pharmacy records: Definitions, data and computation . Int J Med Inform . (2020 ) 136 :104092. doi: 10.1016/j.ijmedinf.2020.104092 32062562

Come citare questo studio

Yu-Chen Cheng, Yuahn-Sieh Huang, Jui-Yi Chen, Vin-Cent Wu, Ruei-Ming Chen, Kuo-Cheng Lu, Ko-Lin Kuo, Chien-Lin Lu, Cai-Mei Zheng, Yi-Chou Hou. (2026). Distinct temporal neuropsychiatric outcomes associated with GLP-1 receptor agonists versus SGLT2 inhibitors after acute kidney disease in type 2 diabetes.. Frontiers in endocrinology. DOI: 10.3389/fendo.2026.1963944

Consulta la pubblicazione originale ↗