Use of the TriNetX platform for this study received approval from the Institutional Review Board of Chi-Mei Hospital (No. 11210-E01) and from the review boards of all participating institutions. The TriNetX platform aggregates de-identified patient data, ensuring compliance with the Health Insurance Portability and Accountability Act and General Data Protection Regulation, with an informed consent waiver granted by the Western Institutional Review Board. This study was conducted in accordance with the Declaration of Helsinki principles.

This observational cohort study used data from the TriNetX Analytics platform, a global federated research network comprising de-identified health records from 128 healthcare organizations (Supplementary Material 1, available online) [16,18–20]. The TriNetX database contains comprehensive patient information, including demographic details, diagnoses (according to International Classification of Diseases, 10th Revision [ICD-10], Clinical Modification [CM] codes), procedures (classified according to the ICD-10, Procedure Coding System or Current Procedural Terminology), medications (coded as per the Veterans Affairs National Formulary), laboratory tests (organized using Logical Observation Identifiers Names and Codes), and healthcare utilization records. Acute kidney disease (AKD) was defined in accordance with the ADQI-16 consensus, that is, persistence of kidney dysfunction for ≥7 days but ≤90 days after an AKI event [13]. The operational definition and timing windows used in this analysis are detailed in the ‘Study Cohort’ section.

The study population comprised patients diagnosed with AKD between January 1, 2012 and December 31, 2022. This timeframe allowed for a detailed analysis of the longitudinal prescribing trends of GLP-1 RAs from their initial availability to the present, providing insights into how their use has evolved over time in response to emerging clinical data and changing guidelines. This study adhered to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for cohort studies (Supplementary Material 2, available online).

The primary outcome was the prescription rate of GLP-1 RAs in patients with AKD. The secondary outcome was the incidence of complications related to the use of GLP-1 RAs.

To mitigate potential differences among groups, we recorded factors pertinent to the study population including key demographic information such as age, sex and ethnicity, as well as preexisting comorbidities and medication histories. To reduce baseline differences across study groups, we used an algorithm to integrate and select high-dimensional covariates, all of which were evaluated within the year preceding the index date. Comorbidities were identified using ICD-10-CM codes (Supplementary Material 1, available online). We also recorded systolic blood pressure and body mass index (BMI), along with data on eGFR, proteinuria, glycated hemoglobin (HbA1c), and serum albumin.

We identified 1,095,318 AKD patients who were admitted to participating healthcare facilities during the study period (Fig. 1). Consistent with ADQI-16, we operationalized AKD as adults who (i) required renal replacement therapy (RRT) during the index hospitalization (dialysis-requiring AKI, stage 3 AKI by Kidney Disease: Improving Global Outcomes [KDIGO]), (ii) were successfully weaned off dialysis before discharge, and (iii) survived to day 90. Throughout, we refer to this analytically defined population as “post-dialysis AKD.” This operationalization was necessary because laboratory and urine-output data needed to stage milder AKI (KDIGO stage 1–2) and to verify persistent dysfunction from day 7 to day 90 are incompletely recorded across the TriNetX network, dialysis procedure codes yield a precise and reproducible time-stamp for AKI onset, thereby maximizing phenotype accuracy in this real-world dataset. This “post-dialysis AKD” group is recognized as a high-risk subset with markedly elevated long-term morbidity and mortality [14,15].

For all participants, the index date was set as 90 days post hospital discharge. The inclusion criteria were age over 18 years and having undergone dialysis during their hospital stay. Patients were excluded if they had been diagnosed with type 1 DM, were still on dialysis at discharge, required long-term or repeated dialysis, or died during the AKD period. GLP-1 RA users were defined as patients whose first prescription was recorded after hospital discharge but on or before day 90, thus entirely within the ADQI-16 AKD window. Patients prescribed a GLP-1 RA during the AKD period were classified as GLP-1 RA users, and the cohort was divided into two groups: the GLP-1 RA user group (n = 8,334), and non-user group (n = 1,086,984). All patients were followed for up to 1 year or until death to monitor the outcomes of interest.

To test whether prior exposure might bias results, we repeated analyses after excluding individuals with any GLP-1 RA prescription in the 180 days preceding the index AKI admission.

Subgroup analyses were conducted to explore the factors associated with GLP-1 RA prescription rates, including subgroups of age, sex, race, and existing comorbidities such as hypertension, DM, CKD, and heart failure. Laboratory results such as HbA1c and albumin were also evaluated, along with the concurrent use of other glycemic control medications and renin-angiotensin-aldosterone system blockers. In addition, temporal trends were analyzed for key prescription variables including T2DM, eGFR, proteinuria, obesity, and those prescribed long-acting GLP-1 RAs (exenatide once-weekly, liraglutide, albiglutide, dulaglutide, semaglutide) versus short-acting GLP-1 RAs (exenatide and lixisenatide) [21].

Descriptive statistics were presented as means with standard deviations for numerical variables, and counts with percentages for categorical variables. The t tests and chi-square tests were used to compare differences between GLP-1 RA users and non-users. In accordance with TriNetX policy, cases with missing data and those lost to follow-up were excluded to ensure data integrity.

Ordinary least squares regression was used for trend analysis to assess changes in the prevalence of GLP-1 RA prescriptions and side effects over time (Supplementary Material 1, available online). In the model, time was the independent variable, while the prevalence of GLP-1 RA prescriptions and side effects were the dependent variables. A positive slope in the ordinary least squares model denoted an upward trend in prevalence over time, whereas a negative slope indicated a decline. To avoid reverse causality, the observation period began the day after the index date and extended up to 1 year. Logistic regression was used to analyze the association between GLP-1 RA use and the underlying reasons for their prescription, with results reported as odds ratios (OR) and 95% confidence intervals (CI). Statistical significance was defined as a two-sided p-value of <0.05. Data analysis was performed using R software, version 3.2.2 (R Foundation for Statistical Computing).

A total of 2,154,847 patients with AKI who required dialysis and were discharged from hospital were identified. Of these patients, 1,095,318 (50.8%) had AKD, of whom 156,008 (14.2%) also had concurrent T2DM. The mean age of the AKD cohort was 61.1 years, 51.6% (n = 561,175) were male, 58.0% (n = 634,908) had hypertension, and 20.1% (n = 219,999) had heart failure. In addition, 12.9% (n = 142,043) had an HbA1c level >6.5%, 50.6% (n = 553,685) had an eGFR <60 mL/min/1.73 m², 1.5% (n = 16,565) had proteinuria, and 26.2% (n = 286,871) were obese (Supplementary Table 1, available online). Sepsis was the leading cause of AKI, accounting for 48.2% of cases, followed by cardiorenal syndrome, which accounted for 26.5% of cases (Supplementary Table 2, available online).

Of the overall AKD cohort, 8,334 (0.8%) used GLP-1 RAs (users group) (Table 1; Supplementary Fig. 1, available online), and the remaining 1,086,984 patients did not (non-users group). The proportion of sex was similar in the two groups. The GLP-1 RA users had a higher proportion of Black or African American individuals, elevated BMI and HbA1c levels, and higher prevalence of hypertension, heart failure, cardiovascular disease, musculoskeletal disease, chronic obstructive pulmonary disease (COPD), liver disease, and anxiety, compared to the non-users. The mean eGFR in the GLP-1 RA users was 70.9 mL/min/1.73 m² compared to 77.3 mL/min/1.73 m² in the non-users. In the patients with T2DM, the GLP-1 RA users (4.5%) were younger, had higher BMI, HbA1c, and higher prevalence of hypertension, liver disease, anxiety, and musculoskeletal diseases. Their mean eGFR was 71.4 mL/min/1.73 m² compared to 70.2 mL/min/1.73 m² in the non-users. Analysis of GLP-1 RA prescription trends revealed a significant upward trajectory across the overall AKD cohort and those with T2DM (Supplementary Table 3, available online). In the overall AKD cohort, the prescription rate increased from 0.04% in 2012–2013 to 1.13% in 2020–2022. The most pronounced increase occurred in the DM subgroup, in which the rate surged from 0.33% to 6.44%. These trends were statistically significant, with p-values of 0.023 for the overall cohort and DM subgroup, reflecting a clear increase over time (Fig. 2). Long acting agents (dulaglutide, semaglutide) accounted for 92% of new prescriptions in 2022.

Subgroup analysis was performed to examine trends in GLP-1 RA prescriptions based on eGFR, proteinuria, obesity, and the use of long- or short-acting formulations. Across all eGFR categories, a significant increase in GLP-1 RA prescriptions was observed from 2012 to 2022 (Supplementary Table 4 and Supplementary Fig. 2, available online). The most substantial increases occurred in the eGFR ≥15 to <30 mL/min/1.73 m² and ≥30 to <60 mL/min/1.73 m² groups, reflecting a substantial increase in prescription rates among patients with kidney impairment. Specifically, the eGFR ≥15 to <30 mL/min/1.73 m² group had an increase from 0.09% to 1.53% (slope, 0.1725; p = 0.031), while the ≥30 to <60 mL/min/1.73 m² group had an increase from 0.08% to 1.44% (slope, 0.1653; p = 0.018). The eGFR <15 mL/min/1.73 m² and eGFR ≥60 mL/min/1.73 m² groups also demonstrated notable, albeit smaller, increases over time. Similarly, GLP-1 RA use increased significantly among the patients with proteinuria, from 0.11% to 4.00% (p = 0.027), while the non-proteinuria group had an increase from 0.04% to 0.96% (p = 0.025) (Supplementary Table 5 and Supplementary Fig. 3, available online). These findings reflected the growing use of GLP-1 RAs, especially in patients with CKD stage 3, CKD stage 4, and proteinuria, indicating more aggressive management strategies in these high-risk groups. Comparisons of GLP-1 RA prescription rates between the patients with obesity (BMI ≥30 kg/m²) and without obesity (BMI <30 kg/m²) from 2012 to 2022 showed a consistent trend of higher prescription rates in both groups (Supplementary Table 6 and Supplementary Fig. 4, available online). In the obese patients, the rate increased from 0.09% to 2.12% (p = 0.015), while in the nonobese patients the rate increased from 0.03% to 0.75% (p = 0.03), indicating a greater use of GLP-1 RAs, particularly among those with higher BMI. In terms of formulations, there was a sharp increase in long-acting GLP-1 RA prescriptions from 0.04% to 1.09% (p = 0.023), while there was a minimal increase in short-acting prescriptions, remaining stable at around 0.04% (p = 0.146) (Supplementary Table 7 and Supplementary Fig. 5, available online).

Across the AKD cohort, dulaglutide and semaglutide were the most frequently prescribed GLP-1 RAs, with dulaglutide accounting for 27.73% of prescriptions overall, followed by semaglutide at 19.05%. This pattern was also found in the DM subgroup. Other GLP-1 RAs such as albiglutide, lixisenatide, and exenatide were prescribed less frequently, with usage rates below 3% across all groups. These findings underscore the growing preference for long-acting GLP-1 RAs in AKD management (Supplementary Fig. 6, available online).

We further performed multivariable logistic regression to analyze the associations between GLP-1 RA use and underlying patient characteristics. In the overall AKD cohort, each additional year of age decreased the odds of prescription (OR, 0.9; 95% CI, 0.94–0.95), while females (OR, 0.6; 95% CI, 0.60–0.67) and those of Black ethnicity (OR, 0.4; 95% CI, 0.41–0.48) or Asian ethnicity (OR, 0.9; 95% CI, 0.78–0.99) were less likely to receive GLP-1 RAs. Comorbidities including T2DM (OR, 6.6; 95% CI, 6.01–7.27), CKD stage 3 (OR, 1.7; 95% CI, 1.61–1.85), obesity (OR, 1.5; 95% CI, 1.46–1.63), and hypertension (OR, 1.4; 95% CI, 1.31–1.49), ischemic heart disease (OR, 1.1; 95% CI, 1.08–1.22), and COPD (OR, 1.1; 95% CI, 1.03–1.19) were associated with increased prescription rates. Conversely, the presence of anxiety (OR, 0.7; 95% CI, 0.64–0.74), cerebrovascular disease (OR, 0.8; 95% CI, 0.74–0.85), and liver disease (OR, 0.8; 95% CI, 0.74–0.87) were associated with lower prescription rates. In addition, the concurrent use of certain other medications significantly influenced GLP-1 RA prescription trends: insulin (OR, 3.4; 95% CI, 3.16–3.73), lipid-modifying agents (OR, 2.7; 95% CI, 2.48–2.83), metformin (OR, 2.2; 95% CI, 2.12–2.37), sodium-glucose cotransporter 2 (SGLT2) inhibitors (OR, 2.0; 95% CI, 1.78–2.13), renin-angiotensin system agents (OR, 1.5; 95% CI, 1.41–1.58), sulfonylureas (OR, 1.4; 95% CI, 1.31–1.50), and dipeptidyl peptidase-4 inhibitors (OR, 1.2; 95% CI, 1.09–1.30) were positively associated with GLP-1 RA use. In contrast, the use of anticoagulants (OR, 0.4; 95% CI, 0.41–0.46) and beta-blockers (OR, 0.8; 95% CI, 0.77–0.86) was associated with reduced prescription rates (Fig. 3). In the DM cohort, age, race, obesity, and CKD stage 3 played significant roles in determining the prescription rates, consistent with the overall AKD cohort. However, female sex (OR, 1.3; 95% CI, 1.22–1.38) was associated with a higher rate of GLP-1 RA prescriptions. Patients with CKD stages 4 (OR, 1.2; 95% CI, 1.04–1.41), proteinuria (OR, 1.4; 95% CI, 1.20–1.56), and those who used diuretics (OR, 1.1; 95% CI, 1.01–1.16) were associated with increased GLP-1 RA prescription rates, while those with cerebrovascular disease (OR, 0.8; 95% CI, 0.68–0.82), heart failure (OR, 0.8; 95% CI, 0.75–0.88), and COPD (OR, 0.9; 95% CI, 0.78–0.93) were associated with lower prescription rates (Supplementary Fig. 7, available online). A parallel analysis for the non‑diabetic subgroup is provided in Supplementary Fig. 8 (available online) for completeness.

Key predictors remained materially unchanged after the 180-day washout (Supplementary Figs. 9, 10; available online), confirming that pre-index users did not drive the primary associations.

The GLP-1 RA users had higher rates of tachycardia, depression, and gastrointestinal issues including abdominal pain, nausea/vomiting, constipation, diarrhea, and gastroparesis compared to the non-users. Fatigue, hernia, hypoglycemia, as well as a comparable incidence of suicide attempts were also more common among the GLP-1 RA users (Table 2). Differences in side effect profiles between the GLP-1 RA users and non-users are shown in Supplementary Fig. 11 (available online).

When stratified by baseline renal function, eGFR-stratified analysis revealed a valley-shaped pattern for nausea/vomiting (10.1% at <15, dipping to 8.5% at 15–29, rising to 9.5% at 30–59, and 10.9% at ≥60 mL/min/1.73 m²) and a monotonic increased in constipation as eGFR declined. Other gastrointestinal events were most frequent below 30 mL/min/1.73 m², whereas hypoglycemia and tachycardia displayed only modest variation across strata (Supplementary Table 8, available online).

Fig. 4 demonstrates that the GLP-1 RA users had a significantly higher risk of side effects, including hypoglycemic events (OR, 5.4; 95% CI, 4.87–5.99), depression (OR, 2.6; 95% CI, 2.51–2.80), and gastroparesis (OR, 2.6; 95% CI, 2.37–2.83). Tachycardia (OR, 2.1; 95% CI, 2.00–2.22), diarrhea (OR, 2.0; 95% CI, 1.85–2.21), and nausea/vomiting (OR, 2.0; 95% CI, 1.86–2.15) were also significantly more frequent in the GLP-1 RA users. The risks of other side effects including constipation, fatigue, hernia, abdominal pain, and acute cholecystitis were also higher in the GLP-1 RA users, with ORs ranging from 1.61 to 1.97. Importantly, the risk of suicide attempts did not differ significantly between the GLP-1 RA users and non-users (OR, 1.0; 95% CI, 0.90–1.11). Conversely, the GLP-1 RA users were associated with a lower risk of dementia (OR, 0.7; 95% CI, 0.60–0.81).

Supplementary Table 9 (available online) shows the temporal changes in the prevalence of associated side effects among the GLP-1 RA users over time. Fig. 5 illustrates these trends compared to the 2012–2013 baseline, showing a significant reduction in nausea/vomiting (slope, –0.65; p = 0.025). Reductions in hypoglycemia, dementia, depression, diarrhea, and fatigue were not statistically significant. On the other hand, a significant increase was observed in the trend for constipation (slope, +0.48; p = 0.044), while abdominal pain also showed an increase, but without significance.

Our results also identified several factors associated with GLP-1 RA prescription rates. Younger patients were more likely to receive GLP-1 RAs, with prescription rates declining as age increased, likely reflecting concerns about tolerability and potential side effects in older individuals, especially those with multiple comorbidities. Conversely, conditions such as anxiety, cerebrovascular disease, and liver disease were associated with lower prescription rates, likely due to concerns about tolerability in these vulnerable populations. Notably, obesity and renal function status played a significant role in influencing higher GLP-1 RA prescription rates, aligning with growing evidence that suggests these agents may offer kidney benefits independent of glycemic control [4,5,7–9]. The most substantial increases were observed in the patients with kidney impairment, specifically in the eGFR ≥30 to <60 mL/min/1.73 m² (CKD stage 3) and eGFR ≥15 to <30 mL/min/1.73 m² (CKD stage 4) groups. However, logistic regression analysis indicated that only CKD stage 3, but not CKD stage 4, was an independent predictor of GLP-1 RA prescription rate. This may reflect the clinical caution advised in current guidelines regarding the use of exendin-based GLP-1 RAs in patients with severe renal insufficiency (eGFR <30 mL/min/1.73 m²) [28,29]. This underscores the need for further clinical studies and revised evidence of their safety in high-risk populations [30].

In patients with DM, CKD stage 3, stage 4, and proteinuria were independently associated with higher GLP-1 RA prescriptions rates. Concurrent use of insulin, metformin, SGLT2 inhibitors, lipid-modifying agents, and renin-angiotensin system agents was also associated with higher utilization, indicating a multi-faceted approach to managing complex AKD cases [28–30]. Detailed analyses for the non diabetic subgroup are provided in Supplementary Material 1 (available online).

The most pronounced differences between the GLP-1 RA users and non-users were in hypoglycemic events, depression, gastrointestinal disturbances, and tachycardia, which is in line with previous reports [31–33]. This underscores the importance of careful monitoring, especially for hypoglycemia, when GLP-1 RAs are used in conjunction with other glucose-lowering medications in AKD patients. Of particular interest is the observed lower risk of dementia in the GLP-1 RAs users, suggesting potential neuroprotective effects, a finding that aligns with preclinical data indicating that GLP-1 RAs may influence neuroprotective pathways [34–38]. Notably, the incidence of suicide attempts did not differ between groups.

Over the study period, we observed a trend of reductions in most adverse events among the GLP-1 RA users, including a significant decrease in nausea/vomiting, which likely reflects wider adoption of gradual dose-titration protocols, increased utilization of once-weekly agents with flatter pharmacokinetic profiles, and physiologic tachyphylaxis to emetogenic stimuli. Rates of hypoglycemia, dementia, depression, diarrhea, and fatigue remained stable. In contrast, there was a significant increase in constipation, which may be due to the mechanism of action of GLP-1 RAs in slowing gastric motility, potentially requiring additional long-term management for these patients [39]. The growing prevalence of constipation, particularly in patients receiving long acting formulations, underscores the need for proactive bowel regimens (adequate hydration, dietary fiber, and early use of osmotic or stimulant laxatives) and systematic reassessment of gastrointestinal tolerability at follow-up visits to optimize adherence.

The valley-shaped curve for nausea/vomiting across eGFR strata suggests dose titration and tachyphylaxis may mitigate symptoms in moderate CKD, whereas very low or preserved eGFR may confer higher susceptibility. By contrast, constipation rose progressively as eGFR declined, indicating that proactive bowel regimens are particularly important in advanced CKD, while hypoglycemia and tachycardia risks remained stable regardless of renal function.

This study highlights the evolving role of GLP-1 RAs in AKD management, including for those with DM. The preference for long-acting formulations, such as dulaglutide and semaglutide, suggests that clinical decisions are strongly influenced by their convenience, more robust clinical data, as well as their favorable safety profile [5,8,26,27]. Nonetheless, the relatively low prescription rates, even among high-risk patients, point to possible barriers such as clinical inertia, limited practical knowledge, and concerns regarding adverse effects [40].

This study has several limitations. First, as an observational study, it assesses trends without establishing causality or directly predicting outcomes. Second, as a real-world study, it does not address personalized factors that may influence individual responses or clinical decisions. Further research in controlled settings is needed to validate our findings and investigate the long-term renal benefits and safety of GLP-1 RAs in diverse AKD populations. Third, the reliance on diagnostic codes for identifying AKD and comorbid conditions may have introduced misclassification bias, potentially leading to an underrepresentation of cases with milder disease. Fourth, while our analysis explored trends in GLP-1 RA prescriptions, we did not account for confounding variables such as socioeconomic status or provider-level factors, all of which could have influenced prescribing patterns. Fifth, the lack of information on specific clinical indications for initiating GLP-1 RA treatment during AKD episodes limits our ability to fully understand the drivers behind prescribing decisions. Sixth, the TriNetX database is derived from electronic health records across participating healthcare organizations, meaning that GLP-1 RA prescriptions in community settings outside of these healthcare organizations may not be fully captured. Although our 180-day washout sensitivity analysis showed consistent results, prescription fills that occur outside participating institutions may still be missed. Finally, because laboratory and urine-output fields required to stage KDIGO 1–2 AKI are incompletely captured across many participating sites, we restricted our analytic AKD cohort to survivors of dialysis-requiring AKI; accordingly, findings may not generalize to survivors of milder AKI who never necessitated RRT. Our findings highlight key trends and associations, however the current gaps underscore the importance of more comprehensive data collection in future research to fully elucidate the impact of GLP-1 RAs on post-AKI outcomes.

This study provides important insights into the real-world use of GLP-1 RAs in AKD patients, particularly among younger, obese patients with CKD stage 3 and T2DM. Despite potential benefits, overall utilization remains modest, and gastrointestinal adverse events, especially rising constipation, necessitate ongoing monitoring and supportive care to ensure safe, effective implementation.