Association between sodium‐glucose cotransporter 2 inhibitor use and clinical outcomes in patients with type 2 diabetes after urinary tract infection

Abstract

Aims

To evaluate whether initiating sodium‐glucose cotransporter 2 inhibitors (SGLT2i) within 3 months after a urinary tract infection (UTI) is associated with improved long‐term outcomes in patients with type 2 diabetes mellitus (T2DM), compared with dipeptidyl peptidase‐4 inhibitors (DPP‐4i).

Materials and Methods

A retrospective cohort study was conducted using the TriNetX US Collaborative Network (2016–2023). Adults with T2DM who initiated SGLT2i or DPP‐4i within 90 days after a UTI or pyelonephritis diagnosis were identified. Propensity score matching (1:1) was applied to balance baseline covariates. The primary outcome was mortality. Secondary outcomes included major adverse kidney events (MAKE), major adverse cardiovascular events (MACE), hospitalisation, dialysis dependence, emergency department (ED) visits, and sepsis. Outcomes were assessed over a 4‐year period using Cox proportional hazards models and Kaplan–Meier analysis.

Results

After matching, 2129 patients were included in each group. Compared with DPP‐4i, SGLT2i use was associated with lower risks of mortality (HR 0.59, 95% CI 0.45–0.77), MAKE (HR 0.78), hospitalisation (HR 0.89), dialysis dependence (HR 0.43), sepsis (HR 0.75), and ED visits (HR 0.88). No significant difference was found for MACE (HR 0.91, 95% CI 0.79–1.04).

Conclusions

In patients with T2DM recovering from UTI, the use of SGLT2 inhibitors was associated with lower risks of mortality, kidney complications, and infection‐related outcomes compared with DPP‐4 inhibitors. These findings support the safety and clinical benefit of continuing or initiating SGLT2i during the post‐infectious phase, even after a recent urinary tract infection.

1. INTRODUCTION

Infections remain a frequent and clinically significant complication in individuals with type 2 diabetes mellitus (T2DM), primarily due to persistent hyperglycaemia, immune dysregulation, and endothelial dysfunction which together impair host defence mechanisms. ¹ , ² , ³ , ⁴ Among these, urinary tract infections (UTIs) and acute pyelonephritis are especially common and may trigger prolonged inflammatory responses, metabolic instability, and organ vulnerability in diabetic patients. ⁵ The period following such infections—often termed the “vulnerable recovery phase”—has been associated with increased risks of systemic complications, renal function deterioration, and cardiovascular events. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ Despite growing awareness of this transitional window, there remains limited evidence to guide the selection of optimal antihyperglycaemic agents that can safely support recovery while minimizing long‐term adverse outcomes.

Sodium‐glucose cotransporter 2 inhibitors (SGLT2i) have emerged as a preferred therapeutic class for patients with T2DM, owing to their demonstrated benefits in reducing cardiovascular mortality, slowing kidney disease progression, and improving heart failure outcomes in multiple randomised controlled trials. ¹¹ , ¹² , ¹³ These organ‐protective effects are thought to arise from pleiotropic mechanisms, including natriuresis, modulation of tubuloglomerular feedback, reduction in oxidative stress, and anti‐inflammatory activity. ¹⁴ , ¹⁵ , ¹⁶ However, due to their known association with genital and urinary tract infections, clinicians have exercised caution regarding their use in patients with a recent history of UTI. While meta‐analyses suggest that the risk of serious infections such as urosepsis remains low with SGLT2i, the perceived infection‐related risk may influence prescribing decisions in real‐world settings. ¹⁷ , ¹⁸ , ¹⁹ Consequently, many clinicians delay or avoid initiating SGLT2i in the post‐UTI period, potentially foregoing the opportunity for cardiometabolic benefit.

Dipeptidyl peptidase‐4 inhibitors (DPP‐4i), by contrast, are often considered a safer choice in this post‐infectious context. With a favourable safety profile and neutral infection risk, they are frequently prescribed during recovery, making them a suitable active comparator. ²⁰ , ²¹ , ²² While both drug classes have comparable glycaemic efficacy, SGLT2i confer superior cardiorenal benefits in high‐risk T2DM populations. ²³ , ²⁴ , ²⁵ , ²⁶ Whether these advantages apply to the post‐infectious period, and whether SGLT2i can be safely used after UTI, remains an important clinical question.

To address this gap, we conducted a retrospective, multicentre cohort study using the TriNetX US Collaborative Network to evaluate real‐world outcomes in patients with T2DM who initiated SGLT2i or DPP‐4i therapy within 3 months after a documented UTI or acute pyelonephritis. Rather than testing superiority, DPP‐4i served as an active comparator to reflect routine practice and account for confounding by indication. Through this active‐comparator design and propensity score matching, we examined whether initiating SGLT2i post‐UTI is associated with better long‐term outcomes—including mortality, kidney function decline, infection‐related complications, and hospitalisation—thus providing a novel evaluation of SGLT2i viability as an advantageous treatment approach during recovery in this high‐risk group.

2. MATERIALS AND METHODS

2.1. Data sources

TriNetX is built on a continually updated network summarising de‐identified patient data from the electronic health records and insurance claims of member healthcare organisations. The platform does not provide users with access to raw patient data and adheres to privacy regulations such as HIPAA and GDPR, ensuring the safeguarding of patient confidentiality. Researchers have access to built‐in tools like Query Builder and Analytics to analyse large sets of summarised patient data and explore associations between treatments and clinical outcomes.

2.2. Ethics statement

This study was approved by the Institutional Review Board of Shin Kong Wu Ho‐Su Memorial Hospital (IRB number: 20240812R) and the patients' informed consent was waived because TriNetX provides de‐identified summaries of patient data based on selection and analytic criteria.

2.3. Study design and study population

This retrospective cohort study employed an intention‐to‐treat and active‐comparator design to evaluate the association between clinical outcomes and SGLT2i use in patients with UTI or acute pyelonephritis, using data obtained from the TriNetX US Collaborative Network (Figure S1). As illustrated in Figure 1, the study began with a source population of 115 192 013 individuals within the TriNetX network. Patients were eligible for inclusion if they had at least two clinical visits, were aged 18 years or older, and had a diagnosis of T2DM between 1 January 2016 and 31 December 2023, resulting in 5 800 230 eligible individuals. From this group, 903 671 patients with UTI or acute pyelonephritis were further identified. The index date was defined as 3 months after the diagnosis of UTI or pyelonephritis. Rigorous exclusion criteria were then applied, removing those with adult‐type polycystic kidney disease (n = 1430), organ transplantation before the index date (n = 28 158), immunological agent use within 6 months prior (n = 33 934), any malignancy before the index date (n = 156 853), CVD within 6 months prior (n = 160 023), cutaneous lupus erythematosus (n = 254), and pregnancy within the preceding year (n = 2368).

FIGURE 1.

Schema of patient enrolment in the study. DPP‐4i, dipeptidyl peptidase‐4 inhibitor; PSM, propensity score matching; SGLT2i, sodium‐glucose co‐transporter 2 inhibitor; T2DM, type 2 diabetes mellitus; UTI, urinary tract infection.

Finally, patients were classified based on receipt of SGLT2i therapy (without DPP‐4i use, n = 4805) or DPP‐4i therapy (without SGLT2i use, n = 8839) within 3 months following their diagnosis of UTI or pyelonephritis. To avoid exposure misclassification, patients who received both SGLT2 inhibitors and DPP‐4 inhibitors during the 3‐month post‐UTI window were excluded from cohort assignment. Medication data from the 1‐year baseline period were retained for covariate adjustment and descriptive analyses. Accordingly, the reported percentages of SGLT2i use in the DPP‐4i group and vice versa in Table 1 reflect historical medication use rather than treatment allocation. We adhered to the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline for observational studies. ²⁷

TABLE 1.

Baseline characteristics of study population before and after propensity score matching.

Parameters	Before matching			After matching a
	SGLT2i users (n = 4805)	DPP‐4i users (n = 8839)	SMD b	SGLT2i users (n = 2129)	DPP‐4i users (n = 2129)	SMD b
Age at index (years)	60.7 ± 12.5	65.2 ± 12.5	0.359	61.4 ± 13	61.2 ± 13.1	0.013
Male, n (%)	1343 (28)	2414 (27.3)	0.014	593 (27.9)	616 (28.9)	0.024
BMI (kg/m2), mean ± SD	33.5 ± 8.1	31.9 ± 8.4	0.188	33.3 ± 8.2	33.5 ± 8.7	0.019
Race, n (%)
White people	3443 (71.7)	5988 (67.7)	0.085	1474 (69.2)	1484 (69.7)	0.01
African American	680 (14.2)	1304 (14.8)	0.017	335 (15.7)	316 (14.8)	0.025
Asian	256 (5.3)	555 (6.3)	0.041	113 (5.3)	118 (5.5)	0.01
Other race	84 (1.7)	260 (2.9)	0.079	41 (1.9)	38 (1.8)	0.01
Comorbidity, n (%)
Nicotine dependence	601 (12.5)	1072 (12.1)	0.012	266 (12.5)	264 (12.4)	0.003
Alcohol related disorders	95 (2)	189 (2.1)	0.011	45 (2.1)	42 (2)	0.01
Hypertensive diseases	3868 (80.5)	7301 (82.6)	0.054	1662 (78.1)	1665 (78.2)	0.003
Dyslipidaemia	3578 (74.5)	6151 (69.6)	0.109	1455 (68.3)	1460 (68.6)	0.005
Heart failure	154 (3.2)	250 (2.8)	0.022	54 (2.5)	53 (2.5)	0.003
Ischemic heart diseases	1006 (20.9)	1750 (19.8)	0.028	402 (18.9)	400 (18.8)	0.002
Cerebrovascular diseases	346 (7.2)	784 (8.9)	0.061	159 (7.5)	170 (8)	0.019
Aortic aneurysm and dissection	59 (1.2)	92 (1)	0.018	22 (1)	23 (1.1)	0.005
PAOD	98 (2)	207 (2.3)	0.021	44 (2.1)	42 (2)	0.007
Disorders of thyroid gland	1131 (23.5)	2112 (23.9)	0.008	458 (21.5)	446 (20.9)	0.014
Pulmonary heart disease	131 (2.7)	243 (2.7)	0.001	52 (2.4)	53 (2.5)	0.003
COPD	430 (8.9)	940 (10.6)	0.057	177 (8.3)	176 (8.3)	0.002
Fatty liver	453 (9.4)	583 (6.6)	0.104	175 (8.2)	174 (8.2)	0.002
Liver cirrhosis	144 (3)	235 (2.7)	0.02	58 (2.7)	63 (3)	0.014
Peptic ulcer	22 (0.5)	45 (0.5)	0.007	10 (0.5)	10 (0.5)	<0.001
Gout	164 (3.4)	401 (4.5)	0.058	70 (3.3)	69 (3.2)	0.003
Anxiety disorders	1102 (22.9)	1759 (19.9)	0.074	439 (20.6)	443 (20.8)	0.005
Medication c , n (%)
Insulin	2487 (51.8)	4345 (49.2)	0.052	960 (45.1)	956 (44.9)	0.004
Sulfonylureas	1259 (26.2)	2648 (30)	0.084	468 (22)	471 (22.1)	0.003
Biguanides	4023 (59.9)	6040 (53.3)	0.135	1236 (45.6)	1241 (45.8)	0.003
Thiazolidinediones	342 (7.1)	510 (5.8)	0.055	105 (4.9)	97 (4.6)	0.018
DPP4i	540 (11.2)	5258 (59.5)	1.169	454 (21.3)	468 (22)	0.016
SGLT2i	2984 (62.1)	545 (6.2)	1.461	443 (20.8)	450 (21.1)	0.008
GLP‐1 RA	1543 (32.1)	655 (7.4)	0.653	312 (14.7)	296 (13.9)	0.021
Statin	3002 (62.5)	4997 (56.5)	0.121	1036 (48.7)	1036 (48.7)	<0.001
Fibrates	314 (6.5)	498 (5.6)	0.038	99 (4.7)	90 (4.2)	0.021
RASB	2684 (55.9)	4601 (52.1)	0.076	938 (44.1)	933 (43.8)	0.005
Beta‐blocker	1786 (37.2)	3298 (37.3)	0.003	653 (30.7)	658 (30.9)	0.005
Calcium channel blockers	1185 (24.7)	2370 (26.8)	0.049	418 (19.6)	427 (20.1)	0.011
Antithrombotic agents	2107 (43.9)	4282 (48.4)	0.092	839 (39.4)	856 (40.2)	0.016
NSAID	2136 (44.5)	3438 (38.9)	0.113	798 (37.5)	813 (38.2)	0.015

Laboratory, mean ± SD							Non‐missing (%) SGLT2i	Non‐missing (%) DPP4i
eGFR (mL/min/1.73 m2)	73.9 ± 29.2	65.4 ± 30.5	0.284	72.2 ± 30.1	69.6 ± 31.9	0.082	86.8%	87.4%
Urea nitrogen (mg/dL)	19.2 ± 11.9	21.7 ± 15.1	0.183	19.8 ± 14	20.6 ± 14.4	0.051	83.8%	84.5%
Sodium (mmol/L)	137.5 ± 3.9	137 ± 4.6	0.13	137 ± 4.3	136.9 ± 4.6	0.038	87.2%	88.0%
Potassium (mmol/L)	4.2 ± 0.5	4.2 ± 0.6	0.046	4.1 ± 0.5	4.2 ± 0.5	0.050	85.9%	86.9%
Calcium (mg/dL)	9.4 ± 0.6	9.3 ± 0.7	0.122	9.3 ± 0.6	9.3 ± 0.7	0.034	85.4%	86.4%
Phosphate (mg/dL)	3.5 ± 0.9	3.4 ± 1.0	0.085	3.4 ± 0.9	3.5 ± 1.2	0.101	17.0%	17.8%
Albumin (g/dL)	4 ± 0.5	3.8 ± 0.6	0.237	3.9 ± 0.6	3.9 ± 0.6	0.035	79.3%	80.4%
Haemoglobin (g/dL)	13.3 ± 2	12.5 ± 2.0	0.407	13.1 ± 2.0	13 ± 2.0	0.059	81.4%	81.9%
ALT (U/L)	28.2 ± 30.7	28.2 ± 35.6	0.002	28.8 ± 26.4	30.6 ± 42.2	0.051	80.7%	82.2%
AST (U/L)	26.9 ± 32.2	27.9 ± 37.5	0.03	28.2 ± 30.6	29.8 ± 50.2	0.038	79.4%	80.8%
LDL (mg/dL)	88.3 ± 38.5	86.6 ± 38.7	0.042	88.9 ± 39	91.1 ± 39.7	0.056	51.9%	52.0%
HDL (mg/dL)	44.9 ± 13.7	44.9 ± 14.8	0.001	44.5 ± 13.9	43.7 ± 14.1	0.055	53.1%	52.8%
Triglyceride (mg/dL)	189.9 ± 157.9	173.7 ± 131.1	0.112	192.4 ± 166.2	184.2 ± 157.6	0.051	52.6%	52.7%
Haemoglobin A1c (%)	8.2 ± 2.1	7.9 ± 2.2	0.139	8.4 ± 2.3	8.3 ± 2.4	0.039	64.6%	62.8%
UPCR (mg/g)	559.6 ± 1235.3	861.2 ± 1853.2	0.191	698.2 ± 1769	973.9 ± 2425.1	0.130	2.3%	2.1%

Abbreviations: AKI, acute kidney injury; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; COPD, chronic obstructive pulmonary disease; DPP‐4i, dipeptidyl peptidase‐4 inhibitor; eGFR, estimated glomerular filtration rate; ED, emergency department; GLP‐1 RA, glucagon‐like peptide‐1 receptor agonist; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; MAKE, major adverse kidney events; MACE, major adverse cardiovascular events; NSAID, non‐steroidal anti‐inflammatory drug; PAOD, peripheral arterial occlusive disease; RASB, renin‐angiotensin system blocker; SD, standard deviation; SGLT2i, sodium‐glucose cotransporter 2 inhibitor; SMD, standardised mean difference; UPCR, urine protein‐to‐creatinine ratio.

^a The propensity score matching includes all parameters in this table to achieve balanced cohorts.

^b A SMD value below 0.1 indicates a negligible difference between groups.

^c Medication use percentages reflect historical prescriptions during the 1‐year baseline period and do not indicate exposure status used for cohort assignment. Patients prescribed both SGLT2i and DPP‐4i during the 90‐day post‐UTI exposure window were excluded.

2.4. Covariates

Baseline parameters were collected during the 1 year preceding the index date to ensure a thorough comparison of patient characteristics between the two cohorts. Demographically, these parameters included age at index in years, sex distribution, body mass index (BMI), and racial categories including White, African American, Asian, and other races. Additionally, various comorbidities were documented, including nicotine dependence, alcohol‐related disorders, hypertensive diseases, dyslipidaemia, heart failure, ischaemic heart diseases (IHD), cerebrovascular diseases, aortic aneurysm and dissection, atherosclerosis of native arteries of the extremities, disorders of the thyroid gland, pulmonary heart disease, chronic obstructive pulmonary disease, fatty liver, liver cirrhosis, peptic ulcer, gout, and anxiety disorders. Furthermore, medication history was thoroughly recorded, encompassing insulin use, sulfonylureas, thiazolidinediones, DPP4is, SGLT2is, glucagon‐like peptide‐1 receptor agonists (GLP‐1 RA), statins, fibrates, renin‐angiotensin system blockade, beta‐blockers (RASB), calcium channel blockers, antithrombotic agents, and non‐steroidal anti‐inflammatory drugs. Moreover, comprehensive laboratory parameters were measured, including estimated glomerular filtration rate (eGFR), urea nitrogen, sodium, potassium, calcium, phosphate, albumin, haemoglobin, alanine aminotransferase, aspartate aminotransferase, low‐density lipoprotein, high‐density lipoprotein, triglycerides, haemoglobin A1c, and protein‐to‐creatinine ratio in urine (UPCR). eGFR was calculated using the CKD‐EPI 2021 equation without race coefficient, following current international guidelines to enhance applicability across racial groups. This extensive collection of baseline characteristics enabled the creation of balanced cohorts through propensity score matching (PSM), thus facilitating more reliable comparisons between treatment groups. The diagnosis codes are shown in Method S1.

2.5. Study outcomes

The primary outcome assessed was mortality. Secondary outcomes included major adverse cardiovascular events (MACE), which comprised myocardial infarction, stroke, and death, and major adverse kidney events (MAKE), which encompassed acute kidney injury, end‐stage renal disease (ESRD), initiation of dialysis, and death. Additionally, all‐cause hospitalisation was systematically measured. For specificity analysis, the study observed the incidence of acute myocardial infarction, heart failure, stroke, and ESRD requiring dialysis. Other important outcomes included the frequency of emergency department (ED) visits, the development of sepsis, and episodes of acute kidney injury. The outcomes of interest were followed up within a specified timeframe of 1 day to 4 years following the index date. If these clinical events did not occur, observation and data collection were continued until the last available data entry or the end of the study period on 9 June 2025, whichever came first, to ensure consistent follow‐up duration across patients. The diagnosis codes used to define all outcomes are provided in Method S1.

2.6. Sensitivity analysis

To ensure the robustness of our findings, we performed three sensitivity analyses. First, we applied a new‐user design for outcome evaluation. Second, we used alternative PSM models, adjusting for multiple confounders to minimise bias. Third, we assessed risk across different time frames to address violations of the proportional hazard assumption, enhancing the temporal validity of our results.

Additionally, a landmark analysis was conducted to examine clinical outcomes associated with SGLT2i at key intervals—specifically within 6 and 12 months post‐UTI or pyelonephritis—providing further insight into the timing and durability of therapeutic effects.

2.7. Statistical analysis

Categorical variables were reported as counts and percentages; continuous variables as means with standard deviations. Standardised mean differences (SMD) were used to assess group balance, with <0.1 indicating good balance and <0.2 suggesting a small effect size. ²⁸ , ²⁹ PSM was used to create comparable SGLT2i and DPP‐4i user groups, reducing baseline confounding. During matching, age at index was analysed as a continuous variable, while sex, race, and comorbidities were treated as categorical (present or absent). Laboratory data were considered present if available within the time window and absent otherwise. ³⁰ , ³¹ , ³² Logistic regression generated propensity scores, and matching was conducted using a greedy nearest‐neighbour algorithm with a 0.1 pooled SD calliper in Python and R 3.4.4. Details are provided in Method S2.

Cox proportional hazards models were used to estimate hazard ratios (HRs), with proportionality verified using the generalised Schoenfeld method in TriNetX. Kaplan–Meier curves with log‐rank tests evaluated event‐free probabilities. Subgroup analyses assessed effect modification across sex, age (≥65, <65), race (White, African American, Asian), BMI (≥30, <30 kg/m²), HbA1c (≥7%, <7%, ≥8%, <8%), eGFR (≥60, <60 mL/min/1.73 m²), CVD, and IHD status.

All analyses were conducted using TriNetX. A two‐sided p‐value <0.05 was considered statistically significant. R version 4.4.2 (Free Software Foundation Inc.) with Forestploter and ggplot2 packages was used for figure generation.

3. RESULTS

3.1. Baseline characteristics of patients

Table 1 shows the baseline characteristics of SGLT2i and DPP‐4i users both before and after PSM. Before matching, there were 4805 SGLT2i users and 8839 DPP4i users, with notable disparities observed between the groups. SGLT2i users were younger (mean age 60.7 vs. 65.2 years; SMD = 0.36), had a higher BMI (33.5 vs. 31.9 kg/m²; SMD = 0.19), and were more likely to be White (71.7% vs. 67.7%). Differences extended to comorbidities: SGLT2i users had higher rates of dyslipidaemia (74.5% vs. 69.6%), anxiety disorders (22.9% vs. 19.9%), and fatty liver (9.4% vs. 6.6%), along with decreased rates of cerebrovascular disease (7.2% vs. 8.9%). Distinct differences were also evident in baseline medication use, with SGLT2i users being more frequently prescribed insulin and GLP‐1 receptor agonists, and less likely to be on sulfonylureas or DPP‐4is. Laboratory values showed that SGLT2i users had a higher eGFR (73.9 vs. 65.4 mL/min/1.73 m²; SMD = 0.28), higher haemoglobin, slightly better albumin, and higher triglyceride levels.

Following PSM, 2129 well‐matched individuals remained in each group. Demographic and clinical characteristics became closely aligned: mean ages were nearly identical at 61.4 and 61.2 years, male proportions were 27.9% and 28.9%, and BMI averaged around 33.4 kg/m² in both groups. Notably, most SMDs were below 0.1, demonstrating excellent balance between the groups.

3.2. Recurrent UTI risk after SGLT2i initiation

To assess whether SGLT2i use increases the risk of recurrent UTI, we performed a subgroup analysis among patients with a history of UTI. As shown in Table 2, the recurrence rate was slightly lower in the SGLT2i group (30.9%) than in the DPP‐4i group (34.3%). The estimated hazard ratio was 0.952 (95% CI: 0.870–1.041; p = 0.283), indicating no statistically significant difference.

TABLE 2.

Clinical outcomes in patients with type 2 diabetes after urinary tract infection.

Clinical outcomes	After propensity score matching within 4‐year follow‐up
	SGLT2i user (n = 2129)		DPP4i user (n = 2129)		SGLT2i vs. DPP‐4i
	Events (n)	Risk (%)	Events (n)	Risk (%)	HR (95% CI)	p value
Primary outcome
Mortality	83	3.9	158	7.4	0.59 (0.40–0.77)	<0.001 a
Secondary outcome
MACE b	375	17.6	449	21.1	0.91 (0.79–1.04)	0.166
MAKE c	349	16.4	469	22	0.78 (0.68–0.90)	0.001
All‐cause hospitalisation	660	31	770	36.2	0.89 (0.80–0.99)	0.029
Specificity analysis
UTI recurrence	910	30.9	1011	34.3	0.95 (0.87–1.04)	0.283
AMI	101	4.7	102	4.8	1.12 (0.85–1.47)	0.431
Heart failure	213	10	247	11.6	0.94 (0.79–1.13)	0.539
Stroke	93	4.4	94	4.4	1.09 (0.82–1.45)	0.575
ESRD on dialysis	24	1.1	62	2.9	0.43 (0.27–0.69)	<0.001
ED visits	768	36.1	910	42.7	0.88 (0.80–0.97)	0.007
Sepsis	186	8.7	265	12.4	0.75 (0.62, 0.91)	0.003
AKI	294	13.8	380	17.8	0.82 (0.71, 0.96)	0.011

Abbreviations: AKI, acute kidney injury; AMI, acute myocardial infarction; CI, confidence interval; DPP4i, dipeptidyl peptidase‐4 inhibitor; ED, emergency department; ESRD, end‐stage renal disease; HR, hazard ratio; MACE, major adverse cardiovascular events; MAKE, major adverse kidney events; SGLT2i, sodium‐glucose cotransporter 2 inhibitor.

^a Indicates that the proportional hazards assumption was violated.

^b MACE includes myocardial infarction, stroke, and death.

^c MAKE includes acute kidney injury, end stage of kidney disease, dialysis, and death.

3.3. Follow‐up duration after matching

The distribution of follow‐up time in each treatment group is illustrated in Figure S2. After propensity score matching, the mean follow‐up duration was 932 ± 489 days in the SGLT2i group and 1045 ± 504 days in the DPP‐4i group. The corresponding median durations were 983 days (interquartile range, IQR: 906) and 1357 days (IQR: 825), respectively. Given the right‐skewed distribution of follow‐up times, the median and IQR are more appropriate for representing central tendency and variability. These comparable follow‐up durations between groups support the validity of the subsequent time‐to‐event analyses.

3.4. SGLT2i effects after UTI: Primary and secondary outcomes

Figure 2 displays four‐year Kaplan–Meier survival curves comparing SGLT2i and DPP‐4i users. For the primary outcome, SGLT2i users had significantly lower mortality (log‐rank p < 0.001). Regarding secondary outcomes, SGLT2i users experienced fewer MAKE (p = 0.001), no significant difference in MACE (p = 0.116), and significantly lower all‐cause hospitalisation rates (p = 0.029), further supporting their favourable outcomes.

FIGURE 2.

Kaplan–Meier survival curves comparing SGLT2i and DPP‐4i users initiated within 3 months after UTI for (A) Mortality, (B) major adverse cardiovascular events (MACE), (C) major adverse kidney events (MAKE), and (D) all‐cause hospitalisation. DPP‐4i, dipeptidyl peptidase‐4 inhibitor; SGLT2i, sodium‐glucose co‐transporter 2 inhibitor.

Table 2 provides detailed comparisons of clinical outcomes between the SGLT2i and DPP‐4i groups. The analysis shows that the primary outcome of mortality was significantly lower in the SGLT2i group (3.9% vs. 7.4%, HR 0.589, 95% CI 0.451–0.769). Among secondary outcomes, SGLT2i users had a reduced risk of MAKE (16.4% vs. 22.0%, HR 0.784, 95% CI 0.682–0.900) and all‐cause hospitalisation (31.0% vs. 36.2%, HR 0.891, 95% CI 0.803–0.988), while there was no significant difference in MACE (17.6% vs. 21.1%, HR 0.908, 95% CI 0.791–1.041).

3.5. SGLT2i effects after UTI: Specificity analysis of outcomes

The Kaplan–Meier survival curves show that there were no significant differences between groups for acute myocardial infarction, heart failure, or stroke, as shown by high log‐rank p‐values (Figure S3). However, SGLT2i users consistently showed significantly better outcomes for kidney and infection‐related events, including a lower incidence of acute kidney injury (p = 0.011), reduced progression to dialysis dependence (p < 0.001) (Figure S4), less sepsis (p = 0.003), and fewer ED visits (p = 0.007) (Figure S5). These findings highlight the renal and infection‐related protective benefits of SGLT2i therapy.

The specificity analysis in Table 2 demonstrates that SGLT2i users had a significantly lower risk of developing ESRD requiring dialysis (1.1% vs. 2.9%, HR 0.43, 95% CI 0.269–0.690). The risk of sepsis was also notably reduced in the SGLT2i group (8.7% vs. 12.4%, HR 0.75, 95% CI 0.62–0.91). Additionally, acute kidney injury occurred less frequently in the SGLT2i cohort (13.8% vs. 17.8%, HR 0.82, 95% CI 0.71–0.96), and ED visits were lower (51.8% vs. 57.1%, HR 0.88, 95% CI 0.80–0.97). However, there were no significant differences between the groups for acute myocardial infarction, heart failure, or stroke.

3.6. Subgroup analysis

Figure 3 displays subgroup analyses of mortality comparing SGLT2i and DPP‐4i users. Significant reductions in mortality with SGLT2i were observed among females, patients under 65 years, those with HbA1c ≥7%, and in patients without CVD or without IHD. In contrast, no significant mortality differences were observed among males, older adults, patients of other races, those with higher BMI, lower HbA1c, or reduced eGFR.

FIGURE 3.

Subgroup analysis of mortality comparing SGLT2i versus DPP‐4i users. Hazard ratios (HRs) with 95% confidence intervals (CIs) are shown for each subgroup. The vertical dashed line indicates the null value (HR = 1). BMI, body mass index; CVD, cardiovascular disease; DPP‐4i, dipeptidyl peptidase‐4 inhibitor; eGFR, estimated glomerular filtration rate; HbA1c, haemoglobin A1c; IHD, ischaemic heart disease; SGLT2i, sodium‐glucose co‐transporter 2 inhibitor.

For the outcome of MACE (Figure S6), there is no significant difference between SGLT2i and DPP‐4i users in any subgroup. For MAKE (Figure S7), SGLT2i shows greater benefit in specific subgroups, particularly among patients under 65 years, White patients, those with HbA1c ≥7% but <8%, and those without CVD or IHD. For all‐cause hospitalisation (Figure S8), SGLT2i is generally favoured, with significant reductions observed among males, White and African American patients, individuals with eGFR <60 mL/min/1.73 m², those without CVD or IHD, and those with HbA1c ≥7%.

3.7. Landmark analysis of different time points

Landmark analyses at various time points consistently showed benefits for SGLT2i users (Table S1). Within 6 months after UTI, SGLT2i was associated with significantly lower risks of mortality (HR 0.37, 95% CI 0.30–0.45), MACE (HR 0.78, 95% CI 0.71–0.86), MAKE (HR 0.78, 95% CI 0.71–0.85), and all‐cause hospitalisation (HR 0.88, 95% CI 0.82–0.94). These favourable associations persisted at 12 months, with SGLT2i use linked to lower mortality (HR 0.69, 95% CI 0.59–0.80), MACE (HR 0.92, 95% CI 0.86–0.99), MAKE (HR 0.78, 95% CI 0.72–0.83), and all‐cause hospitalisation (HR 0.93, 95% CI 0.88–0.99) compared with DPP‐4i.

3.8. Sensitivity analysis

Sensitivity analyses demonstrated consistent findings across various analytic frameworks. In the new‐user design presented in Table S2, results remained directionally consistent, reinforcing the validity of our primary findings. Across alternative propensity score matching models with increasingly comprehensive covariate adjustment (Table S3), the associations between SGLT2i use and reduced risks of mortality, kidney outcomes, and hospitalisation remained robust. Additionally, comparisons across different follow‐up durations from 1 to 3 years (Table S4) showed sustained trends favouring SGLT2i over time.

To evaluate whether baseline proteinuria or infection severity confounded our results, we performed an additional PSM including urine protein and hospitalisation at index UTI as covariates. As shown in Table S5, both variables were well balanced after matching. In this refined cohort, clinical outcomes (Table S6) remained consistent with our main findings, with SGLT2i users showing lower risks of mortality and MAKE.

4. DISCUSSION

In this large, multicentre, propensity score–matched cohort study, we evaluated real‐world outcomes in patients with T2DM who initiated SGLT2i or DPP‐4i within 3 months after UTI or acute pyelonephritis. SGLT2i use was associated with significantly lower risks of mortality, MAKE, and all‐cause hospitalisation compared with DPP‐4i. Although the difference in MACE was not statistically significant, SGLT2i users showed meaningful reductions in acute kidney injury, dialysis dependence, ED visits, and sepsis. Importantly, SGLT2i use was not associated with an increased risk of recurrent UTI, even among patients with a prior infection history, thereby alleviating concerns that have traditionally constrained their use in this population. These protective effects were robust across subgroups and sensitivity analyses, with early divergence in Kaplan–Meier curves suggesting a rapid benefit. This supports timely SGLT2i initiation post‐infection, even among patients without overt CVD but with elevated risk of organ stress. Contrary to concerns, our findings argue against delaying SGLT2i during recovery, as this may forgo substantial cardiorenal and infection‐related protection. By using DPP‐4i—a class with neutral cardiorenal effects—as the comparator, we contextualised SGLT2i's added benefits in a clinically relevant scenario. Overall, the post‐UTI phase may represent a window for effective SGLT2i initiation, offering not only glycaemic control but also mortality and kidney protection.

The observed survival benefit aligns with cardiovascular outcome trials such as Empagliflozin Cardiovascular Outcome Event Trial in Type 2 Diabetes Mellitus Patients (EMPA‐REG OUTCOME), the Canagliflozin Cardiovascular Assessment Study (CANVAS), and the Dapagliflozin Effect on Cardiovascular Events–Thrombolysis in Myocardial Infarction 58 (DECLARE–TIMI 58), which demonstrated all‐cause mortality reductions with SGLT2i, albeit with variability in effect size and consistency. ¹¹ , ¹² , ¹³ Meta‐analyses further support a class‐wide mortality benefit, especially for empagliflozin. ³³ , ³⁴ Our study adds evidence for mortality reduction in a post‐UTI context, a phase marked by heightened systemic stress but often underrepresented in trials. ⁵ Notably, survival benefit was seen even among patients without prior CVD, suggesting that SGLT2i may mitigate early pathophysiological processes—such as inflammation, endothelial dysfunction or autonomic imbalance—triggered by infection. ³⁵ , ³⁶ Subgroup analyses showed stronger effects among women, younger individuals, and those without baseline CVD, who may retain greater physiological reserve. The early separation in survival curves (Figure 2A) underscores the potential value of timely SGLT2i initiation. Although not designed to establish superiority, the comparison with DPP‐4i argues against deferring SGLT2i initiation during recovery.

While major trials have shown that SGLT2i reduce MACE, our study did not observe a statistically significant MACE difference when initiated during the early recovery phase. In EMPA‐REG OUTCOME and CANVAS, empagliflozin and canagliflozin achieved ~14% relative MACE risk reductions, while dapagliflozin did not show statistical significance. ¹¹ , ¹² , ¹³ Observational studies likewise support the cardiovascular benefit of SGLT2i over DPP‐4i. One multi‐database cohort study (>200 000 patients) found a 24% MACE reduction with SGLT2i (HR 0.76, 95% CI 0.69–0.84), ³⁷ and a population‐based analysis in Hong Kong reported decreased stroke (HR 0.64), cardiovascular death (HR 0.39), and all‐cause mortality (HR 0.44) with SGLT2i versus DPP‐4i in over 40 000 patients. ³⁸ Possible reasons for our null MACE finding include low absolute risks limiting statistical power, transient physiological stabilisation post‐infection, ³⁹ and widespread baseline use of statins, RAS blockers, and antithrombotics that may have obscured incremental benefit. Though HRs for MACE were numerically lower among patients with baseline CVD, our study was not powered for formal interaction testing. Further research with enriched high‐risk populations and longer follow‐up is needed. Still, our findings suggest that early post‐UTI SGLT2i use does not increase cardiovascular risk and is likely safe.

We also observed a significant reduction in MAKE with SGLT2i, including lower risks of AKI and ESRD requiring dialysis. This reinforces known nephroprotective properties of SGLT2i and extends them to patients recovering from infections—a group not typically included in RCTs. ⁴⁰ , ⁴¹ Importantly, many patients in our cohort lacked baseline CKD, suggesting that SGLT2i can benefit individuals with preserved renal function during a period of heightened renal stress. Mechanistically, benefits may derive from the restoration of tubuloglomerular feedback, reductions in glomerular hyperfiltration and renal oxygen demand, and the mitigation of interstitial inflammation. ⁴² , ⁴³ These effects are especially relevant in the post‐infectious state. Moreover, SGLT2i reduce albuminuria independent of glycaemic or BP control, highlighting their role as disease‐modifying agents beyond glycaemic control. ⁴⁴ The consistency of our results across time windows and models supports the robustness of these renal benefits.

In addition, SGLT2i users had significantly lower risks of all‐cause hospitalisation, ED visits, and sepsis, indicating broader post‐infectious protective effects (Figure S5). These observations echo prior real‐world studies showing fewer hospitalisations and infectious complications among SGLT2i users versus DPP‐4i. ⁴⁵ , ⁴⁶ Potential mechanisms include improved renal haemodynamics, systemic anti‐inflammatory effects, and fluid balance stabilisation. ⁴⁷ , ⁴⁸ , ⁴⁹ Landmark and sensitivity analyses confirmed early and sustained divergence in hospitalisation risk. Although concerns exist regarding SGLT2i safety in acutely ill patients—such as the risk of euglycaemic ketoacidosis or volume depletion—recent data indicate these events are infrequent with appropriate monitoring. ⁵⁰ Thus, deferring SGLT2i use after infection may forfeit an opportunity to prevent downstream hospitalisations and complications. Our findings support considering SGLT2i not only for long‐term cardiometabolic management but also as a recovery‐phase intervention.

Limitations include potential residual confounding inherent to retrospective observational designs, despite careful matching and adjustment. Data on medication adherence, dosing, and microbiologic profiles were unavailable. The US‐based sample may limit generalisability. We did not differentiate effects among individual SGLT2i agents. MACE analyses may have been underpowered due to low event counts. Lastly, unmeasured behaviours (e.g., health‐seeking patterns) may have influenced the observed infection‐related outcomes through residual confounding. In addition, due to the nature of the TriNetX platform, only summary‐level laboratory data (mean ± SD) were available, precluding assessment of distribution normality or use of more appropriate metrics such as medians or IQRs for non‐normally distributed variables like UPCR. Nonetheless, our results provide real‐world evidence that supports early SGLT2i initiation after UTI as a strategy to reduce mortality, kidney injury, and infection‐related complications in T2DM patients.

5. CONCLUSION

In this real‐world cohort study of patients with T2DM recovering from UTI, the use of SGLT2i within 3 months was associated with significantly lower risks of mortality, kidney complications, and hospitalisation compared with DPP‐4i. Despite historical concerns regarding infection risk, SGLT2i use did not increase adverse infectious outcomes. These findings support early post‐UTI initiation of SGLT2i as a safe and effective therapeutic option for patients with T2DM.

AUTHOR CONTRIBUTIONS

Design: Ming‐Hsien Tsai, Chien‐Lin Lu; Conduct/Data Collection: Kuo‐Cheng Lu, Ming‐Hsien Tsai; Analysis: Ming‐Hsien Tsai, Joshua Wang, Chien‐Lin Lu; Writing Manuscript: Ming‐Hsien Tsai, Chien‐Lin Lu (original draft); All authors (review and editing).

FUNDING INFORMATION

This study was supported by grants from Shin Kong Wu Ho‐Su Memorial Hospital (2024SKHADR020, Ming‐Hsien Tsai) and Fu Jen Catholic University Hospital (PL‐202408022‐V, Chien‐Lin Lu).

CONFLICT OF INTEREST STATEMENT

All the authors declare no competing interests.

PEER REVIEW

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/dom.70003.

ETHICS STATEMENT

The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Shin Kong Wu Ho‐Su Memorial Hospital (IRB number: 20240812R). The requirement for informed consent was waived due to the use of de‐identified retrospective data.

Supporting information

Supporting Information Materials S1.

Method S1. Codes for cohorts, outcomes and baseline covariates.

Method S2. The detail of propensity score matching in TriNetX platform.

Figure S1. Study design and timeline of follow‐up.

Figure S2. Follow‐up time distribution.

Figure S3. Kaplan–Meier curves of cardiovascular outcomes: (A) acute myocardial infarction, (B) heart failure, and (C) stroke.

Figure S4. Kaplan–Meier curves of kidney outcomes: (A) acute kidney injury, (B) dialysis dependence.

Figure S5. Kaplan–Meier curves of other outcomes.

Figure S6. Subgroup analysis of major adverse cardiovascular events.

Figure S7. Subgroup analysis of major adverse kidney events.

Figure S8. Subgroup analysis of all‐cause hospitalisation.

Table S1. Landmark analysis of outcomes at 6 and 12 months following UTI in patients with T2DM.

Table S2. Clinical outcomes analysis with new‐user design.

Table S3. Clinical outcomes with different propensity score matching models.

Table S4. Clinical outcomes analysis with different time frame.

Table S5. Baseline characteristics of study population before and after propensity score matching, with additional variables of hospitalisation at index UTI and urine protein.

Table S6. Clinical outcomes in patients with type 2 diabetes after urinary tract infection, with additional variables of hospitalisation before index day and urine protein level.

Wang H‐W, Tsai M‐H, Fang Y‐W, Lu K‐C, Wang J, Lu C‐L. Association between sodium‐glucose cotransporter 2 inhibitor use and clinical outcomes in patients with type 2 diabetes after urinary tract infection. Diabetes Obes Metab. 2025;27(11):6188‐6199. doi: 10.1111/dom.70003

DATA AVAILABILITY STATEMENT

The data utilised in this study were obtained from the TriNetX Global Health Research Network and are not publicly available due to licencing agreements and privacy regulations. TriNetX provides access to de‐identified, aggregate‐level data sourced from a global consortium of healthcare institutions. Researchers interested in accessing the data may submit a request via the TriNetX website (https://trinetx.com) or by contacting privacy@trinetx.com. Additionally, data may be made available from the corresponding author upon reasonable request.