Urogenital Infection Risk Associated With SGLT2 Inhibitors Versus DPP‐4 Inhibitors in Type 2 Diabetes and Chronic Kidney Disease: A Nationwide Target Trial Emulation

Abstract

Although sodium‐glucose cotransporter‐2 inhibitors (SGLT2i) are recommended for patients with type 2 diabetes mellitus (T2DM) and chronic kidney disease (CKD), concerns regarding urogenital infections persist. Therefore, we evaluated the risk of urinary tract infection (UTI) and genital tract infection (GTI) after SGLT2i initiation in this population. Using nationwide Korean claims data, we emulated a target trial within an active‐comparator, new‐user cohort design. Adults with T2DM and CKD who initiated SGLT2i or dipeptidyl peptidase‐4 inhibitors (DPP‐4i) between 2016 and 2023 were included. After inverse probability of treatment weighting, Cox models estimated hazard ratios (HRs) and 95% confidence intervals (CIs). Among 41,801 eligible initiators, 6385 initiated SGLT2i and 35,416 initiated DPP‐4i. Compared with DPP‐4i initiation, SGLT2i initiation was not associated with increased UTI risk (HR, 0.99; 95% CI, 0.83–1.17); instead, it was associated with a modestly increased GTI risk (HR, 1.31; 95% CI, 1.09–1.58). Severe UTI and GTI requiring hospitalization or emergency department visits did not differ between groups. Moreover, the findings were robust across per‐protocol analyses, Fine–Gray competing‐risks models, propensity score matching, and alternative outcome definitions. Subgroup and landmark analyses showed no consistent increase in UTI risk, whereas excess GTI risk clustered early and was most pronounced among women aged 65–79 years. Overall, SGLT2i initiation was not associated with increased UTI risk and was associated with only a modest, early increase in GTI risk, without an increase in severe urogenital infections. These findings support clinically indicated use of SGLT2i with targeted early counseling, particularly in older women.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

SGLT2 inhibitors have been established as cardiorenal‐protective therapies for patients with type 2 diabetes and chronic kidney disease. In broader populations with diabetes, they consistently increase the risk of genital tract infections, but their association with urinary tract infections remains inconsistent. However, evidence specific to patients with both type 2 diabetes and chronic kidney disease, including Asian populations, remains limited.

• WHAT QUESTION DID THIS STUDY ADDRESS?

Among adults with type 2 diabetes and chronic kidney disease, does initiating an SGLT2 inhibitor, compared with initiating a DPP‐4 inhibitor, increase the risk of urinary tract, genital tract, or severe urogenital infections?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This nationwide active‐comparator, new‐user cohort emulating a target trial showed that SGLT2 inhibitor initiation was not associated with increased urinary tract infection risk but was associated with a modest increase in genital tract infection risk. The incidence of severe urinary and genital tract infections requiring hospitalization or emergency department visits did not increase. These findings remained robust across per‐protocol analyses, Fine–Gray competing‐risk models, propensity score matching, and alternative outcome definitions. Subgroup and landmark analyses further indicated that the excess risk of genital tract infection was concentrated early and was most pronounced among women aged 65–79 years.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

These findings clarify the urogenital infection safety profile of SGLT2 inhibitors in patients with type 2 diabetes and chronic kidney disease. Clinically indicated use should not be deterred by concerns about urinary tract infections, whereas early counseling and symptom monitoring for genital tract infections may be most relevant for older women.

Sodium‐glucose cotransporter‐2 inhibitors (SGLT2i) are now recommended as first‐line pharmacological therapy for patients with type 2 diabetes mellitus (T2DM) and chronic kidney disease (CKD) in current clinical practice guidelines. ¹ SGLT2i inhibit sodium‐glucose cotransport in the proximal tubule, thereby inducing sustained glucosuria, a pharmacological effect that may alter the local urogenital environment. ² Because SGLT2i‐induced glucosuria can create a glucose‐enriched milieu in the genital area and may also modify the urinary microenvironment, SGLT2i therapy has raised differing concerns regarding genital tract infection (GTI) and urinary tract infection (UTI); these concerns are reflected in U.S. Food and Drug Administration safety communications regarding serious urogenital infections and related hospitalizations. ³ , ⁴ Similarly, the Korean Ministry of Food and Drug Safety issued a safety communication regarding the risk of severe genital tract infections for the entire SGLT2i class, and clinical practice guidelines consistently describe urogenital infection risk as a common feature across all SGLT2 inhibitors. ⁵

These urogenital complications are clinically important for patients with CKD for several reasons. UTI and CKD have a bidirectional relationship: CKD predisposes patients to frequent, severe infections, whereas recurrent UTI episodes can precipitate acute kidney injury and accelerate renal decline. ⁶ , ⁷ Conversely, GTI arises when SGLT2i‐induced glucosuria interacts with local mucosal vulnerability. ⁸ , ⁹ Notably, although SGLT2i‐induced glucosuria diminishes as the glomerular filtration rate declines, progressive renal impairment concurrently amplifies uremic vulnerability. This interplay can drive divergent risk patterns for GTI and UTI, requiring stage‐specific safety evaluations across the CKD spectrum.

Although several real‐world studies have examined the urogenital infectious risks associated with SGLT2i, evidence specific to patients with concomitant CKD remains limited, particularly in Asian populations with distinct clinical phenotypes. ¹⁰ , ¹¹ , ¹² Furthermore, the existing literature has seldom characterized these outcomes by clinical severity, leaving the risk profiles of UTI and GTI incompletely defined across progressive stages of renal impairment, as well as across sex and age subgroups. ¹³

To address these gaps, we conducted a nationwide, population‐based cohort study using Korean National Health Insurance claims data to assess the risk of UTI and GTI after SGLT2i initiation among patients with T2DM and CKD. We applied a target trial emulation (TTE) framework and used dipeptidyl peptidase‐4 inhibitors (DPP‐4i), which do not induce pharmacological glucosuria, as active comparators to reduce immortal‐time bias and confounding by indication in this observational setting. ¹⁴ , ¹⁵ We examined both overall and severe urogenital infectious outcomes to define the safety profile, with a primary focus on characterizing risk variations across assessable CKD stages and key demographic strata.

MATERIALS AND METHODS

Study design and data source

This nationwide retrospective cohort study was conducted using the TTE framework ¹⁴ to compare the risk of urogenital infections after initiation of SGLT2i vs. DPP‐4i among patients with T2DM and CKD. The specifications of the target trial are listed in Table S1 . This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guideline ¹⁶ (Table S2 ). This study was approved by the Institutional Review Board of Yonsei University (IRB No. 7001988‐202509‐HR‐2890‐01E), which waived the requirement for informed consent because only de‐identified secondary claims data were used.

Claims data from the Health Insurance Review and Assessment Service (HIRA) for 2015–2024 were used. HIRA covers reimbursement claims generated under South Korea's compulsory National Health Insurance system and captures nearly all outpatient and inpatient visits, emergency department (ED) visits, procedures, and prescription dispensing. ¹⁷ Because all demographic and clinical variables were derived from mandatory reimbursement claims and coded records, no analyzed variables had missing data; thus, no patients were excluded for missingness, and no imputation was performed. The overall study scheme is illustrated in Figure S1 . ¹⁸

Study population

Eligible patients were adults aged ≥ 30 years with T2DM and CKD who newly initiated either an SGLT2i or a DPP‐4i treatment between January 1, 2016, and December 31, 2023. To implement a new‐user design, patients with dispensing records for either drug class during the 1‐year baseline look‐back period were excluded. For patients meeting the initiation criteria for both classes on different dates, the earliest dispensing date was designated as the index date, and patients were assigned to the corresponding treatment strategy.

T2DM was defined as an International Classification of Diseases, 10th Revision (ICD‐10) code E11 recorded in ≥ 2 outpatient claims or ≥ 1 inpatient claim during the baseline period. ¹⁹ CKD (stages 2–4 or unspecified) was identified using the same operational algorithm, based on the ICD‐10 codes N18.2, N18.3, N18.4, and N18.9. ²⁰ , ²¹ , ²² Patients with type 1 diabetes mellitus, chronic dialysis (identified by special copayment reduction registration codes V001 and V003), or pregnancy during the 1‐year look‐back period were excluded. Additional exclusions included simultaneous initiation of both drug classes on the index date and non‐positive follow‐up. All codes are listed in Tables S3 – S5 . Given the nationwide scope of the database, no formal a priori sample size calculation was performed, and all eligible patients meeting the predefined criteria were included.

Treatment strategies, outcome definitions, and follow‐up

Patients were assigned to either the SGLT2i or DPP‐4i group according to their index medication, with DPP‐4i initiators serving as the active comparator. ¹⁵ The SGLT2i class included dapagliflozin, empagliflozin, ipragliflozin, ertugliflozin, and enavogliflozin, which were approved in Korea between 2013 and 2022. The DPP‐4i class comprised linagliptin, gemigliptin, sitagliptin, teneligliptin, evogliptin, vildagliptin, saxagliptin, alogliptin, and anagliptin. The primary outcomes were the first post‐index UTI and GTI, defined as the first claim after the index date with a relevant ICD‐10 diagnosis code in either the primary or secondary diagnosis position. ¹⁰ , ²³ UTI included pyelonephritis, cystitis, urethritis, and other UTIs. GTI included vulvovaginal candidiasis, balanitis, vaginitis/vulvitis, pelvic inflammatory disease, prostatitis, orchitis/epididymitis, and perineal cellulitis. Secondary outcomes were severe UTI and GTI, defined as infections requiring hospitalization or an ED visit with a corresponding diagnosis. The outcome codes are presented in Table S6 .

In the intention‐to‐treat (ITT) analysis, follow‐up extended from the index date to the earliest of outcome occurrence, in‐hospital death, the last observed healthcare claim, or 365 days post‐index. In the per‐protocol (PP) analysis, follow‐up was additionally censored at treatment discontinuation, switching, or addition of the comparator. Treatment discontinuation was defined as failure to receive a subsequent prescription within a 30‐day grace period after depletion of the supplied medication.

Covariates

During the 1‐year baseline look‐back period, baseline covariates, including demographics (sex and age group), insurance type, index year, disease characteristics (CKD stage and disease duration), comorbidities, urogenital procedures and catheterization, prior healthcare utilization, urogenital infection history, and baseline medications (Table 1 ), were comprehensively assessed. Individual baseline comorbidities (detailed in Table 1 ) were identified using the operational algorithm described above.

Table 1.

Baseline characteristics of study participants before and after inverse probability of treatment weighting

Variables	Before IPTWa		aSMDb	After IPTWa		aSMDb
	DPP‐4i	SGLT2i		DPP‐4i	SGLT2i
	N = 35,416	N = 6385		N = 20,904.0	N = 20,623.5
Sex
Female	13,233 (37.4)	2028 (31.8)	0.118	7650.3 (36.6)	8003.5 (38.8)	0.046
Male	22,183 (62.6)	4357 (68.2)		13,253.7 (63.4)	12,620.0 (61.2)
Age group, years
30–64	9297 (26.3)	2229 (34.9)	0.189	5804.3 (27.8)	6257.9 (30.3)	0.057
65–79	17,199 (48.5)	2917 (45.7)	0.058	10,036.8 (48.0)	10,087.0 (48.9)	0.018
≥ 80	8920 (25.2)	1239 (19.4)	0.139	5062.9 (24.2)	4278.6 (20.7)	0.083
Insurance type
Health insurance	30,726 (86.8)	5450 (85.3)	0.041	18,109.6 (86.6)	17,768.2 (86.2)	0.014
Medical Aid	3873 (10.9)	566 (8.9)	0.069	2232.3 (10.7)	2400.5 (11.6)	0.030
Veterans aid	817 (2.3)	369 (5.8)	0.177	562.1 (2.7)	454.8 (2.2)	0.031
Index year
2016–2017	12,104 (34.2)	397 (6.2)	0.743	6250.9 (29.9)	6431.5 (31.2)	0.028
2018–2019	9188 (25.9)	596 (9.3)	0.447	4891.5 (23.4)	4504.2 (21.8)	0.037
2020–2021	7595 (21.5)	1366 (21.4)	0.001	4481.3 (21.4)	4405.8 (21.4)	0.002
2022–2023	6529 (18.4)	4026 (63.1)	1.019	5280.2 (25.3)	5282.0 (25.6)	0.008
CKD stage
2	2436 (6.9)	747 (11.7)	0.167	1611.9 (7.7)	1773.3 (8.6)	0.032
3	11,711 (33.1)	2497 (39.1)	0.126	7066.4 (33.8)	6656.4 (32.3)	0.033
4	4653 (13.1)	391 (6.1)	0.240	2519.3 (12.1)	2208.0 (10.7)	0.042
Unclassified	16,616 (46.9)	2750 (43.1)	0.077	9706.4 (46.4)	9985.7 (48.4)	0.040
Comorbiditiesc
Diabetic complications
Retinopathy	9701 (27.4)	1432 (22.4)	0.115	5552.3 (26.6)	5584.6 (27.1)	0.012
Neuropathy	6484 (18.3)	879 (13.8)	0.124	3681.8 (17.6)	3766.1 (18.3)	0.017
Transplant/HIV status
Kidney transplant	898 (2.5)	129 (2.0)	0.035	516.1 (2.5)	528.0 (2.6)	0.006
Non‐kidney transplant	162 (0.5)	25 (0.4)	0.010	93.2 (0.4)	64.6 (0.3)	0.022
HIV infectiond	36 (0.1)	9 (0.1)	0.011	22.1 (0.1)	16.3 (0.1)	0.009
UG comorbidities
BPH	10,841 (30.6)	1907 (29.9)	0.016	6345.3 (30.4)	6084.3 (29.5)	0.019
Voiding dysfunction	968 (2.7)	148 (2.3)	0.027	560.9 (2.7)	640.6 (3.1)	0.025
Urolithiasis/obstruction	1509 (4.3)	222 (3.5)	0.041	862.2 (4.1)	883.2 (4.3)	0.008
UG cancer	1933 (5.5)	348 (5.5)	0.000	1129.5 (5.4)	976.6 (4.7)	0.030
STI or UG TB	116 (0.3)	27 (0.4)	0.016	71.5 (0.3)	79.5 (0.4)	0.007
Congenital UG abnormalitiesd	9 (0.0)	0 (0.0)	0.023	5.2 (0.0)	0.0 (0.0)	0.022
Other systemic comorbidities
Hypertension	31,705 (89.5)	5759 (90.2)	0.022	18,723.7 (89.6)	18,199.8 (88.2)	0.042
Dyslipidemia	30,391 (85.8)	5866 (91.9)	0.193	18,136.3 (86.8)	18,481.0 (89.6)	0.088
Cardiovascular disease	20,910 (59.0)	3956 (62.0)	0.060	12,421.6 (59.4)	11,820.0 (57.3)	0.043
Obesity	38 (0.1)	21 (0.3)	0.048	34.4 (0.2)	49.5 (0.2)	0.017
COPD	10,791 (30.5)	1814 (28.4)	0.045	6287.8 (30.1)	6263.3 (30.4)	0.006
Moderate liver disease	1093 (3.1)	166 (2.6)	0.029	627.7 (3.0)	489.5 (2.4)	0.039
Severe liver disease	397 (1.1)	44 (0.7)	0.046	219.1 (1.0)	179.6 (0.9)	0.018
Other solid cancers	3368 (9.5)	505 (7.9)	0.057	1943.3 (9.3)	2064.8 (10.0)	0.024
Hematological cancers	345 (1.0)	41 (0.6)	0.037	192.1 (0.9)	170.2 (0.8)	0.010
Other carbohydrate disordersd	0 (0.0)	1 (0.0)	0.018	0.0 (0.0)	1.0 (0.0)	0.010
Neuropsychiatric conditions
Dementia	4047 (11.4)	532 (8.3)	0.104	2278.1 (10.9)	1768.2 (8.6)	0.079
Mental disorder	9768 (27.6)	1324 (20.7)	0.160	5533.3 (26.5)	5151.0 (25.0)	0.034
Alcoholism	855 (2.4)	157 (2.5)	0.003	508.5 (2.4)	506.5 (2.5)	0.002
UG procedures/catheterization
Urological surgery	1506 (4.3)	143 (2.2)	0.114	821.1 (3.9)	708.2 (3.4)	0.026
Genital surgerye	118 (0.3)	30 (0.5)	0.022	73.3 (0.4)	58.3 (0.3)	0.012
Urinary catheterization	5823 (16.4)	654 (10.2)	0.183	3227.4 (15.4)	2502.9 (12.1)	0.096
Urinary incontinence procedure	16 (0.0)	3 (0.0)	0.001	9.3 (0.0)	7.0 (0.0)	0.005
Prior healthcare/UG infection hx
Prolonged hospitalization (≥ 30 d)f	2210 (6.2)	132 (2.1)	0.210	1167.7 (5.6)	684.3 (3.3)	0.110
Hx of UTI	15,676 (44.3)	3227 (50.5)	0.126	9431.9 (45.1)	8945.0 (43.4)	0.035
Hx of GTI	5477 (15.5)	1615 (25.3)	0.246	3564.9 (17.1)	3863.6 (18.7)	0.044
Major surgeryd	23 (0.1)	5 (0.1)	0.005	13.3 (0.1)	12.2 (0.1)	0.002
Disease duration, 2‐y look‐backg
T2DM
< 6 m	3924 (11.1)	1096 (17.2)	0.175	2501.0 (12.0)	2270.2 (11.0)	0.030
≥ 6 m	24,678 (69.7)	5102 (79.9)	0.237	14,902.2 (71.3)	14,582.8 (70.7)	0.013
Look‐back < 2 y	6814 (19.2)	187 (2.9)	0.538	3500.8 (16.7)	3770.5 (18.3)	0.040
CKD
< 6 m	8591 (24.3)	1654 (25.9)	0.038	5123.1 (24.5)	4886.8 (23.7)	0.019
≥ 6 m	20,011 (56.5)	4544 (71.2)	0.309	12,280.1 (58.7)	11966.1 (58.0)	0.015
Look‐back < 2 y	6814 (19.2)	187 (2.9)	0.538	3500.8 (16.7)	3770.5 (18.3)	0.040
Baseline antidiabetic medications
Number of oral agentsh
1	10,895 (30.8)	2505 (39.2)	0.178	6511.3 (31.1)	5852.4 (28.4)	0.061
2	16,666 (47.1)	2918 (45.7)	0.027	10,046.0 (48.1)	8841.5 (42.9)	0.104
≥ 3	7855 (22.2)	962 (15.1)	0.184	4346.7 (20.8)	5929.7 (28.8)	0.185
Specific drug classes
Metforminf	16,380 (46.3)	3439 (53.9)	0.153	9999.5 (47.8)	11,839.9 (57.4)	0.193
Sulfonylureas	12,503 (35.3)	1120 (17.5)	0.411	6821.5 (32.6)	7105.0 (34.5)	0.038
Thiazolidinediones	2669 (7.5)	272 (4.3)	0.139	1478.1 (7.1)	1775.8 (8.6)	0.057
Meglitinides	769 (2.2)	47 (0.7)	0.120	407.7 (2.0)	455.1 (2.2)	0.018
AGI	1260 (3.6)	95 (1.5)	0.132	677.8 (3.2)	605.6 (2.9)	0.018
Insulin	10,729 (30.3)	960 (15.0)	0.371	5837.4 (27.9)	5304.0 (25.7)	0.050
GLP‐1RA	91 (0.3)	38 (0.6)	0.052	66.0 (0.3)	113.1 (0.5)	0.036
Other baseline medications
ACEIs or ARBs	24,683 (69.7)	5015 (78.5)	0.203	14,842.8 (71.0)	14,259.7 (69.1)	0.041
Other antihypertensives	21,112 (59.6)	3709 (58.1)	0.031	12,383.4 (59.2)	11,326.6 (54.9)	0.087
Thiazides	6707 (18.9)	1152 (18.0)	0.023	3945.6 (18.9)	4061.0 (19.7)	0.021
Loop diureticsf	10,994 (31.0)	1539 (24.1)	0.156	6234.7 (29.8)	4978.6 (24.1)	0.128
Statins	24,795 (70.0)	5162 (80.8)	0.254	14,995.3 (71.7)	15,211.2 (73.8)	0.045
Antiplatelets	18,957 (53.5)	3188 (49.9)	0.072	11,072.5 (53.0)	10,747.9 (52.1)	0.017
Systemic glucocorticoids	4907 (13.9)	604 (9.5)	0.137	2755.1 (13.2)	2579.7 (12.5)	0.020
Other immunosuppressants	1497 (4.2)	248 (3.9)	0.017	873.6 (4.2)	848.8 (4.1)	0.003
Systemic antibioticsf	7219 (20.4)	681 (10.7)	0.271	3940.9 (18.9)	3075.3 (14.9)	0.105
Systemic antifungals	431 (1.2)	47 (0.7)	0.049	239.2 (1.1)	178.1 (0.9)	0.028
Antivirals	724 (2.0)	87 (1.4)	0.053	403.6 (1.9)	337.0 (1.6)	0.022
Antihistamines	4032 (11.4)	455 (7.1)	0.147	2236.1 (10.7)	1911.2 (9.3)	0.048
Antimuscarinics	4927 (13.9)	755 (11.8)	0.062	2833.1 (13.6)	2780.8 (13.5)	0.002
Antiepileptic drugs	1894 (5.3)	297 (4.7)	0.032	1091.3 (5.2)	912.4 (4.4)	0.037
Bisphosphonates	369 (1.0)	47 (0.7)	0.033	209.3 (1.0)	247.1 (1.2)	0.019
HRT/male hormones	174 (0.5)	33 (0.5)	0.004	103.2 (0.5)	140.2 (0.7)	0.024
Opioids	8090 (22.8)	1121 (17.6)	0.132	4593.9 (22.0)	4196.7 (20.3)	0.040

Bold values indicate potential imbalance (aSMD > 0.10). ACEI, angiotensin‐converting enzyme inhibitor; AGI, alpha‐glucosidase inhibitor; ARB, angiotensin II receptor blocker; aSMD, absolute standardized mean difference; BPH, benign prostatic hyperplasia; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; d, day; DPP‐4i, dipeptidyl peptidase‐4 inhibitor; GLP‐1RA, glucagon‐like peptide‐1 receptor agonist; GTI, genital tract infection; HIV, human immunodeficiency virus; HRT, hormone replacement therapy; Hx, history; IPTW, inverse probability of treatment weighting; m, month; SGLT2i, sodium‐glucose cotransporter‐2 inhibitor; STI, sexually transmitted infection; T2DM, type 2 diabetes mellitus; TB, tuberculosis; UG, urogenital; UTI, urinary tract infection; y, year.

^a Values are presented as N (%); after IPTW, weighted counts are shown.

^b aSMDs were calculated for covariates in the propensity score estimation, where aSMD > 0.10 indicates potential imbalance. For categorical covariates, aSMDs are shown for each category.

^c Systemic fungal infection was omitted because no cases were observed during look‐back.

^d Prespecified covariates with fewer than 10 patients in either group before weighting were excluded from the propensity score estimation and shown descriptively.

^e Fertility‐related or pregnancy maintenance‐related procedures were excluded.

^f Covariates with residual imbalance after IPTW (aSMD > 0.10) were additionally adjusted for in the IPTW‐weighted Cox proportional hazards model.

^g Duration of T2DM and CKD was classified using the earliest diagnosis within 2 years before the index date. Patients with less than 2 years of available look‐back were classified as insufficient look‐back period (Look‐back < 2 y).

^h Excluded from the propensity score estimation and IPTW‐weighted Cox proportional hazards models to avoid multicollinearity (VIF > 24).

The CKD stage was determined using the most recent code within the 1‐year look‐back period; if multiple codes occurred on the same date, the most severe stage was selected. T2DM and CKD durations were classified as < 6 months or ≥ 6 months using a 2‐year look‐back window, whereas 2016 initiators were categorized as insufficient look‐back. Prior urogenital infections were identified using all available data prior to the index date, and prolonged prior hospitalization was defined as any inpatient stay of ≥ 30 days.

Baseline medication use (Table S4 ) required an active prescription on the index date, defined as a dispensing within 365 days with days of supply covering the index date, or a new dispensing on that date. For injectables (GLP‐1 receptor agonists and insulin), use was defined as ≥ 3 packages dispensed within 90 days before the index date or a new dispensing on the index date. To ensure model stability, prespecified covariates with < 10 patients in either treatment group before weighting were excluded from the inverse probability of treatment weighting (IPTW) model but were retained in Table 1 for descriptive balance assessment.

Statistical analysis

Propensity scores were estimated using multivariable logistic regression that included all eligible baseline covariates. IPTW was applied to balance baseline characteristics between treatment groups. ²⁴ , ²⁵ Weights were calculated as w = 0.5/p for SGLT2i initiators and w = 0.5/(1 − p) for DPP‐4i initiators, where p denotes the estimated propensity score. Weights were rescaled by 0.5 to construct an equal‐allocation pseudo‐population, preserving the original sample size and stabilizing the model against extreme propensity scores. ²⁶ Covariate balance was assessed using absolute standardized mean differences (aSMD), with a threshold of 0.10. ²⁷

Weighted Cox proportional hazards models with robust sandwich variance estimators were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). ²⁸ To establish a doubly robust estimation framework, covariates with residual imbalance (aSMD > 0.10) after weighting, including baseline systemic antibiotic use, prior prolonged hospitalization, loop diuretic use, and metformin use, were additionally adjusted for in the Cox model. The proportional hazards assumption for the treatment effect was evaluated using Schoenfeld residual tests. ²⁹ Incidence rates (IRs) and 95% CIs were calculated per 100 person‐years using normal approximations to Poisson rates. Weighted Kaplan–Meier curves were generated to display the 1‐year cumulative incidence, and 1‐year risk differences were calculated.

Prespecified subgroup analyses were performed to evaluate potential effect modification by key characteristics, sequentially encompassing sex, age group, index year, CKD stage, comorbid urogenital cancer, history of prolonged hospitalization, prior urogenital infection, duration of T2DM and CKD, and baseline medication use. A post hoc subgroup analysis of the primary GTI outcome was stratified by age and sex. Propensity scores and IPTW weights were re‐estimated within each subgroup, with all covariates from the main analysis retained. If any residual imbalance remained after weighting, additional adjustment was conducted for those specific variables.

To assess the robustness of our findings, a PP analysis was performed to examine exposure misclassification, and the analyses were repeated using alternative outcome definitions from prior studies ¹⁰ , ²³ and more restrictive definitions requiring concomitant dispensing of systemic antibacterial or antifungal agents. As alternatives to IPTW, 1:1 and 1:2 propensity score‐matched analyses were conducted. ³⁰ Additionally, Fine–Gray competing‐risk models were used, treating in‐hospital death as a competing event, ³¹ and landmark analyses were performed at 30, 60, 90, and 180 days to examine time‐specific associations. ³²

In post hoc sensitivity analyses, the female GTI outcome was further restricted by requiring an obstetrics and gynecology (OB/GYN) claim accompanied by either (1) clotrimazole dispensing or (2) dispensing of systemic antibacterial or antifungal agents within 3 days of diagnosis. Additionally, to evaluate potential agent‐specific differences, we performed outcome analyses for the most frequently prescribed groups (dapagliflozin and the empagliflozin‐or‐ertugliflozin). All analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC, USA).

RESULTS

Baseline characteristics

The final cohort comprised 41,801 eligible new users: 6385 SGLT2i and 35,416 DPP‐4i initiators (Table 1 ). The most frequently prescribed agents in each group were as follows: SGLT2i dapagliflozin (60.6%) and empagliflozin‐or‐ertugliflozin (37.6%), and DPP‐4i linagliptin (42.1%), gemigliptin (20.0%), sitagliptin (14.3%), and teneligliptin (11.1%) (Table S7 ). Before weighting, SGLT2i initiators were more likely than DPP‐4i initiators to be male (68.2% vs. 62.6%), younger than 65 years (34.9% vs. 26.3%), and indexed in 2022–2023 (63.1% vs. 18.4%). They also had less advanced CKD, with a lower proportion of CKD stage 4 (6.1% vs. 13.1%), but more frequent prior UTI (50.5% vs. 44.3%) and prior GTI (25.3% vs. 15.5%). After IPTW, covariate balance improved substantially, with most weighted aSMDs below 0.10. Variables with residual imbalance (systemic antibiotic use, prior prolonged hospitalization, loop diuretic use, and metformin use) were further adjusted for in the final Cox model.

Urogenital infection outcomes

In the weighted ITT analysis, SGLT2i initiation was not associated with higher UTI risk than DPP‐4i initiation (IR, 12.77 vs. 13.52 per 100 person‐years; HR, 0.99; 95% CI, 0.83–1.17) (Table 2 ). Conversely, SGLT2i initiation was significantly associated with a higher risk of GTI (IR, 6.32 vs. 4.67 per 100 person‐years; HR, 1.31; 95% CI, 1.09–1.58). Severe urogenital infections requiring hospitalization or ED visits were rare (Table 2 ). Compared with DPP‐4i initiation, SGLT2i initiation was not associated with a significantly different risk of severe UTI (HR, 1.00; 95% CI, 0.66–1.53) or severe GTI (HR, 0.72; 95% CI, 0.39–1.33). Schoenfeld residual diagnostics suggested potential time‐varying hazards for UTI, severe UTI, and severe GTI, but not for GTI (Table S8 ). The IPTW‐weighted Kaplan–Meier curves were consistent with the Cox model results (Figure 1 ). The curves showed a higher cumulative incidence of GTI among SGLT2i initiators than among DPP‐4i initiators, whereas no clear separation was observed for UTI or severe urogenital infection outcomes.

Table 2.

Incidence and risk of urogenital infections with SGLT2 inhibitor vs. DPP‐4 inhibitor initiation: intention‐to‐treat analysis

Outcomes	Drug	No. of patientsa	Eventsa	Total person‐yearsa	IRb	aHRc
					(95% CI)	(95% CI)
Primary outcomes
UTI	SGLT2i	20,623.5	2362.6	18,504.7	12.77 (12.25–13.28)	0.99 (0.83–1.17)
	DPP‐4i	20,904.0	2456.8	18,175.7	13.52 (12.98–14.05)	Reference
GTI	SGLT2i	20,623.5	1211.4	19,162.7	6.32 (5.97–6.68)	1.31 (1.09–1.58)
	DPP‐4i	20,904.0	893.3	19,112.5	4.67 (4.37–4.98)	Reference
Secondary outcomes
Severe UTI	SGLT2i	20,623.5	702.0	19,457.6	3.61 (3.34–3.87)	1.00 (0.66–1.53)
	DPP‐4i	20,904.0	802.9	19,218.9	4.18 (3.89–4.47)	Reference
Severe GTI	SGLT2i	20,623.5	35.1	19,872.0	0.18 (0.12–0.24)	0.72 (0.39–1.33)
	DPP‐4i	20,904.0	50.0	19,609.2	0.25 (0.18–0.33)	Reference

Bold values indicate statistical significance. aHR, adjusted hazard ratio; CI, confidence interval; DPP‐4i, dipeptidyl peptidase‐4 inhibitor; GTI, genital tract infection; IR, incidence rate; SGLT2i, sodium‐glucose cotransporter‐2 inhibitor; UTI, urinary tract infection.

^a Weighted patient counts, events, and person‐years were derived using inverse probability of treatment weighting based on propensity scores.

^b IRs were calculated as weighted events divided by weighted person‐years, expressed per 100 person‐years, with 95% CIs derived using the normal approximation to Poisson rates.

^c Adjusted hazard ratios were estimated using Cox proportional hazards models with inverse probability of treatment weighting based on propensity scores. Variables with residual imbalance after weighting (absolute standardized mean difference > 0.1)—including systemic antibiotic use, prior prolonged hospitalization, loop diuretic use, and metformin use—were additionally adjusted.

Figure 1.

Cumulative incidence of urogenital infections after treatment initiation. (a) Urinary tract infection, (b) Genital tract infection, (c) Severe urinary tract infection, and (d) Severe genital tract infection. DPP‐4i, dipeptidyl peptidase‐4 inhibitor; SGLT2i, sodium‐glucose cotransporter‐2 inhibitor.

Subgroup analyses

Subgroup analyses were generally consistent with the primary findings (Figure 2 ). SGLT2i initiation was not associated with increased UTI risk in most prespecified subgroups, including those defined by prior urogenital infection history; however, UTI risk estimates were higher in more advanced CKD stages, particularly among patients with CKD stage 4. For GTI, increased risk estimates were observed in several clinically relevant subgroups, including patients with CKD stage 2, prior urogenital infection, and greater baseline diabetes‐treatment intensity, whereas the risk was attenuated in patients with CKD stages 3–4. Similarly, among patients with unclassified CKD stage (46.3% of the cohort) and across index year strata (2016–2017, 2018–2019, 2020–2021, and 2022–2023), risk patterns remained consistent with the overall findings.

Figure 2.

Subgroup analyses of urogenital infection risk associated with SGLT2i initiation compared with DPP‐4i initiation. (a) Urinary tract infection and (b) Genital tract infection. Propensity scores and IPTW weights were re‐estimated within each subgroup, with all covariates from the main analysis retained. Bold indicates statistical significance at P < 0.05; hazard ratios and 95% CIs are rounded to two decimal places. aHR, adjusted hazard ratio; CI, confidence interval; CKD, chronic kidney disease; DPP‐4i, dipeptidyl peptidase‐4 inhibitor; GTI, genital tract infection; hx, history; IPTW, inverse probability of treatment weighting; m, month; SGLT2i, sodium‐glucose cotransporter‐2 inhibitor; T2DM, type 2 diabetes mellitus; UTI, urinary tract infection.

Because risk elevations were particularly pronounced among female patients and older adults, a post hoc age‐by‐sex analysis was performed (Figure 3 ). In this analysis, increased risk was most evident among women aged 65–79 years (HR, 2.00; 95% CI, 1.44–2.76). Severe urogenital infection outcomes showed no consistent subgroup‐specific increase; CIs were wide across strata because of low absolute event numbers.

Figure 3.

Age‐ and sex‐stratified risk of genital tract infection associated with SGLT2i initiation compared with DPP‐4i initiation. Propensity scores and IPTW weights were re‐estimated within each subgroup, with all covariates from the main analysis retained. Bold indicates statistical significance at P < 0.05; hazard ratios and 95% CIs are rounded to two decimal places. aHR, adjusted hazard ratio; CI, confidence interval.

Sensitivity analyses

Sensitivity analyses largely supported the primary findings (Table S9 ). In the PP analysis, SGLT2i initiation was not associated with increased UTI risk (HR, 0.86; 95% CI, 0.72–1.02), whereas GTI risk remained increased (HR, 1.40; 95% CI, 1.13–1.73). Similar results were observed in the Fine–Gray competing‐risk analysis for both UTI (subdistribution HR, 0.93; 95% CI, 0.82–1.07) and GTI (subdistribution HR, 1.38; 95% CI, 1.16–1.64). In landmark analyses evaluating time‐specific associations, no increased UTI risk was observed at any landmark time point. For GTI, the association was significant at 30 days but attenuated after 60 and 90 days. The primary findings remained robust across both 1:1 and 1:2 propensity score matching (Table S10 ) and multiple alternative or restrictive outcome‐definition sensitivity analyses, including post hoc female‐specific criteria. The analysis results for the predominant agent groups were also consistent with the primary findings (Table S11 ).

DISCUSSION

In this nationwide, active‐comparator, new‐user cohort study emulating a target trial, initiation of an SGLT2i, compared with a DPP‐4i, among patients with T2DM and CKD was associated with an increased risk of GTI (HR, 1.31; 95% CI, 1.09–1.58), but not UTI (HR, 0.99; 95% CI, 0.83–1.17). These findings remained consistent across a range of sensitivity analyses, confirming the robustness of the results. This study provides two distinct contributions. Methodologically, it establishes a TTE framework within a nationwide CKD population to systematically mitigate immortal‐time and indication biases. Clinically, beyond confirming class‐level safety patterns, it shows that excess GTI risk clusters early within 30 days and predominantly affects older women aged 65–79 years, while the incidence of severe infections requiring inpatient or emergency department care is not elevated.

Consistent with our results, prior studies have shown that SGLT2i use is associated with an increased risk of GTI but not UTI. A meta‐analysis of 77 randomized controlled trials reported a three‐fold increase in GTIs with SGLT2i use, with no increase in UTI. ³³ Similarly, a Danish TTE reported 1‐year risk ratios of 0.98 for UTI and 2.95 for GTI (SGLT2i vs. GLP‐1 receptor agonists), ¹⁰ and a large U.S. claims‐based study found an approximately 2.7‐fold higher GTI risk with SGLT2i than with DPP‐4i use. ³⁴ A Korean nationwide cohort study, using the same DPP‐4i comparator, also reported an elevated GTI risk with SGLT2i (HR 2.38). ¹²

The dissociation between elevated GTI risk and null UTI risk is mechanistically coherent. SGLT2i lower the renal threshold for glucose and induce persistent glucosuria, creating a carbohydrate‐rich genital microenvironment that promotes Candida adherence and overgrowth, the most common cause of vulvovaginal candidiasis in women and balanitis in men. ² , ³⁵ Conversely, ascending UTI depends on uropathogen virulence, host urodynamics, and mucosal defenses, for which glucosuria is a weaker and less consistent driver. ³⁶ Although patients with CKD are recognized to be at increased risk of UTI, ⁶ this study confirmed that SGLT2i initiation conferred no additional UTI risk, even in this susceptible population.

Nevertheless, the magnitude of the GTI risk in our study (HR, 1.31) was lower than that reported in previous studies. This discrepancy may reflect differences in the study cohorts: prior studies were generally not restricted to patients with CKD, whereas our analysis focused exclusively on this population. Because the glucosuric effect of SGLT2i is known to diminish as renal function declines, ³⁷ the modest risk increase observed here is biologically plausible. The same mechanism may underlie our subgroup findings: CKD stage 2, in which renal function (and thus glucosuria) was relatively preserved, showed a distinctly higher risk (HR, 1.82), whereas the association was attenuated in stages 3 and 4. Unlike GTI, UTI risk increased with worsening CKD stage (UTI: HR, 0.61 in stage 2 and 1.72 in stage 4), a pattern that also suggests a mechanistic distinction between the two outcomes: GTI appears to arise from glucosuria, whereas UTI arises from host susceptibility. Nevertheless, given the limited number of patients within the CKD stage strata and the possible differences in comorbidities and medication use across stages, these findings should be interpreted with caution. Consequently, this attenuated GTI risk profile offers clinical reassurance and provides additional evidence supporting clinically indicated use of SGLT2i in this specific population.

When the landmark was set to 60 days or more after the index date in the sensitivity analysis, the risk of GTI no longer differed significantly between the two cohorts. This suggests that the excess GTI risk is concentrated in the early period after SGLT2i initiation, consistent with the findings of previous reports. A self‐controlled case‐series study found that GTI risk peaked 15–28 days after initiation, ²³ and a cohort study similarly reported that most first events occurred within the initial months. ³⁴ Nonetheless, these landmark results warrant cautious interpretation for the following reasons. Excluding patients who experienced early events not only reduces the sample size and statistical power but also conditions the analysis on event‐free survival. This conditioning can differentially deplete more susceptible patients from the SGLT2i arm, introducing the built‐in selection bias of period‐specific hazard ratios. ³⁸ In the landmark analysis for severe GTI, the low HR observed after 30 days can be interpreted in a similar context. The number of severe GTI events was relatively small (DPP‐4i: 50.0, SGLT2i: 35.1), and a higher proportion of these events occurred between the index date and the 30‐day landmark (DPP‐4i: 3.6 (7.1%), SGLT2i: 12.4 (35.3%)). Therefore, landmark analysis of severe GTI may be considered further evidence of the early clustering of SGLT2i‐associated GTI.

The subgroup findings were biologically interpretable. The higher risk associated with greater baseline diabetes‐treatment intensity likely reflects longer‐standing, less well‐controlled diabetes, with attendant immune dysfunction and elevated baseline susceptibility. ³⁹ Lastly, the post hoc age‐by‐sex analysis localized the strongest signal to women aged 65–79 years (HR, 2.00). In postmenopausal women, estrogen deficiency produces the genitourinary syndrome of menopause, characterized by vaginal atrophy, depletion of protective lactobacilli, and increased vaginal pH, thereby increasing susceptibility to GTI; combined with SGLT2i‐driven glucosuria, this offers a plausible explanation for the concentration of risk among older women. ⁴⁰

Our study has several strengths. First, the nationwide claims database captured virtually all reimbursed care, thereby minimizing selection bias and loss to follow‐up. Second, the explicit target‐trial specification ¹⁴ with an active‐comparator, new‐user design ¹⁵ reduced confounding by indication and mitigated immortal‐time and prevalent‐user bias. Third, we addressed residual covariate imbalance using a doubly robust approach that combined IPTW with multivariable outcome adjustment, and we confirmed the robustness of our findings across a range of sensitivity analyses. Finally, a key clinical merit of this study is its focused evaluation of SGLT2i‐associated UTI and GTI risks in patients with T2DM and CKD. This population stands to benefit substantially from SGLT2i therapy, yet remains susceptible to urogenital infection.

Despite these strengths, some limitations should be considered. First, dispensing records do not confirm actual medication ingestion. Second, outcomes and covariates were operationally defined using diagnosis and procedure codes, which may introduce inaccuracies in outcome identification. Third, regulatory warnings and clinical guidelines in Korea could theoretically introduce both channeling bias in prescribing and detection bias in outcome ascertainment. Regarding prescribing behavior, SGLT2i initiators exhibited more frequent baseline prior UTI (50.5% vs. 44.3%) and prior GTI (25.3% vs. 15.5%), indicating that treatment decisions were not predominantly driven by infection avoidance. Although these histories were rigorously balanced via IPTW and verified in the stratified analyses, channeling from unmeasured clinical judgments could not be fully excluded. Regarding detection, heightened awareness may prompt SGLT2i users to seek care more readily, potentially compounding the aforementioned outcome misclassifications. To address this, we evaluated restrictive outcome definitions (requiring concurrent antimicrobial dispensing, OB/GYN specialty claims, and severe events [hospitalizations or ED visits]) to mitigate the outcome misclassification inherent to claims‐based definitions. The findings remained consistent, and the lack of an increase in severe events suggested that the observed risk was not merely driven by heightened surveillance. Fourth, laboratory values (estimated glomerular filtration rate and glycated hemoglobin), body mass index, and behavioral factors were unavailable, leaving the potential for residual confounding. Additionally, the CKD stage was code‐based and could not be resolved for 46.3% of the cohort. Estimates in this stratum were consistent with the overall findings, indicating that the primary results were not driven by its inclusion, but stage misclassification nonetheless limits the interpretation of the stage‐specific contrasts. Fifth, ingredient‐specific analyses of SGLT2 inhibitors were limited to dapagliflozin because of HIRA data governance policies and the low prescription volume of certain agents. Nonetheless, considering the distribution of individual SGLT2 inhibitors in the previous literature and the approved label indications for ertugliflozin, ⁴¹ the findings of the empagliflozin‐or‐ertugliflozin group are expected to predominantly reflect empagliflozin characteristics. Sixth, SGLT2i initiation was concentrated in the later study years, coinciding with several changes in Korean policy and market environment. Although the calendar year was adjusted via IPTW (achieving weighted aSMDs < 0.05) and stratified analyses showed stable HRs across index‐year intervals, residual calendar time confounding cannot be completely ruled out. Seventh, Schoenfeld residual diagnostics indicated potential non‐proportional hazards for UTI, severe UTI, and severe GTI, though not for the primary GTI outcome. However, for UTI or severe UTI, given the pattern of cumulative incidence curves, the violation of the proportional hazards assumption is presumed not to have affected the null association with the outcome. Regarding severe GTI, as previously mentioned, non‐proportionality is thought to be associated with an early concentration of events. Lastly, given the renal function‐based dose adjustment profiles of dapagliflozin and empagliflozin, ¹ , ⁴² an outcome assessment according to dosage was not conducted.

In conclusion, this nationwide TTE of patients with T2DM and CKD demonstrated that SGLT2i use was not associated with an increased risk of UTI, even in a population inherently susceptible to these infections. Although the risk of GTI was elevated, the magnitude of the increase was modest. Given the well‐established cardiorenal benefits of SGLT2i in this population, these real‐world findings are reassuring and support clinically indicated SGLT2i use, while underscoring the value of brief, targeted counseling on early GTI for higher‐risk patients, such as older women.

FUNDING

This research was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIT) (No. RS‐2026‐25481749), as well as by the ANCHOR program through the Jeollanamdo ANCHOR Center, funded by the Ministry of Education (MOE) and the Jeonnam‐Gwangju Special Metropolitan City, Republic of Korea (2026‐ANCHOR‐14‐001). The funders had no role in the study design, data collection, data analysis, interpretation, manuscript preparation, or the decision to submit the manuscript.

CONFLICTS OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

Y.‐M.A., J.N., Y.M.Y., and S.K. wrote the manuscript; Y.‐M.A., J.N., J.W.Y., Y.M.Y., and S.K. designed the research; Y.‐M.A., J.N., J.W.Y., Y.M.Y., and S.K. performed the research; Y.‐M.A., Y.M.Y., and S.K. analyzed the data.

Supporting information

Data S1.