Abstract
Background
To evaluate the association between Glucagon-like peptide-1 receptor agonists (GLP-1RA) use and the risk of acute diabetes complications among adults with type 1 diabetes (T1D) who were eligible for anti-obesity medication (AOM) treatment.
Methods
We employed a target trial emulation using EHR data from the OneFlorida+ network (2014–2024) to investigate the association between GLP-1RA initiation and acute diabetes complications among adults with T1D. Eligible participants were adults with a diagnosis of T1D and who met clinical criteria for AOM treatment. GLP-1RA initiators were 1:1 matched to non-initiators using time-conditional propensity scores. The primary outcome was the occurrence of diabetic ketoacidosis (DKA). Secondary outcomes included severe hypoglycemia, all-cause hospitalizations, and emergency department (ED) visits. Cox proportional hazards models were utilized to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). We applied a causal learning approach to explore heterogeneous treatment effects.
Findings
The matched cohort included 651 GLP-1RA users and 651 non-users. For GLP-1RA users and non-users, the incidence rates were 13.5 vs. 21.8 per 1,000 person-years for DKA. Compared to non-users, GLP-1RA use was not significantly associated with incidence of DKA (HR 0.62 [95%CI 0.33–1.17]) or severe hypoglycemia (HR 0.52 [95%CI 0.17–1.55]); notably, GLP-1RA use was significantly associated with fewer hospitalizations (HR 0.74 [95%CI 0.62–0.90]) and ED visits (HR 0.73 [95%CI 0.57–0.92]).
Interpretation
Among adults with T1D and obesity, GLP-1RA use was not associated with an increased risk of DKA or severe hypoglycemia but was linked to fewer ED visits and hospitalizations.
Funding
The study was supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIH/NIDDK) R01DK133465.
Introduction
Type 1 diabetes (T1D) represents approximately 5% to 10% of all diabetes cases in the United States (U.S.).1 Individuals with T1D continue to face challenges avoiding acute complications such as severe hypoglycemia2, and diabetic ketoacidosis (DKA).3 Severe hypoglycemia, characterized by dangerously low blood glucose levels, can lead to confusion, seizures, loss of consciousness, or even death if not promptly treated.2 DKA, classically resulting from insulin deficiency, leads to the accumulation of ketones in the blood, causing metabolic acidosis, dehydration, and, if untreated, can progress to severe pain, coma, and/or death.3 About 4.2% of individuals with T1D experience at least one hypoglycemia-related hospitalization annually,4 and DKA accounts for 61.6 cases per 10,000 diabetes-related hospitalizations in the U.S.5 Although T1D is traditionally considered a disease of lean people, overweight and obesity are becoming increasingly more common in individuals with T1D.6,7
Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) are a well-established class of antihyperglycemic agents approved for T2D. They enhance glucose-dependent insulin secretion, suppress glucagon, and delay gastric emptying, which collectively improve glycemic regulation.8,9 Although GLP-1 RAs are associated with favorable outcomes in T2D and generally well tolerated aside from gastrointestinal adverse events, their use in T1D remains off-label and largely under-investigated. Still, they are increasingly being explored as adjunctive therapy in T1D to reduce insulin requirements and glycemic variability.10,11 However, their potential effects on hypoglycemia and DKA risk remain unclear.
GLP-1 RAs are thought to carry a low intrinsic risk of hypoglycemia due to their glucose-dependent action.8,9 However, hypoglycemia may still occur, especially when combined with insulin.12 Emerging evidence also indicates that some GLP-1 RAs may increase hypoglycemia risk even when not used with other glucose-lowering therapies.13,14 A network meta-analysis of 34 T2D trials (excluding insulin and sulfonylureas) found increased hypoglycemia risk with all GLP-1 RAs except albiglutide, with odds ratios ranging from 1.59 to 3.74.13 Real-world post-marketing data from the FDA Adverse Event Reporting System (FAERS) also showed hypoglycemia signals across six GLP-1 RAs (excluding albiglutide due to a paucity of data), including severe events. Most reports involved T2D, but some occurred in T1D patients as well.14 In parallel, concerns have emerged regarding the potential association between GLP-1 RAs and DKA. In a FAERS-based pharmacovigilance study, most DKA reports linked to GLP-1 RAs involved individuals with T2D. However, the signal disappeared when GLP-1 RAs were used with insulin and was stronger when insulin was absent, indicating that reduced or withheld insulin may contribute to the observed risk.15 Additionally, case reports have documented euglycemic DKA in individuals with T1D or latent autoimmune diabetes in adults following GLP-1 RA use, often triggered by gastrointestinal side effects (nausea, vomiting, and/or diarrhea) and reduced insulin intake; all collectively highlighting potential safety concerns when insulin is withheld or interrupted in patients with T1D.16,17
Given the expanding off-label use of GLP-1 RAs in T1D and the limited availability of real-world evidence, further evaluation on safety and potential benefit is warranted. This study aimed to examine the association between GLP-1 RA use and the risk of DKA among individuals with T1D using real-world data from OneFlorida+ network. To explore heterogeneous treatment effect (HTE), we incorporated a causal learning approach to further identify subgroups of patients who may derive the maximized benefit or experience potential harm from GLP-1RAs, and to uncover key moderators.
Research Design and Methods
Data Source
Data was from the OneFlorida+ clinical research network, which is part of the National Patient-Centered Clinical Research Network (PCORnet). The dataset spans from 2014 through 2024 and includes comprehensive, longitudinal electronic health records (EHRs) for approximately 21 million individuals across Florida, Georgia, and Alabama. OneFlorida+ integrates data from 14 partnering healthcare systems and is formatted according to the PCORnet Common Data Model (CDM).18 The dataset, designated as a HIPAA-limited set, includes key temporal and geographic identifiers (e.g., dates and ZIP codes) and provides robust clinical and demographic information, including diagnoses, procedures, medication use, laboratory results, and vital signs.
Study Design and Study Population
This study employed a retrospective cohort design with a new-user approach. An intention-to-treat analysis was conducted to examine the relationship between GLP-1RAs and the incidence of acute diabetes complications and the occurrence of hospitalization or ED visits in real world T1D patients who were eligible for anti-obesity medications (AOMs). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines (eTable 1 in Supplement) and was approved by the University of Florida Institutional Review Board (# IRB202300903).
The study cohort comprised T1D adults meeting the eligible criteria for anti-obesity medication (AOM) treatment. (1) According to 2013 AHA/ACC/TOS guideline, which is applied to now, AOM eligibility was defined as the presence of a documented diagnosis of obesity (BMI ≥ 30 kg/m2) or a BMI in the range of 27.0 to 29.9 kg/m2 accompanied by at least one weight-related comorbidity.19 (2) Because of the high prevalence of T2D in the adult population, adult-onset T1D is misclassified as T2D in at least one in three cases in clinical practice.20 Simple ICD codes may not provide accurate identification of T1D in real-world data. Therefore, we adopted a validated EHR-based computable phenotype developed through network-wide chart reviews, which achieved >90% sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) in classifying T1D.21,22 To further reduce bias, we additionally restricted the cohort to individuals with documented insulin use.
We then identified GLP-1RA users and non-users. The index date for GLP-1RA users was defined as the date of the first GLP-1RA prescription. The comparator group, defined as patients who did not initiate GLP-1RA therapy during the study period, was constructed by aligning non-users to GLP-1RA users within ±1 week window of the initiation date, to ensure temporal comparability of treatment assignment. Individuals were excluded if they were under 18 years of age at the index date, had not been identified as T1D before the index date, or lacked any clinical encounters in the 12 months preceding the index date. Patients’ follow-up began at the index date and continued until the earliest occurrence of one of the following censoring events: the outcome of interest, death, loss to follow-up (the date of the last recorded clinical encounter), or the end of the study period (January 31, 2024).
Study Outcomes
The primary outcome was hospitalization or ED visit for DKA. The secondary outcomes included severe hypoglycemia (i.e., hospitalization or ED visit for hypoglycemia), any hospitalization, and ED visit. DKA and severe hypoglycemia were identified using ICD-9 and ICD-10 diagnosis codes, based on previously validated computable phenotype definitions23–25 (eTable 2 in Supplement).
Exposure of interest
The exposure of interest was the new use of GLP-1RA, i.e., semaglutide, tirzepatide, and liraglutide, identified using RxNorm codes and National Drug Codes (NDCs) (eTable 3 in Supplement). The comparator group consisted of non-users of GLP-1RAs.
Covariates
Baseline covariates encompassed a range of demographic, clinical, and pharmacological characteristics, as detailed in Table 1. Demographic variables included age, sex, race/ethnicity, smoking status, and insurance type. Clinical covariates comprised comorbid conditions (e.g., cardiovascular disease, biliary disorders and neuropathy), clinical observations (e.g., BMI, blood pressure), laboratory values (e.g., hemoglobin A1c [HbA1c], estimated glomerular filtration rate [eGFR], and lipid profile), and medication (e.g., aspirin, opioids, and statins).
Table 1.
Baseline characteristics of GLP-1RA Users vs GLP-1RA Non-Users cohort after 1:1 propensity score matching
| Characteristics | GLP-1RA Users, No.(%) (n=651) | GLP-1RA Non-Users, No.(%) (n=651) | SMD |
|---|---|---|---|
| Age, yrs, mean(std) | 44.57 (14.64) | 45.01 (17.44) | −0.0274 |
| Race/Ethnicity | |||
| Hispanics | 136 (20.89)% | 122 (18.74)% | 0.0539 |
| NHB | 126 (19.35)% | 126 (19.35)% | 0.0000 |
| NHW | 306 (47.0)% | 305 (46.85)% | 0.0031 |
| Others | 83 (12.75)% | 100 (15.06)% | 0.0384 |
| Sex | |||
| Female | 413 (63.44)% | 415 (63.75)% | −0.0064 |
| Male | 238 (36.56)% | 236 (36.25)% | 0.0064 |
| Vital information | |||
| BMI | 34.11 (6.21) | 34.02 (7.28) | 0.0120 |
| Systolic blood pressure | 129.83 (15.37) | 129.75 (16.13) | 0.0051 |
| Diastolic blood pressure | 77.61 (9.88) | 77.7 (10.07) | −0.0088 |
| Insurance Coverage | |||
| Medicaid | 35 (5.38)% | 48 (7.37)% | −0.0817 |
| Medicare | 67 (10.29)% | 72 (11.06)% | −0.0249 |
| No Payment | 15 (2.3)% | 14 (2.15)% | 0.0104 |
| Others | 76 (11.67)% | 91 (13.98)% | −0.0689 |
| Private | 205 (31.49)% | 191 (29.34)% | 0.0467 |
| Unknown | 253 (38.86)% | 235 (36.1)% | 0.0571 |
| Smoking Status | |||
| Current Smoker | n<11 | n<11 | −0.1300 |
| Former Smoker | n<11 | n<11 | 0.0554 |
| Never Smoker | 53 (8.14)% | 40 (6.14)% | 0.0775 |
| Unknown | 596 (91.55)% | 603 (92.63)% | −0.0398 |
| Lab | |||
| HbA1c | 8.28 (1.75) | 8.33 (2.16) | −0.0278 |
| HDL | 51.92 (11.97) | 52.45 (11.31) | −0.0456 |
| LDL | 88.09 (12.2) | 89.43 (13.06) | −0.1057 |
| Triglycerides | 138.78 (85.73) | 137.16 (83.54) | 0.0192 |
| Total cholesterol | 171.19 (33.89) | 172.1 (33.56) | −0.0269 |
| eGFR | 80.83 (23.14) | 80.85 (24.26) | −0.0009 |
| Medications | |||
| Insulin | 651 (100.0)% | 651 (100.0)% | |
| ACEIs | 163 (25.04)% | 160 (24.58)% | 0.0108 |
| Beta-blocker | n<11 | n<11 | 0.0116 |
| CCB | n<11 | n<11 | |
| Diuretic | 18 (2.76)% | 20 (3.07)% | −0.0098 |
| Angiotensin-receptor blocker | 118 (18.13)% | 113 (17.36)% | −0.0126 |
| Statin | 271 (41.63)% | 271 (41.63)% | −0.0319 |
| Lipid-lowering non-statin | 44 (6.76)% | 39 (5.99)% | 0.0314 |
| NSAIDS | 104 (15.98)% | 111 (17.05)% | −0.0289 |
| Warfarin | n<11 | n<11 | 0.0000 |
| Direct oral anticoagulant | 19 (2.92)% | 26 (3.99)% | −0.0588 |
| Aspirin | 83 (12.75)% | 84 (12.9)% | −0.0046 |
| Non-aspirin antiplatelet agents | 25 (3.84)% | 21 (3.23)% | 0.0333 |
| Proton pump inhibitor | 123 (18.89)% | 113 (17.36)% | 0.0399 |
| Antidepressant | 77 (11.83)% | 76 (11.67)% | 0.0048 |
| Antipsychotics | 43 (6.61)% | 43 (6.61)% | 0.0000 |
| Antiparkinson agent | 53 (8.14)% | 53 (8.14)% | 0.0000 |
| Benzodiazepines | 88 (13.52)% | 86 (13.21)% | 0.0090 |
| HRT | 20 (3.07)% | 17 (2.61)% | 0.0277 |
| Anti-dementia medication | n<11 | n<11 | 0.0210 |
| Oral steroids | 165 (25.35)% | 157 (24.12)% | 0.0285 |
| Opioid | 147 (22.58)% | 147 (22.58)% | 0.0000 |
| TNF inhibitor | n<11 | n<11 | 0.0000 |
| Immunosuppressants | 15 (2.3)% | 7 (1.08)% | 0.0954 |
| Orlistat | n<11 | n<11 | |
| Phentermine topiramate | n<11 | n<11 | 0.0320 |
| Bupropion/Naltrexone | n<11 | n<11 | −0.0248 |
| Lorcaserin | n<11 | n<11 | |
| Weight-related comorbidities | |||
| Cancer | 220 (33.79)% | 205 (31.49)% | 0.0491 |
| Acute/Chronic Pancreatitis | n<11 | n<11 | 0.0321 |
| Adenocarcinoma of the Esophagus | n<11 | n<11 | |
| Anorexia | n<11 | n<11 | 0.0000 |
| Atrial fibrillation | 17 (2.61)% | 24 (3.69)% | −0.0616 |
| Breast cancers | n<11 | 12 (1.84)% | −0.0238 |
| Chronic kidney disease | 79 (12.14)% | 66 (10.14)% | 0.0635 |
| Colorectal cancers | n<11 | n<11 | −0.0185 |
| Coronary Artery Disease | 78 (11.98)% | 87 (13.36)% | −0.0415 |
| Cushing Syndrome | n<11 | n<11 | −0.0210 |
| Depression | 112 (17.2)% | 107 (16.44)% | 0.0205 |
| Dyspepsia | n<11 | n<11 | −0.0298 |
| Dyslipidemia | 401 (61.6)% | 416 (63.9)% | −0.0476 |
| Feeding Difficulties | 59 (9.06)% | 64 (9.83)% | −0.0262 |
| Gallbladder Disease | n<11 | n<11 | −0.0432 |
| GERD | 117 (17.97)% | 125 (19.2)% | −0.0316 |
| Heart Failure | 29 (4.45)% | 42 (6.45)% | −0.0880 |
| Hypertension | 378 (58.06)% | 377 (57.91)% | 0.0031 |
| Inflammatory Bowel Disease | n<11 | n<11 | 0.0554 |
| Kidney cancer | n<11 | n<11 | 0.0320 |
| Liver Cancer | n<11 | n<11 | 0.0320 |
| Meningioma | n<11 | n<11 | 0.0000 |
| Metabolic Syndrome | 29 (4.45)% | 16 (2.46)% | 0.1094 |
| Multiple myeloma | 192 (29.49)% | 182 (27.96)% | 0.0339 |
| NAFLD/NASH | 49 (7.53)% | 49 (7.53)% | 0.0000 |
| Ovaries cancer | n<11 | n<11 | |
| Pancreatic cancer | n<11 | n<11 | 0.0000 |
| Prader-Willi Syndrome | n<11 | n<11 | |
| Pulmonary Embolism | n<11 | n<11 | 0.0000 |
| Rheumatoid Arthritis/Osteoarthritis (RA/OA) | 126 (19.35)% | 145 (22.27)% | −0.0719 |
| Sleep apnea | 82 (12.6)% | 93 (14.29)% | −0.0495 |
| Stroke/TIA | 13 (2.0)% | 20 (3.07)% | −0.0684 |
| Thyroid cancer | n<11 | n<11 | 0.0210 |
| Upper stomach cancer | n<11 | n<11 | |
| Uterus cancer | n<11 | n<11 | 0.0554 |
| Venous Thromboembolism | 17 (2.61)% | 13 (2.0)% | 0.0409 |
| Vitamin D deficiency | 209 (32.1)% | 195 (29.95)% | 0.0465 |
| At least one weight related comorbidities | 629 (96.62)% | 633 (97.24)% | −0.0356 |
| Other Comorbidities | |||
| AD | n<11 | n<11 | 0.0000 |
| ADRD | n<11 | n<11 | −0.0630 |
| Acute kidney injury | 72 (11.06)% | 70 (10.75)% | 0.0098 |
| Alcohol use disorder | n<11 | n<11 | 0.0248 |
| Anemia | 125 (19.2)% | 140 (21.51)% | −0.0572 |
| Anxiety | 158 (24.27)% | 161 (24.73)% | −0.0107 |
| Asthma | 53 (8.14)% | 61 (9.37)% | −0.0435 |
| B12_deficiency | n<11 | n<11 | −0.0210 |
| Benign prostatic hyperplasia | n<11 | n<11 | 0.0000 |
| Biliary disease | 17 (2.61)% | 22 (3.38)% | −0.0450 |
| Bipolar | 15 (2.3)% | 16 (2.46)% | −0.0101 |
| Cardiovascular disease | 90 (13.82)% | 101 (15.51)% | −0.0477 |
| Cataracts | 70 (10.75)% | 65 (9.98)% | 0.0252 |
| COPD | 38 (5.84)% | 43 (6.61)% | −0.0318 |
| Endometrial cancer | n<11 | n<11 | 0.0210 |
| Exercise | 13 (2.0)% | 16 (2.46)% | −0.0312 |
| Gastroparesis | 42 (6.45)% | 34 (5.22)% | 0.0524 |
| Glaucoma | 27 (4.15)% | 28 (4.3)% | −0.0076 |
| Gout | 13 (2.0)% | 18 (2.76)% | −0.0504 |
| Hearing impairment | 22 (3.38)% | 22 (3.38)% | 0.0000 |
| HHF | 29 (4.45)% | 42 (6.45)% | −0.0880 |
| Hip Pelvic Fracture | n<11 | n<11 | 0.0210 |
| Lower extremity ulcers | 48 (7.37)% | 50 (7.68)% | −0.0116 |
| Lung cancer | n<11 | n<11 | −0.0554 |
| Bladder cancer | n<11 | n<11 | 0.0554 |
| MCI | n<11 | n<11 | 0.0784 |
| Neuropathy | 433 (66.51)% | 435 (66.82)% | −0.0065 |
| Obstruction | n<11 | n<11 | −0.0556 |
| OCD | n<11 | n<11 | −0.0352 |
| Osteoporosis | 22 (3.38)% | 19 (2.92)% | 0.0264 |
| Parkinson | n<11 | n<11 | −0.0630 |
| Periodontitis | n<11 | n<11 | 0.0554 |
| Prostate cancer | n<11 | n<11 | 0.0907 |
| PTSD | 13 (2.0)% | 11 (1.69)% | 0.0228 |
| Schizophrenia | n<11 | n<11 | 0.0962 |
| Seizures | 14 (2.15)% | 11 (1.69)% | 0.0336 |
| Sleep disorder | 141 (21.66)% | 148 (22.73)% | −0.0259 |
| Substance use | 82 (12.6)% | 94 (14.44)% | −0.0539 |
| Suicide | n<11 | 12 (1.84)% | −0.0640 |
| TBI | n<11 | n<11 | 0.0557 |
| Hypothyroidism | 185 (28.42)% | 201 (30.88)% | −0.0538 |
| VD | n<11 | n<11 | −0.0554 |
| Vision impairment | n<11 | n<11 | 0.0298 |
| ESRD | 29 (4.45)% | 16 (2.46)% | 0.1094 |
| Bariatric procedure | 134 (20.58)% | 133 (20.43)% | 0.0038 |
| Bariatric surgery | 279 (42.86)% | 276 (42.4)% | 0.0093 |
| Pregnant | 36 (5.53)% | 40 (6.14)% | −0.0262 |
| CCI Comorbidities | |||
| Myocardial infarction | 199 (30.57)% | 190 (29.19)% | 0.0302 |
| Congestive heart failure | 34 (5.22)% | 45 (6.91)% | −0.0708 |
| Peripheral vascular disease | 40 (6.14)% | 44 (6.76)% | −0.0250 |
| Cerebrovascular disease | 29 (4.45)% | 37 (5.68)% | −0.0560 |
| Chronic pulmonary disease | n<11 | n<11 | −0.0630 |
| Rheumatic disease | 80 (12.29)% | 92 (14.13)% | −0.0544 |
| Peptic ulcer disease | 15 (2.3)% | 19 (2.92)% | −0.0385 |
| Mild liver disease | n<11 | n<11 | −0.0168 |
| Diabetes without chronic complication | 64 (9.83)% | 64 (9.83)% | 0.0000 |
| Diabetes with chronic complication | 637 (97.85)% | 640 (98.31)% | −0.0336 |
| Hemiplegia or paraplegia | 225 (34.56)% | 214 (32.87)% | 0.0357 |
| Renal disease | n<11 | n<11 | −0.0745 |
| Any malignancy, including lymphoma and leukemia, except malignant neoplasm of skin | 87 (13.36)% | 73 (11.21)% | 0.0655 |
| Moderate or severe liver disease | 34 (5.22)% | 20 (3.07)% | 0.1079 |
| Metastatic solid tumor | n<11 | n<11 | 0.0000 |
| AIDS/HIV | n<11 | n<11 | 0.0279 |
| CCI Score | 2.3 (2.08) | 2.16 (1.83) | 0.0198 |
| Outcomes | |||
| DKA | 16 (2.5%) | 24 (3.7%)9%) | |
| Hospitalizations | 193 (29.6%) | 237 (36.4%) | |
| ED visits | 124 (19.0%) | 156 (24.0%) |
SMD, standardized mean difference; ADRD, Alzheimer’s Disease and Related Dementias; COPD, Chronic obstructive pulmonary disease; ACEI, angiotensin-converting-enzyme inhibitors; ARB, angiotensin II receptor blockers; NSAIDS, non-steroidal anti-inflammatory drugs; TNF inhibitors, tumor necrosis factor inhibitors; HRT, hormone replacement therapy; GERD, gastroesophageal reflux disease; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; HHF, hypertensive heart failure; MCI, mild cognitive impairment; OCD, Obsessive-compulsive disorder; PTSD, Post-traumatic stress disorder; TBI, traumatic brain injury; ESRD, end-stage renal disease; CCI, Carlson comorbidity index; Others: include American Indian or Alaska Native, Asian, Native Hawaiian or Other Pacific Islander, and individuals identifying as multiracial.
Statistical Analysis
To address potential confounding and enhance comparability between two treatment groups, we implemented a time-dependent 1:1 propensity score (PS) matching strategy.26 Propensity scores were derived from a multivariable logistic regression model incorporating a comprehensive set of baseline covariates. Matching was performed using a nearest-neighbor (KNN) algorithm without replacement and a caliper of 0.2, ensuring that each GLP-1RA user was matched to a non-user with the closest propensity score within the specified threshold.27 Covariate balance between matched pairs was evaluated using standardized mean differences (SMDs), with values below 0.1 considered acceptable balance between GLP-1RA users and non-users.
Furthermore, Cohen’s d28 was used to assess the magnitude of difference in covariates before and after matching, providing additional assurance that confounding was adequately minimized in the final matched cohort used for outcome analyses. Additional details are provided in eMethod 1.
We calculated the incidence rate (IR) of study outcomes and employed Kaplan–Meier survival analysis to illustrate the cumulative incidence per 1000 person-years over time between the matched groups. Additionally, Cox proportional hazard regression models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) by comparing GLP-1RAs users vs. non-users.
Identification of Heterogenous Treatment Effect (HTE) and Potential Moderators
We employed a doubly-robust meta-learner29 following the potential outcome framework. A logistic algorithm was used to estimate the PS for GLP-1RA use, while an XGBoost regression model30 predicted the outcomes. To ensure unbiased estimation, we applied a triple cross-fitting strategy and used bootstrapping for hyperparameter tuning within each round. Additional methodological details are provided in eMethod 2. We estimated absolute risk differences (RDs) with 95% CIs to assess the average exposure effect, with negative RDs indicating benefit and positive values indicating harm. We conducted HTE analysis based on the estimations to explore risk variation using an interpretable decision tree.31 We also identified potential moderators based on key decision nodes that distinguish subpopulations with heterogeneous effects.
To further assess the strength of interaction effects for candidate moderators, we fit Cox proportional hazards models that included interaction terms of the treatment variable and candidate moderators, with robust variance estimation applied.32
Subgroup Analysis
Subgroup analyses were conducted for both prespecified factors and factors identified through a data-driven approach via HTE modeling. The prespecified subgroups included: age groups (<40, 40–64, and ≥65 years), sex (male vs. female), racial and ethnic categories (non-Hispanic White, non-Hispanic Black, Hispanic, and other groups), obesity severity (overweight vs. obesity), baseline Metformin (use vs. non-use), and individual GLP1-RA drugs (liraglutide, semaglutide, and tirzepatide).
The second set of subgroup analyses involved potential moderators identified through HTE modeling. We conducted subgroup analyses to assess whether the treatment effect of GLP-1RA varied meaningfully across distinct patient strata, thereby evaluating whether the identified moderators consistently modified the effect in real-world clinical subgroups.
Results
The flowchart of patient selection is presented in Figure 1. The study population comprised 15,585 T1D patients eligible for AOM treatment in OneFlorida+ from 2014–2024, including 664 GLP-1RA user and 14,921 GLP-1RA non-users. After applying 1:1 PS matching (eFigure 1 & 2 in Supplement), the final cohort included 1,302 matched participants (651 GLP-1RA user vs. 651 non-user). After PS matching, baseline characteristics showed a balanced distribution of key covariates, as detailed in eTable 4 of Supplement. In the matched cohort (Table1), mean age was 44.8 (± 16) years, 63.6% were female, and 43.4% were non-Hispanic White. The mean BMI at baseline was 34.6 kg/m2. The two treatment groups showed similar distributions in baseline HbA1c levels, hypertension prevalence, and other clinical factors.
Figure 1.
Flowchart of patient selection
eFigure 3 illustrates Kaplan-Meier plots depicting the cumulative incidence during follow-up. GLP-1RA users exhibited a lower incidence rate of DKA (13.5 per 1,000 person-years) compared with non-users (21.8 per 1,000 person-years). However, GLP-1RA use was not significantly associated with incidence of DKA (HR 0.62 [95%CI 0.33–1.17]) or severe hypoglycemia (HR 0.52 [95%CI 0.17–1.55]). Additionally, GLP-1RA use was associated with a significantly fewer hospitalizations (HR 0.74 [95%CI 0.62–0.90]) and ED (HR 0.73 [95%CI 0.57–0.92]).
HTE Results
The details regarding data preprocessing and hyperparameter tuning were shown in eResult 1 in the Supplement. The estimated RD for DKA was negative in GLP-1RA users vs non-users (RD: −0.76% [95% CI: −1.44 to 1.29]), suggesting a potentially lower DKA risk associated with GLP-1RA but without statistical significance; the results were consistent with those from Cox regression models.
To further explore HTE, we employed interpretable decision tree models to identify patient subgroups with differential GLP-1RA effects on DKA (see Figure 2). Specifically, patients without hypothyroidism appeared to benefit from GLP-1RA treatment, particularly among those with dyslipidemia. In contrast, patients with hypothyroidism may experience potential harm (e.g., a positive risk difference).
Figure 2.
Interpretable decision tree for DKA.
Effect modification assessment
There was a significant interaction between GLP-1RA and hypothyroidism (interaction HR, 1.48 [95% CI, 1.06–2.08]), suggesting that the benefit of GLP-1RA was attenuated among patients with thyroid conditions. Conversely, the interaction between GLP-1RA and dyslipidemia was not statistically significant (HR, 0.77 [95% CI, 0.56–1.08]), although a trend toward stronger benefit among dyslipidemia patients was observed.
A significant three-way interaction was found between GLP-1RA, hypothyroidism, and dyslipidemia (HR, 1.42 [95% CI, 1.04–1.94]), indicating that the combination of both conditions may further modify the treatment effect. These findings suggest that patients without hypothyroidism, particularly those with dyslipidemia, may derive greater benefit from GLP-1RA therapy in reducing the risk of DKA.
Subgroup Analysis and Sensitivity analysis
The results for DKA are presented in Figure 3. The reduced risk of DKA associated with GLP-1RA use was particularly notable among individuals who had not used metformin at baseline (HR, 0.30 [95% CI, 0.10–0.91]). Additionally, we stratified the cohort based on potential moderators identified through interpretable decision tree analysis. GLP-1RA use was significantly associated with a reduced risk of DKA among patients without hypothyroidism (HR, 0.47 [95% CI, 0.24–0.92]), whereas no benefit was observed among those with hypothyroidism (HR, 1.08 [95% CI, 0.35–3.36]). Notably, the most pronounced benefit was observed among patients without hypothyroidism but with dyslipidemia compared to non-users (HR, 0.12 [95% CI, 0.03–0.55]).
Figure 3.
Subgroup analysis of DKA in patients receiving GLP-1 RA vs. those not receiving GLP-1 RA.
Regarding hospitalizations or ED visits (see eFigure 4), most subgroups demonstrated a significant reduction in risk associated with GLP-1RA use or showed a potential trend toward benefit.
Furthermore, we conducted a sensitivity analysis by composing DKA and severe hypoglycemia. While GLP-1RA use was not significantly associated with a reduced risk, it demonstrated a favorable trend (HR 0.65 [95% CI, 0.37–1.14]).
Discussion
In our target trial emulation study leveraging real-world data from the OneFlorida+ EHR network, GLP-1RA use, compared to non-use, was not associated with a risk of DKA or severe hypoglycemia but linked fewer hospitalizations and ED visits among adults with T1D and obesity. While GLP-1RAs have consistently shown benefits in glycemic control, weight reduction, and cardiovascular outcomes among patients withT2D or obesity,33,34 evidence supporting their safe use in T1D remains limited and mixed, partly due to concerns about acute diabetes complications, such as hypoglycemia and DKA.12,15,35 Our findings contribute important real-world evidence to this understudied area, suggesting that GLP-1RAs may offer broader clinical benefits without introducing safety concerns on acute diabetes complications among individuals with T1D.
Prior studies evaluating GLP-1RA adjunct therapy in T1D have yielded mixed results. Early trials such as ADJUNCT ONE and TWO reported modest glycemic improvements but raised safety concerns, noting higher rates of hyperglycemia with ketosis with a high-dose GLP-1RA (liraglutide 1.8 mg).35,36 These findings tempered enthusiasm of GLP-1 RA use in T1D despite their marked benefits in improving cardiovascular and renal outcomes.37,38 In contrast, recent evidence suggests potential benefits of GLP-1RAs in T1D and Obesity.39 Meta-analyses and real-world studies indicate that adding a GLP-1RA can improve metabolic outcomes (lower HbA1c, weight loss, and reduced insulin requirements) without increasing the incidence of severe hypoglycemia or DKA.40 These prior observations set the stage for our investigation. In this context, our findings contrast with earlier safety concerns and instead underscore the safety of GLP-1RA regarding acute diabetes complications, as well as its broader clinical benefits in reducing hospitalizations and ER visits.
While our prespecified subgroup analyses were consistent with the main analysis overall, we observed that GLP-1RAs may offer the benefit of lowering DKA risk in particularly in patients who did not use metformin at baseline. One potential explanation is that both metformin and GLP-1RAs improve insulin sensitivity and reduce hepatic glucose output; their combined effects might lead to lower exogenous insulin requirements, which—if not carefully adjusted—can increase ketogenic risk.41 However, given the exploratory nature and limited power of these subgroup analyses, additional studies with larger sample size and longer follow-up are warranted to confirm these findings and to better characterize safety and effectiveness differences across subpopulations.
Using a data-driven approach, we identified that hypothyroidism may modify the association between GLP-1RA use and the DKA. Specifically, GLP-1RA use was associated with a lower risk of DKA among patients without hypothyroidism, but this protective association was not observed in those with hypothyroidism. Given that hypothyroidism is a common comorbidity in individuals with T1D—largely due to underlying autoimmune mechanisms—thyroid function may influence both glucose metabolism and the efficacy of adjunctive therapies such as GLP-1RAs.42 Hypothyroidism is linked to reduced insulin sensitivity, impaired glucose and ketone metabolism, and blunted counterregulatory responses, all of which may increase susceptibility to DKA.43 Also, both hypothyroidism and GLP-1RAs slow gastric motility, possibly affecting adherence and glucose control. These factors may attenuate the DKA risk reduction associated with GLP-1RAs observed in T1D patients without thyroid dysfunction.
Several limitations should be acknowledged when interpreting these results. First, despite the rigorous PS matching and adjustments for a broad range of covariates, residual confounding from unmeasured factors such as diet, physical activity, patient adherence, or socioeconomic status remains possible. In addition, the relatively limited number of events within specific subgroups led to wide confidence intervals. These results, therefore, must be interpreted cautiously, and future larger-scale studies or randomized controlled trials are needed to validate these findings.
Conclusion
In conclusion, in this target trial emulation study of adults with T1D who had overweight or obesity, the use of GLP-1RAs was associated with fewer hospitalizations or ED visits, without increasing the risk of acute complications, including DKA and severe hypoglycemia. These findings provide valuable clinical evidence supporting the safety of GLP-1RAs use in this vulnerable patient population. Further research, particularly larger prospective studies and randomized controlled trials is warranted to confirm these findings.
ACKNOWLEDGMENT
Funding/Support:
The study was supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIH/NIDDK) R01DK133465.
Role of the Funder/Sponsor:
The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Funding Statement
The study was supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIH/NIDDK) R01DK133465.
Footnotes
Conflict of Interest Disclosures: None reported.
Data Availability Statement:
Data set Available through OneFlorida+ Clinical Research Network (email, oneFloridaOperations@health.ufl.edu)
Reference
- 1.Holt RIG, DeVries JH, Hess-Fischl A, et al. Correction to: The management of type 1 diabetes in adults. A consensus report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). Diabetologia. 2022;65(1):255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Dib SA. Hypoglycemia in type 1 diabetes: a burden to worry about during treatment. Arch Endocrinol Metab. 2022;66(6):776–779. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Calimag APP, Chlebek S, Lerma EV, Chaiban JT. Diabetic ketoacidosis. Dis Mon. 2023;69(3):101418. [DOI] [PubMed] [Google Scholar]
- 4.Liu J, Wang R, Ganz ML, Paprocki Y, Schneider D, Weatherall J. The burden of severe hypoglycemia in type 1 diabetes. Curr Med Res Opin. 2018;34(1):171–177. [DOI] [PubMed] [Google Scholar]
- 5.Ramphul K, Joynauth J. An update on the incidence and burden of diabetic ketoacidosis in the U.s. Diabetes Care. 2020;43(12):e196–e197. [DOI] [PubMed] [Google Scholar]
- 6.Van der Schueren B, Ellis D, Faradji RN, Al-Ozairi E, Rosen J, Mathieu C. Obesity in people living with type 1 diabetes. Lancet Diabetes Endocrinol. 2021;9(11):776–785. [DOI] [PubMed] [Google Scholar]
- 7.Kueh MTW, Chew NWS, Al-Ozairi E, le Roux CW. The emergence of obesity in type 1 diabetes. Int J Obes (Lond). 2024;48(3):289–301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Drucker DJ. Mechanisms of action and therapeutic application of glucagon-like peptide-1. Cell Metab. 2018;27(4):740–756. [DOI] [PubMed] [Google Scholar]
- 9.Nauck MA, Meier JJ. Incretin hormones: Their role in health and disease. Diabetes Obes Metab. 2018;20:5–21. [DOI] [PubMed] [Google Scholar]
- 10.Shah VN, Peters AL, Umpierrez GE, et al. Consensus report on glucagon-like peptide-1 receptor agonists as adjunctive treatment for individuals with type 1 diabetes using an automated insulin delivery system. J Diabetes Sci Technol. 2025;19(1):191–216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Guyton J, Jeon M, Brooks A. Glucagon-like peptide 1 receptor agonists in type 1 diabetes mellitus. Am J Health Syst Pharm. 2019;76(21):1739–1748. [DOI] [PubMed] [Google Scholar]
- 12.Hirsch IB, Parkin CG, Cavaiola TS, Bergenstal RM. Use of continuous glucose monitoring when initiating glucagon-like peptide-1 receptor agonist therapy in insulin-treated diabetes. Diabetes Obes Metab. 2024;26 Suppl 7(S7):17–26. [DOI] [PubMed] [Google Scholar]
- 13.Htike ZZ, Zaccardi F, Papamargaritis D, Webb DR, Khunti K, Davies MJ. Efficacy and safety of glucagon-like peptide-1 receptor agonists in type 2 diabetes: A systematic review and mixed-treatment comparison analysis. Diabetes Obes Metab. 2017;19(4):524–536. [DOI] [PubMed] [Google Scholar]
- 14.Zhao Z, Tang Y, Hu Y, Zhu H, Chen X, Zhao B. Hypoglycemia following the use of glucagon-like peptide-1 receptor agonists: a real-world analysis of post-marketing surveillance data. Ann Transl Med. 2021;9(18):1482. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yang Z, Yu M, Mei M, et al. The association between GLP-1 receptor agonist and diabetic ketoacidosis in the FDA adverse event reporting system. Nutr Metab Cardiovasc Dis. 2022;32(2):504–510. [DOI] [PubMed] [Google Scholar]
- 16.Widhalm C, Pulido V. Euglycemic diabetic keto acidosis in a type 1 diabetic patient after glucose like peptide-1 administration: A case presentation. J Investig Med High Impact Case Rep. 2023;11:23247096231206332. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Zhang J, Ma Y, Zu Q, Wang X, Zhang Y. GLP-1 receptor agonist-induced diabetic ketoacidosis: A case report. Medicine (Baltimore). 2024;103(39):e39799. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.PCORnet® Common Data Model. The National Patient-Centered Clinical Research Network.
- 19.Saunders KH, Umashanker D, Igel LI, Kumar RB, Aronne LJ. Obesity pharmacotherapy. Med Clin North Am. 2018;102(1):135–148. [DOI] [PubMed] [Google Scholar]
- 20.Evans-Molina C, Oram RA. Type 1 diabetes presenting in adults: Trends, diagnostic challenges and unique features. Diabetes Obes Metab. 2025;27 Suppl 6(S6):57–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Shao H, Thorpe LE, Islam S, et al. Developing a computable phenotype for identifying Children, Adolescents, and Young Adults with diabetes using electronic health records in the DiCAYA network. Diabetes Care. Published online March 31, 2025. doi: 10.2337/dc24-1972 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Li P, Li Z, Staton E, et al. GLP-1 receptor agonist and SGLT2 inhibitor prescribing in people with type 1 diabetes. JAMA. 2024;332(19):1667–1669. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.VanderWeele J, Pollack T, Oakes DJ, et al. Validation of data from electronic data warehouse in diabetic ketoacidosis: Caution is needed. J Diabetes Complications. 2018;32(7):650–654. [DOI] [PubMed] [Google Scholar]
- 24.Hodzic-Santor B, Colacci M, Raissi A, et al. Validation of the diagnostic accuracy levels of international Classification of Diseases, 10th revision codes for diabetic ketoacidosis: A multicentre, cross-sectional study of adults. Can J Diabetes. 2024;48(4):227–232. [DOI] [PubMed] [Google Scholar]
- 25.Karter AJ, Warton EM, Moffet HH, et al. Revalidation of the hypoglycemia risk stratification tool using ICD-10 codes. Diabetes Care. 2019;42(4):e58–e59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Zhang Z, Li X, Wu X, Qiu H, Shi H, written on behalf of AME Big-Data Clinical Trial Collaborative Group. Propensity score analysis for time-dependent exposure. Ann Transl Med. 2020;8(5):246. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Benedetto U, Head SJ, Angelini GD, Blackstone EH. Statistical primer: propensity score matching and its alternatives. Eur J Cardiothorac Surg. 2018;53(6):1112–1117. [DOI] [PubMed] [Google Scholar]
- 28.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083–3107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Kennedy EH. Towards optimal doubly robust estimation of heterogeneous causal effects. Electron J Stat. 2023;17(2). doi: 10.1214/23-ejs2157 [DOI] [Google Scholar]
- 30.Chen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. arXiv [csLG]. Published online March 8, 2016. http://arxiv.org/abs/1603.02754 [Google Scholar]
- 31.Tang H, Guo J, Shaaban CE, et al. Heterogeneous treatment effects of metformin on risk of dementia in patients with type 2 diabetes: A longitudinal observational study. Alzheimers Dement. 2024;20(2):975–985. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Memon MA, Cheah JH, Ramayah T, Ting H, Chuah F, Cham TH. Moderation analysis: Issues and guidelines. Journal of Applied Structural Equation Modeling. 2019;3(1):i–xi. [Google Scholar]
- 33.Marso SP, Daniels GH, Brown-Frandsen K, et al. Liraglutide and cardiovascular outcomes in type 2 diabetes. N Engl J Med. 2016;375(4):311–322. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Wilding JPH, Batterham RL, Calanna S, et al. Once-weekly semaglutide in adults with overweight or obesity. N Engl J Med. 2021;384(11):989–1002. [DOI] [PubMed] [Google Scholar]
- 35.Mathieu C, Zinman B, Hemmingsson JU, et al. Efficacy and safety of liraglutide added to insulin treatment in type 1 diabetes: The ADJUNCT ONE treat-to-target randomized trial. Diabetes Care. 2016;39(10):1702–1710. [DOI] [PubMed] [Google Scholar]
- 36.Ahrén B, Hirsch IB, Pieber TR, et al. Efficacy and safety of liraglutide added to capped insulin treatment in subjects with type 1 diabetes: The ADJUNCT TWO randomized trial. Diabetes Care. 2016;39(10):1693–1701. [DOI] [PubMed] [Google Scholar]
- 37.Kristensen SL, Rørth R, Jhund PS, et al. Cardiovascular, mortality, and kidney outcomes with GLP-1 receptor agonists in patients with type 2 diabetes: a systematic review and meta-analysis of cardiovascular outcome trials. Lancet Diabetes Endocrinol. 2019;7(10):776–785. [DOI] [PubMed] [Google Scholar]
- 38.Sattar N, Lee MMY, Kristensen SL, et al. Cardiovascular, mortality, and kidney outcomes with GLP-1 receptor agonists in patients with type 2 diabetes: a systematic review and meta-analysis of randomised trials. Lancet Diabetes Endocrinol. 2021;9(10):653–662. [DOI] [PubMed] [Google Scholar]
- 39.Shah VN, Akturk HK, Kruger D, et al. Semaglutide in adults with type 1 diabetes and obesity. NEJM Evid. Published online June 23, 2025:EVIDoa2500173. [DOI] [PubMed] [Google Scholar]
- 40.Tan X, Pan X, Wu X, et al. Glucagon-like peptide-1 receptor agonists as add-on therapy to insulin for type 1 diabetes mellitus. Front Pharmacol. 2023;14:975880. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Podobnik J, Prentice KJ. Metabolic interventions as adjunctive therapies to insulin in type 1 diabetes: Current clinical landscape and perspectives. Diabetes Obes Metab. 2025;27(3):1032–1044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Pathophysiology. Thyroid Dysfunction in Patients With Type 1 Diabetes: A Longitudinal Study.; 2003. [DOI] [PubMed]
- 43.Eom YS, Wilson JR, Bernet VJ. Links between thyroid disorders and glucose homeostasis. Diabetes Metab J. 2022;46(2):239–256. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data set Available through OneFlorida+ Clinical Research Network (email, oneFloridaOperations@health.ufl.edu)