{
  "abstract": "Introduction To assess the within-class variation in kidney outcomes following initiation of sulfonylurea therapy.Research design and methods We used claims data of enrollees in commercial, Medicare Advantage, and traditional Medicare health plans between 2014 and 2021 to emulate a target trial including adults ≥21 years with type 2 diabetes at moderate cardiovascular risk to compare initiation of glimepiride, glipizide, or glyburide on the incidence of chronic kidney disease (CKD) stage 3 or worse, including initiation of kidney replacement therapy (primary outcome); secondary outcomes examined incident CKD stages 3-4, kidney failure (including kidney replacement therapy), all-cause mortality, and hypoglycemia requiring emergency department or hospital care. Random treatment assignment was emulated using propensity scores, estimated using the super learner ensemble method, and incorporated as inverse probability of treatment weights into proportional hazards models.Results The weighted study cohort included 295 092 individuals starting glimepiride (n=134 926), glipizide (n=145 984), and glyburide (n=14 182). One year after treatment initiation, stage 3 or worse CKD developed in 2.1% of patients in the glimepiride group, 2.2% in the glipizide group, and 1.8% in the glyburide group. Glyburide was associated with a lower risk of kidney complications compared with both glimepiride (HR 0.84, 95% CI 0.76 to 0.92) and glipizide (HR 0.81, 95% CI 0.73 to 0.89), despite a higher risk of severe hypoglycemia (HR 1.47, 95% CI 1.27 to 1.71 vs glipizide and HR 1.22, 95% CI 1.05 to 1.42 vs glimepiride). In contrast, the risk of kidney complications was modestly increased with glipizide compared with glimepiride use (HR 1.04, 95% CI 1.00 to 1.07).Conclusions Glyburide was associated with a modestly lower risk of kidney complications, despite a higher risk of hypoglycemia, while glipizide was associated with a higher risk of kidney complications. These hypothesis-generating findings suggest important within-class differences that warrant consideration in clinical decision-making and future research. Despite rigorous prespecified causal inference analytic methods, the risk of unmeasured confounding and bias by indication with the use of observational data remains.Trial registration number NCT05214573.",
  "authors": [
    {
      "affiliations": [
        "School of Medicine, University of Maryland Baltimore, Baltimore, Maryland, USA"
      ],
      "name": "Stacey Sklepinski"
    },
    {
      "affiliations": [
        "Division of Gerontology, Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, Maryland, USA",
        "University of Maryland Institute for Health Computing, North Bethesda, Maryland, USA"
      ],
      "name": "Alexandria A Ratzki-Leewing"
    },
    {
      "affiliations": [
        "Section of Cardiovascular Medicine, Department of Medicine, Yale School of Medicine, New Haven, Connecticut, USA"
      ],
      "name": "Jeph Herrin"
    },
    {
      "affiliations": [
        "Mayo Clinic Robert D and Patricia E Kern Center for the Science of Health Care Delivery, Rochester, Minnesota, USA",
        "Optum Labs, Eden Prairie, Minnesota, USA"
      ],
      "name": "Kavya Sindhu Swarna"
    },
    {
      "affiliations": [
        "Mayo Clinic Robert D and Patricia E Kern Center for the Science of Health Care Delivery, Rochester, Minnesota, USA",
        "Optum Labs, Eden Prairie, Minnesota, USA"
      ],
      "name": "Yihong Deng"
    },
    {
      "affiliations": [
        "Department of Public Health Sciences, The University of Chicago, Chicago, Illinois, USA"
      ],
      "name": "Eric C Polley"
    },
    {
      "affiliations": [
        "Department of Pharmacotherapy, Washington State University, Spokane, Washington, USA",
        "Providence Medical Research Center, Spokane, Washington, USA"
      ],
      "name": "Joshua J Neumiller"
    },
    {
      "affiliations": [
        "Division of Endocrinology, Department of Medicine, University of Miami Miller School of Medicine, Miami, Florida, USA"
      ],
      "name": "Rodolfo J Galindo"
    },
    {
      "affiliations": [
        "Division of Endocrinology, Department of Medicine, Emory University School of Medicine, Atlanta, Georgia, USA"
      ],
      "name": "Guillermo E Umpierrez"
    },
    {
      "affiliations": [
        "Department of Internal Medicine, Yale School of Medicine, New Haven, Connecticut, USA",
        "Department of Health Policy and Management, Yale School of Public Health, New Haven, Connecticut, USA"
      ],
      "name": "Joseph S Ross"
    },
    {
      "affiliations": [
        "Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic, Rochester, Minnesota, USA",
        "Knowledge and Evaluation Research Unit, Mayo Clinic, Rochester, Minnesota, USA"
      ],
      "name": "Juan P Brito"
    },
    {
      "affiliations": [
        "Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic, Rochester, Minnesota, USA",
        "Knowledge and Evaluation Research Unit, Mayo Clinic, Rochester, Minnesota, USA"
      ],
      "name": "Victor M Montori"
    },
    {
      "affiliations": [
        "Mayo Clinic Robert D and Patricia E Kern Center for the Science of Health Care Delivery, Rochester, Minnesota, USA"
      ],
      "name": "Mindy M Mickelson"
    },
    {
      "affiliations": [
        "Division of Gerontology, Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, Maryland, USA",
        "University of Maryland Institute for Health Computing, North Bethesda, Maryland, USA",
        "Division of Endocrinology, Diabetes, & Nutrition, University of Maryland School of Medicine, Baltimore, Maryland, USA"
      ],
      "name": "Rozalina Grubina McCoy"
    }
  ],
  "full_text": "WHAT IS ALREADY KNOWN ON THIS TOPIC Limited data exist directly comparing individual sulfonylurea agents with respect to adverse kidney outcomes.WHAT THIS STUDY ADDS This study provides the first head-to-head comparison of kidney outcomes with use of the three most commonly prescribed sulfonylureas in the US. We found that glyburide was associated with a modestly lower risk of kidney complications, while glipizide was associated with a higher risk, among adults with type 2 diabetes at moderate cardiovascular risk.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These results highlight the need to reconsider assumptions about within-class effects of sulfonylurea drugs and may help support clinical decision-making when sulfonylurea therapy is required.Introduction Preventing kidney complications and related sequelae is central to optimal type 2 diabetes (T2D) management. In the USA, diabetes ranks as a leading cause of chronic kidney disease (CKD) and kidney failure, 1 accounting for 39% of kidney failure diagnoses1 and 32% of all CKD-related medical expenditures.2 Furthermore, diabetic kidney disease has been associated with a nearly threefold increase in 10-year cumulative mortality risk compared with individuals with T2D but without kidney disease.3 Glucagon-like peptide-1 receptor agonists (GLP-1RAs) and sodium glucose co-transporter 2 inhibitors (SGLT2is) slow CKD progression and reduce the risk of kidney failure in adults with T2D,4 5 yet their uptake in the USA has been limited by cost and access considerations, and for many patients, they may be insufficient to achieve or sustain glycemic targets. Current guidelines therefore continue to endorse sulfonylureas as monotherapy or add-on therapy in patients without established or high cardiovascular risk, citing their potent glucose-lowering efficacy and low cost. In the USA, sulfonylureas remain among the most frequently prescribed glucose-lowering agents in T2D, second only to metformin, with glimepiride leading in use.6 7However, despite this widespread use, little is known about the long-term effects of sulfonylureas on kidney outcomes or potential within-class differences. This evidence gap is particularly relevant for the majority of adults with T2D who have moderate levels of cardiovascular risk8 and for whom sulfonylureas are considered a reasonable treatment option under current guidelines.4 Prior studies have found that the risk of incident kidney disease is higher among patients treated with sulfonylureas than those treated with metformin, rosiglitazone, dipeptidyl peptidase 4 inhibitor (DPP4i), GLP-1RA, or SGLT2i drugs.9 10 Meanwhile, a 2015 network meta-analysis showed that glimepiride and glipizide had a similar risk of all-cause mortality compared with glyburide.11Currently, there are three sulfonylureas approved by the US Food and Drug Administration and in use in the USA: glimepiride, glipizide, and glyburide. While they are all second-generation sulfonylureas, they have important pharmacological differences that may contribute to differential kidney-related safety effects. Glyburide is the longest-acting sulfonylurea and, as such, carries the highest risk of hypoglycemia.12 Glyburide also has the strongest extrapancreatic effects, binding to sulfonylurea receptor-1 (SUR1; found primarily in pancreatic β cells), SUR2A (found primarily in cardiac muscle cells), and SUR2B (found in vascular smooth muscle and glomerular mesangial cells), and may therefore impact end-organ complications to a greater extent than glipizide, which is both shorter-acting and more pancreas (SUR1) specific.13 Glimepiride also has extrapancreatic effects but with weaker binding affinity across the different KATP channels. We recently showed that glimepiride was associated with the lowest risk of major adverse cardiovascular events (MACE), while glyburide was associated with the highest risk14; however, the comparative safety of different sulfonylurea drugs on kidney outcomes among adults with T2D at moderate cardiovascular risk remains unknown.To address this knowledge gap, we emulated a target trial using US-wide medical and pharmaceutical claims data to compare the incident rates of moderate to severe kidney disease following initiation of glimepiride, glipizide, and glyburide among adults with T2D at moderate cardiovascular risk and without baseline evidence of CKD.Materials and methods Study design  This longitudinal cohort study was conducted under the target trial emulation framework 15 using observational data for adults with T2D at moderate cardiovascular risk. The target trial emulation framework was designed to minimize common biases in observational research,15 including confounding by indication, immortal time bias, and selection bias, by explicitly emulating key design elements of a randomized clinical trial, including randomized treatment assignment via propensity score weighting, intention-to-treat (ITT) analysis, and new-user design with a baseline assessment period. Results reported here are a prespecified secondary analysis of a larger study that compared different classes of T2D medications10 and focused on within-class comparisons of the sulfonylurea drugs. The target trial and emulation methods were preregistered on ClinicalTrials.gov (NCT05214573). Online supplemental table 1 details the target trial and its emulation. The study was reviewed and determined to be exempt by the Mayo Clinic Institutional Review Board.SP110.1136/bmjdrc-2025-005775.supp1Supplementary dataData sources  We used deidentified claims data from the Optum Labs Data Warehouse (OLDW), which were linked to the 100% sample of traditional (fee-for-service) Medicare parts A, B, and D claims using personal identifiers prior to deidentification (and prior to being made available to researchers). The OLDW includes longitudinal medical and pharmacy claims and enrollment records for enrollees in commercial and Medicare Advantage plans. Once linked with traditional Medicare claims, which enabled uninterrupted observation of individuals as they switch between health plans, the analytic dataset is highly heterogeneous and representative of a wide range of ages, racial, ethnic, and socioeconomic groups, health systems, and geographic regions across the USA. 16Study cohort We identified adults ≥21 years of age filling their first prescription for glimepiride, glipizide (both short-acting and extended release), or glyburide between January 1, 2014 and December 31, 2021 ( online supplemental figure 1 and online supplemental table 2). The date of the first fill served as the index date. Prior to the index date, patients must have had 12 months of baseline enrollment without a sulfonylurea fill to ensure a new user design, with a 30-day grace period to confirm early adherence as noted below. Patients were required to have a second fill of the same drug with no more than a 30-day coverage gap to confirm use and not have filled a different sulfonylurea during this time. Exclusion criteria applied to the 12-month baseline period included: (1) missing year of birth, sex, or region; (2) death date before index date, which suggests an error in the record; (3) both OLDW and Medicare coverage on the index date; (4) International Classification of Diseases (ICD)-9/10 codes for type 1 diabetes, metastatic cancer, pregnancy, or stage 3 or worse CKD, including ICD-9/10 or Current Procedural Terminology codes for kidney replacement therapy (KRT) (online supplemental table 3); (5) a fill for a meglitinide or insulin (online supplemental table 2) and (6) estimated low (<1%) or high (>5%) cardiovascular risk, estimated using covariates from the baseline 12 months period using the annualized claims-based MACE estimator.8Outcomes  The primary outcome was a composite of incident CKD stages 3 or worse, including kidney failure and initiation of KRT. Secondary outcomes included a second kidney composite (the primary kidney composite plus all-cause mortality) and the individual components of the second kidney composite outcome. Online supplemental table 4 details the diagnosis codes used to ascertain these outcomes. Mortality data are included in the OLDW and are derived from the Social Security Administration Death Master File, deceased status from OLDW-linked electronic health records, death as a reason for disenrollment from an included health plan, death indicated by inpatient discharge status, obituary information, and Medicare Advantage beneficiary report information.Independent variables Covariates were ascertained from the baseline 12-month period and included demographics (age, sex, race and ethnicity, US census region), clinical comorbidities ( online supplemental table 5), and medications (online supplemental table 2).Statistical analysis  Our primary analytic approach was under the ITT framework, observing patients until death, disenrollment, or the end of the study period (December 31, 2022), whichever came first. We first estimated propensity scores to model the probability of treatment assignment to each sulfonylurea drug (vs the pool of the other two), including the baseline covariates detailed above in a diverse set of individual binomial prediction algorithms included in the SuperLearner ensemble ( online supplemental methods).17 18 The super learner framework, as outlined by van der Lann et al,18 allows for flexible estimation of an ensemble predictor with the theoretical properties of cross-validation for model selection to control for overfitting. To avoid concerns about extreme weights, stabilized weights without trimming were used where the inverse propensity score was multiplied by the marginal frequency of the initiated study drug.19 The final stabilized inverse propensity scores were used to create the weighted study cohort.We evaluated baseline covariate balance across the weighted treatment groups using standardized mean differences (SMD), calculating SMDs for each pairwise comparison and assuming good balance if the maximum SMD across the pairwise comparisons did not exceed 0.1. We calculated the weighted median (IQR) follow-up times for each treatment arm. Cumulative incidence curves by medication were estimated using the inverse probability of treatment weight (IPTW) Kaplan-Meier method.The stabilized inverse propensity scores were incorporated as weights into proportional hazards models, one for each outcome, with all treatment groups. No additional variables were included in the outcomes models as treatment arms were fully balanced (SMD<0.1) on the measured baseline covariates. For each model, we tested the proportional hazards assumption using the Grambsch and Therneau test.20 A single overall omnibus test for the HRs all being equal to each other versus at least one different was conducted at the 0.05 significance level, and simulation was used to estimate 95% CIs. Cumulative incidence curves were estimated by first taking the pseudo values of the Kaplan-Meier estimate at each unique event time and then applying propensity scores weighted averages of the pseudo values.21 22 Isotonic regression was then applied to enforce monotonicity of the cumulative incidence curve.23 Finally, to better interpret the magnitude of the effects, we used these models to estimate the cumulative incidence of each outcome after 1 year, 2 years, and 3 years of treatment. Patients were censored on death (when not included in the outcome), disenrollment, or end of the study period, whichever came first.Sensitivity analyses Two sensitivity analyses were conducted using an as-treated approach, where patients were censored on: (1) discontinuation of the assigned treatment (ie, a gap more than 30 days after the dispensed pill count), death, disenrollment from the health plan, or end of the study period, whichever came first and (2) discontinuation of the assigned treatment, addition of another glucose-lowering drug, death, disenrollment from the health plan, or end of the study period, whichever came first.   Falsification endpoints Falsification endpoint analysis was performed to evaluate the sensitivity of treatment effects to potential unmeasured confounders. Hospitalizations for pneumonia and appendicitis were prespecified as falsification endpoints 24 25 (online supplemental table 4).Results We identified 296 563 adults with T2D, moderate cardiovascular risk, and without baseline evidence of stage 3 or greater CKD who initiated glimepiride (n=134 850), glipizide (n=145 719), or glyburide (n=15 994) between 2014 and 2021 ( online supplemental figure 1). Dose information for these drugs is shown in online supplemental table 6. Baseline characteristics of the unweighted sample are shown in online supplemental table 7. After IPTW, the study cohorts were balanced on all covariates with all SMD<0.10 (table 1 and online supplemental table 8) and included 295 092 patients: 134 926 starting glimepiride (mean age 66.8 (SD 7.9), 76.5% non-Hispanic white, 52.5% male), 145 984 patients starting glipizide (mean age 66.8 (SD 7.9), 76.5% non-Hispanic white, 52.5% male), and 14 182 patients starting glyburide (mean age 66.6 (SD 7.9), 78.0% non-Hispanic white, 52.9% male). Concomitant use of other glucose-lowering medications was also balanced among each group, with approximately 77% on metformin, 5% on thiazolidinediones, 3% on GLP-1RAs, 5% on SGLT2is, and 15% on DPP4is. Median IPTW follow-up times were 1209 days (IQR 1203–1218) for the glimepiride group, 1187 days (IQR 1180–1194) for the glipizide group, and 1149 days (IQR 1122–1175) for the glyburide group.Table 1Baseline patient characteristics, by treatment arm, after inverse probability of treatment weighting. Additional baseline characteristics are presented in online supplemental table 8Glimepiride(N=134 926.0)Glipizide(N=145 984.0)Glyburide(N=14 182.1)Largest SMDAge, years, mean (SD)66.8 (7.9)66.8 (7.9)66.6 (7.9)0.02Age group, N (%)0.05 <45 years922.2 (0.7)1059.8 (0.7)67.3 (0.5) 45–49 years3993.8 (3.0)4353.4 (3.0)430.0 (3.0) 50–54 years8732.2 (6.5)9376.1 (6.4)968.5 (6.8) 55–59 years11 584.6 (8.6)12 544.1 (8.6)1313.9 (9.3) 60–64 years11 862.6 (8.8)12 846.9 (8.8)1330.5 (9.4) 65–69 years41 282.4 (30.6)44 983.4 (30.8)4290.3 (30.3) 70–74 years39 382.9 (29.2)42 273.2 (29.0)3951.9 (27.9) 75–79 years17 062.9 (12.6)18 433.1 (12.6)1815.9 (12.8) ≥80 years102.3 (0.1)114.2 (0.1)13.8 (0.1)Sex, N (%)0.01 Male70 838.6 (52.5)76 693.3 (52.5)7495.6 (52.9) Female64 087.4 (47.5)69 290.7 (47.5)6686.5 (47.1)Race/ethnicity, N (%) *0.04 Asian3991.0 (3.0)4228.5 (2.9)376.5 (2.7) Black12 779.2 (9.5)13 901.8 (9.5)1241.2 (8.8) Hispanic10 996.7 (8.2)11 936.8 (8.2)1113.0 (7.8) White103 195.5 (76.5)111 616.0 (76.5)11 057.0 (78.0) Other or unknown3963.6 (2.9)4300.9 (2.9)394.4 (2.8)Prescriber specialty, N (%)0.03 Cardiology151.7 (0.1)170.7 (0.1)23.7 (0.2) Endocrinology1837.4 (1.4)1852.0 (1.3)146.2 (1.0) Nephrology73.8 (0.1)94.7 (0.1)<11 (0.1) Primary care†97 361.6 (72.2)105 359.0 (72.2)10 278.1 (72.5) Other23 785.7 (17.6)25 780.0 (17.7)2508.6 (17.7) Unknown11 715.8 (8.7)12 727.7 (8.7)1214.9 (8.6)Region, N (%)0.02 Midwest37 782.7 (28.0)40 679.4 (27.9)4034.4 (28.4) Northeast18 402.9 (13.6)19 939.7 (13.7)1903.1 (13.4) South61 027.3 (45.2)66 040.2 (45.2)6353.7 (44.8) West17 482.3 (13.0)19 041.0 (13.0)1868.5 (13.2) Unknown230.8 (0.2)283.7 (0.2)22.4 (0.2)Index year, N (%)0.04 201418 430.9 (13.7)19 835.5 (13.6)2089.0 (14.7) 201518 029.4 (13.4)19 447.0 (13.3)1954.9 (13.8) 201618 662.6 (13.8)20 115.3 (13.8)1955.2 (13.8) 201719 824.0 (14.7)21 470.2 (14.7)2086.6 (14.7) 201819 228.9 (14.3)20 829.5 (14.3)1972.5 (13.9) 201918 568.4 (13.8)20 188.3 (13.8)1928.4 (13.6) 202016 524.0 (12.2)17 916.7 (12.3)1655.4 (11.7) 20215657.9 (4.2)6181.5 (4.2)540.1 (3.8)Data source, N (%)0.05 OLDW49 092.52 (36.4)53 062.86 (36.3)4843.7 (34.2) Medicare fee-for-service85 833.5 (63.6)92 921.1 (63.7)9338.4 (65.8)All numbers are weighted. None of the covariates had maximum SMD between groups exceeding 0.10, the prespecified threshold for a significant difference between groups.*Race is classified in the Optum Labs Data Warehouse database as Asian, Hispanic, non-Hispanic black (black), and non-Hispanic white (white). ‘Other’ is a race and ethnicity choice in the OptumLabs Data Warehouse database, with no additional information available.†Primary care includes internal medicine and family medicine.OLDW, OptumLabs Data Warehouse; SMD, standardized mean difference.The predicted probabilities of each outcome at 1 year, 2 years, and 3 years following initiation of sulfonylurea therapy are shown in table 2. The primary kidney composite outcome was experienced by 2.1%, 2.2%, and 1.8% of patients after 1 year following initiation of glimepiride, glipizide, and glyburide, respectively, increasing to 5.0%, 5.2%, and 4.2% after 3 years, with overall incidence rates of 21.40, 22.69, and 16.43 per 1000 person-years. The proportions experiencing the secondary composite kidney outcome (which additionally included all-cause death) were only slightly higher (3.3–3.7%) after 1 year of treatment but increased considerably after 3 years of treatment: glimepiride: 10.0%, glipizide: 10.5%, and glyburide: 9.5%. Glyburide-treated patients experienced the highest rate of emergency department (ED) visits/hospitalizations for hypoglycemia: 0.4% 1 year after treatment initiation (vs 0.3% of glipizide-treated or glimepiride-treated patients) and 1.1% 3 years after treatment initiation (vs 0.7% of glipizide-treated patients and 0.9% of glimepiride-treated patients). Absolute differences in predicted probabilities at 1 year, 2 years, and 3 years of treatment were modest, as shown in online supplemental figure 2.Table 2Predicted probabilities and event rates of kidney outcomes following sulfonylurea initiation by duration of treatmentPrimary composite*Secondary composite†Stage 3–4 CKDStage 5 CKD/kidney failureDeathSevere hypoglycemia1 year Glimepiride2.1%3.5%1.9%0.3%1.1%0.3%Glipizide2.2%3.7%2.0%0.3%1.1%0.3% Glyburide1.8%3.3%1.5%0.3%1.1%0.4%2 years Glimepiride3.6%6.7%3.2%0.6%2.7%0.6%Glipizide3.7%7.0%3.3%0.7%2.9%0.5% Glyburide3.0%6.4%2.6%0.7%2.9%0.7%3 years Glimepiride5.0%10.0%4.4%1.0%4.6%0.9%Glipizide5.2%10.5%4.6%1.1%4.8%0.7% Glyburide4.2%9.5%3.6%1.0%4.9%1.1%Overall predicted event rates per 1000 person-years Glimepiride21.4037.2418.086.0119.794.76Glipizide22.6938.9919.366.4120.343.85 Glyburide16.4332.4713.016.3719.645.63*Defined as incident stage 3–5 CKD and kidney failure, including kidney replacement therapy.†Defined as incident stage 3–5 CKD, kidney failure (including kidney replacement therapy), or death.CKD, chronic kidney disease.The IPTW cumulative incidence curves for each study outcome are shown in figure 1. When compared with glimepiride, glipizide was associated with a modestly higher risk of all study outcomes: primary kidney composite (HR 1.04, 95% CI 1.00 to 1.07), secondary kidney composite (HR 1.05, 95% CI 1.03 to 1.07), stage 3–4 CKD (HR 1.04, 95% CI 1.01 to 1.07), stage 5 CKD/kidney failure (HR 1.09, 95% CI 1.02 to 1.16), and all-cause mortality (HR 1.06, 95% CI 1.03 to 1.09) (figure 2). Conversely, glyburide was associated with lower risks of the primary kidney composite outcome (HR 0.84, 95% CI 0.76 to 0.92) and incident stage 3–4 CKD (HR 0.80, 95% CI 0.72 to 0.89) compared with glimepiride and a lower risk of the primary kidney composite (HR 0.81, 95% CI 0.73 to 0.89), secondary kidney composite (HR 0.90, 95% CI 0.85 to 0.96) and stage 3–4 CKD (HR 0.77, 95% CI 0.70 to 0.86) compared with glipizide. Glyburide was also associated with a higher risk of severe hypoglycemia: HR 1.47 (95% CI 1.27 to 1.71) versus glipizide, and HR 1.22 (95% CI 1.05 to 1.42) versus glimepiride.Figure 1Cumulative incidence of kidney outcomes with sulfonylurea therapies after inverse probability of treatment weighting. Figures truncated at 72 months due to numbers remaining at risk. The primary composite outcome was defined as incident stage 3–5 CKD, kidney failure, or need for kidney replacement therapy. The secondary composite outcome was defined as incident stage 3–5 CKD, kidney failure, need for kidney replacement therapy, or death. CKD, chronic kidney disease.Figure 2Association between each sulfonylurea comparison and kidney outcomes: intention-to-treat analysis. The primary composite outcome was defined as incident stage 3–5 CKD, kidney failure, or need for kidney replacement therapy. The secondary composite outcome was defined as incident stage 3–5 CKD, kidney failure, need for kidney replacement therapy, or death. CKD, chronic kidney disease.Neither sensitivity analysis detected differences in any study outcome when comparing glipizide versus glimepiride (online supplemental tables 9 and 10). In contrast, when comparing glyburide versus glimepiride and glyburide versus glipizide, sensitivity analyses were consistent with main (ITT) analysis but with a higher magnitude of treatment effects. Additionally, glyburide showed a reduced risk of the secondary kidney composite outcome compared with glimepiride (HR 0.84, 95% CI 0.74 to 0.96 in the first sensitivity analysis; HR 0.81, 95% CI 0.69 to 0.95 in the second sensitivity analysis).There was no difference between the study drugs for the falsification endpoints of hospitalizations for pneumonia or appendicitis (online supplemental table 11).Discussion In this large target trial emulation that examined kidney outcomes following initiation of sulfonylurea therapy by adults with T2D at moderate cardiovascular risk, we found that glipizide was associated with the highest risk of incident moderate to severe kidney disease compared with glimepiride and glyburide. One year after treatment initiation, stage 3 or worse CKD developed in 2.1% of patients in the glimepiride group, 2.2% in the glipizide group, and 1.8% in the glyburide group. By 3 years, rates of stage 3 or worse CKD reached 5.0%, 5.2%, and 4.2% with glimepiride, glipizide, and glyburide use, respectively. Glipizide was associated with very modestly (4–9%) increased risk of all study outcomes (primary kidney composite, secondary kidney composite, incident stage 3–4 CKD, incident kidney failure, and all-cause mortality) compared with glimepiride. In contrast, glyburide was associated with lower risks of the primary kidney composite outcome (16%) and incident CKD stages 3–4 (20%) than glimepiride and lower risks of the primary kidney composite outcome (19%), secondary kidney composite outcome (10%), and incident CKD stages 3–4 (23%) compared with glipizide. However, glyburide was associated with significantly higher risk of hypoglycemia requiring ED or hospital care (22% higher than glimepiride and 47% higher than glipizide). As such, glimepiride may be the safest sulfonylurea drug in adults with T2D at moderate cardiovascular risk.This is the first study to compare head-to-head the three second-generation sulfonylureas that are most often used in clinical practice in the USA. Historically, there had been a lack of robust data surrounding the kidney safety of these agents, particularly as they are used in contemporary practice (ie, as both monotherapy and as an add-on to metformin and other medications). Furthermore, although glipizide, glimepiride, and glyburide have been available for decades, they have not undergone rigorous postmarketing outcome trials to which newer glucose-lowering agents have been subjected. Overall, our findings suggest that when sulfonylureas must be used for glucose-lowering in T2D, clinicians may want to avoid glipizide and prioritize glimepiride, particularly when prioritizing addressing kidney disease complications. Importantly, cardiovascular disease is the leading cause of death among people with diabetes26 and those with CKD,27 and glipizide is also associated with a higher risk of MACE compared with both glimepiride14 28 and glyburide.14 Additionally, glyburide is associated with much higher risk of severe hypoglycemia than either glimepiride or glipizide.14 This supports our finding that glimepiride is likely the safest sulfonylurea drug to use in most contexts.Glyburide and glimepiride are both long-acting sulfonylureas with extrapancreatic effects,29 30 though glyburide has much stronger affinity for SUR2B (with greater expected effects on renal mesangial cells) and SUR2A (impacting cardiac muscle cells),13 and is associated with higher risk of hypoglycemia.31 Somewhat surprisingly, therefore, we found that glyburide was associated with lower risk of adverse kidney outcomes (the primary composite kidney outcome and incident stage 3–4 CKD) compared with glimepiride. The mechanism for this is unclear. It is feasible that glyburide’s SUR2B blockade may reduce mesangial cell contraction in ways that protect glomerular structure, modulate glomerular hemodynamics, and/or reduce intraglomerular pressure, though these were not examined or observed in prior work. Glyburide may also have anti-inflammatory effects vis-à-vis NLRP3 inflammasome inhibition.32 While glyburide is generally not used in individuals with CKD due to its high risk for hypoglycemia and renal clearance, our study focused on individuals with preserved kidney function at baseline who did not have CKD stage 3 or higher. Exclusion of patients with established CKD makes confounding by indication (ie, preferential prescribing of glyburide to patients without CKD and thus at lower risk for hypoglycemia) less likely.Importantly, our results need to be contextualized in the broader landscape of T2D pharmacotherapy. Sulfonylureas remain the least preferred glucose-lowering treatment option compared to DPP4 inhibitors, GLP-1 receptor agonists, and SGLT2 inhibitors due to the higher risks of adverse cardiovascular,33 kidney,10 eye,34 and lower extremity35 complications. All sulfonylureas carry high risk of hypoglycemia14 - including glimepiride, though its risk of hypoglycemia is lower than the other sulfonylureas - and their use should be avoided in individuals at high hypoglycemia risk. Thus, their use should remain limited to situations where other medications cannot be used safely, are inaccessible without undue burden to patients, or are insufficient to maintain individualized glycemic goals.Our study’s strength includes access to longitudinal clinical data from a large, diverse US population with individuals enrolled in multiple private, Medicare Advantage, and traditional Medicare health plans. We were able to examine the real-world use of sulfonylureas both as monotherapy and as adjuvant treatment to other glucose-lowering medications. The target trial emulation framework facilitated adherence to a prespecified analytic approach, while use of rigorous causal inference methods and inclusion of falsification endpoints sought to minimize risk of residual confounding that is always a concern in observational study designs.While our study benefits from rigorous causal inference methodology, we acknowledge that these findings are hypothesis-generating and subject to the inherent limitations of observational research. Residual confounding from unmeasured patient or provider factors cannot be completely excluded despite robust IPTW. However, in the absence of head-to-head randomized trials comparing these sulfonylurea agents, which are unlikely to be conducted given their generic status and non-preferred placement in the T2D treatment algorithm, our findings provide important real-world evidence to inform clinical decision-making.Another important limitation of our study was that diagnosis codes were used to ascertain kidney outcomes, as laboratory data are unavailable in claims. Mild CKD is undercoded in administrative data, so lack of detection of stages 1 and 2 CKD and the inability to assess the gradual decline of kidney function are important potential confounders.36 We also could not assess declining kidney function within the broad range of estimated glomerular filtration rate that comprise stages 3–4 CKD. This may lead to lower range of estimates of differences in treatment effects. Claims data do not contain other certain clinical variables, such as hemoglobin A1c levels, estimated glomerular filtration rate, urine albumin/creatinine ratio, weight, and blood pressure, which may have impacted our observed outcomes. Glyburide was also infrequently used in our cohort, consistent with national prescribing patterns,37 raising concern about potential selection bias and confounding by indication for glyburide comparisons. Glyburide’s small sample size also resulted in less effective power to detect differences between glyburide and the other treatment groups; this is reflected in the wider CIs around glyburide effects. We were unable to specifically analyze mortality resulting from kidney disease, as cause of death was not available in our datasets. Our data sources also did not capture Medicaid beneficiaries or individuals without insurance, nor were we able to capture medications obtained outside of insurance coverage. Lastly, gliclazide, another commonly used second-generation sulfonylurea worldwide, is not available in the USA, and thus, we were unable to include it in our drug comparisons. This limits the clinical impact of our results for international practice settings, though still provides important evidence on the three sulfonylurea agents examined here. We also did not differentiate between short-acting and extended-release glipizide.In summary, the results from our target trial emulation showed that glyburide was associated with modestly lower risk of kidney complications but much higher risk of hypoglycemia, while glipizide was associated with higher risk of kidney complications among adults with T2D and moderate cardiovascular risk. Thus, glimepiride may be the preferred sulfonylurea in patients with T2D when the use of sulfonylureas cannot be avoided. Given the limited data to date that have examined within-class differences of individual sulfonylurea agents, this study offers new, practical guidance for clinicians in their decision-making for glucose-lowering therapy.",
  "title": "Comparative effectiveness of sulfonylureas on kidney outcomes in adults with type 2 diabetes and moderate cardiovascular risk: a target trial emulation",
  "uid": "7cc07efa-b081-5baa-96e3-df004d01142d"
}