JOSLINE 1C.pdf

EMERGING THERAP...

EMERGING THERAPIES: DRUGS AND REGIMENS Diabetes Care Volume 45, April 2022 919 Comparative Safety of Ghadeer K. Dawwas,1,2,3 James H. Flory,1,4 Sean Hennessy,1,2,3,5 Sodium–Glucose Cotransporter 2 Charles E. Leonard,1,2,3 and James D. Lewis1,2,3,6 Inhibitors Versus Dipeptidyl Peptidase 4 Inhibitors and Downloaded from http://diabetesjournals.org/care/article-pdf/45/4/919/672227/dc212177.pdf by Aishwarya Thanekar on 07 September 2024 Sulfonylureas on the Risk of Diabetic Ketoacidosis Diabetes Care 2022;45:919–927 | https://doi.org/10.2337/dc21-2177 OBJECTIVE To assess the association of sodium–glucose cotransporter 2 (SGLT2) inhibitors with diabetic ketoacidosis compared with dipeptidyl peptidase 4 (DPP-4) inhibi- tors and sulfonylureas in patients with type 2 diabetes. RESEARCH DESIGN AND METHODS We conducted a new-user active comparator cohort study to examine two pairwise 1 Center for Pharmacoepidemiology Research comparisons: 1) SGLT2 inhibitors versus DPP-4 inhibitors and 2) SGLT2 inhibitors versus and Training, Center for Clinical Epidemiology sulfonylureas. The main outcome was diabetic ketoacidosis present on hospital admis- and Biostatistics, Perelman School of Medicine, sion. We adjusted for confounders through propensity score matching. We used Cox University of Pennsylvania, Philadelphia, PA 2 proportional hazards regression with a robust variance estimator to estimate hazard Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, ratios (HRs) and corresponding 95% CIs while adjusting for calendar time. University of Pennsylvania, Philadelphia, PA 3 Leonard Davis Institute of Health Economics, RESULTS University of Pennsylvania, Philadelphia, PA In cohort 1 (n 5 85,125 for SGLT2 inhibitors and n 5 85,125 for DPP-4 inhibitors), 4 Endocrinology Service, Department of Subspecialty the incidence rates of diabetic ketoacidosis per 1,000 person-years were 6.0 and Medicine, Memorial Sloan Kettering Cancer Center, New York, NY 4.3 for SGLT2 inhibitors and DPP4 inhibitors, respectively. In cohort 2 (n 5 72,436 5 Department of Systems Pharmacology and for SGLT2 inhibitors and n 5 72,436 for sulfonylureas), the incidence rates of dia- Translational Therapeutics, Perelman School of betic ketoacidosis per 1,000 person-years were 6.3 and 4.5 for SGLT2 inhibitors Medicine, University of Pennsylvania, Philadelphia, and sulfonylureas, respectively. In Cox proportional hazards regression models, PA 6 Division of Gastroenterology and Hepatology, the use of SGLT2 inhibitors was associated with a higher rate of diabetic ketoaci- Perelman School of Medicine, University of dosis compared with DPP-4 inhibitors (adjusted HR [aHR] 1.63; 95% CI 1.36, 1.96) Pennsylvania, Philadelphia, PA and sulfonylureas (aHR 1.56; 95% CI 1.30, 1.87). Corresponding author: Ghadeer K. Dawwas, [email protected] CONCLUSIONS Received 19 October 2021 and accepted 18 In this comparative safety study using real-world data, patients with type 2 dia- January 2022 betes who were newly prescribed SGLT2 inhibitors had a higher rate of diabetic This article contains supplementary material online ketoacidosis compared with DPP-4 inhibitors and sulfonylureas. Clinicians should at https://doi.org/10.2337/figshare.18834131. be vigilant about this association. © 2022 by the American Diabetes Association. Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. Type 2 diabetes affects 34 million individuals in the U.S. (1). The use of antidiabe- More information is available at https://www. tes drugs can prevent or delay the development of macrovascular and microvascular diabetesjournals.org/journals/pages/license. 920 SGLT2 Inhibitor Safety in Diabetic Ketoacidosis Diabetes Care Volume 45, April 2022 complications (2). Evidence from ran- the risk of diabetic ketoacidosis with (i.e., chlorpropamide, tolbutamide, tolaza- domized clinical trials (RCTs) showed that SGLT2 inhibitors compared with DPP-4 mide, glimepiride, glipizide, glyburide), sodium–glucose cotransporter 2 (SGLT2) inhibitors and sulfonylureas using a large who did not have prior use during the inhibitors (vs. placebo) exhibit cardiopro- U.S.-based health insurance database that lookback period. Patients were considered tective (e.g., reduction in hospitalization included sociodemographic data, mortal- exposed if they filled at least one prescrip- due to heart failure) and metabolic bene- ity records, and laboratory measures for tion of one of the study drugs. We fits (e.g., reduction in body weight) (2,3). a subset of the cohort. selected DPP-4 inhibitors and sulfonylureas Observational analyses evaluating SGLT2 as the active comparators since they are inhibitors compared with other com- RESEARCH DESIGN AND METHODS both commonly used second-line alterna- monly used antidiabetes drugs (e.g., Database tives to SGLT2 inhibitors. We examined dipeptidyl peptidase 4 [DPP-4] inhibitors We used Optum’s deidentified Clinfor- the within-class effect of the individual and sulfonylureas) reported findings con- matics Data Mart Database. The data SGLT2 inhibitors, including dapagliflozin, Downloaded from http://diabetesjournals.org/care/article-pdf/45/4/919/672227/dc212177.pdf by Aishwarya Thanekar on 07 September 2024 sistent with these (4,5). However, con- include privately insured individuals in canagliflozin, empagliflozin, compared with cerns about the association between the U.S. and provide information on DPP-4 inhibitors and sulfonylureas. For the these agents and risk of diabetic ketoaci- enrollment, patient demographics, outpa- latter analysis, the initiation of an SGLT2 dosis remain unresolved (6). inhibitor different from the one at cohort tient claims, inpatient claims, and pre- In 2015, the U.S. Food and Drug entry led to censoring (i.e., termination of scription drug claims. Laboratory data Administration (FDA) issued a safety follow-up). were available for a subset of beneficia- warning that use of SGLT2 inhibitors, ries. At the University of Pennsylvania, including canagliflozin, dapagliflozin, and Outcome Ascertainment studies using the Optum Clinformatics empagliflozin, may lead to diabetic ketoa- The primary outcome measure was dia- Data Mart Database are categorized as cidosis, a serious condition where the betic ketoacidosis, defined on the basis exempt from requiring institutional review body produces high levels of ketones, of the presence of diagnosis codes on board approval. which often leads to hospitalization (6). hospital admission (principal position) The FDA later added new warnings to Study Design (ICD-9 code 250.1 and ICD-10 code E1x.1) the labels of all SGLT2 inhibitors and We performed a retrospective new-user (15). This algorithm was validated in the required manufacturers of SGLT2 inhibi- active comparator cohort study of pa- inpatient setting and had a positive pre- tors to conduct postmarketing assess- tients with type 2 diabetes who had one dictive value of 88.9% (95% CI 71.9, 96.1) ment of this safety concern (7). A post or more prescriptions dispensed for (15). hoc analysis of an RCT suggested an increased rate of diabetic ketoacidosis, SGLT2 inhibitors, DPP-4 inhibitors, or sul- fonylureas between January 2013 and Follow-up although not statistically significant, with December 2019. We included patients Eligible patients were followed from the canagliflozin compared with placebo (0.6 vs. 0.3 events per 1,000 person-years; who were 1) aged >18 years at cohort cohort entry date (i.e., treatment initia- entry, 2) new users of SGLT2 inhibitors or tion) until the occurrence of the first of hazard ratio [HR] 2.33; 95% CI 0.76, 7.17) DPP-4 inhibitors in cohort 1 and SGLT2 the following: development of diabetic (8). In the Canagliflozin and Renal Events in Diabetes and Nephropathy Clinical inhibitors or sulfonylureas in cohort 2, 3) ketoacidosis, treatment discontinuation Evaluation (CREDENCE) trial, the rate of had 12 months of continued enrollment defined by the presence of a gap >30 diabetic ketoacidosis was significantly before their first prescription (i.e., look- days between consecutive refills, end of higher in the canagliflozin versus placebo back period), and 4) had a diagnosis of enrollment, initiation of study compara- group (2.2 vs. 0.2 per 1,000 patient- type 2 diabetes during the lookback tor, death, or end of study period (30 years; HR 10.80; 95% CI 1.39, 83.65) (9). period (ICD-9 codes 250.x0 or 250.x2 and June 2019). Similarly, in the Dapagliflozin Effect on ICD-10 code E11.x) (14) (Supplementary Cardiovascular Events–Thrombolysis in Fig. 1). We assigned the date of the first Adjustment for Confounding Myocardial Infarction 58 (DECLARE-TIMI eligible prescription as the cohort entry We ascertained the following confound- 58) trial, diabetic ketoacidosis was more date. We excluded patients who as of the ers during the lookback period: 1) dem- common with dapagliflozin versus pla- index date had end-stage renal disease, ographics (i.e., age, sex, education level cebo (0.3% vs. 0.1%, P 5 0.02) (10). type 1 diabetes, diabetic ketoacidosis (to [census block level], race, income [cen- Observational studies that reported con- minimize the capture of prevalent events), sus block level], geographic location); 2) flicting findings adjusted for a small num- initiated SGLT2 inhibitors or comparators comorbidities (e.g., heart failure, periph- ber of variables (11) or included insulin in on the same date, or used combined eral vascular disease [PVD], hyperten- the comparator group (12). It is worth products that constituted SGLT2 inhibitors sion, cancer, stroke); 3) medications (e.g., noting that the absolute increase in the and DPP-4 inhibitors (in cohort 1 only). ACE inhibitors, b-blockers, diuretics); 4) incidence rate of diabetic ketoacidosis use of other antidiabetes drugs, excluding was relatively small (crude rate difference Exposure Ascertainment the study drugs (e.g., insulin, metformin, 1.2 per 1,000 person-years) (13). We identified new users of the study glucagon-like peptide 1 receptor agonist); Given the limited evidence, compara- drugs, including SGLT2 inhibitors (i.e., 5) Diabetes Complications Severity Index tive safety studies are needed to guide dapagliflozin, canagliflozin, empagliflozin), (DCSI) (16); and 6) measures of intensity treatment selection in clinical practice. DPP-4 inhibitors (i.e., saxagliptin, linagliptin, of health care utilization (e.g., total hospi- In this study, we aimed to investigate sitagliptin, alogliptin), and sulfonylureas talizations during the lookback period). diabetesjournals.org/care Dawwas and Associates 921 Statistical Analysis cholesterol, HbA1c, AST, ALT, and triglyc- SGLT2 Inhibitors Versus Sulfonylureas We used propensity score (PS) matching eride levels. In subgroup analyses using We identified 84,583 new users of (PSM) to adjust for confounding. We Cox proportional hazards regression mod- SGLT2 inhibitors and 296,947 new users calculated PSs through a logistic regres- els, we examined whether differential of sulfonylureas who met inclusion crite- sion model (PROC LOGISTIC in SAS [SAS associations existed for SGLT2 inhibitors ria (Supplementary Fig. 2). Patient demo- Institute, Cary, NC) that included all versus comparators by age, sex, baseline graphic and clinical characteristics before covariates listed in Table 1. We did not insulin (including basal and prandial insu- PSM are summarized in Supplementary include laboratory values for measures Table 1. Supplementary Fig. 3B depicts lin), stroke, and PVD. To assess the poten- such as cholesterol, hemoglobin A1c the distribution of PSs before and after tial for effect modification, we included (HbA1c), AST, ALT, or triglyceride levels in PSM. After matching (n 5 72,436 for an interaction term between exposure the PS since the information was avail- SGLT2 inhibitors and n 5 72,436 for sul- status and indicators for each subgroup. able for a small subset of the cohort. No fonylureas), patient characteristics were To account for multiple testing in sub- Downloaded from http://diabetesjournals.org/care/article-pdf/45/4/919/672227/dc212177.pdf by Aishwarya Thanekar on 07 September 2024 other data were missing in our study. well-balanced between the two groups, The estimated PSs were used to match group analyses, we used Bonferroni including mean age (60 vs. 60 years), edu- patients in each comparison (SGLT2 adjustment, which considered results sta- cation (less than bachelor’s degree 53% inhibitors vs. DPP-4 inhibitors and SGLT2 tistically significant if the corresponding P vs. 53%), race (White 66% vs. 66%), inhibitors vs. sulfonylureas) using 1:1 value was # a (i.e., 0.05)/n where n 5 income ($$100,000 25% vs. 26%), heart fixed ratio matching to the nearest total number of subgroup analyses (18). failure (5% vs. 5%), hypertension (66% vs. neighbor on the basis of a maximum cal- We assessed for the potential of unmeas- 66%), chronic kidney disease (11% vs. iper width of 0.1 of the SD of the logit of ured confounders using the E-value, 11%), ACE inhibitor use (36% vs. 36%), PS. We examined the distribution of PS which provides, conditional on mea- b-blocker use (21% vs. 21%), insulin use values (PROC SGPLOT in SAS) and sured confounders, the minimum needed (20% vs. 20%), and mean inpatient visits assessed the balance before and after PS strength of association among an unmeas- (0.1 vs. 0.1) (Table 1). The median fol- using standardized mean differences. We ured confounder, exposure, and outcome low-up time was 138 days (25th–75th considered variables balanced between to move the observed effect estimates percentile 57–333 days) among SGLT2 treatment groups if they met a threshold toward the null (https://www.evalue- inhibitor users and 128 days (56–322