Risk of acute pancreatitis among new users of empagliflozin compared to sulfonylureas in patients with type 2 diabetes: A post‐authorization safety study

This study was undertaken to evaluate the potential risk of acute pancreatitis with empagliflozin in patients with type 2 diabetes (T2D) newly initiating empagliflozin.

• In patients with T2D on metformin, those newly initiating empagliflozin did not have an increased risk of acute pancreatitis compared with patients initiating a sulfonylurea.
• This voluntary post-approval safety study supports existing evidence that the use of empagliflozin for the treatment of T2D is not associated with an increased risk of acute pancreatitis.

Plain Language Summary
Some medicines for people with type 2 diabetes (T2D) have been linked with a very small chance of developing inflammation of the pancreas (acute pancreatitis).Such medicines include sodium-glucose cotransporter 2 (SGLT2) inhibitors.This study was done to find out whether people taking a type of SGLT2 inhibitor called empagliflozin might have developed acute pancreatitis.The study looked at anonymous information about people in the United States who made claims for healthcare costs.In the study, information was collected about people who had newly started empagliflozin treatment for T2D.This was compared with information about people who were newly taking another type of treatment for T2D (a sulfonylurea).Sulfonylureas were chosen for comparison because this drug has not been linked with developing acute pancreatitis.All people in the study were also taking metformin for T2D.The researchers looked at data from 72 621 new users of empagliflozin compared with the same number of new users of sulfonylureas, matched on their characteristics.The average time that people had been taking their new treatment was 6 months.The results showed that empagliflozin for the treatment of T2D does not lead to an increase in the risk of acute pancreatitis in people taking this medicine.

| INTRODUCTION
3][4] The safety profile of empagliflozin is supported by the findings of a recent pooled analysis of 20 randomized clinical trials and four extension studies of patients with T2D.The pooled analysis showed that empagliflozin, at doses of 10 or 25 mg, was well tolerated in patients with T2D and was not associated with a higher rate of confirmed hypoglycemia versus placebo, except in patients receiving concurrent insulin and/or a sulfonylurea (SU). 5As a consequence of their mechanism of action, SGLT2 inhibitors are associated with transient increases in urinary volume, with a potential risk of volume depletion and hypotension. 6However, for empagliflozin, the risk of volume depletion appears to be similar to placebo, except in patients aged 75 years and over or individuals receiving loop diuretics. 5The SGLT2 inhibitor mechanism of action is also associated with a risk of urinary tract infections and genital mycotic infections as a consequence of glycosuria. 7][10] These include urinary and genital infection, diabetic ketoacidosis, bone fracture, cancers, and foot and toe amputation.1][12][13] A meta-analysis and systematic review of 109 studies of SGLT2 inhibitors (dapagliflozin, canagliflozin, empagliflozin, ipragliflozin) showed no increased risk of acute kidney injury, diabetic ketoacidosis, urinary tract infection (except dapagliflozin) or fracture with these agents versus placebo or active comparators. 35][16][17] Two incretin-based therapies, dipeptidyl-peptidase-4 (DPP-4) inhibitors and glucagon-like peptide-1 receptor agonists (GLP-1 RAs), have been linked with the development of acute pancreatitis, although overall drug-associated acute pancreatitis is relatively uncommon 15,18 and some studies using appropriate new user, active comparator designs to mitigate confounding, have shown no association between incretin-based therapies and acute pancreatitis. 14,16,17For SGLT2 inhibitors, spontaneous reports and case reports have suggested that these agents may be associated with a potential risk of acute pancreatitis, although it is unknown if reporting bias could explain these reports. 19,20Since acute pancreatitis is a relatively rare but potentially serious and occasionally fatal condition, its occurrence is an important consideration for drug safety studies. 21,22In 2016, the US Food and Drug Administration (FDA) identified acute pancreatitis as a potential new safety issue for the SGLT2 inhibitor class in response to data gathered by the Adverse Event Reporting System. 23Although the true prevalence of druginduced pancreatitis is difficult to identify because most data are from individual case reports, this event appears to occur rarely for SGLT2 inhibitors, including empagliflozin. 24,25When drug-induced acute pancreatitis events occur, they tend to occur within 1-6 weeks of drug initiation; 26 however, although these results reflect data across 120 different medications, none of the included drugs were SGLT2 inhibitors.
The mechanism by which SGLT2 inhibitors might cause pancreatitis remains unknown, and it can be difficult to rule out other more common causes of acute pancreatitis among reported cases.Common causes of acute pancreatitis include concomitant medications (500 drugs have been linked with pancreatitis, with at least 30 having a definite association), 21,22 cholelithiasis, alcohol use, and hypertriglyceridemia. 20,21,27 In addition, the presence of T2D 28,29 and/or established cardiovascular (CV) disease 30 increases the risk of acute pancreatitis compared with individuals without these conditions.Potential mechanisms of action for drugs known to cause acute pancreatitis include pancreatic duct constriction, metabolic effects, cytotoxic effects, accumulation of a toxic metabolite or intermediary, idiosyncratic reactions, and hypersensitivity reactions. 18,31,32 further elucidate the potential risk of acute pancreatitis with empagliflozin, the present study was voluntarily undertaken to evaluate the occurrence of this event in patients with T2D newly initiating empagliflozin versus SU on a background of metformin (i.e., concurrently using metformin at the time of initiation of empagliflozin or SU).This comparator was chosen because, unlike SGLT2 inhibitors and incretin-based therapies such as DPP-4 inhibitors and GLP-1RAs, SUs have not been linked with an increased risk of acute pancreatitis.

| Study design
This was a non-interventional study using existing data from two US claims databases with a new-user, active-comparator design.
Therapies for the comparator groups were selected based on clinical guideline recommendations for the management of T2D and current real-world treatment patterns of oral glucose-lowering agents.Since first-line pharmacotherapy in the US and most countries worldwide is generally metformin, 33 with SGLT2 inhibitors generally being used as second-or third-line therapy, a background of metformin was reflected in the patient selection.Previous studies of the safety of empagliflozin have included DPP-4 inhibitors as the comparator group because these agents were relatively new to the market, compared with other oral agents such as metformin and SUs.Because DPP-4 inhibitors and GLP-1 RAs have, more recently, been linked with increased risk of acute pancreatitis, [15][16][17] these agents were not included in the comparator group.To determine the analysis cohorts for the study, an initial feasibility assessment was conducted in the Optum and MarketScan databases using data between August 1, 2014 and the latest data-cut available in MarketScan (September 30, 2020) and Optum (March 31, 2021).Proposed cohorts were: monotherapy consisting of empagliflozin versus metformin; dual therapy with empagliflozin versus SU both on a background of metformin; and triple therapy with empagliflozin or thiazolidinediones (TZDs) on a background of SU and metformin.The dual-therapy cohort reported the highest number of empagliflozin initiators and was the only adequately powered line of therapy, therefore, analyses of mono-and triple-therapy cohorts were not performed, and patients receiving SUs as add-on to metformin were included as the comparator group.

| Patients
For the main analysis, adults (aged ≥18 years) with T2D were selected for inclusion if they were already on metformin therapy and were initiating empagliflozin or an SU, between August 1, 2014 and the latest data-cut available in each database (MarketScan CCAE/MDCR: September 30, 2020; Optum CDM: March 31, 2021).Patients were excluded if, in the 6 months prior to study drug initiation, they had used an SGLT2 inhibitor/DPP-4 inhibitor/GLP-1 RA or the comparator (SU), or had been using insulin (≥60 days of continuous exposure) in the outpatient setting.Exclusion criteria also included claims for acute or chronic pancreatitis, pancreatic cancer, or other disease of the pancreas made any time prior to initiation of the study drug.
Patients were also excluded if they had type 1 diabetes (T1D), or gestational diabetes, or secondary diabetes in the 6 months prior to study drug initiation.No minimum duration of follow-up was required for inclusion in the cohort.

| Assessments and data analyses
Pooled analyses of data from the Optum CDM and MarketScan CCAE/MDCR databases were performed.These two US claims databases provide claims data from different sets of providers (Optum largely from United Healthcare; MarketScan from multiple providers).
To ensure an optimal sample size, the main analysis for this study was a meta-analysis of the dual therapy cohorts from the two databases.
To minimize potential confounding, patients initiating empagliflozin were matched 1:1 within the database to patients initiating SU using propensity scores (PS) that incorporated relevant demographic and clinical characteristics.Matching was performed using a nearestneighbor algorithm and calipers of width equal to 0.2 of the standard deviation of the logit of the estimated propensity score.To account for patient characteristics changing over time, patients were also matched on the calendar year of initiation of the study medication and a variable for calendar quarter-year was included in the propensity score model.To minimize misclassification, where possible, validated codes and algorithms were used to optimize accurate capture of conditions and events.Pre-specified sensitivity analyses were performed for design parameters with the potential for misclassification that could affect the study results (Tables S1-S7).
A modified version of an algorithm developed and validated in the Kaiser Permanente health system was used to identify acute pancreatitis events. 34The original algorithm was reported to have a sensitivity of 89.8%, a specificity of 79.4%, a positive predictive value of 90.0%, and a negative predictive value of 79.0%.In brief, the algorithm requires the presence of an acute pancreatitis diagnosis from any inpatient, outpatient, or emergency department diagnoses (ICD-9-CM 577.0 or ICD-10-CM K85), a lipase measure within ±7 days of the acute pancreatitis diagnosis (using LOINC code of 3040-3 and 2572-6 or CPT code of 83690 for lipase), and an abdominal ultrasound (using CPT codes of 76700 and 76705) within ±7 days of the acute pancreatitis diagnosis.Cox proportional-hazards regression models based on time-to-first acute pancreatitis event were used to estimate hazard ratios (HR) and 95% confidence intervals (CIs) for the dual-therapy cohort (in each database and pooled) in the unmatched and PS-matched cohorts within each database.The degree of overlap of patient data (i.e., duplicate patients) from the two databases used in this study was estimated to be 10%, and this was taken into account in the meta-analysis.The overlap was not expected to impact the results of this study because of the small size of overlap relative to the study population.To correct variance for potential duplicates after pooling the two databases for the primary outcome, the variance correction described by Gruber and Robins was utilized. 35The pooled analyses were conducted using R package rmeta in RStudio. 36To test the impact of overlapping patient data, the use of a 10% versus 30% overlap proportion was compared in the pooled meta-analysis, and the results remained consistent.Where possible, validated codes and algorithms were used to minimize misclassification of acute pancreatitis and to optimize accurate capture of conditions and events.Data were analyzed using an as-treated (AT) approach for the pri- Because this approach could result in the exclusion of some cases of pancreatitis due to censoring criteria (i.e., patients would be classified as not experiencing acute pancreatitis, although the event could have occurred after censoring), a sensitivity analysis was performed using an intention-to-treat (ITT) approach.This approach allows for the possibility of evaluating cases of acute pancreatitis occurring during the study period after discontinuation or switching that might have been missed with an AT analysis.

| RESULTS
Analysis of the MarketScan CCAE/MDCR and Optum CDM databases identified 72 661 new users of empagliflozin and 422 018 new users of SUs, with both patient groups on concurrent metformin therapy.

Cohort attrition for new users of empagliflozin and SU in MarketScan
CCAE/MDCR and Optum CDM databases is shown in Table S8.
Baseline characteristics within treatment groups appeared to be similar across the two databases.In the pooled unmatched cohort, mean age (standard deviation [SD]) was 57 (12) years in the empagliflozin group and 60 (13) years in the SU group, and 58% and 56% of patients, respectively, were male.Compared with patients initiating SU, empagliflozin initiators started the index drug more recently and a smaller proportion were treatment naïve.Baseline characteristics were generally similar for patients initiating empagliflozin and SU (Table 1).Comorbidity indices were similar between treatment groups; however, differences in individual comorbidities were observed.As expected, a larger proportion of empagliflozin users had obesity and obesity-related comorbidities of hypertension, hyperlipidemia, hypertriglyceridemia, and obstructive sleep apnea, in addition to hyperglycemia and a diagnosis code for diabetes mellitus.Empagliflozin users were also less likely to have Stage 3 or 4 chronic kidney disease or infections, as well as a lower mean lipase value.After 1:1 propensity score matching, the groups were well balanced across the 72 621 matched pairs.Mean age (SD) was 57 (12) years, and 58% were male.
Standardized differences for all variables were less than 10% with the exception of certain mean laboratory measures (i.e., creatinine, NT-proBNP, and lipase).This is likely due to the fact that laboratory measures were only available in a subset of patients and were not included in the propensity score model.Typical risk factors for acute pancreatitis were also balanced in the matched cohorts.
The mean follow-up period was 6 months for the matched cohort (185.86 days in the empagliflozin group, 194.22 days in the SU group).Distribution of first censored event in the follow-up period in both the unmatched and matched cohorts in each database is outlined in Table S9.The incidence rates of acute pancreatitis in the pooled unmatched cohorts were 10.33 (95% CI 9.32-11.42)and 14.78 (95% CI 14.32-15.24)events per 1000 patient-years (PY) for empagliflozin and SUs, respectively (Figure 1).Incidence rates of acute pancreatitis in the pooled matched cohort were 10.30 (95% CI 9.29-11.39)events per 1000 PY for empagliflozin and 11.65 (95% CI 10.59-12.77)events per 1000 PY for SUs.On a background of metformin, patients newly initiating empagliflozin did not have an increased risk of acute pancreatitis compared with those initiating an SU (pooled PS matched HR 0.88 [95% CI 0.76-1.02])across 75 621.42PY of follow-up.
Results of the propensity score matched pooled ITT sensitivity analysis were similar to the primary analysis and demonstrated a significantly decreased risk of acute pancreatitis for the empagliflozin group versus SU group (HR 0.84 [95% CI 0.73-0.97]).Furthermore, these results were robust across multiple sensitivity analyses (Tables S1-S7).

| DISCUSSION
The information pooled from the MarketScan CCAE/MDCR and Optum CDM databases in the present study provided a large data source from real-world patients.The analyses showed that for patients on a background of metformin therapy, new users of empagliflozin did not have an increased rate of acute pancreatitis compared with patients initiating SU.Since the mean period of follow-up extended to 6 months, the study is likely to have captured the majority of acute pancreatitis events occurring in the study population.This assumption is based on the findings of previous research that examined the latency periods between initiation of 120 different drugs and onset of pancreatitis, which showed that acute pancreatitis largely occurs within 1-6 weeks of drug initiation. 26The incidence of acute pancreatitis observed in the present study (10.3 cases per 1000 PY) is higher than that reported previously in patients with T2D (4.22 to 5.64 per 1000 PY), 29,37 and might reflect the growing prevalence of acute pancreatitis (estimated to be 3% per year). 38Furthermore, the present study included a larger proportion of older patients than the earlier studies, who may be at increased risk of developing pancreatitis.
Overall, the incidence of acute pancreatitis associated with SGLT2 inhibitors is unknown since few population-based studies have been performed, and available evidence is limited.A recent safety analysis of antidiabetes medications evaluated the risk of pancreatic events based on real-world safety data and systematic review and meta-analysis of randomized controlled trials. 39The findings showed that while evidence from post-marketing safety data indicates a strong association between incretin mimetics, including SGLT2 inhibitors and acute pancreatitis, few events were reported in randomized clinical trials.For empagliflozin, a pooled analysis of clinical trials of patients with T2D who received placebo, empagliflozin 10 mg or 25 mg showed the incidence of acute pancreatitis in these groups was 0.1, <0.1 and 0.1 per 100 PY, respectively. 4Based on more than 15 000 patient-years of empagliflozin exposure, this pooled analysis showed no evidence of an association between empagliflozin and acute pancreatitis.The findings of the present study are consistent with two recent analyses of the occurrence of acute pancreatitis with SGLT2 inhibitor use.A systematic literature review and meta-analysis of 35 randomized clinical trials involving 44 912 patients with T2D included 41 events of acute pancreatitis (19 trials; 32 932 patients). 24 the 19 trials, 10 assessed empagliflozin, 4 canagliflozin, 3 ertugliflozin, and 2 dapagliflozin.The analysis showed no significant increase in the risk of acute pancreatitis associated with the use of SGLT2 inhibitors versus control groups (receiving oral anti-diabetes therapies or insulin).A further study of three US claims databases evaluated the risk of acute pancreatitis in new users of canagliflozin compared with new users of six other classes of glucose-lowering agents. 40Compared with the present study, the PS matched analysis of the canagliflozin groups showed lower incidence rates for acute pancreatitis (1.5 to 2.2 per 1000 PY for canagliflozin in the on-treatment analysis versus 10.30 per 1000 PY with empagliflozin) in the present study, although this could partly reflect methodological differences between the studies and populations.Specifically, the definition of acute pancreatitis used in the present study was based on a validated algorithm selected to detect acute pancreatitis in all clinical settings rather than only at hospital discharge, resulting in more cases identified.In a sensitivity analysis of the present study, use of an alternate outcome definition of acute pancreatitis based only on the primary hospitalization discharge diagnosis produced results that were more comparable to FAZELI FARSANI ET AL.
T A B L E 1 Selected patient characteristics at baseline before and after PS matching (pooled data).the study with canagliflozin (Table S1).Furthermore, the present study was restricted to new users of empagliflozin or SU on a background of metformin, whereas the study of canagliflozin included users of other glucose-lowering therapies.However, consistent with the present study, the results also showed no evidence of an association of an increased risk of acute pancreatitis with canagliflozin compared with the other agents.
As with all non-interventional studies, the present study is associated with several methodological limitations.These include the possibility that patients did not remain on their original treatments during the study period in response to disease progression, adverse events with specific drugs, or other changes in clinical conditions.To account for this possibility, a sensitivity analysis was performed using an ITT approach in which users maintained their original treatment assignment.The results were consistent with the primary analysis and showed a significantly reduced risk of pancreatitis among patients in the empagliflozin group compared with the SU group.Another possible limitation is that the data source only reflects claims for medications and does not indicate how or whether the treatments were taken.Furthermore, the pooling of two claims databases introduces the potential for overlap between the two datasets and the possibility of duplicated data from individual patients.To minimize this outcome, a variance correction was used to correct for potential duplicates.
Within the study methodology, there is also a risk of confounding biasfor example, patients starting treatment with a newer drug might have more severe disease than those receiving treatment with an older drug.In the present study, the empagliflozin group tended to have more comorbidities than the comparator group, although after PS matching, comorbidity indices were generally similar between treatment groups.However, unknown or unmeasured factors cannot be accounted for.The design of this database study also has a potential risk of misclassification of the recorded patient characteristics, such as in the identification of cases of acute pancreatitis outcomes.
However, this risk was minimized by the use of previously validated algorithms and code-lists wherever possible.Finally, it should be noted that the findings might not be generalizable to all patients with T2D, especially to individuals on other treatment regimens that were not included in the analysis.
mary analysis.Study follow-up started the day after initiation of the index medication and continued until: (1) the first occurrence of the outcome (first date of acute pancreatitis event); (2) the date of end of continuous registration in the database; (3) the date of death; (4) the date of study end; (5) the date of discontinuation of index medication (defined as the date of the last date of last dispensing + days of supply + grace period of 30 days); (6) the date of switch from the index medication to another hypoglycemic agent; (7) the date of addition of incretin-based therapy (i.e., GLP-1 RA or DPP-4i; with continuation of index therapy); (8) the date of addition of SGLT2 inhibitor (with continuation of index therapy); (9) the date of addition of insulin (with continuation of index therapy); or(10) the date of addition of the within treatment line comparator index drug (with continuation of index therapy).Discontinuation of the index drug occurred when there was a gap of more than 30 days from the end of days of supply of the current prescription and the next prescription date.The discontinuation date was defined as 30 days after the end of days of supply for the last prescription date prior to the gap exceeding 30 days.The continuous exposure periods of the index drugs were created by linking all prescriptions with days of supply that did not have a gap greater than 30 days (i.e., the patient did not discontinue).Addition of a new medication occurred when the prescription date of a new medication was before the last prescription date of the index continuous exposure.The prescription date of the new medication was the addon date.Switch to a new medication occurred when the prescription date for the new medication was on or after the last prescription date of the index continuous exposure and before the end of discontinuation.The prescription date of the new medication is the switch date.

Nonetheless, despite the 1
potential methodological limitations of this type of study, the use of large claims data sources, such as the MarketScan CCAE/MDCR and Optum CDM databases, provides important data from the US population that can be used to evaluate treatment safety.Although people over 65 years are underrepresented in the MarketScan databases, the use of the Optum CDM data increases the proportion of older people in the analysis, thus enhancing the applicability of the data to the US population as a whole.0.28%; Q(df = 1) = 1.0028;P = 0.3166 *I 2 = 0%; Q(df = 1) = 0.0063; P = 0Incidence rate of acute pancreatitis in 1:1 propensity score, A, unmatched and B, matched patients with T2D receiving empagliflozin or SU therapy as add-on to metformin.CI, confidence interval; HR, hazard ratio; IR, incidence rate; PY, patient-years; SU, sulfonylurea; T2D, type 2 diabetes.The findings of this study can assist therapeutic decision-making when considering SGLT2 inhibitors as add-on therapy to metformin, providing additional evidence of the low risk of acute pancreatitis associated with empagliflozin therapy, based on real-world evidence across over 75 000 PY of follow-up.These findings are of clinical importance since patients with T2D are already at increased risk of acute pancreatitis and, therefore, the presence of other potential risk factors should be carefully evaluated.Overall, the findings of this study are consistent with prior evidence and support the wellestablished safety profile of empagliflozin.The results of this voluntary post-approval safety study provide additional evidence that the use of empagliflozin for the treatment of T2D is not associated with an increased risk of acute pancreatitis.