[Skip to Content]
[Skip to Content Landing]
Figure.
Derivation of the Sample
Derivation of the Sample

Flowchart of the sample. CLI indicates critical limb ischemia; DPP-4, dipeptidyl peptidase 4 inhibitors; GLP-1, glucagon-like peptide 1 agonists; PVD, peripheral vascular disease; SGLT-2, sodium-glucose cotransporter 2 inhibitors; and TZD, thiazolidinediones.

aNot mutually exclusive.

Table 1.  
Characteristics of the Study Samplea
Characteristics of the Study Samplea
Table 2.  
Incidence of Amputation and Other Vascular Outcomes Among New Users of SLGT-2 Inhibitors, DPP-4 Inhibitors, GLP-1 Agonists, or Other Oral Treatments for Type 2 Diabetes
Incidence of Amputation and Other Vascular Outcomes Among New Users of SLGT-2 Inhibitors, DPP-4 Inhibitors, GLP-1 Agonists, or Other Oral Treatments for Type 2 Diabetes
Table 3.  
Adjusted Association Between Use of SGLT-2 Inhibitors and Amputation and Other Vascular Outcomes
Adjusted Association Between Use of SGLT-2 Inhibitors and Amputation and Other Vascular Outcomes
1.
Centers for Disease Control and Prevention.  National Diabetes Statistics Report, 2017. Atlanta, GA: Centers for Disease Control and Prevention, US Dept of Health and Human Services; 2017.
2.
US Food and Drug Administration. FDA confirms increased risk of leg and foot amputations with the diabetes medicine canagliflozin (Invokana, Invokamet, Invokamet XR). https://www.fda.gov/downloads/Drugs/DrugSafety/UCM558427.pdf. Updated May 16, 2017. Accessed October 17, 2017.
3.
Fulcher  G, Matthews  DR, Perkovic  V,  et al.  Efficacy and safety of canagliflozin used in conjunction with sulfonylurea in patients with type 2 diabetes mellitus: controlled trial.  Diabetes Ther. 2015;6(3):289-302. doi:10.1007/s13300-015-0117-zPubMedGoogle ScholarCrossref
4.
Fulcher  G, Matthews  DR, Perkovic  V,  et al; CANVAS Trial Collaborative Group.  Efficacy and safety of canagliflozin when used in conjunction with incretin-mimetic therapy in patients with type 2 diabetes.  Diabetes Obes Metab. 2016;18(1):82-91. doi:10.1111/dom.12589PubMedGoogle ScholarCrossref
5.
Neal  B, Perkovic  V, de Zeeuw  D,  et al; CANVAS Trial Collaborative Group.  Efficacy and safety of canagliflozin, an inhibitor of sodium-glucose cotransporter 2, when used in conjunction with insulin therapy in patients with type 2 diabetes.  Diabetes Care. 2015;38(3):403-411. doi:10.2337/dc14-1237PubMedGoogle ScholarCrossref
6.
Neal  B, Perkovic  V, Mahaffey  KW,  et al; CANVAS Program Collaborative Group.  Canagliflozin and cardiovascular and renal events in type 2 diabetes.  N Engl J Med. 2017;377(7):644-657. doi:10.1056/NEJMoa1611925PubMedGoogle ScholarCrossref
7.
Lee  S.  Update on SGLT2 inhibitors—new data released at the American Diabetes Association.  Crit Pathw Cardiol. 2017;16(3):93-95. doi:10.1097/HPC.0000000000000125PubMedGoogle ScholarCrossref
8.
Liu  J, Lee  TW, DeFronzo  RA.  Why do SGLT2 inhibitors inhibit only 30-50% of renal glucose absorption in humans?  Diabetes. 2012;61:2199-2204. doi:10.2337/db12-0052PubMedGoogle ScholarCrossref
9.
Chang  CC, Chen  YT, Hsu  CY,  et al.  Dipeptidyl peptidase-4 inhibitors, peripheral arterial disease, and lower extremity amputation risk in diabetic patients.  Am J Med. 2017;130(3):348-355. doi:10.1016/j.amjmed.2016.10.016PubMedGoogle ScholarCrossref
10.
Newton  KM, Wagner  EH, Ramsey  SD,  et al.  The use of automated data to identify complications and comorbidities of diabetes: a validation study.  J Clin Epidemiol. 1999;52(3):199-207. doi:10.1016/S0895-4356(98)00161-9PubMedGoogle ScholarCrossref
11.
Bekwelem  W, Bengtson  LG, Oldenburg  NC,  et al.  Development of administrative data algorithms to identify patients with critical limb ischemia.  Vasc Med. 2014;19(6):483-490. doi:10.1177/1358863X14559589PubMedGoogle ScholarCrossref
12.
Chang  HY, Weiner  JP, Richards  TM, Bleich  SN, Segal  JB.  Predicting costs with Diabetes Complications Severity Index in claims data.  Am J Manag Care. 2012;18(4):213-219.PubMedGoogle Scholar
13.
Chang  HY, Weiner  JP, Richards  TM, Bleich  SN, Segal  JB.  Validating the adapted Diabetes Complications Severity Index in claims data.  Am J Manag Care. 2012;18(11):721-726.PubMedGoogle Scholar
14.
Singh  S, Chang  HY, Richards  TM, Weiner  JP, Clark  JM, Segal  JB.  Glucagonlike peptide 1–based therapies and risk of hospitalization for acute pancreatitis in type 2 diabetes mellitus: a population-based matched case-control study.  JAMA Intern Med. 2013;173(7):534-539. doi:10.1001/jamainternmed.2013.2720PubMedGoogle ScholarCrossref
15.
Chang  HY, Hsieh  CF, Singh  S, Tang  W, Chiang  YT, Huang  WF.  Anti-diabetic therapies and the risk of acute pancreatitis: a nationwide retrospective cohort study from Taiwan.  Pharmacoepidemiol Drug Saf. 2015;24(6):567-575. doi:10.1002/pds.3770PubMedGoogle ScholarCrossref
16.
Tseng  YH, Tsan  YT, Chan  WC, Sheu  WH, Chen  PC.  Use of an α-glucosidase inhibitor and the risk of colorectal cancer in patients with diabetes: a nationwide, population-based cohort study.  Diabetes Care. 2015;38(11):2068-2074. doi:10.2337/dc15-0563PubMedGoogle ScholarCrossref
17.
Ou  SM, Shih  CJ, Chao  PW,  et al.  Effects on clinical outcomes of adding dipeptidyl peptidase-4 inhibitors versus sulfonylureas to metformin therapy in patients with type 2 diabetes mellitus.  Ann Intern Med. 2015;163(9):663-672. doi:10.7326/M15-0308PubMedGoogle ScholarCrossref
18.
Chang  HY, Zhou  M, Tang  W, Alexander  GC, Singh  S.  Risk of gastrointestinal bleeding associated with oral anticoagulants: population based retrospective cohort study.  BMJ. 2015;350:h1585. doi:10.1136/bmj.h1585PubMedGoogle ScholarCrossref
19.
Casadaban  L, Rauscher  G, Aklilu  M, Villenes  D, Freels  S, Maker  AV.  Adjuvant chemotherapy is associated with improved survival in patients with stage II colon cancer.  Cancer. 2016;122(21):3277-3287. doi:10.1002/cncr.30181PubMedGoogle ScholarCrossref
20.
Harder  VS, Stuart  EA, Anthony  JC.  Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research.  Psychol Methods. 2010;15(3):234-249. doi:10.1037/a0019623PubMedGoogle ScholarCrossref
21.
Kurth  T, Walker  AM, Glynn  RJ,  et al.  Results of multivariable logistic regression, propensity matching, propensity adjustment, and propensity-based weighting under conditions of nonuniform effect.  Am J Epidemiol. 2006;163(3):262-270. doi:10.1093/aje/kwj047PubMedGoogle ScholarCrossref
22.
Austin  PC.  An introduction to propensity score methods for reducing the effects of confounding in observational studies.  Multivariate Behav Res. 2011;46(3):399-424. doi:10.1080/00273171.2011.568786PubMedGoogle ScholarCrossref
23.
Lin  DY, Wei  LJ.  The robust inference for the Cox proportional hazards model.  J Am Stat Assoc. 1989;84:1074-1078. doi:10.1080/01621459.1989.10478874Google ScholarCrossref
24.
Lin  DY, Wei  LJ, Ying  Z.  Checking the Cox model with cumulative sums of martingale-based residuals.  Biometrika. 1993;80:557-572. doi:10.1093/biomet/80.3.557Google ScholarCrossref
25.
European Medicines Agency. Pharmacovigilence Risk Assessment Committee (PRAC): PRAC assessment report. http://www.ema.europa.eu/docs/en_GB/document_library/Referrals_document/SGLT2_inhibitors_Canagliflozin_20/European_Commission_final_decision/WC500227102.pdf. Published February 9, 2017. Accessed October 23, 2017.
26.
US Food and Drug Administration. Sodium-glucose cotransporter-2 (SGLT2) inhibitors. https://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationforPatientsandProviders/ucm446852.htm. Updated May 16, 2017. Accessed October 27, 2017.
27.
Fadini  GP, Avogaro  A.  SGLT2 inhibitors and amputations in the US FDA Adverse Event Reporting System.  Lancet Diabetes Endocrinol. 2017;5(9):680-681. doi:10.1016/S2213-8587(17)30257-7PubMedGoogle ScholarCrossref
28.
Yuan  Z, DeFalco  FJ, Ryan  PB,  et al.  Risk of lower extremity amputations in people with type 2 diabetes mellitus treated with sodium-glucose co-transporter-2 inhibitors in the USA: a retrospective cohort study.  Diabetes Obes Metab. 2018;20(3):582-589. doi:10.1111/dom.13115PubMedGoogle ScholarCrossref
29.
Hsu  CY, Chen  YT, Su  YW, Chang  CC, Huang  PH, Lin  SJ.  Statin therapy reduces future risk of lower-limb amputation in patients with diabetes and peripheral artery disease.  J Clin Endocrinol Metab. 2017;102(7):2373-2381. doi:10.1210/jc.2016-3717PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    September 2018

    Association Between Sodium-Glucose Cotransporter 2 Inhibitors and Lower Extremity Amputation Among Patients With Type 2 Diabetes

    Author Affiliations
    • 1Center for Drug Safety and Effectiveness, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland
    • 2Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland
    • 3Department of Family Medicine and Community Health, University of Massachusetts Medical School, Worcester
    • 4Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland
    • 5Division of General Internal Medicine, Department of Medicine, Johns Hopkins Medicine, Baltimore, Maryland
    JAMA Intern Med. 2018;178(9):1190-1198. doi:10.1001/jamainternmed.2018.3034
    Key Points

    Question  What is the association between the use of sodium-glucose cotransporter 2 inhibitors and lower extremity amputation, peripheral arterial disease, osteomyelitis, and venous ulceration among patients with type 2 diabetes?

    Findings  In this population-based cohort study of commercially insured patients, the length of follow-up was relatively short, and amputations were rare. There was no statistically significantly increased risk of amputations associated with new use of sodium-glucose cotransporter 2 inhibitors compared with new use of dipeptidyl peptidase 4 inhibitors and glucagon-like peptide 1 agonists, although the hazard ratios were elevated, while new use of sodium-glucose cotransporter 2 inhibitors was associated with a statistically significant increased risk of amputation compared with use of metformin, sulfonylureas, and thiazolidinediones.

    Meaning  Sodium-glucose cotransporter 2 inhibitors may be associated with increased risk of amputation compared with some oral treatments for type 2 diabetes.

    Abstract

    Importance  Results of clinical trials suggest that canagliflozin, a sodium-glucose cotransporter 2 (SGLT-2) inhibitor for treating type 2 diabetes, may be associated with lower extremity amputation.

    Objective  To quantify the association between the use of oral medication for type 2 diabetes and 5 outcomes (lower extremity amputation, peripheral arterial disease, critical limb ischemia, osteomyelitis, and ulcer).

    Design, Setting, and Participants  A retrospective cohort study was conducted using Truven Health MarketScan Commercial Claims and Encounters data on new users between September 1, 2012, and September 30, 2015. The study focused on 2.0 million commercially insured individuals and used propensity score weighting to balance baseline differences among groups. Sensitivity analyses varied statistical models, assessed the effect of combining dipeptidyl peptidase 4 (DPP-4) inhibitors and glucagon-like peptide 1 (GLP-1) agonists as a single referent group, adjusted for baseline use of older oral agents, and included people with baseline amputation.

    Exposures  New use of SGLT-2 inhibitors alone, DPP-4 inhibitors alone, GLP-1 agonists alone, or other antidiabetic agents (sulfonylurea, metformin hydrochloride, or thiazolidinediones).

    Main Outcomes and Measures  Foot and leg amputation, defined by validated International Classification of Diseases, Ninth Revision and Current Procedural Terminology codes.

    Results  Among 2.0 million potentially eligible individuals, a total of 953 906 (516 046 women and 437 860 men; mean [SD] age, 51.8 [10.9] years) were included in the final analyses, including 39 869 new users of SGLT-2 inhibitors (4.2%), 105 023 new users of DPP-4 inhibitors (11.0%), and 39 120 new users of GLP-1 agonists (4.1%). The median observation time ranged from 99 days for new users of GLP-1 agonists to 127 days for those using metformin, sulfonylureas, and thiazolidinediones, while the crude incident rates ranged from 4.90 per 10 000 person-years for those using metformin, sulfonylureas, and thiazolidinediones to 10.53 per 10 000 person-years for new users of SGLT-2 inhibitors. After propensity score weighting and adjustment for demographics, severity of diabetes, comorbidities, and medications, there was a nonstatistically significant increased risk of amputation associated with new use of SGLT-2 inhibitors compared with DPP-4 inhibitors (adjusted hazard ratio, 1.50; 95% CI, 0.85-2.67) and GLP-1 agonists (adjusted hazard ratio, 1.47; 95% CI, 0.64-3.36). New use of SGLT-2 inhibitors was statistically significantly associated with amputation compared with sulfonylureas, metformin, or thiazolidinediones (adjusted hazard ratio, 2.12; 95% CI, 1.19-3.77). These results persisted in sensitivity analyses.

    Conclusions and Relevance  Use of SGLT-2 inhibitors may be associated with increased risk of amputation compared with some oral treatments for type 2 diabetes. Further observational studies are needed with extended follow-up and larger sample sizes.

    Introduction

    Diabetes is common and costly. An estimated 23 million individuals in the United States have received a diagnosis of diabetes, and approximately 1.5 million additional cases are identified each year.1 The morbidity associated with diabetes is due to both macrovascular and microvascular complications, yet the prevalence of these complications can be attenuated by appropriate glycemic control, underscoring the importance of lifestyle modification and pharmacotherapies that reduce glycemic burden.

    Quiz Ref IDIn 2013, the first sodium-glucose cotransporter 2 (SGLT-2) inhibitor was approved by the US Food and Drug Administration (FDA) for the treatment of type 2 diabetes. In contrast to other oral treatments, such as biguanides, sulfonylureas, thiazolidinediones (TZDs), glucagon-like peptide 1 (GLP-1) agonists, or dipeptidyl peptidase 4 (DPP-4) inhibitors, SGLT-2 inhibitors inhibit the reabsorption of glucose in the kidney. Since the approval of canagliflozin, 2 other SGLT-2 inhibitors, dapagliflozin and empagliflozin, have been approved by the FDA.

    In May 2017, the FDA issued a Drug Safety Communication regarding an increased risk of foot and leg amputations with the use of canagliflozin.2 This warning was based on the evidence from 2 clinical trials and resulted in a boxed warning on the labels of canagliflozin products. The Canagliflozin Cardiovascular Assessment Study (CANVAS) program used data from 2 trials3-5 and showed that there was a statistically significantly higher risk of amputation with canagliflozin than with placebo (6.3 vs 3.4 participants with amputations per 1000 patient-years).6,7

    Although both CANVAS and Canagliflozin Cardiovascular Assessment Study–Renal (CANVAS-R) suggested an increased risk for lower limb amputations, they examined only 1 SGLT-2 inhibitor, neither was powered to assess this outcome, and neither offers an assessment of the association among a broad and diverse group of real-world users. Thus, we investigated the association between use of SGLT-2 inhibitors and lower extremity amputation using a new user design among a large cohort of commercially insured individuals in the United States between 2012 and 2015.

    Methods
    Study Design and Data Source

    Quiz Ref IDWe conducted a population-based, retrospective, new-user design cohort study using Truven Health MarketScan Commercial Claims and Encounters data (Truven Health Analytics) from September 1, 2012, through September 30, 2015. MarketScan is one of the largest commercial claims databases in the United States, with information on more than 25 million individuals annually. The data consist of individual-level health care use data, including demographic characteristics and information on medical and pharmacy services provided. The study was exempted from review by a Johns Hopkins Institutional Review Board; as this study was not human participant research, patient consent was not required.

    Cohort Derivation

    Patients with at least 1 of the following antidiabetic medications were considered for the analysis: 3 newer agents (SGLT-2 inhibitors, DPP-4 inhibitors, or GLP-1 agonists) and 3 older agents (sulfonylureas, metformin hydrochloride, and TZDs). Medications were identified using National Drug Codes. The index date for each patient was the first prescription date of the assigned exposure medication on or after the study entry date, March 29, 2013. We included a baseline period of 6 months prior to the index date.

    Patients were excluded if they met any of the following criteria: sex unknown; younger than 18 years of age at cohort entry; having any newer antidiabetic agent during the baseline period; absence of continuous medical and pharmacy enrollment during the baseline period; any insulin use between 2010 and 2015; receipt of 2 or more of the 3 newer agents as a first prescription (eg, concomitant initiation of SGLT-2 and DPP-4 inhibitors); having an index date after September 30, 2015; and having an outcome(s) of interest in the baseline period. Given that amputation is rare and that we wanted to include as many participants as possible, we derived 2 cohorts: 1 excluding individuals with baseline amputation (n = 953 906) and another excluding individuals with any baseline outcome (n = 933 073).

    We defined each individual’s last date of observation as the earliest of the following 6 dates: (1) the last date of continuous exposure to the assigned medication, which was the last day of continuous drug at hand plus 30 days (we used 30 days to account for the clearance of the drug since the half-life of SGLT-2 inhibitors is hours)8; (2) the date prior to the switch to or the addition of other newer agents; (3) the date of the loss of medical or pharmacy enrollment; (4) the first date of non–outcome-related hospitalization, as prescription use during hospitalization was not available; (5) the study end date, September 30, 2015; or (6) the first date of the outcome. Patients were censored for that outcome if the last date of observation was not equal to the first date of the outcome. The distribution of reasons for exiting the cohort are presented in eTable 1 in the Supplement.

    Definition of Exposure

    Patients with any prescription of the newer medications were assigned to a medication group (SGLT-2 inhibitors, DPP-4 inhibitors, or GLP-1 agonists) based on their first prescribed newer medication, while the remaining patients were assigned to a fourth group (“other drugs”) reflecting use of the 3 older agents only (sulfonylureas, metformin, and TZDs) (eTable 2 in the Supplement). Patients assigned to the SGLT-2 inhibitors, DPP-4 inhibitors, or GLP-1 agonists group could have any older agent before the index date, or we would not have had enough patients; for example, more than half of the new users of the newer agents had metformin, and about one-fourth had sulfonylureas prescribed during the baseline period (Table 1). Although exposure to DPP-4 inhibitors was the primary comparison group (relative to SGLT-2 inhibitors), exposure to GLP-1 agonists was included as a second comparison group because a recent observational study suggested that DPP-4 inhibitors may be protective against amputations.9

    Definition of Outcomes

    We defined 5 outcomes: foot and leg amputation, peripheral arterial disease, critical limb ischemia, osteomyelitis, and ulcer. These outcomes were defined using administrative codes after a comprehensive literature review to identify the most accurate administrative codes associated with these conditions. Four of 5 outcomes came from a validation study identifying diabetes-related complications,10 and the predictive positive value was 0.85 for amputation, 0.89 for ulcer, 0.64 for osteomyelitis, and 0.64 for peripheral vascular disease; critical limb ischemia was also similarly identified with a κ coefficient of 80.11 (Detailed definitions can be found in eAppendix 1 in the Supplement.)

    Definition of Confounders

    Based on a comprehensive literature review, we derived 4 groups of confounders: demographics (age and sex), diabetes severity, comorbidities, and medications from the baseline period. The adapted Diabetes Complications Severity Index score measures an individual’s diabetes severity using claims only and is the sum of 7 diabetes complications graded by severity as 0, 1, or 2 (as the severity increases, the score increases); it ranges from 0 to 13 (0 indicates no complications and 13 indicates the most severe complications).12,13 The adapted Diabetes Complications Severity Index score has been shown to be associated with number of hospitalizations and costs among patients with diabetes and was adopted extensively.14-17 We assessed the comorbidities, including cerebrovascular disease, congestive heart failure, ischemic heart disease, hypertension, retinopathy, nephropathy, neuropathy, atrial fibrillation, renal disease, and eye disease (eAppendix 2 in the Supplement). We evaluated the medications, including cardiovascular medications (eg, cardiac drugs, loop diuretics, lipid-lowering agents, and antithrombotic drugs), hormone replacement therapy, antiasthmatic drugs, statins, aspirin, anticoagulants, thiazides, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, and calcium channel blockers (eTable 3 in the Supplement).

    Propensity Score

    For each cohort, we developed 3 propensity score–based weights to achieve a better balance between the SGLT-2 group and each comparison group.15,18,19 We applied logistic regression to determine the possibility of SGLT-2 use among individuals using SGLT-2 inhibitors and those using DPP-4 inhibitors, GLP-1 agonists, and other drugs. All confounders described earlier were included in the propensity score models as independent variables.

    Although there are various methods to apply propensity scores, no single method consistently outperforms others.20,21 We used propensity score weighting because it could produce one interpretable overall treatment effect and would not diminish our sample size. To estimate the average effect of treatment on individuals using SGLT-2 inhibitors, the average treatment effect of the treated (ATT) weighting was applied; that is, we compared the hazards of outcomes among individuals using SGLT-2 inhibitors with the hypothesized situation had they taken DPP-4 inhibitors, GLP-1 agonists, or older agents instead of SGLT-2 inhibitors. This approach is specifically useful when systematic differences likely occur between the study sample and the overall population.22 The balance between individuals using SGLT-2 inhibitors and 3 comparison groups in baseline covariates before and after ATT weighting was compared using the standardized difference, with a standardized difference less than 0.1 considered negligible.22 After ATT weighting, standardized differences of all but 1 covariate (categorical age groups) were reduced from as high as 0.41 to 0.1 or smaller, suggesting that the groups were well balanced (eTables 4-7 in the Supplement).

    Statistical Analysis

    We used χ2 tests for categorical variables and Kruskal-Wallis tests for continuous variables to assess whether patients’ characteristics were different across the 4 medication groups. For each outcome, 3 separate Cox proportional hazards regression models with propensity score ATT weighting were constructed to examine the association between the use of SGLT-2 inhibitors (relative to 3 reference groups) and the outcome. We calculated robust estimates of SEs for all variables in the models.23 All control variables initially entered the models as regression covariates, and the proportional hazards assumption was assessed at the P = .05 level for each control variable (Kolmogorov-type supremum test); if a variable violated the proportional hazards assumption, the variable was included as a stratification factor.24 We continued this process until no regressors violated the proportional hazards assumption (eTable 8 in the Supplement). We used SAS, version 9.3 (SAS Institute Inc) for all analyses. P < .05 (2-sided) was considered significant.

    Sensitivity Analyses

    We performed 4 sets of sensitivity analyses to test the robustness of our findings. First, we evaluated 2 additional models, including all control variables as regression covariates with and without ATT weighting. Second, we combined individuals using DPP-4 inhibitors and GLP-1 agonists as 1 reference group and repeated the main analyses. Third, we added 3 binary indicators of the older agents’ baseline use in the covariate-only model and evaluated the differences in the hazard ratios. Fourth, we included patients with baseline amputation.

    Results
    Patient Inclusion and Characteristics

    The Figure depicts the process whereby the final sample, which included 953 906 individuals (516 046 women and 437 860 men; mean [SD] age, 51.8 [10.9] years), was derived. We started with approximately 26.8 million transactions for diabetes medicines derived between March 29, 2013, and December 31, 2015, coming from 2 042 383 individuals. Of these, 595 377 patients (29.2%) had prescriptions for DPP-4 inhibitors, GLP-1 agonists, or SGLT-2 inhibitors, while the remainder represented patients with prescriptions for other, older oral antidiabetes medicines. After excluding individuals such as those using insulin or the newer agents at baseline, 953 906 patients remained in our first cohort (excluding only those with amputation during baseline), while 933 073 patients remained in our second cohort (excluding those with any outcome during baseline). The first cohort included 39 869 new users of SGLT-2 inhibitors (4.2%), 105 023 new users of DPP-4 inhibitors (11.0%), and 39 120 new users of GLP-1 agonists (4.1%). Among the 39 869 new users of SGLT-2 inhibitors, 28 036 were taking canagliflozin, 8647 were taking dapagliflozin, and 3186 were taking empagliflozin.

    Table 1 depicts the characteristics of study participants from the first cohort and demonstrates that there were statistically significant differences in all baseline covariates across 4 groups of medication users. For example, new users of SGLT-2 inhibitors and DPP-4 inhibitors were less likely to be female (18 842 of 39 869 [47.3%] used SGLT-2 inhibitors, and 45 518 of 105 023 [43.3%] used DPP-4 inhibitors) than those initiating GLP-1 agonists (23 844 of 39 120 [61.0%]) or other diabetes treatments (427 842 of 769 894 [55.6%]). The prevalence of hypertension at baseline was greater among new users of SGLT-2 inhibitors (23 825 of 39 869 [59.8%]), DPP-4 inhibitors (58 725 of 105 023 [55.9%]), and GLP-1 agonists (20 348 of 39 120 [52.0%]) than among those initiating other diabetes medicines (305 551 of 769 894 [39.7%]). Similarly, the prevalence of statin use was greater among users of SGLT-2 inhibitors (14 383 of 39 869 [36.1%]), DPP-4 inhibitors (35 778 of 105 023 [34.1%]), and GLP-1 agonists (12 318 of 39 120 [31.5%]) than among those initiating other diabetes medicines (167 723 of 769 894 [21.8%]). We observed similar results in the second cohort (eTable 9 in the Supplement).

    Crude Association Between Treatment, Amputation, and Other Vascular Outcomes

    Table 2 depicts the incidence of amputation and other vascular outcomes among new users of SGLT-2 inhibitors, DPP-4 inhibitors, and GLP-1 agonists and users of older agents for type 2 diabetes. For example, among the first cohort, amputation was rare, although it was more common among new users of SGLT-2 inhibitors (10.53 per 10 000 person-years; 10.00 per 10 000 person-years for canagliflozin users) than among new users of DPP-4 inhibitors (8.52 per 10 000 person-years), GLP-1 agonists (7.10 per 10 000 person-years), or metformin, sulfonylureas, or TZDs (4.90 per 10 000 person-years). Patterns regarding other vascular outcomes varied, although the crude, or unadjusted, rates of both peripheral vascular disease and critical limb ischemia were greater among new users of SGLT-2 inhibitors, DPP-4 inhibitors, and GLP-1 agonists than among users of metformin, sulfonylureas, or TZDs. The median observation time was short, ranging from 99 days (interquartile range, 61-184 days) for GLP-1 agonists to 127 days (interquartile range, 61-278 days) for users of metformin, sulfonylureas, or TZDs.

    Adjusted Association Between Treatment, Amputation, and Other Vascular Outcomes

    Table 3 describes the adjusted associations of interest. After use of propensity scores as well as additional adjustment for potentially confounding covariates, there was a nonstatistically significant increased risk of amputation associated with new use of SGLT-2 inhibitors compared with new use of DPP-4 inhibitors (adjusted hazard ratio [aHR], 1.50; 95% CI, 0.85-2.67) and GLP-1 agonists (aHR, 1.47; 95% CI, 0.64-3.36). Quiz Ref IDNew use of SGLT2-inhibitors was associated with a statistically significant greater likelihood of amputation compared with the other oral diabetes medicines (aHR, 2.12; 95% CI, 1.19-3.77).

    After adjustment, SGLT-2 inhibitors were associated with some but not all of the other outcomes examined. Quiz Ref IDFor example, compared with users of metformin, sulfonylureas, or TZDs, new users of SLGT-2 inhibitors had higher rates of vascular ulcers (aHR, 1.34; 95% CI, 1.10-1.61), osteomyelitis (aHR, 1.44; 95% CI, 1.02-2.05), and peripheral vascular disease (aHR, 1.11; 95% CI, 1.02-1.22). However, compared with new users of DPP-4 inhibitors, new users of SGLT-2 inhibitors also had lower rates of some outcomes, such as peripheral vascular disease (aHR, 0.88; 95% CI, 0.79-0.96) and critical limb ischemia (aHR, 0.76; 95% CI, 0.64-0.89).

    Sensitivity Analyses

    The results of sensitivity analyses are depicted in eTables 10 to 13 in the Supplement. Models that included all control variables as regression covariates with and without ATT weighting yielded substantively similar results as those depicted in Table 3. Analyses that combined new users of DPP-4 inhibitors and new users of GLP-1 agonists as a single reference group also yielded similar findings to those reported herein, with a nonstatistically significant increased association between SLGT-2 inhibitor use and amputation relative to this combined reference group. Including 3 binary indicators of the use of older agents at baseline did not change our substantive findings or their interpretation. Last, after inclusion of patients with prior amputation, we found a borderline statistically significant increase in the hazard ratio associated with the use of SGLT-2 inhibitors compared with the use of DPP-4 inhibitors (aHR, 1.73; 95% CI, 1.01-2.98; P = .048).

    Discussion

    It remains unclear whether or not SGLT-2 inhibitors, such as canagliflozin, are associated with an increased risk of foot and leg amputations compared with placebo. In this retrospective cohort study using a large commercial claims database in the United States, overall rates of amputation were low, and the length of follow-up was limited. Despite this limitation, we found a nonstatistically significant increased risk of amputation associated with new use of SGLT-2 inhibitors compared with DPP-4 inhibitors and GLP-1 agonists. In addition, new use of SGLT-2 inhibitors was associated with a statistically significant greater likelihood of amputation compared with the use of 3 older oral diabetes medicines examined. These findings are relevant because of how commonly SGLT-2 inhibitors are prescribed, the increased risk of foot and leg amputations with the use of canagliflozin in the CANVAS and CANVAS-R clinical trials, and the importance of this outcome for patients.

    The mechanisms by which SGLT-2 inhibitors may increase the risk of amputation are not known. Some have speculated that by promoting glucosuria and volume depletion and subsequent hemoconcentration in patients with diabetes who are already at risk, these products may increase the risk of peripheral artery disease and amputation. Consistent with this postulated hypothesis are the findings from an exploratory analysis by the European Medicines Agency that patients with chronic kidney disease may be at the highest risk, although, again, definitive evidence is lacking.25

    Our results bear similarities to and differences from recent clinical trials, observational studies, and disproportionality analyses yielding conflicting results. In the CANVAS and CANVAS-R trials, there was an approximate doubling in the risk of amputation of toes, feet, or legs with canagliflozin than with placebo (6.3 vs 3.4 participants with amputation per 1000 patient-years, corresponding to a hazard ratio of 1.97 [95% CI, 1.41-2.75]), with 71% of the affected participants having their highest amputation at the level of the toe or metatarsal in 2 clinical trials of patients with cardiovascular disease after a mean follow-up of 188 weeks.6 The inability to detect a significant effect in our sample compared with individuals taking DPP-4 inhibitors and GLP-1 agonists may reflect a shorter duration of follow-up, as well as a much lower proportion of patients with cardiovascular disease, a fact that is reflected in the overall low rates of amputation among all groups in our study. An increase in the risk of amputation has not been seen with clinical trials of the other SGLT-2 inhibitors, such as dapagliflozin and empagliflozin, although long-term studies are ongoing.26 Another disproportionality analysis of the FDA’s Adverse Event Reporting System noted a significantly increased proportional reporting ratio for amputation associated with canagliflozin at 3.83 (95% CI, 2.39-6.14) but allowed for limited inferences owing to unknown factors influencing reporting trends.27

    In contrast to clinical trials and disproportionality analyses reporting an increased risk of amputation, a cohort study conducted by the manufacturer of canagliflozin reported no statistically significant increased risk of below-knee amputations among patients with type 2 diabetes; the authors reported a hazard ratio of 0.98 (95% CI, 0.68-1.41; P = .92) for the comparison of SGLT-2 inhibitors and non–SGLT-2 antiglycemic agents.28 However, in contrast to our analysis, the authors combined older and newer diabetes treatments (except metformin) into 1 group and conducted an intention-to-treat analysis regardless of whether participants discontinued, switched, or augmented treatment, which may have diluted any difference between groups. It is also unclear if the loss of enrollment or pharmacy coverage was accounted for. Similar to our study, event rates were very low compared with those in clinical trials.

    The FDA has recently placed a boxed warning on products containing canagliflozin; however, regulators have not issued statements on whether this association is a class-wide association. In addition, the warning addresses only lower-limb amputations and does not make a statement on other manifestations of peripheral vascular disease—some, but not all, of which we found were associated with use of SLGT-2 inhibitors. The European Medicines Agency has also taken steps to better understand patient safety with the use of SGLT-2 inhibitors and mandated a review of all peripheral vascular disease outcomes with SGLT-2 inhibitors. Because all SGLT-2 inhibitors share similar mechanisms of action, a warning for amputations as a class effect was applied to all SGLT-2 inhibitors after review of clinical trial data.25 Our study supplements this body of evidence by investigating the risk of lower extremity amputations across 3 SGLT-2 inhibitors and including a variety of additional outcomes of interest to patients, clinicians, and regulators.

    Limitations

    Our study has some limitations. The MarketScan data set consisted of privately insured individuals, so the generalizability to other populations may be questionable. As we noted, our event rates were low, and our duration of follow-up was limited. As data from clinical trials accumulate over time, a meta-analysis of randomized clinical trials may be helpful, although, to our knowledge, amputations have not been prospectively evaluated in trials nor formally ascertained as adverse events. In addition, our sample size is insufficient to examine for evidence of heterogeneity of effects across different SLGT-2 products. The inclusion of prevalent users of older drugs has the potential to introduce immortal time bias. There is also a potential for misclassification of exposure and outcomes. For example, prescriptions may be dispensed yet not taken, although such misclassification of exposure would lead us to underestimate the true magnitude of the association between SGLT-2 inhibitors and amputation. We used previously validated algorithms to derive our outcome variables (eg, peripheral vascular disease), although outcome misclassification also remains possible. Quiz Ref IDConfounding by indication and disease severity could explain the increased risk of amputation seen with the use of SGLT-2 inhibitors compared with the use of older agents. To address this confounding, we used propensity score weighting, adjusted for differences in baseline characteristics, and adopted a new-user design. The higher proportion of statin use among those using SGLT-2 inhibitors may have attenuated our risk estimates because statins may be protective against amputations.29 Finally, we focused on SGLT-2 inhibitors and amputation; further studies are necessary to examine this association and also the potential association between DPP-4 inhibitors and the outcomes we explore.9

    Conclusions

    Given the uncertainty regarding the true nature of the association between SLGT-2 inhibitors and amputation, clinicians and patients will have to navigate treatment choices while balancing the potential risks of these products against their benefits and alternatives. As the regulatory communications from the FDA and European Medicines Agency make clear, there is a pressing need for further information derived from a number of sources, ranging from passive surveillance systems to meta-analyses of large studies, including observational studies of vascular outcomes such as those we examine herein.

    Back to top
    Article Information

    Accepted for Publication: April 20, 2018.

    Corresponding Author: G. Caleb Alexander, MD, MS, Center for Drug Safety and Effectiveness, Johns Hopkins Bloomberg School of Public Health, 615 N Wolfe St W6035, Baltimore, MD 21205 (galexand@jhsph.edu).

    Published Online: August 13, 2018. doi:10.1001/jamainternmed.2018.3034

    Author Contributions: Drs Chang and Alexander had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Chang, Singh, Baksh, Alexander.

    Acquisition, analysis, or interpretation of data: All authors.

    Drafting of the manuscript: Chang, Singh, Mansour, Baksh.

    Critical revision of the manuscript for important intellectual content: All authors.

    Statistical analysis: Chang, Singh.

    Obtained funding: Alexander.

    Administrative, technical, or material support: Mansour, Baksh, Alexander.

    Supervision: Alexander.

    Conflict of Interest Disclosures: Dr Alexander reported serving as Chair of the US Food and Drug Administration’s Peripheral and Central Nervous System Advisory Committee, serving as a paid consultant to QuintilesIMS, serving on the Advisory Board of MesaRx Innovations, holding equity in Monument Analytics, and serving as a member of OptumRx’s Pharmacy and Therapeutics Committee. This arrangement has been reviewed and approved by the Johns Hopkins Bloomberg School of Public Health. Dr Singh reported attending advisory board meetings on the safety of diabetic drugs hosted by Janssen Pharmaceuticals (manufacturer of canagliflozin) and Eli Lilly & Co (manufacturer of dulaglutide) and was compensated for his time.

    Funding/Support: This work was supported in part through contract number U01 FD004977-03 from the Johns Hopkins Center of Excellence in Regulatory Science and Innovation (Dr Alexander).

    Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study, analysis or interpretation of the data, and preparation or final approval of the manuscript prior to publication.

    References
    1.
    Centers for Disease Control and Prevention.  National Diabetes Statistics Report, 2017. Atlanta, GA: Centers for Disease Control and Prevention, US Dept of Health and Human Services; 2017.
    2.
    US Food and Drug Administration. FDA confirms increased risk of leg and foot amputations with the diabetes medicine canagliflozin (Invokana, Invokamet, Invokamet XR). https://www.fda.gov/downloads/Drugs/DrugSafety/UCM558427.pdf. Updated May 16, 2017. Accessed October 17, 2017.
    3.
    Fulcher  G, Matthews  DR, Perkovic  V,  et al.  Efficacy and safety of canagliflozin used in conjunction with sulfonylurea in patients with type 2 diabetes mellitus: controlled trial.  Diabetes Ther. 2015;6(3):289-302. doi:10.1007/s13300-015-0117-zPubMedGoogle ScholarCrossref
    4.
    Fulcher  G, Matthews  DR, Perkovic  V,  et al; CANVAS Trial Collaborative Group.  Efficacy and safety of canagliflozin when used in conjunction with incretin-mimetic therapy in patients with type 2 diabetes.  Diabetes Obes Metab. 2016;18(1):82-91. doi:10.1111/dom.12589PubMedGoogle ScholarCrossref
    5.
    Neal  B, Perkovic  V, de Zeeuw  D,  et al; CANVAS Trial Collaborative Group.  Efficacy and safety of canagliflozin, an inhibitor of sodium-glucose cotransporter 2, when used in conjunction with insulin therapy in patients with type 2 diabetes.  Diabetes Care. 2015;38(3):403-411. doi:10.2337/dc14-1237PubMedGoogle ScholarCrossref
    6.
    Neal  B, Perkovic  V, Mahaffey  KW,  et al; CANVAS Program Collaborative Group.  Canagliflozin and cardiovascular and renal events in type 2 diabetes.  N Engl J Med. 2017;377(7):644-657. doi:10.1056/NEJMoa1611925PubMedGoogle ScholarCrossref
    7.
    Lee  S.  Update on SGLT2 inhibitors—new data released at the American Diabetes Association.  Crit Pathw Cardiol. 2017;16(3):93-95. doi:10.1097/HPC.0000000000000125PubMedGoogle ScholarCrossref
    8.
    Liu  J, Lee  TW, DeFronzo  RA.  Why do SGLT2 inhibitors inhibit only 30-50% of renal glucose absorption in humans?  Diabetes. 2012;61:2199-2204. doi:10.2337/db12-0052PubMedGoogle ScholarCrossref
    9.
    Chang  CC, Chen  YT, Hsu  CY,  et al.  Dipeptidyl peptidase-4 inhibitors, peripheral arterial disease, and lower extremity amputation risk in diabetic patients.  Am J Med. 2017;130(3):348-355. doi:10.1016/j.amjmed.2016.10.016PubMedGoogle ScholarCrossref
    10.
    Newton  KM, Wagner  EH, Ramsey  SD,  et al.  The use of automated data to identify complications and comorbidities of diabetes: a validation study.  J Clin Epidemiol. 1999;52(3):199-207. doi:10.1016/S0895-4356(98)00161-9PubMedGoogle ScholarCrossref
    11.
    Bekwelem  W, Bengtson  LG, Oldenburg  NC,  et al.  Development of administrative data algorithms to identify patients with critical limb ischemia.  Vasc Med. 2014;19(6):483-490. doi:10.1177/1358863X14559589PubMedGoogle ScholarCrossref
    12.
    Chang  HY, Weiner  JP, Richards  TM, Bleich  SN, Segal  JB.  Predicting costs with Diabetes Complications Severity Index in claims data.  Am J Manag Care. 2012;18(4):213-219.PubMedGoogle Scholar
    13.
    Chang  HY, Weiner  JP, Richards  TM, Bleich  SN, Segal  JB.  Validating the adapted Diabetes Complications Severity Index in claims data.  Am J Manag Care. 2012;18(11):721-726.PubMedGoogle Scholar
    14.
    Singh  S, Chang  HY, Richards  TM, Weiner  JP, Clark  JM, Segal  JB.  Glucagonlike peptide 1–based therapies and risk of hospitalization for acute pancreatitis in type 2 diabetes mellitus: a population-based matched case-control study.  JAMA Intern Med. 2013;173(7):534-539. doi:10.1001/jamainternmed.2013.2720PubMedGoogle ScholarCrossref
    15.
    Chang  HY, Hsieh  CF, Singh  S, Tang  W, Chiang  YT, Huang  WF.  Anti-diabetic therapies and the risk of acute pancreatitis: a nationwide retrospective cohort study from Taiwan.  Pharmacoepidemiol Drug Saf. 2015;24(6):567-575. doi:10.1002/pds.3770PubMedGoogle ScholarCrossref
    16.
    Tseng  YH, Tsan  YT, Chan  WC, Sheu  WH, Chen  PC.  Use of an α-glucosidase inhibitor and the risk of colorectal cancer in patients with diabetes: a nationwide, population-based cohort study.  Diabetes Care. 2015;38(11):2068-2074. doi:10.2337/dc15-0563PubMedGoogle ScholarCrossref
    17.
    Ou  SM, Shih  CJ, Chao  PW,  et al.  Effects on clinical outcomes of adding dipeptidyl peptidase-4 inhibitors versus sulfonylureas to metformin therapy in patients with type 2 diabetes mellitus.  Ann Intern Med. 2015;163(9):663-672. doi:10.7326/M15-0308PubMedGoogle ScholarCrossref
    18.
    Chang  HY, Zhou  M, Tang  W, Alexander  GC, Singh  S.  Risk of gastrointestinal bleeding associated with oral anticoagulants: population based retrospective cohort study.  BMJ. 2015;350:h1585. doi:10.1136/bmj.h1585PubMedGoogle ScholarCrossref
    19.
    Casadaban  L, Rauscher  G, Aklilu  M, Villenes  D, Freels  S, Maker  AV.  Adjuvant chemotherapy is associated with improved survival in patients with stage II colon cancer.  Cancer. 2016;122(21):3277-3287. doi:10.1002/cncr.30181PubMedGoogle ScholarCrossref
    20.
    Harder  VS, Stuart  EA, Anthony  JC.  Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research.  Psychol Methods. 2010;15(3):234-249. doi:10.1037/a0019623PubMedGoogle ScholarCrossref
    21.
    Kurth  T, Walker  AM, Glynn  RJ,  et al.  Results of multivariable logistic regression, propensity matching, propensity adjustment, and propensity-based weighting under conditions of nonuniform effect.  Am J Epidemiol. 2006;163(3):262-270. doi:10.1093/aje/kwj047PubMedGoogle ScholarCrossref
    22.
    Austin  PC.  An introduction to propensity score methods for reducing the effects of confounding in observational studies.  Multivariate Behav Res. 2011;46(3):399-424. doi:10.1080/00273171.2011.568786PubMedGoogle ScholarCrossref
    23.
    Lin  DY, Wei  LJ.  The robust inference for the Cox proportional hazards model.  J Am Stat Assoc. 1989;84:1074-1078. doi:10.1080/01621459.1989.10478874Google ScholarCrossref
    24.
    Lin  DY, Wei  LJ, Ying  Z.  Checking the Cox model with cumulative sums of martingale-based residuals.  Biometrika. 1993;80:557-572. doi:10.1093/biomet/80.3.557Google ScholarCrossref
    25.
    European Medicines Agency. Pharmacovigilence Risk Assessment Committee (PRAC): PRAC assessment report. http://www.ema.europa.eu/docs/en_GB/document_library/Referrals_document/SGLT2_inhibitors_Canagliflozin_20/European_Commission_final_decision/WC500227102.pdf. Published February 9, 2017. Accessed October 23, 2017.
    26.
    US Food and Drug Administration. Sodium-glucose cotransporter-2 (SGLT2) inhibitors. https://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationforPatientsandProviders/ucm446852.htm. Updated May 16, 2017. Accessed October 27, 2017.
    27.
    Fadini  GP, Avogaro  A.  SGLT2 inhibitors and amputations in the US FDA Adverse Event Reporting System.  Lancet Diabetes Endocrinol. 2017;5(9):680-681. doi:10.1016/S2213-8587(17)30257-7PubMedGoogle ScholarCrossref
    28.
    Yuan  Z, DeFalco  FJ, Ryan  PB,  et al.  Risk of lower extremity amputations in people with type 2 diabetes mellitus treated with sodium-glucose co-transporter-2 inhibitors in the USA: a retrospective cohort study.  Diabetes Obes Metab. 2018;20(3):582-589. doi:10.1111/dom.13115PubMedGoogle ScholarCrossref
    29.
    Hsu  CY, Chen  YT, Su  YW, Chang  CC, Huang  PH, Lin  SJ.  Statin therapy reduces future risk of lower-limb amputation in patients with diabetes and peripheral artery disease.  J Clin Endocrinol Metab. 2017;102(7):2373-2381. doi:10.1210/jc.2016-3717PubMedGoogle ScholarCrossref
    ×