Confirming the BidirectionalNature of the Association BetweenSevere Hypoglycemic andCardiovascular Events in Type 2Diabetes: Insights From EXSCELDiabetes Care 2020;43:643–652 | https://doi.org/10.2337/dc19-1079

OBJECTIVE

We sought to confirm a bidirectional association between severe hypoglycemicevents (SHEs) and cardiovascular (CV) event risk and to characterize individuals atdual risk.

RESEARCH DESIGN AND METHODS

In a post hoc analysis of 14,752 Exenatide Study of Cardiovascular Event Lowering(EXSCEL) participants, we examined time-dependent associations between SHEsand subsequent major adverse cardiac events (CV death, nonfatal myocardialinfarction [MI] or stroke), fatal/nonfatal MI, fatal/nonfatal stroke, hospitalizationfor acute coronary syndrome (hACS), hospitalization for heart failure (hHF), and all-causemortality (ACM), aswell as time-dependentassociationsbetweennonfatal CVevents and subsequent SHEs.

RESULTS

SHEs were uncommon and not associated with once-weekly exenatide therapy(hazard ratio1.13 [95%CI0.94–1.36],P50.179). In fully adjustedmodels, SHEswereassociated with an increased risk of subsequent ACM (1.83 [1.38–2.42], P < 0.001),CV death (1.60 [1.11–2.30], P5 0.012), and hHF (2.09 [1.37–3.17], P5 0.001), whilenonfatal MI (2.02 [1.35–3.01], P 5 0.001), nonfatal stroke (2.30 [1.25–4.23], P 5

0.007), hACS (2.00 [1.39–2.90],P<0.001), andhHF (3.24 [1.98–5.30],P<0.001)wereall associated with a subsequent increased risk of SHEs. The elevated bidirectionaltime-dependent hazards linking SHEs and a composite of all CV events wereapproximately constantover time,with those individuals atdual risk showinghighercomorbidity scores compared with those without.

CONCLUSIONS

Thesefindings, showing greater risk of SHEs after CV events aswell as greater risk ofCV events after SHEs, validate a bidirectional relationship between CV events andSHEs in patients with high comorbidity scores.

In the post-2008 era of cardiovascular outcome trials (CVOTs) mandated by the U.S.Food and Drug Administration (FDA) for any new glucose-lowering drug, severehypoglycemic events (SHEs) requiring third-party assistance have remained a majorchallenge complicating diabetes therapy (1). Post hoc analyses of several CVOTs have

1Munich Diabetes Research Group e.V. at Helm-holtz Centre, Neuherberg, Germany2Duke Clinical Research Institute, DukeUniversitySchool of Medicine, Durham, NC3Diabetes Trials Unit, Oxford Centre for Diabetes,Endocrinology and Metabolism, University ofOxford, Oxford, U.K.4University of North Carolina School of Medicine,Chapel Hill, NC5AstraZeneca Research and Development, Gai-thersburg, MD6ANMCO Research Center, Florence, Italy

Corresponding author: Eberhard Standl, eberhard.standl@lrz.uni-muenchen.de

Received 29 May 2019 and accepted 21 October2019

Clinical trial reg. no. NCT01144338, clinicaltrials.gov.

This article contains Supplementary Data online athttps://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc19-1079/-/DC1.

M.A.B. is currently affiliated with Eli Lilly and Co.,Indianapolis, IN.

© 2019 by the American Diabetes Association.Readersmayuse this article as longas thework isproperly cited, the use is educational and not forprofit, and the work is not altered. More infor-mation is availableathttps://www.diabetesjournals.org/content/license.

See accompanying articles, pp. 512and 541.

Eberhard Standl,1 Susanna R. Stevens,2

Yuliya Lokhnygina,2 M. Angelyn Bethel,3

John B. Buse,4 Stephanie M. Gustavson,5

Aldo P. Maggioni,6 Robert J. Mentz,2

Adrian F. Hernandez,2 andRury R. Holman,3

for the EXSCEL Study Group

Diabetes Care Volume 43, March 2020 643

CARDIOVASCULA

RANDMETA

BOLIC

RISK

found that SHEs are associated with anincreased risk of subsequent cardiovas-cular (CV) and all-cause mortality (ACM)events (2–9). These findings, and the lackof compelling evidence for CV benefitfrom intensive glycemic control in pa-tients with established CV disease, haveprompted many diabetes managementguidelines to suggest relaxing individualglycated hemoglobin (HbA1c) targets insuch patients in the anticipation that thiswill reduce SHEsandassociatedmortalityevents (10,11). Higher HbA1c targets,however, may increase the risk of mi-crovascular complications in the longerterm,andhavenotbeenshownto reducethe risk of SHEs (4,12,13). Indeed, in theAction to Control Cardiovascular Risk inDiabetes (ACCORD) trial, patients withhigher HbA1c values had a greater risk ofSHEs, irrespectiveof the treatment group(4), and the risk of ACM was greatest inthe participants with the highest averageHbA1c values, especially in the intensivetreatment arm (14). On the other hand,SHEs did not account for the difference inmortality between the two study arms(4), and transferring the intensive treat-ment group to less strict HbA1c targetsbetween 7.0% and 7.9% left the in-creased risk of ACM and CV death un-changed (15). Accordingly, it is unclear towhat extent SHEs may play a causativerole in ACMandCVevents,with commonconfounders being a realistic alternativeexplanation. Of note, a strong reverseassociation, in which nonfatal CV eventswere associated with a subsequent in-creased risk of SHEs, was demonstratedin a post hoc analysis of TECOS (Trial Eval-uating Cardiovascular Outcomes WithSitagliptin) (8).We sought to confirm in this post hoc

analysis of data from EXSCEL (ExenatideStudy of Cardiovascular Event Lowering)(16) that there is a bidirectional relation-ship between an increased risk for CVevents after SHEs and an increased riskfor SHEs after nonfatal CV events, par-ticularly in patients at dual risk withfeatures of “frailty” as evidenced by ahigh summary comorbidity score.

RESEARCH DESIGN AND METHODS

Study Design and ParticipantsEXSCEL was a double-blind, placebo-controlled, randomized, event-driven tri-al conducted at 687 sites in 35 countries.It was designed to assess the CV safety ofextended-release exenatide given once

weekly versus placebo when added tousual care in patients with type 2 di-abetes; the results have been reported indetail (16). The trial was designed suchthat;70% of enrolled patients had pre-vious CV events and 30% did not. EXSCELwas run jointly by the Duke Clinical Re-search Institute (DCRI) and theUniversityof Oxford Diabetes Trials Unit in anacademically independent collaborationwith the sponsor, Amylin Pharmaceuti-cals (a wholly owned subsidiary ofAstraZeneca). The protocol was approvedby the ethics committee at each partici-pating site, and the statistical analyseswere performed by the DCRI, indepen-dentof the sponsor. All patients providedwritten informed consent. Eligible pa-tients were adults with type 2 diabetesand with an HbA1c level of 6.5–10.0%(48–96mmol/mol) inclusive. Previous CVevents were defined as a history ofmajorclinical manifestation of coronary arterydisease, ischemic cerebrovascular dis-ease, or atherosclerotic peripheral arte-rial disease. In addition, the summarycomorbidity score at baseline was as-sessed using the Charlson ComorbidityIndex (17,18). History of two or moreepisodes of severe hypoglycemia (de-finedashypoglycemia forwhich apatientreceived third-party assistance) duringthe preceding 12 months was a principleexclusion criterion.

Randomization and Study MedicationEXSCEL participants were randomlyassigned in a 1:1 ratio to receive sub-cutaneous injections of once-weeklyexenatide (EQW) at a dose of 2 mg ormatching placebo. Participants wererequired to discontinue study medica-tion if they had two or more SHEs be-tween trial visits (despite adjustmentof other glucose-lowering agents), hadirreversible kidney dysfunction (con-firmed by two consecutive estimatedglomerular filtration rate [eGFR] values,30 mL/min/1.73 m2), or received renalreplacement therapy. For minimizationof potential confounding effects of differ-ential glycemic levels on trial outcomes,the use of open-label glucose-loweringagents (including dipeptidyl peptidase 4inhibitors but not including glucagon-likepeptide 1 receptor agonists) was encour-aged to promote glycemic equipoise be-tween the two trial groups and to helppatients reach clinically appropriateHbA1c targets.

SHEsAt screening/enrollment, 1 week, 2months, 6 months, and 12 months, andthen at semiannual visits, the symptomsand appropriate management of hypo-glycemiawere reviewed proactively withparticipants. All SHEs were recordedsystematically as prespecified eventsof clinical interest. SHEs were definedper protocol as episodes in which aparticipant was sufficiently disorientedor incapacitated as to require help fromeither another individual or frommedicalpersonnel, i.e., third-party assistance,irrespective of whether this assistancewas actually provided. It did not suffice,for example, if a family member or otherbystander brought the patient a snack ordrink to help raise his or her bloodglucose if it was not clear that the patientcould not have done this unaided. Thiscategory included all patients who had atleast one SHE that was reported duringthe overall period, which was defined asthe period from the date of randomiza-tion through the lastdate that thepatientwas known to be alive.

Clinical OutcomesMajor adverse CV events (MACE) consti-tuted the EXSCEL primary compositeoutcome, defined as the first occurrenceof CV death, nonfatal myocardial infarc-tion (MI), or nonfatal stroke, which wasevaluated in a time-to-event analysis.Secondary outcomes, also evaluated intime-to-event analyses, included ACM,CV death, the first occurrence of nonfatalor fatal MI, nonfatal or fatal stroke,hospitalization for acute coronary syn-drome (hACS), and hospitalization forheart failure (hHF). An independent clin-ical events classification committeewhose members were unaware of thetrial-group assignments adjudicated allthe components of the primary andsecondary outcomes, ventricular arrhyth-mias that led to intervention, neoplasms,and pancreatitis. Prespecified events ofclinical interest for which informationwas collected systematically at everyfollow-up visit, regardless of serious-ness, were pancreatitis, neoplasm, SHEs,and expected CV or diabetes-relatedcomplications.

Statistical AnalysisCategorical variables are presented as n(%), and continuous variables are pre-sented asmedian (25th, 75th percentile).

644 Severe Hypoglycemia and CV Events Diabetes Care Volume 43, March 2020

Risk of first SHE as a function of time isshown with Kaplan-Meier plots, withstudy treatment differences tested usingCox proportional hazards regressionmodels stratified by prior CV disease.Treatment effects for EQW versus pla-cebo are presented as hazard ratios (HRs)and 95% CIs, with numbers of events andeventsper 100patient-yearsof follow-upreported. Using the same approach, wefurther subdivided both treatment arms,considering patients on “insulinotropic”therapies (insulin, sulfonylureas, or nonsul-fonylurea insulin secretagogues) comparedwith those without such drugs, and therandomized treatment–by–insulinotropictherapy interaction was evaluated.For investigation of the association

between SHEs and CV events, Cox pro-portional hazards regression modelswere fittedwith SHEs as time-dependentvariables and subsequent MACE, as wellas fatal/nonfatalMI or stroke, hACS, hHF,and ACM, as outcomes. For investigationof the reverse association, Cox modelswere fitted with nonfatal CV events astime-dependent variables and subse-quent SHEs as outcomes. For both anal-yses, unadjusted as well as partially andfullyadjustedmodelswereused.Partiallyadjusted models included randomizedtreatment and clinical factors of age,sex, ethnicity, weight, and current smok-ing, while models fully adjusted for in-dependent predictors of MACE and ACMin EXSCEL (19) included age, sex, ethnic-ity, HbA1c, New York Heart Associationclass, current smoking, randomized treat-ment, MI, CV disease, stroke, $50% ste-nosis in carotid artery, atrial fibrillationor flutter, systolic blood pressure, dia-stolic blood pressure, heart rate, height,BMI, eGFR, diabetes duration, insulintherapy at baseline, time-dependent in-sulin use during the trial, chronic res-piratory disease, amputation, diabeticneuropathy, and foot ulcers as covari-ates.Models for SHE outcomeswere alsoadjusted for baseline b-blockers and sul-fonylureas. For the continuous variables,we checked whether nonlinear terms(piecewise splines) needed to be in-cluded in the models. Where piecewisesplines for continuous variables werenecessary, a cut point was selected thatwould work reasonably well for all endpoints. The proportional hazards as-sumption was checked for the full ad-justmentmodelwithnomajor violationsbeing identified.

Models for CV outcomes used only thefirst SHE per patient and assumed thattherewereno time-dependent confound-ers associated with both SHEs and clinicaloutcomes. Events per 100 patient-yearsof follow-up are presented separately forthe time fromfirst SHE to clinical outcomeand for time to clinical outcome or censorwithout an SHE. Results are displayed asforest plots. Analyses of the associationbetween nonfatal CV outcomes and sub-sequent SHEs were conducted and pre-sented similarly.

To investigate the time dependence ofthe riskofCVevents after SHEs, nonlinearrestricted cubic spline functions of timesince SHE were included in fully adjustedproportional hazards regression models.These functions were exponentiated toget HRs associatedwith SHE as a functionof time since SHE and plotted for therange of follow-up times postevent. Riskof SHEas a functionof timesincenonfatalCV events was investigated using thesame approach. P values for whetherthe HRs were constant over time areprovided. Data were analyzed using SASsoftware, version 9.4 (SAS Institute, Cary,NC), with P values ,0.05 consideredstatistically significant.

RESULTS

The EXSCEL intention-to-treat popula-tion consisted of 14,752 patients (ofwhom 10,782 [73.1%] had previous CVdisease) with a median follow-up of 3.2years (interquartile range 2.224.4). Atotal of 14,187 participants (96.2%) com-pleted the trial, and vital status wasobtained for 98.8% of the participants.Themedian durationof exposure to studymedicationwas 2.4 years (1.423.8) in theEQWgroupand2.3years (1.223.6) in theplacebo group. No patients were ex-cluded from the trial because of a historyof two SHEs.

SHEs were relatively uncommon andnot associated with EQW therapy (N 5247 [3.4%], 1.0/100 patient-years, vs.N 5 219 [3.0%], 0.9/100 patient-yearsfor placebo; HR 1.13 [95% CI 0.9421.36],P 5 0.179) (Fig. 1A). Subgroups with orwithout baseline therapy that includedinsulin, sulfonylurea, or nonsulfonylureasecretagogues also showed no differ-ence between the two treatment arms(HR 1.12 [95% CI 0.9321.36]) for thesubgroup with insulin, sulfonylurea, ornonsulfonylurea secretagogues and HR1.24 [95% CI 0.6122.51] for the group

without; P 5 0.80 for interaction) (Fig.1B). Incidence rates for SHEswere higherin the 1st year than thereafter in bothtreatmentgroupsandwerehighest in thesubgroup with insulin, sulfonylurea, ornonsulfonylurea secretagogue therapyat baseline compared with those onother therapies at baseline (HR 4.17[95% CI 2.90–6.02]). Recurrent SHEswere balanced between treatment arms,seen in 71 placebo participants and 62EQW participants. There was a total of450 events (1.8/100 patient-years) in theplacebo group and 404 in the EQW group(1.6/100 patient-years). Two participantsin the placebo group and four in the EQWgroup discontinued study drug because ofSHEs.

Baseline characteristics for those withand without an SHE during the trial arelisted in Table 1. Participantswith an SHE,compared with those without, had lon-ger diabetes duration and lower eGFR,more frequently had prior CV or heartfailure events, and were more oftennonwhite or insulin treated and with ahigher daily dose per kilogram of bodyweight. Furthermore, participants withSHEs were more often treated with di-uretics, statins,b-blockers, and antithrom-botic agents. Of the 466participantswithSHEs, 15 (3.2%) died of non-CV causesand 116 (24.9%) had a CV event (i.e., MI,stroke, hHF, hACS, or CV death). Of these116 participants, 66 had these eventsafter an SHE during the trial (with fourparticipants having an SHE on the sameday) and 50 participants had an SHEafter a nonfatal CV or hHF event duringthe trial. The baseline characteristics ofthe latter two groups are also listed inTable 1, compared with the 14,636 EX-SCEL participants who did not have anSHE or a CV event during the trial. Thosetwo groups with, compared with thosewithout, concomitant SHEs and CVevents during the trial showed rathersimilar characteristics andwere;4 yearsolder, had a 4-year longer duration ofdiabetes, and were almost twice as likelyto be insulin treated. In addition, theirdaily insulin dose at baseline was highercompared with those without the dualrisk (0.82 vs. 0.62 units/kg body wt) (seealso Supplementary Table 1). They com-prised fewer women; were more oftenblack; tended to have a higher BMI andHbA1c; had twice the proportion withchronic kidney disease stage 3 (eGFR 30–59 mL/min/1.73 m2); more frequently

care.diabetesjournals.org Standl and Associates 645

reported a history of CV disease, heartfailure, and prior CV events; showed amarkedly higher Charlson ComorbidityIndex; and were more often treatedwith acetylsalicylic acid, thienopyridines,statins, diuretics, and b-blockers. Thesetrends were most pronounced in thegroup with an SHE before a CV event,with 97% having experienced a prior CVevent, 41% with a prior heart failurehistory, and 46% with chronic kidneydisease stage 3. Those patients alsohad the highest daily insulin dose (0.87units/kgbodywt). Of the fourparticipantswith an SHE and a CV event occurring on

the same day, three were taking insulin atbaseline and the other had commencedinsulin therapyprior to theevents. Twodiedwithin 2 days from CV death and non-CVdeath, while the other two remained freefrom further events.

In analyses adjusted for selected clin-ical factors, SHEs were associated with asubsequent increased risk of the primaryMACE end point (HR 1.43 [95% CI 1.08–1.91], P5 0.013), ACM (HR 2.06 [95% CI1.57–2.71],P,0.001), CVdeath (HR1.84[95% CI 1.28–2.62], P 5 0.001), MI (HR1.68 [95%CI1.18–2.39],P50.004), hACS(HR 1.66 [95% CI 1.19–2.30], P5 0.003),

and hHF (HR 2.88 [95%CI 1.92–4.33],P,0.001) (Fig. 2A). Conversely, nonfatal MIor stroke (HR 2.50 [95% CI 1.75–3.56],P, 0.001), nonfatal MI (HR 2.52 [95% CI1.70–3.74], P , 0.001), nonfatal stroke(HR 2.56 [95% CI 1.40–4.70], P5 0.002),hACS (HR 2.42 [95% CI 1.68–3.48], P ,0.001), and hHF (HR 4.40 [95% CI 2.70–7.15], P , 0.001) were associated with asubsequent increasedriskofSHEs (Fig.2B).

In fully adjusted models, statisticallysignificant associations of SHEs with sub-sequent eventswere limited to CVdeath,ACM, and hHF events (Fig. 2C), whileall associations between nonfatal CVevents and subsequent increased riskof SHEs remained statistically significant(Fig. 2D).

After adjustment for the full list ofcovariates, the elevated hazards ofcombined CV events (MACE/hACS/hHF)following SHEs and of SHEs followingcombined nonfatal CV events (nonfatalMACE/hACS/hHF) were approximatelyconstant over time (Fig. 3A and B). Thehazard for ACM following hypoglycemia,however, changed over time (P, 0.001),beinggreatest(approximatelyfivefold)soonafter an SHE, decreasing over the first 2years to about normal levels and increasingagain after;3yearspost-SHE, although thenumbers at risk beyond 3 years decreasedand 95%CIs widened substantially (Fig. 3C).A similar patterndalbeit only of borderlinesignificancedwas seen with single time-dependent hazard plots for MACE and CVdeath after an SHE (P 5 0.031 for MACEand 0.056 for CV death) (SupplementaryFig. 1) but not for hACS or hHF.

CONCLUSIONS

The primary objective of this workdtovalidate the hypothesis of a bidirec-tional relationship for an increased riskof CV events after SHEs as well as anincreased risk of SHEs after CV events inpatients with type 2 diabetesdwas met.We found consistent evidence againconfirming a greater risk of CV death,ACM, or hHF after SHEs, with also asubstantially higher risk of SHEs after anonfatal CV event such as nonfatal MI,stroke, hACS, or hHF.Moreover, we havedemonstrated that the elevated hazardsof combined CV events following SHEsand SHEs following combined nonfatalCV events were approximately constantover time, while the specific hazard forall-causedeathdandto someextentalso

Figure 1—Kaplan-Meier plots of time tofirst SHE for participants assigned to exenatide or placebo.A: In the intention-to-treat population. B: Split by those with or without therapy with insulin,sulfonylurea (SU), or nonsulfonylurea secretagogues (NSS) at baseline.

646 Severe Hypoglycemia and CV Events Diabetes Care Volume 43, March 2020

Table 1—Baseline patient characteristics according to CV and SHEs occurring during the study

No SHE during trial(N 5 14,286)

SHE during trial(N 5 466)

CV event before SHEduring trial (N 5 50)

SHE before CV eventduring trial (N 5 66)*

DidnothaveSHEandCVeventduring trial (N 5 14,636)

Age at randomization(years) 62 (56, 68) 63 (57, 70) 67 (59, 71) 65 (61, 70) 62 (56, 68)

Female sex 5,422 (38.0) 181 (38.8) 15 (30.0) 16 (24.2) 5,572 (38.1)

RaceWhite 10,865 (76.1) 310 (66.5) 33 (66.0) 50 (75.8) 11,092 (75.8)Black 832 (5.8) 46 (9.9) 6 (12.0) 5 (7.6) 867 (5.9)Asian 1,402 (9.8) 50 (10.7) 4 (8.0) 6 (9.1) 1,442 (9.9)Other 1,182 (8.3) 60 (12.9) 7 (14.0) 5 (7.6) 1,230 (8.4)

Duration of type 2diabetes (years) 12 (7, 17) 14 (9, 22) 16 (11, 22) 17 (10, 24) 12 (7, 18)

History of CV diseaseCoronary artery disease 7,505 (52.5) 289 (62.0) 36 (72.0) 55 (83.3) 7,703 (52.6)Cerebrovascular

disease 2,419 (16.9) 90 (19.3) 11 (22.0) 23 (34.8) 2,475 (16.9)Peripheral artery

disease 2,712 (19.0) 88 (18.9) 13 (26.0) 12 (18.2) 2,775 (19.0)

CV eventPrior CV event 10,421 (72.9) 361 (77.5) 43 (86.0) 64 (97.0) 10,675 (72.9)CV disease without

prior event 143 (1.0) 6 (1.3) 0 (0) 0 (0) 149 (1.0)Neither prior CV event

nor disease 3,722 (26.1) 99 (21.2) 7 (14.0) 2 (3.0) 3,812 (26.0)

Heart failure 2,291 (16.0) 98 (21.0) 16 (32.0) 27 (40.9) 2,346 (16.0)

Charlson ComorbidityIndex 5 (4, 7) 6 (5, 8) 7 (5, 9) 8 (6, 9) 5 (4, 7)

Systolic blood pressure(mmHg) 135 (124, 145) 135 (125, 148) 136 (127, 145) 131 (120, 145) 135 (124, 145)

Diastolic blood pressure(mmHg) 80 (70, 85) 78 (70, 82) 77 (68, 82) 77 (65, 81) 80 (70, 85)

Heart rate (bpm) 72 (66, 80) 72 (65, 80) 74 (64, 80) 73 (66, 84) 72 (66, 80)

BMI (kg/m2) 31.8 (28.3, 36.2) 31.6 (27.9, 36.5) 32.7 (29.7, 36.9) 33.3 (28.6, 37.5) 31.8 (28.2, 36.2)

HbA1c (%) 8.0 (7.3, 8.9) 8.1 (7.5, 8.9) 8.6 (7.7, 9.4) 8.0 (7.3, 9.1) 8.0 (7.3, 8.9)

Total cholesterol (mg/dL) 166 (138, 201) 163 (137, 197) 164 (138, 200) 146 (123, 176) 166 (138, 201)

LDL (mg/dL) 88 (66, 116) 87 (65, 119) 80 (60, 104) 77 (62, 104) 88 (66, 116)

HDL (mg/dL) 42 (35, 50) 42 (35, 49) 42 (34, 49) 38 (32, 44) 42 (35, 50)

Triglycerides (mg/dL) 159 (114, 228) 163 (106, 230) 170 (106, 255) 162 (104, 222) 159 (114, 228)

eGFR (mL/min/1.73 m2) 77 (61, 92) 70 (55, 86) 64 (54, 87) 61 (45, 80) 76 (61, 92)$90 4,172 (29.3) 96 (20.6) 11 (22.0) 9 (13.6) 4,248 (29.1)60–89 7,027 (49.3) 219 (47.1) 19 (38.0) 27 (40.9) 7,200 (49.4)30–59 3,027 (21.3) 150 (32.3) 20 (40.0) 30 (45.5) 3,127 (21.4),30 14 (0.1) 0 (0) 0 (0) 0 (0) 14 (0.1)

CV medicationsAspirin 9,055 (63.4) 325 (69.7) 38 (76.0) 58 (87.9) 9,284 (63.4)Thienopyridines 2,426 (17.0) 98 (21.0) 16 (32.0) 24 (36.4) 2,484 (17.0)Any antiplatelets 10,472 (73.4) 363 (77.9) 45 (90.0) 60 (90.9) 10,730 (73.4)ACEI or ARB 11,024 (77.2) 374 (80.3) 38 (76.0) 52 (78.8) 11,308 (77.3)b-Blockers 7,933 (55.5) 278 (59.7) 36 (72.0) 49 (74.2) 8,126 (55.5)Calcium channel

blockers 4,550 (31.8) 160 (34.3) 14 (28.0) 28 (42.4) 4,668 (31.9)Any antihypertensive 12,894 (90.3) 429 (92.1) 49 (98.0) 65 (98.5) 13,209 (90.3)Statin 10,479 (73.4) 366 (78.5) 41 (82.0) 59 (89.4) 10,745 (73.4)Any lipid-lowering

medication 10,992 (76.9) 378 (81.1) 42 (84.0) 59 (89.4) 11,269 (77.0)Diuretics 6,208 (43.5) 235 (50.4) 34 (68.0) 46 (69.7) 6,363 (43.5)

Continued on p. 648

care.diabetesjournals.org Standl and Associates 647

for CV death and MACEdfollowing anSHE changed in a time-dependent way,with highest excess risk closest to thetime the SHE occurred. Notably, thoseindividuals at dual risk of SHEs and CVeventsdregardless of the time sequencedwere characterized by much higher CharlsonComorbidity Index scores comparedwith those without.

TheseEXSCEL-derived results aremorerobust than those for TECOS (8), as thenumber of patients with concomitantSHEs and a CV event was ;70% larger.As with the TECOS results, but unlikemost other reports regarding SHEs andCV events, these EXSCEL results wereadjusted for all major predictors of CVoutcomes in the trial including prior CV

disease, heart failure, eGFR, duration ofdiabetes, insulin therapy at baseline, andnew insulin use during the trial. Also, theplots of time-dependent HRs weredone with full adjustment. Neverthe-less, the finding of a two- to threefoldincreased risk for SHEs after nonfatal CVevents persisted and remained approx-imately constant over time, while the

Table 1—Continued

No SHE during trial(N 5 14,286)

SHE during trial(N 5 466)

CV event before SHEduring trial (N 5 50)

SHE before CV eventduring trial (N 5 66)*

DidnothaveSHEandCVeventduring trial (N 5 14,636)

Diabetes medicationsBiguanide 10,982 (76.9) 313 (67.2) 29 (58.0) 39 (59.1) 11,227 (76.7)Sulfonylurea 5,253 (36.8) 148 (31.8) 11 (22.0) 14 (21.2) 5,376 (36.7)Insulin 6,509 (45.6) 327 (70.2) 40 (80.0) 52 (78.8) 6,744 (46.1)Dose (units/kg

body wt) 0.62 (0.39, 0.89) 0.73 (0.45,1.07) 0.72 (0.43, 1.12) 0.87 (0.52, 1.19) 0.62 (0.39, 0.90)DPP-4i 2,148 (15.0) 55 (11.8) 1 (2.0) 6 (9.1) 2,196 (15.0)TZD 565 (4.0) 14 (3.0) 0 (0) 4 (6.1) 575 (3.9)Other 610 (4.3) 14 (3.0) 0 (0) 3 (4.5) 621 (4.2)Insulin/sulfonylurea/

NSS 11,005 (77.0) 435 (93.3) 48 (96.0) 63 (95.5) 11,329 (77.4)

Data are n (%) or median (25th, 75th percentile). ACEI, ACE inhibitor; ARB, angiotensin receptor blocker; DPP-4i, dipeptidyl peptidase 4 inhibitor; NSS,nonsulfonylurea secretagogues; TZD, thiazolidinedione. *Four patients with CV event and SHE on the same day are included with the SHE before CVevent group.

Figure2—Forestplots showing theassociationofSHEsandsubsequent riskofCVoutcomes (AandC) andnonfatalCVeventsandsubsequent riskofSHEs(B andD). Models forA and Bwere adjusted for clinical factors (randomized treatment, age, sex, ethnicity, weight, and smoking), whereasmodels for CandDwere fully adjusted for age, sex, ethnicity, HbA1c level, New York Heart Association class, current smoking, history ofMI, coronary artery disease,stroke, stenosisof carotidartery, atrialfibrillation, chronicobstructivepulmonarydisease, amputation,diabeticneuropathy, footulcers,baseline insulintherapy, systolic blood pressure, diastolic blood pressure, heart rate, height, BMI, eGFR, diabetes duration, randomized treatment, and insulin use asa time-dependent variable. Models for panel D were additionally adjusted for baseline b-blockers and baseline sulfonylurea. ACS, acute coronarysyndrome; HF, heart failure.

648 Severe Hypoglycemia and CV Events Diabetes Care Volume 43, March 2020

increased risk of 60% for CV death and;80% for all-cause death after an SHE isquite comparable with other publica-tions including the finding of these haz-ards beingmost frequent close to an SHE

(3,4,6,7,9,20,21). The significant associ-ation of hHF in both directions with SHEsis another interesting new observation.Importantly, assigned treatment withEQW was not associated with different

rates of SHEs between randomized studygroups.

Although associations found in epide-miologic patient cohorts are not able toprove a causal relationship between the

A B

C

Figure 3—Time-dependent HR plots with 95% CIs and numbers at risk for combined CV events (MACE/hACS/hHF). A: Occurring after an SHE. B: For anSHE after a nonfatal CV event (MACE/hACS/hHF). C: For all-cause death after an SHE. P values for the test of constant HR over time are provided.

care.diabetesjournals.org Standl and Associates 649

parameters studied, the bidirectional re-lationship between SHEs and CV out-comes, together with a higher CharlsonComorbidity Index of;8 in those at dualrisk of SHEs and CV events, suggests thatthere may be a common type 2 diabetesphenotype of patients with multiple co-morbidities and likely featuresof “frailty”who are susceptible to both of theseevents. Thus, SHEs in many, if not most,instancesdrather than being causativeof CV death, hHF, or all-cause deatheventsdmay simply be indicative of mul-timorbid or “frail” patients who are athigher risk of both outcomes likely dueto a large accumulation of coexisting riskfactors. That there were four patients inEXSCEL with same-day events does notseem to exclude or confirm a causal roleof SHEs for CV outcomes.These observations of EXSCEL data

also shed light on the emerging charac-teristics of this patient phenotype at dualrisk for SHEs and CV events. Comparedwith thosewithout this dual risk, patientsare considerably older, have a 4-year-longer duration of diabetes, and aremore likely (80% vs. 37%) to be insulintreated (and with a higher dose) at studyentry, with a further increase of insulinusers within the trial. Almost all had aprior CV event either at study baselinepreceding the first SHE in the study oroccurringbefore anSHEduring the study.Some 43% had chronic kidney diseasestage 3 at study entry, and more thanone-third had a history of heart failure,with all these factors adding up to in-creased SHE risk and risk of fatal andnonfatal CV events, hHF events, andACM. The high comorbidity score forthese patients, as evidenced by a highCharlson Comorbidity Index, seems alsoto be reflected by their much greater useof CV medications, especially of antipla-telet therapies,b-blockers, diuretics, andstatins. Of note, baseline HbA1c concen-trations of these patients did not differfrom those without dual risk or evenshowed a trend to higher levels.These characteristics of an emerging

“vulnerable” phenotype at dual risk forSHEs and CV events are mirrored by therecently studied populations at dual SHEand CV event risk in the Liraglutide Effectand Action in Diabetes: Evaluation ofCardiovascular Outcome Results (LEADER)trial and the Trial Comparing Cardiovascu-lar Safetyof InsulinDegludecVersus InsulinGlargine in PatientsWith Type2Diabetes

at High Risk of Cardiovascular Events(DEVOTE), as well as in TECOS and theVeterans Affairs Diabetes Trial (VADT)(7–9,20). Epidemiological observationsfrom the Hong Kong Diabetes Registryalsoseemtosupport thenovel conceptoftheexistenceof suchaparticularly frail orvulnerable patient phenotype (22,23).Remarkably, a recent publication fromthe Atherosclerosis Risk in Communities(ARIC) cohort study group looking atsevere hypoglycemia, mild cognitive im-pairment, dementia, and brain volumesin older adults with type 2 diabetesdescribed a very similar at-risk patientphenotype with longer diabetes dura-tion,olderage, andmorecommon insulinuse (24). Finally, the characteristics ofa vulnerable patient phenotype at dualrisk for SHEs and CV events seeminglyresemble a recently suggested “renal”subphenotype cluster of adult-onsetdiabetes (25).

Mechanistically, SHEs have been as-sociatedwith hypokalemia, prolongationof theQTc interval on electrocardiogram,and neuro-sympathetic overdrive withmarked increase of blood pressure,tachycardia, and various kinds of ar-rhythmias (18,26–29). Moreover, SHEsactivate prothrombotic and proinflam-matory pathways, factors that all addup to a theoretical increased CV risk(21,30–33). Our results are compatiblewith the notion that these pathophysio-logical mechanismsmay have been criticaltosomeindividualpatientsdperhapsevenbeyond the CV contextdespecially whenconsidering the segment section closestto the SHE of the time relationship withall-cause death and, to a lesser extent,with CV death. Alternatively, the closetime association between an SHE andnon-CV death in some patients mayjust indicate excessive comorbidities inthose. In this context, the example of thefour patients with same-day dual SHEsand CV events shows the complexity anddiversity of such a potential interaction.Furthermore, in others, an SHE had oc-curred long before a CV event, render-ing a causative role of SHEs for CVevents in these patients rather unlikely.The gradual recurrenceof an increased riskfor ACM beyond 3 years after an SHEdalthough with numbers at risk declining,which limits any firm conclusionsdwasan unexpected finding potentially alludingto a persisting high comorbidity score thatshould be further explored.

In terms of SHEs occurring after a CVevent including hHF, brain natriureticpeptides may play an important role,as theydamong many other effectsdactivate insulin-sensitizing mechanismsat the level of skeletalmuscle and adiposetissue and increase circulating concentra-tions of the insulin-sensitizing hormoneadiponectin (34,35).Anaugmentingeffectof atrial natriuretic peptide on insulin-induced hypoglycemia has long beenknown (36). The strongest associationsseen in our analyses were between hHFand SHEs in both settings of SHEs, beforeand after a CV event. History of knownheart failure at baseline was rather fre-quent at baseline in both groups withdual SHE and CV events (32% and 41%,respectively). Potentially, a similarly highproportion of patients with heart failurego unrecognized in most CVOTs focusingon patients with established CV diseaseunless natriuretic peptide levels aremea-sured (37). Heart failure patientswith thehighest levels of both natriuretic pepti-des and adiponectin have been found tohave the highest mortality risk (38). Inthis context, continued use of insulin inthose patients may be questioned.

Strengths of our study include the sizeand length of follow-up with proactivecollection and independent, blinded ad-judication of all relevant events in thecontextofaCVOT.Thenumberofpatientswith dual SHEs and CV events (n 5 116)is to our knowledge the largest reportedfrom a CVOT thus far. Moreover, ourresults appear to be particularly robustbecause of adjustment for a rather com-prehensive list of potential confounders.Though an evaluation of an a priori hy-pothesis, study weaknesses include thepost hoc nature of the analysis and thelack of available biomarkers such asmeas-urements of natriuretic peptides to fur-ther substantiate potential pathogeniclinks. Moreover, whether less severe hy-poglycemic episodes also might have im-pacted the results could not be evaluated,as theywere not systematically evaluatedin the pragmatic EXSCEL trial. This seems,however, rather unlikely in light of neutralfindings related to less severe hypoglyce-mia in studies such as the Outcome Re-duction With Initial Glargine Intervention(ORIGIN) or LEADER trials (6,9). Neverthe-less, residual confounding cannot be ex-cluded. Finally, a formal assessment offrailty was not included in the EXSCELstudy protocol.

650 Severe Hypoglycemia and CV Events Diabetes Care Volume 43, March 2020

In summary, we report a robust bi-directional association in EXSCEL partic-ipants confirming a greater risk of SHEsafter CV events as well as a greater risk ofCV or all-cause death events after SHEs.These findings support our view of anexisting common at-risk polymorbid andpotentially frail type 2 diabetes patientphenotype, susceptible to both SHEs andCV events. These vulnerable patientswith high comorbidity scores after a firstCV event in conjunction with heart fail-ure, advanced kidney disease, older age,longer duration of diabetes, and ten-dency to be on insulin and on higherdoses warrant special consideration, in-cluding careful dosing of insulin therapy.

Acknowledgments. Peter Hoffmann, an em-ployee of DCRI, provided editorial support.R.R.H. is an Emeritus National Institute forHealth Research Senior Investigator.Funding and Duality of Interest. EXSCEL wassponsored and funded by Amylin Pharmaceut-icals (San Diego, CA), a wholly owned subsidiaryof AstraZeneca (Gaithersburg, MD). E.S. reportspersonal fees from Oxford Diabetes Trials Unit,Bayer, Berlin Chemie, Boehringer Ingelheim,Menarini, Merck Serono, EXCEMED, Novartis,Novo Nordisk, and Sanofi. Y.L. reports grantsfrom Merck, Janssen Research & Development,AstraZeneca, GlaxoSmithKline, andBayerHealth-Care AG. M.A.B. reports research support fromMerckandAstraZeneca;participatingonadvisoryboards for Boehringer Ingelheim and NovoNordisk; receiving honoraria, personal fees,and other support from Merck, Novo Nordisk,AstraZeneca, and Sanofi; and receiving nonfi-nancial research support from Bayer andMerck Serono; M.A.B. is now an employee ofEli Lilly and Co. J.B.B. reports that contractedadvisory fees are paid to the University ofNorth Carolina by Adocia, AstraZeneca, DanceBiopharm, Eli Lilly, MannKind, NovaTarg, NovoNordisk, Senseonics, vTv Therapeutics, andZafgen; reports grant support from Novo Nordisk,Sanofi, and vTv Therapeutics; is a consultant toNeurimmune AG; and owns stock options inMellitus Health, PhaseBio, and Stability Health.J.B.B. is supported by a grant from the NationalInstitutes of Health (UL1TR002489). S.M.G. isan employee of AstraZeneca. A.P.M. reportshonoraria from Bayer, Novartis, Cardiorentis,and Fresenius Medical Care for participationin study committees. R.J.M. reports grantsfrom Merck, AstraZeneca, and GlaxoSmithKlineand personal fees fromMerck, AstraZeneca, andBoehringer Ingelheim. A.F.H. reports researchfunding from Amgen, Amylin, AstraZeneca,Daiichi Sankyo, Genentech, GlaxoSmithKline,Luitpold, and Merck and consulting fees fromAmgen, AstraZeneca, Bayer, Boehringer Ingelheim,Boston Scientific, Merck, MyoKardia, Novartis, andPluristem Therapeutics. R.R.H. reports researchsupport from AstraZeneca; grants and personalfees from Bayer, Boehringer Ingelheim, and MerckSharp & Dohme, a subsidiary of Merck & Co.;

personal fees from Novartis, Amgen, and Servier;and financial support from Elcelyx Therapeutics,GlaxoSmithKline, Janssen, and Takeda. No otherpotential conflicts of interest relevant to this articlewere reported.Author Contributions. E.S. and R.R.H. equallycontributed to the study design and data analysisand interpretation and drafted and edited themanuscript. S.R.S. and Y.L. performed the sta-tistical analysis and edited the manuscript.M.A.B., J.B.B., S.M.G., A.P.M., R.J.M., andA.F.H. edited the manuscript. E.S. and R.R.H.are the guarantors of this work and, as such, hadfull access to all the data in the study and takeresponsibility for the integrity of the data and theaccuracy of the data analysis.Prior Presentation. Parts of this study werepresented in abstract form at the Diabetes UKProfessional Conference 2019, Liverpool, U.K.,6–8 March 2019, and American College of Car-diology’s 68th Annual Scientific Session & Expo,New Orleans, LA, 16–18 March 2019.

References1. Zaccardi F,DhalwaniNN,WebbDR,DaviesMJ,Khunti K. Global burden of hypoglycaemia-related mortality in 109 countries, from 2000 to2014: an analysis of death certificates. Diabeto-logia 2018;61:1592–16022. DuckworthW,AbrairaC,Moritz T, et al.; VADTInvestigators. Glucose control and vascular com-plications in veterans with type 2 diabetes[published correction appears in N Engl J Med2009;361:1028; N Engl J Med 2009;361:1024–1025]. N Engl J Med 2009;360:129–1393. Zoungas S, Patel A, Chalmers J, et al.; AD-VANCE Collaborative Group. Severe hypoglyce-mia and risks of vascular events anddeath.N EnglJ Med 2010;363:1410–14184. BondsDE,MillerME,Bergenstal RM,et al. Theassociation between symptomatic, severe hypo-glycaemia and mortality in type 2 diabetes:retrospective epidemiological analysis of theACCORD study. BMJ 2010;340:b49095. Mellbin LG, Malmberg K, Waldenstrom A,Wedel H, Ryden L; DIGAMI 2 Investigators. Prog-nostic implications of hypoglycaemic episodesduring hospitalisation for myocardial infarctioninpatientswith type2diabetes: a report fromtheDIGAMI 2 trial. Heart 2009;95:721–7276. Mellbin LG, Ryden L, RiddleMC, et al.; ORIGINTrial Investigators. Does hypoglycaemia increasethe risk of cardiovascular events? A report fromthe ORIGIN trial. Eur Heart J 2013;34:3137–31447. Pieber TR, Marso SP, McGuire DK, et al.;DEVOTE Study Group. DEVOTE 3: temporal rela-tionships between severe hypoglycaemia, car-diovascular outcomes and mortality. Diabetologia2018;61:58–658. Standl E, Stevens SR, Armstrong PW, et al.;TECOS Study Group. Increased risk of severehypoglycemic events before and after cardio-vascular outcomes in TECOS suggests an at-risktype 2 diabetes frail patient phenotype. DiabetesCare 2018;41:596–6039. ZinmanB,MarsoSP,ChristiansenE,CalannaS,Rasmussen S, Buse JB; LEADER Publication Com-mittee on behalf of the LEADER Trial Investiga-tors. Hypoglycemia, cardiovascular outcomes,and death: the LEADER experience. DiabetesCare 2018;41:1783–1791

10. Ryden L, Grant PJ, Anker SD, et al.; Authors/TaskForceMembers; ESCCommittee forPracticeGuidelines (CPG); Document Reviewers. ESCguidelines on diabetes, pre-diabetes, andcardiovascular diseases developed in collabora-tion with the EASD: the Task Force on diabetes,pre-diabetes, and cardiovascular diseases of theEuropean Society of Cardiology (ESC) and de-veloped in collaboration with the EuropeanAssociation for the Study of Diabetes (EASD).Eur Heart J 2013;34:3035–308711. Inzucchi SE, Bergenstal RM, Buse JB, et al.Management of hyperglycemia in type 2 di-abetes, 2015: a patient-centered approach:update to a position statement of the AmericanDiabetes Association and the European Asso-ciation for the Study of Diabetes. Diabetes Care2015;38:140–14912. Zoungas S, Arima H, Gerstein HC, et al.;Collaborators on Trials of Lowering Glucose(CONTROL) Group. Effects of intensive glucosecontrol on microvascular outcomes in patientswith type 2 diabetes: a meta-analysis of in-dividual participant data from randomised con-trolled trials. Lancet Diabetes Endocrinol 2017;5:431–43713. Munshi MN, Slyne C, Segal AR, Saul N, LyonsC,Weinger K. Liberating A1C goals in older adultsmaynot protect against the risk of hypoglycemia.J Diabetes Complications 2017;31:1197–119914. Riddle MC, Ambrosius WT, Brillon DJ, et al.;Action to Control Cardiovascular Risk in DiabetesInvestigators. Epidemiologic relationships be-tweenA1Candall-causemortalityduringamedian3.4-year follow-up of glycemic treatment in theACCORD trial. Diabetes Care 2010;33:983–99015. Gerstein HC, Miller ME, Genuth S, et al.;ACCORD Study Group. Long-term effects of in-tensive glucose lowering on cardiovascular out-comes. N Engl J Med 2011;364:818–82816. Holman RR, Bethel MA, Mentz RJ, et al.;EXSCEL Study Group. Effects of once-weeklyexenatide on cardiovascular outcomes in type 2diabetes. N Engl J Med 2017;377:1228–123917. Charlson ME, Pompei P, Ales KL, MacKenzieCR. A new method of classifying prognostic co-morbidity in longitudinal studies: developmentand validation. J Chronic Dis 1987;40:373–38318. Austin SR, Wong YN, Uzzo RG, Beck JR,Egleston BL. Why summary comorbidity meas-ures such as the Charlson Comorbidity Indexand Elixhauser score work. Med Care 2015;53:e65–e7219. Mentz RJ, BethelMA,Merrill P, et al.; EXSCELStudyGroup. Effect of once-weekly exenatide onclinical outcomes according to baseline risk inpatients with type 2 diabetes mellitus: insightsfrom the EXSCEL trial. J Am Heart Assoc 2018;7:e00930420. Davis SN, Duckworth W, Emanuele N, et al.;Investigatorsof theVeteransAffairsDiabetesTrial.Effects of severe hypoglycemia on cardiovascularoutcomes and death in the Veterans Affairs Di-abetes Trial. Diabetes Care 2019;42:157–16321. International Hypoglycaemia Study Group.Hypoglycaemia, cardiovascular disease, andmortality in diabetes: epidemiology, pathogen-esis, and management. Lancet Diabetes Endo-crinol 2019;7:385–39622. Kong APS, Yang X, Luk A, et al. Hypoglycae-mia, chronic kidney disease and death in type

care.diabetesjournals.org Standl and Associates 651

2diabetes: theHongKongdiabetes registry. BMCEndocr Disord 2014;14:4823. Kong APS, Yang X, Luk A, et al. Severehypoglycemia identifies vulnerable patientswith type 2 diabetes at risk for premature deathand all-site cancer: the Hong Kong diabetesregistry. Diabetes Care 2014;37:1024–103124. Lee AK, Rawlings AM, Lee CJ, et al. Severehypoglycaemia, mild cognitive impairment,dementia and brain volumes in older adultswith type 2 diabetes: the Atherosclerosis Riskin Communities (ARIC) cohort study. Diabeto-logia 2018;61:1956–196525. Ahlqvist E, StormP,KarajamakiA, et al.Novelsubgroups of adult-onset diabetes and theirassociation with outcomes: a data-driven clusteranalysis of six variables. Lancet Diabetes Endo-crinol 2018;6:361–36926. Desouza C, Salazar H, Cheong B, Murgo J,Fonseca V. Association of hypoglycemia andcardiac ischemia: a study based on continuousmonitoring. Diabetes Care 2003;26:1485–148927. Adler GK, Bonyhay I, Failing H, Waring E,Dotson S, Freeman R. Antecedent hypoglycemiaimpairs autonomic cardiovascular function:

implications for rigorous glycemic control.Diabetes 2009;58:360–36628. Tsujimoto T, Yamamoto-Honda R, Kajio H,et al. Vital signs, QT prolongation, and newlydiagnosed cardiovascular disease during severehypoglycemia in type 1 and type 2 diabeticpatients. Diabetes Care 2014;37:217–22529. Stahn A, Pistrosch F, Ganz X, et al. Relation-ship between hypoglycemic episodes and ven-tricular arrhythmias in patients with type 2diabetes and cardiovascular diseases: silent hy-poglycemias and silent arrhythmias. DiabetesCare 2014;37:516–52030. Desouza CV, Bolli GB, Fonseca V. Hypogly-cemia, diabetes, and cardiovascular events. Di-abetes Care 2010;33:1389–139431. Frier BM, Schernthaner G, Heller SR. Hypo-glycemia and cardiovascular risks. Diabetes Care2011;34(Suppl. 2):S132–S13732. Cryer PE. Death during intensive glycemictherapy of diabetes: mechanisms and implica-tions. Am J Med 2011;124:993–99633. ConnellyKA,YanAT,LeiterLA,BhattDL,VermaS.Cardiovascular implications of hypoglycemia in di-abetes mellitus. Circulation 2015;132:2345–2350

34. Jordan J, Birkenfeld AL, Melander O, Moro C.Natriuretic peptides in cardiovascular andmetabolic crosstalk: implications for hyper-tension management. Hypertension 2018;72:270–27635. Coue M, Barquissau V, Morigny P, et al.Natriuretic peptides promote glucoseuptake in acGMP-dependent manner in human adipocytes.Sci Rep 2018;8:109736. Jungmann E, Konzok C, Holl E, FassbinderW,Schoffling K. Effect of human atrial natriureticpeptide on blood glucose concentrations andhormone stimulation during insulin-induced hy-poglycaemia in healthyman. Eur J Clin Pharmacol1989;36:593–59737. Standl E, Schnell O, McGuire DK. Heartfailure considerationsof antihyperglycemicmed-ications for type 2 diabetes. Circ Res 2016;118:1830–184338. Tsutamoto T, Tanaka T, Sakai H, et al. Totaland high molecular weight adiponectin,haemodynamics, and mortality in patientswith chronic heart failure. Eur Heart J 2007;28:1723–1730

652 Severe Hypoglycemia and CV Events Diabetes Care Volume 43, March 2020