Analyses of Results FromCardiovascular Safety Trials WithDPP-4 Inhibitors: CardiovascularOutcomes, Predefined SafetyOutcomes, andPooledAnalysis andMeta-analysisDiabetes Care 2016;39(Suppl. 2):S196–S204 | DOI: 10.2337/dcS15-3024

The U.S. Food and Drug Administration requires that the cardiovascular (CV)safety of all new drugs for diabetes be demonstrated through pooled analyses ofphase III studies or specifically designed trials. This requirement prompted severalplacebo-controlled, noninferiority CV safety trials in high-risk patients; to date, allcompleted trials showed that dipeptidyl peptidase (DPP)-4 inhibitors do notincrease or reduce the risk of major CV events. These results apparently contrastwith those of pooled analyses and meta-analyses of previous, smaller trials withmetabolic end points, which had suggested a reduction of risk. However, thedesign of CV trials, which were required to demonstrate safety, is not adequate(for duration, management of concurrent therapies, etc.) for the assessment ofpotential therapeutic benefits. In addition, CV safety trials enroll patients at highrisk of CV events, who are different from those included in earlier trials withmetabolic end points. Differences in characteristics of patients enrolled probablyaccount for most of the discrepancy in CV outcomes between CV safety studiesand earlier trials. The availability of several large-scale trials with longer durationprovides the unique opportunity for assessment of the safety of DPP-4 inhibitorsnot only with respect to major CV events but also with reference to other safetyissues. For example, CV safety trials can be a source of information for pancreatitis,cancer, or hypoglycemia.

In December 2008, the U.S. Food and Drug Administration (FDA) published guidanceprompting pooled analyses and meta-analyses of cardiovascular (CV) events (some-times with post hoc adjudication) observed in trials with metabolic outcomes (1). Ashift of emphasis occurred from short-term, HbA1c-centered phase II–III trials in“healthy” patients with type 2 diabetes to CV safety studies in a more vulnerablepopulation (2). Results from the first trials according to the FDA guidance werepublished in 2013: the SAVOR-TIMI 53 (Saxagliptin Assessment of Vascular Out-comes Recorded in Patients with Diabetes Mellitus–Thrombolysis in MyocardialInfarction 53) and the EXAMINE (Examination of Cardiovascular Outcomes withAlogliptin versus Standard of Care) trials with the dipeptidyl peptidase (DPP)-4inhibitors saxagliptin and alogliptin, respectively (3,4). The publication of theresults of TECOS (Trial Evaluating Cardiovascular Outcomes with Sitagliptin) (5)with sitagliptin, the presentation of the ELIXA [Evaluation of Lixisenatide in Acute

1Diabetology, Careggi Hospital, Florence, Italy2Internal Medicine, Hadassah Medical Center,Jerusalem, Israel3Endocrinology, University of Padua, Padua, Italy

Corresponding author: Edoardo Mannucci,edoardo.mannucci@unifi.it.

This publication is based on the presentationsat the 5th World Congress on Controversies toConsensus in Diabetes, Obesity and Hyperten-sion (CODHy). The Congress and the publicationof this supplement were made possible inpart by unrestricted educational grants fromAstraZeneca.

© 2016 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.

Edoardo Mannucci,1 Ofri Mosenzon,2 and

Angelo Avogaro3

S196 Diabetes Care Volume 39, Supplement 2, August 2016

RECEN

TOUTC

OMESTUDIES

Coronary Syndrome] trial with theglucagon-like peptide 1 (GLP-1) receptoragonist lixisenatide, and the publicationof the BI 10773 (Empagliflozin) Cardio-vascular Outcome Event Trial in Type 2Diabetes Mellitus Patients (EMPA-REGOUTCOME) trial with the sodium–

glucose cotransporter 2 (SGLT2) inhibi-tor empagliflozin (6) followed in 2015.This wave of information on CV safetyof new antidiabetes drugs (ADD) is onlythe beginning (Fig. 1 and Table 1). In-terpretation of the “mountains” ofnew information provided by theselarge-scale studies presents new chal-lenges for the treating physician, the sci-entific world, the pharmaceutical industry,and eventually patients with diabetesthemselves (7).Unexpectedly, in the case of DPP-4

inhibitors the results of earlier trialshad suggested a reduction in the inci-dence of major CV events (8–10), whichwas not confirmed by the specifically

designed CV studies that followed (3–5). The reasons for this discrepancy arecontroversial and will be discussed inthe article.

Primary and Secondary Outcomes: CVSafety Outcomes

CV Outcome Versus CV Safety Trials

The intensification of diabetes treat-ment appears to improve CV outcomesonly in patients with recent-onset dia-betes and after a prolonged latency(11,12). The FDA guidance requires therecruitment of patients at high CV risk,often with prolonged diabetes duration,advanced age, and renal failure (1), whoare the least likely to benefit from anyintervention regarding CV risk factors. Inaddition, most of those patients alreadyreceive multiple treatments for CV dis-ease prevention. In the Steno-2 trial, theuse of statins at the end of the interven-tion in the intensive versus conventionaltreatment arms was 0 vs. 3% at baseline

and 85 vs. 22% at the end of the inter-vention stage of the trial, respectively(13). In SAVOR-TIMI 53, EXAMINE, andTECOS, statins were used at baseline by78.3, 90.4, and 79.9% of the trial pop-ulation, respectively (3–5). The resid-ual CV risk left, in spite of statintreatment, was smaller, and thereforeany CV superiority was more difficultto prove.

The FDA requirements for new ADDdevelopment programs were to excludean unacceptable increased relative CVrisk with an upper limit of 95% CI of,1.3 (1). For this reason, trials are de-signed to show noninferiority betweentreatment arms with respect to CVevents while maintaining similar glycemiccontrol (14): investigators are free tomodify glucose-lowering medicationsother than the experimental drug in or-der to maintain good glycemic controlin both treatment groups, and in fact,between-group differences in HbA1c

Figure 1—Timing of CV safety trials with drugs for type 2 diabetes.

care.diabetesjournals.org Mannucci, Mosenzon, and Avogaro S197

Table

1—CompletedandongoingCV

safety

trials

withADDs

GroupofADD

andtrial

Tested

drug

(comparator

ifnotplacebo)

Patien

ts(N

)MainCVinclusion

criteria

HbA1cinclusion

criteria(%

)BackgroundADDs

Start–actual/planned

completion,m

ean

durationoffollow-up

Prim

aryoutcome

(number

oftarget

even

ts)

DPP

-4inhibitors

SAVOR-TIM

I53

Saxagliptin

16,492

$50

y.o+established

CVDor

$55

(M)o

r$60

(F)y.o+CVDrisk

factors

6.5–12

duringthe

last6months

Allother

than

incretin

based

5/201

0–5/201

3,2.1

years

CVdeath,nonfatalM

I,nonfatalstroke

(1,040

)

EXAMINE

Alogliptin

5,38

0ACSwithin

15–90

days

6.5–11

Allother

than

incretin

based

10/2009

–6/20

13,

1.5years

CVdeath,nonfatalM

I,nonfatalstroke

(550

)TECOS

Sitagliptin

14,724

PreexistingCVD

6.5–8

Mono-ordualtherapy:

metform

in,SU,p

ioglitazone,

insulin

12/2008

–6/20

13,

3.0years

CVdeath,n

onfatalM

Iorstroke,

hospitalizationforUAP(1,300

)

CAROLINA

Linagliptin

(glim

epride)

6,00

0CVDhistory,en

d-organ

dam

age,

$70

y.o,o

rCVDrisk

factors

6.5–8.5or6.5–7.5if

withSU

/glinides

Drugnaıve

ormono-ordual

therapy:metform

in,SU,A

GI

10/201

0–9/20

18

CVdeath,n

onfatalM

Iorstroke,

hospitalizationforUAP(631

)CARMELINA

Linagliptin

8,30

0Albuminuria+CVDorR

F+specific

albuminuria

6.5–

10.0

Allother

than

incretin

based

andSG

LT2Inh

7/201

3–1/201

8CVdeath,n

onfatalM

Iorstroke,

hospitalizationforUAP

OMNEO

N01

8MK31

02(once

weekly)

4,00

0PreexistingCVD

6.5–10

or7–10

(insulin,SU,glinides)

Allother

than

incretin

based

10/201

2–10

/201

7CVdeath,n

onfatalM

Iorstroke,

hospitalizationforUAP

GLP-1

receptor

agonists

ELIXA

Lixisenatide

6,06

8ACS,18

0days

5.5–

11.0

Allother

than

incretin

based

6/201

0–2/201

5,

25.7

months

CVdeath,n

onfatalM

Iorstroke,

hospitalizationforUAP(844

)LEADER

Liraglutide

9,34

0$50

y.o+priorCVD

even

ts/PAD/CRF/

CHFor

$60

y.o+CVDrisk

factors

.7.0

Allother

than

incretin

based

9/201

0–10

/201

5,all

patients

atleast3.5years

CVdeath,n

onfatalM

I,or

nonfatalstroke

(611

)

SUSTAIN

6SemaglutideQW

3,29

7$50

y.o+priorCVDor

$60

y.o+subclinicalCVD

.7.0

Drugnaıve

ormono-ordual

oraltherapy,long-actingor

premixinsulin

2/201

3–1/201

6,upto

148weeks

CVdeath,nonfatalM

I,ornonfatalstroke

EXSC

ELExen

atideQW

14,000

none

6.5–10

0–3oralA

DDs(includingDPP

-4Inh),insulin

+0–

2oralA

DDs

6/201

0–1/201

8,upto

7.5years

CVdeath,n

onfatalM

I,or

nonfatalstroke

FREEDOM

CVO

ITCA65

0exen

atide

Q3–6months

4,00

0History

ofCAD,cerebrovascular

disease,o

rPA

D.6.5

Allother

than

incretin

based

3/201

3–7/201

8,2

years

CVdeath,n

onfatalM

Iorstroke,

hospitalizationforUAP

REW

IND

DulaglutideQW

9,62

2$50

y.o+established

CVD,$55

y.o+subclinicalvasculardisease,

or$60

y.o+$2CVDrisk

factors

#9.5

0–2oralA

DDs6

basal

insulin

6GLP-1

RA

7/201

1–4/201

9,upto

8years(expectedmean

of6.5years)

CVdeath,nonfatalM

I,ornonfatalstroke

AlbiglutideCV

safety

trial

AlbiglutideQW

9,40

0$40

y.o+established

CVD

.7

Allother

than

GLP-1

RA

7/201

5–5/201

9CVdeath,n

onfatalM

I,or

nonfatalstroke

SGLT2inhibitors

EMPA

-REG

OUTC

OME

Empagliflozin

(2[10mg,25

mg]:1)

7,03

4.18

y.o,established

CVD

7–10

(7–9ifADD

naıve)

ADD-naıve

patients,allADDs,

ortheircombination

7/201

0–4/201

5CVdeath,nonfatalM

I,orn

onfatal

stroke

($69

1)

CANVAS

Can

agliflozin

(2[100

mg,30

0mg]:1)

4,40

7.30

y.owithestablished

CVDor

.50

y.oand$2CVrisk

factors

7–10

.5AllADDs

12/200

9–6/20

17

CVdeath,nonfatalM

I,orn

onfatal

stroke

CANVAS-R

Can

agliflozin

5,86

5Established

vascularcomplication

or$2CVrisk

factors

7–10

.5ADD-naıve

patients,allADDs,

ortheircombination

1/201

4–1/201

7Number

ofpatients

with

progressionofalbuminuria

Con

tinu

edon

p.S199

S198 CV Safety Trials With DPP-4 Inhibitors Diabetes Care Volume 39, Supplement 2, August 2016

are small (3–5). Conversely, in superioritytrials the study is designed to highlightthe benefits of drug treatment: forexample, in lipid trials additional lipid-lowering therapies could not be modifiedduring the study, creating a sizable differ-ence in the surrogate marker (e.g., LDLcholesterol) that would hopefully lead toimproved CV outcome.

The FDA required the CV safety trialsto extend for a minimum of 2 years. Alonger trial duration increases the riskfor missing data (6), possibly affectingthe magnitude and even the directionof results (15). The short duration ofCV trials is also driven by market issues,such as patent duration, the extravagantcost of large clinical trials (the three cur-rently available trials with DPP-4 inhibi-tors alone, taken together, accumulated.85,000 patient-years of follow-up),and competition between pharmaceuti-cal companies.

In summary, these trials should becalled “CV safety trials” and not “CV out-come trials” owing to the trial goals anddesign, the patient population chosen,the minimization of glycemic differencesbetween treatment arms, and the rela-tively short duration. The recent findingof a significant benefit of empagliflozinin a similarly designed CV safety trial(6) was completely unexpected; despitethis unpredictable result, the design ofthose trials remains inadequate for ex-ploring treatment benefits.

Choosing the Primary Outcome: MACE

Versus MACE PlusdIs There a Difference?

The FDA guidance specifies that CV endpoints should be prospectively andblindly adjudicated by an independentcommittee in all trials of clinical devel-opment of new drugs. The CV end pointsthat must be adjudicated are CVmortality,myocardial infarction (MI), and stroke,combined in the composite end point, ma-jor adverse CV events (MACEs). Other CVend points specifically mentioned in theFDA guidance that can also be adjudicatedare hospitalization for acute coronarysyndrome (ACS), urgent revascularizationprocedures, “and possibly other end-points” (1). Some of the companies havechosen to add hospitalization for unstableangina, urgent coronary revascularization,and/or hospitalization for heart failure(hHF) to the definition of MACE (Table 1),although these end points are less “clear-cut” and more difficult to define; this

Table

1—Continued

GroupofADD

andtrial

Tested

drug

(comparator

ifnotplacebo)

Patien

ts(N

)MainCVinclusion

criteria

HbA1cinclusion

criteria(%

)BackgroundADDs

Start–actual/planned

completion,m

ean

durationoffollow-up

Prim

aryoutcome

(number

oftarget

even

ts)

CRED

ENCE

Can

agliflozin

3,62

7.30

y.o,urinaryACR30

0–2,00

0mg/g,eG

FR90

–30

mL/min/BSA

.Stab

lemaxim

um

tolerateddose

ofACEI

orARB

6.5–12

AllADDs

2/201

4–1/202

0ESRD,d

oublingofserum

creatinine,renalorCVdeath

DEC

LARE-TIMI

58Dapagliflozin

17,150

$50

y.o+established

CVDor

$55

(M)o

r$60

(F)y.o+CVDrisk

factors

6.5–12

AllADDsother

than

pioglitazone

4/201

3–4/201

9CVdeath,nonfatalM

I,ornonfatalstroke

Ertugliflozin

CVOT

Ertugliflozin

[2(5

mg,15

mg):1]

3,90

0.40

y.owithestablished

CVD

7–10

.5Naıve

orallA

DDs(specific

substudies)

11/201

3–6/20

20

CVdeath,nonfatalM

I,ornonfatalstroke

Other

drugs

DEV

OTE

Insulin

degludec

vs.

insulin

glargine

7,63

7$50

y.o+established

CVDor

renaldisease

or$60

y.o+CVD

risk

factors

$7or,7if.20

unitsinsulin/day

Oneormore

oralorinjectab

leADDs

10/201

3–12

/201

6CVdeath,nonfatalM

I,ornonfatalstroke

TOSC

A.IT

Pioglitazonevs.SU

(open

label)

3,37

150

–75

y.o,.

2years’diabetes

duration

7–9

.2months,

metform

in.2g

9/200

8–12

//20

18All-cause

mortality,nonfatalM

I,ornonfatalstroke,unplanned

coronaryrevascularization

ACEI,A

CEinhibitor;ACR,albumin-to-creatinineratio;AGI,a-glucosidaseinhibitors;ARB,angiotensinreceptorblocker;CAD,coronaryartery

disease;CANVAS,CANaglifl

ozincardioVascularAssessm

entStudy;

CANVAS-R,A

StudyoftheEffectsofC

anaglifl

ozinonRen

alEndpointsinAdultParticipan

tsWithType2Diabetes

Mellitus;CARMELINA,CArdiovascularandRen

alMicrovascularoutcomEstudywithLINAgliptinin

patientswithtype2diabetes

mellitus;CAROLINA,CARdiovascularOutcomeStudyofLINAgliptinVersusGlim

epirideinPatien

tswithType2Diabetes;C

HF,congestiveheartfailure;C

RED

ENCE,Canagliflozinand

Ren

alEven

tsin

Diabetes

withEstablished

Nep

hropathyClinicalEvaluation;CRF,chronicrenalfailure;CVD,CVdisease;DEC

LARE-TIMI58

,DapagliflozinEffect

onCardiovascuLA

REven

ts;DEV

OTE,D

egludec

CardiovascularOutcomes

Trial;eG

FR,estim

ated

glomerularfiltrationrate;ESR

D,end-stage

renaldisease;EXSCEL,EXe

natideStudyofC

ardiovascularEven

tLowering;F,female;Inh,inhibitor;LEADER

,Liraglutide

Effect

andActionin

Diabetes:EvaluationofcardiovascularoutcomeResultsd

ALongTerm

Evaluation;M,m

ale;

MI,myocardialinfarction;PA

D,p

eripheralarterialdisease;RA,receptoragonist;REW

IND,

ResearchingCardiovascularEven

tsWithaWeeklyINcretin

inDiabetes;R

F,renalfailure;TOSC

A.IT,Thiazolidined

iones

orSulphonylureas

andCardiovascularAcciden

tsInterven

tionTrial;UAP,

unstab

leangina

pectoris;y.o,years

old.

care.diabetesjournals.org Mannucci, Mosenzon, and Avogaro S199

increases overall event rates, thus reducingrequired sample size, but it limits the re-liability of results, augmenting backgroundnoise and biasing results in favor of “nodifference” between treatment arms (16).

Choosing the Trial Population and Its

Effect on Study Conduct and Results

In the FDA 2008 guidance (1), patientswith “higher CV risk” to be included inCV safety analysis were defined as “pa-tients with relatively advanced disease,elderly patients, and patients with somedegree of renal impairment.” Pharma-ceutical companies and researchers fol-lowed these instructions yet chosedifferent inclusion and exclusion criteriain different CV safety trials (Table 1).Patients in EXAMINE and ELIXA (4) hadexperienced recent ACS, whereas thosein TECOS had previous CV disease (5).Conversely, in the SAVOR-TIMI 53 trial21.7% of enrolled patients had no priorCV events but showed at least two riskfactors (17). The selection of a higher-risk population increases CV event rate,thus reducing the sample size and/orthe duration of the trial. For example,if the expected annual event rate is 3%instead of 2%, with a 5-year follow-up,the sample needed to prove CV safetyaccording to the FDA requirement is5,400 patients instead of 8,000 (16).However, the downside of usingsuch a specific high–CV risk populationis the external validity of trial results.Alogliptin was shown to be safe in pa-tients with ACS during the last 90 days(4), but the inference of those results topatients with stable CV disease, or withhigh CV risk without prior events is ofuncertain validity.

Other CV Safety Issues: hHF

Unexpectedly, in the SAVOR-TIMI 53 trial,saxagliptin was associated with an in-creased rate of hHF (3.5 vs. 2.8%; hazardratio [HR] 1.27 [95% CI 1.07–1.51]; P =0.007) (3,18). This phenomenon oc-curred during the first year, with nosignificant difference thereafter. InEXAMINE, in which hHF was a compo-nent of a prespecified exploratory ex-tended MACE end point, there was nostatistically significant increase in therisk of first event of hHF with alogliptinversus placebo (3.1 vs. 2.9%; 1.07 [0.79–1.46]) (19). However, these results arestill not sufficient to exclude a small in-crease in risk (20). In TECOS, in whichhHF was a secondary outcome, the rate

of hHF in the sitagliptin arm (228 [3.1%])was not statistically or numerically differ-ent from that in the placebo arm (229[3.1%]) (HR 1.00 [95% CI 0.83–1.2] P =0.98) (5). The results of saxagliptin onhHF in the SAVOR-TIMI 53 trial couldbe a play of chance, a specific effect ofthat drug, or an effect of the class ofdrugs on specific (and still unidentified)subpopulations of patients. In fact, theTECOS and SAVOR-TIMI 53 populations,although similar in age (66 vs. 65.1 years)and diabetes duration (9.4 vs. 10.3years), differ in other features: for exam-ple, inclusion of patients without priorCV disease (0 vs. 21.6%), HbA1c (7.2 vs.8.0%), ethnicity (whites 67.9 vs. 75.3%),renal impairment (estimated glomerularfiltration rate ,50 mL/min [9.3 vs.15.6%]), and/or use of insulin at baseline(23.2 vs. 41.4%).

None of the major CV safety trialswith DPP-4 inhibitors were specificallydesigned to assess the effect of thesedrugs on heart failure, which was a sec-ondary end point or part of a compositeend point. This is a typical case inwhich a meta-analysis can add relevantinformation (see below).

Other Predefined Safety Outcomes:Hypoglycemia, Weight Changes,Cancer, Pancreatitis, and FracturesCommon adverse events (AEs) can beeasily identified in phase II/III trials; con-versely, rare AEs can be missed by thosestudies due to insufficient sample sizeand/or duration of observation (21). Inaddition, most early trials with diabetesdrugs include only relatively healthypopulations, who may be at lower riskof AEs (22). Large-scale trials designedfor CV end points may yield importantresults for other safety issues.

Hypoglycemia

Targeting intensive glycemic controlsignificantly increases the risk of severehypoglycemia (23), which may increaseCV disease risk (24).

However, there is a consistent het-erogeneity in the incidence of severehypoglycemia among trials, which couldbe attributable to various factors suchas frailty, age, duration of disease,HbA1c targets, and therapeutic strategies.Severe hypoglycemia tends to be posi-tively associated with baseline HbA1c;the higher the starting HbA1c value ofthe patient, the higher the event rate ofsevere hypoglycemia.

Notably, incretin-based therapy shouldbe void of hypoglycemic risk owing to theability of these drugs to stimulate insulinsecretionwith a glucose-dependent effect.However, in SAVOR-TIMI 53, episodes ofsevere hypoglycemia were significantlymore frequent in the saxagliptin arm(2.1 vs. 1.7%, P = 0.047) (3), whereas anyhypoglycemic events were more frequentonly in patients with a baseline HbA1c,7%. Almost all episodes were observedin patients treated with sulfonylureas(SUs), especially in those with HbA1c,7%. In EXAMINE (4) and TECOS (5), noincrease in rates of hypoglycemia wasreported with alogliptin and sitagliptin,respectively.

Weight Changes

The observed adherence to oral hypo-glycemic drugs in patients with type 2diabetes varies from 48.5 to 72.5%(25). Drug-induced weight gain is awell-known predictor of nonadherence(25). In addition, the increase in bodyweight induced by some glucose-loweringdrugs may have an adverse effect onquality of life and on some CV risk factors(26). Weight gain induced by treatmentintensification is influenced by theamount of HbA1c reduction, backgroundtherapy, and treatment strategy. Wheninsulin is added to backgroundmetforminand SUs,weight gain is almost 3 kg; whenTZDs are added the observed weightgain is .4 kg (27). In a meta-analysis of40 randomized controlled trials, the useof secretagogues (SUs plus meglitinides)added to background metformin was as-sociatedwith aweight gain of 1.9 kg (28).

Conversely, the use of DPP-4 inhibi-tors is not associated with weight gain:in SAVOR-TIMI 53, EXAMINE, and TECOS,no significant weight changes were ob-served. The main reasons for this may berelated to lesser degrees of serious hypo-glycemic events, to a delay in gastric emp-tying, and/or toa senseof satiety secondaryto a direct hypothalamic effect (29).

Cancer and Pancreatitis

Type 2 diabetes may be associated withincreased risk of malignancies (30);in addition, it is possible that someglucose-lowering drugs increase the riskfor all cancers or specific forms of cancer.Such a concern has been raised also forGLP-1 receptor agonists and DPP-4 in-hibitors. GLP-1 receptor activation di-rectly promotes cell proliferation andenhances cell survival in several tissues

S200 CV Safety Trials With DPP-4 Inhibitors Diabetes Care Volume 39, Supplement 2, August 2016

(31). It has been observed that GLP-1–based therapies can induce histologicalchanges suggestive of pancreatic dam-age in rodents (32). An analysis of AEsreported to the FDA suggested that GLP-1receptor agonists and DPP-4 inhibitorscould be associated with pancreatitisand pancreatic cancer (33), but that re-sult could have been heavily affected byunderreporting and reporting bias. Infact, cohort studies failed to detect anysignal of risk either for pancreatitis or forpancreatic cancer (34).Reassuring findings are also available

from SAVOR-TIMI 53, EXAMINE, andTECOS, which did not report any signifi-cant increase in risk of either pancreatitisor pancreatic cancer, although a modesttrend toward an increased incidence ofpancreatitis was reported in TECOS (3–5).Finally, it has been observed in pre-

clinical studies that the incidence ofthyroid C-cell tumors is increased inrodents treated with GLP-1 analogs(35). Notably, in mouse and rat thyroidgland C cells, GLP-1 receptor densitiesare much higher than in humans andprimates (36). The available CV safetytrials confirm the safety of DPP-4 inhibitorsin this respect; in fact, no case of thyroidmedullary carcinoma was reported inSAVOR-TIMI 53, EXAMINE, or TECOS (3–5).

Pooled Analyses and Meta-analyses:Combining CV Outcome Studies andTrials With Other OutcomesMeta-analyses are typically performedwhen there is a multiplicity of availablestudies that either provide inconsistentresults or lack adequate statisticalpower because of insufficient samplesize. CV trials performed with moleculesof the same class, which can be assumedto have a similar profile of action on CVrisk, can be combined in a meta-analysis,thus increasing their power to detecteven small differences between treatmentgroups. However, it is unlikely that such ameta-analysis would provide any relevant

additional information on the effect ofDPP-4 inhibitors on major CV events, inwhich the sample size was calculated tohave an adequate power to explore theprincipal end point (i.e., MACE).

On the other hand, a meta-analysisof CV trials could be useful for the ex-ploration of secondary end points or ofindividual components of the principalend points, for which each trial is under-powered. A typical example is hHF,which was significantly increased withsaxagliptin in the SAVOR-TIMI 53 trial(3). In the CV trial with alogliptin, thedifference between treatment groupsfor hHF was not statistically significant,but this could have been the conse-quence of a smaller overall number ofevents, with a lower statistical power(19); on the other hand, no such effectwas detected with sitagliptin in TECOS(5). A meta-analysis of all available trialswith DPP-4 inhibitors (including thosewith a non-CV end point, but not TECOS,whichwas unavailable at the time) con-firmed an increased risk for heart fail-ure (37). However, when trials with anon-CV end point (i.e., phase II–IIIstudies) were considered separately,no signal of risk was detected, with nosignificant difference across molecules(37). This result is consistent with thehypothesis that the increased risk ob-served with saxagliptin in SAVOR-TIMI53 may have been related to the charac-teristics of the patients enrolled. In thecase of heart failure with DPP-4 inhibi-tors, the number of available large-scaletrials is insufficient to draw definitiveconclusions through meta-analysis. Fur-ther insight may be provided by a pooledanalysis of patient-level data from the CVsafety trials (SAVOR-TIMI 53, EXAMINE,and TECOS), exploring the effect ofDPP-4 inhibitor treatment on differentsubgroups of patients categorized fortheir baseline characteristics, with themolecule used as a categorical covariate.However, such an analysis would pose

many technical and organizational chal-lenges, considering differences in proto-cols across studies.

Information on CV events provided bymeta-analyses and pooled analyses isparticularly relevant when no CV out-come study is yet available. Under cur-rent regulations, most new drugs aremarketed on the basis of a clinical pro-gram composed of studies with meta-bolic end points, with a large-scale,noninferiority CV study performed onlyafter marketing, to confirm the safety ofthe drug. While long-term studies areongoing, combined analyses of shorter-term trials can provide some insights onpossible CV actions of new treatments.In the case of DPP-4 inhibitors, pooledanalyses of CV events in randomizedphase II–III studies have been publishedfor all molecules (38–42); all these anal-yses yielded risk estimates,1, althoughdifferences from comparator groupswere not statistically significant (Table2). Data from different analyses arenot comparable, because the definitionof events and the trial entry criteria areheterogeneous. In addition, in some ofthe analyses CV events had not beenadjudicated (38).

Pooled analyses, which combinepatient-level data, can be performed onlyby companies that hold the property ofdata collected during clinical developmentof new drugs. Independent researchers,without access to patient-level data, canonly pool trial-level results from studiesin a post hoc meta-analysis, which allowsthe combination of trials with differentdrugs from the same class, estimating anoverall class effect. Since some of the trialssubmitted to regulatory authorities remainunpublished, it is important that all dis-closed trials, including those not publishedin medical journals, are included in themeta-analysis. Notably, in such analysisonly the crude number of events is avail-able, whereas time to event remains un-known; this produces an imprecision in risk

Table 2—Pooled analyses of major CV events based on patient-level data from phase II–III trials with DPP-4 inhibitors

Reference Drug # Studies# Patients

(drug/comparator)# Events

(drug/comparator) AdjudicatedRisk estimate(HR [95% CI])

Engel et al. (38), 2013 Sitagliptin 25 7,726/6,885 40/38 No 0.83 [0.53–1.30]

Iqbal et al. (39), 2014 Saxagliptin 20 5,701/3,455 43/31 Yes 0.75 [0.46–1.21]

Rosenstock et al. (40), 2015 Linagliptin 19 5,847/3,612 60/62 Yes 0.78 [0.55–1.12]

Schweizer et al. (41), 2010 Vildagliptin 25 7,509/6,061 81/91 Yes 0.84 [0.62–1.14]

White et al. (4), 2013 Alogliptin 11 4,162/1,855 13/10 Yes 0.63 [0.00–1.41]

care.diabetesjournals.org Mannucci, Mosenzon, and Avogaro S201

estimates,which is negligible in short-termstudies but which can be relevant if theduration of treatment is very long. Thelack of adjudication of events and the shortduration of treatment are also major limi-tations for post hoc meta-analyses. Fur-thermore, it is possible that some trialswith unfavorable results for experimentaldrugs remain undisclosed or that CV out-comes are only partially reported, thusproducing a systematic bias. All thesemethodological issues suggest cautionin the interpretation of results frompost hoc meta-analyses. However, it isnotable that all post hoc analyses (8–10,43) reported a significant reductionin the incidence of major CV events,with risk estimates ranging from 0.36to 0.71 (Table 3); the only exceptionis a meta-analysis that included twoCV studies (SAVOR-TIMI 53 and EXAMINE)along with earlier trials, with theformer largely driving the overall result(44). One of the largest of those meta-analyses (8) also reported a significantreduction of all-cause mortalitydtheonly end point unaffected by eventadjudication.

Why Are the Results of PooledAnalyses and Meta-analyses of EarlyTrials Different From Those of StudiesWith a CV End Point?The difference in results betweenpooled analyses and meta-analyses ofearly studies, on one side, and CV trials,on the other, is quite substantial. Themethodological limitation of pooledanalyses and meta-analyses summa-rized above, although relevant, doesnot seem sufficient for inducing such alarge distortion. It has been suggestedthat differences in results could be de-termined by the duration of trials, withDPP-4 inhibitors producing a benefitin the short-term, which is lost withlonger-term treatment (45). This viewis not confirmed by CV trials, in whichno evidence of a temporary benefit ispresent in the first few months afterrandomization; in addition, a signifi-cant reduction of major CV events isobserved in non–CV outcome trialswith DPP-4 inhibitors with duration oftreatment .52 weeks (46). Anotherpossible explanation for the observeddifference is the diversity in the charac-teristics of enrolled patients. Subjectsrecruited in CV studies, who areselected for high CV risk, are typically

older with longer duration of diabetes,greater impairment of renal function,higher comorbidity, and higher numberof concomitant treatments than thoseparticipating in earlier trials with meta-bolic end points. It is possible that DPP-4inhibitors produce a CV benefit in rela-tively low-risk patients, whereas theyare neutral in subjects at very high riskor with prior CV events.

The enrollment of subjects at high CVrisk is a practical strategy for yielding ahigh number of events with a limitedsample size and duration of observation.It is also reasonable to verify safety inthose patients who are at greatest riskof adverse outcomes. On the otherhand, patients enrolled in CV trialswith DPP-4 inhibitors are scarcely/hardly representative of those actuallyreceiving a prescription of those drugsin clinical practice.

ConclusionsThe results of the first CV trials withDPP-4 inhibitors (3–5) did not meet theexpectations of many clinicians, whohoped for a demonstration of a CV bene-fit, as suggested by meta-analyses of pre-vious, shorter-term studies (8–10,43).However, CV trials provided exactly theresult that they had been designed for:noninferiority versus placebo. CV trialsin diabetes simply demonstrate thatdrugs do not catastrophically increaseCV risk. The unexpected finding of a sig-nificantly beneficial effect of the activedrug, as with empagliflozin (6), giventhe design of those studies, is almost acase of serendipity.

The explanation of differences in CVoutcomes between earlier studies andCV safety trials is complex. Our opinionis that methodological issues are not

sufficient to account for such differ-ence and that themore favorable resultof an earlier trial is the effect of theenrollment of younger, healthier pa-tients with fewer complications, with alower proportion of subjects receivingother medications for the reduction ofCV risk. However, available data arenot sufficient to confirm such a hypoth-esis, which could be proven onlythrough a specifically designed CV out-come trial in patients with diabetes atlow riskdwhich is very unlikely to beever performed.

CV trials provide a huge amount ofsafety data. Large-scale trials of appropri-ate duration, although designed formajorCV events, yield relevant information onother safety end points (for example,withDPP-4 inhibitors, pancreatitis). Exposureof several thousand patient-years to anexperimental drug also warrants the de-tection of relatively infrequent AEs. How-ever, in this latter case, unless results areconsistent across trials with molecules ofthe same class, the interpretation of re-sults can be overwhelmingly difficult. Thisis what happened for the unexpected(and problematic) finding of increasedhHF with saxagliptin in SAVOR-TIMI 53.

Although the results of CV safetystudies could underestimate the actualbenefit of DPP-4 inhibitors on CVoutcomes, a recent study with a similardesign did show a significant reductionof CV morbidity and mortality with theSGLT2 inhibitor empagliflozin (6). Al-though the characteristics of patientsenrolled in this latter trial (5) are notexactly the same as in trials with DPP-4inhibitors (3–5), the fact that results withempagliflozin in patients with prior CVevents were much improved comparedwith than those observedwith saxagliptin,

Table 3—Post hoc meta-analyses of major CV events based on trial-level data withDPP-4 inhibitors

Reference # Studies# Patients

(drug/comp)# Events

(drug/comp)Risk estimate(HR [95% CI])

Not including CV safety trialsMonami et al. (37), 2014 63 23,562/16,509 263/232 0.71 (0.59–0.86)Patil et al. (43), 2012 18 4,998/3,546 45/56 0.48 (0.61–0.75)Wu et al. (9), 2014 8a 7,778 6/18a 0.36 (0.15–0.85)a

50b 10/12b 0.54 (0.25–1.19)b

Zhang et al. (10), 2014 12 5,505/5,477 25/43c 0.53 (0.32–0.87)c

Including CV safety trialsAgarwal et al. (44), 2014 80 40,749/32,592 NR 0.95 (0.86–1.04)

NR, not reported. aVersus metformin, monotherapy; bversus placebo, add-on to metformin;cversus SUs.

S202 CV Safety Trials With DPP-4 Inhibitors Diabetes Care Volume 39, Supplement 2, August 2016

alogliptin, or sitagliptin should be clearlyrecognized.Inevitably, information on safety is

excellent for newer agents but largelymissing for older drugs. This creates aninformation disequilibrium, for whichsafety issues have a greater probabilityof being raised for newer drugs, whilethey may remain unnoticed, or under-estimated, for older agents. Anotherproblem with the current approach isthat populations enrolled in CV trialscan be quite different from those that,on average, receive a prescription for adrug in the real-world. This can lead toan overestimation of safety issues withnewer agents.Finally, the most important drawback

of CV safety trials is their excruciatinglyhigh cost, which has an impact on theprice, and therefore the widespreadavailability, of newer agents. Further-more, pharmaceutical companies, beingforced to invest hugebudgets onplacebo-controlled CV trials, tend to shrink theirdevelopment programs on the side ofphase III active-comparator controlled tri-als, which are probably more useful forclinical decision making and which couldbe more cost-efficient.The decision of the FDA to require

safety data from large-scale, specificallydesigned trials had the merit of focusingthe attention of the scientific com-munity on the relevance of CV safetyof glucose-lowering drugs. At the sametime, the unintended negative conse-quences of this decision may outweighthe benefits. As an alternative, CV safetycouldbeassessed throughpooledanalysesof patient-level data from wider programsof active comparator-controlled phase IIItrials of appropriate duration, providedthat a relevant fraction of patients athigh CV risk is included and that MACEsare appropriately adjudicated. This wouldprobably require a safety criterion lessstringent than the present 1.3 for upperconfidence limit (1); however, additionalreal-world data could be obtained throughspecific programs for postmarketing sur-veillance, thus avoiding specific CV safetytrials.

Funding and Duality of Interest. E.M. hasreceived speaking and/or consulting honorariafrom AstraZeneca, Boehringer Ingelheim, EliLilly, Jannsen, Merck, Novartis, Novo Nordisk,Sanofi, and Takeda and research grants fromAstraZeneca, Boehringer Ingelheim, Eli Lilly,

Novartis, Novo Nordisk, Takeda, and Sanofi. O.M.has received speaking fees fromAstraZeneca andBristol-Myers Squibb, Novo Nordisk, Eli Lilly,Sanofi, Novartis, Merck Sharp & Dohme, andKyowa Hakko Kirin; consulting fees from NovoNordisk, Eli Lilly, Sanofi, Novartis, AstraZeneca,Boehringer Ingelheim, and Janssen; and researchgrants from AstraZeneca, Bristol-Myers Squibb,and Novo Nordisk. A.A. has received researchgrants from the Strategic Project DYCENDIfrom the University of Padova STPD11ALFEand has also received speaking honoraria fromNovartis, Eli Lilly, Novo Nordisk, Sanofi, BoehringerIngelheim, AstraZeneca, Mediolanum, Servier,Janssen, and Merck Sharp & Dohme; financialsupport for attending symposia from Novartis,Eli Lilly, NovoNordisk, Sanofi, Boehringer Ingelheim,and AstraZeneca; financial support for educa-tional programs from Novartis; consulting feesfrom Novartis, Eli Lilly, Novo Nordisk, Sanofi,Boehringer Ingelheim, AstraZeneca,Mediolanum,Servier, Janssen, andMerck Sharp&Dohme; andresearch grants from Boehringer Ingelheim andAstraZeneca. No other potential conflicts of interestrelevant to this article were reported.Author Contributions. E.M., O.M., and A.A.contributed to collection of relevant references,discussionof their clinical implications,writingofthe manuscript, and revision of the text.

References1. U.S. Food and Drug Administration. Guid-ance for industry: evaluating cardiovascularrisk in new antidiabetic therapies. Availablefrom http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm071627.pdf. Accessed 3 August20152. Hirshberg B, Raz I. Impact of the U.S. Foodand Drug Administration cardiovascular assess-ment requirements on the development ofnovel antidiabetes drugs. Diabetes Care 2011;34(Suppl. 2):S101–S1063. Scirica BM, Bhatt DL, Braunwald E, et al.;SAVOR-TIMI 53 Steering Committee and Investi-gators. Saxagliptin and cardiovascular outcomesin patients with type 2 diabetes mellitus. N Engl JMed 2013;369:1317–13264. White WB, Cannon CP, Heller SR, et al.;EXAMINE Investigators. Alogliptin after acutecoronary syndrome in patients with type 2 di-abetes. N Engl J Med 2013;369:1327–13355. Green JB, Bethel MA, Armstrong PW, et al.;TECOS Study Group. Effect of sitagliptin on car-diovascular outcomes in type 2 diabetes. N EnglJ Med 2015;373:232–2426. Zinman B, Wanner C, Lachin JM, et al.;EMPA-REG OUTCOME Investigators. Empagliflozin,cardiovascular outcomes, and mortality intype 2 diabetes. N Engl J Med 2015;373:2117–21287. Hirshberg B, Katz A. Cardiovascular outcomestudies with novel antidiabetes agents: scien-tific and operational considerations. DiabetesCare 2013;36(Suppl. 2):S253–S2588. Monami M, Ahren B, Dicembrini I, MannucciE. Dipeptidyl peptidase-4 inhibitors and cardio-vascular risk: a meta-analysis of randomizedclinical trials. Diabetes Obes Metab 2013;15:112–1209. Wu D, Li L, Liu C. Efficacy and safety of dipep-tidyl peptidase-4 inhibitors and metformin as

initial combination therapy and as monother-apy in patients with type 2 diabetes mellitus: ameta-analysis. Diabetes Obes Metab 2014;16:30–3710. Zhang Y, Hong J, Chi J, Gu W, Ning G, WangW. Head-to-head comparison of dipeptidylpeptidase-IV inhibitors and sulfonylureas - ameta-analysis from randomized clinical trials.Diabetes Metab Res Rev 2014;30:241–25611. Skyler JS, Bergenstal R, Bonow RO, et al.;American Diabetes Association; American Col-lege of Cardiology Foundation; AmericanHeart Association. Intensive glycemic controland the prevention of cardiovascular events:implications of the ACCORD, ADVANCE, andVA diabetes trials: a position statement ofthe American Diabetes Association and a sci-entific statement of the American College ofCardiology Foundation and the AmericanHeart Association. Diabetes Care 2009;32:187–19212. Bergenstal RM, Bailey CJ, Kendall DM. Type 2diabetes: assessing the relative risks and benefitsof glucose-lowering medications. Am JMed 2010;123:9–1813. Gaede P, Lund-Andersen H, Parving HH,Pedersen O. Effect of a multifactorial interven-tion on mortality in type 2 diabetes. N Engl JMed 2008;358:580–59114. Ibrahim JG, ChenMH, Xia HA, Liu T. Bayesianmeta-experimental design: evaluating car-diovascular risk in new antidiabetic therapies totreat type 2 diabetes. Biometrics 2012;68:578–58615. Mega JL, Braunwald E, Wiviott SD, et al.;ATLAS ACS 2–TIMI 51 Investigators. Rivaroxabanin patients with a recent acute coronary syndrome.N Engl J Med 2012;366:9–1916. Joffe HV, Parks MH, Temple R. Impact ofcardiovascular outcomes on the developmentand approval of medications for the treatmentof diabetes mellitus. Rev Endocr Metab Disord2010;11:21–3017. Mosenzon O, Raz I, Scirica BM, et al. Base-line characteristics of the patient population inthe Saxagliptin Assessment of Vascular Out-comes Recorded in Patients with Diabetes Mel-litus (SAVOR)-TIMI 53 trial. Diabetes Metab ResRev 2013;29:417–42618. Scirica BM, Braunwald E, Raz I, et al.;SAVOR-TIMI 53 Steering Committee and In-vestigators. Heart failure, saxagliptin, anddiabetes mellitus: observations from theSAVOR-TIMI 53 randomized trial. Circulation2014;130:1579–158819. Zannad F, Cannon CP, Cushman WC, et al.;EXAMINE Investigators. Heart failure and mor-tality outcomes in patients with type 2 diabetestaking alogliptin versus placebo in EXAMINE: amulticentre, randomised, double-blind trial.Lancet 2015;385:2067–207620. Standl E, Schnell O. DPP-4 inhibitors andrisk of heart failure EXAMINEd. Lancet 2015;385:2022–202421. FrenchDD,MargoCE, Campbell RR. Enhancingpostmarketing surveillance: continuing challenges.Br J Clin Pharmacol 2015;80:615–61722. Lakey WC, Barnard K, Batch BC, Chiswell K,Tasneem A, Green JB. Are current clinical trialsin diabetes addressing important issues in dia-betes care? Diabetologia 2013;56:1226–1235

care.diabetesjournals.org Mannucci, Mosenzon, and Avogaro S203

23. Hemmingsen B, Lund SS, Gluud C, et al. In-tensive glycaemic control for patients withtype 2 diabetes: systematic review with meta-analysis and trial sequential analysis of randomisedclinical trials. BMJ 2011;343:d689824. Goto A, Arah OA, Goto M, Terauchi Y, NodaM. Severe hypoglycaemia and cardiovasculardisease: systematic review and meta-analysiswith bias analysis. BMJ 2013;347:f453325. Mateo JF, Gil-Guillen VF, Mateo E, OrozcoD, Carbayo JA,Merino J.Multifactorial approachand adherence to prescribed oral medicationsin patients with type 2 diabetes. Int J Clin Pract2006;60:422–42826. Strandberg TE, Strandberg A, Salomaa VV,Pitkala K, Tilvis RS, Miettinen TA. The associa-tion between weight gain up to midlife, 30-yearmortality, and quality of life in older men. ArchIntern Med 2007;167:2260–226127. Gross JL, Kramer CK, Leitão CB, et al.; Di-abetes and Endocrinology Meta-analysis Group(DEMA). Effect of antihyperglycemic agentsadded to metformin and a sulfonylurea on gly-cemic control and weight gain in type 2 diabe-tes: a network meta-analysis. Ann Intern Med2011;154:672–67928. McIntosh B, Cameron C, Singh SR, et al.Second-line therapy in patients with type 2diabetes inadequately controlled with metforminmonotherapy: a systematic review and mixed-treatment comparison meta-analysis. Open Med2011;5:e35–e4829. Toft-Nielsen MB, Madsbad S, Holst JJ. Con-tinuous subcutaneous infusion of glucagon-likepeptide 1 lowers plasma glucose and reducesappetite in type 2 diabetic patients. DiabetesCare 1999;22:1137–114330. Tsilidis KK, Kasimis JC, Lopez DS, Ntzani EE,Ioannidis JP. Type 2 diabetes and cancer: umbrella

review of meta-analyses of observational studies.BMJ 2015;350:g760731. Brubaker PL, Drucker DJ. Minireview:glucagon-like peptides regulate cell prolifera-tion and apoptosis in the pancreas, gut, andcentral nervous system. Endocrinology 2004;145:2653–265932. Butler PC, Dry S, Elashoff R. GLP-1-basedtherapy for diabetes: what you do not knowcan hurt you. Diabetes Care 2010;33:453–45533. Elashoff M, Matveyenko AV, Gier B,Elashoff R, Butler PC. Pancreatitis, pancreatic,and thyroid cancer with glucagon-like peptide-1-based therapies. Gastroenterology 2011;141:150–15634. Montilla S, Marchesini G, Sammarco A,et al.; AIFA Anti-diabetics Monitoring Group.Drug utilization, safety, and effectiveness of ex-enatide, sitagliptin, and vildagliptin for type 2diabetes in the real world: data from the ItalianAIFA Anti-diabetics Monitoring Registry. NutrMetab Cardiovasc Dis 2014;24:1346–135335. ChiuWY, Shih SR, Tseng CH. A review on theassociation between glucagon-like peptide-1 re-ceptor agonists and thyroid cancer. Exp Diabe-tes Res 2012;2012:92416836. Waser B, Beetschen K, Pellegata NS, ReubiJC. Incretin receptors in non-neoplastic andneoplastic thyroid C cells in rodents and hu-mans: relevance for incretin-based diabetestherapy. Neuroendocrinology 2011;94:291–30137. Monami M, Dicembrini I, Mannucci E. Dipep-tidyl peptidase-4 inhibitors and heart failure: ameta-analysis of randomized clinical trials. NutrMetab Cardiovasc Dis 2014;24:689–69738. Engel SS, Golm GT, Shapiro D, Davies MJ,Kaufman KD, Goldstein BJ. Cardiovascular safetyof sitagliptin in patients with type 2 diabetes

mellitus: a pooled analysis. Cardiovasc Diabetol2013;12:339. Iqbal N, Parker A, Frederich R, Donovan M,Hirshberg B. Assessment of the cardiovascularsafety of saxagliptin in patients with type 2 di-abetes mellitus: pooled analysis of 20 clinicaltrials. Cardiovasc Diabetol 2014;13:3340. Rosenstock J, Marx N, Neubacher D, et al.Cardiovascular safety of linagliptin in type 2 di-abetes: a comprehensive patient-level pooledanalysis of prospectively adjudicated cardiovas-cular events. Cardiovasc Diabetol 2015;14:5741. Schweizer A, Dejager S, Foley JE, CouturierA, Ligueros-Saylan M, Kothny W. Assessing thecardio-cerebrovascular safety of vildagliptin:meta-analysis of adjudicated events from alarge Phase III type 2 diabetes population. Di-abetes Obes Metab 2010;12:485–49442. White WB, Pratley R, Fleck P, et al. Cardio-vascular safety of the dipetidyl peptidase-4 in-hibitor alogliptin in type 2 diabetes mellitus.Diabetes Obes Metab 2013;15:668–67343. Patil HR, Al Badarin FJ, Al Shami HA, et al.Meta-analysis of effect of dipeptidyl peptidase-4inhibitors on cardiovascular risk in type 2 diabetesmellitus. Am J Cardiol 2012;110:826–83344. Agarwal S, Parashar A, Menon V. Meta-analysis of the cardiovascular outcomes withdipeptidyl peptidase 4 inhibitors: validation ofthe current FDA mandate. Am J CardiovascDrugs 2014;14:191–20745. Savarese G, Perrone-Filardi P, D’Amore C,et al. Cardiovascular effects of dipeptidylpeptidase-4 inhibitors in diabetic patients: ameta-analysis. Int J Cardiol 2015;181:239–24446. Dicembrini I, Mannucci E. Risk of myocardi-al infarction in trials with dipeptidyl peptidase-4inhibitors: is duration of study a real issue? Int JCardiol 2015;184:543–544

S204 CV Safety Trials With DPP-4 Inhibitors Diabetes Care Volume 39, Supplement 2, August 2016