The dawn phenomenon in type 2 diabetes: How to assess it in clinical practice?

HAL Id: hal-02164831

https://hal.archives-ouvertes.fr/hal-02164831

Submitted on 25 Jun 2019

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

The dawn phenomenon in type 2 diabetes: How to

assess it in clinical practice?

L. Monnier, C. Colette, S. Dejager, D. Owens

To cite this version:

Availableonlineat

ScienceDirect

www.sciencedirect.com

Diabetes&Metabolism41(2015)132–137

Original

article

The

dawn

phenomenon

in

type

2

diabetes:

How

to

assess

it

in

clinical

practice?

L.

Monnier

,

C.

Colette

,

S.

Dejager

,

D.

Owens

a_{InstituteofClinicalResearch,UniversityMontpellier1,France} b_{DepartmentofEndocrinology,HospitalPitié-Salpétrière,Paris,France}

c_{DiabetesResearchGroup,SwanseaUniversity,UnitedKingdom}

Received20August2014;receivedinrevisedform2October2014;accepted2October2014 Availableonline6November2014

Abstract

Aim.–Thestudywasaimedatdeterminingwhetherthedawnphenomenonintype2diabetes(T2D)canbepredictedandquantifiedusing simpleandeasilyaccessibleglucosedeterminations.

Methods.–Atotalof210non-insulin-treatedpersonswithT2Dunderwentcontinuousglucosemonitoring(CGM).Thedawnphenomenonwas quantifiedastheabsoluteincrementfromthenocturnalglucosenadirtothepre-breakfastvalue(dawn,mg/dL).Pre-lunch(preL)andpre-dinner (preD)glucose, andtheiraveragedvalues(preLD),werecomparedwiththenocturnalnadir.Thesepre-mealvaluesweresubtractedfromthe pre-breakfastvalues.Thedifferencesobtained(pre-mealL,pre-mealDandpre-mealLD)werecorrelatedwithdawnvalues.Thereceiver operatingcharacteristic(ROC)curvewasusedtoselecttheoptimalpre-mealvaluethatbestpredictedadawnphenomenon,setatathresholdof 20mg/dL.

Results.–Allpre-mealglucoselevelsanddifferencesfrompre-breakfastvalues(pre-meal)significantlycorrelated(P<0.0001)withthe nocturnalnadiranddawnvalues,respectively.ThestrongestcorrelationswereobservedfortheparametersaveragedatpreLandpreDtime points:r=0.83forpreLDandr=0.58forpre-mealLD.ROCcurveanalysisindicatedthatthedawnphenomenonatathresholdof20mg/dL canbesignificantlypredictedbyapre-mealLDcutoffvalueof10mg/dL.Therelationshipbetweendawn(Y,mg/dL)andpre-mealLD(X, mg/dL)wasY=0.49X+15.

Conclusion.–Theself-monitoringofpreprandialglucosevaluesatthethreemainmealtimescanpredictthepresence/absenceofthedawn phenomenon,andpermitsreliableassessmentofitsmagnitudewithoutrequiringcontinuousovernightglucosemonitoring.

©2014ElsevierMassonSAS.Allrightsreserved.

Keywords: Dawnphenomenon;Glucosemonitoring;Type2diabetes

1. Introduction

The dawn phenomenon has been extensively investigated for more than 30 years [1,2]. By using continuous glucose monitoring(CGM),ithasrecentlybeendemonstratedthat,in non-insulin-treated persons with type 2 diabetes (T2D), the dawnphenomenon–definedasanexcessiveincrementfromthe nocturnal glucose nadirtothe pre-breakfast glucosevalue–is

∗_{Correspondingauthor.InstituteofClinicalResearch,641,avenuedu}

Doyen-Giraud,34093Montpelliercedex5,France.Tel.:+33411759891; fax:+33411759898.

E-mailaddress:louis.monnier@inserm.fr(L.Monnier).

aglycaemicdisorderwithamagnitudeandfrequencythatare relatively stable across all groups of subjects, irrespective of treatment, HbA1c or age[3].Its frequencycanbe as highas 40% [3],andits impacton HbA1c levels approximates0.4% (4mmol/mol)[4]andthereforecannotbeignoredinthe man-agementofpersonswithT2D.Currently,healthcareproviders failtotakethedawnphenomenonintoconsiderationasoneof thetargetsinthemanagementofthedisorder.Thisisverylikely linkedtothefactthatitsquantificationrequiresCGMto accu-ratelydetectthenocturnalglucosenadir[4–7],andpermitthe calculation of the absolutedifferences between the nocturnal glucosenadirandpre-breakfastglucosevalues[3,4].Accessto CGMremainscostlyandisthereforeonlyavailabletophysicians workingatclinicslocatedinspecializeddiabetesunits[8]. http://dx.doi.org/10.1016/j.diabet.2014.10.002

L.Monnieretal./Diabetes&Metabolism41(2015)132–137 133

Thus,anyattempttoincrease awarenessof thedawn phe-nomenonand its management requires asimpler method for itsquantification.Suchamethodwouldideallybebasedonan appropriatecost-effectivefrequencyofself-monitoringofblood glucose(SMBG)inpeoplewithT2D[9].However,structured SMBG is inadequate for detecting glucosenadirs, especially thoseoccurring duringnocturnal periods[3,4].Consequently, the questionis whetherSMBG atspecifictime pointsof the diurnalperiodcanprovideanalternativeinclinicalpracticefor detecting the presence of the dawn phenomenon and further quantifying its magnitude. Using CGM through an observa-tionalstudy,thepresentstudysetouttoascertain,first,whether oneorseveralpreprandialorinterprandialglucosevaluescould approximatethenocturnalglucosenadirand,second,whetherit ispossibletopredictthepresenceorabsenceofthedawn phe-nomenonbycalculatingthe decrementor incrementbetween pre-breakfastglucoselevels and thoseobservedat other pre-mealtimepoints.

2. Methods

Atotalof210personswithT2D(meanage=60.1years,mean bodymassindex[BMI]=30.4kg/m2andmeanHbA1c=7.5% [58mmol/mol])wereselectedafterscreeningforeligibilityfrom atotalpopulationof242non-insulin-treatedpatientswithT2D whounderwent 3-dayambulatoryCGM. Thekey criteria for exclusionfromtheinitialscreenedlistofpotentialparticipants includedhavingabnormaleatinghabitsoreatingpatterns char-acterizedbyunexpectedfoodintakesduring bothdiurnal and nocturnal periods. All participants who reportedat least one clinical hypoglycaemic event over the test period were also excludedtoavoidanymisinterpretation duetoeither glucose risesinthe earlymorningor excessive glucosereboundafter correctionof thehypoglycaemic episode. Toavoidany inter-ferenceofcarbohydrateintakeonpre-mealglucosevalues,all individualsforwhomthetimeintervalsbetweentwoconsecutive mealswere<4hwerealsoexcluded.Accordingly,allsubjects whoreportedhavingamid-afternoonsnackwereexcludedfrom thefinalanalysis.Additionalcriteriaforexclusionwerearecent illnessortreatmentwithsteroidsduringthepreceding3months, andanydisruptioninglucosemonitoringoraninsufficient num-berofbloodglucosetestsforcalibrationduringCGM(fourtests adayarerequiredforthispurpose).Unacceptablecalibration meantanaccuracycriterionwithacoefficientcorrelation<0.79. All study participants were investigated from 2003 to 2011 at the outpatients facilities of the University Hospi-tal in Montpellier, France, and placed on a stable treatment regimen with either dietary measures alone or the addition of oral hypoglycaemic agents (OHAs) for at least 3 months prior to CGM. Modalities of treatment were classified into threecategories:(i)dietarymeasuresalone(n=6);(ii)insulin sensitizersalone (metforminand/or pioglitazone,n=82); and (iii)insulinsecretagogues(sulphonylureas,glinides)orincretin enhancers(dipeptidylpeptidase[DPP]-4inhibitors)takenalone or in combination with insulin sensitizers (n=122). DPP-4 inhibitorswerecategorizedalongwithsulphonylureasand glin-idesbecausethesedrugsareallinsulinotropicagents,although

Table1

Characteristicsofstudypatients.

Patientstested(n) 210

Age(years) 60.1(0.7)

Bodymassindex(kg/m2) 30.4(0.5) Genderratio(males/females) 134/76 Meandailyenergyintake(kcal) 2193(23) Meandailycarbohydrateintake(g) 274(3)

HbA1c(mmol/mol) 58(0.9)

HbA1c(%) 7.5(0.1)

Average24hmeanglucose concentrations(mg/dL)

144.8(2.4) Meanglucosevalue(mg/dL)atthe

followingtimepoints

Nocturnalnadir 113.0(2.3)

Pre-breakfast 131.4(2.5)

Pre-lunch 126.5(2.7)

Pre-dinner 121.1(2.9)

MeanpreLDa_{(mg/dL) 123.8(2.4)}

Diabetestreatment(patients,n)

Dietalone 6

Insulinsensitizersalone(metformin and/orglitazones)

82 Insulinsecretagogues/insulin

enhancersaloneorincombination withinsulinsensitizers

122

Dawnb_{(mg/dL) 18.4(1.3)}

Timeofnocturnalglucosenadirc_{(h) 0341[0157]}

Dataareexpressedasmeans(SEM)unlessstatedotherwise.

a_{Averagedmeanglucosevaluesatpre-lunchandpre-dinnertimepoints.} b _{Differencebetweenpre-breakfastandnocturnalnadirglucosevalues.} c_{Calculatedonlyforpatients(n=80)exhibitinganovertdawnphenomenon,}

expressedasmean[SD].

themechanismofactionofDPP-4inhibitorsismuchbroader thanitseffectsoninsulinsecretionalone[10].This categoriza-tionhaspreviouslybeenused[4]andjustifiedinaletter[11]in responsetoaquestionraisedbyCarretal.[12].Inaddition,no differenceswereobservedbetweenglycaemicprofilesinthetwo maingroupsselectedbycategoriesoftreatment(insulin sensi-tizersalone,andeitherinsulinsecretagogues/insulinenhancers aloneor incombinationwithinsulinsensitizers).Thislackof difference(datanotshown)hasreinforcedourstrategyto ana-lyzeourstudypopulationasawhole,andnotasseparategroups categorizedbytypeofantidiabetictreatment.

routinelow-levelphysicalactivitywhileavoidinganyvigorous exercise.

Allinvestigationswereroutinelyperformedatdiabetes out-patients clinics and were in compliance with the Helsinki Declaration[14].Thestudywas conductedafter eachsubject hadgivenoralinformedconsentinaccordancewithEuropean directivesthatrequirenoapprovalfromanethicscommitteefor thenon-interventionalprotocoldescribedbelow[15].

2.1. Clinicalinvestigationsandlaboratorydeterminations

AllparticipantsunderwentambulatoryCGMfor3 consecu-tivedays,usingthesametechnology,between2003and2011 (asecond-generationMiniMedsystem, Medtronic,Inc., Min-neapolis,MN,USA).Theglucosesensorwasinsertedonday0 andremovedmid-morningonday3.SensorinsertionforCGM wasmadeonMonday/TuesdayandremovalonThursday/Friday toavoidany overlap intothe weekend.All calculations were derivedfromdataobtainedovera48hperiodduringdays1and 2 toavoid any bias duetoeither insertion or removal of the sensor. Status of chronic hyperglycaemia (ambient hypergly-caemia)wasassessedonstudyday0byHbA1clevelsthatwere determined using a high-performance liquid chromatography (HPLC)assay[16–18].Theintra-andinterassaycoefficientsof variation(CVs)were<2%,withanon-diabeticreferencerange of16.7–32.1mmol/mol(3.7–5.1%)forHbA1c.

2.2. AnalysisofdataobtainedbyCGM

CGMwasusedtodetectthenocturnalglucosenadirandto quantifyglucoselevelsatpre-mealtimepoints.Thelatterwere determinedinaccordancewiththedatarecordedbyparticipants intheirownSMBGlogbooks.AllCGMreadingswere concomi-tantlymadebytwoinvestigatorsusingthesamemethodology for data analysis and interpretationof the glycaemic profiles obtained from 2003 to2011. All measurements provided by CGMwerepresentedafteraveragingthedatarecordedonstudy days1and2.Thisstrategywasusedtoattenuatetheday-to-day fluctuationsseeninglucoseprofiles,includingthoseofthedawn phenomenon.

Thedawnphenomenon–definedasthespontaneousglucose rise during the early morning hours–was quantified by sub-tractingtheglucosenadirfromtheglucosevalueobservedjust before breakfast. Although measurements of glucose in the interstitial fluid from CGM readings are known tobe lower thanvenousplasma glucoseconcentrations [19],itshouldbe noted that, in the present study, the dawn phenomenon was assessedonlyfromdifferencesbetweeninterstitialvalues.The upwardglycaemicvariationthatoccursatnightwasonly con-sidered a dawn phenomenon if its magnitude exceeded the thresholdof20mg/dL,chosenonthebasisoftwomain obser-vations/principlesthatreflectwhatweknowofthespontaneous variabilityandfluctuationsofglucoselevels[4].First,avalue of 20mg/dL approximates the 95% confidence interval (CI) of total intra- andinterassay variability seen in relationto a truefastingglucoseconcentrationof126mg/dL[20].Second, thisvalueof 20mg/dLisabovethespontaneousbetween-day

Fig.1.Graphicrepresentationofcalculationsofthedifferencesbetween glu-cosevaluesatpre-mealtimepointsandthenocturnalnadir.Glucosevalues (GVs)atthenocturnalnadirandpre-breakfast(BF),pre-lunch(L)and pre-dinner(D)timepointsareindicatedbynumbers1,2,3and4,respectively.GVs fordawn,pre-mealLandpre-mealDareindicatedbya,bandc, respec-tively.preLD:averageofpre-lunch(preL)andpre-dinner(preD)GVs;dawn: differencebetweenpre-breakfastandnocturnalnadirGVs;pre-mealL: dif-ferencebetweenpre-breakfastandpreLGVs;pre-mealD:differencebetween pre-breakfastandpreDGVs;pre-mealLD:differencebetweenpre-breakfast andpreLD(averageofpreLandpreD)GVs.

fluctuationinmeandifferencefromnocturnalglucosenadirto pre-breakfastglucosevalueswhich,atanindividuallevel,was 15mg/dL(interquartile range [IQR]:0–21.0mg/dL)[4]. Pre-lunch(preL),pre-dinner(preD)andtheaveragedglucosevalues atpreLandpreDtimepoints(preLD)weretestedassurrogates of nocturnal glucosenadirswhen testedagainstpre-breakfast glucosevaluestoquantifythemagnitudeofdawnglucoserises duringtheearlymorninghours.

2.3. Calculationsandstatisticalanalyses

Meanglucosevalueswerecalculatedatthenocturnalnadir andatpre-mealtimepoints,andcomparedbyanalysisof vari-ance (ANOVA) testing. The relationships between pre-meal glucosevaluesandnocturnalnadirswerecalculatedusing Pear-son’stest. Differences(pre-meal)betweenthepre-breakfast andotherpre-mealvalues,suchaspreL,preDandtheaveraged preLD,werecalculated(mg/dL)andpresentedaspre-mealL,

pre-meal Dandpre-meal LD,respectively.Details of the calculationareillustratedinFig.1.Allpre-mealvalueswere furtheranalyzedforstatisticalcomparisonandcorrelationwith themagnitudeofthedawnglucoserise(dawn,mg/dL), deter-minedaccordingtothemethodmentionedabove(bycalculating theglucoseincrementfromnocturnalnadirtopre-breakfasttime point).Itshouldbenotedthat,inallcases,thedawnglucoserise (dawn) wasquantifiedas≥0while thedifferencesbetween pre-mealvalues(pre-meal)couldbeaffectedbyeithera pos-itiveornegativesign,dependingonwhethertheglucosevalues measuredatpreLandpreDtimepointsandcalculatedfromthe preLD were belowor above the pre-breakfastglucoselevels. Relationshipsbetweenthedawnandpre-mealvalueswere furthertestedusingPearson’scorrelationcoefficient.

L.Monnieretal./Diabetes&Metabolism41(2015)132–137 135

Fig.2.a:Relationshipbetweennocturnalnadir(Y,mg/dL)andtheaverageofpreLandpreDglucosevalues(X,preLD,mg/dL):r=0.83,P<0.0001.Therelationship isY=0.82X+11;b:Relationshipbetweenthedawn(Y,mg/dL)andthepre-mealLD(X,mg/dL):r=0.58,P<0.0001.TherelationshipisY=0.49X+15. presenceorabsenceofadawnphenomenonwassetat20mg/dL,

avalueselectedonapreviouslydescribedbasis[4].Sensitivity andspecificityforpredictingthepresenceorabsenceofadawn phenomenonwerecalculatedatdifferentlevels ofpre-meal (differencesbetweenthe pre-breakfastandothermeasured or calculatedpre-meal glucose values,using step-by-step incre-mentsfromanegative[–30mg/dL]toapositive[+30mg/dL] value).Sensitivity(thetrue-positivefraction,TPF)wasdefined as the proportion of subjects to have a dawn phenomenon (>20mg/dL) inrelation tothosewho indeed hadone, while specificitywasdefinedastheproportionofsubjectspredictedto nothaveadawnphenomenon(≤20mg/dL)inrelationtothose whoactuallydidnot.Theoptimalcutoffdifferencesbetweenthe pre-breakfastandothermeasuredorcalculatedpre-mealvalues (pre-meal)wasselectedbyanROCcurveconstructedby plot-tingsensitivityagainstthefalse-positivefraction(1–specificity, FPF) over ranges of the pre-meal cut off points [21]. The optimalcut offpointbalancingsensitivityandspecificitywas selectedattheshoulderoftheROCcurve.Theareaunderthe curve (AUC)was calculatedusing ROCKIT software (avail-ablefromtheUniversityofChicagoDepartmentofRadiology website)[22].

3. Results

Clinicalandlaboratorydatafortheentirestudypopulation (n=210)areincludedinTable1.

3.1. Glucosevaluesatpre-mealtimepointsvs.nocturnal nadirs

Mean pre-mealglucose values wereseen toprogressively decrease throughout the day from pre-breakfast to preD time points when the study population was considered as a whole (Table 1). All mean pre-meal glucose values were significantly greater (P<0.0001) than the nocturnal nadir of 113.0±2.3mg/dL(mean±SEM).

3.2. Relationshipsbetweennocturnalnadirandmeasured oraveragedpre-mealglucosevalues

There was a strongly significant correlation (P<0.0001) between pre-meal glucose values,whether measured at preL (r=0.78) or preD timepoints(r=0.77),when tested against nocturnalglucosenadirs.Whenglucosevalueswereaveraged overpreLandpreDtimepoints(preLD),anevenbetter corre-lationwasseenwiththenocturnalnadir(r=0.83,P<0.0001). Multiple comparisonsacross thethreecorrelationcoefficients showednosignificantdifferences.AsillustratedinFig.2a,the relationship betweenthe preLDand nocturnal glucose nadir, consideredthe“predictor”(Xaxis,mg/dL)and“response”(Y axis,mg/dL) variables,respectively, wasclosetothe identity line:Y=0.82X+11.Asaconsequence,thepre-mealLDwas usedtocalculatethecutoffvaluethatwasabletopredict,with satisfactoryandbalanced sensitivity andspecificity,the pres-enceorabsenceofthedawnphenomenonwhentheincrement todefinethisglycaemicdisorderwassetat20mg/dL.

3.3. Relationshipsbetweenmagnitudeofdawn phenomenon(dawn)andmeasuredoraveraged pre-mealvalues

Allmeasuredandaveragedpre-mealvaluescorrelatedwith the dawn value. However, the correlation withthe greatest significance wasobservedwhendawn(Yaxis,mg/dL)was correlatedwiththecalculatedpre-mealLD(Xaxis,mg/dL):

r=0.58,P<0.0001.Fig.2bpresentsagraphoftherelationship: Y=0.49X+15.

The ROC curve calculated from the stepwise increase in

Fig.3.Receiveroperatingcharacteristic(ROC)curvederivedfromthestepwise increaseinpre-mealLD.Thethresholdfordefiningtheabsence/presenceof adawnphenomenon(dawn)wassetat20mg/dL.Thecutoffvalueof pre-mealLDforoptimalpredictionoftheabsence/presenceofadawnphenomenon (theoptimalbalancebetweenhighestsensitivityandspecificity)is10mg/dL, indicated in thebox atthe shoulder of theROC curve. The true-positive fraction(Y=sensitivity)andthefalse-positivefraction(X=1–specificity) indi-cate/reflectthespecificityandsensitivity,respectively,ofthepre-mealLDfor predictingadawnphenomenonatathresholdof20mg/dL.

beinginthefairrange(0.70–0.80)accordingtothetraditional academicpointsystem[21].

4. Discussion

Thepresentresultsindicatethatthenocturnalglucosenadir occurred atan averageclock timeof 03h41inthe morning, which has previously been reported as corresponding to the nocturnalgrowthhormonepeakandthebeginningofthe pro-gressiveincreaseincortisol,epinephrineandnorepinephrinein theearlymorningperiod/endofthenight[23].Thisnocturnal glucose nadir was well correlatedwith all daytime pre-meal values,althoughthe correlationimprovedwhen thepreL and preDvalueswereaveraged.Basedontheseresults,we tested theaccuracyoftherelationshipbetweenthemagnitudeofthe dawn glucose rise (dawn; the difference between the noc-turnalnadirandpre-breakfast value)vs. theotherpre-meal values,especiallythe pre-meal LD(the difference between thepre-breakfastvalueandaveragedglucosevaluesatpreLand preDtimepoints).Asmentionedabove,thepresentstudygave particular attention tothelatter difference,as ourresults had shownthatthisparameterwasbettercorrelatedwiththe mag-nitudeofthedawnglucoseriseduringtheearlymorninghours thanwithany of the otherindividual pre-mealvalues. Fur-thermore, using stepwise analysis of the pre-meal LD and calculating the ROC curves on the basis of the presence or absenceofadawnphenomenonatathresholdlevelof20mg/dL, it was foundthat the dawn phenomenon could be fairly pre-dictedbyacutoffvalueof10mg/dLfortheaveragedpre-meal LD.

For obvious reasons, nocturnal glucose values cannot be easily measuredinclinical practiceby discontinuousSMBG. Therefore,itappearsthatdefiningacutoffpointderivedfromthe averagedpre-mealLDasapredictorofthedawnphenomenon maybeofimportance,althoughmanyotherbarriersareyettobe overcomebeforethemajorityofhealthcareproviderscanobtain easyaccesstosuchnewtechnologiesasCGM[8].

Consequently, as discontinuousself-monitoring ofglucose concentrationsatthethreepre-mealtimepointsiseasily avail-able in clinical practice, the difference between preLD and pre-breakfast glucose levels can be used as a substitute for detectingthepresenceorabsenceofthedawnphenomenonand as areliablesurrogateforits quantification.Measurements of preD,preLandpre-breakfastglucosevaluescaneasilybe inte-gratedintoapatient-centredprogrammeof structuredSMBG, eveninpersonswithT2Dnottreatedwithinsulin. Such mea-surementswouldalsobehelpfulformakingbothpatientsand theirhealthcareprovidersawareofthepresenceofadawn phe-nomenon. Indeed, such a glycaemic disorder should not be ignored, given its global impact on HbA1c of approximately 0.4%(4mmol/mol)[4].Thedawnphenomenonisprobablyone oftheearlierdisordersinthenaturalhistoryofT2D[23,24].In addition,through oneof itsconsequences–theextendeddawn phenomenon,characterizedbyexcessivepost-breakfastglucose excursionsduring themorningperiod–thedawn phenomenon can contribute to overall glucose exposure in T2D [3–5,25]. Consequently,thedawnphenomenonshouldbetakeninto con-siderationinthemanagementofT2D,anditspresenceshould instigate the useof more aggressive therapyas earlyas pos-sible inthetimecourseofthedisease.Atpresent,little orno considerationisgiventothedawnphenomenonbythevarious organizations that havepublished clinical guidelines or stan-dardsofmedicalcareindiabetes[26,27],notevenwhenthese recommendations aremade withinapatient-centredapproach

[28].

L.Monnieretal./Diabetes&Metabolism41(2015)132–137 137

Disclosureofinterest

Theauthors declare that theyhaveno conflictsof interest concerningthisarticle.

Acknowledgementsandfunding

LouisMonnier,MD,(UniversityMontpellier1,Montpellier, France)takesfullresponsibilityforallthecontentsofthisarticle. Louis Monnier (University Montpellier 1, Montpellier, France)contributedtothe designofthe study,participatedto theclinicalcareofthepatients,analyzedthedataofthestudy, andwrote,reviewedandeditedthemanuscript.

Claude Colette (University Montpellier 1, Montpellier, France)contributedtotheanalysisofdataandreviewed/edited themanuscript.

SylvieDejager,MD,PhD(HospitalPitié-Salpétrière,Paris, France)contributedtothedesignandreviewedthemanuscript. DavidOwens,MD,FRCPcontributedtothedesign ofthe studyandreviewedthemanuscript.

Thisstudywassupportedbyacademicfundsprovidedbythe UniversityofMontpellier1.

References

[1]SchmidtMT,Hadji-GeorgopoulosA,RendellM,MargolisS,KowarskiA. Thedawnphenomenon,anearlymorningglucoserise:implicationsfor diabeticintradaybloodglucosevariation.DiabetesCare1981;4:579–85. [2]BolliGB,GerichJE.The“dawnphenomenon”–acommonoccurrence

inbothnon-insulindependentandinsulin-dependentdiabetesmellitus.N EnglJMed1984;310:746–50.

[3]MonnierL,ColetteC,SardinouxM,BaptistaG,Regnier-ZerbibA,Owens D.Frequencyandseverityofthedawnphenomenonintype2diabetes. DiabetesCare2012;35:2597–9.

[4]MonnierL,ColetteC,DejagerS,OwensD.Magnitudeofthedawn phe-nomenonanditsimpactoftheoverallglucoseexposureintype2diabetes. Isthisofconcern?DiabetesCare2013;36:4057–62.

[5]MonnierL,ColetteC,DunseathGJ,OwensDR.Thelossofpostprandial glycemiccontrolprecedesstepwisedegradationoffastingwithworsening diabetes.DiabetesCare2007;30:263–9.

[6]CarrollMF,SchadeDS.Thedawnphenomenonrevisited:implicationsfor diabetestherapy.EndocrPract2005;11:55–64.

[7]CarrollMF,HardyKJ,BurgeMR,SchadeDS.Frequencyofthedawn phenomenonintype2diabetes:implicationsfordiabetestherapy.Diabetes TechnolTher2002;4:595–605.

[8]MonnierL,ColetteC,BoegnerC,PhamTC,LapinskiH,BonifaceH. Con-tinuousglucosemonitoringinpatientswithtype2diabetes:why?when? whom?DiabetesMetab2007;33:247–52.

[9]InternationalDiabetesFederationGuidelineonself-monitoringofblood glucosein non-insulin treated type2 diabetes. Brussels:International DiabetesFederation;2009.Availablefrom:http://www.idf.org/webdata/ does/SMBGEN2.pdf

[10]DruckerDJ,Nauck MA.Theincretinsystem:glucagon-likepeptide-1 receptoragonistsanddipeptidylpeptidase-4inhibitorsintype2diabetes. Lancet2006;368:1696–705.

[11]MonnierL,ColetteC,DejagerS,OwensD,Responsetocommenton Monnier,etal.Magnitudeofthedawnphenomenonanditsimpactonthe overallglucoseexposureintype2diabetes:isthisofconcern?Diabetes Care2013;36:4057–62[DiabetesCare2014;37:e163].

[12]CarrRD,AlexanderCM.CommentonMonnieretal.Magnitudeofthe dawnphenomenonanditsimpactontheoverallglucoseexposureintype 2diabetes:isthisofconcern?DiabetesCare2013;36:4057–62[Diabetes Care2014;37:e161-e162].

[13]SchofieldWN,SchofieldC,JamesWPT.Basalmetabolicrate-reviewand prediction,togetherwithanannotadedbibliographyofsourcematerial. HumNutrClinNutr1985;39C(Suppl.1):1–96.

[14]WorldMedical Association DeclarationofHelsinki: recommendations guiding physicians in biomedical research involving human subjects. JAMA1997;277:925–6.

[15]Directive2001/20/CEoftheEuropeanParliamentandofthecouncilof4 April2001ontheapproximationofthelaws,regulationsandadministrative provisionsoftheMemberStatesrelatingtotheimplementationofgood clinicaltrialsonmedicinalproductsforhumanuse.OffJEurCommunities 2001;L121:33–44.

[16]Gorus F, Mathieu C, Gerlo E. How should HbA1c measurementbe reported?Diabetologia2006;49:7–10.

[17]SacksDB,BrunsDE,GoldsteinDE,MaclarenNK,McDonaldJM,Parrott M.Guidelinesandrecommendationsforlaboratoryanalysisinthe diag-nosisandmanagementofdiabetesmellitus(positionstatement).Diabetes Care2002;25:750–86.

[18]SacksDB.MeasurementsofhemoglobinA1c.Anewtwistonthepastto harmony.DiabetesCare2012;35:2674–80.

[19]MonsodTP,FlanaganDE,RifeF,CaprioS,SherwinRS,Tamborlane WV. Dosensorglucoselevelsaccurately predictplasma glucose con-centrationsduringhyperglycemiaandhyperinsulinemia?DiabetesCare 2002;25:889–93.

[20]OllertonRL,PlayleR,AhmedK,DunstanFD,LuzioSD,OwensDR. Day-to-dayvariabilityoffastingplasmaglucoseinnewlydiagnosedtype 2diabeticsubjects.DiabetesCare1999;22:394–8.

[21]HanleyJA,McNeilBJ.Themeaninganduseoftheareaunderareceiver operatingcharacteristic(ROC)curve.Radiology1982;143:29–36. [22]MetzCE.StatisticalanalysisofROCdatainevaluatingdiagnostic

per-formance.In:HerbertD,MyersR,editors.Multipleregressionanalysis: applicationinthehealthscience.NewYork:AmericanInsituteofPhysics; 1986.p.365–84.

[23]BolliGB,DeFeoP,DeCosmoS,PerrielloG,VenturaMM,CalcinaroF, etal.Demonstrationofadawnphenomenoninnormalhumanvolunteers. Diabetes1984;33:1150–3.

[24]PorcellatiF,LucidiP,BolliGB,FanelliCG.Thirtyyearsofresearchonthe dawnphenomenon:lessonstooptimizebloodglucosecontrolindiabetes. DiabetesCare2013;36:3860–2.

[25]MonnierL,ColetteC,Rabasa-LhoretR,CaubelC,AvignonA,etal. Morn-inghyperglycemicexcursions:aconstantfailureinthemetaboliccontrolof non-insulinpatientswithtype2diabetes.DiabetesCare2002;25:737–41. [26]IDF2012ClinicalGuidelinesTaskForce.Globalguidelinefortype2 dia-betes.DiabetesResClinPract2012,http://dx.doi.org/10.1016/j.diabres. 10.001.

[27]AmericanDiabetes,Association.Standardsofmedicalcarein Diabetes-2014.DiabetesCare2014;37(Suppl.1):S14–80.