ltfYwx001FYY86Y475P89YY7 - PDF document

1 ltf,YwYY86/Y475-P89Y-Y.-7 )y ?*=
ltf,YwYY86/Y475-P89Y-Y.-7 )y ?*== REVIEWOpenAccessEffectoffenugreek(Trigonellafoenum-graecumL.)intakeonglycemia:ameta-analysisofclinicaltrialsNithyaNeelakantan,MadanagopalNarayanan,RussellJdeSouzaandRobMvanDamBackgroundandaim:FenugreekisaherbthatiswidelyusedincookingandasatraditionalmedicinefordiabetesinAsia.Ithasbeenshowntoacutelylowerpostprandialglucoselevels,butthelong-termeffectonglycemiaremainsuncertain.Wesystematicallyreviewedclinicaltrialsoftheeffectoffenugreekintakeonmarkersofglucosehomeostasis.PubMed,SCOPUS,theCochraneTrialsRegistry,WebofScience,andBIOSISweresearchedupto29Nov2013fortrialsofatleast1weekdurationcomparingintakeoffenugreekseedswithacontrolintervention.Dataonchangeinfastingbloodglucose,2hourpostloadglucose,andHbA1cwerepooledusingrandom-effectsmodels.Atotalof10trialswereidentified.Fenugreeksignificantlychangedfastingbloodglucoseby-0.96mmol/l(95%CI:-1.52,-0.40;I=80%;10trials),2hourpostloadglucoseby-2.19mmol/l(95%CI:-3.19,-1.19;I=71%;7trials)andHbA1cby-0.85%(95%CI:-1.49%,-0.22%;I=0%;3trials)ascomparedwithcontrolinterventions.Theconsiderable *Correspondence:nidhineel@gmail.comSawSweeHockSchoolofPublicHealth,NationalUniversityofSingaporeandNationalUniversityHealthSystem,Singapore,SingaporeFulllistofauthorinformationisavailableattheendofthearticle ©2014Neelakantanetal.;licenseeBioMedCentralLtd.ThisisanopenaccessarticledistributedunderthetermsoftheCreativeCommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycited.Neelakantanetal.NutritionJournalhttp://www.nutritionj.com/content/13/1/7 reviewandmeta-analysisoftheeffectsoffenugreekonglucosehomeostasisbasedonacomprehensiveliteraturesearchleadingtotheidentificationofareasonablylargenumberoftrialswithanevaluationofpotentialexplana-tionsfordifferencesinstudyresults.DatasourcesandsearchesToidentifyarticlesontheeffectoffenugreekonglucosehomeostasiswesearchedMEDLINE(PubMed),SCOPUS,WebofScience,BIOSIS,andCochraneTrialsRegistryfrominceptionthroughNov29,2013usingkeysearchtermsrelatedtofenugreek(fenugreektrigonella),anexperimentalstudydesign(trialclinicaltrialinter-ventiontherapy),toidentifypotentiallyrelevantarti-cles.Thesearchstrategyutilizedbothindextermsandfreetexttosearchforsynonymsoftrigonella,fenugreekanddiabetes/healthysubjects,andwaslimitedtohumanstudies.Greyliteraturesuchasconferenceproceedings,abstracts,dissertationsandtechnicalreportswasidenti-fiedusingthesamekeytermsthroughtheelectronicsearchenginesGoogleScholar,SCIRUS,CINAHL,andProQuest.Nolanguagerestrictionwasapplied.Theresults(titles,abstractsandcitations)ofelec-tronicsearchesweredownloadedintoEndNotesoftware(EndNoteX5,2011,ThomsonReuters,Philadelphia)andinitialscreeningforeligibilitywasperformedbytwoinde-pendentreviewers(NithyaNeelakantan,MadanagopalNarayanan).Whenassessmentofeligibilitybasedonthetitleandabstractwasinsufficient,thefulltextofthearti-cleswasobtained.Thesecondscreeningofthosefulltextarticleswasthenindependentlyperformedbyatleasttworeviewers(NithyaNeelakantan,MadanagopalNarayanan,RobMvanDam).Disagreementswereresolvedbycon-sensus.Thekappafortheinter-reviewerreliabilitywas0.78.Studyauthorswerecontactedtoverifyresultsandmethodologicalqualityofretrievedarticleswhereneces-sary.WeusedthePreferredReportingItemsforSystem-aticRe

2 viewsandMeta-Analysis(PRISMA)statementto
viewsandMeta-Analysis(PRISMA)statementtoreportourfindings[30].StudyselectionWeincludedclinicaltrialsthatcomparedsingleherbpreparationsoffenugreekinanydoseorformwithacontrolinterventionthatwaseitherplaceboornotreat-mentandevaluatedeffectsonmarkersofglycemia[fast-ingbloodglucose,2hrpostloadglucose,glycosylatedhemoglobin(%HbA1c)and/orfastingseruminsulinlevels].Weexcludedtrialsthatusedcombinationprepa-rationsoffenugreekwithotherherbs,non-humanstud-ies,observationalstudies,literaturereviews/editorials/letters/casereports,andarticlesnotreportingtheout-comesofinterest.Wealsoexcludedtrialswithinterven-tionsthatlastedlessthan7days.ThenumberofarticlesthatdidnotmeettheeligibilitycriteriaandthereasonsfortheirexclusionareshowninFigure1.DataextractionandqualityassessmentDetailsoftrialdesign,studysetting,population,ran-domization,blinding,samplesize,durationoffollow-up,participantcharacteristics,interventions,totaldailydoseandoutcomecharacteristicswereindependentlyextractedbytworeviewers(NithyaNeelakantan,MadanagopalNarayanan),usingastandardizeddataextractionform.Differencesindataextractionwereresolvedbyathirdre-viewer(RobMvanDam).ThequalityassessmentwasconductedusingtheCONSORTstatementforherbaltrials[31]bytworeviewers(NithyaNeelakantan,MadanagopalNarayanan),withdisagreementsresolvedbyconsensus.Fromeachtrial,dataonmeanandSDforalloutcomesofinterestwereextracted.Iftrialsreportedfastingbloodglucoseand2hrpostloadglucose(glucoseconcentra-tions2hoursafterthestartoftheoralglucosetolerancetest)inunitsofmg/dL,thiswasconvertedtothestan-dardizedinternationalunit[32]ofmmol/Lbymultiplyingtheglucosevaluesinmg/dLby0.0555;forfastingseruminsulin,wedividedtheseruminsulinvaluesreportedinpmol/Lby6.945andreportedtheresultsinmU/L.Paralleltrialsgenerallyreportedthebaselinemeanandstandarddeviationandfollow-upmeanandstandardde-viation,butnotthestandarddeviation(SD)ofchangefortheinterventionandcontrolgroups.Forparalleltri-als,thenetchangesineachoutcomemeasurewerecal-culatedasthechangeintheinterventiongroupminusthechangeinthecontrolgroup.Forcrossovertrials,netchangesintheoutcomemeasureswerecalculatedasthevalueoftheoutcomemeasureattheendoftheinter-ventionperiodminusthevalueoftheoutcomemeasureattheendofcontrolperiod.WeestimatedtheSDofthechangeonthebasisofreportedpvaluesfordifferencesinmeans,ifavailable[33].Weusedthep-valuescutoffifitwasonlyreportedthatap-valuewasbelowathresh-old(e.g.,0.05ifp0.05wasreported)leadingtoconser-vativeestimates[34,35].Ifp-valueswerenotreported,weimputedSDofthechangebyusingapooledcorrel-ationcoefficientbetweenbaselineandfinalmeasure-mentsfromameta-analysisofcorrelationcoefficientsfromthosetrialsreportingsufficientdata.Wederivedcorrelationcoefficientsforindividualtrialsaccordingastandardformula[33]andwethenimputedthesecorre-lationsintothemeta-analysisastransformedzscores(±SEs)toestimatethepooledcorrelationcoefficient[36].ForHbA1candfastingseruminsulinmeasures,duetosmallnumberoftrials,weestimatedtheSDsofthechangeassumingaconservative0.5correlationandper-formedasensitivityanalysisassumingalternativevaluesof0.25or0.75.Toinvestigatetheeffectofimputedwithin-personcorrelationcoefficients,weperformedsensitivityNeelakantanetal.NutritionJournalPage2of11http://www.nutritionj.com/content/13/1/7 analyseswitharangeofcorrelationcoefficients(0.25,0.50and0.75)[37],thepooledestimatesdidnotchangesubstantially.

3 Meta-analysisThemeta-analysiswasperforme
Meta-analysisThemeta-analysiswasperformedaccordingtothemethodsdescribedbyCurtinetal.[38].Inthecombineddesignmeta-analysis,thepooledestimateoftreatmenteffectcombiningparallelandcrossovertrialresultswastheweightedsumoftheseparatetreatmenteffectsesti-mated,respectively,fromparallelandcrossovertrialsdividedbythesumoftheassociatedweights.Weantici-patedlargedifferencesinthefenugreekdrugpreparationformat,activecomponents/chemicalcomposition,ad-ministrationofsupplements,anddosagesaswellasvariationinthestudypopulationandstudydesign.Therefore,weaprioridecidedtousearandomeffectsmodelforthismeta-analysis.Hence,foreachoutcomemeasure,weightedmeandifferencesandcorresponding95%confidenceintervals(CI)werecalculatedbyusingDerSimonianandLairdrandom-effectsmodels.Wealsoconductedseparatemeta-analysesforparallelandcross-overtrialsfortheprimaryoutcomemeasures,fastingbloodglucose,and2hrpostprandialglucose.HeterogeneityinstudyresultswastestedbyusingtheCochranstatistic(andassociatedpvalue),andwasquantifiedbythestatistic.Theprovidesanestimateofthepercentageofvariationinstudyresultsthatisex-plainedbybetween-studyheterogeneityratherthansam-plingerror[39].Potentialsourcesofheterogeneitywereinvestigatedusingaprioridefinedstratifiedanalysesbystudydesign(parallelorcrossover),dailydoseoffenu-greekextract(g,510gor-20;.80;10g),studyduration(daysor0-6;
.4;=30days),randomization(yesorno),blinding(yesorno),baselineBMI(5or-8.;က=25kg/mstudyprecision(SEoftheeffectestimateaboveorbelowthemedian),geographicalregion(Indiavs.othercoun-tries)andage(aboveorbelowthemedianmeanageofallstudies).Meta-regressionanalyseswereusedtoassessthesignifi-canceofdifferencesintheeffectsoffenugreekbetweenstrata.p-valuesfortheoverallF-testforacommonmean 161 Potentially relevant articles screened 38 Duplicates 199 Potentially relevant articles identified 32 full text articles assessed for eligibility 12 Cross-sectional study, survey 23 Excluded10 Acute studies ()6 Combination preparation of fenugreek with other herbs4 Inappropriate /no control group2 Unrelated study outcomes1 Short review 9 articles reporting 10 trials included in meta- Figure1Selectionoftrialsformeta-analysis.(Searchwasconductedtoidentifyarticlesupto29Nov2013).Neelakantanetal.NutritionJournalPage3of11http://www.nutritionj.com/content/13/1/7 amongstthreeormoregroupswereobtainedusingonewayANOVA.PublicationbiaswasinvestigatedbyvisualinspectionoffunnelplotsandbytheEggerregressiontestandtheBeggadjustedcorrelationtest[40].Therobust-nessofthefindingsofthemeta-analysistodifferentas-sumptionswereexaminedinasensitivityanalysisusingbothfixedandrandomeffectsmodels,inwhichthemeta-analysisestimateswerecomputedomittingonetrialatatimetoassesstheinfluenceofeachindividualtrial[41].Alltestsweretwo-sidedandp0.05wasconsideredsta-tisticallysignificant.ThedatawereanalyzedbyusingStataversion11(StataCorp,CollegeStation,Texas).SearchresultsWeidentified161potentiallyrelevantarticlesandscreenedtheabstractsforeligibility.Theflowoftrialselectionisre-portedinFigure1.Weevaluated32fulltextarticlesinde-tail.Ninearticlesreporting10trialsmettheinclusioncriteriaforthemeta-analysis.Ofthese10trials,onetrial[42]reportedresultsseparatelyforparticipantswithmildandseveretype2diabetes,andthus11datapointsarepresentedinTable1.Fiveofthesehadaparalleldesignandsixhadacrossoverdesign.Trialcharacterist

4 icsThemeanageofparticipantsinthetrialsra
icsThemeanageofparticipantsinthetrialsrangedfrom22.0to54.4years(median:43.1y),andthemedianpercentageofmaleswas76%.Mosttrialsincludedpar-ticipantswithtype2diabetestreatedwithdietororalanti-diabeticmedication(Table1).Onetrialwascon-ductedinpersonswithtype1diabetesandtwotrialsin-cludedoverweightornon-overweightparticipantswithoutdiabetes.Thesamplesizerangedfrom5to15participantsforcrossovertrialsandfrom25to69participantsforpar-alleltrials.Thesamplesizeforalltrialscombinedwas278.Thedailydoseoffenugreekseedrangedfrom1gto100g(median:25g),andthestudydurationfrom10to84days(median:30days).Fenugreeksupplementswereadministeredaspowderedfenugreekseeds,debitterizedpowderedfenugreekseeds,orhydro-alcoholicseedextracteitherinformofcapsulesorasaningredientofunleav-enedbread.Thesewereprovidedinequaldoses2to3timesperday.All10trials[7,9,17,22,23,25,42-44](11datapoints)reportedfastingbloodglucose,7trials[7,9,17,22,23,42,44](8datapoints)reported2hrglucose,5trials[22,23,25,43,44]reportedfastingseruminsulinand3trials[9,17,23]reportedHbA1c.Withregardtostudyquality,onetrial[42]didnotreportwhethergroupswererandom-ized,noneofthetrialsreporteddetailsonthemethodofrandomizationorallocationconcealment,andmosttrials[7,17,22,42,44]didnotreporttheblindingstatus.Onlyfourtrialsreportedthepercentageofdrop-outs[17,23,25,43].Ofwhichthreestudies[23,25,43]havereportedlowpercentagedrop-outrangingbetween0to5%andonestudy[17]hasreported25%drop-out.Thesesubjectswerenotincludedinthestatisticalanalysesthatwereconductedfortheprimarystudies.Mosttrialsprovidedinformationontheconcentrationofcomponentsoftheusedfenugreekpreparations.Reportedcomponentsin-cludeddiosgenine,saponins,trigonelline(1.4%)and4-hydroxyisoleucine(1.5%)forthehydro-alcoholicextract[25,43];alkaloids,carpaine,erythricine,trigonelline,mele-tin,andsaponinsforpowderedfenugreekseeds[22,44];andlipids(0.1%),protein(28.3%),starch(6.5%),totalfiber(51.7%[gum19.2%])fordebitterizedfenugreekseeds[22,44].EffectonglucosehomeostasisFastingbloodglucoseAll10trials(11datapoints)wereincludedinthemeta-analysisoffastingbloodglucose.TheindividualtrialresultsandthepooledestimatesbytrialdesignsareshowninFigure2.Basedontheoverallpooledestimate,fenugreeksignificantlyreducedbloodglucoselevelsascomparedwithcontroltreatments(pooledmeandiffer-ence=-0.96mmol/l;95%CI:-1.52,-0.40;p=0.001).Therewaslargeheterogeneityinstudyresults(I=80%;p0.001).Therewerenosignificantdifferencesintheeffectsoffenugreekonfastingglucosebystudydesign,studyduration,geographicalregion,meanageandmeanBMIofthestudypopulation(Table2).However,theef-fectoffenugreekonfastingbloodglucosedifferedsig-nificantlybydiabetesstatuswithsubstantialeffectsinpersonswithtype1diabetesandtype2diabetes,butnotinpersonswithoutdiabetes.Therewasalargevariationinthedoseoffenugreekusedrangingfrom1gperdayto100gperdayoffenugreekseeds.Theeffectsizedif-feredsignificantlybythedoseoffenugreekusedsuggest-ingnoeffectsforthestudiesusinglowdoses(g/day)andgreatereffectswithhigherdosesoffenugreek.Ef-fectsoffenugreekonfastingglucosealsodifferedbypreparationmethodofthefenugreeksupplement.Stud-iesusingdebitterizedfenugreekpowderedshowedthegreatestreductioninglucoselevels,buttheseweremostlythesametrialsthatadministeredthehighestdoseoffenugreek.Similarly,thetrialsusingahydro-alcoholicextractwerethesametrialsthatusedthelowestdose.Heterogeneityineffectsonfastingglu

5 cosewaspartlyex-plainedbyfenugreekdose(I
cosewaspartlyex-plainedbyfenugreekdose(Iresidual=69%;adjustedR47%),fenugreekpreparationmethod(Iresidual=38%;ad-justedR=83%),anddiabetesstatusofthestudypopu-lation(I=55%;adjustedR=61%).Thefunnelplotforeffectsoffenugreekonfastingbloodglucosebystudyprecisionappearedtobeasym-metrical(Additionalfile1:FigureS1,Beggtest,p=0.10)andtheEggertestwassignificant(p=0.03)suggestingpotentialpublicationbias.However,thesetestsarebasedNeelakantanetal.NutritionJournalPage4of11http://www.nutritionj.com/content/13/1/7 Table1Characteristicsofthe10trials(11datapoints)includedinthemeta-analysisoftheeffectsoffenugreekseedsonglycemiaLeadauthor,yearandcountryPopulation,medicationDesignAge,yMale(%)Samplesize,NDuration,Fenugreek,preparationDailydose,gControlOutcomesStudyqualityRCTBlindingDropoutBordiaetal.,1997[],IndiaMildT2DM,NRPLNRNR4030Capsule,PS5PlaceboFBG,2hrglucoseNRNRNRBordiaetal.,1997[],IndiaSevereT2DM,NRPLNRNR4030Capsule,PS5PlaceboFBG,2hrglucoseNRNRNRChevassusetal.,2010[],FranceOverweight,NRPL38.01004042Capsule,H1.176PlaceboFBG,FSIYesDB5%Guptaetal.,2001[],IndiaT2DM,SU,BIPL51.0762556Capsule,H1PlaceboFBG,2hrglucose,HbA1c,FSIYesDB4%Luetal.,2008[],ChinaT2DM,SUPL54.4556984Capsule,PS6.3PlaceboFBG,2hrglucose,YesDBNRAlamdarietal.,2009[],IranT2DM,Diet,OADCO43.11001256PS8UnspecifiedFBG,2hrglucose,YesNR25%Chevassusetal.,2009[],FranceHealthy,NRCO22.01001214Capsule,H1.176PlaceboFBG,FSIYesDB0%Raghurametal.,1994[],IndiaT2DM,BICO46.6NR1015Chapati,PS25ChapatiFBG,1hrglucose,YesNRNRSharmaetal.,1990[],IndiaT1DM,InsulintherapyCO22.7701010Chapati,DPS100ChapatiFBG,2hrglucose,FSIYesNRNRSharmaetal.,1990[],IndiaT2DM,BI/metforminCO46.0671510Chapati,DPS100ChapatiFBG,2hrglucose,FSIYesNRNRSharmaetal.,1990[],IndiaT2DM,NRCO42.0NR520Chapati,DPS100ChapatiFBG,2hrglucoseYesNRNRnotreported,type1diabetesmellitus,type2diabetesmellitus,oralantidiabeticdrug,crossovertrials,Dailydose,fenugreektotaldailydoseingrams,powderedfenugreekseeds,Hydro-alcoholicextractoffenugreekseeds,Debitterizedfenugreekseedpowder,fastingbloodglucose,%glycosylatedhemoglobin,seruminsulin,randomizedcontroltrial,doubleblinded,Fenugreekincorporatedintochapati(unleavenedbread).Neelakantanetal.NutritionJournalPage5of11http://www.nutritionj.com/content/13/1/7 ondetectinganassociationbetweenstudyprecision(lowerSEofeffectestimates)andeffectsize.Lessprecisestudiesalsotendedtouseagreaterdoseoffenugreek(r=0.51betweendoseandSEoftheeffectestimates)andwecouldthereforenotdistinguishbetweenpoten-tialpublicationbiasandthedoseoffenugreekused.2hrglucoseSeventrials(8datapoints)reportedeffectsoffenugreekon2hrglucosevalues.Forestplotsof2hrglucoseef-fectsinindividualtrialsandthepooledanalysesareshowninFigure3.Meta-analysisofthetrialsyieldedapooledestimatefortheeffectoffenugreekon2hrglu-coseof-2.19mmol/l(95%CI:-3.19,-1.19,p1).Therewaslargeheterogeneityinstudyresults(I=71%;p=0.001).Stratifiedanalysesofeffectsoffenugreekon2hrglucoseaccordingtostudycharacteristicsarepre-sentedinTable2.Asobservedforfastingglucose,ahigherdosewasassociatedwithgreatereffectson2hrglucoseconcentrations.Wealsoobservedstrongeref-fectsfortrialswithstudydurationlessthan30daysthanfortrialswithalongerdurationandfortrialsthatad-ministereddebitterizedfenugreekseedpowderthanfortrialsusingotherfenugreekpreparations.However,thetrialsofshortdurationthatuseddebitterizedfenugreekpowderwerethesamethreetrialsthatadministereda

6 substantiallyhigherdoseoffenugreekthanth
substantiallyhigherdoseoffenugreekthantheothertri-als.Heterogeneityineffectsoffenugreekon2hrglucosewaspartlyexplainedbyfenugreekdose(Iresidual=42%;adjustedR=72%)andfenugreekpreparationmethod=50%;adjustedR=61%).Thefunnelplotfor2hrglucosedataappearedtobeasymmetrical(Additionalfile1:FigureS2,Beggtest,p=0.05)andtheEggertest(p=0.03)alsoprovidedevidenceforpublicationbias.Again,lessprecisestudiesalsotendedtouseagreaterdoseoffenugreek(r=0.54betweendoseandSEoftheeffectestimates)andwecouldnotdistinguishbetweenpotentialpublicationbiasandthedoseoffenugreekOtheroutcomemeasures(HbA1candfastingseruminsulin)Threetrials(2paralleland1crossovertrial)reportedtheeffectsoffenugreekonHbA1c.Fenugreeksignifi-cantlyreducedHbA1cvaluesascomparedwithcontroltreatment(pooledmeandifference-0.85%;95%CI:-1.49%,-0.22%,p=0.009)(Additionalfile1:FigureS3) NOTE: Weights are from random effects analysis .. ParallelCrossoverSharma et al.(1990),T2DM-ISubtotal (I-squared = 77.2%, p = 0.001)Chevassus et al.(2009),HealthyGupta et al.(2001),T2DMSharma et al.(1990),T1DMLu et al.(2008),T2DMAlamdari et al.(2009),T2DMStudyIDBordia et al.(1997),Severe T2DMChevassus et al.(2010),OverweightSubtotal (I-squared = 85.8%, p = 0.000)Sharma et al.(1990),T2DM-IIBordia et al.(1997),Mild T2DMRaghuram et al.(1994),T2DM-0.96 (-1.52, -0.40)-2.32 (-4.44, -0.20)-1.20 (-2.03, -0.38)-0.20 (-0.38, -0.02)-0.22 (-2.08, 1.64)-4.20 (-6.73, -1.67)-1.77 (-2.47, -1.07)-0.84 (-1.58, -0.10)ES (95% CI)-0.82 (-2.10, 0.46)0.38 (-0.08, 0.84)-0.78 (-1.92, 0.37)-2.27 (-3.88, -0.66)-1.53 (-3.08, 0.02)-0.54 (-1.52, 0.44)52.6815.035.663.7112.2912.01%Weight8.4713.8647.326.757.0510.39-0.96 (-1.52, -0.40)-2.32 (-4.44, -0.20)-1.20 (-2.03, -0.38)-0.20 (-0.38, -0.02)-0.22 (-2.08, 1.64)-4.20 (-6.73, -1.67)-1.77 (-2.47, -1.07)-0.84 (-1.58, -0.10)ES (95% CI)-0.82 (-2.10, 0.46)0.38 (-0.08, 0.84)-0.78 (-1.92, 0.37)-2.27 (-3.88, -0.66)-1.53 (-3.08, 0.02)-0.54 (-1.52, 0.44)52.6815.035.663.7112.2912.01%Weight8.4713.8647.326.757.0510.39 favors fenu g reek favors control 0 -7 -5 -3 -1 0 1 3 Figure2Forestplotoftheeffectoffenugreekonfastingbloodglucose.Theeffectsinindividualtrialsaredepictedasopensquareswith95%confidenceintervals(CIs).Pooledestimateswith95%CIsaredepictedasopendiamonds.Neelakantanetal.NutritionJournalPage6of11http://www.nutritionj.com/content/13/1/7 Table2Stratifiedmeta-analysesoftheeffectsoffenugreekonfastingbloodglucoseand2hourpostloadglucoseaccordingtotrialandparticipantcharacteristicsCharacteristicsFastingbloodglucose2hr-postloadglucosenPooledestimateI(95%CI)PnPooledestimateI(95%CI)POverall110.96(0.40)80(65,89)82.19(1.19)71(40,86)0.001StudydesignParallel50.78(1.93,0.37)86(69,94)0.4841.71(0.70)59(0,86)0.070.33Crossover61.20(0.38)77(49,90)0.0143.32(0.75)82(54,93)0.001India71.43(0.60)48(0,78)0.070.2162.60(1.13)73(39,88)0.0020.50Others40.55(1.25,0.16)89(75,95)21.59(0.10)74(0,94)StudyprecisionBelowmedian50.54(1.15,0.07)86(69,94)0.1041.55(0.57)67(2,89)0.030.11Abovemedian61.68(0.74)43(0,77)0.1243.42(1.28)67(4,89)0.03StudypopulationHealthy20.05(0.51,0.61)82(23,96)0.02T1DM14.20(0.0113.20(6.92,0.52)T2DM81.21(0.73)26(0,67)0.2272.14(1.09)74(45,88)0.001RandomizationYes90.94(0.32)83(69,91)0.8362.54(1.21)73(37,88)0.0030.49Unknown21.11(0.12)0(0,100)0.4921.58(3.61,0.45)81(17,95)0.02BlindingstatusDouble-blinded40.44(1.22,0.34)88(72,95)0.1521.95(0.88)16(0,55)0.280.74Un

7 known71.37(0.67)47(0,78)0.0862.43(1.04)7
known71.37(0.67)47(0,78)0.0862.43(1.04)78(50,90)0.000Dailydose(grams)30.02(0.46,0.50)63(0,89)0.0711.34(2.83,0.14)5-1041.27(0.74)21(0,88)0.280.0141.55(0.57)67(2,89)0.040.02-10;ٰ.;怀1042.07(0.55)68(9,89)0.0234.42(2.89)8(0,35)0.58Powderedseed51.12(0.61)29(0,72)0.2341.55(0.57)67(2,89)0.03Hydro-alcoholicextract30.02(0.46,0.50)63(0,89)0.070.00211.34(2.83,0.14)Debitterizedseedpowder32.68(1.54)0(0,90)0.4234.42(2.89)0(0,90)0.58Studyduration(days)51.48(0.32)80(52,91)0.0010.3934.42(2.89)8(0,35)0.580.020-1;Ƒ&#x.600;=3060.78(1.68,0.11)83(65,92)51.49(0.70)55(0,84)0.06MeanAge(years)40.73(1.61,0.16)86(67,94)24.14(2.47)0(0,100)0.58�.00;
-47;7.7;=43.151.11(0.47)45(0,80)0.120.6041.90(0.61)67(4,89)0.030.21MeanBMI(kg/m31.62(0.02)93(82,97)0.2922.52(1.30)0(0,100)0.710.17�.00;
-13;⤀=2530.19(1.14,0.76)74(12,92)0.0221.01(0.30)0(0,100)0.62bodymassindex,confidenceinterval,pvalueforheterogeneity,pvalueforeffectmodification.Thecutoffsarebasedonthemedianofstandarderroroftheeffectsizeforfastingbloodglucoseand2hrglucoserespectively;Formeanageandstudyduration,thecutoffsarebasedonthemedianvaluesof11datapoints.-valuefortheoverallF-testforacommonmeanamongstthethreegroups(i.e.,studypopulation,andfenugreekdrugpreparationformat),theoverallp-valuefordailydosewasobtainedbymodelingthisasacontinuousvariableinmeta-regressionanalysis.Neelakantanetal.NutritionJournalPage7of11http://www.nutritionj.com/content/13/1/7 withoutsignificantheterogeneityinstudyresults(I0%;p=0.78).Fivetrialsreportedfastingseruminsulinconcentra-tions.However,fastinginsulinmayhavebeenaffectedbyexogenousinsulinuseinpersonswithtype1diabetesandfastinginsulinvalueshadastronglyskeweddistri-butioninpersonswithtype2diabetes.Therefore,weonlyincludedtwotrials[25,43]inpersonswithoutdia-betesinourmeta-analysisoftheeffectsoffenugreekonfastinginsulin.Thepooledeffectoffenugreekonfastingseruminsulinwasnotstatisticallysignificant(pooledmeandifference=-1.42mU/L;95%CI:-3.04,0.19mU/L;p=0.08)(Additionalfile1:FigureS4).Heterogeneityforeffectsonfastinginsulinwassubstantial,butnotstatisti-callysignificant(I=62%;p=0.10).SubgroupanalyseswerenotperformedforHbA1candfastingseruminsu-linbecauseofthelimitednumberoftrialsfortheseoutcomes.SensitivityanalysesWeconductedasensitivityanalysisexcludingthetrialthatdidnotreportrandomizationstatus.Thisexclusionhadlittleeffectonthepooledeffectonfastingbloodglucose(-0.94;95%CI:-1.56,-0.32)or2hrglucose(-2.54;95%CI:-3.87,-1.21).Wealsoconductedsensitiv-ityanalysesfortheeffectsoffenugreekonfastingbloodglucoseand2hrglucoseconcentrationsomittingonestudyatatime.Noneoftheindividualtrialsdramaticallyinfluencedpooledeffectestimates,whichrangedfrom-0.78mmol/l(95%CI:-1.31,-0.25)to-1.20(95%CI:-1.85to-0.56)forfastingbloodglucoseandfrom-1.79mmol/l(95%CI:-2.69,-0.90)to-2.52(95%CI:-3.64to-1.39)for2hrglucose.Wealsoconductedsensitivityanalyseswithsimultaneousexclusionoftrialsconductedbythesameresearchers.ThethreetrialsreportedbySharmaetal.[22,44]hadamuchlargerdoseoffenu-greekthantheothertrialsandalsothelowestprecision.Inasensitivityanalysisthatexcludedthesetrials,thepooledeffectonfastingbloodglucose(-0.62mmol/l,95%CI:-1.14,-0.10)and2hglucose(-1.49mmol/l,95%CI:-2.28,-0.70)remainedstatisticallysignificant.WealsoconductedasensitivityanalysisafterexcludingtheChevassustrials([25,43])thatweretheonlytrialsinper-sonswithoutdiabetes,

8 usedalowdose,andhadthelar-gestprecision.
usedalowdose,andhadthelar-gestprecision.Afterexclusionofthesestudies,thepooledeffectestimateforfastingglucosewaslarger(-1.35mmol/L;95%CI:-1.92,-0.74).AdverseeffectsChevassusetal.reported2casesofspecificurinesmelland1caseofabdominalpaininonetrial[25]and4casesofmildgastrointestinalsymptoms,and1caseof NOTE: Weights are from random effects analysis .. Lu et al.(2008),T2DMIDAlamdari et al.(2009),T2DMCrossoverSubtotal (I-squared = 58.5%, p = 0.065)StudySharma et al.(1990),T2DM-ISharma et al.(1990),T1DMSubtotal (I-squared = 82.1%, p = 0.001)Gupta et al.(2001),T2DMBordia et al.(1997),Severe T2DMBordia et al.(1997),Mild T2DMSharma et al.(1990),T2DM-II-2.19 (-3.19, -1.19)-2.44 (-3.74, -1.14)ES (95% CI)-0.91 (-1.72, -0.10)-1.71 (-2.73, -0.70)-6.10 (-10.12, -2.08)-3.20 (-6.92, 0.52)-3.32 (-5.90, -0.75)-1.34 (-2.83, 0.14)-0.61 (-1.67, 0.46)-2.68 (-4.12, -1.24)-4.37 (-6.23, -2.51)100.0015.13Weight17.9360.01%4.775.3539.9914.0616.5414.2911.95-2.19 (-3.19, -1.19)-2.44 (-3.74, -1.14)ES (95% CI)-0.91 (-1.72, -0.10)-1.71 (-2.73, -0.70)-6.10 (-10.12, -2.08)-3.20 (-6.92, 0.52)-3.32 (-5.90, -0.75)-1.34 (-2.83, 0.14)-0.61 (-1.67, 0.46)-2.68 (-4.12, -1.24)-4.37 (-6.23, -2.51)100.0015.13Weight17.9360.01%4.775.3539.9914.0616.5414.2911.95 favors fenu g reek favors control 0 -11 -9 -7 -5 1 3 Figure3Forestplotoftheeffectoffenugreekon2hourpostloadglucose.Theeffectsinindividualtrialsaredepictedasopensquareswith95%confidenceintervals(CIs).Pooledestimateswith95%CIsaredepictedasopendiamonds.Neelakantanetal.NutritionJournalPage8of11http://www.nutritionj.com/content/13/1/7 specificurineandsweatsmellinboththetreatmentgroupandthecontrolgroupintheirothertrial[43].Ofthe12participantsinthetreatmentgroupinthestudybyGuptaetal.[23],5developeddyspepsiaandmildab-dominaldistentionforthefirstfewdaysoftherapy.Thissubsidedoncontinuationofthetherapy.Norenalorhepaticsideeffectswerereportedandtherewerenowithdrawalsduetothesideeffects.Luetal.[9]reportedthat2outof46participantsofthetreatedgroupsuf-feredfromstomachdiscomfortandnausea,andonefromdiarrheaduringthetreatmentperiod.Thesesymp-tomsdisappearedafter2daydrugwithdrawalwithoutspecialtreatment.NoadversereactionwasfoundafterthetreatmentresumedandallparticipantsfinishedtheDiscussionInourmeta-analysisof10clinicaltrials,intakeoffenu-greekseedsresultedinasignificantreductioninfastingbloodglucose,2hrglucose,andHbA1c.However,weobservedsubstantialheterogeneityinstudyresults.Dif-ferencesinthediabetesstatusofparticipantsandthelargevariationindoseoffenugreekseedextractusedandtypeofpreparationappearedtobecontributorstovariationinstudyresults.Nomajorharmfulsideeffectsoffenugreekwerereportedinallincludedstudies.Weonlyfoundasignificantreductioninglucosepa-rametersfortrialsthatadministeredmediumtohighdoses(5g)offenugreekseedpowderandnotfortrialsthatadministeredlowdoses(2g)ofhydro-alcoholicextracts.Mediumtohighdoses(range:525g)offenu-greekseedpowderalsoloweredpostprandialglucoselevelsinacutestudies[6,18,19,21,45,46].Lowerdoses,asusedinthreeofthetrialsinourmeta-analyses,werenotevaluatedinacutestudiesoffenugreek.Themechanismsbywhichfenugreekmaylowerbloodglucoselevelshavenotbeenwellestablishedinhumans.Acutehypoglycemiceffectsoffenugreekseedsanditsextracthavebeenevaluatedinindividualswithandwith-outdiabetes[18,22,44].Wholefenugreekrawseeds,extractedseedpowder,cookedseeds(25g)andgumiso-lateofse

9 eds(5g)decreasedpostprandialglucoselevel
eds(5g)decreasedpostprandialglucoselevels,whereasdegummedseeds(25g)showedlittleeffect[18].Thesefindingssuggestthatacuteeffectsoffenu-greekseedsaremainlyduetothegumfraction,butdonotexcludealongertermeffectofotherfenugreekcom-ponentsonglycemia.Animalstudiesalsoindicatethatthesolublefiberfractionoffenugreekseedsreducestherateofenzymaticdigestionandtheabsorptionofglu-cosefromthegastrointestinaltract[8].However,datafromotherstudiessuggestaneffectofotherfenugreekcomponentsonglucosehomeostasis.Indiabeticrats,tri-gonellineingestionincreasedinsulinsensitivityandre-ducedbloodglucoselevels[47].Inaddition,anovelaminoacidderivativeextractedfromfenugreekseeds,4-hydroxyisoleucine,stimulatedglucose-dependentinsulinreleaseinisolatedratandhumanpancreaticisletcells[14].Inatrialofacuteeffectsinhealthyvolunteers,tri-gonellinereducedtheearlyglucoseresponseduringanOGTT[48].Theonlypreviousmeta-analysisoftheeffectsoffenu-greekonglycemiaincludedonlytwoclinicaltrialsascomparedwith10inthecurrentmeta-analysis[26].Strengthsofourstudyincludedthecomprehensivelit-eraturesearchleadingtotheidentificationofareason-ablylargenumberoftrialsandadetailedanalysisofpotentialsourcesofheterogeneityinstudyresults.Ourstudyalsohasseverallimitationsthatneedtobeconsid-eredintheinterpretationoftheresults.First,thequalityoftheincludedtrialswasgenerallypoor.Noneofthetri-alsreportedthemethodsofrandomizationorallocationconcealment,andonlyafewtrialsprovidedinformationonblindingstatusanddrop-outrates.Inaddition,withsomeexceptions[9]itwasunclearwhetherotherdia-betesmedicationremainedconstantduringthetrial.Mostoftheincludedcrossovertrialsdidnottestthecarryovereffectorreportawashoutperiod.However,wedidnotfindadifferenceinresultsbetweenparallelandcross-overtrialssuggestingthatcarryovereffectsdidnotsubstantiallyaffecttheresults.Second,testsforpub-licationbiassuggestedthatsuchbiasmayhavebeenpresent.Testsforpublicationbiasarebasedondetectingdifferencesineffectsizesbystudyprecisionwithagreatereffectsizeforlessprecise()studiessug-gestingthepresenceofpublicationbias.Inourmeta-analysislessprecisestudieswerealsomorelikelytouselargerdosesoffenugreek.Differencesindosearethusapossiblealternativeexplanationfortheobservedstudyeffect,butwewereunabletodistinguishbetweentheeffectsofdoseandpublicationbiasoneffectsizes.Finally,weonlyfoundasignificanteffectonglycemiaforpowderedfenugreekseedsandourfindingsdonotapplytootherformsoffenugreekandmaydifferforotherstrainsasaresultofnaturalvariationinactiveingredients.Oursystematicreviewandmeta-analysissuggestthatfenugreekseedsmaycontributetobetterglycemiccon-trolinpersonswithdiabetesmellituswithasimilarmagnitudeofeffectasintensivelifestyle[49]orotherpharmaceuticaltreatmentaddedtostandardtreatment[50].Fenugreekiswidelyavailableatlowcostandgener-allyacceptedinresourcepoorcountriessuchasIndiaandChinawherealargeproportionofpersonswithdia-betesintheworldreside.Therefore,fenugreekmaybeapromisingcomplementaryoptionfortheclinicalman-agementofdiabetes.Thepreviouslyreportedlipidlow-eringeffectoffenugreekmaybeanadditionalbenefit[19,44,51].However,giventhelimitedqualityoftheNeelakantanetal.NutritionJournalPage9of11http://www.nutritionj.com/content/13/1/7 includedtrialsandpotentialforpublicationbias,alargerdoubleblindrandomizedtrialshouldbeconductedac-cordingtorigorousstandardsforherbalinterventions[31]withanappropriateran

10 domizationprocedure,anadequatemethodofal
domizationprocedure,anadequatemethodofallocationconcealmentandtrans-parentreportingofthesemethods.Thefenugreekherbalproductmustbestandardizedandtestedforthecom-positionandcanbeadministeredintheformofcapsuleswitharecommendeddoseofatleast5gperday.Inordertoprovidemoreconclusiveevidenceonthebene-fitoffenugreekforglucosehomeostasis,atrialinatleast100(50subjectsineachofthestudyarms)personswithdiabetesiswarranted.ThedurationshouldpreferablybeatleastthreemonthstobeabletoevaluateeffectsonHbA1clevelsandgiventhelongerdurationaparalleltrialappearsmostappropriate.AdditionalfileAdditionalfile1:FigureS1.Funnelplotforeffectoffenugreekonfastingbloodglucose.Thesolidlinerepresentsthepooledeffectestimateexpressedastheweightedmeandifferenceandthedashedlinesrepresentpseudo-95%confidencelimits.FigureS2.Funnelplotforeffectoffenugreekon2hourpostloadglucose.Thesolidlinerepresentsthepooledeffectestimateexpressedastheweightedmeandifferenceandthedashedlinesrepresentpseudo-95%confidencelimits.FigureS3.ForestplotoftheeffectoffenugreekonHbA1c.Theeffectsinindividualtrialsaredepictedasopensquareswith95%confidenceintervals(CIs).Thepooledestimatewith95%CIisdepictedasanopendiamond.FigureS4.Forestplotoftheeffectoffenugreekonfastingseruminsulin.Theeffectsinindividualtrialsaredepictedasopensquareswith95%confidenceintervals(CIs).Pooledestimatewith95%CIisdepictedasanopendiamond.CompetinginterestTheauthorsdeclarethattheyhavenocompetinginterest.NN,MNandRMvDcontributedtotheconceptionanddesignofthestudy.NNandMNconductedtheliteraturesearchanddataextraction.NNperformedthestatisticalanalyses.NNandMNdraftedthemanuscript.RMvDsupervisedthestudy.NN,RJDandRMvDcontributedtotheinterpretationofdataandcriticallyrevisedthemanuscriptforimportantintellectualcontent.Allauthorsgavefinalapproval.NNandRMvDaretheguarantorsofthisworkand,assuch,hadfullaccesstoallthedatainthestudyandtakeresponsibilityfortheintegrityofthedataandtheaccuracyofthedataAcknowledgementsThisworkwassupportedbySawSweeHockSchoolofPublicHealth,NationalUniversityofSingapore.SawSweeHockSchoolofPublicHealth,NationalUniversityofSingapore.AuthordetailsSawSweeHockSchoolofPublicHealth,NationalUniversityofSingaporeandNationalUniversityHealthSystem,Singapore,Singapore.DepartmentofMedicine,YongLooLinSchoolofMedicine,NationalUniversityofSingaporeandNationalUniversityHealthSystem,Singapore,Singapore.DepartmentofClinicalEpidemiology&Biostatistics,McMasterUniversity,Hamilton,ON,ClinicalNutritionandRiskFactorModificationCenter,St.MichaelHospital,Toronto,ON,Canada.DepartmentofNutrition,HarvardSchoolofPublicHealth,HarvardUniversity,Boston,MA,USA.Received:23July2013Accepted:16January2014Published:18January2014InternationalDiabetesFederation.5thedition.Brussels,Belgium:IDFDiabetesAtlas;2011.2.KirtikarKR,BasuBD:Indianmedicinalplants.3rdedition.Allahabad,India:LalitMohanPrakashan;2000.3.SaxenaA,VikramNK:RoleofselectedIndianplantsinmanagementoftype2diabetes:areview.JAlternComplementMed4.WangE,Wylie-RosettJ:ReviewofselectedChineseherbalmedicinesinthetreatmentoftype2diabetes.DiabetesEduc5.Smith-SpanglerCM,BhattacharyaJ,Goldhaber-FiebertJD:Diabetes,itstreatment,andcatastrophicmedicalspendingin35developingDiabetesCare6.NeerajaA,RajyalakshmiP:Hypoglycemiceffectofprocessedfenugreekseedsinhumans.JFoodSciTechnol7.RaghuramTC,SharmaRD,SivakumarB,SahayBK:Effectoffenugreekseedsonintravenousglucosedispos

11 itioninNon-insulin-dependentPhytotherRes
itioninNon-insulin-dependentPhytotherRes8.HannanJM,AliL,RokeyaB,KhalequeJ,AkhterM,FlattPR,Abdel-WahabSolubledietaryfibrefractionofTrigonellafoenum-graecum(fenugreek)seedimprovesglucosehomeostasisinanimalmodelsoftype1andtype2diabetesbydelayingcarbohydratedigestionandabsorption,andenhancinginsulinaction.BrJNutr9.LuF,ShenL,QinY,GaoL,LiH,DaiY:ClinicalobservationonTrigonellaFoenum-graecumL.totalSaponinsincombinationwithsulfonylureasinthetreatmentoftype2diabetesmellitus.ChinJIntegrMed10.SauvaireY,BaissacY,LeconteO,PetitP,RibesG:Steroidsaponinsfromfenugreekandsomeoftheirbiologicalproperties.AdvExpMedBiol11.MoorthyR,PrabhuKM,MurthyPS:Anti-hyperglycemiccompound(GII)fromfenugreek(Trigonellafoenum-graecumLinn)seeds,itspurificationandeffectindiabetesmellitus.IndianJExpBiol12.UemuraT,HiraiS,MizoguchiN,GotoT,LeeJY,TaketaniK,NakanoY,ShonoJ,HoshinoS,TsugeN,etalDiosgeninpresentinfenugreekimprovesglucosemetabolismbypromotingadipocytedifferentiationandinhibitinginflammationinadiposetissues.MolNutrFoodRes13.SinghAB,TamarkarAK,NarenderT,SrivastavaAK:effectofanunusualaminoacid(4-hydroxyisoleucine)inC57BL/KsJ-db/dbmice.NatProdRes14.SauvaireY,PetitP,BrocaC,ManteghettiM,BaissacY,Fernandez-AlvarezJ,GrossR,RoyeM,LeconteA,GomisR,RibesG:4-Hydroxyisoleucine:anovelaminoacidpotentiatorofinsulinsecretion.15.VijayakumarMV,SinghS,ChhipaRR,BhatMK:Thehypoglycaemicactivityoffenugreekseedextractismediatedthroughthestimulationofaninsulinsignallingpathway.BrJPharmacol16.BrocaC,ManteghettiM,GrossR,BaissacY,JacobM,PetitP,SauvaireY,RibesG:4-Hydroxyisoleucine:effectsofsyntheticandnaturalanaloguesoninsulinsecretion.EurJPharmacol17.AlamdariKA,ChoobinehS,JadidiJP:AntidiabeticeffectsofexerciseandfenugreeksupplementationinmaleswithNIDDM.MedicinaDelloSport18.SharmaRD:Effectoffenugreekseedsandleavesonblood-glucoseandserum-insulinresponsesinhuman-subjects.NutrRes19.KassaianN,AzadbakhtL,ForghaniB,AminiM:Effectoffenugreekseedsonbloodglucoseandlipidprofilesintype2diabeticpatients.IntJVitamNutrRes20.GopalpuraPB,JayanthiC,DubeyS:EffectofTrigonellafoenum-graecumseedsontheglycemicindexoffood:aclinicalevaluation.IntJDiabDev21.MadarZ,AbelR,SamishS,AradJ:Glucose-loweringeffectoffenugreekinnon-insulindependentdiabetics.EurJClinNutr22.SharmaRD,RaghuramTC:HypoglycemiceffectoffenugreekseedsinNon-insulin-dependentdiabeticsubjects.NutrRes23.GuptaA,GuptaR,LalB:EffectofTrigonellafoenum-graecum(fenugreek)seedsonglycaemiccontrolandinsulinresistanceintype2diabetesmellitus:adoubleblindplacebocontrolledstudy.JAssocPhysiciansIndiaNeelakantanetal.NutritionJournalPage10of11http://www.nutritionj.com/content/13/1/7 24.MathernJR,RaatzSK,ThomasW,SlavinJL:Effectoffenugreekfiberonsatiety,bloodglucoseandinsulinresponseandenergyintakeinobesePhytotherRes25.ChevassusH,MolinierN,CostaF,GaltierF,RenardE,PetitP:AfenugreekseedextractselectivelyreducesspontaneousfatconsumptioninhealthyEurJClinPharmacol26.SuksomboonN,PoolsupN,BoonkaewS,SuthisisangCC:Meta-analysisoftheeffectofherbalsupplementonglycemiccontrolintype2diabetes.JEthnopharmacol27.YehGY,EisenbergDM,KaptchukTJ,PhillipsRS:Systematicreviewofherbsanddietarysupplementsforglycemiccontrolindiabetes.DiabetesCare28.XieW,ZhaoY,ZhangY:Traditionalchinesemedicinesintreatmentofpatientswithtype2diabetesmellitus.EvidBasedComplementAlternat29.UlbrichtC,BaschE,BurkeD,CheungL,ErnstE,GieseN,FoppaI,HammernessP,HashmiS,KuoG,etal

12 Fenugreek(Trigonellafoenum-graecumL.Legu
Fenugreek(Trigonellafoenum-graecumL.Leguminosae):anevidence-basedsystematicreviewbythenaturalstandardresearchcollaboration.JHerbPharmacother30.MoherD,LiberatiA,TetzlaffJ,AltmanDG:Preferredreportingitemsforsystematicreviewsandmeta-analyses:thePRISMAstatement.AnnIntern269.W264.31.GagnierJJ,BoonH,RochonP,MoherD,BarnesJ,BombardierC:randomized,controlledtrialsofherbalinterventions:anelaboratedCONSORTstatement.AnnInternMedTheInternationalSystemofUnits.8thedition.Paris:BIPM;2006.33.HigginsJPT,GreenS(Eds):Cochranehandbookforsystematicreviewsofinterventionsversion5.1.0,TheCochraneCollaboration.2011.34.ElbourneDR,AltmanDG,HigginsJP,CurtinF,WorthingtonHV,VailA:Meta-analysesinvolvingcross-overtrials:methodologicalissues.IntJ35.FollmannD,ElliottP,SuhI,CutlerJ:Varianceimputationforoverviewsofclinicaltrialswithcontinuousresponse.JClinEpidemiol36.HedgesLV,OlkinI:Statisticalmethodsformeta-analysis.London:AcademicPress,Inc.;1985.37.AbramsKR,GilliesCL,LambertPC:Meta-analysisofheterogeneouslyreportedtrialsassessingchangefrombaseline.StatMed38.CurtinF,AltmanDG,ElbourneD:Meta-analysiscombiningparallelandcross-overclinicaltrialsI:Continuousoutcomes.StatMed39.HigginsJP,ThompsonSG:Quantifyingheterogeneityinameta-analysis.StatMed40.EggerM,DaveySmithG,SchneiderM,MinderC:Biasinmeta-analysisdetectedbyasimple,graphicaltest.41.EggerM,SmithGD,AltmanDG:Systematicreviewsinhealthcare:meta-analysisincontext.2ndedition.London:BMJbooks;2001.42.BordiaA,VermaSK,SrivastavaKC:Effectofginger(ZingiberofficinaleRosc.)andfenugreek(TrigonellafoenumgraecumL.)onbloodlipids,bloodsugarandplateletaggregationinpatientswithcoronaryarteryProstaglandinsLeukotEssentFattyAcids43.ChevassusH,GaillardJB,FarretA,CostaF,GabillaudI,MasE,DupuyAM,MichelF,CantieC,RenardE,etalAfenugreekseedextractselectivelyreducesspontaneousfatintakeinoverweightsubjects.EurJClin44.SharmaRD,RaghuramTC,RaoNS:EffectoffenugreekseedsonbloodglucoseandserumlipidsintypeIdiabetes.EurJClinNutr45.FaqihAM,Al-NawaisehFY:Theimmediateglycemicresponsetofourherbalteasinhealthyadults.JMedJ46.BawadiHA,MaghayadahSN,TayyemRF,TayyemRF:Thepostprandialhypoglycemicactivityoffenugreekseedandseedsextractintype2diabetics:apilotstudy.PharmacognMag47.ZhouJ,ChanL,ZhouS:Trigonelline:aplantalkaloidwiththerapeuticpotentialfordiabetesandcentralnervoussystemdisease.CurrMed48.vanDijkAE,OlthofMR,MeeuseJC,SeebusE,HeineRJ,vanDamRM:effectsofdecaffeinatedcoffeeandthemajorcoffeecomponentschlorogenicacidandtrigonellineonglucosetolerance.DiabetesCare49.WingRR:Long-termeffectsofalifestyleinterventiononweightandcardiovascularriskfactorsinindividualswithtype2diabetesmellitus:four-yearresultsoftheLookAHEADtrial.ArchInternMed50.DeaconCF,MannucciE,AhrenB:Glycaemicefficacyofglucagon-likepeptide-1receptoragonistsanddipeptidylpeptidase-4inhibitorsasadd-ontherapytometformininsubjectswithtype2diabetes-areviewandmetaanalysis.DiabObesMetab51.PrasannaM:Hypolipidemiceffectoffenugreek:aclinicalstudy.IndianJdoi:10.1186/1475-2891-13-7Citethisarticleas:etal.Effectoffenugreek(Trigonellafoenum-graecumL.)intakeonglycemia:ameta-analysisofclinicaltrials.NutritionJournal2014:7. and take full advantage of: € Thorough peer review No space constraints or color “gure charges Research which is freely available for redistribution www.biomedcentral.com/submit Neelakantanetal.NutritionJournalPage11of11http://www.nutritionj.com/con