Several, possibly associated, dimensions of smoking behavior may be distinguished, e.g., smoking initiation, number of cigarettes smoked per day and nicotine dependence (Mayhew et al., 2000) and each
HeritabilityofSmokingInitiationandNicotineDependence401
TableI.PredictedProbabilitiesforaTwinPairundertheSingleLiabilityDimension(SLD),theIndependentLiabilityDimension(ILD)and
theCombinedModel(CM)
Twin2
Twin1 FTND 6
ModelSLDILDCMSLDILDCMSLDILDCMSLDILDCM
FTND 6y11y11x11y11x11y21y11x21y11x21y31y11x31y11x31y41y21x.1y11x41
FTND3–5y12y11x12y11x12y22y11x22y11x22y32y11x32y11x32y42y21x.2y11x42
FTND0–2y13y11x13y11x13y23y11x23y11x23y33y11x33y11x33y43y21x.3y11x43
Non-smoker
y14y12x1.
y12x1.+y12x14y24y12x2.
y12x2.+y11x24y34y12x3.
y12x3.+y11x34
y44y22
y22+y11x44+y21x.4+y12x4.
FTND3–5
FTND0–2
Non-smoker
UndertheSLDmodel,yjk=theprobabilitythatatwinpairfallsinthej,k-thcategoryofsmokingbehavior.Forexample,y11istheprobabilitythatbothtwin1andtwin2fallinthe rstcategory(FTND 6).UndertheCMandILDmodel,yjk=theprobabilitythatatwinpairfallsinthej,k-thcategoryoftheinitiationdimension;xjk=theprobabilitythatatwinpairfallsinthej,k-thcategoryofthenicotinedependencedimension;xj=theprobabilitythatthe rsttwinfallsinthej-thcategoryofthenicotinedependencedimension;x.k=theprobabilitythatthesecondtwinfallsinthek-thcategoryofthenicotinedependencedimension.
UndertheSLDmodelonetwincorrelationforeachzygositygroup(5)andthreethresholdsformalesandthreethresholdsforfemaleswereestimated,givingintotal11parameterstobeesti-mated.UndertheILDmodel,separatetwincorre-lationsfortheinitiationanddependencedimensionswereestimatedforeachzygositygroup.Fortheini-tiationdimensiononethresholdformalesandoneforfemaleswasestimated.Therewasno‘‘non-smoker’’categoryforthedependencedimension,leavingtwothresholdsformalesandtwoforfemalestobeesti-mated.Thismeansthat16parameterswereestimatedundertheILDmodel.UsingtheCMmodel,thesameparameterswereestimatedasintheILDmodel,ex-ceptforthethresholds:inthedependencedimension,threethresholdsformalesandthreeforfemaleswereestimatedbecausenon-smokerswerealsoincludedinthedependencedimension.SofortheCMmodel,18parameterswereestimated.
ThepredictedprobabilitiesforatwinpairunderthethreemodelsarepresentedinTableI.UndertheSLDmodel,y11denotestheprobabilitythatbothtwinsarehighlynicotine-dependent,y12denotestheprobabilitythatthe rsttwinishighlydependentandthesecondtwinismoderatelydependentonnicotine,andsoon.UndertheCMandILDmodel,y11de-notestheprobabilitythattwinsbothinitiatedsmoking,y22denotestheprobabilitythatbothtwins
didnotinitiatesmoking,andy12andy21denotetheprobabilitiesthattwinsarediscordantforsmokinginitiation.Theconditionalprobabilitiesthatbothtwinsarehighlydependentonnicotine,the rsttwinishighlydependentandthesecondtwinismoderatelydependentisrepresentedbyx11,x12etc.UndertheCMmodel,therearetworoutesto‘‘non-smoker’’.Forexample,y11x24givestheprobabilitythatbothtwinsaresmokersontheinitiationdimension(y11)andthe rsttwinismoderatelydependentonnico-tine,whilethesecondtwinisanon-smokeronthedependencedimension,y12x1.givestheprobabilitythatthe rsttwinisasmokerontheinitiationdimensionwhilethesecondtwinisanon-smokerontheinitiationdimension(y12),andthe rsttwinishighlydependentonnicotine(x1.).
Thethreemodelswere ttedtothedata,esti-matingseparatepolychoriccorrelationsforeachzygositygroup.Thegoodness-of- tofnestedmodelswasassessedwiththelikelihood-ratiostatistic,thisstatisticwhichisdistributedaschi-square.GeneticModels
Thethreemodelswere ttedtothedata,andforthemodelthatgavethebestdescriptionofthedata,thetwincorrelationsinliabilitywereexpressedasafunctionofgeneticandenvironmentalfactors(Neale
doi10.1017S0033291705005374 Printed in the United Kingdom Heritability of ... regular exercise, alcohol consumption and smoking behaviour (Boomsma et al....
doi10.1017S0033291705005374 Printed in the United Kingdom Heritability of ... regular exercise, alcohol consumption and smoking behaviour (Boomsma et al....
热门文档