Investigating a Genetically Informed Enhanced Reinforcement Model of Alcohol Involvement: A Prospective Analysis

Enhanced reinforcement has figured prominentl y in et iolog ic models of alcohol invo lvement and may be of particular relevance in adolescence and emerging adulthood. We prospectively exam ined an enhanced reinforcement model in a sample of emerging adults and augmented the model with a promising cand idate gene, OPRMl, which has demonstrated associations with both alcohol outcomes and psychosocial factors compri sing the enhanced reinfor cement model. We examined whether a putatively functional polymorphism in the OPRM I gene was associated with heavy episodic drinking (HED) and negative consequences from drinkin g across four year s via a number of intervening psychosocial factors , including behavioral undercontrol , subject ive responses to alcohol , and cognitive factors (a lcohol


LIST OFT ABLES
days) and 36.7 percent of young adu lts in this age bracket being "b inge" , or heavy episodic , drinkers (5 or more drinks in a row in the last 2 weeks) (Johnston et al.,20 I 0). Young adults enrolled in co llege full time are more likely than thei r peers not enro lled full time to use alcohol (Johnston et al.,20 IO; Substance Abuse and Mental Health Services Admini stration , 2009). According to Knight et al. college students report high rates of alcohol abuse and dependence (2002) . Specifically , they found that 3 1 percent of college students endorsed criteria for alcohol abuse , 6 percent endorsed criteria for alcohol dependence and students who were frequent heavy episodic drinkers had 13 times greater odds for abuse and 19 times greater odd s for dependence (Knight et al., 2002) .
Matriculation into college is typified by increases in alcohol consumption and associated negative consequences (Sher & Rutledge, 2007). Of the college students who do drink alcohol, half report experiencing serious negative consequences including, but not limited to, doing something they later regretted, getting in trouble with the police, having unwanted sex and physically injuring themselve s and others (American College Health Association , 2009). In the U.S., it is estimated that approximately 1,825 annua l deaths and more than 796,000 violent and sexua l assaults are linked to alcohol use in the college student population (Hingson, Zha, & Weitzman, 2009) .
These epidemiologic data highlight the alarming prevalence of alcoho l use and misuse among "emerging adults" between 18 and 25 years of age, particularly among college students. They also underscore the importance of better understanding the complex etiologic pathways of alcohol use and misuse among emerging adults to assist in further developing and refining interventions aimed at reducing the acute and chronic effects of alcohol misuse in this population. To this end, the current study will prospectively investigate an enhanced reinforcement etiologic sub-model of alcohol involvement proposed by Sher ( 1991 ). We will also investigate further whether variation in a candidate gene (OPRM I), associated with alcohol reinforcement , is associated with the psychosoc ial factors comprising the enhanced reinforcement model, as we ll as alcohol use and problems. Next, models of vulnerability will be reviewed along with research examining assoc iations between alcohol outcomes and the variables comprising the etiologic model proposed in this study, the enhanced reinforcement sub-model.

Models of Vulnerability
The notion that there is no single "type" of alcoholi sm or simple et iologic pathway for the development of alco hol misuse is wide ly accepted (Leonard & Blane , 1999;Sher , 199 1). Year s of research on alcohol vulnerab ility has inves tigated the contribution s of biologica l, psyc ho log ica l and socia l influen ces on a lco ho l use. This research has deve loped the co nsensus among contempor ary etiologic mode ls that alcoho l misuse and alcohol use disorders (AU Ds) are caused and exacerbated by an arra y of biopsyc hosoc ial factors whose com bined assoc iation s and influences vary across the life span (Leo nard & Blane , 1999;Sher , 199 1;Zucker, 1987;Zucker , 2006).
Vulnerability studies have sought to identify mechani sms of risk for alcohol misuse and given the long-recog nized heightened risk of childr en of a lcoho lics (COAs) , this group has often been the focus of these studie s. Behavior al genetics studie s have consistentl y supported a role for genetic factor s in fa milial tran smiss ion of alcoholism (Cloninger, Bohman , & Sigvardsson , 1981;Cotton , 1979;Goodwin , 1988;Kaprio , Koshenvuo , & Langinvaini o, 1987;Sher, 1991 ). The evidence that genetic factors are associated with responses to drinkin g and the deve lopment of alcohol mi suse is beyond dispute (Ball & Murray , 1994;McG ue, 1994;1999;Merikan gas, 1990). Howeve r, the most influential genes affect ing alco hol outcomes have not been we ll e luc idated and research linking candidate genes, genes implicated in co ntributin g to a particular phenotype (e.g. disease) (National Institut e on Alcohol Abu se and Alcoholi sm, 2003) , w ith etio log ica lly releva nt psyc hosoc ia l factors is one approac h to substantiating the se genes of interest (Dick,Latt endre sse,& Riley,20 I I ).
Accor din gly, substa ntia l gaps remain in understandin g the spec ific genetic comp onents influ entia l to a lcoho l misuse , but the ca ndid ate gene approach is promi sing and continues to elucidate new genes of interest relevant to alcohol studies.
Genetics alone do not acco unt for the direct ex pressio n of behav ior, howeve r.
According to McGue " ther e are numerou s intervening ste ps between primary gene produ ct (prote in synthes is) and obse rvable behavior. Genetic influen ces on alcoholism risk might reflect mechani sms ranging from ethanol sensitivity to heritable perso na lity characteristics " ( 1999, p. 3 73).
Prospec tiv e examination of etiologic pathwa ys through which susce ptibi lity (e.g., family histo ry and spec ific genoty pe) factors affect intervenin g psyc hosocia l variables and a lcoho l outcomes is criticall y imp ortant for both furthering knowledge of the etiolo gy of alcohol use and mi suse and for informin g preventive interventi ons.
Sher (199 1) introduced a comprehensive mode l of simple and more com plex pathways throu gh which familial risk may be transm itted. In thi s ove rarchin g mode l, fami ly history is medi ated by seve ral broad cate go ries includin g per sonalit y, cog nitive proce sses, bio log ical influen ces and familial and other environmental conside rations ( 1991 ) . Sher ' s ov erarchin g model is also broken dow n into less complex inter-rel ated sub-model s labe led "e nhan ced reinforcement ", "deviance-pronene ss" , and " negative affect " . As noted , the current resea rch w ill focus on the enhanced reinforcement submod el and w ill incorporate exa min at ion of a candid ate gene, OPRM I, which has been linked to reinforcing effects to alcohol (Ray & Hutc hiso n, 2004)  Integrating Sher' s ( 199 1) model with research on the candidate gene OPRM 1, Figure I proposes a genetically-informed enhanced reinforcement sub-model of alcohol involvement among an emerging adult population. An emerging adult sample is particularly well suited to such an examination given the high levels of alcohol use and misuse in this subpopulation and due to evidence that positive reinforcement motives may be particularly important in adolescent and emerging adult populations (Kuntsche, Knibbe, Gmel, & Engles, 2006;Read, Kahler, Wood, Maddock, & Palfai, 2003). Next, we briefly describe the overall sub-model. As can be seen in Figure I , family history of alco holism and genet ic var iation in the OPRMl ge ne are exogenous facto rs w ith assoc iations on alcoho l invo lvemen t (alcoho l consumpt ion and negat ive co nsequences) via a numb er of interven ing psychosocial factors. Specifically , consiste nt wi th Sher 's (199 1) enhancement reinforcem ent model , fam ily histo ry effect s on alco ho l outco mes are purported ly mediated by "behavioral undercontr o l" per sonality traits and subject ive effect s to alco hol (ethano l sensitiv ity). In turn , associa tions between per sonality and subjective effects and alco ho l outcome s are mediated by a lco hol expectan cies that have been purp orted to be final com mon path ways through wh ich more distal biop syc hosoc ial factors influ ence alcohol use and misuse (Coope r, Fro ne, Russe ll, & Mudar , 1995;Go ldman, 1994 ). While not a pa rt of Sher's ( 199 1) formu lat ion, the inclus ion of 6 OPRM 1 in the model depicted in Figure 1 bui Ids on a n~t but growing body of resea-ch on caidi date genes i mportait to understanding a cohol use aid misuse. This model hypothesizes that genet ic veri ati on in the OPRM 1 gene a ong with f a-ni I y hi story wi 11 be related to a cohol outcomes i ndi rectl y vi a behavi ora undercontrol perronaity trats, subjective effects to drinking, aid acohol expectaides. A review of the existing resea-ch in support of the hypothesized asood ai ons between constructs incorporated within the enhaiced rei nforcernerit model aid their eti ol ogi c evi derice wi 11 be di scuS9:d in greater deta I be ow.
According to Adi noff the compulsive drive towad drug use is compl emmted by deficits in impulse control aid decision ma<ing (2004) aid Derringer et a. s,owed that dopaminegmesaeasrodcted with sensation-seeking (Derringer et a ., 2010) which substaiti ctes the hypothes zed pcthway from OPRM 1 to behavi ora under control in this study's model. Adinoff also states that "dopaminergic activation occurs in the pre9:nce of unexpected aid novel stimuli (either pl easura::>I e or aversive) and appears to determine the motivational state of wanting or expectation " (2004 , p 305), which supports OPRM I ' s pathway to motivation.
In OPRM 1 the G alele isfunctionaly differmt. TheG alelevaiait binds the amino ocid bet&mdorphin three times more strongly thai the A vaiait (Bond et a., 1998). The bet&mdorphin isai mdogmousmorphine in the body, or ai opioid 9 peptide neurotraiS'Tli tte-found in the centra aid peri phe-a nervous syste-ns. One of its f undi ons is to numb the body aid modulate the feeling of pain afte-experi end ng tra.ima. lndividuaswith the G alele may display behaviora diffe-enc:e beca.Jre of a hei ghtenoo rensi ti vi ty to µ-receptors. This meais tha i ndi vi dua s with the G a I el e may be more rensitive to the SJbjective effects of acohol SJch as euphoria, as was ob:x:rvoo by Ray aid Hutchioon (2004).Accordingly, individuaswith theG aleleare considered to be " at ri sk" for highe-rensitivity to ethaiol in SJpport of the hypothesi zoo pathway from OPRM 1 to ethaiol rensi ti vi ty in our model.

Pers:nality.
In more conte-nporar y eti ol ogi c models, peroona i ty is vi e.voo as one of mai y foctors infl uenci ng a cohol miSJre (She-et a., 1999). Whil e it is w idely agreoo that there is no such thing as an " alcoholic personality" , decades of research attest to the etiologic relevaice of perronaity trats to acohol ure aid mis.ire (Leonerd & Blaie, 1999;Schucki t, KI ei n, Twi tdiel I , & Smith, 1994). Evidence from both cross-recti ona aid prospective r~d7 for the influence of pa-rona i ty on a coho! ure aid mis.ire is considerable , particularly with respect to traits related to " behavi oral undercontrol " ( deocri bed below).
The behavi ora under control factor in the proposai eihancoo rei nforcemeit model will becomprisai of individua facets of impulsivity, sensciion 9:leking and psydloti d STI . The i nfl ueice of behavi ora undercontrol peroona i ty tra ts on a cohol mi ruoo has beal shown to be mooi atoo by expect and es ( Schucki t & Smith, 2006a; Sher, Waitzer , Wood, & Breit , 1991), thereforetheproposaierihancoo reinforcemerit model pos its th at behav iora l undercon tro l' s influ ence on alcohol outco mes w ill be indirect through a cohol expectand es as deta I oo in a rub5e1uerit recti on.
A I ow I e.tel of response to a rohol is de:ui bed by either a I ow i nterisi ty of reaction to ethaiol a a gi veri blood a rohol ronca,trai on (BA C) vi a a, administered  According to Sher, individuals with ai inherited heightened 9:Jlsitivity to alcohol tend to have i ncreared expectations of reinforcement from drinking. Acc:ordi ngl y, the enhanced reinforcement model in our study is congruent with Sher's hypothesis by positioning ethaiol 9:Jlsitivity beween family history aid cognitive factors, ruch as alcohol expectaici es, which are brief I y revi evved next.

Alcohol Expectancies
Alcohol expectaici es, defined as beliefs roout the c:ogni ti ve, behavioral , aid afroive effrosof drinking alcohol, have long been ascribed a significarit role in the etiology of alcohol use and mi ruse (Goldmai, Del Boca, & Darkes, 1999;Leigh, 1989 Wood, 1994). While a number of fcdo rs relaro to postive c11d nega ive alcohol expectc11des have been identifiro in this resea-ch , of grootest rel eva,ce for i nvesti ga i on in the context of enha,coo reinforcement are positive alcohol expectancies such as " activity enhancement" (Kushner et al., 1994). Expecta,d esfor cd ivit y enha,cement indude statements like, "drinking makes many activities more enj oyable", and relate to postive hroonic expe-i ences . Enha,cement expect a,d es have recei vro cros.5 ~ti onal a,d pro~i ve rupport in their inf I uence on alcohol use anong emerging ooul ts ( Sher et al., 1996).
The col I ecti on of cross-soct i onal a,d pro~i ve resea-ch rupports the eti ol ogi c rel evaice of a cohol expectaid es aid the ind usi on of octi vi ty aihaicement expectaid es in I a-ger aihaiced rei nforcemait models of a cohol ure aid mi sure.
Accordingly, the purpore of the prerent study is to prospectively test ai infl uentia eti ol ogi c sub-model propos ng that enhaiced ra nforcanent is i mportait in undffstaiding the development of acohol ure aid mirure. As noted, given the preva ence of a cohol ure aid mi rure in emffgi ng ooul thood, aid previous rerea-ch ruggesti ng that pos ti ve ra nforcanent motivations a-e pa-ti cul a-I y salient in oool es::ence aid emffgi ng ooul thood, the prerent sanpl e is pa-ti cul a-I y wel I-sui ted to oodress this question. Moreover , the extension of Sher ' s ( 1991)  in OPRM 1, behaviora undffcontrol perronaity trats, rubjective effects to acohol , a cohol expectaid es on multiple a coho! outcomes.
Cons stent with the Ii terature revi e.ved, re.;era hypothe9:l5 ere forwcrded. It is hypotheszed that the relationship between OPRM 1 aid fanily history of acohol ism aid a cohol-rel ated outcomes (negative conse:iuences aid heavy drinking) wi 11 be mediated by "behavioral und ercontrol " per sonality traits and subjective effects to acohol (ethaiol S:nstivity). In turn, asrociciions between peroonaity aid subjective af ects aid a cohol outcomes wi 11 be mediated by a cohol expectaid es. It is a oo hypothes zed thci OPRM 1 wi 11 be pos ti vel y asroci ated with the trcj ectory of heavy drinking aid a cohol problems, such that i ndi vi dua s with the G a I el e wi 11 exhibit higher I eves of a cohol use aid problems. These aia yfxS wi 11 a::ld to the genera scientific understaidi ng of a cohol use trcj ectori es i n ai emergi ng a::lul t/ col I ege student populciion. To our knowledge, the proposed resecrch is both the first to comprehens vel y exani ne ai enhaicoo re nf orcernent sub-model pro~i vel y in a populciion with enhaicoo risk for acohol misuse aid to incorporateafunctionaly rel evait aid ernpi ri ca I y supported ca,di date gene within the sub-model. Addi ti ona I y, understaidi ng how these sped f i c factors influence a cohol use may assist in the development of interventions a med at redud ng the ocute aid chronic a'f ects of a cohol misuse in this population perhaps through tcrgeti ng factors ind uded in this model.

Sample
Aspai of a la-ge-study {Wood et a, 2010) in two ruccessvecohorts, paiidpants (N = 1,014 pa-ent-student dyoos) were recruitoo during the rummer prior to matri a.JI ati on ct a mi d-si zoo public northeastern university. The ta-get population was entering f ul I-ti me or part-ti me students ages 17-21.
At Wave 1, a s ngl e i tern ma39...lre recommended by Cretvs aid  was usa:l to as'2RSS fanily history of acoholis-n. Studentswa-e asked cbout both parents, "Do you think your (father/mother) is/was an alcoholic"? Crews and Sher (Cretvs& Sher, 1992) demonstrated that thisgloba rating of both pcia-na aid mata-na acoholis-n had excellent test-retest stcbility (8S'2R£,SOO twice ova-a 10-day to 20 3-week test-retest interva), high inter-sibling agreement , aid mcx:lerately high agreement w ith the co rrespondin g se lf rating from the parent 's report of problems w ith drinking . Crews and Sher's results also showed that thi s g lobal item had acceptable sensitivity, specificity, kappa, aid Y vaues. Fanily history status was coded asa categorical vcri role ( 0, 1, or 2) for i ndi vi dua s with neither pcrent identified as having aroholism (0), either father or mother having aroholism (1) or both father aid mother having a rohol ism ( 2). In the sa11pl e 79 poopl e had a f ani I y hi story of aroholism (15.56 percent of tota sa11ple, with 14.87 percent= one pcrent aid 0.59 percent= both pcrents) .

OPRM1 .
Oragene DNA Self-Col I ection kit procedures were usad to genotype 524 DNA-providing pcrtidpantsat Wave 4. Whenever possible , DNA was obtained on site by project stcif using written protorol s to ensure sa11pl e vi roi Ii ty aid strict a:tention to ronfidentiaity . When this was not possible, DNA was obtained by selfrollection aid returned using procedures for protecting pcrtidpait ronfidentiaity .
Single nud ooti de polymorphisms ( SN Ps) were genotyped using ai 111 umi na BeEdX press 384-snp paiel fol I owing estrol i shed protorol s. Genetic risk is defined as having at least one ropy of the OPRM 1 putative risk alele , the G alele (Miraida et a . , 201 0; Ray & H utchi oon, 2004). Data were dummy coded to i ndicate which individu als were " at risk " ( I = risk by pre sence of the G a llele). Of tho se w ho provided DNA sa11ples 140 had at least one ropy of the G alele (27%), a proportion rompcrrole to other sa11ples (Kidd, 2011;Oroszi, & Goldmai, 2004).

1993)
, with the sum score of 7 items measu ring impulsivit y (a= .67) and I I items measuring sensation seeking (a = . 75). There were originally 8 items of the i mpul si vi ty &:a e, but one had to be del eta:! due to a typogr~hi ca a-ror on the survey given to partidpa,ts . The items for both &:aes had a true-fase response format for whether the questi on "is true as applied to you" or " if false as applied to you" and includes questions such as, " I often do things on impulse" and " I like to do things just for the thrill of it". Psychoticism (tough mindedness) was measured at Wave 3 as part of the Revised Eysa,ck Pa-ronaity Questionnaire (EFQ-R) (Eysa,ck, 1988), a 57 i tern measure. The sum srore from a psychoti d sn sub&:a e of 17 i terns tci<en from the EPQ-R was used to measure psychotici sm in the behavioral undercontrol factor ( a = .54).

Ethanol Sensitivity (&ibjective Effects).
Etha,ol sa,sitivity was colla:::too a Wwe 3 via self report. Four items of the Self-Rcting of the Effects of Alc ohol (SRE) scale (a = .87) about the fir st five times oneeva-draik was used . Spedficaly, partidpaits were asked how many drinks it took for them to: " begin to feel different", " feel a bit dizzy, or begin to slur your speech", " begin stumbling or walking in an uncoordinated manner " , and " pass out, or fall asleep when you did not want to" wit h reference to the first five ti mes one eve-draik a cohol . Response options wa-e coda:! a::cordi ng to staidard drinks. One staidard drink was def i noo as one shot of Ii quor, 12 ounces of bee-, or one 4-ounce glass of wine, and " in a row" was defined as one occasion without aiy brea<sof ai hour or longa-(W<XXJ. et a. , 2010). The Self-Rating of the Effects of Alcohol scae has been shown to posses.s gocxJ. test-retest relicbil ity in studies conducted ova-multiple yea-s aid ind udi ng multiple fol I ow up as9:)SS1la,ts (Schuck it et a l., 1997). Labo ratory tests know n as the "alcohol challenge", in w hich pa-ti ci paits conrume a cohol inside the I cborat ory aid the result ing effects do~ y trocked, have been strongly correlated to SRE ~f report measJre (.82, p < .0001) . To be consistent w ith Sher's ( 199 1) sub model, lowe r SRE s:::ores cl'e i ndi cati ve of a hi gha-ethaiol 9:lllsi ti vi ty (Iowa-s:::ores cl'e ~ui va a,t to fe.va-drinks nea:led for ai effect which is inf a-red as a high 9:lllsitivity to acohol).

Activity Enhancement Alcohol Expectancies.
From a lcl'ger &:ate of acohol expectaicies, activity 01haicement acohol expectaici es wa-e ~ ct Wave 4 using a 9 i tern activity 01haicem01t rubsca e mea sure (Ku shner et al., 1994) (a = .87). Sample items includ e, "drinkin g makes many activiti es more enjoyable" and "Drin king ma kes sports events (like foo tba ll, basketball , car races) more enj oyable. " Individuals we re asked to rate their perso nal expectations of the effects of a cohol on a 5 point sea e ( 1 = Not ct a I, 2 = A little bit, 3 = Some.vhat , 4 = Quite a bit, aid 5 = A lot). As 93€11 in Figure2 the nine itemswa-e pcl'celed into four indicctors that loaded on the lata,t factor. Each pcl'cel contaned ai ava-age s:::ore .

Alcohol consequences.
Alcohol con9::quences Wffe assessa:! with a 17-item Vffsion of the Young Adult Alcohol Problems Screening Test (YAAPST) (Hurlbut & Shff , 1992). The Y AA PST Wcx, admi ni stffoo at ba921 i ne, 1 0 months 22 aid 46 months. The sea e assessa:! the past 3-month frequaicy of alcohol con9::quaices with response options raigi ng from 1 ( no, not in the past 3 months) to 5 ( 10 or more ti mes in the past 3 months). Consequences include , "have you gotte n into physical fights when drinkin g?" and "have you ever been pressured or forced to hav e sex w ith someo ne because you were too drunk to prevent it?", for example. The responses of each item Wffe r~e:i as an estimate of the number of occurrences ( e.g., respon se option "2" (1-3 time s in past 3 months) recoded to 1.5 and response option "5 " (10 or more times)
First, missing data aid deocri pti ve statistics ere revi e.Noo, tha, the results of confirmatory foctor aialysis (CFA) for the foctor structure of the lciait verial:>les in this model eredeocriba:l. CFA resultserefollowoo by latait growth curve modeling.
I ni ti al models exani noo the model depi ctoo in Figure 1, which assumes f ul I mooi ati on of f ani I y hi story aid OPRM 1. In addition to evaluation of i ni ti al models based on overall mode l fit criteria (detailed under "Latent Growt h Curve Model s," tests of hypothesizoo indirect €ff eds wff"e computoo. Subse:iuent to theoo aia yres, the f ul I mooiation models wff"e compa-oo to more saturatoo models (e.g., with a:lditiona direct paths to a coho! outcomes) in nestoo model compa-i rons.

Missing Data
Anal y'i:£!3 for systematic attrition for the first three wave:, have been conducted for this data (Wood et al., 2010). To extend these aialy'i:£!3to the final wave, Wave4 rurvey completa-s (n = 627) wa-e compa-ed with non-completa-s (n = 387). Using x 2 aid t-tests for cciegorical aid continuous va-ici:>les, we ob9:!rved no significait baselinediffa-enCffion genda-, race, ethnicity, weekly drinking, heavy drinking, or alcohol prob I ems. Geno typed i ndi vi duals conrumed si gni f i caitl y f ewa-peal< drinks ci base! i ne (p = .048) but rurvey com pl eta-s did not di ffa-from non-compl eta-son peal< drinking. In rum, wefound va-y limited evidence of systematic attrition (Wood et. al., 20 I 0). Latent growth curve models testing the study' s subs tantive hypotheses were estimated using al I ava I ci:>I e data, aid al I pa-a-neta-s wa-e estimated using maxi mum Ii kel i hood estimation.

Descriptive Statistics
Uni Three obrervai ons in quest:i on two of the SRE oca e we-e i dentifi oo as extreme outl i e-s. These three obrervai ons we-e adj ustoo to be within 1.0 of the furthest: va ue aJOve the 75th pe-ca1ti I e that sti 11 I ay within the thresiol d of obrervati ons. The va-i cbl e was norma I y di stri butoo ate-the adjustment was made.
As seen in T role 1 men ave-aged 6. 00 drinks pa-week at b893l i ne. The numbeof drinks pa-week i ncrea:a:I at eoch ti me point with men reporting having 15.19 drinks pa-week a Wave 4. The ave-age drinks pa-week in women a oo i ncrea:a:t from b893l i ne to Wave 4 (5.35 at b893l i ne to 8.65 drinks pa-week a the Wave 4 ), though the dlaige was sma I er for women. Men reportoo sightly I ess thai one epi oode of heavy drinking pamonth at the b893l i ne, but the numbe-of epi rodes increa:a:I for men to ~proximately 2.7 epioodes pa-month at Wave 4. This increasing trend was a oo obrervoo for women who reportoo slightly I ess thai one heavy drinking epi oode a month at b893l i ne to ~proximately 1. 7 epi rodes a month at Wave 4. As sea, in T ct>I e 2, at e-JffY ti me point at I ea& ha f the entire sanpl e reported oongiquences of i ntoxi cation, ruch as experi end ng haigovffs, aid wa<i ng in the morning to find they hcd forgotten pai of the pre,;i ous e,;eni ng. At e-JffY ti me point ovff on&thi rd of the enti re sanpl e reported feeling s ck or throwing up aft ff drinking , aid saying things they I ciff regrffted . More thai ten percait of the entire sanp l e, ci e-JffY ti me point, reported getting into a sexua s tuati on they I atff regrffted. Al ro, the percait of men in this sanpl e who neglected to use birth oontrol or protection from STD's was 7.95% at baseline, a rate that increase d to 13.59% at Wave 4 . Women reported a high of 7.26% in neglecting to use birth oontrol or protection from STD's at Wave3. No te. Pe rce nt of samp le w ith prob lem at leas t once in the past 3 mon ths .
The CFI rEfl ects the degree to which the sanpl e veri aices aid coveri aices ere reproduced by the hypothesized model structure. CFI raiges from Oto 1.0, with higher vaues, prEferroly greater thai .90, reflecting better cpproximation of the data (Ullmai , & Bentler, 2003). SRMR isa residua-based index. Spa::ificaly it is the staiderdi zed difference between the observed carrel ati on aid the predicted correlation. Lower vaues, prEferroly below .08, indicate a good model fit (Hu, & Bentler, 1998). Factor loadings were aoo examined in addition to overal fit indices to as.<e3S the adequocy of model sped fi cation. Factor I oadi ng estimates ere given t-va ues aid si gni fi ca,ce is ca cul ated occordi ng to the t di stri buti on.
Fit stai sti cs for the measurement model of behavi ora under control could not be determined in confirmatory factor aia ysi s. This was a result of the factor structure being 'J ust-id entified", meaning that the numb er of param eters spec ified in the model equa ed the number of esti mrol e perameters in the veri aice-covari aice matrix resulting in O degrees of freedom (Kline, 2005). A just identified model yields a trivia ly perfect fit making overal model fit statistics uninteresting (Rigdon, 1997).
Howev er, signifi can t tests on the mode l's pathways from the latent variable to the indicators(or loadings) ca, becacu lated (Kline, 2005;Rigdon, 1997) . Coeffidentsof the pathways from the i ndi vi dua i terns to the I atent behavi ora undercontrol veri a:> I e raiged from .57 to .76 aid were al significa,t at p < .001. In addition, Ea"l ier sections here re.ri e.ved eoch measure of behavi ora under control ind udi ng i mpul si vi ty, 31 ~sati on reeking, aid psychoti d sm aid showoo the thooreti ca support for the93 items' inclusion in the personality factor proposed in this study.

Latent Gro.vth Curve Models
Latent growth curve (LGC) modeling is used to ~turegrowth in a construct ova-time using raidom coefficients thci reflect initial stcius (inta-cept) aid growth rcie (slope). Following Bollen aid Curra, (2006) a two step cpproach was used to exarni ne the tr~ ectori es of two di ff a-ent models, one model for heavy epi s:xli c drinking aid the otha-model for alcohol consequences.
The L GC models for heavy epi s:xli c drinking aid consequences in this study ind udoo data from al I four assessments (baseline, 10 month fol I ow up, 22 month fol I ow up aid 46 month fol I ow up). Growth curve aial yses wa-e conductoo using Mplus6 .1 (Muthe, & Muthe1, 2010) on al avalrole data using full-information maximum likelihood estimation for missng data (Arbuckle, 1996;Schafer & Graham , 2002).  Table 3. The f ul I y mooi atoo hypothesi zoo models ( cons stent with Figure 1) were testoo first for heavy epi SJCli c drinking. As ca, be9::al in Table 3, this model showoo good overal fit indices(CFI = 0.92; TLI = 0.9; RM SEA = 0.058; aid SRM R = .082). For this model the R2 va ues for the intercept, linea-slope aid qua:lraticfoctorswere .34 (p < .001), .013 (ns), aid .10 (ns) respective! y. Note. DF = degrees of freedom; CFI = compartive fit index; TLI = Tucker-Lewis index; RMSEA = root mean square e1TOr of approximation; SRMR = standardized root mean square residual ; As hypothesized, fanily history was positively related to perronality, however, in contrast to our hypotheESS f ani I y hi story was not related to subjective effects ( see Figure3). Alro contra-y to our hypotheESStherewasno assJciation of OPRM 1 with either perronality or subjective effects a,d subjective effects were not related to expecta,ci es a,d hoo no si gnifi ca,t indirect assJci ati ans ba:w001 f ani I y hi story a,d hesvy drinking. aid linea-slopefoctorswffe .30 (p < .001), aid .22 (p < .001) respectively. Given that the exogaious aid mooi ati ng f octors a-e i denti ca ocro~ the heavy epi s:Jdi c drinking aid con9:quence models, these asooci ati ons a-e depi ctoo in Figure 5 but not di ocusred.
As~ in Figure 5, the significa,ce of the pathways is quite simila-to the heavy epi s:Jdi c drinking model .  .28 (p < .001) respectively . Therefore this model was retaned as the final model.
Subjective effects were posi ti vel y assxi ated with the i ni ti al I evel of oonsaquences , but negative! y ass::>ciated with the chaige in problems i ndi cati ng that less 9311sit ivity to al oohol was ass:)ciated with less growth in problems over ti me.
There were no significait direct effects from fani ly history on the intercept or slope; however, family history' s influence on the intercep t and slope of problems was mediat ed through both behavioral undercontrol and expectancies (P = .024 , SE= .0 I 0, p < .05, for the intercept; p = .029, SE= .012, p < .05 for the slope). There was also mediation between fanily history aid the intercept through behavioral underoontrol alone (P = .034, SE= .0 15, p < .05). Mediation of family history and the intercept and slope through expect aid es al one was not si gni f i cait.

DISCUSSION
The m~ or purpore of the present study wa:, to exani ne a geneti ca I y-i nformoo model of enhancoo rei nforcanent in the prospective prooi cti on of heavy drinking and a cohol problems in emerging adulthood. Whi I e the speci fi oo models provi doo very good fit to thedcta, overal support for hypotheszoo mediating pa:hwayswa:, mixed.
Consistent with our predictions, we found that behaviora undercontrol /disinhibitoo perrona i ty tra ts and a cohol expectancies si gni fi cantl y mooi atoo rel cti ons between a f ani I y hi story of a cohol ism and both heavy drinking and a cohol problems. In contrast, we found no support for a hypothes zoo mooi ati ng role for behavi ora undercontrol perrona i ty tra ts and subject ive effects of a cohol in OPRM 1 -a cohol outcome relations. Moreover, a hypothes zoo indirect effect ba:ween f ani I y hi story and a cohol outcomes vi a subjective effects of a cohol wa:, a ro not obrervoo.
We investigctoo the robustness of the enhancoo reinforcanent model by estimating additiona paths from both exogenous (family history, OPRM 1) and intervening (behavi ora undercontrol, subjective effects) factors to a cohol outcomes.
Ba:a:I on significant increments in model fit, these more saturated models were Next we el cborcte on study findings a,d pl ace them in the oontext of the I a-ger theoreti ca model aid prior resea-ch.

Strengths and Limitatioos.
The lack of ethanol sensitivity' s involvement in the model as hypothesized might relate to thewctf this foctor was merouroo in our study. Ethanol 9:nsitivity was not measured consistent with its explanation in Sher' s original model. The SRE scale u~ ha-e purportool y meroures innate tolerance more than 9:nsi ti vi ty. Otha-meroures of 9:nsi ti vi ty mctf be more cppropri ate. For exa-npl e, in his model Sha-stcies that "decrea sed sensitivity on the descendin g limb of th e BAC can easily be accommodated to the model (to measure ethanol sensitivi ty)" (p. 135, 1991 ) . A measure that a:ptures bi phasi c respon~ to a cohol , such as the Bi phasi c A I cohol This study was a post hoc aia ysi s of data col I ectoo as pai of a, intervention study conducted with col I ege students aid was not desi gnoo to ~fi ca I y measure the enhaiced reinforcement model . The post hoc design may ra se issues of genera i zabi Ii ty, paii cul a-I y given that the sanpl e was c:ompri red of a, ethni ca I y homogenous popul cii on dra.vn from a single canpus at a public, northeastern university. As notoo, the use of the SRE to mExS.Jre ethaiol sensitivity aid the la::!< of a high-riS< far,ily desgn ere further limit ciions.
Despite the:e Ii mi tati ons, the current resea-ch replicates aid extends prior resea-ch in multiple ways, nota::>I y with respect to the a::>i Ii ty to conduct a fa rl y comprehensive tes of ai i nfl uenti a eti ol ogi c model with a I a-ge prospective Sclllpl e of ernergi ng adults.

Condusion
The inc lus ion of a candidate ge ne assoc iated with alcohol ' s reinfor cing effects wro not supportoo aid, ro notoo, support for the integration of the caidi date gene ~prooch into etiologic models hro beerl modes. Commenting on this issue, Dick et a.
contend that the i ncorporcii on of genetics into I ongi tudi na devel opmenta resea-ch hro greet potenti a for furthering current understaidi ng of the mErllaii sns by which genetic suocepti bi Ii ty may or may not influence the development of ri 9< behaviors such roacohol misuse . Nonetheless, they ca.Jtion that given the r~idly evolving state of genetic resea-ch , i mportait des gn consi dercii ons (e.g., caidi dcie gene vs. genome wide asrod cii ons vs. requend ng), aid unique consi dercii ons in the aia ysi s of genetic dcia, dose col I a:>orcii on betwea, geneti d sts aid ood a scientists ere necessa-y to achieve ma31i ngf ul advaices. Genetic studies have the ~a::>i Ii ty of doing much more thai simply I ooki ng a single indicators of ri 9< or suocepti bi Ii ty. Results of this study do not supp ort OPRM l ' s singular influe nce on the vu lnerab ility towa rd a lcohol misuse aid it may not be ad~uate to single out OPRM 1 ro a m~ or i ndi vi dua contributor to a cohol misuse. Rather ; it may be better to examine OPRM 1 within a caidi date systens cpprooch, in which multiple si ngl e-nud ooti de polymorphisms with f uncti ona rel e.,a,ce ere cggregated to form a oomposi te genetic risk &:Ore that may better explan aoohol outoomes (Derringer et a ., 2010).
Despite the noted Ii mi tati ons, the current res.JI ts contribute to understa,d i ng the psychoooci a oorrel ates of hea1y drinking aid a oohol probl ens among a de.ielopmentaly at-risk population, energing adults. Support for elenentsof the enha,ced rei nforcenent model ruch as the medi cti ng role of peroona i ty aid a oohol expecta,d es inf ami I y hi story a oohol outoome relations extends prior reg:gch as do the ob9:ll"Ved prospective assxi ati ons baween rubj ecti ve effects of a oohol aid hea1y drinking and aoohol problens. Aside from their etiologic significaice, thES:lfindings may further inform the de.iel opment of more ta I ored pre.ienti ve interventions that sped fi ca I y ta-get eti ol ogi ca I y rel e.ia,t factors ruch as rubj ecti ve response to a oohol (Schuckit et a. , in press) or peroonaity (Conrod, Castellaios-Rya,, & Stra,g, 2010).