Anda di halaman 1dari 13

Journal of Substance Abuse Treatment 38 (2010) 178 – 190

Regular article

Genetic, personality, and environmental predictors of drug use


in adolescents
Bradley T. Conner, (Ph.D.) a , Gerhard S. Hellemann, (Ph.D.) b ,
Terry L. Ritchie, (Ph.D.) c , Ernest P. Noble, (Ph.D., M.D.) c,⁎
a
Department of Psychology, Temple University, Philadelphia, PA 19122, USA
b
Biostatistics Core, Semel Institute for Neuroscience & Human Behavior, David Geffen School of Medicine, University of California Los Angeles,
Los Angeles, CA 90095, USA
c
Alcohol Research Center, Department of Psychiatry & Biobehavioral Sciences, Semel Institute for Neuroscience & Human Behavior,
David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA 90095, USA
Received 3 February 2009; received in revised form 23 June 2009; accepted 21 July 2009

Abstract

During adolescence there is a significant increase in risk-taking behavior, including experimenting with alcohol and drugs, which can lead
to drug dependence. A new hypothesis regarding the genetic mechanisms that lead to drug use is tested using adolescent Caucasian children
of alcoholics (57 males, 54 females; mean age = 14.5 years) data. Variables included in the study were dopaminergic genes (ANKK1 TaqI A,
DRD2 C957T, DRD4 7R, COMT Val/Met substitution, and SLC6A3 9R) and a GABAergic gene (GABRB3), all combinations of genes, a
count of the number of hypodopaminergic genotypes, personality traits, neurocognitive factors, depressive symptoms, and environmental
factors. Separate predictive models were found for males and females. Hypodopaminergic functioning predicted drug use in males; however,
in females, a deleterious environment was the salient predictor. This preliminary study suggests that it is possible to identify children at risk
for problematic drug use prior to the onset of drug dependence. © 2010 Elsevier Inc. All rights reserved.
Keywords: Dopamine genes; GABA genes; Hypodopaminergic functioning; Drug use; Recursive partitioning; Decision tree analysis

1. Introduction understanding of the role genetics plays in the propensity


to engage in risk-taking behavior. Behavioral genetic
As children enter the critical developmental period of research indicates that as much as 40% to 60% of the
adolescence, there is a significant increase in their risk-taking vulnerability to develop dependence on substances is
behavior, including experimenting with alcohol and drugs heritable (Button et al., 2006; Han, McGue, & Iacono,
(Arnett, 1992; Spear, 2000; Steinberg, 2008). These 1999). Identification of the specific genetic mechanisms of
behaviors often lead to devastating outcomes, including this transmission would be invaluable in the prevention and
drug abuse and dependence (Chen, Storr, & Anthony, 2009; treatment of these disorders.
Grant & Dawson, 1998). These negative outcomes have the Use of most drugs of abuse, including alcohol, is
potential to impact adolescents for the rest of their lives. associated with release of dopamine in the mesocorticolim-
Although there has been considerable research on adolescent bic system or “reward” pathway of the brain (Di Chiara &
risk-taking behavior, there are still many gaps in our Imperato, 1988). Activation of this dopamine system induces
feelings of reward and pleasure (Bressan & Crippa, 2005;
⁎ Corresponding author. UCLA Alcohol Research Center, 760 West-
Spanagel & Weiss, 1999). However, reduced activity of the
wood Plaza, Los Angeles, CA, USA. Tel.: +1 310 825 1891; fax: +1 310 206 dopamine system (hypodopaminergic functioning) can
7309. trigger drug-seeking behavior (Bowirrat & Oscar-Berman,
E-mail address: epnoble@ucla.edu (E.P. Noble). 2005; Melis, Spiga, & Diana, 2005). Variant alleles can

0740-5472/09/$ – see front matter © 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jsat.2009.07.004
B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190 179

cause hypodopaminergic functioning through reduced do- (VNTR) affects DRD4 functioning by modulating the
pamine receptor density, blunted response to dopamine, or expression and efficiency of maturation of this receptor
enhanced dopamine clearance or metabolism in the reward (Schoots & Van Tol, 2003). The 7 repeat (7R) VNTR
pathway (Lachman, 2006). Cessation of chronic drug use requires significantly higher levels of dopamine to produce a
induces a hypodopaminergic state that prompts drug-seeking response of the same magnitude as other size VNTRs (Oak,
behavior in an attempt to address the withdrawal-induced Oldenhof, & Van Tol, 2000). This reduced sensitivity to
state (Melis et al., 2005). dopamine leads to hypodopaminergic functioning. The 7R
These observations are the basis for the development of a VNTR has been associated with drug use disorders
functional hypothesis of drug-seeking and drug use. That (McGeary, Esposito-Smythers, Spirito, & Monti, 2007;
hypothesis is that the presence of a hypodopaminergic state, Oak et al., 2000). However, not all studies support this
regardless of its source, is a primary cause of drug-seeking relationship (Lusher, Chandler, & Ball, 2001).
behavior. Thus, genetic polymorphisms that cause hypodo- The dopamine transporter protein regulates dopamine-
paminergic functioning may be the causal mechanism of a mediated neurotransmission by rapidly accumulating dopa-
genetic predisposition to drug use. This hypothesis is unique mine that has been released into the synapse (Vandenbergh
in that it attempts to move the field away from studying the et al., 1992). The dopamine transporter gene (SLC6A3 or
relationships of genetic polymorphisms to drug use as DAT1) is localized to chromosome 5p15.3. Within the 3′
associations and, instead, suggests the mechanism that noncoding region of DAT1 lies a VNTR polymorphism
explains these associations. (Vandenbergh et al., 1992). The DAT1 9 repeat (9R) VNTR
The D2 dopamine receptor (DRD2) gene has probably has been shown to affect gene expression and to enhance
received the most attention as a candidate for the genetic transcription of the dopamine transporter protein (Michel-
transmission of drug use disorders (see Noble, 2003 for haugh, Fiskerstrand, Lovejoy, Bannon, & Quinn, 2001).
review). The TaqI A is a single nucleotide polymorphism This results in enhanced clearance of synaptic dopamine,
(SNP rs: 1800497) originally thought to be located at the 3′ leaving reduced amounts of dopamine to activate the
region of the DRD2 but which was later found to be located postsynaptic neuron, leading to hypodopaminergic func-
within exon 8 of an adjacent gene, the ankyrin repeat and tioning. Presence of the 9R VNTR has been associated with
kinase domain containing 1 (ANKK1; Neville, Johnstone, & drug use disorders (Guindalini et al., 2006; Köhnke et al.,
Walton, 2004). Neville et al. indicate that the mislocation of 2005); however, not all studies support this relationship
the TaqI A may be attributable to the ANKK1 and the DRD2 (Vandenbergh et al., 2002).
being on the same haplotype or the ANKK1 being involved in Catechol-O-methyltransferase (COMT) is an enzyme
reward processes through a signal transduction pathway. The involved in the metabolism of dopamine (Lachman et al.,
TaqI A is expressed as two alleles: the A1 and the A2 (Grandy 1996). The 108/158 codon results in a valine (Val) to
et al., 1989). Presence of the A1+ genotype (A1/A1, A1/A2), methionine (Met) substitution, which reduces COMT
compared to the A1− genotype (A2/A2), is associated with thermostability, even at physiological temperatures, and
reduced DRD2 density (Jönsson et al., 1999; Noble, Blum, significantly decreases enzymatic activity (Lachman et al.,
Ritchie, Montgomery, & Sheridan, 1991; Pohjalainen, Rinne, 1996). The Val/Val homozygote is four times more active in
Någren, Lehikoinen, et al., 1998). This reduction causes metabolizing dopamine than the Met/Met homozygote
hypodopaminergic functioning in the dopamine reward (Lachman et al., 1996). Faster metabolism results in reduced
pathway. In addition, research supports a predictive relation- dopamine availability at the synapse, which diminishes
ship from the A1+ genotype to drug use disorders (see Noble, postsynaptic activation, causing hypodopaminergic func-
2003 for review). However, not all studies have supported tioning. Some studies have shown an association between
this relationship (see Noble, 2003 for review). the Val variant and drug use disorders (Vandenbergh,
Another polymorphism that has received recent attention Rodriguez, Miller, Uhl, & Lachman, 1997; Wang et al.,
is the C957T, a SNP (rs: 6277) at exon 7 of the DRD2 gene 2001); however, not all studies support this association
(Duan et al., 2003). This SNP results in the presence of two (Samochowiec et al., 2006).
alleles: the 957T and the C957. Compared to the T− genotype Gamma-aminobutyric acid (GABA) receptor genes have
(C/C), the T+ genotype (T/T, T/C) is associated with also received some attention as candidates for drug use
decreased translation of DRD2 mRNA and diminished disorders. One reason for this is that the dopamine and
DRD2 mRNA (Duan et al., 2003), leading to decreased GABA systems are functionally interrelated (White, 1996).
DRD2 density (Hirvonen et al., 2004). This results in Research suggests that dopamine neurons projecting from
hypodopaminergic functioning, which renders the T + the anterior ventral tegmental area to the nucleus accum-
genotype a good candidate for the transmission of drug use bens are tonically inhibited by GABA through its actions at
disorders. In accord with this hypothesis, a recent study has the GABAA receptor (Ikemoto, Kohl, & McBride, 1997).
shown a predictive relationship between the T+ genotype and Moreover, it has been shown that alcohol (Theile,
alcohol dependence (Hill et al., 2008). Morikawa, Gonzales, & Morrisett, 2008) or opioid
There is evidence that the length of the D4 dopamine (Johnson & North, 1992) enhancement of GABAergic
receptor (DRD4) exon 3 variable number of tandem repeats (through GABAA receptor) transmission inhibits the release
180 B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190

of dopamine in the mesocorticolimbic system. Thus, a participant could fail to model the complicated association of
hyperactive GABA system, by inhibiting dopamine release, dopaminergic genes with drug use disorders.
could also lead to hypodopaminergic functioning. Because A number of additional variables have been reported as
of this, GABA genes are of interest in the search for causes significant predictors of drug use disorders. Among these
of drug use disorders. A dinucleotide repeat (DNR) are personality traits, positive and negative life events,
polymorphism of the GABA receptor β3 subunit gene negative affect, and neurocognitive functioning (Berman &
(GABRB3) results in either the presence of the 181-bp G1 Noble, 1995; Conner et al., 2005; Wills, Vaccaro, &
or 11 other repeats designated as non-G1 (NG1). Research McNamara, 1992). This further complicates etiological
indicates that the NG1 is more prevalent in children of understanding of drug use disorders because any or all of
alcoholics (COAs; Namkoong, Cheon, Kim, Jun, & Lee, these variables could interact with genes to better explain
2008). Presence of the NG1 has been associated with the causal mechanisms underlying the development of
alcohol dependence (Noble et al., 1998; Song et al., 2003). these disorders. Thus, rather than studying the association
In addition, other GABA receptor genes have also been of any single gene with drug use, as has been the
associated with this disorder (Edenberg et al., 2004). prevailing zeitgeist, researchers should study associations
To date, the replication of research that indicates that among complex gene–gene and gene–nongene (e.g.,
single polymorphisms predict drug use has been inconsis- psychological, neurological, and environmental variables,
tent, leading to confusion and, at times, lack of confidence in which may have their own genetic basis) interactions and
findings of genetic vulnerability to drug use disorders. This drug use disorders.
has also been true for association of single genes with other The goal of this preliminary, or pilot, study was to
complex behavioral disorders (NCI-NHGRI Working Group explore the complex relationships among genetic, personal-
on Replication in Association Studies et al., 2007). Although ity, mood, neurocognitive, and environmental variables that
population stratification has been the popular scapegoat for contribute to the development of drug use to test the
the mixed findings in gene association studies (Hutchison, hypodopaminergic functioning hypothesis proposed above
Stallings, McGeary, & Bryan, 2004), there are a number of in a sample of adolescent COAs. This is a good sample for
other reasons that might explain these difficulties (Colhoun, this exploratory research, as COAs are at an increased risk
McKeigue, & Smith, 2003; Nestler, 2000). for developing drug use disorders when compared to
One reason is that any single gene is likely to have a children whose parents are not dependent on alcohol
relatively small effect on a complex disorder, making (Heath et al., 1997). It is expected, given the previous
experimental detection difficult (Nestler, 2000). A recent research described above, that predictors will vary based on
study (Berggren et al., 2006), using the largest number of sex of the participant; so, in addition to conducting this
participants to date, found that the TaqI A A1+ genotype was exploratory analysis within the total sample, similar analyses
significantly associated with alcohol dependence; however, were run on the two sexes separately.
the effect size was small. Mixed findings are expected when
dealing with small effect sizes. The same holds true for the
association of any single gene with other behavioral 2. Materials and methods
disorders (NCI-NHGRI Working Group on Replication in
Association Studies et al., 2007). Further, complex gene– 2.1. Participants
gene and gene–environment interactions, although frequent-
ly hypothesized as the causal mechanism of drug use Participants' parents were primarily recruited through
disorders (Colhoun et al., 2003; NCI-NHGRI Working advertisements on alcoholism treatment center bulletin
Group on Replication in Association Studies et al., 2007; boards and in treatment community newspapers. Parents
Nestler, 2000), are rarely tested. However, when they are who were interested in having their children participate
tested, the findings support the notion that complex gene– contacted the UCLA Alcohol Research Center. On this
gene relationships may account for inconsistent findings telephone call, they were given more detailed information
across different single gene studies (Yang et al., 2008). and a brief initial screening interview. During a subsequent
Another reason for the difficulties in replicating single- interview conducted at the UCLA Alcohol Research
gene associations with drug use disorders is sex-based Center, the Structured Clinical Interview for DSM-IV
differences in neurochemistry and neuroanatomy (de Disorders (First, Spitzer, Gibbon, & Williams, 1996) was
Courten-Myers, 1999). More specifically, significant differ- administered to all parents to establish the presence or
ences have been found in brain dopaminergic activity absence of alcohol dependence as well as the presence or
between males and females (Kaasinen, Nägren, Hietala, absence of other drug abuse/dependence and psychiatric
Farde, & Rinne, 2001; Pohjalainen, Rinne, Nägren, disorders. To be included in the study, participants had to
Syvälahti, & Hietala, 1998). These sexually dimorphic have at least one parent who met the Diagnostic and
brain differences could explain differential prevalence of Statistical Manual of Mental Disorders, Fourth Edition
drug use disorders between sexes (Feldstein & Miller, 2006; (DSM-IV-TR; American Psychiatric Association, 2000)
Limosin et al., 2002). Studies that do not consider sex of the criteria for alcohol dependence.
B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190 181

As part of the overall study protocol, adolescent (R) included the 2R (379 bp), 3R (427 bp), 4R (475 bp), 5R
participants were excluded from the study if they had (523 bp), 6R (571 bp), 7R (619 bp), and 8R (667 bp) alleles.
hearing or visual impairment (unless corrected to normal), The presence of the 7R homozygote (7R/7R) or heterozygote
color blindness, or a history of head injury with a loss of (7R/N7R) was considered a genetic risk marker for
consciousness greater than 5 minutes or other evidence of hypodopaminergic functioning compared to the absence of
organic brain damage. These exclusions were necessary the 7R (N7R/N7R).
because of the event-related potential electroencephalogra-
phy data that were collected but were not used in this study. 2.2.4. Dopamine transporter (DAT1) VNTR
Ethnicity/Race was reported by the COAs' parents about The DAT1 VNTRs were determined by a standard PCR
themselves, their children, and their parents. In this study, procedure (Vandenbergh et al., 1992). The VNTRs in the
to control for population stratification effects, the sample PCR product were resolved by 2.5% agarose gel electro-
used was restricted to non-Hispanic Caucasian participants phoresis. VNTRs of interest included the 7R (360 bp), 8R
of European descent. This left a final sample of 111 (400 bp), 9R (440 bp), 10R (480 bp), 11R (520 bp), and 12R
Caucasian adolescent COAs (57 males and 54 females, (560 bp). The presence of the 9R homozygote (9R/9R) or
mean age = 14.5 years). heterozygote (9R/N9R) was considered a genetic risk marker
All children, as well as their parents, gave informed for hypodopaminergic functioning compared to the absence
assent. Participants were monetarily compensated for their of the 9R (N9R/N9R).
participation. The research protocol had the approval of the
UCLA Institutional Review Board. 2.2.5. COMT Val108Met SNP
COMT polymorphism was determined by a standard
2.2. Genotyping procedures PCR procedure (Hoda et al., 1996). The resultant PCR
product was digested by N1aIII restriction endonuclease,
A 10-ml blood sample was drawn from each participant. and the DNA fragments resolved by 4% agarose gel
The amplification of DNA was carried out using a Perkin electrophoresis. The Val-108 allele gives DNA fragments of
Elmer GeneAmp 9600 thermocycler. 136 and 81 bp. The Met-108 allele gives DNA fragments of
96, 81, and 40 bp. The Val homozygote (Val/Val) or
2.2.1. ANKK1 TaqI A SNP heterozygote (Val/Met) was considered a genetic risk
The ANKK1 TaqI A polymorphism was determined marker for hypodopaminergic functioning compared to the
using a standard polymerase chain reaction (PCR) procedure Met homozygote (Met/Met).
(Grandy, Zhang, & Civelli, 1993). The PCR product was
digested with TaqI restriction endonuclease, and the resultant 2.2.6. GABRB3 CA-dinucleotide repeat
products were resolved by 2% agarose gel electrophoresis. The GABRB3 CA-dinucleotide repeat polymorphism
Two alleles were obtained: the A1 allele (the uncleaved 310- was determined by a modified (Noble et al., 1998) PCR
bp fragment) and the A2 allele (the cleaved 180- and 130-bp procedure (Mutirangura, Ledbetter, Kuwano, Chinault, &
fragments). The presence of the A1 homozygote (A1/A1) or Ledbetter, 1992). The PCR products were resolved using
heterozygote (A1/A2) was considered a genetic risk marker polyacrylamide gel electrophoresis and the 33P-labeled DNA
for hypodopaminergic functioning compared to the A2 fragments revealed by autoradiography. Twelve CA-repeat
homozygote (A2/A2). alleles of interest were designated as follows: G1 (181 bp)
and NG1: G2 (183 bp), G3 (185 bp), G4 (187 bp), G5 (189
2.2.2. DRD2 C957T SNP bp), G6 (191 bp), G7 (193 bp), G8 (195 bp), G9 (197 bp),
The DRD2 C957T SNP was determined by a standard G10 (199 bp), G11 (201 bp), and G12 (203 bp). The
PCR procedure (Duan et al., 2003). The PCR product was presence of either the NG1 homozygote (NG1/NG1) or
digested with TaqI restriction endonuclease, and the resultant heterozygote (NG1/G1) was considered a genetic risk
products were resolved by 2.5% agarose gel electrophoresis. marker for hypodopaminergic functioning compared to the
Two alleles were obtained: the T allele (uncleaved 196-bp G1 homozygote (G1/G1).
fragment) and the C allele (the cleaved 174- and 22-bp
fragments). The presence of the T homozygote (T/T) or 2.2.7. Genotype combinations
heterozygote (T/C) was considered a genetic risk marker for In addition to examining each of the genotypes listed
hypodopaminergic functioning compared to the C homozy- above, all possible genotype combinations (pairs, triplets,
gote (C/C). quadruplets, quintuplets, and sextuplets) were entered as
predictors. In addition, the presence of each genotype that
2.2.3. DRD4 Exon III VNTR coded for hypodopaminergic functioning was scored as 1,
The DRD4 VNTRs were determined using a standard and the presence of the alternative genotype was scored as 0.
PCR procedure (Lichter et al., 1993). The VNTRs in the Scores were then summed across locations similar to the
PCR product were resolved by 2.5% agarose gel electro- method used by Harlaar et al. (2005). This resulted in a count
phoresis. A number of VNTRs were obtained. The repeats of the number of hypodopaminergic genotypes per
182 B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190

participant, hereafter referred to as the hypodopaminergic 2.3.4. Depressive symptoms


genetic risk variable. Depressive symptoms were assessed using the Beck
Depression Inventory-II (BDI-II; Beck, 1996). The BDI-II is
2.3. Predictor variables a 21-item state-based depressive symptoms self-report
questionnaire. Multiple research studies suggest that depres-
A number of predictor variables were included in this sive symptoms are predictive of drug use (Aseltine, Gore, &
exploratory analysis in addition to the genetic data. These Colten, 1998; Fleming, Mason, Mazza, Abbott, & Catalano,
measures were selected because previous research has 2008; King, Iacono, & McGue, 2004).
suggested that these variables are significant predictors of
the onset of drug use in adolescence. Although this list of 2.4. Outcome variable—drug use
predictors is certainly not exhaustive, it represents a robust
collection of variables that could better account for variations The self-report Health Behavior Questionnaire (Jessor,
in drug use in adolescence. Donovan, & Costa, 1992) was administered to the
participants. On this questionnaire, participants indicated
2.3.1. Visuospatial ability drugs that they have tried at least once from the following
Visuospatial ability, assessed using a computerized list of drugs: alcohol, nicotine, marijuana, cocaine/crack,
version (Berman & Noble, 1995) of the Judgment of Line heroin or other opiates, phenylcyclidine, hallucinogen,
Orientation Test (JLOT; Benton, Varney, & Hamsher, nitrous oxide, gamma-hydroxybutyric acid, amphetamine,
1978), was used as a measure of neurocognitive and barbiturates. A count variable was computed by scoring
functioning. Previous research suggests links between “I have tried…” as 1 and “I have not tried…” as 0 and
neurocognitive functioning and drug use (Shoal & summing across all drug types. This method of measuring
Giancola, 2001). drug use severity in adolescents was developed by Wills
et al. (1992) and used by Conner et al. (2005) in prior
2.3.2. Personality research. The number of drugs tried has been hypothesized
Psychoticism (P), extraversion (E), and neuroticism to be a precursor for the later development of drug use
(N) were assessed with the Eysenck Personality Ques- disorders (Robins & Przybeck, 1991). Finally, recent
tionnaire-Junior (Eysenck & Eysenck, 1975). Novelty research conducted to create a substance use disorder
Seeking (NS), Harm Avoidance (HA), and Reward (SUD) severity scale using item response theory found that
Dependence (RD) were assessed with the Junior Tem- SUD severity significantly predicted the number of different
perament and Character Inventory (Luby, Svrakic, drugs tried by the age of 19, which the authors used as a test
McCallum, Przybeck, & Cloninger, 1999). Previous of construct and concurrent validity for their SUD scale
research indicates that some of these personality variables (Kirisci et al., 2006).
are predictive of drug use, including P (Conner et al.,
2005; Sáiz et al., 2003), E (Merenäkk et al., 2003), and 2.5. Statistical analyses
N (Florsheim et al., 2007). Research also supports
predictive relationships between NS (Cloninger, Sigvards- Data were analyzed using recursive partitioning (RP), a
son, & Bohman, 1988; Mâsse & Tremblay, 1997), HA type of exploratory decision tree analysis (Zhang & Singer,
(Cloninger et al., 1988; Tcheremissine, Lane, Cherek, & 1999). RP is a non- or semiparametric data mining procedure
Pietras, 2003), and RD (Tcheremissine et al., 2003) and that builds a “tree” that identifies exhaustive orthogonal
drug use in adolescence. subgroups of a population, called “leaves” or “nodes,” whose
members share common characteristics that influence the
2.3.3. Positive and negative life events dependent variable (Lemon, Roy, Clark, Friedmann, &
Positive and negative life events were measured using Rakowski, 2003). The data are then split into smaller and
the Life Stressors and Social Resources Inventory-Youth smaller nodes using a recursive procedure until each group is
scale (Moos & Moos, 1994). These variables were perfectly homogenous. This recursive splitting takes place
counts of positive (range = 0–45) and negative (range = separately for each “branch” of the tree.
0–78) life events across multiple domains (i.e., in RP is an ideal method when searching for candidate genes
school, at home) that occurred in the 12 months prior and haplotypes (Zaykin & Young, 2005) and has been used
to the interview. Previous research suggests that both successfully in drug abuse research (Hellemann, Conner,
positive (Scheier, Botvin, & Miller, 1999; Wills et al., Anglin, & Longshore, 2009). The strength of this method is
1992) and negative (Taylor, 2006; Wills et al., 1992) life that it allows for analysis of highly dimensional and highly
events are related to drug use among adolescents. correlated data in small samples (Breiman, Freidman,
Although stress scores were also available for use as Olshen, & Stone, 1984). It detects and describes complex
predictor variables, almost all of the participants had interactions between predictor variables even in extreme
high stress scores, which significantly restricted their cases where the number of predictors is larger than the
utility in these analyses. number of independent observations. It identifies a
B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190 183

parsimonious subset of predictor variables that is sufficient could be an important moderating variable (Flannery,
to predict the specified outcome variable. Vazsonyi, Torquati, & Fridrich, 1994; Jang, Livesley, &
As splitting continues until perfect homogeneity is Vernon, 1997; Kaasinen et al., 2001; Pohjalainen, Rinne,
reached, the final RP tree is likely overfit to the data from Nägren, Syvälahti, et al., 1998). The cut-point estimation
the particular sample under investigation. The goal, was based on the Poisson distribution to account for the
however, is to specify a model that will be applicable effect of the right skew of count data of rare events. The
to any random sample from the population. As with most choice of the splitting rule was verified by evaluating
stepwise procedures, the question arises of when to stop alternative splitting rules using the cross-validated error as
fitting the model to the data. The second stage of the RP the criterion. The size of the terminal nodes was not preset.
procedure consists of using cross validation to prune the The cross validation was based on 100 bootstrapped
full tree to an “appropriate size” to reach this goal. This samples. The rpart package version 3.1-33 (Therneau &
is ideally done using the decision rules specified by the Atkinson, 2006), as implemented in version 2.4 of the R
tree on a new sample from the population (test sample) software package, was used to estimate the RP trees.
and evaluating its ability to accurately predict group
membership. When such a dataset is not available,
however, other methods must be employed. One com- 3. Results
monly used method is to bootstrap the original sample.
Bootstrapping consists of repeated reestimation of the RP Table 1 presents descriptive data on the variables in the
tree using random samples (with replacement) from the study. Table 2 presents the gene frequency data. Table 3
original dataset. The bootstrap cross-validation procedure presents the correlations between each of the predictor
estimates the cross-validated prediction error, and related variables and the outcome variable. Results from Hardy–
standard error, for different numbers of splits. Appropriate Weinberg Equilibrium analyses revealed that genotype
tree size is usually determined by cost, model complexity, frequencies in this sample did not differ significantly from
or cross validation. Breiman et al. (1984) suggested using what would be expected in the general population.
a ±1 standard error rule to choose appropriate tree size. In
other words, a good-fitting tree is one with the least 3.1. Analysis of the entire sample
number of splits and the smallest cross-validation error,
given that the tree's cross-validation error plus its The RP analysis of the entire sample of COAs (n = 111)
standard error is less than 1 (per definition, the relative produced a model that had a relative error = 0.77 and a cross-
error of a model with no splits). validation error = 0.91 (SE = 0.16), indicating that this model
In this study, RP was used to determine the best explains approximately 23% of the variance in the entire
predictors of drug use among 75 potential predictor sample and, on average, 9% in the bootstrap cross-validation
variables: age, six genotypes, 57 specific combinations of procedure. However, given the size of the standard error, the
the six genotypes, the count of hypodopaminergic geno- 9% variance explained is not significantly different from
types, six personality scale scores, BDI-II scores, JLOT zero, indicating a model with unacceptable fit. Based on
scores, and positive and negative life events scores. The previous research, this result was not unexpected and
data were analyzed in three groups: the entire sample, supported the decision to evaluate separately the two sex-
males, and females, as previous research indicates that sex based subsamples.

Table 1
Variables in the analysis of predictors of drug use in COAs
Variable Total (N = 111) Male (n = 57) Female (n = 54)
Age, M (SD) 14.5 (1.3) 14.5 (1.2) 14.6 (1.4)
Personality, M (SD)
Psychoticism 3.06 (2.83) 4.02 (3.09) 2.07 (2.15)
Extraversion 19.07 (3.59) 19.59 (3.13) 18.54 (3.98)
Neuroticism 9.65 (4.46) 9.57 (4.72) 9.74 (4.22)
NS 8.77 (3.13) 8.48 (3.06) 9.08 (3.20)
RD 4.86 (2.19) 4.36 (2.17) 5.40 (2.10)
HA 6.56 (3.93) 6.89 (3.97) 6.21 (3.89)
Visuospatial Ability, M (SD) 23.64 (4.34) 24.40 (4.81) 22.90 (3.73)
Depressive Symptoms, M (SD) 4.07 (6.08) 3.91 (6.45) 4.24 (5.72)
Negative Life Events, M (SD) 8.52 (5.12) 9.04 (4.81) 8.00 (5.58)
Positive Life Events, M (SD) 10.06 (3.48) 10.26 (3.51) 9.85 (3.47)
Number of drugs tried
M (SD) 1.76 (1.93) 1.79 (1.72) 1.72 (2.15)
Median 1.00 2.00 1.00
184 B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190

Table 2 0.18). The best predictor of drug use was the hypodopami-
Genotype frequencies nergic genetic risk variable. Males who had 4 or more
Gene Total (N = 111) Male (n = 57) Female (n = 54) hypodopaminergic genotypes tried a mean of 3.02 drugs
DRD2 TaqI A SNP compared to males who had less than four hypodopaminer-
A2/A2 67 (61) 38 (67) 29 (54) gic genotypes, who tried a mean of 1.29 drugs. These two
A1/A2 39 (35) 18 (31) 21 (39) groups were further split into seven distinctive groups. Males
A1/A1 5 (4) 1 (2) 4 (7)
who had four or more hypodopaminergic genotypes and a
DRD2 C957 SNP
C/C 29 (26) 17 (30) 12 (22) RD score equal to or greater than 5.5 tried a mean of 4.27
C/T 58 (52) 28 (49) 30 (56) drugs, the highest drug use. Males who had less than four
T/T 24 (22) 12 (21) 12 (22) hypodopaminergic genotypes, a Psychoticism score of less
DRD4 exon III VNTR than 2.5, and a RD score of less than 6 tried a mean of 0.10
N7R/N7R 73 (66) 36 (63) 37 (68)
drugs, the lowest drug use.
N7R/7R 33 (30) 18 (32) 15 (28)
7R/7R 5 (4) 3 (5) 2 (4) In addition to the groups with the highest and lowest
DAT1 VNTR drug use, the male COAs were further divided into three
N9R/N9R 59 (53) 32 (56) 27 (50) intermediate nodes. Males who had less than four
N9R/9R 48 (43) 22 (39) 26 (48) hypodopaminergic genotypes and a Psychoticism Score
9R/9R 4 (4) 3 (5) 1 (2)
of 2.5 or more, and the DRD4 heterozygote (7R/N7R)
COMT Val108Met SNP
Met/Met 27 (24) 15 (27) 12 (23) tried a mean of 0.66 drugs, compared to males who had
Val/Met 62 (56) 31 (54) 31 (57) the DRD4 homozygotes (7R/7R, N7R/N7R), who had
Val/Val 22 (20) 11 (19) 11 (20) tried a mean of 2.39 drugs. The latter group was further
GABRB3 DNR a divided on their number of hypodopaminergic genotypes.
G1/G1 14 (14) 4 (8) 10 (20)
Participants who had 3 hypodopaminergic genotypes tried
G1/NG1 41 (41) 21 (42) 20 (41)
NG1/NG1 44 (45) 25 (50) 19 (39) a mean of 3.25 drugs compared to those who had less than
Hypodopaminergic genotype count three hypodopaminergic genotypes, who tried a mean of
M (SD) 3.23 (0.97) 3.14 (1.00) 3.31 (0.95) 1.07 drugs.
Median 3.00 3.00 3.00
Values are expressed as number (percentage). 3.3. Analysis of the female subsample
a
Twelve participants (seven males, five females) were missing
GABRB3 data. The RP tree for the female subsample of COAs is
presented in Fig. 2. The overall fit of the model was good in
3.2. Analysis of the male subsample that this model explained 53% of the variability in this
sample, and on average, 20% in the bootstrap cross-
The RP tree for the male subsample of COAs is shown in validation procedure (relative error = 0.47, cross-validation
Fig. 1. The overall fit of the model was good in that it error = 0.80, SE = 0.15). The best predictor of drug use was
explained 68% of the variability in this sample and, on Negative Life Events. Female COAs who reported less than
average, 21% in the bootstrap cross-validation procedure 7.5 Negative Life Events in the previous 12 months tried a
(relative error = 0.32, cross-validation error = 0.79, SE = mean of 0.85 drugs. Those that reported 7.5 or more

Table 3
Correlations among the predictors and the outcome variable
Variable combination Total (N = 111) Male (n = 57) Female (n = 54)
Age M (SD) and NDT 0.34 ⁎⁎⁎ 0.20 0.43 ⁎⁎⁎
Personality
Psychoticism and NDT 0.16 0.18 0.23
Extraversion and NDT 0.21 ⁎ 0.24 0.21
Neuroticism and NDT 0.12 0.02 0.21
NS and NDT 0.22 ⁎ 0.09 0.31 ⁎
RD and NDT 0.16 0.24 0.10
HA and NDT −0.18 −0.16 −0.19
Visuospatial Ability and NDT −0.19 −0.11 −0.25
Depressive Symptoms and NDT 0.04 −0.01 0.08
Negative Life Events and NDT 0.09 0.29 −0.08
Positive Life Events and NDT −0.19 −0.05 −0.30
Hypodopaminergic Genotype Count and NDT 0.94 0.41 ⁎⁎ −0.14
Numbers are Pearson r values. NDT = number of drugs tried.
⁎ p b .05.
⁎⁎ p b .01.
⁎⁎⁎ p b .001.
B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190 185

Fig. 1. Decision tree for the male subsample (n = 57). The first number within each node represents the mean number of drugs tried for the group, and the second
number represents the size of the group. Circles represent internal nodes that are split further, and squares represent terminal nodes. Relative error = 0.32, cross-
validation error = 0.79 (SE = 0.18).

Negative Life Events in the previous 12 months tried a mean 3.5 on the HA scale, who tried a mean of 5.62 drugs, the
of 2.81 drugs. The female COAs were further divided into highest drug use. Among those females that had less than 7.5
four separate nodes based on two additional splits. Among Negative Life Events, those who scored 9.5 or more on the
female COAs that reported 7.5 or more Negative Life NS scale tried a mean of 1.62 drugs compared to those that
Events, those who scored 3.5 or more on the HA scale tried a scored lower than 9.5 on the NS scale, who tried a mean of
mean of 1.78 drugs compared to those who scored less than 0.28 drugs, the lowest drug use.

Fig. 2. Decision tree for the female subsample (n = 54). The first number within each node represents the mean number of drugs tried for the group, and the
second number represents the size of the group. Circles represent internal nodes that are split further, and squares represent terminal nodes. Relative error = 0.47,
cross-validation error = 0.80 (SE = 0.15).
186 B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190

4. Discussion type, affected performance on reward-based learning tasks


(Cohen, Krohn-Grimberghe, Elger, & Weber, 2007) and
Using RP to search through a large set of genetic, individual ability to learn from errors (Klein et al., 2007).
personality, neurocognitive, affective, and life event vari- This indicates that decreased sensitivity to negative con-
ables, this exploratory study was not able to identify a stable sequences may help explain increased risk for drug use
model of predictors of drug use, nor support the hypodopa- disorders. The findings of this study are also bolstered by the
minergic hypothesis proposed in the introduction in the total research that indicates that hyperdopaminergic functioning
sample of adolescent Caucasian COAs. However, stable serves as a protective factor against the development of drug
models were identified for males and females when analyzed use disorders (Volkow et al., 2006).
separately. In males, hypodopaminergic genetic risk was the The limitations of this study should be considered when
salient predictor of the number of drugs tried, whereas in interpreting the findings. This study was exploratory,
females, it was an environmental factor, Negative Life meaning that replication of the findings in confirmatory
Events. These findings are consistent with a behavioral– studies needs to occur prior to extensive inference of the
genetic study (Jang et al., 1997) of twins that showed that meaning of these relationships. The sample was composed of
drug use in males was significantly determined by genetic Caucasian adolescent COAs, limiting the generalizability of
factors, whereas in females, it was wholly determined by the findings to this population. In addition, there is the
environmental factors. This finding is also consistent with possibility that certain nonmeasured or nonincluded vari-
additional research suggesting significant differences in the ables could explain the findings reported herein, as with all
pathways to drug use between males and females (Derringer, nonexperimental studies. An additional limitation of this
Krueger, McGue, & Iacono, 2008; Han et al., 1999; study was the criterion variable. Although previous research
Kashdan, Vetter, & Collins, 2005). suggests that the number of drugs tried is a good proxy
In this study, the decision tree for males identified the measure of substance use severity (research reviewed
number of hypodopaminergic genotypes as the most above), quantity or frequency of drug use may have provided
important variable in determining drug use. The other more relevant information on adolescent substance use
important variables in the male model were the personality severity. Unfortunately, because this sample was composed
traits Psychoticism and Reward Dependence, with higher of young adolescents, the frequency and quantity of use
scores on each being predictive of a greater number of drugs variables were too highly skewed and poorly distributed to
tried. Individuals who score high compared to those who be of use in these analyses. Future research should include
score average to low on measures of Psychoticism tend to these variables as outcomes whenever possible. Using an
have less control over their emotions and exhibit excessive SUD scale (e.g., Kirisci et al., 2006) may also improve the
aggressive or impulsive behaviors (Eysenck & Eysenck, generalizability of these findings.
1975). Psychoticism has been shown to be positively Additional ways to build on the current results in future
correlated with drug use and drug use disorders (Conner research would be to include other candidate genes that
et al., 2005). Individuals who score high compared to those influence how reward is interpreted in the brain, additional
who score low on measures of Reward Dependence tend to measures of environmental factors that affect drug use
show a greater bias in the maintenance of behavior in behaviors, and more comprehensive measures of neurocog-
response to cues of social reward (Cloninger, 1987). nitive functioning in a confirmatory analysis. Future
Previous research has found a positive predictive relation- analyses should also specifically test the hypodopaminergic
ship between higher reward dependence and drug use among hypothesis in multiple and varied samples to continue to
adolescents (Deas & Thomas, 2002). provide support for hypodopaminergic functioning as a
For the females, two personality traits, NS and HA, were causal mechanism for the onset of drug use and the
identified as important in predicting drug use after dividing development of drug use disorders.
the sample based on the number of Negative Life Events they It is also important to consider the strengths of this study.
had experienced. These splits were not surprising, as One of the strengths is the use of RP in the analysis of the
previous studies have found higher NS and lower HA in data. This allowed for the identification of a pattern of
children to be strongly predictive of drug use and abuse relationships that predicts drug use, which not only holds
(Cloninger et al., 1988). true for this modest sample but generalizes to the entire
The present exploratory study supported the hypothesis population from which this sample was drawn, as shown by
developed in the introduction by identifying hypodopami- the boot-strapped cross validation.
nergic genotypes as the best predictor of drug use behavior in An additional strength of this study is that this is the first
male adolescent COAs. It is in accord with prior research that time that the cumulative effect of multiple genotypes coding
identified a hypodopaminergic state as a causal mechanism for hypodopaminergic functioning, regardless of their
in the development of drug use disorders (Melis et al., 2005). genomic location, has been identified as an important
It is also consistent with recent functional Magnetic predictor of drug use in males. This has important implica-
Resonance Imaging (fMRI) studies indicating that lower tions regarding the mixed findings observed when only single
DRD2 levels, determined by hypodopaminergic A1 geno- genes are studied. The findings suggest that a source of
B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190 187

variation in the predictive relationship from any single Arnett, J. (1992). Reckless behavior in adolescence: A developmental
candidate gene to drug use, beyond itself, is the variation of perspective. Developmental Review, 12, 339−373.
Aseltine, R. H., Jr., Gore, S., & Colten, M. E. (1998). The co-occurrence of
other hypodopaminergic genotypes. In other words, in any depression and substance abuse in late adolescence. Development and
study assessing a single gene, the strength of the predictive Psychopathology, 10, 513−529.
relationship or its effect size, will depend on the distribution Beck, A. T. (1996). Beck depression inventory-II (BDI-II). San Antonio,
of the unobserved hypodopaminergic genotypes. Variations TX: The Psychological Corporation.
in these distributions across different studies are a likely Benton, A. L., Varney, N. R., & Hamsher, K. D. (1978). Visuospatial
judgment. A clinical test. Archives of Neurology, 35, 364−367.
cause of nonreplicable single gene findings. Thus, studying Berggren, U., Fahlke, C., Aronsson, E., Karanti, A., Eriksson, M., Blennow,
multiple polymorphic locations that lead to hypodopaminer- K., Thelle, D., Zetterberg, H., & Balldin, J. (2006). The Taq1 DRD2 A1
gic functioning in the reward pathway should lead to larger allele is associated with alcohol-dependence although its effect size is
effect sizes and increased replicability of findings of the small. Alcohol and Alcoholism, 41, 479−485.
Berman, S. M., & Noble, E. P. (1995). Reduced visuospatial performance in
genetic contribution to drug use disorders. In addition, at least
children with the D2 dopamine receptor A1 allele. Behavioral Genetics,
for males, this study supports a hypothesis that describes the 25, 45−58.
mechanism by which genetic polymorphisms contribute to Bowirrat, A., & Oscar-Berman, M. (2005). Relationship between dopami-
drug seeking and drug use, not just associations between nergic neurotransmission, alcoholism, and reward deficiency syndrome.
genotypes and complex human behavior. American Journal of Medical Genetics Part B, Neuropsychiatric
Genetics, 132B, 29−37.
A further strength of this study was the identification of
Breiman, L., Freidman, J. H., Olshen, R. A., & Stone, C. J. (1984). Clas-
sex differences and the development of two separate sification and regression trees. Belmont, CA: Wadsworth International.
predictive models. Males with hypodopaminergic function- Bressan, R. A., & Crippa, J. A. (2005). The role of dopamine in reward and
ing are more likely to use drugs that stimulate the pleasure behaviour—Review of data from preclinical research. Acta
mesocorticolimbic system than those with normal dopami- Psychiatrica Scandinavica, 111(Suppl 427), 14−21.
Button, T. M., Hewitt, J. K., Rhee, S. H., Young, S. E., Corley, R. P., &
nergic functioning. In contrast, among females, those living
Stallings, M. C. (2006). Examination of the causes of covariation
in environments rife with negative events are at increased between conduct disorder symptoms and vulnerability to drug
risk for using more drugs and more types of drugs, which dependence. Twin Research & Human Genetics, 9, 38−45.
increases their risk of developing drug use disorders. These Chen, C. -Y., Storr, C. L., & Anthony, J. C. (2009). Early-onset drug use and
findings also have important theoretical implications. First, risk for drug dependence problems. Addictive Behaviors, 34, 319−322.
Cloninger, C. R. (1987). Neurogenetic adaptive mechanisms in alcoholism.
males and females may start using drugs and develop drug
Science, 236, 410−416.
use disorders for entirely different reasons. Second, studies Cloninger, C. R., Sigvardsson, S., & Bohman, M. (1988). Childhood
that fail to consider these sex differences may have difficulty personality predicts alcohol abuse in young adults. Alcohol: Clinical
detecting significant findings or may misrepresent the and Experimental Research, 12, 494−505.
etiology of drug use disorders if the sample is dominated Cohen, M. X., Krohn-Grimberghe, A., Elger, C. E., & Weber, B. (2007).
Dopamine gene predicts the brain's response to dopaminergic drug.
by one sex.
European Journal of Neuroscience, 26, 3652−3660.
There are practical implications of both the method used Colhoun, H. M., McKeigue, P. M., & Smith, G. D. (2003). Problems of
and the findings of this study. Identifying those at risk prior reporting genetic associations with complex outcomes. Lancet, 361,
to the onset of drug use disorders allows for early diversion 865−872.
and prevention and, thus, avoidance of the negative Conner, B. T., Noble, E. P., Berman, S. M., Ozkaragoz, T., Ritchie, T.,
Antolin, T., & Sheen, C. (2005). DRD2 genotypes and substance use in
outcomes associated with such use. The decision trees
adolescent children of alcoholics. Drug and Alcohol Dependence, 79,
generated by RP identify those at increased risk for 379−387.
developing drug use disorders during an early stage of Deas, D., & Thomas, S. (2002). Comorbid psychiatric factors contributing to
their lives. In addition, the results suggest that different adolescent alcohol and other drug use. Alcohol Research & Health, 26,
groups of at-risk individuals would benefit from interven- 116−121.
de Courten-Myers, G. M. (1999). The human cerebral cortex: Gender
tions that are specifically tailored to their risk profile.
differences in structure and function. Journal of Neuropatholology and
Experimental Neurology, 58, 217−226.
Acknowledgments Derringer, J., Krueger, R. F., McGue, M., & Iacono, W. G. (2008). Genetic
and environmental contributions to the diversity of substances used in
This research was supported by the Christopher D. adolescent twins: A longitudinal study of age and sex effects. Addiction,
Smithers Foundation, Inc., New York. The authors would 103, 1744−1751.
like to thank Ms. Bonita Porch for her editorial assistance. Di Chiara, G., & Imperato, A. (1988). Drugs abused by humans
preferentially increase synaptic dopamine concentrations in the
This study was presented in part at the 115th Annual
mesolimbic system of freely moving rats. Proceedings of the
Convention of the American Psychological Association, San National Academy of Sciences of the United States of America, 85,
Francisco, CA, August 19, 2007. 5274−5278.
Duan, J., Wainwright, M. S., Comeron, J. M., Saitou, N., Sanders, A. R.,
References Gelernter, J., & Gejman, P. V. (2003). Synonymous mutations in the
human dopamine receptor D2 (DRD2) affect mRNA stability and
American Psychiatric Association. (2000). Diagnostic and statistical synthesis of the receptor. Human Molecular Genetics, 12, 205−216.
manual of mental disorders, 4th ed, Text Revision (DSM-IV-TR). Edenberg, H. J., Dick, D. M., Xuei, X., Tian, H., Almasy, L., Bauer, L. O.,
Washington, DC: American Psychiatric Press, Inc. Crowe, R. R., Goate, A., Hesselbrock, V., Jones, K., Kwon, J., Li, T. -K.,
188 B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190

Nurnberger, J. I., Jr., O'Connor, S. J., Reich, T., Rice, J., Schuckit, M.A., Hoda, F., Nicholl, D., Bennett, P., Arranz, M., Aitchison, K. J., Al-Chalabi,
Porjesz, B., Foroud, T., & Begleiter, H. (2004). Variations in GABRA2 A., Kunugi, H., Vallada, H., Leigh, P. N., Chaudhuri, K. R., & Collier,
encoding the alpha 2 subunit of the GABAA receptor, are associated with D.A. (1996). No association between Parkinson's disease and low-
alcohol dependence and with brain oscillations. American Journal of activity alleles of catechol-O-methyltransferase. Biochemical and
Human Genetics, 74, 705−714. Biophysical Research Communications, 228, 780−784.
Eysenck, H. J., & Eysenck, S. B. G. (1975). Manual of the Eysenck Personality Hutchison, K. E., Stallings, M., McGeary, J., & Bryan, A. (2004).
Questionnaire (Junior and Adult). London: Hodder & Staughton. Population stratification in the Candidate Gene Study: Fatal threat or
Feldstein, S. W., & Miller, W. R. (2006). Substance use and risk-taking red herring? Psychological Bulletin, 130, 66−79.
among adolescents. Journal of Mental Health, 15, 633−643. Ikemoto, S., Kohl, R. R., & McBride, W. J. (1997). GABAA receptor
First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1996). blockage in the ventral tegmental area increases extracellular levels of
Structured clinical interview for DSM-IV Axis I disorders, clinician dopamine in the nucleus accumbens of rats. Journal of Neurochemistry,
version (SCID-CV). Washington, DC: American Psychiatric Press, Inc. 69, 137−143.
Flannery, D. J., Vazsonyi, A. T., Torquati, J, & Fridrich, A. (1994). Ethnic Jang, K. L., Livesley, W. J., & Vernon, P. A. (1997). Gender-specific
and gender differences in risk for early adolescent substance use. etiological differences in alcohol and drug problems: A behavioural
Journal of Youth and Adolescence, 23, 195−213. genetic analysis. Addiction, 92, 1265−1276.
Fleming, C. B., Mason, W. A., Mazza, J., Abbott, R. D., & Catalano, R. F. Jessor, R., Donovan, J. E., & Costa, F. M. (1992). Health Behavior
(2008). Latent growth modeling of the relationship between depressive Questionnaire. Boulder, CO: University of Colorado Institute of
symptoms and substance use during adolescence. Psychology of Behavioral Science.
Addictive Behaviors, 22, 186−197. Johnson, S. W., & North, R. A. (1992). Opioids excite dopamine neurons by
Florsheim, P., Shiozaki, T., Hiraoka, R., Tiffany, S. T., Heavin, S., Hall, S., hyperpolarization of loval interneurons. Journal of Neuroscience, 12,
Teske, N., & Clegg, C. (2007). Craving among polysubstance-using 483−488.
adolescents. Journal of Child & Adolescent Substance Abuse, 17, Jönsson, E. G., Nöthen, M. M., Grünhage, F., Farde, L., Nakashima, Y.,
101−124. Propping, P., & Sedvall, G. C. (1999). Polymorphisms in the
Grandy, D. K., Litt, M., Allen, L., Bunzow, J. R., Marchionni, M., Makam, H., dopamine D2 receptor gene and their relationships to striatal dopa-
Reed, L., Magenis, R. E.., & Civelli, O. (1989). The human dopamine D2 mine receptor density of healthy volunteers. Molecular Psychiatry,
receptor gene is located on chromosome 11 at q22-q23 and identifies a TaqI 4, 290−296.
RFLP. American Journal of Human Genetics, 45, 778−785. Kaasinen, V., Nägren, K., Hietala, J., Farde, L., & Rinne, J. O. (2001). Sex
Grandy, D. K., Zhang, Y., & Civelli, O. (1993). PCR detection of the Taq A differences in extrastriatal dopamine D2-like receptors in the human
RFLP at the DRD2 locus. Human Molecular Genetics, 2, 2197. brain. American Journal of Psychiatry, 158, 308−311.
Grant, B. F., & Dawson, D. A. (1998). Age at onset of drug use and its Kashdan, T. B., Vetter, C. J., & Collins, R. L. (2005). Substance use in
association with DSM-IV drug abuse and dependence: Results from the young adults: Associations with personality and gender. Addictive
National Longitudinal Alcohol Epidemiologic Survey. Journal of Behaviors, 30, 259−269.
Substance Abuse, 10, 163−173. King, S. M., Iacono, W. G., & McGue, M. (2004). Childhood externalizing
Guindalini, C., Howard, M., Haddley, K., Laranjeira, R., Collier, D., Ammar, and internalizing psychopathology in the prediction of early substance
N., Craig, I., O'Gara, C. O., Bubb, V. J., Greenwood, T., Kelsoe, J., use. Addiction, 99, 1548−1559.
Asherson, P., Murray, R. M., Castelo, A., Quinn, J. P., Vallada, H., & Kirisci, L., Tarter, R. E., Vanyukov, M., Martin, C., Mezzich, A., & Brown,
Breen, G. (2006). A dopamine transporter gene functional variant S. (2006). Application of item response theory to quantify substance use
associated with cocaine abuse in a Brazilian sample. Proceedings of the disorder severity. Addictive Behaviors, 31, 1035−1049.
National Academy of Sciences of the United States of America, 103, Klein, T. A., Neumann, J., Reuter, M., Hennig, J., von Cramon, D. Y., &
4552−4557. Ullsperger, M. (2007). Genetically determined differences in learning
Han, C., McGue, M. K., & Iacono, W. G. (1999). Lifetime tobacco, from errors. Science, 318, 1642−1645.
alcohol, and other substance use in adolescent Minnesota twins: Köhnke, M. D., Batra, A., Kolb, W., Köhnke, A. M., Lutz, U., Schick, S., &
Univariate and multivariate behavioral genetic analyses. Addiction, 94, Gaertner, I. (2005). Association of the dopamine transporter gene with
981−993. alcoholism. Alcohol and Alcoholism, 40, 339−342.
Harlaar, N., Butcher, L. M., Meaburn, E., Sham, P., Craig, I. W., & Plomin, Lachman, H., Papolos, D. F., Saito, T., Yu, Y. M., Szumlanski, C. L., &
R. (2005). A behavioural genomic analysis of DNA markers associated Weinshilboum, R. M. (1996). Human catechol-O-methyltransferase
with general cognitive ability in 7-year-olds. Journal of Child pharmacogenetics: Description of a functional polymorphism and its
Psychology and Psychiatry, 46, 1097−1107. potential application to neuropsychiatric disorders. Pharmacogenetics,
Heath, A. C., Bucholz, K. K., Madden, P. A. F., Dinwiddie, S. H., Slutske, 6, 243−250.
W.S., Bierut, L. J., Rohrbaugh, J. W., Statham, D. J., Dunne, M. P., Lachman, H. M. (2006). An overview of the genetics of substance use
Whitfield, J. B., & Martin, N. G. (1997). Genetic and environmental disorders. Current Psychiatry Reports, 8, 133−143.
contributions to alcohol dependence risk in a national twin sample: Lemon, S. C., Roy, J., Clark, M. A., Friedmann, P. D., & Rakowski, W.
Consistency of findings in women and men. Psychological Medicine, 27, (2003). Classification and regression tree analysis in public health:
1381−1396. Methodological review and comparison with logistic regression. Annals
Hellemann, G., Conner, B. T., Anglin, M. D., & Longshore, D. (2009). of Behavioral Medicine, 26, 172−181.
Seeing the trees despite the forest: Applying recursive partitioning to the Lichter, J. B., Barr, C. L., Kennedy, J. L., Van Tol, H. H. M., Kidd, K. K.,
evaluation of drug treatment retention. Journal of Substance Abuse and & Livak, K. J. (1993). A hypervariable segment in the human
Treatment, 36, 59−64. dopamine receptor D4 (DRD4) gene. Human Molecular Genetics, 2,
Hill, S. Y., Hoffman, E. K., Zezza, N., Thalamuthu, A., Weeks, D. E., 767−773.
Mathews, A. G., & Mukhopadhyay, I. (2008). Dopaminergic mutations: Limosin, F., Gorwood, P., Loze, J. -Y., Dubertret, C., Gouya, L., Deybach,
Within-family association and linkage multiplex alcohol dependence J. -C., & Andes, J. (2002). Male limited association of the dopamine
families. American Journal of Medical Genetics Part B, Neuropsychi- receptor D2 gene TaqI A polymorphism and alcohol dependence.
atric Genetics, 147B, 517−526. American Journal of Medical Genetics, 112, 343−346.
Hirvonen, M., Laakso, A., Någren, K., Rinne, J. O., Pohjalainen, T., & Luby, J. L., Svrakic, D. M., McCallum, K., Przybeck, T. R., & Cloninger,
Hietala, J. (2004). C957T polymorphism of the dopamine D2 receptor C.R. (1999). The Junior Temperament and Character Inventory:
(DRD2) gene affects striatal DRD2 availability in vivo. Molecular Preliminary validation of a child self-report measure. Psychological
Psychiatry, 9, 1060−1061. Reports, 84, 1127−1138.
B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190 189

Lusher, J. M., Chandler, C., & Ball, D. (2001). Dopamine D4 receptor gene Robins, L. N., & Przybeck, T. R. (1991). Age of onset of drug use as a factor
(DRD4) is associated with novelty seeking (NS) and substance abuse: in drug and other disorders. In C. L. Jones, & R. J. Battjes (Eds.),
The saga continues. Molecular Psychiatry, 6, 497−499. Etiology of drug abuse: Implications for prevention. National Institute
Mâsse, L. C., & Tremblay, R. E. (1997). Behavior of boys in kindergarten on Drug Abuse Research Monograph 56 (pp. 178−192). Rockville, MD:
and the onset of substance use during adolescence. Archives of General U.S. Department of Health and Human Services.
Psychiatry, 54, 62−68. Sáiz, P. A., G-Portilla, M. P., Paredes, B., Delgado, J., Martínez, S.,
McGeary, J. E., Esposito-Smythers, C., Spirito, A., & Monti, P. M. (2007). Bascarán, M. T., & Bobes, J. (2003). Use of cocaine by secondary school
Association of the dopamine D4 receptor gene VNTR polymorphism students in northern Spain. European Addiction Research, 9, 138−143.
with drug use in adolescent psychiatric inpatients. Pharmacology, Samochowiec, J., Kucharska-Mazur, J., Grzwacz, A., Jabłoński, M.,
Biochemistry, and Behavior, 86, 401−406. Rommelspacher, H., Samochowiec, A., Sznabowicz, M., Horodnicki,
Melis, M., Spiga, S., & Diana, M. (2005). The dopamine hypothesis of drug J., Sagan, L., & Pelka-Wysiecka, J. (2006). Family-based and case-
addiction: Hypodopaminergic state. International Review of Neurobiology, control study of DRD2, DAT, 5HTT, COMT genes polymorphisms in
63, 101−154. alcohol dependence. Neuroscience Letters, 410, 1−5.
Merenäkk, L., Harro, M., Kiive, E., Laidra, K., Eensoo, D., Allik, J., Scheier, L. M., Botvin, G. J., & Miller, N. L. (1999). Life events, neighborhood
Oreland, L., & Harro, J. (2003). Association between substance use, stress, psychosocial functioning, and alcohol use among urban minority
personality traits, and platelet MAO activity in preadolescents and youth. Journal of Child & Adolescent Substance Abuse, 30, 19−50.
adolescents. Addictive Behaviors, 28, 1507−1514. Schoots, O., & Van Tol, H. H. M. (2003). The human dopamine D4 receptor
Michelhaugh, S. K., Fiskerstrand, C., Lovejoy, E., Bannon, M. J., & Quinn, repeat sequences modulate expression. Pharmacogenomics Journal, 3,
J. P. (2001). The dopamine transporter gene (SLC6A3) variable number 343−348.
of tandem repeats domain enhances transcription in dopamine neurons. Shoal, G., & Giancola, P. (2001). Cognition, negative affectivity, and
Journal of Neurochemistry, 79, 1033−1038. substance use in adolescent boys with and without a family history of a
Moos, R. H., & Moos, B. S. (1994). Life stressors and social resources substance use disorder. Journal of Studies on Alcohol, 62, 675−686.
inventory: Youth form manual. Odessa, FL: Psychological Assessment Song, J., Koller, D. L., Foroud, T., Carr, K., Zhao, J., Rice, J., Nurnberger, J. I.,
Resources. Jr., Begleiter, H., Porjesz, B., Smith, T. L., Schuckit, M. A., & Edenberg,
Mutirangura, A., Ledbetter, S. A., Kuwano, A., Chinault, A. C., & Ledbetter, H. J. (2003). Association of GABAA receptors and alcohol dependence
D. H. (1992). Dinucleotide repeat polymorphism at the GABAA receptor and the effects of genetic imprinting. American Journal of Medical
β3 (GABRB3) locus in the Angelman/Prader-Willi region (AS/PWS) of Genetics Part B, Neuropsychiatric Genetics, 117B, 39−45.
chromosome 15. Human Molecular Genetics, 1, 67. Spanagel, R., & Weiss, F. (1999). The dopamine hypothesis of reward: Past
Namkoong, K., Cheon, K. -A., Kim, J. -W., Jun, J. -Y., & Lee, J. -L. (2008). and current status. Trends in Neurosciences, 22, 521−527.
Association study of dopamine D2, D4 receptor gene, GABAA receptor Spear, L. P. (2000). The adolescent brain and age-related behavioral
β subunit gene, serotonin transporter gene polymorphism with children manifestations. Neuroscience and Biobehavioral Reviews, 24, 417−463.
of alcoholics in Korea: A preliminary study. Alcohol, 42, 77−81. Steinberg, L. (2008). A social neuroscience perspective on adolescent risk-
NCI-NHGRI Working Group on Replication in Association Studies, Chanock, taking. Developmental Review, 28, 78−106.
S. J., Manolio, T., Boehnke, M., Boerwinkle, E., Hunter, D.J., Thomas, G., Taylor, J. (2006). Life events and peer substance use and their relation to
Hirschhorn, J. N., Abecasis, G., Altshuler, D., Bailey-Wilson, J. E., substance use problems in college students. Journal of Drug Education,
Brooks, L. D., Cardon, L. R., Daly, M., Donnelly, P., Fraumeni, J. F., Jr, 36, 179−191.
Freimer, N. B., Gerhard, D. S., Gunter, C., Guttmacher, A. E., Guyer, M. Tcheremissine, O. V., Lane, S. D., Cherek, D. R., & Pietras, C. J. (2003).
S., Harris, E. L., Hoh, J., Hoover, R., Kong, C. A., Merikangas, K. R., Impulsiveness and other personality dimensions in substance use
Palmer, L. J., Phimister, E. G., Rice, J. P., Roberts, J., Rotimi, C., Tucker, disorders and conduct disorders. Addictive Disorders & Their
M. A., Vogan, K. J., Wacholder, S., Wijsman, E. M., Winn, D. M., & Treatment, 2, 1−7.
Collins, F. S. (2007). Replicating genotype–phenotype associations. Theile, J. W., Morikawa, H., Gonzales, R. A., & Morrisett, R. A. (2008).
Nature, 447, 655−660. Ethanol enhances GABAergic transmission onto dopamine neurons in
Nestler, E. J. (2000). Genes and addiction. Nature Genetics, 26, 277−281. the ventral tegmental area of the rat. Alcohol: Clinical & Experimental
Neville, M. J., Johnstone, E. C., & Walton, R. T. (2004). Identification and Research, 32, 1040−1048.
characterization of ANKK1: A novel kinase gene closely linked to Therneau, T. M., & Atkinson, B. (2006). The rpart package. Retrieved
DRD2 on chromosome band 11q23.1. Human Mutation, 23, 540−545. October 6, 2006, from University of California, Los Angeles
Noble, E. P. (2003). D2 dopamine receptor gene in psychiatric and Statistics Department Comprehensive R Archive Network http://
neurological disorders and its phenotypes. American Journal of Medical cran.stat.ucla.edu.
Genetics Part B, Neuropsychiatric Genetics, 116B, 103−125. Vandenbergh, D. J., Bennett, C. J., Grant, M. D., Strasser, A. A.,
Noble, E. P., Blum, K., Ritchie, T., Montgomery, A., & Sheridan, P. J. O'Connor, R., Stauffer, R. L., Vogler, G. P., & Kozlowski, L. T.
(1991). Allelic association of the D2 dopamine receptor gene with (2002). Smoking status and the human transporter variable number of
receptor binding characteristics in alcoholism. Archives of General tandem repeats (VNTR) polymorphism: Failure to replicate and finding
Psychiatry, 48, 648−654. that never-smokers may be different. Nicotine & Tobacco Research, 4,
Noble, E. P., Zhang, X., Ritchie, T., Lawford, B. R., Grosser, S. C., Young, 333−340.
R.M., & Sparkes, R. S. (1998). D2 dopamine receptor and GABAA Vandenbergh, D. J., Persico, A. M., Hawkins, A. L., Griffin, C. A., Li, X.,
receptor β3 subunit genes and alcoholism. Psychiatry Research, 81, Jabs, E. W., & Uhl, G. R. (1992). Human dopamine transporter gene
133−147. (DAT1) maps to chromosome 5p15.3 and displays a VNTR. Genomics,
Oak, J. N., Oldenhof, J., & Van Tol, H. H. M. (2000). The dopamine D4 14, 1104−1106.
receptor: One decade of research. European Journal of Pharmacology, Vandenbergh, D. J., Rodriguez, L. A., Miller, I. T., Uhl, G. R., & Lachman,
405, 303−327. H. M. (1997). High-activity catechol-O-methyltransferase allele is more
Pohjalainen, T., Rinne, J. O., Någren, K., Lehikoinen, P., Anttila, K., prevalent in polysubstance abusers. American Journal of Medical
Syvälahti, E. K. G., & Hietala, J. (1998). The A1 allele of the human D2 Genetics Part B, Neuropsychiatric Genetics, 74, 439−442.
dopamine receptor gene predicts low D2 receptor availability in healthy Volkow, N. D., Wang, G. -J., Begleiter, H., Porjesz, B., Fowler, J. S.,
volunteers. Molecular Psychiatry, 3, 256−260. Telang, F., Wong, C., Ma, Y., Logan, J., Goldstein, R., Alexoff, D., &
Pohjalainen, T., Rinne, J. O., Nägren, K., Syvälahti, E., & Hietala, J. (1998). Thanos, P. K. (2006). High levels of dopamine D2 receptors in
Sex differences in the striatal dopamine D2 receptor binding unaffected members of alcoholic families: Possible protective factors.
characteristics in vivo. American Journal of Psychiatry, 155, 768−773. Archives of General Psychiatry, 63, 999−1008.
190 B.T. Conner et al. / Journal of Substance Abuse Treatment 38 (2010) 178–190

Wang, T., Franke, P., Neidt, H., Cichon, S., Knapp, M., Lichtermann, D., Yang, B. -Z., Kranzler, H. R., Zhao, H., Gruen, J. R., Luo, X., &
Maier, W., Propping, P., & Nöthen, M. M. (2001). Association study of Gelernter, J. (2008). Haplotypic variants in DRD2, ANKK1, TTC12,
the low-activity allele of catechol-O-methyltransferase and alcoholism and NCAM1 are associated with comorbid alcohol and drug
using a family-based approach. Molecular Psychiatry, 6, 109−111. dependence. Alcoholism: Clinical and Experimental Research, 32,
White, F. J. (1996). Synaptic regulation of mesocorticolimbic dopamine 2117−2127.
neurons. Annual Review of Neuroscience, 19, 405−436. Zaykin, D. V., & Young, S. S. (2005). Large recursive partitioning analysis
Wills, T. A., Vaccaro, D., & McNamara, G. (1992). The role of life events, of complex disease pharmacogenetic studies. II. Statistical considera-
family support, and competence in adolescent substance use: A test of tions. Pharmacogenomics, 6, 77−89.
vulnerability and protective factors. American Journal of Community Zhang, H, & Singer, B. (1999). Recursive partitioning in the health
Psychology, 20, 349−374. sciences. New York: Springer-Verlag.

Anda mungkin juga menyukai