Content uploaded by Laurie Chassin
Author content
All content in this area was uploaded by Laurie Chassin
Content may be subject to copyright.
The Roles of Familial Alcoholism and Adolescent Family Harmony in
Young Adults’ Substance Dependence Disorders:
Mediated and Moderated Relations
Qing Zhou, Kevin M. King, and Laurie Chassin
Arizona State University
This study examined the prospective relations among family history density of alcoholism (FHD),
adolescent family harmony, and young adults’ alcohol and drug dependence. Family harmony was rated
by mothers and fathers in adolescence, and young adults’ substance dependence diagnoses were obtained
through structured interviews. Higher FHD predicted lower adolescent family harmony, which in turn
increased young adults’ odds of being diagnosed with drug dependence (with and without alcohol
dependence) compared to no diagnoses or to alcohol dependence only. Family harmony also interacted
with FHD such that the protective effect of family harmony on young adults’ drug dependence with or
without alcohol dependence decreased as FHD rose, and was nonsignificant at high levels of FHD. The
findings suggest the importance of distinguishing among alcohol and drug dependence disorders and
examining their differential etiological pathways, and also suggest that the protective effects of harmo-
nious family environments on substance dependence may be limited at high levels of FHD.
Keywords: alcoholism, family harmony, substance use disorders
The high prevalence of substance use disorders (SUDs) that
occur in young adulthood (Newman et al., 1996) makes this an
important developmental period for studying the etiology of sub-
stance dependence. Although previous research indicates that char-
acteristics of the family in childhood and adolescence can elevate
or reduce risk for later SUDs (see reviews by Chassin, Ritter, Trim,
& King, 2003; Hawkins, Catalano, & Miller, 1992), few studies
have examined the mediational and moderational mechanisms
underlying these links. Moreover, despite the importance of dis-
tinguishing between alcohol and drug use outcomes (Chassin,
Flora, & King, 2004; McGue, Slutske, & Iacono, 1999), few
studies have examined differential etiological pathways to alcohol
versus drug use disorders or their combination.
The present study tested whether family harmony in adoles-
cence mediated the relation between familial alcoholism and
young adults’ alcohol and drug dependence disorders, and whether
the relation between adolescent family harmony and later sub-
stance dependence was moderated by the density of familial
alcoholism.
Familial Alcoholism and Young Adults’ Substance
Dependence Disorders
Longitudinal studies have shown that having alcoholic parent(s)
creates significant risk for substance use problems among off-
spring (e.g., Chassin, Pitts, DeLucia, & Todd, 1999; Sher, Wal-
itzer, Wood, & Brent, 1991). Moreover, individuals with high
family history density of alcoholism (FHD) are at especially high
risk for substance use problems (Curran et al., 1999; Johnson &
Pickens, 2001; Stoltenberg, Mudd, Blow, & Hill, 1998; Windle,
1996). However, despite relatively strong evidence for parental
and familial alcoholism as a risk factor for SUDs, there is still
much to know about the processes underlying this risk, which may
include genetic and environmental pathways (McGue, 1999). Re-
cently, Walden, McGue, Iacono, Burt, and Elkins (2004) found
that two environmental factors—parent– child relationships and
peer deviance—accounted for more than 70% of the variance in
early substance use, which increases risk for later SUDs (Hawkins
et al., 1992). Therefore, environmental liabilities associated with
growing up in families with alcoholic members may increase risk
for SUDs (Jacob et al., 2003), and thus mediate the associations
between familial alcoholism and offspring substance use.
Adolescent Family Harmony as a Mediator in the
Relation Between Familial Alcoholism and Young
Adults’ Substance Dependence Disorders
Family environmental factors that have been linked to off-
spring’s substance use-related problems include parental social
support (Wills, Resko, Ainette, & Mendoza, 2004), parental mon-
itoring or discipline (Chassin, Curran, Hussong, & Colder, 1996;
King & Chassin, 2004), and family structure (Eitle, 2005). Several
family environment factors have also been found to mediate the
effect of parental or familial alcoholism on offspring’s substance
use problems, including parental monitoring (Chassin et al., 1996)
and parental discipline (King & Chassin, 2004). Family dishar-
mony, characterized by high levels of conflict (including interpa-
rental and parent– child conflict, or family conflict in general), may
be an additional environmental mechanism by which familial
Qing Zhou, Kevin M. King, and Laurie Chassin, Department of Psy-
chology, Arizona State University.
This work was supported by NIDA Grant DA05227 from the National
Institute on Drug Abuse to Laurie Chassin. The authors thank Kate Morse
and Pam Schwartz for coordinating the data collection and David B. Flora
and Jenn-Yun Tein for consultation on these analyses.
Correspondence concerning this article should be addressed to Laurie
Chassin, Department of Psychology, Arizona State University, P. O. Box
871104, Tempe, AZ 85287-1104. E-mail: Laurie.Chassin@asu.edu
Journal of Abnormal Psychology Copyright 2006 by the American Psychological Association
2006, Vol. 115, No. 2, 320 –331 0021-843X/06/$12.00 DOI: 10.1037/0021-843X.115.2.320
320
alcoholism influences SUDs in offspring. Family disharmony may
promote children’s aggressive or disruptive behaviors and deficits
in emotion regulation (Cummings & Davies, 1996). Family con-
flict may also interfere with effective parenting, which affects risk
for later substance abuse and dependence (Cummings & Davis,
1996; Patterson, DeBaryshe, & Ramsey, 1989; Sher, 1991). Con-
sistent with these theories, both concurrent and prospective rela-
tions have been found between family disharmony and substance
use problems among offspring (e.g., Guo, Hill, Hawkins, Catalano,
& Abbott, 2002; Sher, Gershuny, Peterson, & Raskin, 1997; Wills,
Sandy, Yaeger, & Shinar, 2001). Moreover, the link between
familial alcoholism and family disharmony has also been sup-
ported. Familial alcoholism is related to marital conflict (Heyman,
O’Leary, & Jouriles, 1995), negative communications among fam-
ily members (Jacob, Leonard, & Haber, 2001), and parent– child
conflict (El-Sheikh & Flanagan, 2001). Therefore, we posited a
mediational pathway in which familial alcoholism lowers family
harmony, which in turn increases risk for SUDs.
Only a few researchers have directly examined the mediating
role of family disharmony in the link between familial alcoholism
and developmental outcomes in childhood through young adult-
hood, and they have not studied SUDs as an outcome. For exam-
ple, family disharmony mediated risk for difficulty in leaving
home for children of alcoholics in the transition from adolescence
to young adulthood (Hussong & Chassin, 2002). Family cohesion
and adaptability mediated the link between parental problem
drinking and school-age children’s adjustment problems (El-
Sheikh & Buckhalt, 2003; El-Sheikh & Flanagan, 2001). Family
disharmony in toddlerhood mediated the link between parent an-
tisocial behavior (but not parent alcohol problems) and boys’ later
externalizing problems (Loukas, Fitzgerald, Zucker, & von Eye,
2001). Thus, family disharmony mediates familial alcoholism ef-
fects on a range of negative offspring outcomes, but this has not
been tested for young adult SUDs. The current study provides the
first prospective test of this mediating relation.
Differential Prediction of Alcohol Dependence, Drug
Dependence, and Their Combination
Most research on young adults’ SUDs either examines alcohol
or drug use outcomes in isolation or aggregates them into an
overall index of “substance use disorder.” However, recent work
suggests that alcohol and drug use disorders and their combination
may have distinct antecedents. For example, drug disorders, with
or without alcohol disorders, have been associated with the
externalizing-spectrum problems including conduct problems and
antisocial behavior (Krueger et al., 2002; Taylor, Iacono, &
McGue, 2000; Taylor, Malone, Iacono, & McGue, 2002). Com-
pared to alcohol disorders, drug disorders tend to be more strongly
related to behavioral undercontrol (disinhibition, McGue et al.,
1999; impulsivity and low agreeableness, Chassin et al., 2004),
whereas alcohol disorders in the absence of drug disorder tend to
be more strongly related to negative emotionality (McGue et al.,
1999) and neuroticism (Chassin et al., 2004). Moreover, Chassin et
al. (2004) found that impulsivity mediated the links between
parental alcoholism and drug dependence with or without alcohol
dependence, but not the link between parental alcoholism and
alcohol dependence alone.
Thus, previous research suggests that drug problems (compared
to alcohol problems in the absence of drug problems) are more
closely tied to behavioral undercontrol and externalizing-spectrum
problems such as conduct disorder and aggression. Given these
findings, the family factors that are thought to produce behavioral
undercontrol and externalizing behaviors may also be more closely
related to drug problems than to alcohol problems in the absence
of drug problems. Sher (1991) described a deviance proneness
pathway to SUDs by which dysfunctional family environments
interact with difficult temperament to produce behavioral under-
control, externalizing behaviors, school failure, and deviant peer
affiliations. Similarly, Patterson’s (Patterson, 1982; Patterson et
al., 1989) developmental model of antisocial behavior suggests
that dysfunctional family processes, including family conflict,
produce children’s conduct problems, which in turn lead to rejec-
tion by normal peers, school failure, and involvement in deviant
peer groups whose norms promote antisocial behaviors. Given
these theories, and the link between family conflict and offspring’s
behavioral undercontrol and conduct problems (Grych & Fincham,
1990), we hypothesized that family conflict would be more pre-
dictive of young adults’ drug dependence with or without alcohol
dependence than of alcohol dependence in the absence of drug
dependence, and that the mediational role of family harmony
would be specific to the link between familial alcoholism and drug
dependence disorders (with or without alcohol dependence).
To our knowledge, the current study is the first to examine the
differential prediction of alcohol versus drug dependence and their
combination from family environment characteristics. Moreover,
because personality has been shown to be an important predictor of
these differential diagnoses, we also adjusted for the effects of
personality when testing for these relations. Finally, because de-
velopmental pathways to substance use may be moderated by
gender or ethnicity (e.g., Baker & Yardley, 2002; Chassin & Ritter,
2001; Mahaddian, Newcomb, & Bentler, 1988), we included gen-
der and ethnicity as covariates in the models, and tested the
interactions between gender and ethnicity and our predictors.
Is the Relation Between Adolescent Family Harmony and
Young Adults’ Substance Dependence Disorder
Moderated by Family History Density of Alcoholism?
It is also possible that the same environmental characteristic
(e.g., family harmony) may have a different impact on the off-
spring’s SUDs depending on the density of familial alcoholism.
Several studies have found that familial SUD interacts with envi-
ronmental factors in predicting children’s and adolescents’ adjust-
ment (El-Sheikh & Buckhalt, 2003; El-Sheikh & Flanagan, 2001)
or substance use problems (Legrand, McGue, & Iacono, 1999).
However, it is unclear whether these interactions differentially
affect alcohol versus drug disorder. Moreover, there are important
differences in the forms of the interactions that have been found.
In a classic stress-buffering interaction (Rutter, 1990), a protective
factor (e.g., family harmony) buffers the response to risk such that
in the presence of a risk factor (e.g., familial alcoholism) individ-
uals with higher levels of a protective factor exhibit better out-
comes than do individuals with lower levels of the protective
factor. Consistent with this type of interaction, El-Sheikh and
Buckhalt (2003) found that the positive effect of high family
cohesion and adaptability on children’s adjustment problems was
stronger among families with more rather than fewer parental
drinking problems. Similarly, Legrand et al. (1999) found that the
positive effect of a low-risk peer environment on male adoles-
321
SUBSTANCE DEPENDENCE DISORDERS IN YOUNG ADULTS
cents’ substance use was stronger at higher than at lower levels of
familial SUD. Assuming that the protective factor is modifiable,
classic stress-buffering interactions support the importance of in-
terventions that increase levels of the protective factor.
In contrast, findings from other studies (including findings
based on the current sample) suggest a different pattern of inter-
action, which has been termed “protective but reactive” (Luthar,
Cicchetti, & Becker, 2000, p. 5). In these cases, although the
protective factor buffers against risk, the effect of the protective
factor becomes weaker at higher levels of risk (compared to lower
levels of risk). For example, King and Chassin (2004) found that
the positive effect of parental support on drug disorder was weaker
among individuals with higher rather than lower levels of behav-
ioral undercontrol. Similarly, El-Sheikh and Flanagan (2001)
found that the positive effect of low parent– child conflict on
children’s internalizing problems was weaker in families with
higher rather than lower parental problem drinking. For preventive
interventions, protective but reactive interactions point to potential
limitations of interventions aimed at increasing protective factors,
because their effectiveness may be reduced at high levels of risk.
Thus, the present study tested whether the relation between ado-
lescent family harmony and young adults’ substance dependence
disorder was moderated by FHD of alcoholism, showing either
classic stress buffering or a protective but reactive interaction.
Method
Participants
Participants were from an ongoing study of parental alcoholism (Chassin
et al., 2004; Chassin, Rogosch, & Barrera, 1991). At Time 1, there were
454 adolescents (M age ⫽ 13.2 years, range ⫽ 10.5–15.5), 246 of whom
had at least one biological alcoholic parent who was also a custodial parent
(children of alcoholics or COAs) and 208 demographically matched ado-
lescents with no biological or custodial alcoholic parents (control partici-
pants). At a young adult follow-up (Time 4), full biological siblings were
included if they were in the age range of 18–26 (and all of these siblings
were again invited to participate at Time 5, five years later). A total of 327
siblings (78% of eligible participants) were interviewed at Time 4, while
350 siblings (83%) were interviewed at Time 5 (n ⫽ 378 interviewed at
either wave). The combined sample of original targets and their siblings
was n ⫽ 734 at Time 4 (M age ⫽ 21.1), n ⫽ 762 at Time 5 (M age ⫽ 26.6),
and n ⫽ 817 with at least one wave of measurement. Retention in young
adulthood was excellent, with 407 (90%) of the original target sample
interviewed at Time 4 and 411 (91%) interviewed at Time 5 (96% had data
at either time point). We use parent report data from the three annual
adolescent assessments (Time 1–Time 3), and self-report data from the two
five-year young adult follow-ups (Times 4 and 5).
Details of sample recruitment are reported elsewhere (Chassin, Barrera,
Bech, & Kossak-Fuller, 1992). COA families were recruited using court
records of DUI arrests, health maintenance organization wellness question-
naires, and community telephone screening. Parental lifetime Diagnostic
and Statistical Manual of Mental Disorders (3rd ed.; DSM–III; American
Psychiatric Association, 1980) alcohol abuse or dependence was confirmed
with a structured interview. Demographically matched control families
were recruited using telephone interviews. When a COA was recruited,
reverse directories were used to locate families in the same neighborhood.
Families were screened to match the COA participant in ethnicity, family
structure, target child’s age (within one year), and socioeconomic status
(SES), using the property value code from the reverse directory. Structured
interviews were used to confirm that neither parent met lifetime DSM–III
criteria for alcohol abuse or dependence.
A complete description of sample representativeness is reported else-
where (Chassin et al., 1991). The sample was unbiased with respect to
alcoholism indicators available in archival records (e.g., blood alcohol
levels recorded at the time of the arrest; see Chassin et al., 1992, for
details). Moreover, the alcoholic sample had rates of other psychopathol-
ogy similar to those that were reported for a community-dwelling alcoholic
sample (Helzer & Pryzbeck, 1988). However, those who refused partici-
pation were more likely to be Hispanic, suggesting some caution in
generalization to more diverse samples.
Procedure
Data were collected with computer-assisted interviews either at families’
homes or on campus, or by telephone for out-of-state participants. Inter-
viewers were unaware of the family’s group membership. Interviews
required one to three hours, and participants were paid up to $70 over the
waves. To encourage honest responding, we reinforced confidentiality with
a Department of Health and Human Services Certificate of Confidentiality.
Selection of the Current Subsample
A total of 732 participants from 393 families (84% of total sample; 90%
of Time 4–5 sample) had complete data on parental and grandparental
alcoholism, at least partial family harmony data from Times 1 to 3, and
substance dependence diagnosis information at Waves 4 or 5. Because
some research has shown that DSM–III–R diagnoses of substance abuse are
more ambiguous and less reliable than those of substance dependence
(Pollock, Martin, & Langenbucher, 2000, 56), participants who had a
lifetime diagnosis of alcohol or drug abuse (but not dependence) were
dropped from analyses, which resulted in a final sample of 678 young
adults (see Table 1 for the descriptive data). Compared to the original
targets, the siblings were older and more likely to be married at Time 5, but
no other differences were found. Compared to control participants, the
COAs had a lower proportion of non-Hispanic Caucasians, were less likely
to be married at Time 5, and were less likely to receive higher education.
We next compared those who were retained for analysis (n ⫽ 678) with
those who were dropped due to missing data or an abuse-only diagnosis
(n ⫽ 99). The two groups did not differ in gender, ethnicity, or familial
alcoholism. However, those who were dropped were more likely to have a
parent with antisocial personality disorder (16% of those dropped vs. 6%
of those retained,
2
⫽ 8.93, p ⬍ .01), and had mothers with slightly less
education (high school graduates vs. some postsecondary education, t ⫽
2.48, p ⬍ .05). Because these differences are small, some caution is
warranted in generalization.
Measures
The measures used in the current study were part of the larger interview
battery.
Adolescent family harmony. At Times 1, 2, and 3, mothers and fathers
reported on their perception of family harmony during the past three
months using the five-item family conflict scale from Bloom’s Family
Processes Scale (BFPS; Bloom, 1985). The items assess the extent to
which family members fought a lot, got angry, threw things, lost their
tempers, hit each other, and criticized each other (
␣
⫽ .68, .68, and .69 for
fathers’ reports at Times 1, 2, and 3, and .69, .73, and .70 for mothers’
reports). The items were reverse-scored such that a high score reflected
greater family harmony. The BFPS has been widely used in research and
its psychometric properties (including factor integrity) have been well
established (see Bloom & Naar, 1994, for a review). In two previous
studies of the current sample (Hussong & Chassin, 1997, 2002), this
measure was related to lowered likelihood of adolescent substance use
initiation, better young adult psychological adjustment (i.e., fewer exter-
nalizing and internalizing symptoms), and fewer difficulties in the leaving-
home transition.
We examined the factor structure of the measure using Mplus Version
2.12 (Muthe´n & Muthe´n, 2003) to estimate a confirmatory factor model
322
ZHOU, KING, AND CHASSIN
with three correlated latent factors: Family Harmony at T1, T2, and T3,
each indicated by fathers’ and mothers’ reports of harmony at each time
point. The measurement model fit the data well, ⌬
2
(df ⫽ 3, n ⫽ 454
families) ⫽ .24, p ⫽ .97, comparative fit index (CFI) ⫽ 1.00, root-mean-
square-error of approximation (RMSEA) ⫽ .000, standardized root-mean-
square residual (SRMR) ⫽ .004. All the model-estimated loadings were
significant in a positive direction (the standardized loadings ranged from
.50 to .91). Moreover, the three latent factors (i.e., Family Harmony at T1,
T2, and T3) were highly correlated (r ranged from .69 to .83). Based on
these results, we created a composite score of family harmony by first
averaging across fathers’ and mothers’ reports of family harmony within
time, and then averaging the computed scores across the three time points.
We also tested for invariance of this measurement model across the COA
versus non-COA groups in a two-group SEM that constrained the loadings,
correlations among latent factors, and correlations among error terms to be
invariant across the two groups. This constrained two-group model fit the
data well (
2
(df ⫽ 18, n ⫽ 246 and 208 for COA and non-COA families) ⫽
22.57, p ⫽ .21, CFI ⫽ .99, RMSEA ⫽ .033, SRMR ⫽ .073), indicating
that the measurement model was invariant across the COA versus non-
COA groups.
Parental alcoholism, parental psychopathology, and family history den-
sity of alcoholism. At Times 1 and 4, parent’s lifetime DSM–III diag-
noses of parent alcoholism (abuse or dependence), affective disorder (ma-
jor depression or dysthymia), and antisocial personality disorder were
assessed by direct interview using the Diagnostic Interview Schedule (DIS,
Version III; Robins, Helzer, Croughan, & Ratcliff, 1981). For noninter-
viewed parents (24% of fathers, 13% of mothers), lifetime alcoholism
diagnoses were established using Family History–Research Diagnostic
Criteria (FH–RDC) (Version 3, Endicott, Andreason, & Spitzer, 1975)
based on spouse’s report.
1,2
Diagnoses of grandparental alcoholism were established using FH–RDC
(Version 3, Endicott et al., 1975) based on parent report at Times 2 and 4.
Parents reported on each of the child’s four biological grandparents, with
acceptable agreement across reporters (pooled
⫽ .56). A grandparent was
considered to have a positive diagnosis of alcoholism if he or she met the
FH–RDC criterion based on reports from either parent; a grandparent had
a negative diagnosis if he or she did not meet FH–RDC criteria according
to reports from both parents (unless only one reporter was available).
FHD scores were created by considering alcoholism in both parents and
all four grandparents and assigning weights to alcoholic relatives based on
their familial relatedness (Stoltenberg et al., 1998; Zucker, Ellis, & Fitzger-
ald, 1994). Each nonalcoholic relative was given a score of 0. Each
alcoholic parent scored .50, and each alcoholic grandparent scored .25.
These scores were then summed (possible range ⫽ 0 –2). The overall mean
of .45 (SD ⫽ .39) indicated that participants averaged one alcoholic
relative, but they varied from zero (N ⫽ 195, FHD score ⫽ 0) to six
alcoholic relatives (N ⫽ 3, FHD score ⫽ 2).
As expected, COAs had higher FHD scores than did non-COAs (t ⫽
⫺35.17, p ⬍ .001, df ⫽ 558.27, see Table 1), and COAs also had greater
variation in FHD scores (SD
COA
⫽ .28, SD
non-COA
⫽ .18, Levene’s test for
1
Noninterviewed parents were considered not to meet criteria (except
for alcoholism, where FH–RDC criteria were used for diagnosis based on
spousal reports). This allowed us to include single-parent families, but it
underestimates the prevalence of parental psychopathologies other than
alcoholism, which could produce negatively biased estimates of their
effects. Note that such underestimates could not occur when the inter-
viewed parent met diagnostic criteria because in those cases parent psy-
chopathology was coded as present. Thus, these errors could occur only in
cases where the interviewed parent did not meet criteria and the noninter-
viewed parent would have. Given our high interview rates for parents, this
occurrence was not frequent. On the basis of data from our two-
interviewed-parent families, estimates of potential misclassification errors
were only 1% for antisocial personality diagnoses and 3% for depression.
Thus, misclassification error should not substantially affect the findings.
2
Because parental affective disorder and antisocial personality disorder
are also possible risk factors for offspring’s SUD (e.g., Chassin et al., 2003;
Hawkins et al., 1992), we also tested our mediation and moderation models
by adding Time 1 measures of parental affective disorder and antisocial
personality disorder as covariates. We also tested the current models
without FHD in the model to examine the specific effect of parental
affective disorder and antisociality. Results indicated that parental affective
disorder and antisocial personality disorder did not predict young adults’
substance dependence contrasts adjusting for the effects of other predictors
in the models, either with or without FHD in the models. Moreover, the
results of the mediational and moderational analyses were unchanged.
Therefore, parental antisocial personality and affective disorder were not
considered further.
Table 1
Demographic Characteristics of the Young Adult Sample
Total
(N ⫽ 678)
Original targets vs. siblings COAs vs. Non-COAs
Original targets
(n ⫽ 354)
Siblings
(n ⫽ 324)
COAs
(n ⫽ 332)
Non-COAs
(n ⫽ 346)
Demographics
% Females 46.6 45.8 47.5 45.8 47.4
% Non-Hispanic Caucasian 27.6 25.7 29.6 32.5* 22.8*
% COA 49.0 52.0 45.7 — —
Mean age at T5 26.6 25.6* 27.6* 26.5 26.6
% Married at T5 41.0 31.9* 43.5* 34.1* 47.4*
% Employed 91.1 93.2 88.8 92.2 90.2
% Completed a college degree 30.5 31.1 29.9 26.5* 34.4*
% Attended some college 61.2 61.3 61.0 55.4* 66.8*
Mean (SD) of FHD .45 (.39) .46 (.40) .43 (.39) .77 (.28)* .13 (.18)*
Mean (SD) of family harmony 3.43 (.57) 3.46 (.57) 3.40 (.58) 3.26 (.58)* 3.61 (.53)*
Ns for substance dependence
diagnosis groups
ALC group 137 68 69 87* 50*
DRUG group 44 25 19 25 19
ALC ⫹ DRUG group 102 55 47 72* 30*
NON group 395 206 189 148* 247*
Note. The * indicates significant difference between groups.
323
SUBSTANCE DEPENDENCE DISORDERS IN YOUNG ADULTS
equality of variances, p ⬍ .05). Moreover, there was minimal overlap
between COAs and non-COAs in their FHD scores. By definition, the
minimum FHD score for COAs was .50, and only 10.6% of non-COAs had
scores that reached or exceeded this level. Thus, the FHD score largely
maintained the distinction between COAs and non-COAs, but provided
additional variation within the COA group in terms of density of familial
alcoholism.
Young adults’ alcohol and drug dependence diagnoses. At Times 4
and 5, participants’ lifetime DSM–III–R alcohol and drug dependence
diagnoses were obtained with a computerized version of the DIS (CDIS–
III–R, Robins et al., 1981). Rates of lifetime dependence were 35.2% for
alcohol and 21.6% for drugs. As expected in a study that oversamples
individuals at high risk, these prevalences are higher than national data. For
example, National Comorbidity Survey participants aged 18 –25 showed
17.5% lifetime alcohol dependence and 9% lifetime drug dependence
(Kessler, 2002). For the current analyses, the outcome was a four-category
nominal variable: no diagnoses at either wave (henceforth called NON;
n ⫽ 395, 58.3%), alcohol dependence only at either wave (henceforth
called ALC; n ⫽ 137, 20.2%), drug dependence only at either wave
(henceforth called DRUG; n ⫽ 44, 6.5%), and both alcohol and drug
dependence disorders diagnosed at either wave (henceforth called
ALC⫹DRUG; n ⫽ 102, 15.0%).
3
To examine variation in severity of disorders as well as the possibility of
subclinical problems within the nondiagnosed group, we compared the
groups on their alcohol and drug dependence symptoms (the maximum
number of lifetime dependence symptoms at Time 4 or 5) using ANOVA.
The ALC⫹DRUG group reported the most alcohol (M ⫽ 3.59, SD ⫽ 1.94)
and drug dependence (M ⫽ 4.31, SD ⫽ 2.14) symptoms, significantly more
alcohol symptoms than the ALC group but not significantly more drug
symptoms than the DRUG group ( p ⬎ .05). The ALC group had more
alcohol dependence symptoms than did the DRUG group (1.22 vs. .69 p ⬍
.05), and the DRUG group had more drug dependence symptoms than did
the ALC group (2.61 vs. .45, p ⬍ .05). Thus, the ALC⫹DRUG group had
somewhat more severe alcohol dependence than did the ALC group, but
not more severe drug disorder than did the DRUG group.
Moreover, supporting the distinctiveness among the groups, the ALC
group averaged less than one drug symptom (.45) and did not significantly
differ from the NON group in their drug symptoms (.13, p .05). Con-
versely, the DRUG group averaged less than one alcohol symptom (.69)
and did not significantly differ from the NON group in their alcohol
symptoms (.26, p ⬎ .05). Finally, alcohol and drug symptoms in the NON
group did not significantly differ from zero (.26 for alcohol and .13 for
drugs), suggesting that they did not have subclinical levels of disorder.
Young adult personality. At Times 4 and 5, young adults self-reported
their personality characteristics (neuroticism, extraversion, agreeableness,
openness, conscientiousness) using the NEO Five-Factor Inventory (NEO–
FFI, Costa & McCrae, 1992). Self-reported personality was relatively
consistent over the two occasions (correlations from .53 to .63). Thus,
scores from the two waves were averaged unless one score was missing. In
that case the available wave of measurement was used. Internal consisten-
cies ranged from .72 to .86 across the scales and measurement waves.
Data Analytic Strategy
We used multilevel modeling with MIXNO software (Hedeker, 1999)
because of the nonindependence of the sibling data. To ensure that the
effects of the predictors were independent of gender and ethnicity as
covariates, we also tested the two-way and three-way interactions involv-
ing these covariates. However, because none of these interactions was
significant, they were dropped from the final regression models.
Mediated effects occur when, in the context of a theoretical rationale, the
relation between two variables can be explained, at least in part, by a path
through an intervening variable. This effect can be tested in two ways: by
testing the significance of the difference between the coefficients of the
predictor to the outcome with and without the mediator in the model
(Baron & Kenny, 1986) and by testing the significance of the product of the
coefficients for paths from the predictor to the mediator and the mediator
to the outcome. Recent work has shown that the Baron and Kenny method
is seriously underpowered compared to the product of coefficients method
(Mackinnon, Lockwood, Hoffman, West, & Sheets, 2002; Shrout &
Bolger, 2002). Thus, we used the regression coefficients from ordinary
least squares (OLS) regression (predicting family harmony from FHD) and
mixed effects multinomial logistic regressions (predicting substance de-
pendence diagnoses from harmony and FHD) to calculate the significance
of the indirect effect (FHD 3 family harmony 3 substance dependence)
using the techniques of MacKinnon and colleagues
4,5
(MacKinnon &
Dwyer, 1993; MacKinnon, Lockwood, & Williams, 2004).
We examined moderation by testing the significance of the FHD ⫻
Family Harmony interaction when added to our base model. When a
significant interaction was found, we probed the nature of the interactions
using simple slope analyses (Aiken & West, 1991). Furthermore, we tested
whether the mediated path (FHD 3 family harmony 3 substance depen-
dence) varied as a function of FHD level (high, medium, and low) by
conducting the simple mediational analyses using the methods of Tein,
Sandler, McKinnon, and Wolchik (2004).
Results
Sibling Correlations
To examine the effects of the nestedness due to sibling data, we
tested a mixed multinomial logistic regression model predicting
alcohol and drug dependence with a single predictor (gender) and
examined the intraclass correlation (ICC). There was a significant
effect of sibling relationship (ICC ⫽ 0.33) such that siblings were
3
Because the siblings were not assessed in adolescence, there is a
possibility that some SUDs already existed in adolescence and did not
precede the measurement of adolescent family environment. However, for
the original target sample, there were only 8 participants who met criteria
for alcohol or drug disorders in adolescence (using the Diagnostic Inter-
view for Children and Adolescents, Parent Version; Herjanic & Reich,
1982) and removing these participants did not influence the findings. Thus,
it is likely that the rates of SUDs for all participants in adolescence was
low, and it is unlikely that preexisting adolescent SUDs could explain the
current results.
4
Although each coefficient can be assumed to be normally distributed,
their product is not. Rather, the product tends to be skewed and highly
kurtotic (MacKinnon et al., 2004), which suggests that critical values from
the standard normal table will be incorrect. Thus traditional tests (e.g.,
Sobel, 1982) of the significance of the mediated effect are underpowered.
However, confidence limits and thus a test of significance can be obtained
from this asymmetric distribution by a method which has been shown to
have better power and Type I error rates than the Sobel approach to
significance testing (MacKinnon et al., 2004). Critical values from an
asymptotic distribution based on the product of two coefficients can be
obtained from Meeker, Cornwell, and Aroian (1981), and based on these
values, confidence intervals can be constructed. Significant mediated ef-
fects are indicated by confidence intervals not containing zero. Thus, for
the present analyses we tested the significance of the indirect effect using
this method, and we report asymmetric confidence intervals.
5
Because multinomial logistic regression uses maximum likelihood
estimation to produce estimates of coefficients, the scale of the outcome
across analyses differs and thus coefficients are not directly comparable.
Using coefficients from different analyses to compute the coefficient for
the indirect effect would incorrectly estimate the magnitude of the effect.
MacKinnon and Dwyer (1993) describe a method for standardizing coef-
ficients from logistic regression equations that allows the comparison of
coefficients across equations and thus properly computes the coefficient for
an indirect effect.
324
ZHOU, KING, AND CHASSIN
more similar than were nonsiblings. This supports our use of a data
analytic procedure that accounted for the nestedness of the data.
Family Harmony as a Mediator of Familial Alcoholism’s
Effects
We first tested the relation of FHD to family harmony using
OLS regression. After adjusting for ethnicity and gender, higher
FHD significantly predicted lower family harmony during adoles-
cence (

⫽⫺.41, p ⬍ .001, ⌬R
2
⫽ .11). Next, we estimated a
mixed-effects multinomial logistic regression predicting SUD di-
agnostic group from gender, ethnicity, FHD, and family harmony.
Results are shown in Table 2. Gender (but not ethnicity) predicted
substance dependence. Males had higher odds of being in the ALC
group or the ALC⫹DRUG group versus the NON group and males
had lower odds of being in the DRUG group or the ALC⫹DRUG
group versus the ALC group. Family harmony significantly de-
creased the odds of being in the DRUG group or the ALC⫹DRUG
group versus the NON group. Moreover, compared to the ALC
group, family harmony significantly decreased the odds of being in
the ALC⫹DRUG group and marginally decreased the odds of
being in the DRUG group.
In the same model (after adjusting for the effects of family
harmony), FHD remained a significant predictor of substance
dependence,
6
increasing the odds of being in the ALC group or the
ALC⫹DRUG group versus the NON group, and the odds of being
in the ALC⫹DRUG group versus the DRUG group. Moreover,
FHD marginally increased the odds of being in the DRUG versus
NON group and the ALC group (see Table 2).
We then used the coefficients from the OLS regression and
multinomial regression to test the significance of the indirect
(mediated) effects (i.e., FHD 3 family harmony 3 SUDs), for
each pairwise diagnostic group comparison. To test the signifi-
cance of the indirect effects, the logistic coefficients were stan-
dardized. Then, for each indirect effect, asymmetric upper and
lower confidence intervals were constructed (MacKinnon et al.,
2004). We also calculated the proportion of the total effect of FHD
on the outcome that was indirect through family harmony (Mac-
Kinnon, Warsi, & Dwyer, 1995; see Figure 1, in which significant
indirect effects are represented by bold lines).
There was a significant indirect effect of FHD through family
harmony for the contrasts of DRUG versus NON (upper confi-
dence limit (UCL) ⫽ .84, lower confidence limit (LCL) ⫽ .15,
32% mediated) and ALC⫹DRUG versus NON (UCL ⫽ .81,
LCL ⫽ .25, 20% mediated). The indirect effect of FHD through
harmony was also significant for the contrast between
ALC⫹DRUG and ALC (UCL ⫽ .57, LCL ⫽ .07, 91% mediated),
but not between ALC and DRUG.
In summary, a denser family history of alcoholism was associ-
ated with lower family harmony in adolescence, which in turn
increased young adults’ risk for drug dependence (with and with-
out alcohol dependence) compared to having no diagnosis, as well
as increased the risk for having both alcohol and drug dependence
diagnoses compared to alcohol dependence only.
To ensure that these findings were robust to different operation-
alizations of familial alcoholism and not an artifact of combining
the COA and control groups, we repeated the analyses using
parental alcoholism (COA vs. non-COA) in place of FHD scores
and obtained the same pattern of results. Specifically, adolescent
family harmony significantly mediated the relations between pa-
rental alcoholism and contrasts between the DRUG and the NON
groups (UCL ⫽ .71, LCL ⫽ .05, p ⬍ .01, 27% mediated), between
the ALC⫹DRUG and the NON groups (UCL ⫽ .69, LCL ⫽ .15,
6
Because parental drug problems may also account for the prediction of
offspring’s SUDs, it is important to consider their effects on the present
models. However, previous analyses of the same sample have shown that
parental drug problems in predicting patterns of alcohol and drug depen-
dence were nonsignificant over and above the effects of FHD (Chassin et
al., 2004). Therefore, parental drug problems were not included in the
regression models.
Table 2
Mixed-Effects Multinomial Logistic Regressions Predicting the Contrasts Among Substance
Dependence Diagnosis Groups From the Covariates, FHD, and Adolescent Family Harmony
Independent variables
ALC vs. NON DRUG vs. NON ALC⫹DRUG vs. NON
B Adjusted OR
a
B Adjusted OR
a
B Adjusted OR
a
Intercept ⫺1.23 .29 .63 1.88 .85 2.34
Gender 1.47*** 4.35 .30 1.35 .70* 2.01
Ethnicity ⫺.39 .68 ⫺.54 .58 ⫺.18 .84
FHD 1.85*** 6.36 1.00† 2.72 1.94*** 6.96
Family harmony ⫺.45 .64 ⫺.99** .37 ⫺1.05*** .35
Independent variables
DRUG vs. ALC ALC⫹DRUG vs. ALC ALC⫹DRUG vs. DRUG
B Adjusted OR
a
B Adjusted OR
a
B Adjusted OR
a
Intercept 2.32† 10.18 2.43* 11.36 .20 1.22
Gender ⫺1.27** .28 ⫺.88** .41 .45 1.57
Ethnicity ⫺.15 .86 .22 1.25 .42 1.52
FHD ⫺.92† .40 .02 1.02 .93† 2.53
Family harmony ⫺.60† .55 ⫺.63* .53 .03 1.03
a
Adjusted OR ⫽ adjusted odds ratio, or the odds ratio adjusted for the effects of other predictors in the regression
model. FHD ⫽ family history density of alcoholism.
† p ⬍ .10. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
325
SUBSTANCE DEPENDENCE DISORDERS IN YOUNG ADULTS
p ⬍ .01, 21% mediated), and between the ALC⫹DRUG and the
ALC groups (UCL ⫽ .48, LCL ⫽ .01, p ⬍ .01, 86% mediated).
Compared to the control participants, COA families had lower
harmony, which in turn increased young adults’ risk for drug
dependence disorders with or without alcohol dependence com-
pared to no diagnoses, as well as increased the risk for developing
both alcohol and drug dependence compared to alcohol depen-
dence only.
Personality as a Competitive Mediator
Although the present study focused on family harmony, person-
ality has also been shown to mediate the effect of family history on
diagnosis (Chassin et al., 2004). Accordingly, we repeated the
above models including personality factors as competitive medi-
ators to test whether the effects of family harmony were robust to
the effects of personality. To do this, we used a subsample of
participants (N ⫽ 647) with personality (NEO–FFI) data. As
Chassin et al. (2004) found, NEO–FFI agreeableness and neurot-
icism predicted substance dependence diagnoses (both p ⬍ .01)
and mediated the effect of FHD on dependence diagnoses (all p ⬍
.05, mediated effect ranged from 11% to 28%). After including
personality, the main effects of harmony on diagnosis were gen-
erally unchanged, reducing the magnitude but not the direction of
the mediational effects of adolescent family harmony. Family
harmony remained a significant mediator of the effects of FHD in
differentiating ALC⫹DRUG from NON ( p ⬍ .05, 16% mediated).
The effect of family harmony became marginally significant in
differentiating ALC⫹DRUG from ALC ( p ⬍ .10) and DRUG
from NON ( p ⬍ .10). Correspondingly, the indirect (mediated)
effect of harmony became marginally significant in distinguishing
ALC⫹DRUG from ALC ( p ⬍ .10, 41% mediated) and DRUG
from NON ( p ⬍ .10, 31% mediated).
The Interaction Between FHD and Adolescent Family
Harmony in Predicting Later SUDs
To examine whether the effect of family harmony on SUDs was
moderated by FHD, we added the two-way interaction (FHD ⫻
Family Harmony) to the multinomial regression models. FHD and
family harmony were mean centered (Aiken & West, 1991) to
reduce collinearity among the predictors (see Table 3 for results).
The FHD ⫻ Family Harmony interaction was significant in the
contrasts between DRUG and NON ( p ⬍ .05), and between
ALC⫹DRUG and NON ( p ⬍ .01, see Table 3). The form of these
interactions is shown in Figure 2. Adolescent family harmony was
significantly associated with decreased odds of being in the DRUG
group versus the NON group at both low (–1 SD) and medium
(mean) levels of FHD (simple slopes ⫽ –1.98 and –1.04, p ⬍ .001
and .004, adjusted odds ratio (OR) ⫽ .14 and .35 for low and mean
levels of FHD, respectively). However, as the density of familial
alcoholism increased, this relation weakened such that family
harmony was unrelated to the odds of DRUG versus NON for
those from families with high (⫹1 SD) FHD. The same pattern was
seen for the contrast between the ALC⫹DRUG group versus the
NON group. Family harmony significantly reduced the odds of
being in the ALC⫹DRUG group versus the NON group at low (–1
SD) and medium (mean) levels of FHD (simple slopes ⫽ –2.00
Figure 1. Adolescent family harmony as a mediator in the relations between FHD and young adults’ substance
dependence. The numbers presented are the unstandardized regression coefficients obtained from the OLS
regression predicting family harmony from FHD and the multinomial logistic regressions predicting the contrasts
among young adults’ substance dependence categories from FHD and family harmony (gender and ethnicity
were controlled in both regressions). The thick lines represent significant mediational pathways. *p ⬍ .05.
**
p ⬍
.01.
***
p ⬍ .001.
326
ZHOU, KING, AND CHASSIN
and ⫺1.20, p ⬍ .001, adjusted OR ⫽ .14 and .30). However, as the
density of familial alcoholism increased, this relation weakened
and became nonsignificant at high (⫹1 SD) levels of FHD. To
ensure that the lack of association between family harmony and
substance dependence at high FHD was not due to the restricted
variability in family harmony, we compared the variance in family
harmony among those with high versus low FHD and found no
significant differences (F for Levene’s test for equality of vari-
ances ⫽ .54, p ⫽ .46).
Because these interactions suggested that the effects of family
harmony on the DRUG versus NON and the ALC⫹DRUG versus
NON contrasts differed by levels of FHD, it is likely that the
strength of the indirect effect (FHD 3 family harmony 3 sub-
stance dependence) also differed as a function of FHD. To exam-
ine this hypothesis, we tested the simple mediated and indirect
effects at low (–1SD), medium (mean), and high (⫹1 SD) FHD
using the methods of Tein et al. (2004).
7
The indirect effect of FHD on the DRUG versus NON contrast
through family harmony was significant at low (–1 SD) FHD
(UCL ⫽ 1.79, LCL ⫽ .31, p ⬍ .01, 42% mediated) and medium
(mean) FHD (UCL ⫽ 1.03, LCL ⫽ .08, p ⬍ .01, 27% mediated).
However, the indirect effect became nonsignificant at high (⫹1
SD) FHD. Similarly, the indirect effect of FHD on the
ALC⫹DRUG versus NON contrast through family harmony was
significant at low (–1 SD) FHD (UCL ⫽ 1.66, LCL ⫽ .43, p⬍ .01,
42% mediated) and mean FHD (UCL ⫽ 1.03, LCL ⫽ .24, p ⬍ .01,
20% mediated), but nonsignificant at high (⫹1 SD) FHD.
As above, we repeated these moderational analyses with COA
status rather than FHD score as a predictor. Parental alcoholism
significantly interacted with family harmony in predicting the
contrast between the ALC⫹DRUG and NON group (B ⫽ 1.29,
p ⬍ .05). The simple slope analyses showed that family harmony
was more strongly associated with decreased odds of developing
both alcohol and drug dependence versus having no diagnosis in
nonalcoholic families (simple slope ⫽ –1.88, p ⬍ .001) than in
alcoholic families (simple slope ⫽ –.59, p ⫽ .09). The parental
alcoholism ⫻ family harmony interaction was marginally signifi-
cant for the DRUG versus NON contrast (b ⫽ 1.28, p ⬍ .10).
Family harmony was significantly associated with decreased odds
of developing drug dependence disorder only versus having no
diagnosis in nonalcoholic families (simple slope ⫽ –1.69, p ⬍
.001), but it was unrelated to the DRUG versus NON contrast in
alcoholic families (simple slope ⫽ –.40, p ⫽ .41).
Discussion
The present study examined family harmony as a mediator of
familial alcoholism effects on the development of drug versus
alcohol disorders in young adulthood. We tested the mediational
role of family harmony with and without adjusting for personality,
and we tested whether the relation between family harmony and
young adults’ SUDs was moderated by FHD.
Family Harmony as a Mediator
The first finding of note was that a higher density of familial
alcoholism was associated with lower family harmony during
7
In this method, the FHD variable was rescaled so that the zero point
corresponded to 1 SD below the mean (i.e., rescaled FHD ⫽ [FHD centered
at mean] – [–1 SD of FHD]) and created the FHD ⫻ family harmony
interaction term with the rescaled FHD. The models were then reestimated
with the covariates, family harmony, the rescaled FHD, and the interaction
term, and the main effect of family harmony (above and beyond other
predictors in the model) now corresponds to the simple effect of family
harmony at low (–1 SD) FHD. We then tested the significance of the simple
indirect effect. The above steps were then repeated with FHD centered at
the mean and at 1 SD above the mean.
Table 3
Mixed-Effects Multinomial Logistic Regressions Predicting the Contrasts Among Substance
Dependence Diagnosis Groups From the Covariates, FHD, Adolescent Family Harmony, and
FHD ⫻ Family Harmony Interaction
Independent variables
ALC vs. NON DRUG vs. NON ALC⫹DRUG vs. NON
B Adjusted OR
a
B Adjusted OR
a
B Adjusted OR
a
Gender 1.46*** 4.31 .32 1.38 .70 2.01
Ethnicity ⫺.41 .66 ⫺.59 1.80 ⫺.22 .80
FHD 1.92*** 6.82 1.34* 3.82 2.25*** 9.49
Family harmony ⫺.42 .66 ⫺1.03** 2.80 ⫺1.20*** .30
FHD ⫻ family harmony 1.01 2.75 2.38* 10.80 2.03** 7.61
Independent variables
DRUG vs. ALC ALC⫹DRUG vs. ALC ALC⫹DRUG vs. DRUG
B Adjusted OR
a
B Adjusted OR
a
B Adjusted OR
a
Gender ⫺1.25** .29 ⫺.87** .42 1.82** 6.17
Ethnicity ⫺.18 .83 .21 1.23 ⫺.11 1.12
FHD ⫺.61 .54 .29 1.34 .49 1.63
Family harmony ⫺.65† .52 ⫺.80** .45 1.27† 3.56
FHD ⫻ family harmony 1.35 3.86 .95 1.16 ⫺2.56 .08
a
Adjusted OR ⫽ adjusted odds ratio, or the odds ratio adjusted for the effects of other predictors in the regression
model. FHD ⫽ family history density of alcoholism.
† p ⬍ .10. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
327
SUBSTANCE DEPENDENCE DISORDERS IN YOUNG ADULTS
adolescence, which in turn increased risk for substance depen-
dence disorders in young adulthood. Thus, the effects of familial
alcoholism on young adults’ SUDS can be accounted for, in part,
by their earlier exposure to high conflict in their families. Impor-
tantly, this mediational path (i.e., FHD 3 family harmony 3
alcohol ⫹ drug dependence) remained significant after adjusting
for the effects of personality (although the mediational path to drug
dependence became only slightly weaker). This suggests that the
risks conferred by personality and family environment are unique
and additive so that both personality and family environment serve
to account for familial alcoholism effects on the development of
SUDs. These results are consistent with previous studies on the
mediating role of family harmony in the link between familial
alcoholism and adjustment (e.g., El-Sheikh & Buckhalt, 2003;
El-Sheikh & Flanagan, 2001; Hussong & Chassin, 2002; Loukas et
al., 2001) and extend these studies by demonstrating the effect of
family harmony on young adults’ substance dependence diag-
noses. However, because family harmony only partially mediated
familial alcoholism effects, other mechanisms are also necessary to
fully explain the relation between familial alcoholism and SUDs.
Differential Pathways From Familial Alcoholism to
Alcohol, Drug, and Combined Dependence
An important contribution of the present study is the distinction
between alcohol and drug dependence disorders and their combi-
nation in examining the mediational pathways. We found that
family harmony during adolescence decreased the risk of drug
Figure 2. The interactions between FHD and family harmony in predicting the DRUG versus NON and the
ALC⫹DRUG versus NON contrasts.
328
ZHOU, KING, AND CHASSIN
dependence with or without alcohol dependence compared to
either alcohol dependence only or no diagnosis. However, family
harmony did not differentiate young adults with alcohol depen-
dence only from those with no diagnoses. These results support the
suggestions by McGue et al. (1999), Taylor et al. (2000, 2002,
2002), and Chassin et al. (2004) that separate mechanisms may be
involved in the development of drug dependence (with and without
alcohol dependence) compared to alcohol dependence in the ab-
sence of drug dependence. The current findings extend those
studies by identifying family harmony as an additional differenti-
ating predictor.
Why should family harmony be more specifically linked to drug
dependence than to alcohol dependence in the absence of drug
dependence? One possibility rests with the illegal nature of drug
use and its link to behavioral undercontrol. Because young adults
with drug dependence (with or without alcohol dependence) nec-
essarily engage in illegal behavior, they may be more behaviorally
undercontrolled (i.e., more “deviance prone”) than individuals
with alcohol disorders alone. This is consistent with previous
findings that behavioral undercontrol was more strongly related to
drug use disorders than to alcohol use disorders alone (Chassin et
al., 2004; McGue et al., 1999). Given the large literature that
connects behavioral undercontrol and externalizing problems to
family conflict and dysfunctional family processes (Grych &
Fincham, 1990; Patterson et al., 1989), the results of the current
study further suggest that family processes may be especially
important to the development of drug use disorders with or without
alcohol use disorders because highly conflictual families are likely
to produce adolescents with higher levels of undercontrol and
externalizing problems. An alternative interpretation focuses not
on the illegal and deviance prone nature of drug use compared to
alcohol use, but rather on the relative severity of disorder in the
different diagnostic subgroups. Because those with both alcohol
and drug dependence tended to have the more severe disorders
than those with alcohol dependence alone, it may also be that
individuals who are behaviorally undercontrolled and who live in
conflictual families develop both a broader spectrum of SUDs and
more severe disorders.
In contrast, although familial alcoholism was associated with
heightened risk for alcohol dependence only compared to no
diagnoses, this relation was not mediated through family harmony
(see Figure 1). This suggests that those from alcoholic families
who develop alcohol dependence in the absence of drug depen-
dence may follow a different pathway—a pathway that is less
comorbid with externalizing or behavioral undercontrol, that is less
related to family harmony, and that leads to less severe alcoholism.
The pathway to alcohol dependence in the absence of drug disor-
der only may be more alcohol-specific, and may involve other
mediators such as a heritable sensitivity to alcohol effects.
The Interaction Between FHD and Family Harmony
Another important contribution of the current study is the dem-
onstration that the effects of family harmony on SUDs varied with
the density of familial alcoholism. Specifically, family harmony
decreased risk for developing drug dependence with or without
alcohol dependence only at low to moderate (but not high) levels
of FHD. Moreover, the mediated effects of FHD on drug depen-
dence (with or without alcohol dependence) through family har-
mony also disappeared at high levels of FHD.
The form of this interaction is what Luthar, Cicchetti, and
Becker (2000) called “protective but reactive” in that a protective
factor (such as a favorable family environment) generally provides
benefits, but the protective effect turns weaker at higher levels of
risk (such as FHD) (p. 5). A similar protective but reactive inter-
action was found between parenting and offspring temperament in
another study based on the same sample (King & Chassin, 2004),
where the protective effects of parental support on the risk for drug
use disorder disappeared at high levels of behavioral undercontrol.
Individuals with a strong diathesis for SUD (indicated either by
dense familial alcoholism or high levels of behavioral undercon-
trol) may be less influenced by qualities of the family environment
(e.g., parenting and family conflict) because that diathesis inhibits
bonding with the family, because other pathways may override
familial influences, or both.
It is interesting to speculate about why the current findings (and
other findings from this data set, King & Chassin, 2004) produce
interactions that are protective but reactive whereas some other
studies find classic buffering effects (e.g., El-Sheik & Buckhalt,
2003; Legrand et al., 1999). The different findings may reflect
differing degrees of risk across samples. That is, samples with
greater representation of the highest risk levels (e.g., denser alco-
holic pedigrees, more undercontrolled participants) may produce
protective but reactive interactions whereas lower-risk samples
may produce classic buffering effects. Consistent with this inter-
pretation, using a clinical sample of young children, Wootton,
Frick, Shelton, and Silverthorn (1997) found a protective but
reactive interaction such that the protective effects of effective
parenting against conduct problems diminished among children
with high personality risk. In contrast, studies that did not over-
sample high-risk individuals (Legrand et al., 1999; Stice & Gonza-
les, 1998) have found classic buffering interactions. Interestingly,
studies of a sample in which COAs were selected using a less strict
criterion for parental alcoholism (i.e., the Michigan Alcoholism
Screening Test) than the current DSM diagnoses have reported
both the classic buffering and protective but reactive interactions
(El-Sheikh & Buckhalt, 2003; El-Sheikh & Flanagan, 2001). How-
ever, the sampling difference interpretation is speculative, and
other method or design differences (e.g., differing ages of the
participants, differing outcome variables, and differing protective
factors) may be responsible for differences in findings across
studies.
Finally, although the current study makes an important contri-
bution by providing the first prospective test of the role of family
harmony as a differential mediator of familial alcoholism effects
on drug and alcohol dependence, it is also necessary to consider
some of its limitations. First, our measure of FHD included only
grandparents and parents, and future research could expand the
measure to include more relatives. Second, different diagnostic
criteria might produce different findings, and our use of RDC
criteria might underestimate rates of grandparental disorder,
whereas DSM–III–R criteria may overdiagnose SUDS. Third, our
community sample of alcoholic families had low rates of comorbid
antisocial personality disorder, and different findings might be
produced in samples with more familial antisociality. Fourth, our
findings may be specific to SUDs, and not necessarily similar for
other mental health problems. Finally, we assessed SUDs in young
adulthood, and the relations among family harmony, FHD, and
substance dependence diagnoses may differ at different ages. For
example, at younger ages when alcohol use is more uncommon
329
SUBSTANCE DEPENDENCE DISORDERS IN YOUNG ADULTS
and thus more “deviant,” it might be more difficult to detect unique
predictors of alcohol compared to drug outcomes. At ages when
alcohol use becomes statistically common and therefore less de-
viant than drug use (as in young adulthood), it might be easier to
detect a difference between alcohol and drug use disorders.
In sum, the current study found that adolescent family harmony
decreased young adults’ risk for drug dependence but did not
predict alcohol dependence in the absence of drug dependence.
Family harmony partially mediated the effect of familial alcohol-
ism on young adults’ combined alcohol and drug dependence, and
this effect was unique, after adjusting for the effects of personality.
However, this mediated effect varied with the density of familial
alcoholism, such that protective benefits of family harmony were
lost at high levels of familial alcoholism density. These findings
suggest both the importance of distinguishing between alcohol and
drug outcomes and that interventions designed to improve adoles-
cent family environment may be insufficient for COAs with a high
density of familial alcoholism.
References
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and
interpreting interactions. Newbury Park: Sage.
American Psychiatric Association. (1980). Diagnostic and statistical man-
ual of mental disorders (3rd ed.). Washington, DC: Author.
Baker, J. R., & Yardley, J. K. (2002). Moderating effect of gender on the
relationship between sensation seeking-impulsivity and substance use in
adolescence. Journal of Child and Adolescent Substance Abuse, 12,
27– 43.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychology,
51, 1173–1182.
Bloom, B. L. (1985). A factor analysis of self-report measures of family
functioning. Family Process, 24, 225–239.
Bloom, B. L., & Naar, S. (1994). Self-report measures of family function-
ing: Extensions of a factorial analysis. Family Process, 33, 203–216.
Chassin, L., Barrera, M., Bech, K., & Kossak-Fuller, J. (1992). Recruiting
a community sample of adolescent children of alcoholics: A comparison
of three subject sources. Journal of Studies on Alcohol, 53, 316 –319.
Chassin, L., Curran, P. J., Hussong, A. M., & Colder, C. R. (1996). The
relation of parent alcoholism to adolescent substance use: A longitudinal
follow-up study. Journal of Abnormal Psychology, 105, 70 –80.
Chassin, L., Flora, D., & King, K. M. (2004). Trajectories of alcohol and
drug use and dependence from adolescence to adulthood: The effects of
familial alcoholism and personality. Journal of Abnormal Psychology,
113, 483– 498.
Chassin, L., Pitts, S. C., DeLucia, C., & Todd, M. (1999). A longitudinal
study of children of alcoholics: Predicting young adult substance use
disorders, anxiety, and depression. Journal of Abnormal Psychology,
108, 106 –119.
Chassin, L., & Ritter, J. (2001). Vulnerability to substance use disorders in
childhood and adolescence. In R. E. Ingram & J. M. Price (Eds.),
Vulnerability to psychopathology: Risk across the life span (pp. 107–
134). New York: Guilford Press.
Chassin, L., Ritter, J., Trim, R., & King, K. (2003). Adolescent substance
use. In R. Barkley & E. Mash (Eds.), Handbook of child psychopathol-
ogy (pp. 119 –232). New York: Plenum Press.
Chassin, L., Rogosch, F., & Barrera, M. (1991). Substance use and symp-
tomatology among adolescent children of alcoholics. Journal of Abnor-
mal Psychology, 100, 449 – 463.
Costa, P., & McCrae, R. (1992). Revised NEO Personality Inventory
(NEO-PI–R) and NEO Five-Factor Inventory (NEO–FFI) professional
manual. Odessa, FL: Psychological Assessment Resources.
Cummings, M. E., & Davies, P. (1996). Emotional security as a regulatory
process in normal development and the development of psychopathol-
ogy. Development and Psychopathology, 8, 123–139.
Curran, G. M., Stoltenberg, S. F., Hill, E. M., Mudd, S. A., Blow, F. C., &
Zucker, R. A. (1999). Gender differences in the relationships among
SES, family history of alcohol disorders and alcohol dependence. Jour-
nal of Studies on Alcohol, 60, 825– 832.
Eitle, D. (2005). The moderating effects of peer substance use on the
family-structure-adolescent substance use association: Quantity versus
quality of parenting. Addictive Behaviors, 30, 963–980.
El-Sheikh, M., & Buckhalt, J. A. (2003). Parental problem drinking and
children’s adjustment: Attachment and family functioning as moderators
and mediators of risk. Journal of Family Psychology, 17, 510 –520.
El-Sheikh, M., & Flanagan, E. (2001). Parental problem drinking and
children’s adjustment: Family conflict and parental depression as medi-
ators and moderators of risk. Journal of Abnormal Child Psychology, 29,
417– 432.
Endicott, J., Andreason, N., & Spitzer, R. (1975). Family history diagnosis
criteria. New York: Biometrics Research, New York Psychiatric
Institute.
Grych, J. H., & Fincham, F. D. (1990). Martial conflict and children’s
adjustment: A cognitive-contexual framework. Psychological Bulletin,
108, 267–290.
Guo, J., Hill, K. G., Hawkins, J. D., Catalano, R. F., & Abbott, R. D.
(2002). A developmental analysis of sociodemographic, family, and peer
effects on adolescent illicit drug initiation. Journal of American Acad-
emy of Child and Adolescent Psychiatry, 41, 838 –845.
Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective
factors for alcohol and other drug problems in adolescence and early
adulthood: Implications for substance abuse prevention. Psychological
Bulletin, 112, 64 –105.
Hedeker, D. (1999). MIXNO: A computer program for mixed-effects
nominal logistic regression. Journal of Statistical Software, 4, 1–92.
Helzer, J. E., & Pryzbeck, T. R. (1988). The co-occurrence of alcoholism
with other psychiatric disorders in the general population and its impact
on treatment. Journal of Studies on Alcohol, 49, 219 –224.
Herjanic, B., & Reich, W. (1982). Development of a structured psychiatric
interview for children: Agreement between parent and child on individ-
ual symptoms. Journal of Abnormal Child Psychology, 10, 307–324.
Heyman, R. E., O’Leary, K. D., & Jouriles, E. N. (1995). Alcohol and
aggressive personality styles: Potentiators of serious physical aggression
against wives? Journal of Family Psychology, 9, 44 –57.
Hussong, A. M., & Chassin, L. (1997). Substance use initiation among
adolescent children of alcoholics: Testing protective factors. Journal of
Studies on Alcohol, 58, 272–279.
Hussong, A. M., & Chassin, L. (2002). Parent alcoholism and the leaving
home transition. Development and Psychopathology, 14, 139 –157.
Jacob, T., Leonard, K. E., & Haber, J. R. (2001). Family interactions of
alcoholics as related to alcoholism type and drinking condition. Alco-
holism: Clinical and Experimental Research, 25, 835– 843.
Jacob, T., Waterman, B., Heath, A., True, W., Bucholz, K. K., Haber, R.,
et al. (2003). Genetic and environmental effects on offspring alcoholism.
Archives of General Psychiatry, 60, 1265–1272.
Johnson, E. O., & Pickens, R. W. (2001). Familial transmission of alco-
holism among nonalcoholics and mild, severe, and dyssocial subtypes of
alcoholism. Alcoholism: Clinical and Experimental Research, 25, 661–
666.
Kessler, R. C. (2002). National Comorbidity Study, 1990 –1992. Retrieved
February 10, 2003, from the University of Michigan Survey Research
Center, Inter-university Consortium for Political and Social Science
Research Web site: http://www.icpsr.umich.edu
King, K. M., & Chassin, L. (2004). Mediating and moderated effects of
adolescent behavioral undercontrol and parenting in the prediction of
drug use disorders in emerging adulthood. Journal of Addictive Behav-
iors, 18, 239 –249.
330
ZHOU, KING, AND CHASSIN
Krueger, R. F., Hicks, B. M., Patrick, C. J., Carlson, S. R., Iacono, W. G.,
& McGue, M. (2002). Etiologic connections among substance depen-
dence, antisocial behavior and personality: Modeling the externalizing
spectrum. Journal of Abnormal Psychology, 111, 411– 424.
Legrand, L. N., McGue, M., & Iacono, W. G. (1999). Searching for
interactive effects in the etiology of early-onset substance use. Behavior
Genetics, 29, 433– 443.
Loukas, A., Fitzgerald, H. E., Zucker, R. A., & von Eye, A. (2001).
Parental alcoholism and co-occurring antisocial behavior: Prospective
relationships to externalizing behavior problems in their young sons.
Journal of Abnormal Child Psychology, 29, 91–106.
Luthar, S. S., Cicchetti, D., & Becker, B. (2000). The construct of resil-
ience: A critical evaluation and guidelines for future work. Child De-
velopment, 71, 543–562.
MacKinnon, D. P., & Dwyer, J. H. (1993). Estimating mediated effects in
prevention studies. Evaluation Review, 17, 144 –158.
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., &
Sheets, V. (2002). A comparison of methods to test mediation and other
intervening variable effects. Psychological Methods, 7, 83–104.
MacKinnon, D. P., Lockwood, C. M., & Williams, J. (2004). Confidence
limits for the indirect effect: Distribution of the product and resampling
methods. Multivariate Behavioral Research, 39, 99 –128.
MacKinnon, D. P., Warsi, G., & Dwyer, J. H. (1995). A simulation study
of mediated effect measures. Multivariate Behavioral Research, 30,
41– 62.
Mahaddian, E., Newcomb, M. D., & Bentler, P. M. (1988). Risk factors for
substance use: Ethnic differences among adolescents. Journal of Sub-
stance Abuse, 1, 11–23.
McGue, M. (1999). The behavioral genetics of alcoholism. Current Direc-
tions in Psychological Science, 8, 109 –115.
McGue, M., Slutske, W., & Iacono, W. G. (1999). Personality and sub-
stance use disorders: II. Alcoholism versus drug use disorders. Journal
of Consulting and Clinical Psychology, 67, 394 –404.
Meeker, W. Q., Cornwell, L. W., & Aroian, L. A. (1981). Selected tables
in mathematical statistics: Vol. VII. The product of two normally dis-
tributed random variables. Providence, RI: American Mathematical
Society.
Muthe´n, L. K., & Muthe´n, B. O. (2003). Mplus (Version 2.12) [Computer
software]. Los Angeles: Author.
Newman, D. L., Moffitt, T. E., Caspi, A., Magdol, L., Silva, P. A., &
Stanton, W. R. (1996). Psychiatric disorder in a birth cohort of young
adults: Prevalence, comorbidity, clinical significance, and new case
incidence from ages 11 to 21. Journal of Consulting and Clinical
Psychology, 64, 552–562.
Patterson, G. R. (1982). Coercive family process. Eugene, OR: Castalia.
Patterson, G. R., DeBaryshe, B. D., & Ramsey, E. (1989). A developmental
perspective on antisocial behavior. American Psychologist, 44, 329–
335.
Pollock, N. K., Martin, C. S., & Langenbucher, J. W. (2000). Diagnostic
concordance of DSM–III, DSM–III–R, DSM–IV and ICD–10 alcohol
diagnoses in adolescents. Journal of Studies on Alcohol, 61, 439 –446.
Robins, L. N., Helzer, J. E., Croughan, J., & Ratcliff, K. S. (1981). National
Institute of Mental Health Diagnostic Interview Schedule: Its history,
characteristics, and validity. Archives of General Psychiatry, 38, 381–
389.
Rutter, M. (1990). Psychological resilience and protective mechanisms. In
J. Rolf, A. S. Masten, D. Cicchetti, K. H. Nuechterlein, & S. Weintraub
(Eds.), Risk and protective factors in the development of psychopathol-
ogy (pp. 181–214). New York: Cambridge University Press.
Sher, K. J. (1991). Children of alcoholics: A critical appraisal of theory
and research. Chicago: University of Chicago Press.
Sher, K. J., Gershuny, B. S., Peterson, L., & Raskin, G. (1997). The role of
childhood stressors in the intergenerational transmission of alcohol use
disorders. Journal of Studies on Alcohol, 58, 414 –427.
Sher, K. J., Walitzer, K. S., Wood, P. K., & Brent, E. E. (1991). Charac-
teristics of children of alcoholics: Putative risk factors, substance use and
abuse, and psychopathology. Journal of Abnormal Psychology, 100,
427– 448.
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and non-
experimental studies: New procedures and recommendations. Psycho-
logical Methods, 7, 422– 445.
Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in
structural equation models. In S. Leinhard (Ed.), Sociological method-
ology 1982 (pp. 290–312). Washington, DC: American Sociological
Association.
Stice, E., & Gonzales, N. (1998). Adolescent temperament moderates the
relation of parenting to antisocial behavior and substance use. Journal of
Adolescent Research, 13, 5–31.
Stoltenberg, S. F., Mudd, S. A., Blow, F. C., & Hill, E. M. (1998).
Evaluating measures of family history of alcoholism: Density versus
dichotomy. Addiction, 93, 1511–1520.
Taylor, J., Iacono, W. G., & McGue, M. (2000). Evidence for a genetic
etiology of early-onset delinquency. Journal of Abnormal Psychology,
109, 634 – 643.
Taylor, J., Malone, S., Iacono, W. G., & McGue, M. (2002). Development
of substance dependence in two delinquency subgroups and nondelin-
quents from a male twin sample. Journal of the American Academy of
Child and Adolescent Psychiatry, 41, 386 –393.
Tein, J., Sandler, I. N., MacKinnon, D. P., & Wochik, S. A. (2004). How
did it work? Who did it work for? Mediation in the context of a
moderated prevention effect for children of divorce. Journal of Consult-
ing and Clinical Psychology, 72, 617– 624.
Walden, B., McGue, M., Iacono, W. G., Burt, S. A., & Elkins, I. (2004).
Identifying shared environmental contributions to early substance use:
The respective roles of peers and parents. Journal of Abnormal Psychol-
ogy, 113, 440 – 450.
Wills, T. A., Resko, J. A., Ainette, M. G., & Mendoza, D. (2004). Role of
parental social support and peer support in adolescent substance use: A
test of mediated effects. Psychology of Addictive Behaviors, 18, 122–
134.
Wills, T. A., Sandy, J. M., Yaeger, A., & Shinar, O. (2001). Family risk
factors and adolescent substance use: Moderation effects for tempera-
ment dimensions. Developmental Psychology, 37, 283–297.
Windle, M. (1996). On the discriminative validity of a family history of
problem drinking index with a national sample of young adults. Journal
of Studies on Alcohol, 57, 378 –386.
Wootton, J. M., Frick, P. J., Shelton, K. K., & Silverthorn, P. (1997).
Ineffective parenting and childhood conduct problems: The moderating
role of callous-unemotional traits. Journal of Consulting and Clinical
Psychology, 65, 301–308.
Zucker, R. A., Ellis, D. A., & Fitzgerald, H. E. (1994). Developmental
evidence for at least two alcoholisms: I. Biopsychosocial variation
among pathways into symptomatic difficulty. In T. F. Babor, V. Hes-
selbrock, R. E. Meyer, & W. Shoemaker (Eds.), Types of alcoholics:
Evidence from clinical, experimental and genetic research (pp. 134 –
146). New York: New York Academy of Sciences.
Received February 2, 2005
Revision received November 19, 2005
Accepted November 28, 2005 䡲
331
SUBSTANCE DEPENDENCE DISORDERS IN YOUNG ADULTS
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Abnormal Psychology
This content is subject to copyright. Terms and conditions apply.
