Drinking alcohol to improve mood partially mediates the relation between major depression and alcohol dependence

PDF

Download

Open document

Flip pages

Summarize

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

DRINKING ALCOHOL TO IMPROVE MOOD PARTIALLY MEDIATES THE
RELATION BETWEEN MAJOR DEPRESSION AND ALCOHOL DEPENDENCE

by

Kelly Young-Wolff

A Thesis Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF ARTS
(PSYCHOLOGY)

August 2008

Copyright 2008 Kelly Young-Wolff

ii
Table of Contents

List of Tables iii

List of Figures vi

Abstract v

Chapter 1: Introduction 1
1.1 Epidemiological Studies 2
1.2 Family and Twin Studies 3
1.3 Alcohol Expectancies, Motives and AD 8
1.4 MD, AD, and Drinking to Manage Mood 10
1.5 Familial Transmission of Expectancies and Drinking Motives 13
1.6 Potential Mechanisms 14
1.7 Current Study Aims and Hypotheses 16

Chapter 2: Method 18
2.1 Participants 18
2.2 Measures 20
2.3 Covariates 22
2.4 Statistical Method 23
2.5 Model Fitting 27
2.6 Twin Models 27

Chapter 3: Results 31
3.1 Descriptive Statistics 31
3.2 Bias from Incomplete Data 33
3.3 Phenotypic Mediation Analyses 34
3.4 Twin Models 39

Chapter 4: Discussion 54
4.1 General Discussion 54
4.2 Strengths and Limitations 60
4.3 Conclusion 62

References 64

Appendices
Appendix A: Items on Mood Management Scale (MOT) and Factor 72
Loadings
Appendix B: Correlation Table With and Without Covariates 73

iii
List of Tables

Table 1: Means, Standard Deviations and Range AD, MD, MOT 31
Table 2: Correlations between Key Variables and Covariates by Gender 32
Table 3: Correlations for Cross-Twin Lifetime MD, AD and MOT 33
Table 4: Model Fitting Results for Phenotypic Mediation Analyses (N=5506) 35
Table 5: Standardized Estimates from Mediation Analyses (N=5506) 36
Table 6: Results from Univariate Twin Analyses of MOT (N=3344) 39
Table 7: Fit Indices for MD Univariate Nested Models (N=3344) 40
Table 8: Fit Indices for AD Univariate Nested Models (N=3344) 40
Table 9: Fit Indices MOT-AD and MD-AD Bivariate Models (N=3344) 44
Table 10: Parameter Estimates from Baseline Bivariate MOT-AD (N=3344) 44
Table 11: Model Fitting Results from Best Fitting Bivariate MD-AD AE Model 45
Table 12: Model Fitting Results for Multivariate Twin Models 48
Table 13: Best Fitting Multivariate Model (I): Standardized Parameter Estimates 49

iv
List of Figures

Figure 1: Direct Mediation Model 25

Figure 2: Indirect Mediation Model 26

Figure 3: Phenotypic Mediation Results 38

Figure 4a: Males: No Mediation 50

Figure 4b: Males: Mediation Via the Factors Underlying MOT 50

Figure 5a: Females: No Mediation 50

Figure 5b: Females: Mediation Via the Factors Underlying MOT 51

v
Abstract
The purpose of this study was to examine whether the relation between lifetime
major depression (MD) and lifetime alcohol dependence (AD) is mediated by drinking
to manage mood. Participants were 5506 individuals from the Virginia Adult Twin
Study of Psychiatric and Substance Use Disorders (VATSPSUD). First, it was
hypothesized that drinking to manage mood partially mediates the association between
MD and AD within individuals. Second, it was hypothesized that the overlapping
familial risk for MD and AD is partially mediated by drinking to manage mood.
Both hypotheses were confirmed. Drinking to manage mood partially mediated
the relation between MD and AD in both men and women. More specifically, after
accounting for the genetic and environmental factors shared among drinking to manage
mood, MD, and AD, there was little remaining overlapping genetic and environmental
variance between AD and MD.

1
Chapter 1: Introduction
Major depression (MD) is a pressing public health problem that can lead to
significant impairment in occupational and psychosocial functioning. Similarly, alcohol
problems, including alcohol dependence (AD), are highly prevalent and have a
detrimental impact on daily life. In addition, MD and AD co-occur at higher rates than
expected by chance in both treatment and community-based studies. In general,
individuals with comorbid MD and AD enter treatment earlier, have more
hospitalizations, more legal problems, and higher risk of blackouts and suicide (for a
review, see Merikangas & Gelernter, 1990). Although the higher than expected
association between AD and MD is recognized, there is little agreement about the
relation between the two disorders (for a review, see Swendsen & Merikangas, 2000;
Davidson & Ritson, 1993). It is likely that the association between MD and AD is
heterogeneous, and a better understanding of which individuals with MD are at risk for
AD and the circumstances wherein risk is likely to be expressed is needed (Holahan et
al., 2003).
A great deal of research has indicated that alcohol expectancies and motives are
key factors in predicting drinking. Drinking to cope with negative affect and to manage
mood increases the risk for alcohol consumption, DSM-IV alcohol abuse and
dependence, frequent intoxication, and lifetime alcohol problems (Prescott et al., 2004;
Peirce, et al., 1994; Fromme et al., 1993; Carey & Correia, 1997). Although research
has demonstrated that drinking to manage mood is a significant genetic risk factor for

2
AD (Prescott et al., 2004), the potential role of drinking to improve mood as a mediator
of the association between lifetime MD and AD has not been examined.
1.1 Epidemiological Studies
Epidemiological studies are necessary to determine rates of AD and MD in the
general population. The National Comorbidity Survey (NCS), a nationally
representative household survey of 8,098 individuals ages 15-54, found that 15.8% of
their sample met lifetime criteria for MD, 9.4% met lifetime criteria for alcohol abuse,
and 14.1% met lifetime criteria for AD (Kessler et al., 1997). In a NCS replication
(NCS-R) of individuals ages 18-60

and older, the 12-month prevalence of MD, alcohol
abuse, and AD was 6.7%, 3.1% and 1.3% (Kessler et al., 2005). Recently, the National
Survey on Drug Use and Health (NSDUH) indicated that of adults aged 18 and older,
7.6% experienced at least one MD episode in the past year, and 8% experienced AD in
the past year (SAMHSA, 2004 & 2005 NSDUHs).
In addition, nationally representative studies typically indicate that there is two
to four times the risk of MD or AD given the occurrence of the other (Grant and
Harford, 1995; Kessler et al., 1997; Regier et al., 1990). The NCS study found that
among individuals ages 15-54, AD was significantly higher in men and women with
MD compared to men and women without MD (ORs, 2.95 and 4.05) (Kessler et al.,
1997). The Epidemiologic Catchment Area survey (ECA), based on 14,480 community
members ages 18-79 from five sites, found that AD was strongly associated with
lifetime mood disorders (OR, 2.9) (Regier et al., 1990). In a follow-up of analyses from
the ECA, Crum et al. (2001) found that depressed female problem drinkers were twice

3
as likely as non depressed drinkers to become daily drinkers, and depressed male
problem drinkers were three times as likely to become daily drinkers at a one year
follow-up.
Although lifetime prevalence of MD and AD varies slightly by study, nationally
representative surveys typically find higher rates and earlier onset of MD in women
compared to men, and higher rates and earlier onset of AD in men compared to women.
The National Longitudinal Alcohol Epidemiologic Survey (NLAES) found that among
individuals ages 18-95

and older, 8.6% of men and 11% of women had lifetime MD,
and 25.5% of men and 11.4% of women had lifetime AD (Grant & Harford, 1995). In
addition, there are typically more women classified as having both disorders. Using
data from the ECA, Helzer and Pryzbeck (1988) found that alcoholism preceded the
onset of depression in 78% of cases in men, and MD preceded the onset of alcoholism
in 66% of cases in women. However, Hettema et al. (2003) showed similar rates of
disorder onset for men and women ages 21-62: 45.5% of males and 50.8% of females
had AD following MD onset. More longitudinal research is needed to better understand
the trajectories of alcohol consumption and depression between genders across the
lifespan.
1.2 Family and Twin Studies
Both MD and AD are genetically influenced. Children of alcoholics are up to
four times more likely to develop alcohol related problems than individuals in the
general population (Cotton, 1979), and children of individuals with MD are almost
three times more likely to develop MD than individuals in the general population

4
(Sullivan et al., 2000). There are several possible explanations for the association of
AD and MD including: MD increases the risk for AD; AD increases the risk for MD;
and there is a common etiologic factor underlying both disorders, such as shared social
and environmental or genetic factors. Similar rates of AD would be expected in
relatives of individuals with and without MD if the two disorders were independent
(Grant et al., 1996). If MD directly increases the risk for AD there should be increased
rates of AD in relatives of probands with MD when the relative also has MD. If AD
directly increase the risk for MD (i.e. AD leads to MD through neurobiological
adaptations or the negative social consequences of drinking) increased rates of MD in
relatives of probands with AD would only be expected when the relative also has AD.
If the two disorders have a shared liability (i.e. they share the same set of genes or
family environmental factors), relatives of a proband with MD will be at an increased
risk for AD, and relatives of a proband with AD will be at an increased risk for MD.
Family and twin studies are used to assess possible mechanisms for the association
between AD and MD within families. Inconsistent findings have emerged from studies
that examine MD in relatives of alcoholics and alcoholism in relatives of individuals
with MD (for review, see Merikangas & Gelernter, 1990).
Several studies that include family history suggest that MD and AD are not
etiologically distinct. Grant et al. (1996) and Dawson and Grant (1996) demonstrated
that relatives of individuals with both AD and MD had a higher prevalence of AD than
relatives of individuals with AD only. In addition, individuals with MD without AD
were more likely to have alcoholic relatives compared with individuals without either

5
disorder which suggests that AD and MD have a shared underlying liability. Finn et al.
(1990) found an increased risk for MD in non-alcoholic men with a family history of
alcoholism, which also supports the shared liability model. These studies suggest that
there may be a familial overlap between AD and MD; however, other studies do not
support this conclusion. Several studies find that MD and AD segregate independently
in families and do not arise from shared familial liability (Meier & Merikangas, 1996)
and that children of adults with MD are not at an increased risk of alcoholism compared
to controls (Merikangas et al., 1985; Weissman et al., 1984; Maier et al., 1994).
Schuckit et al. (2006) found that independent episodes of MD (i.e. not substance
induced depression) in individuals with a family history of alcoholism enhanced risk
for alcohol problems, but a family history of alcoholism did not increase risk for MD
by itself.
Adoption studies are useful for separating the effects of genetic influences on
MD and AD from environmental influences. In adopted individuals, similarities to
biological parents are likely to be the result of heritable genetic factors. Adoption
studies have found evidence both for and against shared genetic overlap between AD
and MD. Ingraham & Wender (1992) examined risk for substance abuse in the
biological relatives of adopted individuals with MD and observed an increased
frequency of substance abuse in biological relatives even when the depressed proband
didn’t have substance abuse. However, other studies have found no evidence of an
increased risk of MD in adopted children of alcoholic parents (Goodwin et al., 1977;
Goodwin et al., 1973).

6
The twin design is especially instructive because it allows for the estimation of
the degree to which environmental and genetic factors contribute to the overlapping
risk for MD and AD. An overlap of genetic and environmental risk factors for AD and
MD supports the view that the liability to MD and AD is correlated both within
individuals and across relatives. In an earlier study using a subset of the sample used in
this study, Prescott et al. (2000) found that there are genetic and specific environmental
factors that influence both MD and AD providing evidence for the correlated liability
model. This study assessed lifetime MD and AD in 3,755 complete twin pairs in the
VATSPSUD sample and found that approximately 9%-14% of the variation in liability
overlapped between MD and AD, with approximately 50%-60% due to shared genetic
factors. An epidemiological study of female twins from the VATSPSUD found that
liability to alcoholism in one twin was significantly correlated with liability to
depression in cotwin. Kendler et al. (1993) concluded that it is likely that genetic
factors contribute to the liability of alcoholism, the liability of MD, and jointly to
influence the liability of both disorders.
Kuo et al. (2006) assessed the ages of onset of AD and MD using a sample from
the VATSPSUD study. They found that the onset of AD was typically in young
adulthood, and the onset of MD was more spread out across age. The majority of
individuals with lifetime AD and lifetime MD had an onset of AD following an onset
of MD. In this sample, prior MD significantly increased the risk of developing AD,
although the risk decreased with time; 45% of women and 20.5% of men with AD had
pre-existing MD. Prior AD had only slight effects on the risk for future MD. This

7
study indicates that MD may lead to AD in both genders, but AD may not lead to MD
(Kuo et al. 2006). In addition, after controlling for within disorder familial transmission
there was no evidence for across disorder transmission suggesting that there may not be
a shared underlying liability to both disorders. These results are especially clinically
significant because they provide support for the self-medication hypothesis. However,
this approach looked at subsequent onsets of MD or AD after already having either MD
or AD, rather than assessing liability to MD or AD (as indicated by ever having it, or a
cotwin ever having it) so the results may not generalize to individuals who have not yet
had an onset of MD or AD.
Some studies provide evidence for differences in family liability depending on
gender. Coryell et al. (1992) found that alcoholism in depressed female probands, but
not depressed male probands, increased the risk for MD in relatives. Dawson & Grant
(1996) demonstrated that alcoholism in paternal male and maternal female relatives
was more highly correlated with MD than alcoholism in maternal male and paternal
female relatives. In addition, there may be gender differences in familial risk for AD
(for review, see Prescott, 2002). Winokur & Coryell (1991) found higher rates of
alcoholism in families of women with MD but not in families of men with MD.
Prescott et al. (2000) found that a history of MD in one twin was a risk factor for AD
among identical twins, and to a smaller degree among fraternal same-sex twins, but not
among opposite sex pairs. This suggests that there may be a sex-specific transmission
of MD and AD. One limitation to these approaches is their reliance on clinical
diagnoses. Binary measures ignore individuals with sub-threshold symptoms and

8
reduce the power to find cross gender differences in transmission of MD and AD. It
would be useful to study continuous measures that look at mechanisms for risk
transmission that are measurable in all individuals.
1.3 Alcohol Expectancies, Motives, and AD
Alcohol expectancies refer to the anticipated behavioral, cognitive and
emotional consequences of alcohol consumption. Alcohol expectancies can develop
prior to drinking and predict the likelihood of both drinking onset and level of
consumption once drinking begins (for review, see Goldman et al., 1999). Expectancies
are among the strongest predictors of drinking, even after controlling for other variables
(for review, see Goldman, 1994). In addition to predicting drinking practices,
expectancies also influence recovery and relapse (for review, see Jones et al., 2001).
Brown (1985) found that alcohol patients entering treatment with strong alcohol
expectancies had worse outcomes at a one-year follow-up.
A great deal of research has indicated that alcohol expectancies are key factors
in motivating drinking behavior (for a review, see Leigh, 1989). Drinking motives are
based on situation and personal experience and they are considered to be more
proximal influences of use than expectancies (Cox and Klinger, 1988; Galen et al.,
2001). Expectancy theory suggests that people will drink to manage their mood to the
extent that they expect that alcohol will reduce negative affect (Cooper et al., 1995). In
theory, an individual must believe that alcohol will improve his/her negative mood
before he/she will use it instrumentally to manage mood. Whereas expectancies can be
assessed in those who drink and those who have never tried alcohol, drinking motives

9
can only be assessed in individuals who have consumed alcohol; however, there are
very few operational differences between the two (Goldman, Del Boca, & Darkes,
1999).
Drinking motives are important in the initiation, maintenance and relapse to
drinking. The motivational model assumes that an individual makes a decision to
consume alcohol or not to consume alcohol based on the affective change that he or she
is motivated to achieve, and the decision to drink may be unconscious and automatic.
Four key motives for alcohol consumption have been recognized in the literature:
drinking for coping reasons, drinking for enhancement reasons, drinking for social
reasons, and drinking for conformity reasons (Cooper, 1994; Cooper et al., 1992b).
According to motivational models of alcohol consumption, there are two types of
internal motives for drinking; coping motives to alleviate the experience of distress, and
enhancement motives to enhance positive emotions (for a review, see Cox & Klinger,
1988). Reduction of negative affect is commonly cited as a motive for using alcohol
(Park & Levenson, 2002; Cooper et al., 1992a; 1995), and several studies have
demonstrated that coping motives mediate the relation between expectancies and
consumption problems in both community and treatment samples (Cooper et al., 1995;
Galen et al., 2001).
Many studies have found an association between drinking to improve mood and
formal DSM-IV alcohol use disorders (Prescott et al., 2004; Carpenter & Hasin, 1998a;
Carpenter & Hasin, 1999; Cooper et al., 1992b), consumption and drinking problems
(Peirce, et al., 1994; Holahan et al., 2003), lifetime alcohol problems (Carey & Correia,

10
1997), and frequency of intoxication (Fromme et al., 1993; Carpenter & Hasin, 1998b).
Affect regulation theory asserts that drinking is purposeful and individuals consume
alcohol in order to cope with or regulate negative emotions. Past research has shown
that 10% to 25% of drinkers drink alcohol to regulate negative emotions (Peirce et al.,
1994). In addition, drinking to cope and alcohol problems are still associated after
adjusting for consumption (e.g. Cooper et al., 1988), which indicates that the
correlation between drinking to cope and alcohol problems is not completely explained
by increases in alcohol consumption.
The role of gender in drinking to regulate affect is equivocal. Many studies
report that men are more likely to drink to cope with negative emotions than women,
(Ratliff & Burkhart, 1984; Klein, H., 1992; Peirce et al., 1994) and some research
indicates that although positive alcohol expectancies are associated with increased risk
for problem drinking in both genders, the risk for men is greater than for women
(Cooper et al., 1992a; Cooper et al., 1995). In contrast, other investigators find that
alcoholic women are more likely then alcoholic men to use alcohol to alter mood
(Olenick and Chalmers, 1991; Timko, Finney & Moos, 2005), and that drinking to
reduce tension is a stronger predictor of frequency of drinking in women than in men
(Mooney et al., 1987). Still other studies find no gender differences (Chalder et al.,
2006; Carey & Correia, 1997).
1.4 MD, AD, and Drinking to Manage Mood
The self-medication hypothesis (SMH; Khantzian, 1985) suggests that
individuals who are motivated to drink alcohol in order to reduce unpleasant emotional

11
states should have improvements in psychiatric symptoms, at least temporarily,
following the use of alcohol. However, there are inconsistencies in the literature about
how depression impacts the clinical course of alcohol use, or how alcohol use
influences subsequent depression. Directionality is difficult to determine because MD
can both lead to and result from alcohol use. It is likely that the process is cyclical in
some cases and individuals who drink to achieve the mood regulating effects of alcohol
will be at a risk for greater distress, which may then lead to additional regulation.
Hussong et al. (2001) found that individuals who experienced more sadness than usual
drank more alcohol than they usually do, which then predicted increases in subsequent
sadness. It is probable that the motive of drinking to improve mood is an important
predictor of AD, even in the absence of an actual mood improvement.
Several studies have demonstrated that coping motives mediate the relation
between negative affect and alcohol problems (Cooper et al, 1995; Gaher et al., 2006;
Peirce et al., 1994). However, the correlations between negative affect and coping
motives are typically moderate which suggests that other things also influence coping
motives (Gaher et al., 2006). Research examining drinking to improve mood indicates
that high levels of negative emotionality are linked to coping reasons for drinking
(Cooper et al., 1995; Holahan et al., 2004), suggesting that individuals with higher
levels of negative emotions are at an increased risk of drinking to self-medicate
(Colder, 2001). Zack et al. (1999) found that among individuals with drinking
problems, those with high psychiatric distress demonstrated stronger cognitive
associations between negative affect and alcohol concepts. Cooper et al. (1995) tested a

12
motivational model of alcohol use in adults and adolescents and found that drinking for
coping reasons mediated the relation between drinking and the experience of negative
emotion. This study concluded that individuals who drink to cope with negative
emotion might be motivated to do so due in part to their generally elevated levels of
depression. In a study that assessed the relation of financial strain to alcohol use,
Peirce et al. (1994) found that drinking to manage negative emotions partially mediates
the relation between depressed mood and alcohol consumption. There was a positive
direct effect of depressed mood on drinking to cope, and of drinking to cope on alcohol
consumption.
Holahan et al. (2003) assessed drinking to cope with distress over a ten-year
period in depressed patients and found that baseline drinking to cope was a risk factor
for alcohol consumption at 1, 4 and 10-year follow-up and a risk factor for alcohol
problems at the 1 and 4 year follow-up. In addition, 67% of individuals who drank to
cope at baseline experienced these drinking consequences in the context of a depressive
episode, whereas only 40% of the individuals who did not drink to cope experienced
these consequences in the context of a depressive episode. Johnson & Gurin (1994)
used four items related to drinking motives adapted from a Spanish version of the
Alcohol Expectancy Questionnaire (Brown et al., 1987) to assess the relation between
depressed mood and drinking problems in over 1000 Puerto Rican Americans. This
study found that drinking to improve mood moderated the relation between drinking
and negative affect, such that co-occurrence was more likely when individuals believed
that alcohol would improve their mood. The available evidence suggests that drinking

13
to regulate affect may partially explain the relation between depressed mood and
alcohol consumption.
1.5 Familial Transmission of Expectancies and Drinking Motives
There is relatively little research on familial influences on drinking expectancies
or motives. Many studies indicate that individuals with an alcoholic parent have more
positive expectations about alcohol than those without an alcoholic parent (Brown et
al., 1987; Sher et al., 1991; Brown et al., 1999), which suggests that alcohol
expectations may run in families. Studies have also shown that twins have similar
beliefs about alcohol and much of this similarity can be attributed to shared genes
(Perry, 1973; Prescott et al. 2004; Vernon et al., 1996). Perry (1973) demonstrated that
identical twins had more similar alcohol expectancies than fraternal twins, which
suggests the influence of shared genes. Prescott et al. (2004) found that approximately
40% of the variation in drinking with the motivation of improving mood in both sexes
is attributed to genetic factors. Recently, Agrawal et al. (2007) found that genetic
factors contributed to 18% of the total variance in coping motives in a sample of young
adult same-sex female twins, with the majority of variance attributable to non-shared
individual specific environment. In general, past research suggests that there is a strong
familial similarity (possibly genetic) for drinking expectancies and drinking motives.
Chalder et al. (2006) showed that children of parents with alcohol problems
report higher frequency and consumption and stronger motives for drinking than
children whose parents do not have alcohol problems. This study also found that
drinking to cope moderated the influence of parental drinking problems on alcohol

14
consumption, such that drinking to cope intensified the effect of parental alcohol
problems on alcohol consumption. There are two main reasons for the correlation
between familial risk for drinking motives and AD. It is possible that familial risk could
be translated into AD via drinking motives, or familial risk could contribute to both
drinking motives and risk for AD.
Several studies have tested whether expectancies and motives mediate familial
risk for AD. Prescott et al. (2004) found that some of the genetic variation in drinking
to alter mood overlapped with alcoholism risk for both genders in a study using a
sample from the VATSPSUD. Drinking to manage mood was associated with risk for
alcoholism; 31% of AD risk was shared with drinking to manage mood in women, and
35% of AD risk was shared with drinking to manage mood in men. The greater part of
this was due to genetic factors that influenced both drinking to manage mood and AD,
especially in men. The current study expands on the Prescott et al. (2004) study to
determine if the genetic overlap of MD and AD is partially explained by drinking to
manage mood. If drinking to manage mood is genetically influenced and affects risk for
MD and AD, it could create shared genetic variation for MD and AD. To our
knowledge, this is the first study examining whether drinking motives mediate the
association between familial risk for MD and AD.
1.6 Potential Mechanisms
The mechanisms behind the relation between MD and AD have not yet been
established. However, there are several possible reasons for the high comorbidity of
MD and AD. MD could be a result of neurobiological adaptations or physiological

15
changes due to chronic alcohol use, and problematic drinking can lead to disruptions in
work, relationships and general functioning that may eventually cause MD. On the
other hand, attempts to cope with MD may lead some people to drink excessively; if
this were the case MD would be expected to precede AD. If MD causes AD, relatives
of a proband with MD should be at an increased risk for AD if they have a lifetime
history of MD. It is also possible that the disorders have shared risk factors or a
common genetic predisposition, in which case relatives of depressed probands would
be at an increased risk for AD, and relatives of probands with AD at an increased risk
for MD (Swendsen & Merikangas, 2000). Although it is not possible to directly test
temporal causality with the current cross-sectional data, it is possible to assess whether
drinking to manage mood explains some of the overlapping variance between MD and
AD. If drinking to manage mood partially mediates the relation between MD and AD
there should be a significant correlation between MD and drinking to manage mood,
and between drinking to manage mood and AD. In addition, if drinking to manage
mood has a substantial genetic overlap with MD and AD, drinking to manage mood
would be expected to partially explain some of the overlapping familial risk. Research
has shown that individuals with an alcoholic parent have more positive expectations
about alcohol than those without an alcoholic parent (Brown et al., 1987) and prior
research in the current sample supports genetic influences on drinking to cope with
negative mood and with social anxiety (Prescott et al., 2004).
There are several ways in which drinking to manage mood might explain some
of the shared familial risk for MD and AD. There could be a genetic predisposition to

16
experience alcohol as physiologically pleasing, leading to drinking in response to
negative mood, and eventually causing AD. It is also possible that drinking motives are
transmitted through common environment. Individuals who grow up in an environment
where alcohol is used as a mechanism to cope with MD and manage mood may learn to
cope with alcohol in the same way. An alternative model would be one in which MD,
drinking to manage mood, and AD are correlated because they are caused by a shared
familial vulnerability. If this were the case, drinking to manage mood would not
necessarily directly affect risk for AD.
1.7 Current Study Aims and Hypotheses
In summary, research has established that MD and AD co-occur within
individuals and families at a higher than expected rate. In addition, drinking to improve
mood is associated with higher frequency of drinking and alcohol related problems.
What has not yet been investigated is whether drinking with the motive of improving
mood mediates the relation between MD and AD, and what are the genetic and
environmental bases for this relation.
The goals of this research are twofold: 1) to examine the impact of drinking to
manage mood on the relation between MD and AD within individuals, and 2) to
estimate the genetic and environmental contributions to MD-AD comorbidity
accounted for by mood drinking motives. It was hypothesized that drinking to manage
mood would partially explain the association between MD and AD. Drinking to
manage mood was expected to have a substantial genetic overlap with MD and AD and

17
partially explain some of the phenotypic and familial overlapping risk for co-occurring
lifetime MD and AD.

18
Chapter 2: Method
2.1 Participants
The sample is comprised of individuals from the Virginia Adult Twin Study of
Psychiatric and Substance Use Disorders (VATSPSUD, Kendler & Prescott, 2006).
This study is comprised of twins from the Virginia Twin Registry (VTR), a database of
all twins born in Virginia since 1918. The sample includes Caucasian twins born
between 1934 and 1974 who participated in personal clinical interviews. The
VATSPSUD study was chosen because it is a large population-based sample of twins
that allows the testing of explicit hypotheses about the role of genetic and
environmental risk factors in the relation between drinking to manage mood, MD, and
AD.
Data for this study were collected from the Female-Female (FF) sample and the
Male-Male/Male-Female (MF) sample of the VATSPSUD. The FF study consisted of
female twins born between 1934 and 1970 who were first interviewed in person
between 1987 and 1989 (FF1). Wave two of data collection (FF2) started in 1989 and
was completed in 1990. Wave three of data collection (FF3) began in 1992 and
continued through 1995. Finally, wave four of data collection (FF4) began in 1995 and
continued through 1997. Twins were eligible to participate in FF4 if they had been part
of FF1, were still living, and had not refused to be contacted after their previous
interview. Of the 2,295 eligible to participate in FF4, 1939 (84.5%) were interviewed
(for details of non-participation, see Kendler & Prescott, 2006). Individuals who

19
participated in FF4 were interviewed over the phone. Our sample included the 1442
females age 30 and older at the time of the FF4 interview.
The first wave of data for the MM/MF sample of the VATSPSUD was collected
between 1993 and 1996 and complete interviews were obtained from 6,812 individuals
(5092 men and 1720 women). The current analyses were based on the second wave of
MM/MF data collected for MM/MF subjects through in-person interviews, usually
conducted in the homes of participants, between 1995 and 1998 (approximately 5% of
interviews were conducted over the phone). Individuals were eligible for the second
wave of data if they had completed wave one. Wave two of the MM/MF study
consisted of 4225 men and 1396 women. The sample for the current analyses included
the 3044 males and 1020 females who were ages 30 and older at the time of the MF2
interview. Individuals who reported that they had never used alcohol were excluded
(N=325). Although MD and AD were assessed on multiple occasions, the MF2 and
FF4 assessments were chosen for this study because the assessment in the two samples
was most comparable across these two waves of data. Furthermore, these provide the
most recent information about lifetime MD and AD.
This study is based on data from a total of 5506 individuals (including 4324
individuals from complete pairs and 1182 individuals whose cotwin did not
participate), ages 30 and older, who completed the clinical interview (for details of
subject ascertainment and characteristics, see Kendler & Prescott, 2006). Given that the
diagnoses of interest are lifetime MD and lifetime AD, individuals who were not yet
past the primary risk period for developing either disorder were excluded. Although

20
limiting our sample to individuals ages 30 and older reduces our power, it is expected
to increase the accuracy of our results. Because the ages of onset for MD and AD are
different across gender (Kuo et al, 2006), and our sample includes opposite-sex pairs, it
is especially important that our sample consists of individuals who have already passed
through the highest risk period so that the cross-twin estimates in the opposite-sex pairs
are not biased in a different way than the bias that might exist for same-sex pairs. Age
30 was used as the cut off age because previous research with this sample shows that
after age 30 rates of onset of MD are about equal in males and females.
2.2 Measures
Lifetime AD and MD diagnoses were made using DSM-IV criteria based on a
structured diagnostic interview assessment, adapted from the Structured Clinical
Interview for DSM Disorders (SCID) (Spitzer and Williams, 1985) and administered by
clinically-trained interviewers. Participants were considered to meet the criteria for
lifetime diagnoses if they had ever met the DSM-IV criteria in their lives. Measuring
lifetime AD and MD, rather than current diagnosis, allows for the assessment of
overlapping liability to either disorder. Participants were informed about the purpose
of the study and gave verbal informed consent before telephone interviews and written
informed consent before in-person interviews. Waves of interviews were conducted so
that at least one year had passed in between assessments. Because these measures
assess AD and MD through self-report, multiple precautions were taken in order to
ascertain accurate information. Repeat interviews were conducted by a second
interviewer within 2 to 8 weeks for a portion of respondents in order to estimate test-

21
retest reliability. Test-retest reliability for DSM-IV MD among 375 participants was
ρ=0.66 (95% CI = 0.58-0.74) (Kendler & Prescott, 2006). Test-retest reliability for
DSM-IV AD among 382 individuals was ρ=0.72 (95% CI = 0.61, 0.82) (Kendler &
Prescott, 2006).
The AUI is a self-report measure developed in alcoholism treatment centers to
characterize reasons for drinking (Horn and Wanberg, 1983). The VATSPSUD study
used four of the 16 AUI scales that were relevant to drinking motives in the general
population: the Mood Management scale, the Social Gregarious scale, the Social
Anxiety scale, and the Mental scale. The current study utilizes the mood management
scale (MOT) because it is most relevant to the questions being addressed. Scores on the
AUI are calculated by adding the item scores (0=no, 1=yes) and dividing by the number
of items (one item was coded 0-2) to get an average score. The MOT scale includes
seven items that assess drinking to overcome depression (see Appendix A). In a
previous VATSPSUD study the internal consistency estimate (Cronbach alpha) for the
MOT scale was .82 and the test-retest correlation based on 256 subjects selected
randomly to complete the AUI a second time within 10 weeks was r= .85 (Prescott et
al., 2004).
A principal components analysis (PCA) was conducted of the 7 MOT items to
confirm that the items were measuring the same construct. The PCA yielded one large
factor (eigenvalue=3.5), that accounted for 50 % of the total variance. The eigenvalue
of the second factor was 0.95. Factor loadings of the MOT items from the one-factor
solution are shown in Appendix A. It was concluded that the 7 items of the MOT scale

22
are measuring a common underlying construct, and all items for the MOT scale were
kept based on adequate factor loadings in the one factor solution.
There are two reasons for missing AUI data. First, because drinking motives are
based on personal experience with drinking alcohol, lifetime abstainers from alcohol
cannot have motives for drinking. Unlike alcohol expectancies, motives for alcohol use
are only measurable after an individual has experienced drinking and develops motives
based on experience. Second, females from the FF4 study were interviewed by
telephone and returned the questionnaire containing the AUI through the mail. It is
likely the higher AUI non-response rates for females from FF4 is due to the additional
effort involved in returning the AUI in the mail rather than filling it out in an in person
interview.
2.3 Covariates
It is likely that additional overlapping risk factors contribute to AD besides MD
and drinking to cope. Gender differentially predicts risk for MD and AD, and previous
research has shown that the association between drinking to cope with negative affect
and AD is stronger in women (Prescott et al., 2004). Gender (male/female), mean-
centered age and mean-centered age-squared (to assess for linear and non-linear age
effects), education (highest grade completed ranging from 1 (elementary school) to 20
(doctoral degree or higher)), family income, and religious denomination
(Fundamentalist Protestant, Non-Fundamentalist Protestant, Catholic, None/Other)
were examined to determine if they are associated with MD, AD, and MOT.

23
2.4 Statistical Method
Analyses were conducted in four steps. First, biases due to attrition and
incomplete data were tested using SAS version 9.1. Demographic variables were
examined to see if they were associated with AUI questionnaire return, and/or predicted
being missing on MOT among non-abstainers. MOT scores and diagnoses of
individuals from complete pairs were compared to those from subjects whose cotwin
did not participate. A logistic regression analysis of AD on MOT and MD scores, with
a main effect of complete status (complete pair vs. incomplete pair) and an interaction
of completion status and MOT was conducted to see if individuals from complete pairs
had different risk for AD than incomplete pairs, and to see whether complete status
moderated the AD-Motive relation. AD was regressed on MD and all demographic
variables to test which demographic variables needed to be included in subsequent
analyses. Any demographic variables that significantly predicted AD were included.
Covariates were gender, age, age-squared, years of education, family income and
religious denomination.
Second, phenotypic associations between lifetime MD and lifetime AD were
investigated. In the baseline correlational model AD was regressed on MD and
correlated with MOT, and MOT and MD were also allowed to be correlated. All
regressions were conducted separately for males and females after adjusting for
covariates using Mplus version 4.2 (Muthen and Muthen, 2006). Third, the mediating
role of MOT (Hypothesis 1) was investigated by multiple regression analysis.
Mediation was evaluated based on the regression estimate changes in the baseline

24
correlational model versus a partial mediation model that regressed AD on MD and
MOT, and MOT on MD. The Mplus MODEL INDIRECT statement was used to assess
the statistical significance of the indirect effect of MOT on the MD-AD relation.
Fourth, evidence for the mediation hypothesis at the familial level was
investigated (Hypothesis 2). Univariate twin models were fit to MD, AD, and MOT,
then bivariate twin models were fit to evaluate the sources of genetic and
environmental overlap of AD with MD and AD with MOT. These models serve as a
basis for interpreting the subsequent models. Two types of trivariate twin models were
fit to partition the covariance between MOT, MD, and AD into genetic and
environmental components. The first is a fully-saturated (Cholesky) model that
assesses indirect mediation. The sources of overlap were assessed by conducting the
following tests: no overlapping sources of variation, full mediation of MD-AD overlap,
and various tests of partial mediation (i.e. genetic mediation of MD-AD overlap,
environmental mediation of MD-AD overlap). The second model is a twin mediation
model that regressed AD directly on MOT. This model tests the hypothesis that MOT
(rather than the factors underlying MOT) directly mediates AD-MD comorbidity.

In the indirect mediational model, risk for co-occurring lifetime MD and AD
arises from genetic and environmental factors shared with MOT. If the risk for co-
occurring lifetime MD and AD arises from both an environmental and genetic overlap,
both paths (a
2a
, e
2a
) would be expected to be significantly reduced after accounting for
the variance they share with MOT. However, if risk for co-occurring lifetime MD and
AD is not mediated by MOT, and there is no environmental or genetic overlap, the

25
relation between MD and AD would not be reduced after accounting for the variance
shared with MOT (see Figure 1).

Figure 1.
Indirect Mediation Model

A1 E1 A2 E2

a1md a1a a2a e2a
e1md e1a

A3 E3

Finally, a direct mediation model was tested that directly regressed AD on
MOT, and MOT on MD, rather than regressing the factors underlying MOT (see Figure
2). This model is different from the multivariate models described above because it
requires that the MOT-AD and MOT-MD covariance must be proportional to the MOT
variance components (a, c, e). However, some of the E variance observed in MOT is
not expected to be correlated with AD and MD because it includes measurement error.
Thus the proportionality test would not be a fair test. To address this, a fixed estimate
of MOT error variance as 1 minus the test-retest correlation within the current sample
(r=.85) was included. This allows the MOT-AD and MOT-MD covariance due to E to
be estimated after removing the measurement error in MOT.
AD

MD

MOT

26
Figure 2.
Direct Mediation Model

A E A E

Standard likelihood-based calculations using the MISSING command in Mplus
were used to account for individuals who are missing on MOT. Essentially, this
command splits the data into groups based on missing variables (i.e. MOT), one group
with the missing variable (MOT treated as a latent variable), and the other group with
complete data (MOT is an observed variable) and provides estimated means and
correlations adjusted for missing data. A basic assumption of likelihood-based
calculations is that the incomplete data are missing at random (MAR). The additional
information added from people with incomplete data is useful, and provides more data
to test the direct effects (of MD with AD) and less biased estimates.
Males and females were found to differ on prevalence of AD, MD and on MOT
scores, with males having greater rates of AD and higher MOT scores, and females
having greater rates of MD. Because of these differences, and prior evidence that males
and females differed in the strength of the association between AD and MOT, the
MD

AD

MOT

27
individual analyses were conducted separately by gender. In the twin models, means,
thresholds, variances and covariances were allowed to vary across males and females.
2.5 Model Fitting
Twin models were fit directly to the raw data using Mplus version 4 (Muthen
and Muthen, 2006). Twin models were based on pair-wise data and included
individuals from both complete and incomplete pairs. The Missing command was used
to allow information from individuals missing on MOT to be included in the analyses.
Because this model evaluates mediation using three variables and estimates the
relation between all three, the baseline mediation model is fully saturated and provides
a perfect fit to the covariances. By selectively forcing different paths between MD, AD
and MOT to 0, the degree of misfit for different models can be investigated using the
DIFFTEST option in Mplus. For example, if forcing the regression of AD on MD to 0
does not lead to a significant misfit, but forcing the regression of AD on MOT, and
MOT on MD to 0 does, there is complete mediation. If all paths lead to a significant
misfit when selectively forced to 0, all three paths are needed, which signifies partial
mediation. On the other hand, if forcing the regression of AD on MD to 0 leads to a
significant misfit, and forcing the regression of MOT on AD and the regression of MD
on MOT to 0 does not, this indicates that MOT does not mediate the MD-AD relation.
2.6 Twin Models
Standard threshold liability models were used to estimate the genetic and
environmental contributions to twin-pair resemblance for their liability to MD and AD.
Liability is an inferred trait that is assumed to be continuous and normally distributed,

28
and those who exceed a theoretical threshold are affected. Individual differences in
liability arise from three sources: additive genetic variation (a) from genes whose allelic
effects combine additively; common environment (c) includes all shared environments
that make family members more similar, and specific environment (e) which includes
all remaining environmental factors and measurement error. Monozygotic (MZ) twins
are similar because they share all of their genetic and shared environmental factors,
whereas dizygotic (DZ) twins resemble each other because they share on average half
of their genetic variation and all of their shared environment. Comparing the
resemblance of MZ and DZ twins allows for the estimation of each source’s
contribution to individual differences in liability to a disorder, or to the covariance
between disorders.
Twin studies have several assumptions including: random mating, equal
environment assumption (EEA), and additivity and independence of genetic and
environmental components. Random mating assumes that when choosing a mate,
individuals do not select a mate with a similar genetic history. The failure of this
assumption is known as assortative mating and leads to incorrect genetic and
environmental estimates. If parents are genetically correlated, their DZ twin offspring
will be correlated more than the predicted 0.50, and results from twin models will
underestimate the amount of genetic influence and overestimate the amount of shared
environment influence. Although there is evidence that assortative mating exists for
MD and AD (e.g., Maes et al., 1998), it does not seem to be a large enough effect to
substantially bias the estimates in the current analyses.

29
The equal environment assumption refers to the assumption in twin studies that
the shared environment of MZ pairs is no different than that of DZ pairs. If this
assumption is not true, and MZ pairs have a more similar shared environment than DZ
pairs, genetic influences will be overestimated. Previous studies using the VATSPSUD
sample examined the EEA by testing whether, among same-sex pairs, similarity for
lifetime MD and AD was associated with the twin pair’s similarity of childhood and
adolescent environmental experiences including childhood treatment and co-
socialization (Kendler & Gardner, 1998). The results indicated that differential
environmental experiences of MZ and DZ twins are unlikely to represent a substantial
bias in twin studies of MD and AD, findings that support the EEA.
The third assumption is additivity and independence of genetic and
environmental factors. In a traditional twin design without additional information on
family members, assumptions of additivity and independence cannot be evaluated.
Studies of alcohol consumption in extended families suggest that gene-environment
correlations only contribute to a small amount of variance in frequency of alcohol use
(Maes, Neale, Martin, Heath & Eaves, 1999).
The logic underlying the univariate and bivariate models was extended to
trivariate models by dividing the covariance between MOT, MD, and AD into genetic
and environmental components (See Figure 2). In these multivariate model, the first set
of factors represent factors shared between MOT, MD and AD (A1, E1). The second
set of factors represents factors shared only between to MD and AD (A2, E2), and the
third set of factors represents factors unique to AD (A3, E3). Because the ordering of

30
variables in this model will influence the interpretation of the results, the AE of MOT
was placed first in order to see how much shared variance remains between MD and
AD after taking out the part shared with MOT. Cross-twin MD-AD correlations were
computed with and without adjusting for covariates (See Appendix B). There was no
effect of covariates on cross-twin MD-MD and MD-AD correlations. For cross-twin
AD-AD and within twin MD-AD correlations, the correlations changed with the
inclusion of covariates at most 0.04. Given the small influence of covariates on the AD-
MD correlations, and the complexity of adding covariates to the models, twin models
were fit without covariates.

31
Chapter 3: Results
3.1 Descriptive Statistics
Descriptive statistics by sex for AD, MD, MOT, and demographic variables are
listed in Table 1. Prevalence rates for lifetime MD were significantly lower in males
compared to females (29.4% vs. 35.9%), prevalence rates for lifetime AD were
significantly higher in males compared to females (25.1% vs. 9.2%), and MOT scores
were higher in males than females (1.98 vs. 1.61).
Table 1.
Means, Standard Deviations and Range AD, MD, MOT.

Variable Mean Std. Dev. Range

Age
Males 41.1 7.0 30-58
Females 40.2 6.8 30-62
Years Education
Males 13.4 2.8 2-20
Females 13.8 2.4 3-20
Family Income (1000s)
Males 56.4 36.8 0-230
Females 55.5 36.8 0-230
MOT
Males 1.98 2.10 0-7.6
Females 1.61 1.85 0-7.7
MD
Males 29.4%
Females 35.9%
AD
Males 25.1%
Females 9.3%

Correlations of AD with MD, MOT, and demographic variables by sex are
shown in Table 2. AD was significantly correlated with MD and MOT for both males
and females. Logistic regression was used to predict AD from MD and various

32
demographic variables. Age-squared, education, and family income were significant
predictors of AD. Next, interaction terms for MD with demographic covariates were
added to the model and no MD-demographic interactions significantly predicted AD in
either gender. Logistic regression predicting AD from religious denomination, MD, and
the four denomination by MD interactions indicated that none of the interactions were
significant, which suggests that denomination did not alter the association between MD
and AD. There was no effect of religious denomination after accounting for other
covariates and religious denomination was not included in subsequent analyses.
Correlations of key variables by zygosity are shown in Table 3. MZ pairs were more
correlated than DZ pairs for all main variables indicating genetic influences. Cross-twin
MD-MOT and AD-MOT correlations were significant across all zygosity groups.
Table 2.
Correlations between key variables and covariates by Gender.

Ed Family
Income

Age

Age
2

MD

AD

MOT
Ed

0.31
(0.01)
-0.10
(0.02)
0.09
(0.02)
-0.04
(0.02)
-0.06
(0.03)
-0.04
(0.02)
Family
Income
0.41
(0.01)
0.06
(0.02)
-0.08
(0.02)
-0.08
(0.02)
-0.14
(0.03)
-0.05
(0.02)
Age

-0.09
(0.02)
0.14
(0.02)
0.99
(0.02)
-0.001
(0.03)
-0.13
(0.04)
-0.08
(0.03)
Age
2

-0.08
(0.02)
0.05
(0.01)
0.99
(0.02)
0.09
(0.03)
-0.15
(0.04)
-0.09
(0.02)
MD

-0.03
(0.02)
-0.05
(0.02)
-0.05
(0.02)
0.03
(0.02)
0.44
(0.04)
0.22
(0.03)
AD

-0.18
(0.03)
-0.17
(0.03)
-0.10
(0.03)
-0.09
(0.03)
0.33
(0.03)
0.42
(0.03)
MOT

-0.10
(0.02)
-0.12
(0.02)
-0.07
(0.02)
-0.06
(0.02)
0.34
(0.02)
0.45
(0.03)

Note: Values shown are pearson, tetrachoric, and polychoric correlations and
(se). Females (N=2462) are above diagonal and males (N=3044) below.

33
Table 3.
Correlations for Cross-Twin Lifetime MD, AD and MOT.

Variable Cross Twin Correlations

Male MZ
Pairs

Male
DZ Pairs
Female
MZ Pairs
Female DZ
Pairs
Opposite-Sex
DZ Pairs
M F

N 661 623 449 373 1238
MOT

0.38
(0.002)
0.14
(0.002)
0.43
(0.002)
0.09 (0.002) 0.15
(0.002)
MD

0.26
(0.08)
0.23 (0.09) 0.46
(0.07)
0.08 (0.05) 0.13
(0.06)
AD

0.61
(0.05)
0.36
(0.002)
0.65
(0.10)
0.20 (0.18) 0.17
(0.08)
MOT -
MD
0.19
(0.01)
0.11 (0.01) 0.14
(0.01)
0.03 (0.01) 0.13
a
(0.01)

0.05
b
(0.01)
MOT -
AD

0.37
(0.01)
0.20 (0.01) 0.43
(0.01)
0.11 (0.02) 0.04
a
(0.01)
0.11
b
(0.01)
MD-AD

0.22
(0.06)
0.21 (0.07) 0.20
(0.09)
0.06 (0.10) 0.13
a
(0.08)

0.18
b
(0.06)

3.2 Bias from Incomplete Data

Possible biases due to incomplete data were tested in several ways using SAS
Version 9.2. Overall, the results of these analyses provided some evidence of bias.
Demographic variables were examined to predict missing on the MOT variable of the
AUI among non-abstainers (complete abstainers cannot have motives for drinking)
using multiple logistic regression. Among females, those from the FF study were more
likely to be missing compared to females from the MF study (OR=4.3, p<.0001), and
those with lower education (OR=1.12, <.0001, for each additional year) were more
likely to be missing on MOT. It is likely that the additional effort of returning the AUI
via mail in the FF study, rather than in person in the MF study, contributes to the

34
differential completion of the AUI in females across studies. None of the variables
significantly predicted AUI questionnaire return in males.
Next, bias from incomplete twin pairs was evaluated by comparing MOT scores
and diagnoses of participants from complete pairs were compared to those from
subjects whose cotwin did not participate. Subjects from complete pairs did not differ
significantly from subjects whose twin did not participate on AD, MD or MOT scores.
Finally, a logistic regression analysis of AD on MOT and MD scores, with a main
effect of complete status (complete pair vs. incomplete pair) and an interaction of
completion status and MOT was conducted to see if individuals from complete pairs
had different risk for AD than incomplete pairs, and to see whether complete status
moderated the AD-Motive relation. There was not a significant main effect of complete
status or a significant interaction for complete status.
3.3 Phenotypic Mediation Analyses
Phenotypic analyses included all participants regardless of whether their co-
twins participated in the study. In order to test the mediation hypothesis, the
phenotypic relations among MD, AD and MOT were investigated after adjusting for
covariates. History of MD was positively associated with AD and higher MOT scores,
and MOT was positively related to AD. Several different models were tested to
evaluate mediation. Table 4 presents the results from a series of regression analyses in
Mplus. In the baseline model (Model I), AD was regressed on MD, AD was correlated
with MOT, and MOT and MD were correlated. This is a fully-saturated model and fits
perfectly. Parameter estimates from this model are provided in Table 5. The

35
correlations between MD and MOT, and MOT and AD were significant for males and
females (p<0.001). MD was a significant predictor of AD for males (b=0.34, p<0.001)
and females (b=0.44 p<0.001). Next, model Ia tested whether the estimates of the MD-
AD-MOT associations for this baseline model could be equated for males and females.
This significantly worsened the model fit, with a chi-square difference test in Mplus of
χ
2
(3, 5506) = 15.8, p< 0.0013.
Table 4.
Model Fitting Results for Phenotypic Mediation Analyses (N=5506).

Difference Tests

Chi-
Square
(DF)
Model
comparison

χ χ χ χ
2

(DF)

P-
Value

I. Baseline Correlational Model
a) Correlational 0.00 (2)
b) Correlational (M=F) 14.7 (5) Ib. vs Ia. 15.76
(3)
0.0013
II. No mediation
a) MD→MOT@0 252.3 (4) IIa vs III 252.3
(2)
0.001
b) AD→MOT@0 345.6 (4) IIb vs III 413.6
(2)
0.001
c)
MD→MOT@0/AD→MOT@0
916.5 (6) IIc vs III 931.2
(4)
0.001
III. Partial Mediation 0.00 (2)
IV. Complete Mediation 133.3 (9) IV. vs III. 135.7
(7)
0.001

36
Table 5.
Standardized Estimates from Mediation Analyses (N=5506).

Test AD-MD
Path Males Females
Model I Est (SE) Est (SE)
c
dm
MD↔MOT

0.34 (0.02) 0.22 (0.03)
c
ma
MOT↔AD

0.42 (0.03) 0.43 (0.03)
b
da
MD→AD

0.33 (0.03) 0.44 (0.04)
Model III
b
dm
MD→MOT

0.34 (0.03) 0.22 (0.03)
b
ma
MOT→AD

0.45 (0.03) 0.43 (0.03)
b
da
MD→AD

0.17 (0.03) 0.33 (0.04)

Various models were tested to see if any of the parameters could be dropped by
selectively forcing each regression to 0. Forcing b
da
to 0 worsened the model fit, with a
chi-square difference test of χ
2
(7, 5506) = 135.7, p< 0.001, which indicated that there
is not full mediation. Forcing both b
dm
and b
ma
to 0 did not fit the data well compared to
the partial mediation model (Model III) χ
2
(4, 5506) = 931.2, p< 0.001, suggesting that
MD is significantly related to MD. Independently forcing b
dm
and b
ma
also fit the data
poorly (p<0.001). The results from Models IIa-IIc (Table 4) provided evidence for
partial mediation, and all 3 paths were necessary for good model fit. Next, estimates
from the baseline model were compared to estimates from the partial mediation model
of AD regressed on all 3 variables (Model III, Table 5). There was a significant
reduction in b
da
after regressing AD on MOT, and MOT on MD (rather than simply
allowing them to be correlated). For males, b
da
was reduced from 0.33 to 0.17, and for

37
females b
da
was reduced from 0.44 to 0.33 (See Figure 3). For males, 11% of the
variance in AD was explained by MD before regressing AD on MOT. After accounting
for the variance in AD due to MOT, less than 3% of the variance in AD was explained
by MD (a reduction of 73%). For females, 19% of the variance in AD was explained by
MD before adding the regression of AD on MOT. After accounting for the variance in
AD explained by MOT, only 11% of the variance in AD was explained by MD (a
reduction of 42%).

38
Figure 3.
Phenotypic Mediation Results.
Males.

0.33

0.34 0.42

0.17

0.34 0.45

Females.
0.44

0.22 0.43

0.33

0.22 0.43

The Mplus MODEL INDIRECT statement was used to further assess the
indirect effect of MOT on the MD-AD relation. A significant indirect effect was found
for MD on AD via MOT for males (beta= .17, p<0.001) and females (beta=0.10,
p<0.001), confirming significant mediation of the MD-AD relation by MOT.

MD
MOT
AD
MD
MOT
AD
MD
MOT
AD
MD AD
MOT

39
3.4 Twin Models
Univariate structural equation models were used to provide estimates of the
proportions of variance attributable to genetic, shared environmental, and/or individual
specific environmental influences for each key variable. Tables 6-8 provide the fit
indices and sequential chi-square difference tests for each univariate nested model, as
well as the genetic, shared environmental, and individual specific estimates for the full
and best fitting models.
Table 6.
Results from Univariate Twin Analyses of MOT (N=3344).

Model Fit Difference Tests
Chi-Square (DF) P-value χ χ χ χ
2
(DF) Comparison

I. ACE (free r
a
)

22.5 (16) 0.13
II. ACE

23.4 (17) 0.14 0.9 (1) II. vs. I.
IIa. ACE (M=F) 50.4 (20) 0.002 27.0 (3) IIa. vs II.
III. AE 24.4 (19) 0.18 1.0 (2) III. vs II.
IV. CE 53.1 (19) 0.00 29.7 (2) IV. vs II.
V. E 136.4 (21) 0.00 54.4 (4) V. vs II.

Parameter Estimates
Full Model
II. a a
2
c c
2
e e
2

Males
Females
0.58 (0.13) 34% 0.18 (0.33) 3% 0.79 (0.06) 62%
0.62 (0.11) 38% 0.15 (0.30) 2% 0.77 (0.06) 59%
Best Fitting
Model

III. a a
2
c c
2
e e
2

Males
Females
0.60 (0.08) 36% 0.00 0% 0.80 (0.06) 64%
0.63 (0.08) 39% 0.00 0% 0.78 (0.06) 61%

40
Table 7.
Fit Indices for MD Univariate Nested Models (N=3344).
Model Fit Difference Tests
Chi-Square (DF) χ χ χ χ
2
(DF) P-Value Comparison

I. ACE (free r
a
)

34.9 (8)
II. ACE

34.9 (9) 0.07 (1) 0.79 II. vs. I.
IIa. ACE (M=F) 37.9 (12) 3.27 (3) 0.35 IIa. vs II.
III. AE 36.2 (11) 1.37 (2) 0.50 III. vs II.
IV. CE 46.5 (11) 11.5 (2) 0.003 IV. vs II.
V. E 96.3 (13) 60.8 (4) 0.00 V. vs II.
Parameter Estimates
Full Model
II. a a
2
c c
2
e e
2

Males 0.35 (0.34) 12% 0.39 (0.25) 15% 0.85 (0.06) 72%
Females 0.65 (0.08) 42% 0.04 (0.31) 0% 0.76 (0.07) 58%
Best Fitting
II. a a
2
c c
2
e e
2

Males 0.53 (0.05) 28% 0.00 0% 0.85 (0.04) 72%
Females 0.65 (0.05) 42% 0.00 0% 0.76 (0.04) 58%

Table 8.
Fit Indices for AD Univariate Nested Models (N=3344).
Model Fit Difference Tests
Chi-Square (DF) χ χ χ χ
2
(DF) P-Value Comparison
I. ACE(free r
a
) 20.8 (8)
II. ACE 20.9 (9) 0.2 (1) 0.64 II. vs. I.
IIa. ACE (M=F) 23.5 (12) 2.8 (3) 0.49 IIa. vs II.
III. AE (free r
a
) 21.0 (10) 0.35 (2) 0.84 III. vs II.
IIIa AE 23.8 (11) 2.9 (2) 0.23 IIIa. vs II.
IV. CE 45.7 (11) 24.3 (2) 0.00 IV. vs II.
V. E 190.4 (13) 166.8 (4) 0.00 V. vs II.
Parameter Estimates
Full Model
II. a a
2
c c
2
e e
2

Males 0.68 (0.16) 46% -0.38 (0.24) 14% 0.63 (0.05) 39%
Females 0.77 (0.09) 59% 0.20 (0.28) 4% 0.61 (0.08) 37%
Best Fitting (III)
Males 0.79 (0.05) 62% 0.00 -- 0% 0.62 (0.05) 38%
Females 0.80 (0.06) 64% 0.00 -- 0% 0.60 (0.07) 36%

41
Starting with the univariate twin analyses of MOT, a full ACE model that
allowed the genetic correlation of opposite twins (r
a
) to be estimated (Model I in Table
6) was compared to a full ACE model that forced r
a
to 0.5 (II). The misfits in Model I
are associated with equating across zygosity and across twin 1 and 2 within sex, and
indicate variability, but are not significant overall. Fixing opposite sex twin pairs at 0.5
provided a more parsimonious fit to the data, χ
2
(1, 3344) = 0.90, p<0.14, and this
model was considered the baseline model for comparisons with subsequent models.
Next, males and females were equated for the full ACE model (IIa), which provided a
significantly worse fit compared to model II, χ
2
(3, 3344) = 27.0, p<0.01. Next, the fit
of an AE model (III) was evaluated after dropping C from the model. The AE model
provided a significantly better fit compared to model II, χ
2
(2, 3344) = 1, p<0.18.
Finally, to evaluate the power of this sample to reject alternative hypotheses, models
were fit that dropped the genetic path (IV, CE), and both the genetic and shared
environmental paths (V, E Only); these both fit poorly. The best fitting model (III)
included genetic and individual specific environment and allowed males and females to
differ. Based on the estimates from the best fitting model for MOT, 36% and 39% of
variation in MOT was attributable to additive genetic factors in males and females,
respectively. The remaining 64% and 61% were attributable to specific environmental
factors and measurement error.
These model-building steps were repeated for MD and AD (Tables 7 & 8). The
misfits in baseline models are associated with equating across zygosity and across twin
1 and 2 within sex, and indicate variability, but are not significant overall. As with

42
MOT, individual differences on these variables were explained largely by additive
genetic and individual specific influences. Dropping the common environment path
provided a more parsimonious fit to the data for each univariate model. The best fitting
univariate model for MD (III) included individual specific environment and additive
genetic paths and allowed for sex differences. About 28% and 42% of variation in MD
was attributable to additive genetics in males and females, and 72% and 58%
attributable to individual specific environment for males and females. The best fitting
univariate model for AD (III) included individual specific environment and additive
genetic factors, allowed the genetic correlation to be free in opposite sex pairs (r
a
,
estimated at 0.27), and also allowed for sex differences in the variance proportions.
Approximately 62% and 64% of variance in AD was attributable to additive genetics in
males and females, and 38% and 36% to individual specific environment in males and
females. The estimates for males and females are very close and could be equated
without much loss of fit, but in the context of a free genetic correlation, it’s not very
meaningful to test a model that equates males and females.
Next, bivariate models were fit to the pair-wise data. First, the fit indices and
sequential chi-square difference tests for MOT-AD, and MD-AD bivariate nested
models were compared (See Table 9). Three models were compared to determine the
best fit. The first model allowed for sex differences and free r
a
for AD based on the
best-fitting univariate model (Model A). This did not fit the data as well as the
bivariate models with r
a
fixed at 0.5 (Model I). Therefore, the full model (ACE) with
sex differences was used as a baseline model. Based on the results for the univariate

43
models, models that dropped all common environment (c) paths were fit next. Based
on the chi-square difference tests, Model II fit best for MOT-AD, χ
2
(4, 3344) = 3.79,
p< 0.43, and for the MD-AD bivariate models, χ
2
(6, 3344) = 6.05, p< 0.42. Estimates
from the best fitting MOT-AD model are provided in Table 10. The first set of factors
represents factors that contribute to both MOT and AD (A1, E1). The second set of
factors is specific to AD (A2, E2). In males, there was a considerable contribution to
AD from the genetic factors underlying MOT (0.61), as well as from genetic factors
specific to AD (i.e., not shared with MOT) (0.49). In addition, there was an overlap of
AD individual specific environment factors with MOT (0.22) as well as AD specific
individual specific environment (0.59). The pattern was similar for females (See Table
10). The Total Variance is the percentage of influences due to a or e, calculated by
squaring and summing the estimates in each row. For example, a
2
for AD is the total
percentage of variance attributable to any of the genetic factors. In table 10, the total
estimate of a
2
for AD for males is (0.61
2
+ 0.49
2
) = 61% and the total estimate for
females is (0.57
2
+ 0.52
2
) = 59%.

44
Table 9.
Fit Indices for MOT-AD and MD-AD Bivariate Nested Models (N=3344).

MOT-AD MD-AD

Model

Comparison

Chi-
Sq
(DF)
Difference Test

χ χ χ χ
2
(DF) P-
Value

Chi-
Sq
(DF)
Difference Test
χ χ χ χ
2
(DF) P-
Value

A. Full
Model, AD
ra free
38.1
(24)
61.9
(29)

I. Full
Model
I. vs. A. 37.5
(24)
0.1 (1) 0.73 62.4
(30)
0.4 (1) 0.53
II. No
Common
Environment
II. vs. I. 39.6
(27)
3.8 (4) 0.43 66.5
(35)
6.1 (6) 0.42

Table 10.
Parameter Estimates from Baseline Bivariate MOT-AD Model (N=3344).
Males
Factor 1 (Overlapping) Factor 2 (AD Specific) Total Variance
A1 A2 a
2

MOT 0.59 (0.01) 35%
AD 0.61 (0.06) 0.49 (0.06) 61%
E1 E2 e
2

MOT 0.81 (0.01) 65%
AD 0.22 (0.05) 0.59 (0.05) 39%
Females
A1 A2 a
2

MOT 0.61 (0.01) 37%
AD 0.57 (0.09) 0.52 (0.10) 59%
E1 E2 e
2

MOT 0.79 (0.01) 63%
AD 0.24 (0.07) 0.59 (0.07) 41%

Estimates from the best fitting bivariate MD-AD model (II in Table 9) are
shown in Table 11. For males, there was a moderate overlap of AD genetic factors
with MD (0.45), as well as genetic factors specific to AD (0.64). There was also a small

45
overlap of AD individual specific environmental factors with MD (0.12) and individual
specific environment not shared with MD (0.62) (See Table 11). The estimated total of
genetic influences on AD was 61%. For females, there was a small overlap of AD
genetic factors with MD (0.29), in addition to genetic factors specific to AD (0.72).
There was also an overlap of AD individual specific environmental factors with MD
(0.32), and individual specific environment not shared with MD (0.54). The estimated
total of genetic influences on AD is 60%.
Table 11.
Model Fitting Results from Best Fitting Bivariate MD-AD AE Model.
Variable Factor 1
(Overlapping)
Factor 2 (AD
Specific)
Total Variance
Males
A1 A2 a
2

MD 0.51 (0.07) 26%
AD 0.45 (0.09) 0.64 (0.07) 61%
E1 E2 e
2

MD 0.86 (0.05) 74%
AD 0.12 (0.06) 0.62 (0.05) 39%
Females
A1 A2 a
2

MD 0.65 (0.07) 42%
AD 0.29 (0.12) 0.72 (0.08) 60%
E1 E2 e
2

MD 0.76 (0.07) 58%
AD 0.32 (0.10) 0.54 (0.09) 40%

Several possible multivariate twin models were fit to test the mediation
hypothesis (See Table 12). First, a full trivariate model that estimated all possible
unique and common parameters was fit, and allowed r
a
for alcohol to be free (Model A)
based on the univariate AD model (r
a
was estimated as 0.502). This model was
compared to a full trivariate model that estimated all possible unique and common

46
parameters and fixed r
a
at 0.5 (Model I). Model I fit the data better than model A (χ
2
(1,
3344) = 2.66, p< 0.10) and served as the baseline model for the difference tests in
Mplus (Model I in Table 12). The misfit in this model arises from equating twin 1 and
twin 2 within pairs and across zygosity groups for means, thresholds, and variances,
and is not related to the hypotheses of interest. Next, male and females were equated on
the unique and common parameters underlying MOT, MD, and AD (Model Ia). This
did not fit the data well χ
2
(9, 3344) = 32.8, p< 0.01. Next, a No Overlap model that
forced the overlapping paths of Factor 1 and Factor 2 to 0 was fit, representing the
hypothesis of no shared sources of variation across variables (Model II), in order to see
how much power there was to reject the no overlap hypothesis. As expected, this
provided a very bad fit to the data χ
2
(10, 3344) = 999.7, p< 0.01, indicating that there
is significant shared variance between variables.
Seven other models were fit to test for the sources of this overlap. Model III
forced the shared factors underlying the MD→MOT relation to 0. This fit the data
poorly, χ
2
(5, 3344) = 275.5, p< 0.01. Full mediation (Model IV) was tested by forcing
the genetic and environmental MD→AD estimates to 0, and this fit worse than the
baseline model, χ
2
(5, 3344) = 29.3, p< 0.01. Models IVa and IVb tested two versions
of partial mediation. Full mediation of the genetic overlap (Model IVa) was tested by
forcing the common genetic MD-AD estimate to 0, χ
2
(3, 3344) = 7.6, p< 0.05. Full
mediation of the environmental overlap (Model IVb) was tested by forcing the specific
environment MD-AD estimate to 0, and did not fit as well as the baseline model, χ
2
(3,
3344) = 14.5, p< 0.002. Of Models IV-IVb, full mediation of the genetic overlap

47
provided the most parsimonious fit to the data, χ
2
(3, 3344) = 7.6 compared to Model I,
p< 0.05. The “best” fitting model depends on the fit criterion used. Based on chi-square
it’s borderline fit (p=0.05), and based on AIC criteria, any χ
2
change greater than 6 for
3 degrees of freedom is a worse fit. The full model was chosen (rather than full
mediation of the genetic overlap) as the best fitting model based on the AIC criteria.
Based on the parameter estimates obtained in Model IVa, two additional partial
mediation models were added. The first tested complete mediation of the genetic MD-
AD overlap in males (Model IVc), and the fit was χ
2
(2, 3344) = 8.3, p< 0.02 (Table
12). The second tested complete mediation of the genetic MD-AD overlap in females
(Model IVd) and the fit was χ
2
(2, 3344) = 5.7, p< 0.06. Finally, a more stringent
direct mediation model was tested that directly regressed AD on MOT, and MOT on
MD, rather than regressing AD onto the factors underlying MOT, and the factors
underlying MOT onto MD (see Figure 1). The model fitting results of this direct
mediation model indicated a worse fit compared with the correlated causes mediation
model, χ
2
(5, 3344) = 11.3 (Table 12).

48
Table 12.
Model Fitting Results for Multivariate Twin Models for MD, AD and MOT.

Chi-Sq #Par RMSEA
Difference Test
χ χ χ χ
2
(DF) P-Value

A). Full Model (r
a
free for AD) 76.0 28 0.024
(r
a
=0.502)

I. Full Model

76.8 27 0.024

2.66 (1) 0.10
I. vs. A.
Ia. Full Model (M=F)

104.8 14 0.033 32.8 (9) 0.001
Ia. vs. I.
II. Null Model (no overlap)
(a
1md
, e
1md
, a
1a
, e
1a
, a
2a
, e
2a,
@0)
982.6 14 0.149

999.7 (10) 0.001
II.vs. I.
III. No MD-MOT overlap
(a
1md
, e
1md,
@0)
270.2 22 0.073 275.5 (5) 0.001
III. vs. I.
IV. No MD-AD overlap Full
Mediation (a
2a
, e
2a,
@0)
92.9 22 0.030 29.3 (5) 0.001
IV. vs. I.
IVa. No Genetic MD-AD
overlap (Full Mediation of A
Overlap, a
2a
=0)

81.5 24 0.025 7.6 (3) 0.05
Iva. vs. I.
IVb. No specific Environ.
MD-AD overlap (Full
Mediation of E Overlap, e
2a
=0)

85.4 24 0.027 14.5 (3) 0.002
IVb. vs. I.
IVc. No genetic MD-AD
overlap Males (male a
2a
=0)
80.9 25 0.026 8.3 (2) 0.02
IVc. vs. I.
IVd. No genetic MD-AD
overlap Females (female
a
2a
=0)
79.9 25 0.025 5.7 (2) 0.06
IVd vs. I.
V. Direct Mediation

88.06 22 0.028 11.3 (5)* V. vs. I.

Table 13 provides estimates from the best fitting twin mediation model (I) and
Figures 5 & 6 portray the parameter estimates for males and females. In males, the
genetic loadings (a1
m
, a1
d
, a1
a
) from the genetic factor shared between all three
variables (Factor A1) were moderate to small, MOT (0.60), MD (0.30), and AD (0.60).
This factor accounted for 33% of the genetic variance in MD and 59% of the genetic
variance in AD. The genetic factor shared by MD and AD (Factor A2) loaded on MD

49
(0.42) and AD (0.09). This factor accounted for less than 2% of the genetic variance in
AD. This indicates that the genetic variance common to MD and AD was almost
entirely carried by the common factor that influences MOT, MD and AD. The genetic
influence unique to AD (Factor A3) was moderate (0.49), indicating that there are
genetic factors specific to AD.
Table 13.
Best Fitting Multivariate Model (I): Standardized Parameter Estimates.
Variable Factor 1
(MOT,MD,AD)
Factor 2
(MD,AD)
Factor 3
(AD specific)
Total Estimate
Males
A1 % A2 % A3 % % a
2

MOT 0.60 (0.08) 36 36%
MD 0.30 (0.06) 9 0.42 (0.08) 18 27%
AD 0.60 (0.06) 36 0.09 (0.10) 1 0.49 (0.06) 24 61%
E1 % E2
%
E3
%
e
2

MOT 0.80 (0.06) 64 64%
MD 0.20 (0.05) 4 0.83 (0.05) 69 73%
AD 0.21 (0.05) 4 0.09 (0.05) 1 0.58 (0.05) 34 39%
Females
A1 % A2 % A3 % % a
2

MOT 0.59 (0.09) 35 35
MD 0.27 (0.08) 7 0.58 (0.08) 34 41
AD 0.57 (.08) 32 0.13 (0.12) 2 0.52 (0.10) 27 61
E1 % E2 % E3 % % e
2

MOT 0.80 (0.06) 64 65
MD 0.09 (0.06) 1 0.77 (0.07) 59 58
AD 0.26 (0.07) 7 0.14 (0.08) 2 0.55 (0.09) 30 39

50
Figure 4a.
Males: No Mediation.

A E A E

0.51 0.86 0.64 0.62
0.45 0.12

Figure 4b.
Males: Mediation Via the Factors Underlying MOT.

A1 E1 A2 E2

0.80 0.20 0.21
0.60 0.30 0.60 0.42 0.09 0.83 0.09

0.49 0.58

A3 E3
Figure 5a.
Females: No Mediation.

A E A E

0.65 0.76 0.72 0.54
0.29 0.32

AD

MD

MOT

MD

AD

MD

AD

51
Figure 5b.
Females: Mediation Via the Factors Underlying MOT.

A1 E1 A2 E2

0.80 0.09 0.26
0.59 0.27 0.57 0.58 0.13 0.77 0.14

0.52 0.55

A3 E3

The individual specific loadings (e1
m
, el
d
, e1
a
) from the environment factor
shared between all three variables (Factor E1) ranged from large to small: MOT (0.80),
MD (0.20), and AD (0.21). Factor E1 accounted for about 5% of the environmental
variance in MD, and 10% of the environmental variance in AD. The environmental
factor common only to MD and AD (Factor E2) showed large and small loadings on
MD (0.83) and AD (0.09). After removing the variance due to Factor E1, Factor E2
accounted for less than 3% of the environmental variance in AD, suggesting that the
environmental MD-AD overlap in males stems primarily from a factor common to
MOT, MD and AD. MD and AD had only a small amount of shared individual specific
environment distinguishable from the environment they share with MOT. The
individual specific environment unique to AD was moderate (0.58) after accounting for
the environmental variance shared with MOT and MD.
AD

MD

MOT

52
In females, genetic loadings (a1
m
, a1
d,
a1
a
) from the genetic factor shared with
all three variables (Factor A1) were moderate to small, MOT (0.59), MD (0.27), and
AD (0.57). This factor accounted for 17% of the genetic variance in MD and 52% of
the genetic variance in AD. The genetic factor specific to MD and AD (Factor A2)
loaded on MD (0.58) and AD (0.13). This factor only accounted for 3% of the genetic
variance in AD. This indicates that the genetic variance shared between MD and AD
was almost entirely carried by the shared factor that influences MOT, MD and AD. The
genetic influence unique to AD (Factor A3) was moderate (0.52), indicating that there
are genetic factors specific to AD.
The individual specific loadings (e1
m
, el
d
, e1
a
) from the environment factor
shared with all three variables (Factor E1) ranged from large to small: MOT (0.80), MD
(0.09), and AD (0.26). Factor E1 accounted for about 2% of the environmental variance
in MD, and 18% of the environmental variance in AD. The environmental factor
specific only to MD and AD (Factor E2) showed large to small loadings on MD (0.77)
and AD (0.14). Factor E2 accounted for 5% of the environmental variance in AD,
suggesting that MD and AD have little shared individual specific environment
distinguishable from the environmental influences shared with MOT. The individual
specific environment unique to AD was moderate (0.55) after accounting for the
environmental variance shared with MOT and MD. The total proportion of variance
due to genetic and environmental factors is shown in the far right column of Table 13.
As expected, these totals correspond very closely to the variability ascribed to genetic
and individual specific influences in the univariate models.

53
Table 13.
Final Best Fitting MOT-MD-AD Model (I): Standardized Parameter Estimates and
Standard Errors.
Variable Factor 1
(MOT,MD,AD)
Factor 2
(MD,AD)
Factor 3
(AD specific)
Total
Estimate
Males
A1 % A2 % A3 % % a
2

MOT 0.60 (0.08) 36 36%
MD 0.30 (0.06) 9 0.42 (0.08) 18 27%
AD 0.60 (0.06) 36 0.09 (0.10) 1 0.49 (0.06) 24 61%
E1 % E2 % E3 % e
2

MOT 0.80 (0.06) 64 64%
MD 0.20 (0.05) 4 0.83 (0.05) 69 73%
AD 0.21 (0.05) 4 0.09 (0.05) 1 0.58 (0.05) 34 39%
Females
A1 % A2 % A3 % % a
2

MOT 0.59 (0.09) 35 35
MD 0.27 (0.08) 7 0.58 (0.08) 34 41
AD 0.57 (.08) 32 0.13 (0.12) 2 0.52 (0.10) 27 61
E1 % E2 % E3 % % e
2

MOT 0.80 (0.06) 64 65
MD 0.09 (0.06) 1 0.77 (0.07) 59 58
AD 0.26 (0.07) 7 0.14 (0.08) 2 0.55 (0.09) 30 39

54
Chapter 4: Discussion
4.1 General Discussion
Drinking to manage mood has been widely studied, yet few investigators have
expanded the role of drinking motives to examine the relation between depression and
alcohol dependence. Both depression and drinking to manage mood share genetic risk
with alcohol dependence, yet no study has examined the role of mood-related drinking
motives as a mediator of the familial association between depression and alcohol
dependence. The evidence that some of the depression-alcohol dependence overlap was
familial, and that some of the drinking motives-alcohol dependence overlap was
familial, raises the question of whether drinking motives mediate the familial
depression-alcohol dependence overlap. The goals of this research were twofold: 1) to
examine the impact of drinking to manage mood on the relation between depression
and alcohol dependence within individuals, and 2) to estimate the genetic and
environmental contributions to depression-alcohol dependence comorbidity accounted
for by mood drinking motives.
First, it was hypothesized that drinking to manage mood partially mediates the
association between depression and alcohol dependence within individuals. This
hypothesis was based on prior research that demonstrated that coping motives mediate
the relation between expectancies and consumption problems in both community and
treatment samples (Cooper et al., 1995; Galen et al., 2001). Results indicate that even
after adjusting for the effects of education, income and age, drinking to manage mood
mediates the depression-alcohol dependence relation in both males and females. These

55
findings lend some support to research by Cooper et al. (1995), which found drinking
for coping reasons mediated the relation between drinking and the experience of
negative emotion. Similarly, Peirce et al. (1994) found that drinking to manage
negative emotions partially mediated the relation between depressed mood and alcohol
consumption; however, this is the first study to directly assess the role of drinking to
manage mood as a mediator of DSM-IV depression and alcohol dependence diagnoses.
Although the depression-alcohol dependence association is reduced by 8% in
both genders, proportionally drinking to manage mood is a stronger mediator of the
depression-alcohol dependence relation for males: 73% of the overlapping variance is
attributable to the overlap with drinking to manage mood for males, and 42% of the
overlapping variance is attributable to the overlap with drinking to manage mood for
females. These results are in line with some previous findings indicating that the risk of
problem drinking associated with drinking motives is greater for men than women
(Ratliff & Burkhart, 1984; Klein, H., 1992; Pierce et al., 1994; Cooper et al., 1995).
However, other investigators have found that alcoholic women are more likely to use
alcohol to alter mood than alcoholic men (Olenick & Chalmers, 1991; Timko, Finney
& Moos, 2005), and that drinking to regulate affect is a stronger predictor of drinking
frequency in women than in men (Mooney et al., 1987).
There are several potential mechanisms behind the gender differences in
mediation. The degree to which an individual drinks alcohol to regulate emotion is
partially determined by the availability of alterative emotion regulation strategies.
Depressed males might be more likely to use alcohol to regulate their negative affect

56
than depressed females if they lack more adaptive strategies for emotion regulation.
Because seeking treatment for depression may be more socially acceptable for females
(Moller-Leimkuhler, 2002), drinking to cope with negative emotions may not be as
fundamental a mediator of the depression-alcohol dependence relation for females.
Although studies indicate that individuals who drink alcohol to regulate and
improve their affective state drink more frequently and are at a higher risk for
developing alcohol dependence than individuals without this motivation, it is also likely
that for some individuals the relation works in the opposite direction. Given that males
tend to develop alcohol dependence before the onset of depression, it is possible that
when males first begin experiencing symptoms of depression they drink to regulate
mood, which leads to primary alcohol dependence (and eventual depression when self
medication is ineffective). Depression could also be a result of neurobiological
adaptations or physiological changes due to chronic alcohol use, and problematic
drinking can lead to disruptions in work, relationships and general functioning that may
eventually cause depression. Some research suggests that the social consequences of
alcohol dependence develop at a faster pace in females than in males (Diehl et al.,
2007). If this is the case, it is possible that the social consequences of drinking play a
greater role in the development of depression for females than for males. For example,
if females are more likely than males to experience social or work-related consequences
due to their drinking, it would make sense that they would be more likely to develop
depression. Especially given that most of the residual AD-MD overlap was
environmental after accounting for shared factors with MOT, it is possible that social

57
factors could explain the differential additional shared depression-alcohol dependence
variance not attributable to drinking to manage mood for females.
Given the multiple potential avenues to depression-alcohol dependence
comorbidity, and that individual differences in depression, alcohol dependence and
drinking to manage mood are partially heritable, a common genetic diathesis
influencing drinking motives, lifetime depression-and lifetime alcohol dependence was
assessed. If drinking to manage mood is genetically influenced and affects risk for
depression and alcohol dependence, it could create shared genetic variation for
depression and alcohol dependence. It was hypothesized that drinking to manage mood
would have a substantial genetic overlap with depression and alcohol dependence, and
partially explain some of the familial overlapping risk for co-occurring lifetime
depression and alcohol dependence.
As predicted, the familial factors underlying drinking to manage mood share
genetic variance with both MD and AD, and explain a large proportion of the shared
genetic risk for co-occurring lifetime MD and AD. In the bivariate MD-AD model,
33% of the genetic variance in AD overlaps with genetic influences on MD in males,
and 13% is shared with MD in females. In the trivariate model, after accounting for the
genetic factors shared among drinking motives, MD and AD, there is only 2% of
genetic variance in AD shared with MD in males, and 3% in females. That is,
accounting for the factors underlying motives explains 94% of the shared genetic risk
for MD and AD in males, and 77% of the shared genetic risk for MD and AD in
females. These results are an expansion on prior research in the VATSPSUD that found

58
genetic influences on drinking to cope with negative mood and with social anxiety, and
found evidence of overlapping genetic variation in drinking to alter mood and
alcoholism (Prescott et al., 2004).
The association between MD and AD is modestly influenced by non-shared
environmental sources overlapping with Motives. In the bivariate MD-AD model, 4%
of the individual specific environmental variance in AD is shared with MD in males,
and 25% is shared with MD in females. After accounting for the environmental factors
shared among drinking motives, MD and AD, there is only 3% of individual specific
environmental variance in AD shared with MD in males, and 5% in females. The
factors underlying Motives explain 25% of the shared environmental risk for MD and
AD in males, and 80% of the shared environmental risk in females.
Overall, the most noteworthy result is that the genetic and environmental factors
underlying drinking to manage mood are primarily the same factors responsible for the
overlap of depression and alcohol dependence. The results are not consistent with
drinking to manage mood acting as a direct cause of depression-alcohol dependence
comorbidity. Instead, the results indicate that drinking to manage mood may be a useful
index of vulnerability among individuals at risk for depression and alcohol dependence.
These findings have implications for prevention and intervention for alcohol use
disorders. Although there are many interventions that target challenging and changing
alcohol expectancies and drinking motives, reductions in alcohol expectancies and
drinking motives do not always lead to reductions in alcohol consumption (Jones,
Corbin, Fromme, 2001). Because the current results indicate an indirect familial effect

59
of drinking to manage mood, rather than a direct effect on risk for lifetime MD and AD,
interventions that target drinking motives would not necessarily directly lead to long-
term changes in risk for alcohol dependence. If drinking to manage mood is an index of
genetic vulnerability to alcohol dependence and comorbid depression, individuals who
report strong motivation to drink in order to manage mood could be seen as a
vulnerable population at risk for the development of depression and alcohol
dependence.
One possibility is that there is an inherited and/or learned tendency to avoid
pain and psychological discomfort. If individuals have a familial sensitivity to negative
affectivity, and find it hard to tolerate negative emotions, they would be more likely to
use avoidance coping to deal with feelings of depression. Because the degree to which
an individual drinks alcohol to regulate emotion is partially determined by the
availability of alterative emotion regulation strategies, interventions that provide
adaptive alternative methods for emotion regulation could benefit this at-risk
population.
Given the substantial genetic overlap between drinking to manage mood and
alcohol dependence, it is also likely that there is a genetic predisposition to experience
alcohol as physiologically pleasing, leading to increased likelihood of drinking in
response to negative mood. Interventions that attempt to challenge alcohol expectancies
and motives would be particularly difficult for these individuals. Perhaps a better
strategy would be to emphasize alternative, more adaptive, methods for experiencing

60
pleasure (e.g. participation in physical activity), in addition to tools for emotion
regulation and coping with negative affect.
Some of these interpretations are speculative and await replication.
Longitudinal, genetically informative studies that assess both positive and negative
motives for drinking are needed to clarify these explanations.
4.2 Strengths and Limitations
There are several strengths to the current study. This is the first study to assess
the role of drinking to manage mood as a mediator of DSM-IV depression-alcohol
dependence comorbidity. The twin design is especially instructive because it allows for
the estimation of the degree to which environmental and genetic factors contribute to
the overlapping risk for depression, alcohol dependence and drinking to manage mood.
By limiting the sample to individuals ages 30 and older, this study included only
individuals who are past the primary risk period for developing depression and alcohol
dependence. This increases the accuracy of the results, especially for the cross-twin
estimates in opposite-sex pairs (given the differential ages of onset of depression and
alcohol dependence across gender). In addition, studies based on the VATSPSUD are
useful because of the large genetically informative sample. Although the current sample
is restricted to Caucasian Virginia-born twins, Virginia is both culturally and
geographically diverse, suggesting that the results from our study can be generalized to
individuals from other regions.
There were several limitations to this study. One potential limitation is recall
bias and accuracy of self-reported lifetime MD and AD symptoms based on one

61
measurement occasion. However, test-retest reliability in this sample is good and
supports the reliability of the interview procedures. Another limitation is the emphasis
on clinical diagnoses. Although DSM-IV diagnoses are clinically informative, analyses
with binary MD/AD diagnoses overlook individuals who experience symptoms that do
not meet full diagnostic criteria.
Another conceptual difficulty is the lack of available drinking motives for
complete abstainers. Unlike alcohol expectancies, motives for alcohol use are only
measurable after an individual has experienced drinking and develops motives based on
experience. Because individuals who are lifetime abstainers from alcohol have not
developed drinking motives, they could not complete the AUI, and their Motives scores
could not be included in the analyses. In addition, approximately 576 people were
missing Motives scores because they did not return their AUI questionnaires. Although
questionnaire return was not associated with any of the demographic variables for
males, being in the FF study and lower levels of education were associated with not
returning the AUI for females. Missing AUI scores were accounted for in Mplus using
the Missing command to account for missing data. This procedure allowed us to use the
depression, alcohol dependence, and demographic information of those missing
Motives to reduce the possible bias due to questionnaire return.
Because the present study was cross-sectional, the current models cannot be
used to extrapolate to the temporal association between depression and alcohol
dependence. However, because drinking to manage mood is a genetically influenced
variable that shares risk for both depression and alcohol dependence, it is uncertain

62
even the best-designed prospective studies would be able to distinguish the direction of
effect. Given that having ever had an episode of depression is an indicator of an
underlying liability that is partially shared with drinking to manage mood and alcohol
dependence, it is not clear how important the temporal relation would be. For example,
if having ever had an episode of depression represents a proxy for trait depression,
knowing that depression occurred before alcohol dependence might be more of an
index of risk for alcohol dependence than a causal precipitant. Most likely, both
temporal causation (i.e. depression →alcohol dependence or alcohol
dependence→depression) and an underlying vulnerability to develop both disorders are
jointly important.
4.3 Conclusion
The purpose of this study was to examine whether drinking to manage mood
mediates the association between lifetime depression and alcohol dependence. As
predicted, drinking to manage mood was a significant mediator of the depression-
alcohol dependence relation at both the phenotypic and genotypic levels. The twin
analyses expanded upon the phenotypic analyses by estimating the sources of
covariation among drinking motives, depression and alcohol dependence. A substantial
proportion of the relation between depression and alcohol dependence can be attributed
to genetic and environmental influences shared between all three variables. After
accounting for the shared variance between the three, there was little remaining shared
genetic and specific environmental variance between depression and alcohol
dependence. The current study makes a significant contribution toward understanding

63
the underlying shared vulnerability to the co-occurrence of lifetime depression and
alcohol dependence via the factors underlying drinking to manage mood. Future
prospective studies that measure drinking motives, depressive symptoms and alcohol
consumption at multiple time-points and assess the development of depression and
alcohol dependence over time are needed.

64
References
Arpana, A., Dick, D. M., Bucholz, K. K., Madden, P. A. F., Cooper, M. L., Sher, K. J., &
Heath, A. C. (2007). Drinking expectancies and motives: a genetic study of young adult
women. Addiction, 103, 194-204.
Berkson, J. (1946). Limitations of the application of the 4-fold table analyses to hospital data.
Biometrics, 2, 47-53.

Burns, L. & Teesson, M. (2002). Alcohol use disorders comorbid with anxiety, depression and
drug use disorders. Findings from the Australian National Survey of Mental Health and
Well Being. Drug and Alcohol Dependence, 68, 299-307.

Brown, S. A. (1985). Reinforcement expectancies and alcoholism treatment outcome after a
one-year follow-up. Journal of Studies on Alcohol, 49, 412-417.

Brown, S. A., Christiansen, B. A., & Goldman, M. S. (1987). Alcohol expectancy
questionnaire: an instrument for the assessment of adolescent and adult alcohol
expectancies. Journal of Studies on Alcoholism, 48, 483-491.

Brown, S. A. & Schuckit, M. A. (1988). Changes in depression in abstinent alcoholics. Journal
of Studies on Alcohol, 49, 412-417.

Brown, S. A., Tate, S. R., Vik, P. W., Haas, A. L., & Aarons, G. A. (1999). Modeling of
alcohol use mediates the effect of family history of alcoholism on adolescent alcohol
expectancies. Experimental and Clinical Psychopharmacology, 7, 20-27.

Carey, K. B. & Correia, C. J. (1997). Drinking motives predict alcohol-related problems in
college students. Journal of Studies on Alcoholism, 58, 100-105.

Carpenter, K. M. & Hasin, D. S. (1998a). Reasons for drinking alcohol: Relationships with
DSM-IV alcohol diagnoses and alcohol consumption in a community sample.
Psychology of Addictive Behavior, 12, 168-184.

Carpenter, K. M. & Hasin, D. S. (1998b). A prospective evaluation of the relationship between
reasons for drinking and DSM-IV alcohol use disorders. Addictive behavior, 23, 41-46.

Carpenter, K. M. & Hasin, D. S. (1999). Drinking to cope with negative affect and DSM-IV
alcohol use disorders: a test of three alternative explanations. Journal of Consulting
and Clinical Psychology, 51, 93-99.

Chandler, M., Elgar, F. J. & Bennett, P. (2006). Drinking and motivations to drink among
adolescent children of parents with alcohol problems. Alcohol & Alcoholism, 41, 107-
113.

Cohen, J. W. (1988). Statistical power analysis for the behavioral sciences (2
nd
Ed). New York:
Lawrence Erlbaum Associates.

65

Colder, C. R. (2001). Life stress, physiological and subjective indexes of negative emotionality,
and coping reasons for drinking: is there evidence for a self-medication model of
alcohol use? Psychology of Addictive Behaviors, 15, 237-245.

Cooper, M. L. (1994). Motivations for alcohol use among adolescents: Development and
validation of a four-factor model. Psychological Assessment, 6, 117-128.

Cooper, M. L., Frone, M. R., Russell, M., & Mudar, P. (1995). Drinking to regulate positive
and negative emotions: A motivational model of alcohol use. Journal of Personality
and Social Psychology, 69, 990-1005.

Cooper, M. L., Russell, M. & George, W. H. (1988). Coping, expectancies, and alcohol abuse:
A test of social learning formulations. Journal of Abnormal Psychology, 97, 218-230.

Cooper, M. L., Russell, M., Skinner, J. B., Frone, M. R. & Mudar, P. (1992a). Stress and
alcohol use: Moderating effects of gender, coping, and alcohol expectancies. Journal of
Abnormal Psychology, 101, 139-152.

Cooper, M. L., Russell, M., Skinner, J. B. & Windle, M. (1992b). Development and validation
of a three-dimensional measure of drinking motives. Psychological Assessment, 4, 123-
132.

Cornelius, J. R., Salloum, I. M., Mezzich, J., Cornelius, M. D., Fabrega, H. J., Ehler, J. G.,
Ulrich, R. F., Thase, M. E. & Mann, J. J. (1995). Disproportionate suicidality in
patients with comorbid major depression and alcoholism. American Journal of
Psychiatry, 152, 358-364.

Coryell, W., Winokur, G., Keller, M., Scheftner, W. & Endicott, J. (1992). Alcoholism and
primary major depression: a family study approach to co-existing disorders. Journal of
Affective Disorders, 24, 93-99.

Cotton, N. S. (1979). The familial incidence of alcoholism: a review. Journal of Studies on
Alcoholism, 40, 89-116.

Cox, W. M. & Klinger, E. (1988). A motivational model of alcohol use. Journal of Abnormal
Psychology, 97, 168-180.

Crum, R. M., Brown, C., Kung-Yee, L. & William, W. E. (2001). The association of depression
and problem drinking: analyses from the Baltimore ECA follow-up study. Addictive
Behaviors, 26, 765-773.

Davidson, K. M. (1995). Diagnosis of depression in alcohol dependence: changes in prevalence
with drinking status. British Journal of Psychiatry, 166, 199-204.

Davidson, K. M. & Ritson, B. E. (1993). The relationship between alcohol dependence and
depression. Alcohol & Alcoholism, 28, 147-155.

66
Dawson, D. A. & Grant, B. F. (1996). Family history of alcoholism and gender: their combined
effects on DSM-IV alcohol dependence and major depression. Journal of Studies on
Alcohol, 59, 97-106.

Derogatis, L. R., Lipman, R. S., Rickels, K., Uhlenhuth, E. H. & Covi, L. (1974). The Hopkins
Symptom Checklist (HSCL): a self-report symptom inventory. Behavioral Science, 19,
1-15.

Diehl, A., Croissant, B., Batra, A., Mundle, G., Nakovics, H. & Mann, K. (2007). Alcoholism
in women: Is it different in onset and outcome compared to men? European Archives of
Psychiatry and Clinical Neuroscience, 257, 344-351.

Finn, P. R., Kleinman, I. & Pihl, R. O. (1990). The lifetime prevalence of psychopathology in
men with multigenerational family histories of alcoholism. Journal of Nervous and
Mental Disease, 178, 500-504.

Fromme, K., Stroot, E. & Kaplan, D. (1993). Comprehensive effects of alcohol: Development
and psychometric analysis of a new alcohol expectancy questionnaire. Psychological
Assessment, 5, 19-26.

Gaher, R. M., Simons, J. S., Jacobs, G. A., Meyer, D. & Johnson-Jimenez, E. (2006). Coping
motives and trait negative affect: Testing mediation and moderation models of alcohol
problems among American Red Cross disaster workers who responded to the
September 11, 2001 terrorist attacks. Addictive Behaviors, 31, 1319-1330.

Galen, L. W., Henderson, M. J. & Coovert, M. D. (2001). Alcohol expectancies and motives in
a substance abusing male treatment sample. Drug and Alcohol Dependence, 62, 205-
214.

Goldman, M. S. 1994. The alcohol expectancy concept: applications to assessment, prevention,
and treatment of alcohol abuse. Applied and Preventive Psychology, 3, 131-144.

Goldman, M. S., Del Boca, & F. K. Darkes, J. (1999). Alcohol expectancy theory: the
application of cognitive neuroscience, in Psychological Theories of Drinking and
Alcoholism, 2
nd
ed (Leonard, K. E., Blaine, H. T. eds), pp 203-246. Guilford, New
York.

Grant, B. F. (1997). The influence of comorbid major depression and substance use disorders
on alcohol and drug treatment: results of a national survey. NIDA Research
Monograph, 172, 4-15.

Grant, B. F. & Harford, T. C. (1995). Comorbidity between DSM-IV alcohol use disorders and
major depression: results of a national survey. Drug and Alcohol Dependence, 39, 197-
206.

Grant, B. F., Hasin, D. S. & Dawson, D. A. (1996). The relationship between DSM-IV alcohol
use disorders and DSM-IV major depression: examination of the primary-secondary
distinction in a general population sample. Journal of Affective Disorders, 38, 113-128.

67
Grant, B. F., Stinson, F. S., Dawson, D. A., Chou, P., Dufour, M. C., Compton, W., Pickering,
R. P. & Kaplan, K. (2004). Prevalence and co-occurrence of substance use disorders
and independent mood and anxiety disorders. Archives of General Psychiatry, 61, 807-
816.

Goodwin, D. W., Schulsinger, R., Hermansen, L., Guze, S. B. & Winokur, G. (1973). Alcohol
problems in adoptees raised apart from alcoholic biological patents. Archives of
General Psychiatry, 28, 238-255.

Goodwin, D. W., Schulsinger, F., Knop, J., Mednick, S. & Guze, S. B. (1977). Alcoholism and
depression in adopted-out-daughters of alcoholic. Archives of General Psychiatry, 34,
751-755.

Hasin, D. S., Goodwin, R. D., Stinson, F. S. & Grant, B. F. (2005). Epidemiology of major
depressive disorder: Results from the national epidemiologic survey on alcoholism and
related conditions. Archives of General Psychiatry, 62, 1097-1106.

Hasin, D. S. & Grant B. F. (2002). Major depression in 6050 former drinkers: Association with
past alcohol dependence. Archives of General Psychiatry, 59, 94-800.

Hasin, D. S., Tsai, W., Endicott, J., Mueller, T. I., Coryell, W. & Keller, M. (1996). Five-year
course of major depression: Effects of comorbid alcoholism. Journal of Affective
Disorders, 41, 63-70.

Helzer, J. & Pryzbeck, T. (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.

Hettema, J. M., Prescott, C. A. & Kendler, K. S. (2003). The effects of anxiety, substance use
and conduct disorders on risk of major depressive disorder. Psychological Medicine,
33, 1423-1432.

Holahan, C. J., Moos, R. H., Holahan, C. K., Cronkite, R. C. & Randall, P. K. (2003). Drinking
to cope and alcohol use and abuse in unipolar depression: a 10-year model. Journal of
Abnormal Psychology, 112, 159-165.

Holahan, C. J., Moos, R. H., Holahan, C. K., Cronkite, R. C., & Randall, P. K. (2004). Unipolar
depression, life context vulnerabilities, and drinking to cope. Journal of Consulting and
Clinical Psychology, 72, 269-275.

Horn, J. L. & Wanberg, K. W. (1983). Assessment of alcohol use with multidimensional
concepts and measures. American Psychologist, 38, 1055-1069.

Hussong, A. M., Hicks, R. E., Levy, S. A. & Curran, P. J. (2001). Specifying the relations
between affect and heavy alcohol use among young adults. Journal of Abnormal
Psychology, 110, 449-461.

68
Ingraham, L. J. & Wender, P. H. (1992). Risk for affective disorder and alcohol and other drug
abuse in the relatives of affectively ill adoptees. Journal of Affective Disorders, 26, 45-
52.

Jackson, K. M. & Sher, K. J. (2003). Alcohol use disorders and psychological distress: a
prospective state-trait analysis. Journal of Abnormal Psychology, 112, 599-613.

Johnson, P. B. & Gurin, G. (1994). Negative affect, alcohol expectancies and alcohol-related
problems. Addiction, 89, 581-586.

Jones, B. T., Corbin, W. & Fromme, K. (2001). A review of expectancy theory and alcohol
consumption. Addiction, 96, 57-72.

Khantzian, E. J. (1985). The self-medication hypothesis of addictive disorders: Focus on heroin
and cocaine dependence. American Journal of Psychiatry, 142, 1259-1264.

Kendler, K. S., Andrew, Heath, A. C., Neale, M. C., Kessler, R. C. & Eaves, L. J. (1993).
Alcoholism and major depression in women: a twin study of the causes of comorbidity.
Archives of General Psychiatry, 50, 690-698.

Kendler, K. S. & Gardner, C. O. (1998). Twin studies of adult psychiatric and substance
dependence disorders: are they biased by differences in the environmental experiences
of monozygotic and dizygotic twins in childhood and adolescence? Psychological
Medicine, 28, 625-633.

Kendler, K. S. & Prescott, C. A. (2006). Genes, Environment, and Psychopathology. Guilford,
New York.

Kessler, R. C., Chiu, W. T., Demler, O. & Walters, E. E. (2005). Prevalence, severity, and
comorbidity of 12-month DSM-IV disorders in the National Comorbidity Survey
Replication. Archives of General Psychiatry, 62, 627-627.

Kessler, R. C., Crum, R., Warner, L. A., Nelson, C. B., Schulenberg, H. & Anthony J. C.
(1997). Lifetime co-occurrence of DSM-III-R alcohol abuse and dependence with other
psychiatric disorders in the national comorbidity survey. Archives of General
Psychiatry, 54, 313-321.

Kessler, K. S. & Price, R. H. (1993). Primary prevention of secondary disorders: a proposal and
agenda. American Journal of Community Psychology, 21, 607-633.

Klein, H. (1992). Self-reported reasons for why college students drink. Journal of Alcohol and
Drug Education, 37, 14-28.

Kuo, P., Gardner, C. O., Kendler, K. S. & Prescott, C. A. (2006). The temporal relationship of
alcohol dependence and major depression: using a genetically informative study
design. Psychological Medicine, 36, 1153-1162.

Leigh, B. C. (1989). In search of the seven dwarves: Issues of measurement and meaning in
alcohol expectancy research. Psychological Bulletin, 105, 361-373.

69

Maes, H. H., Neale, M. C., Kendler, K. S., Hewitt, J. K., Silberg, J. L., Foley, D. L., Meyer, J.
M., Rutter, M., Simononoff, E., Pickles, A., Eaves, L. J. (1998). Assortative mating for
major psychiatric diagnoses in two population-based samples. Psychological Medicine,
28, 1389-1401.

Maier, W. Lichtermann, D. & Minges, J. (1994). The relationship between alcoholism and
unipolar depression- a controlled family study. Journal of Psychiatric Research, 28,
303-317.

Maier, W. & Merikangas, K. (1996). Co-occurrence and cotransmission of affective disorders
and alcoholism in families. British Journal of Psychiatry, 168, 93-100.

Marlatt, G. A. & Gordon, J. R. (1980). Determinants of relapse: Implications for the
maintenance of behavior change. In P.O. Davidson & S. M. Davidson (Eds.),
Behavioral medicine: Changing health lifestyles. New York: Brunner/Mazel.

Merikangas, K. R. & Gelernter, C. S. (1990). Comorbidity for alcoholism and depression.
Psychiatric Clinics of North America, 13, 613-631.

Merikangas, K. R., Risch, N. J. & Weissman, M. M. (1994). Comorbidity and cotransmission
of alcoholism, anxiety and depression. Psychological Medicine, 24, 69-80.

Merikangas, K. R., Weissman, M. M., Prusoff, B. A., Pauls, D. L. & Leckman, J. F. (1985).
Depressives with secondary alcoholism: psychiatric disorders in offspring. Journal of
Studies on Alcohol, 46, 199-204.

Moller-Leimkuhler, A. M. (2002). Barriers to help-seeking by men: a review of sociocultural
and clinical literature with particular reference to depression. Journal of Affective
Disorders, 71, 1-9.

Mooney, D. K., Fromme, K., Kivlahan, D. R. & Marlatt, G. A. (1987). Correlates of alcohol
consumption: sex, age, and expectancies relate differentially to quantity and frequency.
Addictive behaviors, 12, 235-240.

Muthen, L. K. & Muthen, B. O. (2001). Mplus User’s Guide. 2
nd
ed. Muthen and Muthen, Los
Angeles.

Neale, M. C., Eaves, L. J. & Kendler, K. S. (1994). The power of the classical twin study to
resolve variation in threshold traits. Behavioral Genetics, 24, 239-258.

Office of Applied Studies. (2006). Results from the 2005 National Survey on Drug Use and
Health: National findings. Rockville, MD: Substance Abuse and Mental Health
Services Administration.

Office of Applied Studies. (2005). Results from the 2004 National Survey on Drug Use and
Health: National findings. Rockville, MD: Substance Abuse and Mental Health
Services Administration.

70
Olenick, N. L. & Chalmers, D. K. (1991). Gender-specific drinking styles in alcoholics and
nonalcoholics. Journal of Studies on Alcohol, 52, 325-330.

Park, C. L. & Levenson, M. R. (2002). Drinking to cope among college students: prevalence,
problems and coping processes. Journal of Studies on Alcohol, 63, 486-497.

Peirce, R. S., Frone, M. R., Russell, M. & Cooper, M. L. (1994). Relationship of financial
strain and psychosocial resources to alcohol use and abuse: the mediating role of
negative affect and drinking motives. Journal of Health and Social Behavior, 35, 291-
308.

Perry, A. (1973). The effect of heredity on attitudes toward alcohol, cigarettes and coffee.
Journal of Applied Psychology, 58, 275-277.

Preisig, M., Fenton, B. T., Stevens, D. E. & Merikangas, K. R. (2001). Familial relationship
between mood disorders and alcoholism. Comprehensive Psychiatry, 42, 87-95.

Prescott, C. A. (2002) Sex differences in the genetic risk for alcoholism. Alcohol Research &
Health, 26, 269-273.

Prescott, C. A., Aggen, S. H. & Kendler, K. S. (1999). Sex differences in the sources of genetic
liability to alcohol abuse and dependence in a population-based sample of U.S. twins.
Alcoholism: Clinical and Experimental Research, 23, 1136-1144.

Prescott, C. A., Aggen, S. H. & Kendler, K. S. (2000). Sex-specific genetic influences on the
comorbidity of alcoholism and major depression in a population-based sample of U.S.
twins. Archives of General Psychiatry, 57, 803-811.

Prescott, C. A., Cross, R. J., Kuhn, J. W., Horn, J. L. & Kendler, K. S. (2004). Is risk for
alcoholism mediated by individual differences in drinking motivations? Alcoholism:
Clinical and Experimental Research, 28, 29-39.

Ratliff, K. G. & Burkhart, B. R. (1984). Sex differences in motivation for and effects of
drinking among college students. Journal of Studies on Alcohol, 45, 26-32.

Regier D. A., Farmer, M. E., Rae D. S., Lock, B., Keith, S. J., Judd, L. L. & Goodwin, F. K.
(1990). Comorbidity of mental disorders with alcohol and other drug abuse: results
from the Epidemiologic Catchment Area Study. Journal of the American Medical
Association, 264, 2511-2519.

Schuckit, M. A., Smith, T. L. & Chacko, Y. (2006). Evaluation of a depression-related model
of alcohol problems in 430 probands from the San Diego prospective study. Drug and
Alcohol Dependence, 82, 194-203.

Sher, K. J., Wood, P. K. & Raskin, G. (1996). Alcohol outcome expectancies and alcohol use: a
latent variable cross-lagged panel study. Journal of Abnormal Psychology, 100, 427-
448.

71
Slutske, W. S., Cronk, N. J., Sher, K. J., Madden, P. A. F., Bucholz, K. K. & Heath, A. C.
(2002). Genes, Environment, and individual differences in alcohol expectancies among
female adolescents and young adults. Psychology of Addictive Behaviors, 16, 308-317.

Spitzer R. L. & Williams, J. B. W. (1985). Structured Clinical Interview for DSM-III-R (SCID).
Biometrics Research Department, New York State Psychiatric Institute, New York.

Sullivan, P. F., Neale, M. C., & Kendler, K. S. (2000). The genetic epidemiology of major
depression: Review and meta-analysis. American Journal of Psychiatry, 157, 1552-
1562.

Swendsen, J. D. & Merikangas, K. R. (2000). The comorbidity of depression and substance use
disorders. Clinical Psychology Review, 20, 173-189.

Timko, C., Finney, J. W. & Moos, R. H. (2005). The 8-year course of alcohol abuse: gender
differences in social context and coping. Alcoholism: Clinical and Experimental
Research, 29, 612-621.

Vernon, P. A., Lee, D., Harris, J. A. & Jang, K. L. (1996). Genetic and environmental
contributions to individual differences in alcohol expectancies. Personality and
Individual Differences, 21, 183-187.

Wang, J. & Patten, S. B. (2001). A prospective study of sex-specific effects of major depression
on alcohol consumption. Canadian Journal of Psychiatry, 46, 422-425.

Weissman, M. M., Gershon, E. S., Kidd, K. K., Prusoff, B. A., Leckman, J. F., Dibble E.,
Hamovit, J., Thompson, W. D., Pauls, D. L. & Guroff, J. J. (1984). Psychiatric
disorders in the relatives of probands with affective disorders: the Yale University
National Institute of Mental Health Collaborative Study. Archives of General
Psychiatry, 41, 13-21.

Winokur, G. & Coryell, W. (1991). Familial alcoholism in primary unipolar major depressive
disorder. American Journal of Psychiatry, 148, 184-188.

Worobec, T. G., Winston, M. T., O’Farrell, T. J., Cutter, H. S., Bayog, R. D. & Tsuang, M. T.
(1990). Alcohol use by alcoholics with and without a history of parental alcoholism.
Alcoholism: Clinical and Experimental Research, 14, 887-892.

Zack, M., Toneatto, T. & MacLeod, C. M. (1999). Implicit activation of alcohol concepts by
negative affective cues distinguishes between problem drinkers with high and low
psychiatric distress. Journal of Abnormal Psychology, 108, 518-531.

72
Appendix A.
Items on Mood Management Scale (MOT) and Factor Loadings.

1.Do you drink to change your mood (drink when bored, angry, flat)? .69
2.Do you drink to manage mood swings from periods of happiness to periods .76
of despair?
3.Do you drink to relieve tension and stress? .67
4.Do you drink to let down? .48
5.Do you drink to forget? .77
6.Do you frequently begin drinking because things pile up? .71
7.Do you start drinking to get over being depressed? .81

73
Appendix B.
Correlation Table With and Without Covariates.

Cross Twin
MD-MD

Cross Twin
AD-AD
Within
Twin AD-
MD
Cross Twin
AD-MD

N
MZF
no
covariates
(w/
covariates)
449
0.46
(.46)

0.65
(0.62)

0.43
(0.43)

0.20
(0.20)
DZF
no
covariates
(w/
covariates)
373
0.08
(0.08)

0.20
(0.18)

0.44
(0.43)

0.06
(0.06)
MZM
no
covariates
(w/
covariates)
661
0.26
(0.26)

0.60
(0.56)

0.33
(0.30)

0.16
(0.16)
DZM
no
covariates
(w/
covariates)
623
0.23
(0.23)

0.36
(0.32)

0.33
(0.33)

0.19
(0.19)
DZO
Male- no
covariates
(w/
covariates)
1238
0.13
(0.13)

0.17
(0.14)

0.33
(0.30)

0.13
(0.13)
Female
no
covariates
(w/
covariates)

--

--

0.43
(0.43)

0.18
(0.18)

Abstract (if available)

Abstract The purpose of this study was to examine whether the relation between lifetime major depression (MD) and lifetime alcohol dependence (AD) is mediated by drinking to manage mood. Participants were 5506 individuals from the Virginia Adult Twin Study of Psychiatric and Substance Use Disorders (VATSPSUD). First, it was hypothesized that drinking to manage mood partially mediates the association between MD and AD within individuals. Second, it was hypothesized that the overlapping familial risk for MD and AD is partially mediated by drinking to manage mood.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Breaking it down to make it stronger: examining the role of source credibility and reference group specificity in the influence of personalized normative feedback on perceived alcohol use norms a...

Asset Metadata

Creator Young-Wolff, Kelly (author)

Core Title Drinking alcohol to improve mood partially mediates the relation between major depression and alcohol dependence

School College of Letters, Arts and Sciences

Degree Master of Arts

Degree Program Psychology

Publication Date 07/31/2010

Defense Date 05/13/2008

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag alcohol dependence,drinking motives,major depression,OAI-PMH Harvest,Twins

Language English

Advisor Prescott, Carol A. (committee chair), Dawson, Michael E. (committee member), Gatz, Margaret (committee member)

Creator Email kellyyw@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1468

Unique identifier UC1461689

Identifier etd-YoungWolff-20080731 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-90584 (legacy record id),usctheses-m1468 (legacy record id)

Legacy Identifier etd-YoungWolff-20080731.pdf

Dmrecord 90584

Document Type Thesis

Rights Young-Wolff, Kelly

Type series, texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu