Addictive Behaviors 31 (2006) 948 – 961
Psychiatric comorbidity and progression in drug use in adult male
twins: Implications for the design of genetic association studies
Alexandre A. Todorov a,*, Michael T. Lynskey a, Julia D. Grant a, Jeffrey F. Scherrer a,b,
Richard D. Todd a, Kathleen Keenan Bucholz a
a
Department of Psychiatry, Midwest Alcoholism Research Center, Washington University School of Medicine,
660 South Euclid Avenue, Box 8134, St. Louis, MO 63110-1502, United States
b
Research Service, St. Louis VAMC, St. Louis, MO, United States
Abstract
Psychiatric comorbidity with drug dependence has been widely documented. In the present study, we reanalyze
DSM-III-R diagnostic data on middle-aged male twin pairs from the VETR study using latent class methods. We
identify four subtypes based on 15 diagnostic categories. We then show that these subtypes are strongly associated
with differential rates of transitions in drug use histories, with increased risks in relatives for depression, alcohol,
drug and ASPD, as well as with a variety of non-normative and deviant behaviors in youth and in adulthood. We
use the result of these analyses to show how the use of a particular drug disorder phenotype for selecting cases
could impact final sample composition.
D 2006 Published by Elsevier Ltd.
Keywords: Alcohol use disorder; Substance use disorder; Psychiatric co-morbidity; Genetic association study
1. Introduction
Substance dependence is a complex disorder with multiple causes and manifestations. Psychiatric
comorbidity is common, including antisocial personality disorder, depression and anxiety disorders, and
has been correlated with poorer treatment outcome (e.g., Grant, 1995; Grant & Harford, 1995; Helzer &
* Corresponding author.
E-mail address: todorov@wustl.edu (A.A. Todorov).
0306-4603/$ - see front matter D 2006 Published by Elsevier Ltd.
doi:10.1016/j.addbeh.2006.03.046
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
949
Pryzbeck, 1988; Kessler et al., 1997; Merikangas, Mehta et al., 1998; Merikangas, Stolar et al., 1998;
Myrick & Brady, 1997; Regier et al., 1990; Rounsaville et al., 1991a,1991b, 1998; Swendsen et al.,
1998; Tomasson & Vaglum, 1998a,1998b; Ulzen & Hamilton, 1998). Extensive polysubstance use has
been reported in multiple populations, including, for example, the somewhat paradoxical simultaneous
dependence on both opioids and stimulants (e.g., Conway, Cane, Ball, Poling, & Rounsaville, 2003;
Hopfer, Khuri, Crowley, & Hooks, 2002; Kaye & Darke, 2000; Kosten, Rounsaville, & Kleber, 1998;
Lauzon et al., 1994).
These observations have led to the postulate of a common liability for substance use disorders (Glantz
& Pickens, 1992; Vanyukov et al., 2003) and to the suggestion that this liability may overlap with that
for other psychiatric disorders, such as antisocial personality disorder and depression (e.g., Helzer &
Pryzbeck, 1988; Kendler, Jacobson, Prescott, & Neale, 2003,Kendler, Prescott, Myers, & Neale, 2003;
Merikangas, Mehta et al., 1998; Merikangas, Stolar et al., 1998; Merikangas, Risch, & Weissman, 1994;
Vanyukov et al., 2003). Kendler, Jacobson et al. (2003) for example, argue in their analysis of twin data
there may be four general genetic risk factors: liability to externalizing disorders (e.g., adult antisocial
behavior, and conduct disorder), liability to internalizing disorders (major depression, generalized
anxiety disorder, and phobia), a factor specific for alcohol dependence and one for drug abuse/
dependence). Nurnberger et al. (2004) show aggregation of antisocial personality disorder, drug
dependence, anxiety disorders, mood disorders and alcohol dependence within some families. Previous
analyses of data from male twins who served in the military during the Vietnam era suggested significant
genetic correlations between the liabilities for generalized anxiety disorder and panic disorder (Scherrer
et al., 2000) and between these syndromes and posttraumatic stress disorder (Chantarujikapong et al.,
2001). Others have reported that the overlap may be explained, at least in part, by measures of
personality (neuroticism in particular) (Bienvenu et al., 2001; Clark, Watson, & Mineka, 1994; Khan,
Jacobson, Gardner, Prescott, & Kendler, 2005; Krueger, 2005; Markon & Krueger, 2005; Sher & Trull,
1994).
The question of psychiatric comorbidity among those with substance use disorders is especially
salient given the strong momentum to map genes associated with substance use risk. It is now well
established that genetic factors are important determinants of the susceptibility to substance use
disorders. Indeed, family histories of substance abuse problems are reliably associated with an increased
risk of the disorder in close relatives (Hill, Cloninger, & Ayre, 1977; Luthar, Anton, Merikangas, &
Rounsaville, 1992a, Luthar, Anton, Merikangas, & Rounsaville, 1992b; Maddux & Desmond, 1989;
Merikangas et al., 1992; Pickens et al., 2001; Rounsaville et al., 1991) as well as with poorer substance
abuse treatment outcomes (Pickens et al., 2001) Adoption studies (Cadoret, Yates, Troughton,
Woodworth, & Stewart, 1995; Cadoret, Yates, Troughton, Woodworth, & Stewart, 1996; Langbehn,
Cadoret, Caspers, Troughton, & Yucuis, 2003; Troughton, O’Gorman, & Heywood, 1986) and twin
studies (Jang, Livesley, & Vernon, 1995; McGue, Elkins, & lacono, 2000; Newlin, Miles, van den Bree,
Gupman, & Pickens, 2000; Prescott & Kendler, 1999; Rhee et al., 2003; Tsuang et al., 1998; van den
Bree, Johnson, Neale, & Pickens, 1998) which have the ability to some extent to separate the effects of
gene and environment, demonstrate that this familiality is to a large degree determined by genetic
factors, with estimates of the overall heritability of substance abuse vulnerability typically ranging from
40% to 60% (Kendler, Karkowski, Corey, Prescott, & Neale, 1999; Kendler, Jacobson et al., 2003;
Kendler, Prescott et al., 2003; Tsuang et al., 1998). While no single linkage study alone provides
conclusive results, a comparison of the results of several studies suggests 15 reproducible chromosomal
loci (Uhl, Li, Walther, Hess, & Naiman, 2001, Uhl, Liu, & Naiman, 2002). Of special significance is the
950
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
recent report of linkage between conduct disorder symptoms and substance dependence vulnerability to
the same region on chromosome 9q34 (Stallings et al., 2005). Animal studies provide a firm scientific
basis for several candidate genes, with some emerging association findings in humans (reviewed in
Kreek, Bart, Lilly, LaForge, & Nielsen, 2005).
Given the current ability to collect sufficiently large samples for genetic epidemiological studies of
substance use disorders, we now have the prospect of not only acknowledging the complexity of
substance use disorders in genetic epidemiological studies, but modeling it as well. For practical reasons,
participants are typically selected on one particular phenotype, such as severe alcohol, opioid or cocaine
dependence. Given the high level of comorbidity with other psychiatric disorders, many of which have
themselves been shown to be modulated by genetic factors, the question remains as to what population is
studied when selecting on one phenotype, as well as whether other genetic traits are unknowingly being
enriched for at the same time (for example, genes for IQ; Todd, 2005).
In the present study, we examine this issue through a reanalysis of an unusually comprehensive data
set that is based on information obtained from over 8000 middle aged male twins in a telephone
administration of a structured psychiatric interview that covered the full range of substance and nonsubstance use disorders. These data afford certain advantages. Lifetime diagnostic criteria for DSM-3R
abuse and dependence are obtained for a full range of licit (alcohol and nicotine) and illicit (marijuana,
cocaine, other stimulants, sedatives, opiates, hallucinogens/PCP) drugs. Moreover, stages of use of each
illicit drug are also obtained, permitting several transitions in drug use progression to be studied. A wide
range of non-substance psychiatric disorders are covered, including depression, anxiety, and PTSD,
among others, the full range of which is rarely included in a national community sample. For example,
both large-scale national surveys–the National Longitudinal Alcohol Epidemiology Survey (NLAES,
conducted in 1991) and National Epidemiologic Survey of Alcohol and Related Disabilities (NESARC,
conducted in 2001)–have limited coverage of non-substance psychiatric disorders. Therefore, these data
present a rare opportunity to investigate the full extent of psychiatric comorbidity in tandem with
substance use disorders, and progression of substance use disorders.
In the present, we adopted a categorical approach in our application of latent class analysis to these
data to determine if subtypes of individuals can be clearly distinguished with distinctive profiles on 15
psychiatric and substance use diagnoses. Using the subtypes obtained, we examined them in association
with various transitions in drug use; increased risks to relatives (fathers, mothers, siblings, and children)
of depression, alcohol, drug and antisocial problems, and with an array of deviant youthful and adult
behaviors. Finally, we use the results of these analyses to address the impact of selecting individuals
affected with a particular substance use disorder on the final composition of the sample, using the
example of opiate dependence versus general drug dependence (i.e. dependence on any illicit drug).
2. Methods
2.1. Sample
The Vietnam Era Twin Registry (VETR) is a national sample of male–male twin pairs both of whom
served in the military during the Vietnam War era (1965–1975). Approximately 1/3 of participants
actually served in South East Asia. From Department of Defense files of about 5.5 million men who
were on active duty during that time, 7369 twin pairs were identified using algorithms which matched
951
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
subjects on similarity of social security number, date of birth, and first and last names. Construction of
the Registry and method of determining zygosity are described in detail elsewhere (Eisen et al., 1989;
Eisen et al., 1987; Henderson et al., 1990). Twins were recruited for a mailed questionnaire study in
1987, to which 10,300 individuals responded. Data in the present report are from a 1992 follow-up
interview of this twin cohort, in which all twin pairs where at least one member participated in the 1987
questionnaire were invited to participate. Of the 10,300 twins from the 1987 study, 8169 (79.3%)
participated in the 1992 follow-up (3372 complete twin pairs; 66.1% pair-wise response rate) and 1425
singletons. (Tsuang et al., 1998).
2.2. Psychiatric assessment
The structured interview used in the 1992 study was a computerized telephone version of the
Diagnostic Interview Schedule, Version III Revised (DIS-3R) (Robins et al., 1989), a comprehensive
structured interview used to derive psychiatric diagnoses according to DSM-III-R lifetime criteria. A
standard research instrument in the psychiatric field, the DIS has been shown to be highly reliable and
valid (Robins, Helzer, Croughan, & Ratcliff, 1981; Robins, Helzer, Ratcliff, & Seyfried, 1982; Robins et
al., 1985). In the present report, 15 lifetime diagnoses are considered (Table 1) in four general domains
(mood, personality, anxiety, and substance dependence). Twins were also asked to report on the presence
of depression, alcoholism, drug dependence and ASPD in their first-degree relatives (mothers, fathers,
children and sibs), based on thumbnail sketches of these disorders. Interviews were administered by
experienced staff from the Institute for Survey Research at Temple University, who were trained in the
telephone administration of the interview. This work was conducted with approvals from the Institutional
Review Boards of participating universities. All subjects gave informed consent prior to their
participation.
Table 1
DSM-III-R lifetime considered in these analyses
Diagnosis
N
%
Major depressive disorder (excludes bereavement)
Mania
Dysthymia
Panic disorder
Generalized anxiety disorder
Post-traumatic stress disorder
Antisocial personality disorder
Alcohol dependence
Nicotine dependence
Cannabis dependence (marijuana, hashish, etc.)
Stimulant dependence (amphetamines, khat, ice, etc.)
Opiate dependence (heroin, morphine, opium, methadone, codeine, etc.)
Hallucinogen dependence (LSD, PCP, mescaline, peyote, etc.)
Sedative dependence (sleeping pills, tranquilizers, barbiturates, valium, etc.)
Cocaine dependence (Cocaine, crack, coca leaves, etc.)
8150
8146
8151
8149
8149
8107
8147
8148
8169
8135
8135
8135
8135
8134
8135
10.2
0.8
2.5
1.8
2.6
10.1
2.9
35.8
47.6
6.5
2.6
1.3
1.0
1.4
2.6
Prevalence for descriptive purposes only. Both twins included. The mild, moderate and severe dependence categories specified
by DSM-III-R were combined. Except for nicotine dependence, baseline category includes individuals with a DSM-III-R
diagnostic of babuseQ.
952
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
2.3. Transitions in drug use
Because the focus of the data collection effort was on progression to illicit substance dependence,
questions were added to the DIS to obtain use patterns, and the ages at which each usage pattern
began. As well, opportunity to use each drug type (termed bexposureQ) was assessed for all. Ever use,
use more than 5 times, daily use for several weeks (bregular useQ), and drug problems that reflected
DSM-III-R criteria for dependence and abuse were elicited. This information was used to define
transitions—from ever use (bexperimentationQ) to more than experimental use (N5 times), to regular
use, to dependence.
2.4. Statistical methods
Latent class analysis (LCA; McCutcheon, 1987) is a special form of mixture modeling that can be
used to investigate patterns of association among sets of categorical responses. It is predicated on the
assumption that a relatively small number of mutually exclusive classes exist, each class having a
distinctive profile of item endorsement probabilities. The parameters of the latent class model are the
number of classes, the class membership probabilities, which may be thought of as prevalence, and, for
each class, the symptom endorsement probabilities given membership in that class. On the assumption of
local independence, the conditional probabilities of endorsing a set of items are independent within a
given latent class. We verified the validity of this assumption through an examination of residuals
(roughly, comparison of the predicted and observed joint distribution for two items). In the present
analyses, we fitted LCA models with 2 to 6 classes and selected the model with the lowest value of the
Bayesian Information Criterion (Biernacki et al., 1998). The latent class analyses were conducted using
LatentGold, with several restarts to minimize the problems of local minima and boot-strapping for the
determination of empirical p-values to further substantiate the choice of a K-class model. Follow-up
analyses were performed using SAS and S-Plus.
The VETS sample consists of twin pairs which are not independent for the traits considered in this
analysis. Since the primary focus of these analyses presented here is on modeling the joint distribution of
the 15 diagnoses at the individual level, rather than on modeling the joint distribution in twin pairs, we
randomly assigned one member of each twin pair to a btrainingQ set and the other to a breplicationQ set.
Latent class analyses were run on both the training and the replication sets (and reported separately) in
order to ensure that the solution obtained was not a result of the non-independent nature of the twins.
The large sample number of pairs ensures adequate power even when only one twin is considered. At
this stage of the analyses, it would have been premature to bring in the additional complexity, and
associated modeling assumptions, of multivariate twin data analysis. This is reserved for future
investigations.
3. Results
3.1. Sample
The mean age of respondents was 44.6 years (S.D. + 2.8, range 36 to 55 years); 90.4% were nonHispanic white, 4.9% African-American, 2.7% Hispanic, 1.3% Native American/ Alaskan Native, and
953
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
Table 2
Summary statistics for the latent class analyses
Number classes
Number parameters
BIC (Twin 1, N = 4044)
BIC (Twin 2, N = 4006)
2
3
4
5
6
31
47
63
79
95
22,533
22,268
22,096
22,143
22,236
23,614
23,412
23,244
23,321
23,404
Bootstrap p-value for a test of the significance of the difference in BIC value for the 3 and 4 classes models: p b 0.0001, each,
for analyses using only twin 1 data and analyses using only twin 2 data. Empirical p-values for the difference for 4 and 5 classmodels: p b 0.01 for twin 1 data and p = 0.12 for twin 2 data.
0.7% botherQ; 72% were at least high school graduates with 38.6% being college graduates. At the time
of interview, 92.6% were employed full-time and 1.8% part-time.
3.2. Latent class analyses
The latent class analyses suggest that a four class solution was the most parsimonious (Tables 2 and
3). Class I, with an estimated frequency of 56.2%, is essentially a baseline class characterized by low
rates of psychiatric disorders and substance dependence (licit or illicit). Class II (33.6%) is characterized
by elevated prevalence of alcohol and nicotine dependence (67% and 75%), but the rates of illicit
substance dependence remain low when compared to Classes III and IV. Class III (6.8%) and Class IV
(3.4%) are distinguished by high rates of depression, PTSD, alcohol and nicotine dependence. When
Table 3
Estimated class prevalences and diagnostic probabilities within each class
Class
I
II
Class prevalence
Psychiatric comorbidity
PTSD
Depression
Panic
Anxiety
Mania
Dysthymia
ASPD
56.23
33.55%
III
4.7
2.2
0.2
~0
0.1
~0
~0
21.0
4.8
1.2
~0
0.9
~0
~0
75.1
77.0
13.7
8.5
7.8
1.8
1.5
55.1
46.5
8.9
6.0
5.3
1.5
1.7
Substance dependence
Alcohol
Tobacco
Cannabis
Stimulants
Cocaine
Sedatives
Hallucinogens
Opiates
10.8
25.3
0.3
~0
0.1
~0
~0
0.1
67.0
75.1
9.1
1.5
3.6
0.7
0.3
1.2
61.8
71.9
10.9
~0
3.6
2.0
0.2
0.6
88.6
88.9
74.5
62.2
30.8
29.9
26.4
23.9
6.81%
IV
3.40%
954
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
compared to Classes I and II, these two classes also exhibited elevated rates of antisocial personality
disorder, anxiety and panic disorders, as well as higher prevalence estimates of mania and dysthymia.
However, Classes III and IV are strikingly different in the prevalence of illicit substance dependence. In
this respect, Class IV clearly stands out from all other classes.
3.3. Characterization of the four latent classes
We further characterized the four classes according to specific indicators of drug use patterns;
conduct and/or personality disorders symptoms in both youth and adulthood (for which we used the
standard DIS/DSM conceptualization of adult behaviors as those occurring at or after age 15) We
also examined prevalence of depression, alcohol and drug dependence and ASPD in first-degree
relatives. These were based on affirmative responses to thumbnail sketches of these disorders from
the twins for mothers, fathers, siblings, and own children. For the purpose of these analyses,
individuals were assigned to the most likely class given their observed diagnostic profiles. The
posterior probabilities of class membership were calculated using the estimated four-class model
parameters. Overall (all twins considered), 4860 individuals were assigned to Class I, 2077 to class
II, 468 to class 3, and 226 to Class IV. The ability to assign an individual to a particular class was
generally high. The proportion of subjects for which the ratio of the largest to next most largest
posterior class probabilities was greater than two ranged from a low of 82% (individuals assigned to
Class II) to a high of 99% (Class I).
By definition, Class IV is defined as one with a high prevalence of dependence on one or more
substances. However, the other classes are not and yet, we observe that substantial numbers of
individuals in each class have experimented with illicit drugs (Table 4). Thus, differences in the
prevalence of substance dependence between classes are not due to abstinence alone. Cannabis
experimentation was most common. One third of Class I, the unaffected class, experimented with
cannabis, while over 70% of the other classes did so. Although experimenting with substances other than
cannabis was less common in Class I, this behavior was much more common in Classes II and III and
nearly universal in Class IV. For example, about 35% of Classes II and III reported experimenting with
stimulants, compared to 97% in Class IV. Experimentation with sedatives, cocaine and hallucinogens
was reported by 20–28% of Classes II and III, but by over 80% of Class IV. Opiate experimentation was
lowest among the substances.
The classes do differ in the proportions of individuals who have progressed from experimentation to
dependence. For the purposes of the study, an episode of bregular useQ was defined as a period where the
substance was used daily for two weeks or more. Generally, compared to Classes II and III, a lower
proportion of individuals from Class I will progress from using a substance 6 times or more to regular
use. However, the differences are not particularly striking.
In contrast, there are sharp between-class differences in the probabilities to progress from regular use
to dependence, with Class I the lowest, roughly comparable proportions in Classes II and III, and the
highest proportion in Class IV. Using opiates as an illustration, the probability of progressing from
regular use to dependence is 5.3% in Class I, 37.9% in Class II, 15.4% in class III and 76.7% in Class IV.
In Class II compared to Class III, the transition probabilities may be higher for stimulants and opiates,
and lower for cocaine.
Keeping in mind that these transition probabilities are conditional on the prior use level, the transition
to dependence was very low among Class I—under 5% across all substances. In contrast, a very high
955
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
Table 4
Transition probabilities, by drug and class
Class assignment
I
II
Cannabis
Ever use
Used N5 times
Regular use
Dependence
36.8
53.5
59.1
1.4
(4329)
(1590)
(850)
(501)
Stimulants
Ever use
Used N5 times
Regular use
Dependence
9.6
48.3
38.5
0
Sedatives
Ever use
Used N5 times
Regular use
Dependence
72.7
70.72
73.6
31.1
III
IV
(1969)
(1431)
(1012)
(744)
72.9
68.8
69.4
38.7
(431)
(314)
(216)
(150)
100.0
97.8
95.9
82.6
(226)
(226)
(221)
(212)
(4328)
(414)
(200)
(77)
35.2 (1967)
62.6 (693)
54.3 (433)
11.9 (235)
34.6
61.1
52.8
2.1
(431)
(149)
(91)
(48)
97.4
95.9
89.5
83.5
(226)
(219)
(210)
(188)
5.1
40.0
36.4
0
(4328)
(220)
(88)
(32)
21.8
58.6
48.2
12.5
(1967)
(428)
(251)
(120)
24.4
51.4
63.0
29.4
(431)
(105)
(54)
(34)
81.0
79.2
73.1
66.0
(226)
(183)
(145)
(106)
Cocaine
Ever use
Used N5 times
Regular use
Dependence
7.3
41.6
40.2
5.7
(4326)
(317)
(132)
(53)
27.1
53.8
55.1
55.1
(1968)
(534)
(287)
(158)
24.6
50.0
56.6
53.3
(431)
(106)
(53)
(30)
81.9
70.8
68.7
80.0
(226)
(185)
(131)
(90)
Opiates
Ever use
Used N5 times
Regular use
Dependence
2.3
29.6
65.5
5.3
(4326)
(98)
(29)
(19)
12.7
41.2
64.1
37.9
(1968)
(250)
(103)
(66)
14.9
32.8
61.9
15.4
(431)
(64)
(21)
(13)
70.4
57.9
80.4
76.7
(226)
(159)
(92)
(73)
Hallucinogens
Ever use
Used N5 times
Regular use
Dependence
6.0
34.1
27.3
0
(4326)
(258)
(88)
(24)
23.0
44.4
42.3
8.24
(1968)
(453)
(201)
(85)
20.7
36.0
43.8
7.1
(431)
(89)
(32)
(14)
85.0
74.5
70.6
61.0
(226)
(192)
(143)
(100)
Transition probabilities, conditional upon preceding stage having been met. The number of observations on which the
probability is calculated is in parentheses. bRegularQ use is defined as daily use for a minimum of 2 weeks.
percentage of Class IV transitioned to dependence, (greater than 75% for 4 substances, and over 60% for
2 substances), suggestive of a higher susceptibility, not only to experimentation with drugs but also to
progression to dependence. Different patterns among Classes II and III were evident. Transitions to
dependence were moderate for cannabis for both classes–about one third, and quite high for cocaine–
over one half. In comparison, the probability of transition to dependence for stimulants was quite low
(12% or less in Classes I to III). Also of note was the sharp difference in transition probabilities observed
956
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
Table 5
Twin report of problems in relatives, within each class
Problem
Depression
Alcohol problems
Drug problems
ASPD
Father
Mother
Siblings
Children
Father
Mother
Siblings
Children
Father
Mother
Siblings
Children
Father
Mother
Siblings
Children
Class I, N = 4860
Class II, N = 2077
Class III, N = 468
Class IV, N = 226
18.5
19.9
16.1
9.1
23.3
5.4
9.2
3.4
0.88
1.2
3.8
2.4
3.96
0.4
4.0
3.9
29.5
30.7
24.8
15.4
36.1
8.0
18.4
5.1
1.94
2.7
8.8
2.6
6.1
1.0
7.8
4.9
44.8
52.3
52.7
40.8
35.2
12.1
22.3
6.7
0.48
5.1
12.6
7.7
9.1
0.5
14.6
10.6
46.3
44.4
40.7
32.5
58.3
23.3
34.6
7.5
2.4
7.8
28.1
10.0
15.1
2.2
15.9
17.5
for opiate dependence between Class II and Class III—38% of Class II, but only 15% of Class III,
progressed to dependence.
Second, we observed a nearly linear response relationship between class membership and the
prevalence of symptoms of antisocial personality disorder and social maladjustment, both in youth and
in adulthood. For example, the proportion of respondents reporting having committed acts of
vandalism before age 15 are 2.7%, 7.8%, 11.5% and 19.5% in Classes I, II, III and IV, respectively.
Even though Classes III and IV have comparable proportions of antisocial personality disorder (1.5%
versus 1.6%, Table 3), overall, symptom endorsement probabilities tend to be higher in Class IV.
Third (Table 5), we observed that class membership is associated with an elevated rate of depression,
alcohol, drug, and legal problems in first degree relatives, suggesting, if not heritability, at least
familiality. Additional analyses suggest that class membership appears substantially heritable. Indeed,
the polychoric correlations for alcohol dependence were 54% for monozygotic twins and 32% for
dizygotic twins, compared to 56% and 37% for membership into Class IV (versus all others).
Table 6
Distribution of cases of DSM-III-R dependence on particular substance, across the classes
Selection dependence
phenotype, DSM-III-R
N
Alcohol
Tobacco
Cannabis
Stimulants
Sedatives
Cocaine
Opiates
Hallucinogens
2486
3678
483
196
108
188
93
80
Proportion of cases of dependence in each class
Class I
Class II
Class III
Class IV
15.1
34.5
1.5
0
0
1.6
1.1
0
64.8
51.0
49.1
15.3
16.7
50.0
30.1
8.8
11.9
9.0
12.2
0
11.1
8.5
3.2
1.3
8.3
5.6
37.3
84.7
72.2
39.9
65.6
90.0
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
957
3.4. Implications for sample selection
Finally, we considered the implication of this class structure for epidemiological studies that select
probands based on substance dependence diagnosis (Table 6). In this sample, if we select cases to be
dependent on any illicit substance (one or more), then most of the observations would come from the
lower progression risk Class II (54.1%) and fewer from the high progression risk Class IV (31.9%).
However, if we select cases to be opioid dependent, then these proportions are reversed: 65.6% from the
high-risk Class IV and 30.1% from the low-risk Class II. In general, the results suggest that most of the
sample would come from Classes I and II if selected on alcohol or nicotine dependence; from Classes II
and IV if selecting on cannabis or cocaine dependence; and primarily from Class IV if selecting on
opiate, hallucinogen, stimulant or sedative dependence.
4. Discussion
Much progress has been made in the last few years in the understanding of the substance dependence
phenotype. There are several lines of research in this regard. In addition to detailed work that examines
the structure of narrow phenotypes, such as developmental trajectories for specific drug classes, there is
another rapidly developing branch of study that takes a broader perspective in a consideration of
substance dependence in context of other psychiatric and personality related factors. The present study
falls in this latter category.
We adopted a categorical approach and focused on modeling psychiatric and substance use
comorbidity within individuals, rather than between twins. Our results suggest four distinctive groups
of psychiatric comorbidity. One class is composed of essentially unaffected individuals (Class I).
Another reflects dependence on licit substances (alcohol and nicotine). Two additional classes both
show elevations of mood and anxiety disorders but very different prevalence of illicit drug
dependence. Rates of illicit drug dependence across all drug categories are very high in Class IV, but
in contrast, are negligible in Class III. For the most part, the observed classes are suggestive of a
latent trait model. This is especially salient when one considers the relationship between class
membership and the selected symptoms of personality disorder and social maladjustment, both in
youth and in adulthood. However, we do not observe a regular increase in the prevalence of all
symptoms across classes. These symptom endorsement profiles are consistent with the hypothesis of
two broad groups of diagnostic comorbidity, one comprised of internalizing disorders (Class III), the
other with a trend to externalizing disorder and illicit drug dependence (Class IV) or to licit drug
dependence (Class II).
Whether the underlying process is due to the existence of distinct bsyndromesQ as opposed to a
continuum in liability (along one or more dimensions) is of lesser interest for the purpose of
designing a sampling scheme for a genetic epidemiological studies of substance use disorders.
Indeed, the key consideration there is not how these phenotypes co-segregate but rather the end
result—the expected profile of study participants. In the present study population, sampling on
stimulant, sedative, opiate or hallucinogen dependence would likely result in a sample that is
characterized by elevated rates of depression and PTSD, and which exhibits a strong tendency to
transition from experimental drug use to dependence. On the other hand, sampling on alcohol,
tobacco, cannabis or cocaine dependence, or sampling on any drug dependence, would result in
958
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
sample with slightly elevated depression and PTSD rates, but otherwise not altogether different from
baseline.
There are several limitations to the present study. Because the sample was composed only of middleaged males, who are primarily Caucasian, the degree to which the findings reported here apply to
women, to younger individuals, or to a more ethnically diverse sample is not known. The current study
examined DSM-III-R diagnoses only, since data collection predated DSM-IV. Whether the change in
nosology would affect the results is uncertain. However, despite the fact that this is a sample of military
veterans, which were likely screened for the most extreme externalizing disorders, we still observe an
association between externalizing behaviors and substance dependence.
At the same time, there are several advantages to the sample used. The exposure of the men to illicit
drugs was pervasive, by virtue of the era and place of military duty for some. In particular,
experimentation with opiates is far higher than reported in other studies (e.g., NESARC), resulting in
greater expression of an underlying vulnerability to opioid dependence. All participants, being in their
40s when interviewed, are past the crucial stages of drug problem development, so that we can expect
little censoring in the observations. Despite the limitations of the male-only nature of the sample and the
availability of DSM-III-R (not DSM-IV) disorder, the results reported here are broadly consistent with
those of previous studies of comorbidity of common psychiatric conditions and substance dependence.
We report here on lifetime diagnoses. Even though the overall sample is large, given the relative rarity of
each phenotype (e.g., ~ 10% for major depressive disorder), there was relatively little information to
order events, which would have allowed us to assess, for example, if a depressive episode was substance
related. Lastly, the intriguing differences in transition probabilities to dependence across classes and
types of substances may suggest different propensities to addiction based on comorbidity profiles. It was
particularly striking for opiates. However, the data do not permit distinguishing among different types of
opiates. For example, individuals reporting use of prescription medications may have a lower probability
of becoming dependent compared to their counterparts who use heroin. Observed differences might be
due to different forms of the drug taken.
Thus, even though much of the liability to substance dependence may be shared, it appears that the
choice of selection phenotype (e.g., opioid dependence versus cannabis dependence) will strongly affect
the ultimate sample composition, not only in terms, e.g., the propensity to transition to dependence from
experimental use, but also in terms of psychiatric comorbidity as well. Analyses that focus on any
substance dependence as the phenotypic endpoint may be introducing substantial heterogeneity in the
analyses. Our results suggest that a more controlled sampling scheme should ascertain samples for
genetic association studies based on diagnostic profiles rather than individual diagnoses.
Acknowledgments
This study was supported in part by grants DA-14363, DA-14632, DA-18660, AA-13640 and
AA-11998.
References
Bienvenu, O. J., Nestadt, G., Samuels, J. F., Costa, P. T., Howard, W. T., & Eaton, W. W. (2001). Phobic, panic, and major
depressive disorders and the five-factor model of personality. Journal of Nervous and Mental Disease, 189, 154 – 161.
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
959
Biernacki, C., Celeux, G., & Govaert, G. (1998). Assessing a mixture model for clustering with the integrated classification
likelihood. Annals of Statistics, 26, 1614 – 1635.
Cadoret, R. J., Troughton, E., O’Gorman, T. W., & Heywood, E. (1986). An adoption study of genetic and environmental
factors in drug abuse. Archives of General Psychiatry, 43, 1131 – 1136.
Cadoret, R. J., Yates, W. R., Troughton, E., Woodworth, G., & Stewart, M. A. (1995). Adoption study demonstrating two
genetic pathways to drug abuse. Archives of General Psychiatry, 52, 42 – 52.
Cadoret, R. J., Yates, W. R., Troughton, E., Woodworth, G., & Stewart, M. A. (1996). An adoption study of drug abuse/
dependency in females. Comprehensive Psychiatry, 37, 88 – 94.
Chantarujikapong, S. I., Scherrer, J. F., Xian, H., Eisen, S. A., Lyons, M. J., Goldberg, J., et al. (2001). A twin study of
generalized anxiety disorder symptoms, panic disorder symptoms and post-traumatic stress disorder in men. Psychiatry
Research, 103, 133 – 145.
Clark, L. A., Watson, D., & Mineka, S. (1994). Temperament, personality, and the mood and anxiety disorders. Journal of
Abnormal Psychology, 103, 103 – 116.
Conway, K. P., Kane, R. J., Ball, S. A., Poling, J. C., & Rounsaville, B. J. (2003). Personality, substance of choice, and
polysubstance involvement among substance dependent patients. Drug and Alcohol Dependence, 71, 65 – 75.
Eisen, S., Neuman, R., Goldberg, J., Rice, J., & True, W. (1989). Determining zygosity in the Vietnam Era Twin Registry: An
approach using questionnaires. Clinical Genetics, 35, 423 – 434.
Eisen, S., True, W., Goldberg, J., Henderson, W., & Robinette, C. D. (1987). The Vietnam Era Twin (VET) Registry: Method of
construction. Acta Geneticae Medicae et Gemellologiae, 36, 61 – 66.
Glantz, & Pickens (Eds.) 1992. Vulnerability to drug abuse (pp. 149 – 178). Washington, DC7 American Psychological
Association.
Grant, B. F. (1995). Comorbidity between DSM-IV drug use disorders and major depression: Results of a national survey of
adults. Journal of Substance Abuse, 7, 481 – 497.
Grant, B. F., & Harford, T. C. (1995). Comorbidity between DSM-IV alcohol use disorders and major depression: Results of a
national study. Drug and Alcohol Dependence, 39, 197 – 206.
Helzer, J. E., & Pryzbeck, T. R. (1988). The co-occurrence of alcoholism with other psychiatric disorders in the general
population and its impact on treatment. Journal of Studies on Alcohol, 49, 219 – 224.
Henderson, W. G., Eisen, S., Goldberg, J., True, W. R., Barnes, J. E., & Vitek, M. E. (1990). The Vietnam Era Twin Registry: A
resource for medical research. Public Health Reports, 105, 368 – 373.
Hill, S. Y., Cloninger, R., & Ayre, F. R. (1977). Independent familial transmission of alcoholism and opiate abuse. Alcoholism,
Clinical and Experimental Research, 1, 335 – 342.
Hopfer, C. J., Khuri, E., Crowley, T. J., & Hooks, S. (2002). Adolescent heroin use: A review of the descriptive and treatment
literature. Journal of Substance Abuse Treatment, 23, 231 – 237.
Jang, K. L., Livesley, W. J., & Vernon, P. A. (1995). Alcohol and drug problems: A multivariate behavioural genetic analysis of
co-morbidity. Addiction, 90, 1213 – 1221.
Kaye, S., & Darke, S. (2000). A comparison of the harms associated with the injection of heroin and amphetamines. Drug and
Alcohol Dependence, 58, 189 – 195.
Kendler, K. S., Karlowski, L. M., Corey, L. A., Prescott, C. A., & Neale, M. C. (1999). Genetic and environmental
risk factors in the aetiology of illicit drug initiation and subsequent misuse in women. British Journal of Psychiatry,
351 – 356.
Kendler, K. S., Jacobson, K. C., Prescott, C. A., & Neale, M. C. (2003). Specificity of genetic and environmental risk factors for
use and abuse/dependence of cannabis, cocaine, hallucinogens, sedatives, stimulants, and opiates in male twins. American
Journal of Psychiatry, 160, 687 – 695.
Kendler, K. S., Prescott, C. A., Myers, J., & Neale, M. C. (2003). The structure of genetic and environmental risk factors for
common psychiatric and substance use disorders in men and women. Archives of General Psychiatry, 60, 929 – 937.
Kessler, R. C., Crum, R. M., Warner, L. A., Nelson, C. B., Schulenberg, J., & 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.
Khan, A. A., Jacobson, K. C., Gardner, C. O., Prescott, C. A., & Kendler, K. S. (2005). Personality and comorbidity of common
psychiatric disorders. British Journal of Psychiatry, 186, 190 – 196.
Kosten, T., Rounsaville, B., & Kleber, H. A. (1998). 2.5 year follow-up of abstinence and relapse to cocaine abuse in opioid
addicts. NIDA Research Monograph, 81, 231 – 236.
960
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
Kreek, M. J., Bart, G., Lilly, C., LaForge, K. S., & Nielsen, D. A. (2005, Mar.). Pharmacogenetics and human molecular
genetics of opiate and cocaine addictions and their treatments. Pharmacological Reviews, 57(1), 1 – 26.
Krueger, R. F. (2005). Continuity of axes I and II: Toward a unified model of personality, personality disorders, and clinical
disorders. Journal of Personality Disorders, 19, 233 – 261.
Langbehn, D. R., Cadoret, R. J., Caspers, K., Troughton, E. P., & Yucuis, R. (2003, Mar. 1). Genetic and environmental risk
factors for the onset of drug use and problems in adoptees. Drug and Alcohol Dependence, 69(2), 151 – 167.
Lauzon, P., Vincelette, J., Bruneau, J., Lamothe, F., Lachance, N., Brabant, M., et al. (1994, Sep.–Oct.). Illicit use of methadone
among i.v. drug users in Montreal. Journal of Substance Abuse Treatment, 11(5), 457 – 461.
Luthar, S. S., Anton, S. F., Merikangas, K. R., & Rounsaville, B. J. (1992a). Vulnerability to drug abuse among opioid addicts’
siblings: Individuals, familial, and peer influences. Comprehensive Psychiatry, 33, 190 – 196.
Luthar, S. S., Anton, S. F., Merikangas, K. R., & Rounsaville, B. J. (1992b). Vulnerability of substance abuse and
psychopathology among siblings of opioid abusers. The Journal of Nervous and Mental Disease, 180, 153 – 161.
Maddux, J., & Desmond, D. (1989). Family and environment in choice of opioid dependence or alcoholism. American Journal
of Drug and Alcohol Abuse, 15, 117 – 134.
Markon, K. E., & Krueger, R. F. (2005). Categorical and continuous models of liability to externalizing disorders. Archives of
General Psychiatry, 62, 1352 – 1359.
McCutcheon, A. C. (1987). Latent Class Analysis. Beverly Hills7 Sage Publications.
McGue, M., Elkins, I., & lacono, W. G. (2000). Genetic and environmental influences on adolescent substance use and abuse.
American Journal of Medical Genetics, 96, 671 – 677.
Merikangas, K. R., Mehta, R.L, Molnar, B. E., Walters, E. E., Swendsen, J. D., Aguilar-Gaziola, S., et al. (1998). Comorbidity
of substance use disorders with mood and anxiety disorders: Results of the international consortium in psychiatric
epidemiology. Addictive Behaviors, 23, 893 – 907.
Merikangas, K. R., Risch, N. J., & Weissman, M. M. (1994). Comorbidity and co-transmission of alcoholism, anxiety and
depression. Psychological Medicine, 24, 69 – 80.
Merikangas, K. R., Rounsaville, B. J., & Prusoff, B. A. (1992). Familial factors in vulnerability to substance abuse. In M. D.
Glantz, & R. Pickens (Eds.), Vulnerability to Drug Abuse (pp. 75 – 98). Washington, DC7 American Psychological Association.
Merikangas, K. R., Stolar, M., Stevens, D. E., Goulet, J., Preisig, M. A., Fenton, B., et al. (1998). Familial transmission of
substance use disorders. Archives of General Psychiatry, 55, 973 – 979.
Myrick, H., & Brady, K. T. (1997). Social phobia in cocaine-dependent individuals. American Journal on Addictions, 6(2),
99 – 104.
Newlin, D. B., Miles, D. R., van den Bree, M. B. M., Gupman, A. E., & Pickens, R. W. (2000). Environmental transmission
of DSM-IV substance use disorders in adoptive and step families. Alcoholism, Clinical and Experimental Research, 24,
1785 – 1794.
Nurnberger, J. I., Wiegand, R., Bucholz, K. K., O’Connor, S., Meyer, E. T., Reich, T., et al. (2004). A family study of alcohol
dependence: Coaggregation of multiple disorders in relatives of alcohol-dependent probands. Archives of General
Psychiatry, 61, 1246 – 1256.
Pickens, R. W., Preston, K. L., Miles, D. R., Gupman, A. E., Johnson, E. O., Newlin, D. B., et al. (2001, Feb. 1). Family history
influence on drug abuse severity and treatment outcome. Drug and Alcohol Dependence, 61(3), 261 – 270.
Prescott, C. A., & Kendler, K. S. (1999). Genetic and environmental contributions to alcohol abuse and dependence in a
population-based sample of male twins. American Journal of Psychiatry, 156, 34 – 40.
Regier, D. A., Farmer, M. E., Rae, D. S., Locke, B. Z., Keith, S. J., Judd, L. L., et al. (1990). Comorbidity of mental disorders
with alcohol and other drug abuse: Results from the Epidemiologic Catchment Area (ECA) Study. Journal of the American
Medical Association, 264, 2511 – 2518.
Rhee, S. H., Hewitt, J. K., Young, S. E., Corley, R. P., Crowley, T. J., & Stallings, M. C. (2003). Genetic and environmental
influences on substance initiation, use, and problem use in adolescents. Archives of General Psychiatry, 60, 1256 – 1264.
Robins, L. N., Helzer, J. E., Cottler, L. B., & Goldring, E. (1989). National Institute of Mental Health Diagnostic Interview
Schedule Version III Revised.
Robins, L. N., Helzer, J. E., Croughan, J. L., & Ratcliff, K. S. (1981). The NIMH Diagnostic Interview Schedule: Its history,
characteristics and validity. Archives of General Psychiatry, 38, 381 – 389.
Robins, L. N., Helzer, J. E., Orvaschel, H., Anthony, J. C., Blazer, D. G., Burnam, A., et al. (1985). The Diagnostic Interview
Schedule. In W. W. Eaton, & L. G. Kessler (Eds.), Epidemiologic field methods in psychiatry: The NIMH Epidemiologic
Catchment Area Program (pp. 143 – 170). Toronto7 Academic Press.
A.A. Todorov et al. / Addictive Behaviors 31 (2006) 948–961
961
Robins, L. N., Helzer, J. R., Ratcliff, K. S., & Seyfried, W. (1982). Validity of the Diagnostic Interview Schedule, version II:
DSM-III diagnoses. Psychological Medicine, 12, 855 – 870.
Rounsaville, B. J., Kosten, T. R., Weissman, M. M., Prusoff, B., Pauls, D., Anton, S. F., et al. (1991). Psychiatric disorders in
relatives of probands with opiate addiction. Archives of General Psychiatry, 48, 33 – 42.
Rounsaville, B. J., Kosten, T. R., Weissman, M. M., Prusoff, B. A., Pauls, D., Anton, S. F., et al. (1991). Psychiatric disorders in
relatives of probands with opiate addiction. Archives of General Psychiatry, 48, 2 – 33.
Rounsaville, B. J., Kranzler, H. R., Ball, S., Tennen, H., Poling, J., & Triffleman, E. (1998). Personality disorders in substance
abusers: Relation to substance use. Journal of Nervous and Mental Disease, 186, 87 – 95.
Scherrer, J. F., True, W. R., Xian, J., Lyons, M. J., Eisen, S. A., Goldberg, J., et al. (2000). Evidence for genetic influences
common and specific to symptoms of generalized anxiety and panic. Journal of Affective Disorders, 57, 25.
Sher, K. J., & Trull, T. J. (1994). Personality and disinhibitory psychopathology: Alcoholism and antisocial personality disorder.
Journal of Abnormal Psychology, 103, 92 – 102.
Stallings, M. C., Corley, R. P., Dennehey, B., Hewitt, J. K., Krauter, K. S., Lessem, J. M., et al. (2005, Sep.). A genome-wide
search for quantitative trait Loci that influence antisocial drug dependence in adolescence. Archives of General Psychiatry,
62(9), 1042 – 1051.
Swendsen, J. D., Merikangas, K. R., Canino, G. J., Kessler, R. C., Rubio-Stipec, M., & Angst, J. (1998). The comorbidity of
alcoholism with anxiety and depressive disorders in four geographic communities. Comprehensive Psychiatry, 39, 176 – 184.
Todd, R. D. (2005). Genetic advances in autism hinge on the method of measuring symptoms. Current Psychiatry Reports, 7,
133 – 137.
Tomasson, K., & Vaglum, P. (1998a, May–Jun.). The role of psychiatric comorbidity in the prediction of readmission for
detoxification. Comprehensive Psychiatry, 39(3), 129 – 136.
Tomasson, K., & Vaglum, P. (1998b, Mar.). Social consequences of substance abuse: The impact of comorbid psychiatric
disorders. A prospective study of a nation-wide sample of treatment-seeking patients. Scandinavian Journal of Social
Medicine, 26(1), 63 – 70.
Tsuang, M. T., Lyons, M. J., Meyer, J. M., Doyle, T., Eisen, S. A., Goldberg, J., et al. (1998). Co-occurrence of abuse of
different drugs in men: The role of drug-specific and shared vulnerabilities. Archives of General Psychiatry, 55, 967 – 972.
Uhl, G. R., Li, Q. -R., Walther, D., Hess, J., & Naiman, D. (2001). Polysubstance abuse-vulnerability genes: Genome scans for
association, using 1004 subjects and 1494 single-nucleotide polymorphisms. American Journal of Human Genetics, 69,
1290 – 1300.
Uhl, G. R., Liu, Q. R., & Naiman, D. (2002). Substance abuse vulnerability loci: Converging genome scanning data. Trends in
Genetics, 18, 420 – 425.
Ulzen, T. P., & Hamilton, H. (1998, Feb.). The nature and characteristics of psychiatric comorbidity in incarcerated adolescents.
Canadian Journal of Psychiatry, 43(1), 57 – 63.
van den Bree, M. B. M., Johnson, E. G., Neale, M. C., & Pickens, R. W. (1998). Genetic and environmental influences on drug
use and abuse/dependence in male and female twins. Drug and Alcohol Dependence, 52, 231 – 241.
Vanyukov, M. M., Tarter, R. E., Kirisci, L., Kirillova, G. P., Maher, B. S., & Clark, D. B. (2003, Oct.). Liability to substance use
disorders: Common mechanisms and manifestations. Neuroscience and Biobehavioral Reviews, 27(6), 507 – 515.