FLSEVIER Psychiatry Research 56 (1995) 213-220

PSYCHIATRY

RESEARCH

A twin study of non-alcohol substance abuse

Lynn M. Gynther*a, Gregory Careya, Irving I. Gottesmanb, George P. Vogler”

‘Institute for Behavioral Genetics, Campus Box 447, University of Colorado, Boulder. CO 80309-0447. USA

bDepartment of Psychology, University of Virginia, Charlottesville, VA 22901, USA

‘Division of Biostatistics, Washington University School of Medicine, St. Louis, MO 63110. USA

Received 25 February 1994; accepted 22 June 1994

Abstract

Except for alcohol abuse, little is known about the familial aggregation for substance abuse. Here we report twin

resemblance for non-alcohol substance use in the Washington University Twin Series, wherein probands were iden-

tified by consecutive admission to psychiatric facilities in the St. Louis area. A 5-point substance abuse scale was con-

structed with values anchored by never used drugs (1) to drug dependence (5). Year of birth was the most powerful

predictor of drug use - younger twins scored far higher than older twins. Either heritability or common environment

had to be included in the regression model to avoid a significant drop in explained variance, but which was more impor-

tant could not be resolved. The correlation for identical twins exceeded that for fraternal twins, suggesting the possibili-

ty of a heritable factor.

Keywords: Drug dependence; Genetics; Age; Sex; Race

1. Introduction

A familial resemblance for alcohol problems has

been observed for centuries. Twin and adoption

studies conducted during the last 30 years have

demonstrated that part of the liability toward alco-

hol abuse and dependence is heritable (Kaij, 1960;

Goodwin, 1979, 1985; Bohman et al., 1981; Clon-

inger et al., 1981; Murray et al., 1983; Merikangas,

1990; Kendler et al., 1992; McGue et al., 1992;

Prescott et al., 1994). The patterns of familial

transmission of non-alcohol substance abuse have

* Corresponding author, Tel: +l 303 492 2826; Fax: +I 303

492 8063; e-mail: gynther@colorado.edu.

received less attention (Pickens and Svikis, 1988,

1991).

Familial background factors such as broken or

unstable homes, urban poverty, mental illness, and

parental alcohol or drug abuse are known cor-

relates of drug dependence (Bell and Trethowan,

1961; Vaillant, 1966; Rosenberg, 1969; Egger et

al., 1978; Cadoret et al., 1986; Miller et al., 1989;

Mirin et al., 1991; Rounsaville et al., 1991). Sub-

stance abusers are also likely to suffer from comor-

bid psychiatric illness, including alcohol abuse or

dependence, antisocial personality, anxiety disor-

der, and affective disorder (Ross et al., 1988;

Regier et al., 1990; Anthony and Helzer, 1991;

Mirin et al., 1991).

01651781/95/$09.50 0 1995 Elsevier Science Ireland Ltd. All rights reserved

SSDI 0165 I78 I (94)02609-M

214 L.M. Gynther et al. /Psychiatry Research 56 (1995) 213-220

Vertical familial transmission that may be spe-

cific for substance abuse has been reported in

several studies (Scherer, 1973; Annis, 1974; Fawzy

et al., 1983; Malhotra, 1983; Meller et al., 1988).

Horizontal transmission, as might occur among

sibling pairs, is less well studied but may be more

important. For example, Vaillant (1966), in a

follow-up study of narcotic addicts treated in the

early 195Os, found 24% of the sample had an ad-

dicted relative, but he considered a hereditary fac-

tor unlikely because the addicted relatives were

almost all siblings rather than parents.

Although many of these correlates of substance

abuse are well documented, the source of the corre-

lation is often vague. Familial transmission may

represent both genetic and environmental in-

fluences, so a sample of twins or adoptees is re-

quired to separate shared heredity from shared

culture. There are two relevant studies that give

genetic information. Using a large but also older

Swedish adoption cohort and their parents, von

Knorring et al. (1983) showed some transmission

of substance abuse that overlapped with affective

disorder; however, specific transmission of sub-

stance abuse could not be determined.

A more detailed analysis of an adoption cohort

in Iowa by Cadoret et al. (1986) found that drug

abuse in adoptees was correlated with a biological

background of alcohol problems in non-antisocial

probands and correlated with an antisocial bio-

logical background in antisocial probands. The

study was unable to resolve the possibility of a spe-

cific vulnerability to drug abuse due to a lack of in-

formation on drug use in biological relatives.

One major problem of examining genetic and

environmental transmission from parent-offspring

relationships is the large cohort effect for sub-

stance abuse (Robins et al., 1984; Anthony and

Helzer, 1991). Many parents, particularly in older

samples, will not have had the same exposure to

drugs as a younger cohort. Hence, it would be

desirable to study twins or adoptive siblings where

cohort effects can be controlled. Unlike parent-

offspring relationships, the importance of drug

availability and drug use by peers should be

reflected in the shared environment of both iden-

tical (monozygotic) and fraternal (dizygotic) twins.

The familial factors implicated in drug abuse

would also be similar for each twin. Monozygotic

twins, however, share the full complement of

genes, whereas dizygotic twins share only half of

their segregating genes on average. If one assumes

the effects of shared environment to be equal for

monozygotic and dizygotic twins, a greater corre-

lation between monozygotic twins would suggest a

heritable component for drug use.

2. Methods

2.1. Subjects

The Washington University (St. Louis, MO.,

USA) twin sample consists of 295 twin pairs iden-

tified from 1980 to 1986 by consecutive admissions

to adult inpatient and outpatient psychiatric units

throughout the St. Louis area. The twins were se-

lected irrespective of diagnosis, although there was

a slight oversampling of probands from addiction

treatment centers.

The sample contained 30% monozygotic (MZ)

twins, 35% same-sex dizygotic (DZ-SS) twins, and

32% opposite-sex dizygotic (DZ-OS) twins.

Zygosity was determined by history of confusabili-

ty and photographs; if there was any question of

zygosity assignment, red blood cell polymorphism

similarity was used. There were no significant dif-

ferences in age, sex, or race between the MZ and

DZ twins. The sample was 53% female and 66%

Caucasian; 98% of the non-Caucasians were

African-American. Most of the non-Caucasians

entered the catchment net from impoverished

inner city areas, whereas the majority of Cauca-

sians were ascertained from more affluent areas.

Therefore, race was strongly confounded with

socioeconomic class in this sample.

Probands and cotwins were administered the

Diagnostic Interview Schedule (DIS), a structured

psychiatric interview (Robins et al., 1981), as well

as life-history and family-history schedules, an in-

terview about the home environment, and a bat-

tery of psychometric questionnaires. The proband

and cotwin were always interviewed separately

and by different interviewers. Records from hospi-

tals, clinics, and private physicians were sought for

all relevant illnesses.

The case history for an individual, including in-

formation from all sources, was presented to a

L.M. Gynther et al. i Psychiatry Research 56 j199S) 21.1-220 215

panel of judges who discussed the case but made tionship), and the product of P and R. R equals 1

independent diagnostic ratings. The ratings were for identical twins and 0.5 for fraternal twins. The

done without knowledge of twin relationship or regression coefficient for the product PR provides

proband/cotwin status. The diagnostic informa- an estimate of heritability, whereas the proportion

tion from these ratings and from the DIS formed of variance due to common environmental influ-

the basis for the drug use scale, used without a ences is obtained from the coefficient of P.

diagnostic hierarchy. Therefore,

2.2. Measures B3 = 2(RMZ - RDZ) = h2 and B, = c2

We constructed a 5-point scale of non-alcohol

substance use, excluding tobacco use disorder. The

scale was coded as follows: (1) never used any

drugs; (2) used any drug to get high, without a

prescription, or used more than prescribed at least

once but not more than 5 times; (3) used any drug

more than 5 times, but had no problems or had an

insufficient number of problems to merit a LLSM-

III diagnosis of abuse; (4) consensus DSM-III

diagnosis of drug abuse; (5) consensus DSM-III

diagnosis of drug dependence. Overall, 57% of the

sample reported some experience with drugs, and

29% had a diagnosis of either abuse or depend-

ence.

One advantage of using the multiple regression

approach is that year of birth can be entered as a

separate variable in the equation, thus testing for

the presence of common environmental effects

separate from a strong secular trend for increasing

drug use in younger subjects. A second advantage

of the regression method is that the polychoric cor-

relation may not be an appropriate estimate when

cell sizes become small, a fact that affects this

study.

The full model used in our analysis was

represented by the equation

C= B, + BIP + B2R + B3PR + BdAge

2.3. Methods of analysis

There are two methods of estimating the corre-

lation for MZ and DZ pairs. The first method is to

treat the 5-point scale as if the points were a series

of thresholds and compute the polychoric correla-

tion. Heritability (h2), the proportion of observed

variance due to genetics, is assessed as twice the

difference between the MZ correlation and the DZ

correlation. Common environment (c2), or the

proportion of observed variance due to en-

vironments that siblings share, is estimated by sub-

tracting the heritability from the MZ correlation.

Common environment is an index of the extent to

which twins resemble each other for environmen-

tal reasons.

where the variable Age indicates the year of birth.

Hence, a positive coefficient would indicate higher

ratings on the drug scale among younger subjects.

These models assume additive genetic effects

and no assortative mating between parents based

on similarity of drug use. Issues in twin method-

ology have been discussed in detail elsewhere

(Gottesman and Carey, 1983; Hrubec and

Robinette, 1984).

3. Results

3.1. Sample characteristics

A second method is to treat the 5-point scale as

a quantitative score and use multiple regression

(DeFries and Fulker, 1985). The multiple regres-

sion model uses the equation

C = B, + B,P + B2R + B3PR

where C (the cotwin’s score) is predicted from P

(the proband’s score), R (the coefficient of rela-

Table 1 presents the mean ages of the subjects

and ratings on the drug use scale. The subjects

ranged in age from 16 to 83, with a mean of 35

years. Non-Caucasians were younger than Cauca-

sians in the total sample (F = 4.33; df = 1, 553; P

< 0.05), but there was no age difference between

races when those subjects who reported any expe-

rience with drugs (a drug-scale rating 22) were

considered.

The mean scores on the drug use scale exhibited

216 L.M. Gynther et al. /Psychiatry Research 56 (1995) 213-220

Table 1

Characteristics of sample: means for age and ratings on drug

use scale

Race&sex

Non-Caucasian 183 34.0 10.9 2.6 1.5

Female 105 34.9 11.2 2.2 1.4

Male 78 32.8 10.4 3.2 1.5

Caucasian 374 36.4 13.8 2.5 1.6

Female 198 36.2 14.9 2.4 1.6

Male 176 36.7 12.4 2.6 1.5

n Agea Drug useb

Mean SD Mean SD

%ignificant effect of race (F= 4.33; df= 1, 553; P < 0.05).

bDrug use scale ranges from 1 to 5. Significant effect of age

(F= 173.71;df = 1,552; P < O.OOl), sex(F= 20.86;df = 1,552;

P < 0.001) and sex x race interaction (F= 7.77; df = 1, 552; P

< 0.01).

a main effect of sex, with males scoring higher than

females (F = 20.86; df = 1, 552; P c 0.001). A

sex x race interaction was also present (F = 7.77;

df = 1, 552; P < 0.01). Non-Caucasian females

had the lowest drug use scores of the four groups,

whereas non-Caucasian males had the highest

scores.

Information on specific drug use was obtained

only from subjects who reported using any drug

more than 5 times (a rating 1 3 on the drug scale)

100, 1

90

VI T 80

0 70

.’ -J 60 5 % 50

E 40

: 30

b a 20

10

0

i?ZZ Females

CAN AMP BAR TRQ COK HER OPI PSY OTH

Type Drugs Used More Than Five Times

Fig. 1. Percent of individuals reporting use more than 5 times

for each specific drug category. CAN, cannabis; AMP, amphet-

amines; BAR, barbiturates; TRQ, tranquilizers; COK, cocaine;

HER, heroin; OPI, other opiates; PSY, psychedelics; OTH,

other.

and only for those drugs used more than 5 times.

Fig. 1 provides a comparison of the prevalence of

specific drugs. Cannabis was the most frequently

reported drug used by both males and females; it

had been used by 91% of drug users. Amphet-

amines were the second most popular drug. The

most pronounced difference between the sexes was

in the proportion of subjects who reported use of

cocaine, heroin, and other opiates. Polydrug use

was prevalent, as 59% of the drug users reported

using more than one drug.

Overwhelmingly, year of birth was one of the

strongest predictors of both drug use and number

of drugs used, and correspondingly, problems of

abuse or dependence. Fig. 2 depicts the dramatic

increase across birth cohorts in the proportion of

males and females who had ever experimented

with drugs (a rating 1 2 on the drug use scale).

Although 50% of the women overall reported no

experience with drugs, compared with 34% of the

men, the difference between the sexes decreased in

later cohorts.

3.2. Genetic anaIyses

Table 2 gives the proband and cotwin

similarities for MZ, DZ-SS, and DZ-OS pairs

across the different levels of the drug use scale.

The polychoric correlations indicate a fairly high

similarity for both MZ and DZ-SS twins, with MZ

males showing the greatest resemblance among the

Secular trends in drug use

1 .0 , 7

% 2

m 0.8.- / / ,.P., ‘-&--_~

‘I’ 72

$ A’

A/’ .a, /‘I

0.6 -- //

2 3 d

0.4 -- / i

6 / 1 6 /’

‘Z

/ b A---A males

z

0.2~- / / k_”

o-0 females

IL

0.0

00-39 40-44 45-49 50-54 55-59 60-68

Birth Cohort

Fig. 2. Secular trends in drug use: proportion of the sample

reporting any illicit drug use as a function of birth cohort.

L.M. Gynther et al. /Psychiatry Research 56 (1995) 213-220 217

Table 2

Proband-cotwin similarity for illicit drug use

MZ males (n = 36) - Polychoric correlation = 0.70

Proband

Cotwin None Tried >5 times Abuse Dependence

None 7 1 6 0 1 Tried 1 0 0 I 2

>5x I 1 2 2 2

Abuse 0 0 2 I 4

Dependence 0 0 0 0 2

DZ mates (n = 36) - Polychoric correlation = 0.55

PrObMd

Cotwin None Tried >5 times Abuse Dependence

None 12 0 1 I 1 Tried 2 1 0 0 2

>5x 2 2 3 0 3

Abuse 1 0 I 0 2

Dependence 2 0 0 0 3

DZ-OS (n = 31) - Polychoric correlation = 0.60

Female proband

Male cotwin None Tried >5 times Abuse Dependence

None 7 I 1 0 0

Tried 2 0 2 0 1

>5x I 0 2 I 3

Abuse 0 0 0 I 2

Dependence 1 1 2 1 2

MZfemates (n = 46) - Polychoric correlation = 0.59

PrOM

Cotwin None Tried >5 times Abuse Dependence

None 17 2 I 4 0

Tried 3 0 0 0 0

>5x 2 I 2 I 4

Abuse 0 0 1 I 2

Dependence 2 0 1 0 2

DZ femates (n = 53) - Polychoric correlation = 0.63

PrObd

Cotwin None Tried >5 times Abuse Dependence

None 22 0 2 0 4

Tried 1 0 0 4 4

>5 times 2 3 1 0 2

Abuse 0 I 0 2 1 Dependence I 0 0 0 3

218 L.&i. Gynther et al. /Psychiatry Research 56 (1995) 213-220

Table 2 (continued)

DZ-US (n = 55) - Polychoric correlation = 0.33

Male proband

Female cotwin None Tried >5 times Abuse Dependence

None 11 6 3 2 7

Tried 2 0 2 0 0

>5 times 1 0 4 2 2

Abuse 1 0 0 1 2

Dependence 1 2 2 1 4

Note. Entries are the number of pairs.

six groups. The difference between the correlations

for MZ males and for DZ males indicates a possi-

ble genetic factor, but the smaller difference and

reversed direction for the female twin types sug-

gests that common environment is more impor-

tant. The DZ-OS pairs have a sex-related trend

toward differential correlations; female probands

have a higher correlation with their male cotwins

than male probands have with their female co-

twins. Such an effect cannot be due to genes and

must reflect intrapair imitation or contrast. The

number of empty cells cautions against overinter-

pretation of the results from the polychoric cor-

relations, but the overall thrust of these analyses

suggests that common environment is indeed

important for twin similarity.

The differences due to sex and race suggest that

an overall genetic analysis should simultaneously

control for these variables and their interaction. In

addition, it would be desirable to control for the

secular trend of increasing drug use. Different

statistical methods used to accomplish this control

resulted in the same general pattern of results. The

following analyses used z scores standardized for

probands and cotwins by sex and race.

Table 3 shows the twin correlations for the stan-

dardized drug use variable. The MZ twin correla-

tion between proband and cotwin was higher than

the DZ twin correlation, but not substantially so.

Table 4

Regression model (C = Bc + B, P + B2R + &PR + lIdAge) for

genetic and environmental factors

Table 3

MZ and DZ correlations for score on drug use scale

Proband score

MZ twins (n = 82)

Cotwin score 0.47*

Birth year 0.49*

Partial correlation for drug use: 0.33

Birth year

0.43;

DZ twins (n = 179)

Cotwin score 0.38*

Birth year 0.52.

Partial correlation for drug use: 0.19

0.46*

Note. Scores are standardized for race and sex.

*P c 0.001.

Model R2 Parameter estimates

Age h2 C2

Full 0.25* 0.33* 0.11 0.13

c2 = 0 0.25* 0.33* 0.23* -

h*=O 0.25* 0.34* - 0.24*

c*=O, h*=O 0.21’3 0.45* - -

Age=0 0.17*a 0.10 0.32

Note. C, cotwin score on drug use scale, standardized for race

and sex; R, coefficient of relationship (0.5 for DZ; 1 for MZ);

P, proband score on drug use scale, standardized for race and

sex (estimates c*); PR, interaction of P and R (estimates h2);

Age, birth year.

‘P < 0.001.

“Significant F change.

L.M. Gynther et al. /Psychiatry Research 56 (1995) 213-220 219

The correlation between year of birth and drug use

was similar in both MZ and DZ twins.

Table 4 shows the genetic and shared environ-

ment components assessed by the regression

model. The values of R2, the proportion of vari-

ance accounted for, were compared for a full

model and for models where one or more effects

were set at zero.

When the measures for both heritability (h2)

and shared environment (c2) were included in the

model, neither was significant. When one of these

measures was removed from the model, however,

the other became significant. There was also a sig-

nificant F change for the overall model when both

were dropped from the analysis. The removal of

the variable for age created the largest loss in R2.

Although it could not be statistically determined

whether it was c2 or h2 that was important, the

partial correlations controlling for year of birth

(0.33 for MZ twins and 0.19 for DZ twins) were

consistent with genetic effects.

4. Discussion

A high degree of similarity was found among

twins for substance use in this sample, part of

which could be attributed to sharing the same

birth date. Year of birth was one of the strongest

predictors of drug use; younger subjects were

much more likely to have tried drugs or to have

received a diagnosis of abuse or dependence.

When analyses controlled for year of birth, the re-

lative importance of genetic and environmental

contributions could not be distinguished, although

the measure of either heritability or shared envi-

ronment needed to be included in the equation to

avoid a significant decrease in the proportion of

explained variance.

Sex and race differences in likelihood of drug

use and liability of drug problems may indicate

differential modes of inheritance or divergent

cultural patterns. Interpretation of the polychoric

correlations, although moderated by a number of

empty cells, would appear to substantiate the

possibility of separate forms of transmission in

males and females. Differential genetic and en-

vironmental factors have been implicated in trans-

mission of antisocial personality and alcohol abuse

(Cloninger et al., 1978; Pickens and Svikis, 1991).

Both alcohol abuse and abuse of other drugs tend

to cluster with diagnoses of antisocial personality,

a disorder shown to be genetically transmitted

separately from alcohol abuse. Antisocial person-

ality is manifest at a consistently higher frequency

in males, suggesting that the presence of antisocial

personality may somehow moderate or be associ-

ated with the genetic expression of drug use.

The partial twin correlations for the combined

sample, with controls applied for year of birth,

were higher for MZ twins than for DZ twins, but

the difference between them was not significant.

This suggested a possible influence of genetic

effects, but a greater number of subjects would be

needed to provide the power to test the hypothesis.

Acknowledgements

This study was supported by U.S. Public Health

Service grants MH-31302, AA-03539, and DA-

05131. The help of Dr. Jack Knesevich is ap-

preciated.

References

Annis, H.M. (1974) Patterns of intra-familial drug use. Er J Ad-

dict 69, 361-369.

Anthony, J.C. and Helzer, J.E. (1991) Syndromes of drug abuse

and dependence. In: L.N. Robins and D.A. Regier (Eds.),

Psychiatric Disorders in America: The Epidemiologic Catch-

ment Area Study. The Free Press, New York, pp. 116-I 54.

Bell, D.S. and Trethowan, W.H. (1961) Amphetamine addic-

tion. J Nerv Ment Dis 133, 489-496.

Bohman, M., Sigvardsson, S. and Cloninger, C.R. (1981) Ma-

ternal inheritance of alcohol abuse: cross-fostering analysis

of adopted women. Arch Gen Psychiatry 38, 965-969.

Cadoret, R.J., Troughton, E., O’Gonnan, T.W. and Heywood,

E. (1986) An adoption study of genetic and environmental

factors in drug abuse. Arch Gen Psychiatry 43. I I3 I-1 136.

Cloninger, C.R., Bohman, M. and Sigvardsson. S. (1981) ln-

heritance of alcohol abuse: cross-fostering analysis of

adopted men. Arch Gen Psychiatry 38. 861-868.

Cloninger, CR., Christiansen, K.O., Reich, T. and Gottesman,

1.1. (1978) Implications of sex differences in the prevalences

of antisocial personality, alcoholism, and criminality for

familial transmission. Arch Gen Psychiatry 35, 941-995.

&Fries, J.C. and Fulker, D.W. (1985) Multiple regression

analysis of twin data. Behav Genet 15, 467-473.

Egger, G.J., Webb, R.A.J. and Reynolds, 1. (1978) Early ado-

lescent antecedents of narcotic abuse. In1 .I Addict 13,

773-78 I.

220 L.M. Gynther et al. /Psychiatry Research 56 (1995) 213-220

Fawxy, RI., Coombs, R.H. and Gerber, B. (1983) Genera-

tional continuity in the use of substances: the impact of par-

ental substance use on adolescent substance use. Addict

Eehav 8, 109-l 14.

Goodwin, D.W. (1979) Alcoholism and heredity. Arch Gen

Psychiatry 36, 57-61.

Goodwin, D.W. (1985) Alcoholism and genetics: the sins of the

fathers. Arch Gen Psychiatry 42, 171- 174.

Gottesman, 1.1. and Carey, G. (1983) Extracting meaning and

direction from twin data. Psychiatr Dev 1, 35-50.

Hrubec, Z. and Robinette, C.D. (1984) The study of human

twins in medical research. N Engr J Med 310, 435-441.

Kaij, L. (1960) Alcoholism in Twins: Studies on the Etiology and

Sequels of Abuse of Alcohol. Vol. I. Almqvist & Wiksell In-

ternational, Stockholm.

Kendler, K.S., Heath, A.C., Neale, M.C., Kessler, R.C. and

Eaves, L.J. (1992) A population-based twin study of alco-

holism in women. JAMA 268, 1877-1882.

Malhotra, M.K. (1983) Familial and personal correlates (risk

factors) of drug consumption among German youth. Acta

Paedopsychiatr 49, 199-209.

McGue, M., Pickens, R.W. and Svikis, D.S. (1992) Sex and age

effects on the inheritance of alcohol problems: a twin study.

J Abnorm Psycho1 101, 3-17.

Meller, W.H., Rinehart, R., Cadoret, R.J. and Troughton, E.

(1988) Specitic familial transmission in substance abuse. Int

J Addict 23, 1029-1039.

Merikangas, K. (1990) The genetic epidemiology of alcoholism.

Psychiatr Med 20, 1 l-22.

Miller, N.S., Gold, M.S., Belkin, B.M. and Klahr, A.L. (1989)

Family history and diagnosis of alcohol dependence in co-

caine dependence. Psychiatry Res 29, 113- 12 1.

Mirin, SM., Weiss, R.D., Griffin, M.L. and Michael, J.L.

(1991) Psychopathology in drug abusers and their families.

Compr Psychiatry 32, 36-51.

Murray, R., Clifford, C. and Gurling, H. (1983) Twin and

adoption studies: how good is the evidence for a genetic

role? In: Galanter, M. (Ed.), Recent Developments in Alco-

holism. Vol. 1. Plenum Press, New York, pp. l-58.

Pickens, R.W. and Svikis, D.S. (1988) Genetic vulnerability to

drug abuse. Nat1 Inst Drug Abuse Res Monogr 89, l-8.

Pickens, R.W. and Svikis. D.S. (1991) Genetic influences in

human substance abuse. J Addict Disord 10, 205-213.

Prescott, CA., Hewitt, J.K., Truett, K.R., Heath, A.C., Neale,

M.C. and Eaves, L.J. (1994) Genetic and environmental in-

fluences on lifetime alcohol-related problems in a volunteer

sample of older twins. J Stud Alcohol 55, 184-202.

Regier, D.A., Farmer, ME., Rae, D.S., Locke, B.Z., Keith,

S.J., Judd, L.L. and Goodwin, F.K. (1990) Comorbidity of

mental disorders with alcohol and other drug abuse: results

from the Epidemiologic Catchment Area (ECA). JAMA

264, 2511-2518.

Robins, L.N., Helxer, J.E., Croughan, J. and Ratcliff, K.S.

(1981) National Institute of Mental Health Diagnostic In-

terview Schedule: its history, characteristics and validity.

Arch Gen Psychiatry 38, 381-389.

Robins, L.N., Hekrer, J.E., Weissman, M.M., Orvaschel, H.,

Gruenberg, E., Burke, J.D. and Regier, D.A. (1984)

Lifetime prevalence of specific psychiatric disorders in three

sites. Arch Gen Psychiatry 41, 949-958.

Rosenberg, CM. (1969) Young drug addicts: background and

personality. J Nerv Ment Dis 148, 65-73.

Ross, H.E., Glaser, F.B. and Germanson, T. (1988) The preva-

lence of psychiatric disorders in patients with alcohol and

other drug problems. Arch Gen Psychiatry 45, 1023-1032.

Rounsaville, B.J., Kosten, T.R., Weissman, M.M., Prusoff, B.,

Paula D., Anton, S. and Merikangas, K. (1991) Psychiatric

disorders in relatives of probands with opiate addiction.

Arch Gen Psychiatry 48, 33-42.

Scherer, SE. (1973) Self-reported parent and child drug use. Br

J Addict 68, 363-364.

Vaillant, G.E. (1966) A 12-year follow-up of New York narcot-

ic addicts: II. some social and psychiatric characteristics.

Arch Gen Psychiatry 15, 599-609.

von Knorring, A-L., Cloninger, C.R., Bohman, M. and

Sigvardsson, S. (1983) An adoption study of depressive dis-

orders and substance abuse. Arch Gen Psychiatry 40,

943-950.