Intra- and extra-familial in uences on alcohol and drug misuse: a twin study of gene environment correlation

Transcription

1 Addiction (001) 96, RESEARCH REPORT Intra- and extra-familial in uences on alcohol and drug misuse: a twin study of gene environment correlation KERRY L. JANG, 1 PHILIP A. VERNON, W. JOHN LIVESLEY, 1 MURRAY B. STEIN 3 & HEIKE WOLF 4 1 Division of Behavioural Science, Department of Psychiatry, University of British Columbia, Vancouver, British Columbia, Department of Psychology, University of Western Ontario, London, Ontario, Canada, 3 Department of Psychiatry, University of California San Diego, La Jolla, California, USA & 4 Department of Psychology, Universität Bielefeld, Bielefeld, Germany Abstract Aims. Genotype environment correlation refers to the extent to which individuals are exposed to environments as a function of their genetic propensities. These correlations are important in the study of psychopathology because they identify environments that may maintain the expression of underlying genetic liabilities for a disorder. The present study examined the correlation between genetic liabilities for alcohol and drug misuse with perceptions of the social environments of the family of origin and the classroom. Design. Postal survey data were collected from monozygotic and dizygotic twin pairs. Setting. Twin pairs were recruited from Vancouver, British Columbia, Canada using newspaper advertisements and media stories. Participants. Eighty- ve monozygotic and 77 dizygotic twin pairs were recruited from the general population. Measurements. Twin pairs completed self-report measures of alcohol and drug misuse contained in the Dimensional Assessment of Personality Pathology, the Family Environment Scale, the Classroom Environment Scale, and the Traumatic Events Questionnaire. Findings. Genetically indexed alcohol and drug misuse scores were regressed on the environmentally indexed FES and CES scales. Genetic liabilities for alcohol and drug misuse were associated with decreased perceived family moral-religious emphases, family cohesion and classroom task orientation and increased perceptions of classroom order and organization (strictness). Conclusions. Genotype environment correlations, in particular, moral-religious emphases in the home, appear to be important in the development of substance misuse. Introduction One of the primary purposes of behavioural genetic studies is to partition and estimate the degree to which individual differences in behaviour can be explained by heritable and non-heritable differences. Not unexpectedly, twin, adoption and family studies of major behavioural indices such as personality scales (e.g. Bouchard, 1997) 1 1 Correspondence to: Dr K. L. Jang, Division of Behavioural Science, Department of Psychiatry, University of British Columbia, 55 Wesbrook Mall, Vancouver, BC, Canada. V6T A1. Tel: 1 (604) ; fax: 1 (604) ; kjang@unixg.ubc.ca Submitted 19th June 000; initial review completed 13th December 000; nal version accepted 9th March 001. ISSN print/issn online/01/ Ó Society for the Study of Addiction to Alcohol and Other Drugs Carfax Publishing, Taylor & Francis Limited DOI: /

2 1308 Kerry L. Jang et al. and most measures of substance use (e.g. McGue, 1994) show that genetic factors play a role in the expression of these phenotypes. What was not expected to be heritable was a person s report of the external environmental conditions surrounding him or herself. These have, however, been demonstrated reliably to have a substantial heritable basis (e.g. Plomin et al., 1990; Vernon et al., 1997). Subsequent analyses determined that the genetic in uences accounting for the individual differences in personality trait scales also in uenced self-reports of an individual s environment. For example, Saudino et al. (1997) showed that all genetic variance on controllable, desirable and undesirable life events in women was common to the genetic in uences underlying Neuroticism and Extraversion from the Eysenck Personality Questionnaire, and the Openness to Experience scale from a short version of the NEO Personality Inventory (NEO-PI: Costa & McCrae, 1985). Genetic in uences underlying personality scales had little in uence on uncontrollable life events simply because this variable was not heritable. In another study, Jang, Livesley & Vernon (000) estimated the degree to which self-report measures of personality dysfunction, heritable scales from the Family Environment Scale (FES: Moos & Moos, 1994), and present-day alcohol misuse shared a common genetic basis. The study found that the genetic factors in uencing personality also in uenced perceptions of the family environment, but their in uence did not extend to alcohol misuse. Although these studies assessed the nature and magnitude of pleiotropic in uences between measures, that is, the degree to which two or more variables share a common aetiological basis, they also raised the possibility that the environment does not have an effect independent of pre-existing genetic factors (see Reiss et al., 000 for an extensive review). For example, an individual s personality or psychopathology plays a signi cant role in the selection and creation of their own environment. This genotype environment relationship is described by a term from quantitative genetic theory referred to as genotype environment correlation, the extent to which individuals are exposed to environments as a function of their genetic propensities. Three general types of genotype environment correlation have been hypothesized: passive, active and reactive (Plomin, DeFries & Loehlin, 1977). Passive genotype environment correlation occurs because children share heredity and environments with members of their family and can thus passively inherit environments correlated with their genetic propensities. Reactive genotype environment correlation refers to experiences of the child derived from reactions of other people to the child s genetic propensities. Active genotype environment correlation is known as niche-building or niche picking (Plomin, DeFries & McClearn, 1990, p. 51). This occurs when children actively select or create environments commensurate with their underlying genetic propensities. Genotype environment correlations are central to studies of psychopathology, such as substance use, because they identify sources of environmental in uence that either activate or were selected to maintain the underlying genetic liabilities for the behaviour in question. A straightforward way to test for genotype environment correlation for substance use would be to correlate a true measure of the environment, that is, a scale that assesses a person s perception of external conditions that has no demonstrable heritability, with a measure of substance misuse that is completely heritable. Such an approach could be tested if validated scales with these characteristics existed. In this regard, there exist several measures assessing a person s perceptions of the environment that show no heritable component. For example, environmental factors accounted for all of the variance in seven of 10 FES scales (e.g. Vernon, et al., 1997; Jang, Vernon & Livesley, 000). Identi cation of measures of substance use that are entirely heritable is more problematic. Although there is disagreement regarding the relative magnitude of the in uence of genetic and environmental in uences on alcohol use (e.g. Pickens et al., 1991; McGue, Pickins & Svikis, 199), the studies agree that both variance components are required to account for the variability on these measures. Deriving sets of scores that re ect only the genetic and environmental in uences on substance use solves this problem. Such genetically pure scores could then be correlated with the non-heritable scales of the environment to assess the presence of genotype environment correlations between the measures of interest. Computational methods to derive scores that re ect only the genetic and environmental in uences have been appearing in the be-

3 Gene environment correlation 1309 havioural genetic literature (e.g. Sham et al., 000; Thomis et al., 000). These methods make it possible to estimate scores for the common genetic and environmental factors extracted among several items or scales. The value of this factor score approach is that several multivariate genetic analyses of substance use measures exist (e.g. Jang et al., 1995), based on large sample sizes, that may be used to derive the genetic and environmental factor scores for any comparable sample which has completed the same measures. It should be noted that unless the items composing a scale are genetically unifactorial in nature, it is not possible to derive a score for the scale per se, but rather only for its underlying factorial structure. The environmentally pure scores of substance use also provide a means to address the question of the nature of non-shared environment. As noted above, most heritability estimates of psychiatric disorder account for about half of the total variance, with the remaining half attributable to environmental in uences. There are two types of environmental variance that caused by experiences and conditions that are in common to or shared by all members of the family, and those that are unique or nonshared to each member of the family. Twin studies of substance use have shown that most of the environmental variance is of the non-shared variety (e.g. Prescott et al., 1994). However, despite the known in uence of non-shared environmental in uences, identi cation of what they are remains elusive (e.g. Hetherington, Reiss & Plomin, 1994; Turkheimer & Waldron, 000). Thus, it follows that the correlation of substance use scores that were transformed to only index the in uence of non-shared environment with a broad range of non-heritable scales of the environment, would make it possible to identify non-shared environmental factors that have a direct impact on behaviour. The purpose of the present study is to describe a method to test for gene environment correlation and identify speci c aspects of the nonshared environment that have a direct impact on substance misuse. This was accomplished by using behavioural genetic techniques to create measures of alcohol and drug misuse that re ect only the in uences of either genetic or environmental factors. These genetically and environmentally indexed factor scores of substance misuse were then related to non-heritable reports of the environment assessing a range of intraand extra-familial environmental conditions. The primary advantage of the method described and used here is that it provides a means to derive indices of substance use and perceptions of the environmental conditions that are not confounded by competing genetic and environmental in uences. Any interrelationships uncovered between aetiologically distinct substance misuse and environmental variables begin to address the nature of the interplay between genes and the environment on the development of substance misuse. Method Subjects The sample comprises 86 identical or monozygotic (MZ) and 77 fraternal or dizygotic (DZ) twin pairs. The MZ sample consists of 58 sister pairs (mean age years, SD years, range years) and 8 brother pairs (mean age years, SD 5 8.9, range years). The DZ sample includes 37 sister pairs (mean age years, SD , range years), 16 DZ brother pairs (mean age years, SD , range years) and 4 sister brother pairs (mean age years, SD , range years). Zygosity was determined through a questionnaire (Nichols & Bilbro, 1966) that has been shown to be at least 95% as accurate as red blood cell polymorphism analyses (Kasriel & Eaves, 1976). All twin pairs were volunteers recruited through newspaper advertisements and media stories from Vancouver, Canada. A condition of participation in the study was that the twin pairs must have been raised together in the same home with no signi cant separations greater than 1 month. Portions of the data used in this study were previously published in Jang et al. (000). The data obtained from twins who participated in the present study were recruited independently of the twins who completed the FES and CES reported in Vernon et al. (1997). Measures All twin pairs completed a battery of questionnaires at home and were instructed to complete the questionnaires independently of one another in a non-distracting setting. Questionnaires were returned by mail and all participants received a

4 1310 Kerry L. Jang et al. cash honorarium for their participation. A total of 07 pairs were mailed the questionnaire battery. Data from only one member of a pair (discordant-participant) was received from 18 pairs (8.7%) and data from both pairs (concordant-participant) was received from 16 pairs (78.3%). Comparison of the raw scale scores between the discordant-participant and concordant-participant pairs provides a broad means to test the representativeness of the sample (Neale & Cardon, 199). In the usual twin design, both members of a twin pair will initially volunteer to participate in the study. If one member of the pair is dysfunctional, only the healthy (or healthier) member of the pair may complete the study. Because they are twins, there is an expectation that data from the participant twin from a discordant-participant pair will be similar (especially among the MZ pairs) to the non-participant twins, and this could be used as a rough estimate of the level of psychopathology existing on the discordant-participant pair. Signi cant differences in test scores between concordant- and discordant-participant pairs would suggest that the concordant-participant pairs are not representative of the general population (that is, they are too healthy). In the present sample no signi cant differences were observed on the raw substance misuse items (described below) between the concordant-participant and the discordant-participant of either zygosity. The primary measure of substance misuse was eight items that compose the Addictive Behaviours subscale of the Dimensional Assessment of Personality Pathology Differential Questionnaire (DAPP-DQ: Livesley & Jackson, 001; Livesley, Jang & Vernon, 1998). The eight items composing this scale are: I often drink too much; I have consumed so much alcohol at times that I could not remember what happened; There have been times in my life I have repeatedly used alcohol to excess; Alcohol has got me into trouble on a number of occasions; I have used a number of illicit drugs; I use drugs frequently; Drugs have sometimes caused me to take time off work or school; I rarely use drugs unless prescribed by a doctor. Study participants were instructed to answer the questions with reference to their behaviour in the past 6 8 weeks or -month period. The internal consistency (Cronbach s a ) for the drug misuse items was estimated at 0.75 and for the alcohol misuse items at 0.9. Cronbach s a for the total eightitem Additive Behaviours scale was 0.87 in the present sample. One twin was selected randomly from each pair for the reliability computations. Intra-familial environment was measured by the Family Environment Scale (FES: Moos & Moos, 1994): a 90-item true/false scale of a person s perceptions of his or her conjugal or nuclear family environment. For the present study, subjects were instructed to rate the family environment of the family of origin (see also Vernon et al., 1997). The 10 FES subscales are divided into three domains: Relationships (Cohesion, Expressiveness, and Con ict), the degree to which Personal Growth is emphasized in the family (Independence, Achievement Orientation, Intellectual-Cultural Orientation, Active-Recreational Orientation and Moral- Religious Emphasis) and system maintenance issues (Organization and Control). The internal consistencies of the FES scales reported in the manual (Cronbach s a ) ranged from 0.61 (Independence) to 0.78 (Moral-Religious Emphasis), and 4-week test retest reliability (Kuder- Richardson 0) ranged from 0.54 (Independence) to 0.91 (Moral-Religious Emphasis). The Classroom Environment Scale (CES: Moos & Moos, 1993) was used to assess extrafamilial environment. This scale was designed to measure students perceptions of the classroom environment. The scale contains 90 true/false items that, like the FES, assess three broad domains of classroom environment. The rst domain assesses relationship issues with the Involvement, Af liation, and Teacher Support subscales. The second domain assesses personal growth and goal orientation issues with the Task Orientation and Competition subscales. The nal domain is known as system maintenance and change, de ned by the Task Orientation, Rule Clarity, Teacher Control and Innovation scales. The test manual reports satisfactory psychometric properties for each of the subscales. Cronbach s a ranges from 0.67 (Competition) to 0.86 (Teacher Control), and 6-week test retest reliabilities (Kuder-Richardson 0) were reported to range from 0.7 (Rule Clarity) to 0.90 (Innovation). For the present study, the subjects were asked to rate the environment of the classroom they remembered best from high school. The Traumatic Events Questionnaire is a list of traumatic life experiences that assesses lifetime occurrences of traumatic events derived from an

5 Gene environment correlation 1311 item set used in a community survey (Stein et al., 1997). The present study used eight items from this survey to assess the occurrence of two broad forms of traumatic events: assaultative and non-assaultative events. Items are scored on a two-point scale (1 5 yes, 0 5 no/don t know) for three different time periods between the ages of: 0 6, 7 16 and The score for each item is the sum of the responses for each age category. The non-assaultative events subscale is composed of four items: Did a close friend or family member ever die because of an accident, homicide or suicide?; Were you ever in a motor vehicle accident serious enough to cause injury to one or more of the passengers?; Did you ever suffer injury or property damage because of re?; Did you ever suffer property damage because of severe weather, like tornado, or a natural disaster like an earthquake or ood? Cronbach s a for these items in the present sample was estimated at 0.4, which supports the assumption that the occurrence of these events is largely random with respect to one another. The assaultative events scale is composed of the following four items: Did anyone ever take something from you by force or threat of force, such as in a robbery, mugging or hold-up?; Were you ever threatened with a weapon, held captive, or kidnapped?; Did anyone ever beat you up or attack you?; Did anyone ever make you have unwanted sexual contact by using force, pressure, or threats to harm you? Cronbach s a for these items in the present sample was estimated at 0.70, supporting the assumed interdependent nature of the items. Biometric analyses The rst step in the data analysis was to compute the genetically and environmentally indexed factor scores for the alcohol and drug misuse items. Jang et al. (1995) tted multivariate genetic models to the eight items that had additive genetic in uences (a), environmental in uences shared in common (c) or non-shared environmental in uences (e) in common. Additive genetic in uences represent the extent to which genotypes breed true from parent to offspring. Shared environmental in uences distinguish the general environment of one family from another and in uence all children within a family to the same degree (Rowe, 1994). Non-shared environmental factors (see Hetherington et al., 1994) include events that have differential effects on individual family members (e.g. pre- and postnatal traumas, differential parental treatment). It should be noted that e is not estimated directly but constitutes the residual variance after the effects of a and c have been removed. As such, this component also contains random error variance. The analyses presented in Jang et al. (1995) extracted three additive genetic factors and three environmental factors. The rst genetic factor was de ned by the items: I have consumed so much alcohol at times that I could not remember what happened; Alcohol has got me into trouble on a number of occasions; and Drugs have sometimes caused me to take time off work or school. These items describe pathological substance misuse. The second genetic factor is de ned primarily by the drug misuse items: I have used a number of illicit drugs; I use drugs frequently; and I rarely use drugs unless prescribed by a doctor. These items describe nonpathological drug misuse. All the alcohol misuse items, describing genetically based nonpathological alcohol misuse, largely de ne the third genetic factor. The rst environmental factor extracted in these analyses had signi cant loadings on all eight items, indicating that there is a class of environmental variables shared by all members of a family (c) that in uence alcohol and drug misuse ( shared environmental substance misuse ). The next two environmental factors described events that are unique to each member of the family. The rst is de ned by signi cant loadings from all alcohol and drug misuse items ( non-shared environmental substance misuse ). Only the drug misuse items loaded on the third factor, identifying it as nonshared environmental drug misuse. Scores for each of these genetic and environmental factors were computed using the following equation: y 5 c R 1 x Where y 5 factor score for the common genetic or environmental factor, c 5 vector of estimated path coef cients which represent the correlations between the common genetic or environmental factor and the observed measures, R 1 5 inverse of the correlation matrix of the observed measures, and x 5 column vector of observed values on the measures (Sham et al., 001). The values of c were taken from Jang et al. (1995) and

6 131 Kerry L. Jang et al. multiplied by elements of R 1 and values of x obtained from subjects in the present sample to yield a value of y for each subject on each of the genetic and environmental factors. The computer program Mx (Neale et al., 1999) was used to carry out the matrix functions. The distribution of y for each of the factors was normally distributed. Heritability analyses A primary assumption of the twin method is the assumption of equal environments in which it must be shown that the similarity of MZ twins relative to the DZ twins is due to the two-fold greater genetic similarity of MZ twins to DZ twins and not because MZ twins were treated more similarly than DZ twins that is to say, the environments of the MZ twins are no more similar than the environments of the DZ twins. This was tested by endorsement rates of several questions assessing the degree to which twins were treated similarly. We assessed the MZ and DZ similarity of treatment by summing the responses to the seven items (true/false) that ask if: twins spend most of their time together; attend the same school; have the same friends; tend to dress alike; are in most of the same classes at school; if parents treat them very much the same; and had the same teachers. MZ twins were on average found to be treated signi cantly more similarly than the DZ twins (t , p , df 5 34). Correlating similarity of treatment with each FES, CES subscale and genetic and environmental factor scores tested the impact of this violation of the equal-environments assumption. This variable was found to have a signi cant correlation only with FES: Family Cohesion (r , p, 0.01), accounting for 3.40% of the total variance on the scale. Demographic variables such as age and gender may in uence the magnitude of scores, which has the principal effect of biasing heritability estimates (see McGue & Bouchard, 1984). All data were adjusted for these effects by computing the standardized residual of the simultaneous regression of each score on age and gender prior to heritability analysis and the computation of the genetic and environmentally indexed factor scores to correct for this potentially biasing effect. With the adjustments made to the scale scores, univariate heritability analyses of the FES, CES and TEQ subscales proceeded to identify scales that were in uenced by large environmental (shared and non-shared environmental in uences) in uences; that is, not in uenced by any genetic in uences to a signi cant degree in the present sample. Although the present sample is modest in size (85 MZ and 77 DZ pairs), power analyses using the method detailed in Neale and Cardon (199) show that the present sample is suf ciently large to detect scales whose environmental in uences comprise 91% of the total variance (e.g. c 5 41% and e 5 50%) with 95% con dence. The signi cance (p, 0.05) of each of each of the a, c and e parameters was tested by t-test. The heritabilities of the FES, CES and TEQ scales in the present sample were estimated employing the standard biometrical modelling methods as described in Neale & Cardon (199). The LISREL 8 computer program (Jöreskog & Sörbom, 1993a) was used to t a model specifying additive genetic, shared and non-shared environmental in uences to the MZ and DZ twin covariances by the method of maximum likelihood. The MZ and DZ intra-pair covariances were estimated using the computer program PRELIS (Jöreskog & Sörbom, 1993b). Overall model- t was assessed using likelihood ratio v and root mean square error of approximation (RMSEA) that estimates the goodness-of-approximation to a true model in the population (see Steiger, 1989; Loehlin et al., 1998). Note that the parameter estimates a, c, and e must be squared to yield the familiar h, c and e statistics that index the proportion of the total variance on each of the scales attributable to additive genetic, shared and non-shared environmental in uences. Gene environment correlation was tested with a stepwise multiple regression of the FES, CES, and TEQ subscales in which the additive genetic in uence was non-signi cant (p, 0.05, t-test) on the genetically indexed substance use factor scores. A second set of stepwise multiple regression analyses regressed the shared and nonshared substance use factors on each of the not signi cantly heritable FES, CES and TEQ subscales to identify speci c features of the shared and non-shared environment of siblings directly related to substance misuse. Results The proportion of the sample that endorsed each

7 Table 1. Alcohol and drug misuse item endorsement rates Gene environment correlation Very Moderately Somewhat Moderately Very Item* like me like me unlike me like and unlike me like me Alcohol misuse Alc 1: Drunk too much 48.7% 14.1% 14.9% 11.3% 11.1% Alc : Could not remember 65.4% 1.4% 7.8% 6.8% 7.6% Alc 3: Repeated excessive 56.9% 10.8% 9.5% 13.3% 9.5% Alc 4: Got into trouble 67.0% 11.1% 9.0% 7.1% 5.8% Drug misuse Drg 1: Used a number of illicit drugs 64.7% 10.0% 9.5% 10.3% 5.6% Drg : I use drugs frequently 80.3% 7.0% 6.3% 3.5%.9% Drg 3: Take time off work or school 86.1% 6.1% 3.8%.5% 1.5% Drg 4: Rare use unless prescribed 8.5% 6.1% 8.% 1.4% 67.0% * Refer to text for full text of the item; N 5 34 twin individuals. response category for the four alcohol and four drug misuse items are presented in Table 1. The MZ and DZ intra-pair correlations for the 10 FES, nine CES and two TEQ subscales are presented in Table. The MZ and DZ intra-pair correlations provide a broad indication of the magnitude of any heritable effects that may be present on any scale. Examination of the rst two columns of Table show that the MZ correlations exceed the DZ correlation on most Table. Twin correlations (Pearson s r) and parameter and standard error estimates for a model estimating additive genetic, shared and non-shared environmental variance on the Family Environment Scale, Classroom Environment Scale and the Traumatic Events Questionnaire Scales r MZ r DZ a c e Family Environment Scale Cohesion * * * Expressiveness NS * * Con ict NS * * Independence NS * * Achievement orientation * 0.00 NS * Intellectual-cultural orientation * NS * Active-recreational orientation * * * Moral-religious emphasis NS * * Organization NS * * Control NS * * Classroom Environment Scale Involvement NS * * Af liation * 0.00 NS * Teacher support NS * * Task orientation NS * * Competition NS * * Task orientation NS * * Rule clarity NS * * Teacher control NS * * Innovation * 0.00 NS * Traumatic Events Questionnaire Assaultative events * NS * Non-assaultative events NS * * 1 r MZ. r DZ, p, 0.05; r MZ, r DZ, p, 0.05; a, additive genetic in uences; c, shared environmental in uences; e, non-shared environmental in uences; a, c and e are parameter estimates and must be squared to yield the familiar h, c, and e statistics that index the proportion of the total variance on each of the scales attributable to additive genetic, shared and non-shared environmental in uences; ns non-signi cant p. 0.05; * p, 0.05; N MZ 5 85 pairs; N DZ 5 77 pairs.

8 1314 Kerry L. Jang et al. FES scales, but that the MZ correlation signi cantly (p, 0.05 Fisher s Z r transformation) exceeds the DZ correlations on only two scales: Achievement Orientation and Intellectual- Cultural Orientation, suggesting the presence of large heritable in uences on these two scales. For the remaining scales, the MZ and DZ twin correlations are not signi cantly different suggesting that heritable in uences, if any (as indicated when r MZ, r DZ), are negligible and that environmental in uences account for the greatest proportion of the total variance. Environmental in uences also appear to account for the majority of the variance on the CES scales. Only on the Innovation scale did the MZ correlation signi cantly exceed the DZ correlation, suggesting that this scale has a signi cantly large heritable component. The twin correlations also suggested the presence of a sizable heritable in uence on the Assaultative Events scale, but not the Non-Assaultative scale from the TEQ. Table also presents the results of the heritability analyses. The table presents the maximum likelihood parameter estimates and their standard errors for additive genetic (a) and shared (c) and non-shared (e) environmental in uences for each subscale. These analyses con rmed the broad indications drawn from the examination of the MZ and DZ intra-pair twin correlations alone. Heritable in uences were found to account for a signi cant (p, 0.05, t-test) proportion of the variance in FES Intellectual- Cultural Orientation, Active-Recreational Orientation, Cohesion and Achievement Orientation. With the CES, signi cant heritable in uences were detected on the Innovation and Af liation scales. For the TEQ, signi cant genetic in uence was detected on the Assaultative Events scale as expected. Given the signi cant presence of genetic in uence on these seven environmental measures subscales, they were excluded from the next step in the analyses. The top of Table 3 presents the results of the stepwise multiple regression of each genetic substance use factor on all of the non-signi cantly heritable FES, CES, and TEQ subscales used to test for gene environmental correlation between substance misuse and the environment. Scores from one randomly selected member of each pair were included for analysis. The rst genetic factor that indexes pathological substance misuse was found to be associated with decreased FES Moral-Religious Emphasis, FES Cohesion, CES Task Orientation and increased Classroom Organization and Orderliness, F , p , accounting for approximately 4% of the variance in genetically based pathological alcohol and drug misuse. For the second genetic factor indexing non-pathological drug misuse, no FES or CES subscale emerged as having a signi cant Table 3. Stepwise multiple regression of the genetic and environmental factor scores for alcohol and drug misuse on measures of the environment that are not in uenced by heritable factors Factors Predictors b R Adjusted Genetic factors 1. Pathological alcohol and drug misuse FES: Moral-religious emphasis FES: Family cohesion 0.5 CES: Task orientation 0.40 CES: Order and organization Non-pathological drug misuse 3. Non-pathological alcohol misuse FES: Family organization 0.75 FES: Moral-religious emphasis Shared environmental factors 1. All alcohol and drug misuse FES: Moral-religious emphasis Non-shared environmental factors 1. All alcohol and drug misuse FES: Moral-religious emphasis CES: Task orientation CES: Order and organization FES: Family cohesion All alcohol misuse only N MZ 5 85 pairs; N DZ 5 77 pairs.

9 Gene environment correlation 1315 association with this factor. Two variables accounted for a signi cant proportion (14%) of the variance on the third genetic factor, nonpathological alcohol misuse. These subscales are FES Moral-Religious Emphasis and Family Organization (F , p ). The remainder of Table 3 presents the results of the stepwise multiple regression analyses designed to identify sources of shared and nonshared environmental in uence on substance misuse. These analyses regressed each of the shared and non-shared substance use factors on each of the non-signi cantly heritable FES, CES and TEQ subscales. The regression of the shared environmental factor indexing all alcohol and drug misuse items was found to have a negative association with FES Moral-Religious Orientation. This relationship accounted for approximately 9% of the total variance on this factor, F 5 8.7, p The rst non-shared environmental substance misuse factor was found to have a negative relationship with FES Moral-religious Emphasis, CES Task Orientation, FES Family Cohesion and a positive association with CES Classroom Organization and Orderliness. These environmental subscales accounted for approximately 3% of the total variance on this factor, F , p The second non-shared environmental factor indexing non-shared environmental in uences on drug misuse only was not associated with any FES, CES or TEQ scale. Discussion Genotype environment correlation describes the extent to which individuals are exposed to environments as a function of their genetic propensities (e.g. Plomin et al., 1977). The present study assessed genetic liabilities to alcohol and drug misuse with perceptions of the social environments of the family of origin and the classroom. Genotype environment correlations are important in the study of psychopathology because they identify the environments that may activate or maintain the expression of underlying genetic liabilities for disorder. The present study identi ed two genotype environment correlations for alcohol and drug misuse. The primary nding of the present study was that pathological alcohol and drug misuse was correlated with decreases in perceived family moral-religious emphases, family cohesion, and classroom task orientation and increased perceptions of classroom order and organization (strictness). It was also shown that non-pathological alcohol misuse was associated with perceived breakdown or lack of family organization and family moral and religious emphases in the family. It is noteworthy that these results highlight the importance of family values and the in uence that parents have on the development of behaviours in their children. The results run counter to some theories that a person s peers as opposed to families are important in development (e.g. Harris, 1998). The regression of the alcohol and drug misuse scores directly attributable to this variance component on the measures of family and classroom environments begins to identify the nature of the non-shared environment. The present analyses have tentatively identi ed some of the sources as the degree to which issues of a moral-religious nature are emphasized in the family, the degree of cohesion, and the degree to which classroom task orientation, order and organization are emphasized. It is interesting to note that the same FES and CES variables, particularly FES Moral- Religious Emphasis, emerge over most of the analyses with both the genetic and environmental factor scores. The plausibility of variables such as moralreligious emphasis as a factor in alcohol use has been demonstrated in several studies and the present results are not inconsistent with previous studies. For example, religious af liation and participation is signi cantly associated with degrees of alcohol use and abstinence (e.g. Mindanik & Clark, 1994; Heath et al., 1999). Moreover, a recent study by Koopmans et al. (1999) examined the role of religious upbringing, which has no heritable basis (e.g. Boomsma et al., 1999) as a moderator of genetic in uences on the risk of alcohol use initiation. They report some evidence for a gene environment interaction among females: the relative magnitude of the genetic in uences for risk of alcohol use initiation was signi cantly higher in families without a religious upbringing than in families with a religious upbringing, suggesting that religious upbringing moderates the magnitude of genetic in uence on alcohol use initiation. Given the key nding in this and other studies of the role of moral-religious background and education on substance use, it would be interesting to determine if speci c religious af liations or philosophical doctrines have a differentially

10 1316 Kerry L. Jang et al. greater impact on genetically and environmentally based substance use. Unfortunately, this was not testable in the present study because the items comprising the FES Moral-Religious Emphasis subscale were designed to be useable by the widest possible number of subject populations, and as such they do not make reference to speci c religious or philosophical preferences. The present study has several limitations. With regard to the measures of substance misuse, our scales do not assess usage of speci c types of drugs (e.g. marijuana vs. heroin) or alcohol (beer and wine vs. spirits). Furthermore, the substance misuse scales used here only measured substance use in the past 6 8 weeks. Another suggestion for future research would be to try to assess the actual amounts of substance use at different points in a person s life. Information as to what environmental conditions were salient and the level of substance use at any point during development would provide a temporal framework to determine if the relationship between substance use and environmental conditions re ects passive, reactive or active genotype environment correlation (Scarr & MacCartney, 1983). It should also be noted that because the study is based on volunteers from the general population, it is possible that substance users are under-represented in the sample because they are less likely to volunteer. We attempted to test broadly for this problem by comparing the substance use scores from pairs of twins in which both members of a pair completed the study (concordant-participant) against the scores from monozygotic twin pairs in which only one member of a pair completed the study (discordant-participant). No signi cant differences were found between these two groups, suggesting that the present sample is representative of the general population as discussed in the method section. Nevertheless, the present results must be replicated on a larger and broader independent sample. The characteristics of the environmental measures are a second area of concern. First of all, the present study was also limited to a small number of family and classroom environmental measures that limit the generalizability of the results. At the very least, future studies should include other known risk factors for substance use. Another set of limitations that must be considered when interpreting the results concerns how the study participants assessed the environments of their classrooms. Our methods asked participants to report on the best-remembered classroom, but this bestremembered classroom might not be representative of the typical classroom experienced by the person. Nevertheless, the results are clear that classroom task orientation is an important variable whether it was characteristic of all classrooms experienced by the person or one that simply stood out. The issue of selective recall of classrooms does not end with salience of the experience; recall is probably mediated in some way by other factors, such as a genetically based personality trait. This is an issue that clearly requires additional research, and the method described in Jang et al. (000) might be one approach to the problem. As noted in the introduction, this study performed a simultaneous analysis of family environment, personality and alcohol misuse. What was interesting about this study was the nding that that there is not always a clear genetically or environmentally based path from personality and family environment to substance use. Despite the present study s limitations, the general method provides a means to identify risk factors for substance abuse and other psychopathologies. The regression of genetic and environmental factor scores on non-heritable measures of environmental conditions permits the direct assessment of the impact of gene environment correlations the environments conducive to the expression of the genotype that are so often ignored because of the lack of methods to assess them (see Plomin et al., 1990). References BOOMSMA, D. I., DE GEUS, E. J. C., VAN BAAL, G. C. M. & KOOPMANS, J. R. (1999) A religious upbringing reduces the in uence of genetic factors on disinhibition: evidence for interaction between genotype and environment on personality, Twin Research,, BOUCHARD, T. J. JR (1997) The genetics of personality, in: BLUM, K. & NOBLE, E. P. (Eds) Handbook of Psychiatric Genetics, pp (Boca Raton, FL, CRC Press, Inc.). COSTA, P. T. & MCCRAE, R. R. (1985) Manual for the NEO Personality Inventory (Odessa, FL, Psychological Assessment Resources). HARRIS, J. R. (1998) The Nurture Assumption: why children turn out the way they do (New York, NY, The Free Press). HEATH, A. C., MADDEN, P. A. F., GRANT, J. D., MCLAUGHLIN, T. L., TODOROV, A. A. & BUCHOLZ, K. K. (1999) Resiliency factors protecting against

11 Gene environment correlation 1317 teenage alcohol use and smoking: in uences of re ligion, religious involvement and values, and ethnicity in the Missouri Adolescent Female Twin Study, Twin Research,, HETHERINGTON, E. M., REISS, D. & PLOMIN, R. (Eds) (1994) Separate Social Worlds of Siblings: the impact of non-shared environment on development (Hillsdale, NJ, Lawrence Erlbaum Associates, Inc.). JANG, K. L., LIVESLEY, W. J. & VERNON, P. A. (1995) Alcohol and drug problems: a multivariate behavioural genetic analysis of co-morbidity, Addiction, 90, JANG, K. L., VERNON, P. A. & LIVESLEY, W. J. (000) Personality disorder traits, family environment and alcohol misuse: a multivariate behavioural genetic analysis, Addiction, 95, JÖRESKOG, K. G. & SÖRBOM, D. (1993a) PRELIS, a Preprocessor for LISREL (Chicago, IL, Scienti c Software International). JÖRESKOG, K. G. & SÖRBOM, D. (1993b) LISREL 8 (Chicago, IL, Scienti c Software International). KASRIEL, J. & EAVES, L. J. (1976) The zygosity of twins: further evidence on the agreement between diagnosis by blood groups and written questionnaires, Journal of Biosocial Science, 8, KOOPMANS, J. R., SLUTSKE, W. S., VAN BAAL, C. M. & BOOMSMA, D. I. (1999) The in uence of religion on alcohol use initiation: evidence for a genotype X environment interaction, Behavior Genetics, 9, LIVESLEY, W. J. & JACKSON, D. N. (001) Manual for the Dimensional Assessment of Personality Problems Basic Questionnaire (DAPP) (London, ON, Research Psychologists Press), in press. LIVESLEY, W. J., JANG, K. L. & VERNON, P. A. (1998) Phenotypic and genetic structure of traits delineating personality disorder, Archives of General Psychiatry, 55, LOEHLIN, J. C., MCCRAE, R. R., COSTA, P. T. JR & JOHN, O. P. (1998) Heritabilities of common and measure-speci c components of the Big Five personality factors, Journal of Research in Personality, 3, MCGUE, M. & BOUCHARD, T. J. (1984) Adjustment of twin data for the effects of age and sex, Behavior Genetics, 14, MCGUE, M. (1994) Genes, environment, and the etiology of alcoholism, in: The Development of Alchohol Problems: exploring the biopsychosocial matrix of risk, National Institute on Alcohol abuse and Alcoholism Research Monograph no. 6, NIH Publ. no , pp (Bethesda, MD, National Institutes of Health). MCGUE, M., PICKINS, R. W. & SVIKIS, D. S. (199) Sex and age effects on the inheritance of alcohol problems: a twin study, Journal of Abnormal Psychology, 101, MIDANIK, L. T. & CLARK, W. B. (1994) The demographic distribution of U.S. drinking patterns in 1990: description and trends from 1984, American Journal of Public Health, 84, MOOS, R. H. & MOOS, B. S. (1994) Family Environment Scale (Palto Alto, CA, Consulting Psychologists Press). MOOS, R. H. & MOOS, B. S. (1993) Classroom Environment Scale (Palto Alto, CA, Consulting Psychologists Press). NEALE, M., BOKER, S. M., XIE, G. & MAES, H. H. (1999) Mx: Statistical Modelling, 5th edn (Richmond, VA, Department of Psychiatry). NEALE, M. C. & CARDON, L. R. (199) Methodology for Genetic Studies of Twins and Families (London, Kluwer). NICHOLS, R. C. & BILBRO, W. C. JR (1966) The diagnosis of twin zygosity, Acta Geneticae Medicae et Gemellollogiae, 16, PICKENS, R. W., SVIKIS, D. S., MCGUE, M., LYKKEN, D. T., HESTON, L. L. & CLAYTON, P. J. (1991) Heterogeneity in the inheritance of alcoholism: a study of male and female twins, Archives of General Psychiatry, 48, PLOMIN, R., DEFRIES, J. C. & MCCLEARN, G. E. (1990) Behavioral Genetics: a primer, nd edn (New York, NY, W. H. Freeman and Co.). PLOMIN, R., DEFRIES, J. C. & LOEHLIN, J. C. (1977) Genotype environment interaction and correlation in the analysis of human behavior, Psychological Bulletin, 84, PLOMIN, R., LICHTENSTEIN, P., PEDERSEN, N. L. & MCCLEARN, G. E. (1990) Genetic in uence on life events during the last half of the life span, Psychology and Aging, 5, PRESCOTT, C. A., HEWITT, J. K., TRUETT, K. R. & HEATH, A. C. (1994) Genetic and environmental in uences on lifetime alcohol-related problems in a volunteer sample of older twins, Journal of Studies on Alcohol, 55, REISS, D., NEIDERHISER, J. M., HETHERINGTON, E. M. & PLOMIN, R. (000) The Relationship Code: deciphering genetic and social in uences on adolescent development (Cambridge, MA, Harvard University Press). ROWE, D. C. (1994) The Limits of Family In uence: genes, experience, and behavior (New York, Guilford Press). SAUDINO, K. J., PEDERSEN, N. L., LICHTENSTEIN, P. & MCCLEARN, G. E. (1997) Can personality explain genetic in uences on life events? Journal of Personality and Social Psychology, 7, SCARR, S. & MCCARTNEY, K. (1983) How people make their own environments: a theory of genotype environment effects, Child Development, 54, SHAM, P. C., STERNE, A., PURCELL, S., CHERNY, S., WEBSTER, M., RIJSDIJK, F., ASHERSON, P., BALL, D., CRAIG, I., ELEY, T., GOLDBERG, D., GRAY, J., MANN, A., OWEN, M. & PLOMIN, R. (000) GENESiS: creating a composite index of the vulnerability to anxiety and depression in a community-based sample of siblings, Twin Research, 4, STEIGER J. H. (1989) EZPath: a supplementary module for SYSTAT and SYGRAPH (Evanston, IL, SYSTAT, Inc.). STEIN, M. B., WALKER, J. R., HAZEN, A. L. & FORDE, D. R. (1997) Full and partial posttraumatic stress disorder: ndings from a community survey, American Journal of Psychiatry, 154,

12 1318 Kerry L. Jang et al. THOMIS, M. A., VLIETNICK, R. F., MAES, H. H., LIMKIE, C. J., VAN LEEMPUTTE, M., CLAESSENS, A. L., MARCHAL, G. & BEUNEN, G. P. (000) Predictive power of individual genetic and environmental factor scores, Twin Research, 3, TURKHEIMER, E. & WALDRON, M. (000) Non-shared environment: a theoretical, methodological, and quantitative review, Psychological Bulletin, 16, VERNON, P. A., JANG, K. L., HARRIS, J. A. & MCCARTHY, J. M. (1997) Environmental predictors of personality differences: a twin and sibling study, Journal of Personality and Social Psychology, 7,