Cannabis and Other Illicit Drugs: Comorbid Use and Abuse ...

4 downloads 17 Views 90KB Size Report
Cannabis and other illicit drugs are often used or abused comorbidly. Two competing theories to explain this comorbidity are (i) the phenotypic causation ...
481147.qxd

2/18/04

10:58 AM

Page 217

Behavior Genetics, Vol. 34, No. 3, May 2004 (© 2004)

Cannabis and Other Illicit Drugs: Comorbid Use and Abuse/Dependence in Males and Females Arpana Agrawal,1,3 Michael C. Neale,1,2 Carol. A. Prescott,2 and Kenneth S. Kendler1,2 Received 16 April 2003—Final 13 Oct. 2003

Cannabis and other illicit drugs are often used or abused comorbidly. Two competing theories to explain this comorbidity are (i) the phenotypic causation (gateway) model and (ii) the correlated liabilities model. We used data from 1191 male and 934 female same-sex twin pairs to test 13 genetically informative models of comorbidity. Models were fit separately for use and abuse/dependence in both sexes. The correlated liabilities model provided a good fit to the data for cannabis and other illicit drug use, as well as abuse/dependence. The relationship between the use or abuse of cannabis and other illicit drugs is not entirely phenotypic, as depicted by the random multiformity of cannabis model, which is an adaptation of the gateway model. The comorbidity appears to arise from correlated genetic and environmental influences. There is some evidence for a model in which high-risk cannabis users may be at increased risk for other illicit drug use. For abuse/dependence, a model with causal pathways between the liability for cannabis and other illicit drug abuse/dependence also fits well. Overall, our results suggest that the use and abuse/dependence of cannabis and other illicit drugs are strongly linked via common risk factors that jointly influence their individual liabilities. KEY WORDS: Twins; comorbidity; cannabis; illicit drugs; use; abuse/dependence.

INTRODUCTION

1992). These studies suggest that use of cannabis usually precedes initiation of other illicit drugs such as cocaine, hallucinogens, and amphetamines but may follow prior use of nicotine and alcohol. There is also some evidence that the abuse/dependence of cannabis may co-occur with the abuse/dependence of other illicit drugs (Tsuang et al., 1998). Several hypotheses have been proposed to explain the patterns of comorbid substance use. The predominant theories prevalent in the current literature are the gateway theory and the model of correlated liabilities. According to the gateway model, the use of cannabis regularly precedes the use of other illicit psychoactive substances and acts to increase an individual’s risk for using other illicit psychoactive substances (Kandel and Yamaguchi, 1985; Kandel et al., 1992; Yamaguchi and Kandel, 1984). The correlated liabilities model, on the other hand, proposes that cannabis use and other illicit drug use is influenced by correlated genetic and environmental factors (Ellickson et al., 1992; Fergusson and Horwood, 2000; Huba et al., 1981). The

The frequent use of multiple illicit psychoactive substances has been documented by both epidemiological and clinical studies. In particular, the use of illicit psychoactive substances is often seen to co-occur with cannabis use (Adler and Kandel, 1981; Blaze-Temple and Lo, 1992; Brook et al., 1992, 1998; Donovan and Jessor, 1983; Golub and Johnson, 1994; Kandel et al., 1981, 1992, 2001; Merrill et al., 1999; Mills and Noyes, 1984; Voss and Clayton, 1987; Yu and Williford, 1

2

3

Department of Human Genetics, Medical College of Virginia of Virginia Commonwealth University, Virginia Institute for Psychiatric and Behavioral Genetics, Richmond, Virginia. Department of Psychiatry, Medical College of Virginia of Virginia Commonwealth University, Virginia Institute for Psychiatric and Behavioral Genetics, Richmond, Virginia. To whom correspondence should be addressed at Virginia Commonwealth University, Virginia Institute for Psychiatric and Behavioral Genetics, Department of Human Genetics, Box 980003, Suite 1-154, Richmond, Virginia 23298-0003. Tel: 1-804-828-8157; Fax: 1-804-828-1471. e-mail: [email protected]

217 0001-8244/04/0500-0217/0 © 2004 Plenum Publishing Corporation

481147.qxd

2/18/04

10:58 AM

Page 218

218 gateway hypothesis relates to a phenotypic and causal relationship between the two classes of psychoactive drugs; the correlated liabilities model suggests that there is an inherent biological and environmental association between factors that influence the use of either drug. One prior study tested both these models for abuse/dependence in a genetically informative sample and found support for the correlated liabilities model but not for the cannabis gateway model (Tsuang et al., 1998). The phenotypic causation (gateway) model and correlated liabilities model are only 2 of 13 potential models of comorbidity proposed by Klein and Riso (1994). These models are particularly appropriate for examining drug use as well as abuse/dependence because of the substantial evidence for familial/genetic influences on these traits (Cadoret et al., 1995; Karkowski et al., 2000; Kendler and Prescott, 1998; Kendler et al., 1999, 2000; Pickens et al., 1995; Tsuang et al., 2001). Neale and Kendler showed that in a genetically informative design, these models can be clearly articulated and empirically discriminated given adequate power (Neale and Kendler, 1995). The goal of the current analysis is to examine 13 models of comorbidity (including an adaptation of the gateway and correlated liabilities) for cannabis and other illicit drugs. We sought to find the model that explained the comorbid use and abuse/dependence of cannabis and other illicit drugs in males and females.

METHODS

Agrawal, Neale, Prescott, and Kendler subjects were briefed about the aims of the study and informed consent was obtained. Sample Characteristics The sample consists of 556 MZ and 378 DZ female same-sex twin pairs and 702 MZ and 489 DZ male same-sex twin pairs. Mean age of the female twins was 35.8 years (range 21–62 years) at the fourth wave of interviews, with a mean education level of 14.3 years. The male twin pairs had a mean age of 35.5 years (range 20–58 years) and a mean education level of 13.6 years at the second wave of interviews. 47.6% of the female twins used cannabis, and 42.0% used other illicit drugs (cocaine, 14%; sedatives, 8%; stimulants, 10%; hallucinogens, 11%; and opiates, 5%). Of the female twins using cannabis, 15.8% were diagnosed with abuse/dependence. Of the female other illicit drug users, 21.4% were diagnosed with abuse/dependence of other illicit drugs (cocaine, 23%; sedatives, 22%; stimulants, 32%; hallucinogens, 9%; and opiates, 9%). In the males, 54% of the twins used cannabis and 33.3% of these users were diagnosed with abuse/dependence. For other illicit drugs, 62% of the male twins were users (cocaine, 17%; sedatives, 11%; stimulants, 20%; hallucinogens, 15%; and opiates, 8%) and 30.9% (cocaine, 30%; sedatives, 27%; stimulants, 40%; hallucinogens, 23%; and opiates 11%) of the users were diagnosed with abuse/dependence. Within-person correlations for use of cannabis and other illicit drug use was 0.78 in both males and females, and for abuse/ dependence, this correlation was 0.78 in males and 0.79 in females.

Sample This study utilized drug use and abuse/dependence information from 1191 male and 934 female same-sex monozygotic (MZ) and dizygotic (DZ) Caucasian twin pairs from the Virginia Twin Registry (now a part of the Mid-Atlantic Twin Registry). Respondents who agreed to participate were interviewed in the first wave of personal interviews. Zygosity was initially determined by standard questions, and zygosity of a subsample of twins was later confirmed by polymerase chain reaction analyses. Three follow-up telephone interviews have been completed in the females, and one follow-up has been completed for the males. The data for the present analyses come from the fourth and second wave of interviews for the females and males, respectively. Ascertainment details for data collection are available elsewhere (Kendler and Prescott, 1999). As approved by the institutional review board of Virginia Commonwealth University, before the interviews,

Measures Use, abuse, and dependence were assessed using an adaptation of the Structured Clinical Interview for DSM-IIIR-Patient Version (Spitzer et al., 1987). Use and abuse/dependence were coded as binary variables. Cannabis use was defined as lifetime use of cannabis (e.g., hashish and marijuana). We constructed the other illicit drug category by combining the responses to the categories of cocaine, sedatives, stimulants, hallucinogens, and opiates. Participants responding positively to any of these categories were scored positively for other illicit drug use. For the categories of drugs that may be obtained legally (sedatives, stimulants, and opiates), use was defined by any of the following criteria: (i) use without a prescription, (ii) in greater amounts than prescribed, (iii) more frequently than prescribed, or (iv) for reasons other than those for which it was prescribed.

481147.qxd

2/18/04

10:58 AM

Page 219

Cannabis and Other Illicit Drugs: Comorbid Use and Abuse/Dependence in Males and Females If the participant was diagnosed with abuse or dependence for cocaine, sedatives, stimulants, hallucinogens, or opiates, they were diagnosed with abuse/dependence for the other illicit drug category. Additional details are available elsewhere (Kendler et al., 1992). Twin Modeling Thirteen models of comorbidity proposed by Neale and Kendler (1995) were fit separately to use and abuse/dependence data in males and females. These include the chance model, which assumes that

219

the co-occurrence of two disorders simultaneously in an individual may occur by chance alone; alternate forms model; random multiformity (with two submodels) models; extreme multiformity (with two submodels) models; three independent disorders models; the correlated liabilities model; and causal models three models). A brief description of each model is presented in Table I. The chance model tests whether the comorbid use (or abuse/dependence) of cannabis and other illicit drugs occurs because of random factors, whereas the three independent disorders model tests if cannabis use, other illicit drug use, and their comorbid use are independent conditions, each with a unique

Table I. Description of Neale and Kendler Models of Comorbidity with Cannabis and Other Illicit Drug Use as Example Number

Model name

1

Chance

2

Alternate Forms Multiformity

3

Random Multiformity

4

Random Multiformity of Cannabis Random Multiformity of Other Illicit Drugs Extreme Multiformity

5 6

7 8

9 10

Extreme Multiformity of Cannabis Extreme Multiformity of Other Illicit Drugs Three Independent Disorders Correlated liabilities Causality

11

Reciprocal Causation

12

Cannabis Causes Other Illicit Drugs Other Illicit Drug Causes Cannabis

13

Description Comorbidity is due to randomness alone—cannabis and other illicit drug use co-occur by chance. Crossing a single, common threshold leads to use of cannabis or other illicit drugs. The liability for cannabis use and other illicit drug use are unrelated. Use of one drug serves to increase the risk for use of the other drug. Multiformity may be random, with a single threshold, or extreme, with a second threshold accounting for users at increased risk. Each disorder has a distinct liability threshold. Being above one threshold (e.g., cannabis use) increases risk for presenting other disorder (e.g., other illicit drug use). This model allows for use of either drug category to increase the risk of other drug category. Being above the threshold for cannabis use (user of cannabis) leads to an increased risk for other illicit drugs. Also known as gateway or the phenotypic causation model. Being above the threshold for use of other illicit drugs leads to an increased risk for cannabis use. Each disorder has two distinct thresholds. The second threshold accounts for heavy use. Individuals will be at increased risk for use of other illicit drugs if they are at increased risk for cannabis use. Being above the first threshold for cannabis only leads to cannabis use. Individuals are at increased risk for other illicit drug if above second threshold for cannabis. Being above the first threshold for other illicit drug use only leads to other illicit drug use or abuse/dependence. Individuals are at increased risk for cannabis use or abuse/dependence if above second threshold for other illicit drugs. Cannabis use, other illicit drug use, and comorbid use of cannabis and other illicit drugs are independent, unrelated disorders. Cannabis use and other illicit drug use are influenced by genetic and environmental factors that are correlated across the drugs. Refers to causal pathways that influence the liability to use one drug when there is a preexisting liability to use the other drug. Unlike multiformity models, these are not causal at a phenotupic level. There are causal pathways between cannabis use and use of other illicit drugs. Liability to use of one drug has a causal influence on the liability to use the other. Causal pathway from the liability to use cannabis influences the liability to use of other illicit drugs. Causal pathway from the liability to use other illicit drugs has an effect on the liability to use cannabis.

Note: Additional details and diagrammatic representations available elsewhere (Neale and Kendler, 1995).

481147.qxd

2/18/04

10:58 AM

Page 220

220 liability. An alternate forms model assumes a single liability with a single threshold. In contrast, the multiformity models, independent disorders model, correlated liabilities models, and causal models ascribe individual, independent liabilities to cannabis use (or abuse/dependence) and other illicit drug use (or abuse/dependence). Each liability distribution has a unique threshold, and an individual above the threshold would use the drug that the distribution describes. Multiformity suggests that being above a certain threshold for one trait leads to an increased risk for the comorbid trait. If we allow for cannabis use alone to depict this multiformity, the model captures the essential features of the gateway hypothesis of drug use whereby users of cannabis (above the liability threshold for cannabis use) are at an increased risk for use of other illicit drugs only because of their cannabis use. According to the gateway hypothesis, cannabis use is regularly initiated before use of other illicit drugs and users of cannabis are at an increased risk to use other illicit drugs. The random multiformity of cannabis model allows us to model the increased risk of other illicit drug use imposed on a proportion of the cannabis users. The model also allows for some heterogeneity whereby a proportion of users of other illicit drugs do so because they are above the threshold for the risk of other illicit drug use. The remaining proportion of users are below the threshold for their liability to use other illicit drugs but are at increased risk for other illicit drug use because of their prior cannabis use. We propose this as an efficient adaptation of the gateway model. The gateway model has evolved over the years to accommodate varying etiological pathways leading to the use of illicit drugs such that the gateway mechanism may operate in a certain proportion of the population alone (Kandel, 2002; Morral et al., 2002a). The extreme multiformity model is an adaptation of the random multiformity approach that accounts for severity of risk by imposing two thresholds on the liability. Individuals above the first threshold are at risk for use of cannabis and are at no risk for using other illicit drugs. However, an individual crossing the second threshold is at increased risk for using cannabis and also develop a risk for the use of other illicit drugs. Whereas the random multiformity model captures a phenotypic relationship in which use (or abuse/dependence) of cannabis directly increases the risk of use (or abuse/dependence) of other illicit drugs, the causal models suggest a causal mechanism that influences the underlying liability for use or abuse/ dependence of each drug. In contrast, the correlated

Agrawal, Neale, Prescott, and Kendler liabilities model proposes that cannabis and other illicit drugs have individual and independent liabilities that are influenced by genetic and environmental (shared and unique) factors that are correlated across the two drug categories. All models were fit using the structural equation modeling program Mx (Neale, 1990). Models were fit to observed frequencies of the cells of a contingency table of MZ and DZ twins. Contingency tables were constructed separately for use and abuse/dependence. For example, for drug use, each member of the twin pair was categorized in the following way: neither cannabis nor other illicit drug use, cannabis use but no use of other illicit drugs, use of other illicit drugs but no cannabis use, or both cannabis use and use of other illicit drugs. Twins were cross-tabulated with their cotwin, and frequencies of twin pairs falling into each category were used to construct frequency tables for the analyses. This expected pattern of frequencies was compared with the frequencies observed under the model being tested using a minimum chi-square test. Like traditional modeling of twin data, the comorbidity models allow the estimation of additive genetic influences (A), shared environmental influences (C), and unique environmental influences (E) for cannabis and other illicit drug use and abuse/dependence. Model selection was based on the Akaike’s Information Criterion (AIC) which is an index of parsimony and goodness of fit. (Akaike H, 1987). The AIC is calculated by subtracting twice the degrees of freedom of a model from the chi-square value and the lowest AIC (largest negative value) (Williams & Holahan, 2002) indicates the best fit. The major classes of comorbidity models are not nested. This means that the phenotypic models cannot be obtained by adjustments to a liability model. This can be contrasted with the types of multiformity models that are closely related in themselves. AIC allows a comparison of the relative fit of these non-nested models. Finally, parameter estimates from the males and females were equated (without equating thresholds) to assess for sex differences.

RESULTS Twin Analyses Table IIA and IIB present the chi-square (2) fit statistic, degrees of freedom, p value, and AIC fit statistic for comorbid use and abuse/dependence of cannabis and other illicit drugs, respectively Table IIIA and IIIB provides parameter estimates for the bestfitting model and second best fit in each case.

481147.qxd

2/18/04

10:58 AM

Page 221

Cannabis and Other Illicit Drugs: Comorbid Use and Abuse/Dependence in Males and Females

221

Table IIA. Fit Indices:  2 , Degrees of Freedom (df) and p Value for Model-Fitting in Both Sexes for Use of Cannabis and Other Illicit Drugs (Akaike’s Information Criterion [ 2 − 2df = AIC] is also presented) Male Model 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

Chance Alternate Forms Random Multiformity Random Multiformity of Cannabis (Gateway) Random Multiformity of Other Illicit Drugs Extreme Multiformity Extreme Multiformity of Cannabis Random Multiformity of Other Illicit Drugs Three Independent Disorders Correlated Liabilities Reciprocal Causation Liability to Cannabis Affects Other Illicit Drug Liability to Other Illicit Drug Affects Cannabis

df

2

10 13 08 09

582.4 178.9 89.4 150.6

09

Female p value

2

AIC

p value

562.4 152.9 73.4 132.6

Suggest Documents