The familial aggregation of cannabis use disorders

Abstract

Aims

The aim of this paper is to examine the familial aggregation of cannabis use disorders and other psychiatric conditions among first-degree relatives and spouses of probands with a cannabis use disorder.

Design

Controlled family study methods.

Setting

Out-patient psychiatric clinics and the local community (same geographic area).

Participants

Two hundred and sixty-two probands with a life-time history of cannabis use disorder, alcohol dependence, anxiety disorders or no history of any disorder, and their first-degree relatives and spouses.

Measurements

Cannabis use disorders and other DSM-III-R disorders in the relatives and spouses using the Schedule for Affective Disorders and Schizophrenia.

Findings

Results reveal an elevated risk of life-time history of cannabis use disorders among siblings [odds ratio (OR: 3.6), adult offspring (OR): 6.9], and spouses (OR: 4.4) of probands with cannabis use disorders. There is a latent familial factor underlying cannabis use disorders that was shared partially with alcohol abuse/dependence. Comorbid mood and anxiety disorders aggregated independently from cannabis use disorders in families. Equal elevation in the magnitude of the association among the first-degree adult relatives and spouses of probands with a cannabis use disorder suggests the probable contribution of both environmental and genetic factors.

Conclusions

These findings support a family-based approach to drug abuse intervention and the importance of future research concerning environmental mediators of familial transmission of drug abuse.

INTRODUCTION

Cannabis is the most widely available illicit psychoactive drug in the United States [1]. Nearly 40% of 12th graders in the United States reported marijuana use during the past year [2] and 6% of the general population over age 12 used marijuana or hashish over the previous month [3]. The life-time prevalence rate of cannabis abuse or dependence has been estimated recently at 8.5%, and both conditions are associated strongly with other psychiatric disorders [4,5]. The risk factors for cannabis use, abuse and dependence are similar to other substances, and include characteristics such as age of exposure, early reaction to cannabis, economic factors, substance availability, religious involvement, delinquent behaviors, peer relationships and attitudes towards the future [5–12]. The risk factors for use differ in some respects from those that may lead to heavy use, abuse or dependence [11,13–18], but the cumulative number of risk factors an individual possesses is a significant predictor of substance-related problems [10–12,14,17,19].

Controlled family studies [20], involving the identification of individuals with and without a particular disorder followed by blinded assessment of disorder rates in their relatives, have confirmed many of the findings from clinical or epidemiological investigations as well as demonstrated the extent to which substance use disorders are familial [16,21–26]. While some specificity of familial aggregation is found for particular drugs [8,27–29], parental substance abuse is associated with a wide range of behavior problems and drug use among exposed children. Other pathways of familial aggregation include sibling substance abuse [27,29] as well as assortative mating, where a high degree of concordance is observed for substance use disorders between spouses [23,30–42]. Concerning the underlying causes of familial aggregation, twin studies have shown that genetic factors explain a portion of the variance in sibling concordance for cannabis abuse/dependence [43–48]. Adoption studies have confirmed the heritability of these conditions, but also the importance of interactions between genetic and environmental factors in the development of a substance use disorders [16,28,43–46,49–51]. With few exceptions [43], genetic factors have been found to play a more potent role in the etiology of severe patterns of drug use, particularly abuse or dependence, than early stages of use, which appear to be determined more strongly by environmental influences [28,44,46].

In a previous controlled family study of several predominant drugs of abuse, we found an eightfold increased risk of drug use disorders among relatives of affected probands compared to those of psychiatric and unaffected controls [8]. The current study was designed to examine more specifically the patterns of cannabis use disorders in the first-degree relatives and spouses of probands with a life-time history of cannabis use disorder compared to relatives of probands with alcohol dependence, anxiety disorders and controls without these disorders. We address the following questions: (i) what is the magnitude of familial aggregation of cannabis use disorders in relatives of probands with these disorders, compared to relatives of probands with alcohol dependence, anxiety disorders or no disorder; (ii) is there concordance between spouses for cannabis use disorders; and (iii) are there shared familial factors underlying cannabis use disorders, alcohol dependence and other mental disorders?

METHODS

Sample characteristics

Probands

Probands in the present analyses are a subset of participants in a large family study examining comorbidity of substance use disorders and anxiety disorders [52]. In order to investigate the specificity of familial aggregation of cannabis use disorders, the present analyses are restricted to 36 probands with predominant cannabis use disorder, 89 probands with alcohol dependence (47 with and 42 without anxiety disorders), 76 probands with anxiety disorder only, 61 controls without a life-time history of disorder and 215 first-degree relatives and spouses of the different proband groups. Probands with opioid or cocaine dependence, recruited as part of the larger study, were excluded from the present analyses. All probands were white, because we have examined familial aggregation of substance abuse in Hispanics and African Americans in separate studies. Table 1 reports the socio-demographic characteristics of the proband groups studied. The probands were 40 years old on average, with an equal sex distribution.

Table 1.

Characteristics of proband groups.

Characteristic	Cannabis use disorder (n = 36)	Alcohol dependence (n = 89)	Anxiety disorders alone (n = 76)	Controls (n = 61)
Age (years) (mean)	35.7	40.4	39.9	40.8
Male (%)	58.3	70.7	26.3	42.6
Socio-economic status (mean)	46.8	43.8	34.0	30.8
Percentage employed	86.1	82.0	80.3	88.5
Household income (/$1000) (%)
<29	33.3	42.6	23.7	6.7
30–59	44.4	32.6	50.0	58.3
60+	22.2	23.5	26.3	35.0
Religion (%)
Catholic	52.8	57.3	58.7	57.4
Protestant	11.1	19.1	20.0	23.0
Jewish	2.8	1.1	4.0	9.8
Other	8.3	6.7	9.3	4.9
Not affiliated	25.0	15.7	8.0	4.9
Education (%)
Partial university or higher	38.9	47.7	63.1	78.7
High school	38.9	32.6	34.2	19.7
Partial high school or lower	22.2	19.1	2.6	1.6

Probands were recruited through several out-patient clinics for substance and anxiety disorders, and through community random digit dialing in the greater New Haven, Connecticut area. After providing a complete description of the study to the subjects, written informed consent was obtained. Probands were assigned to lifetime diagnostic groups based on an algorithm designed to reflect predominant level of psychopathology: DSM-III-R diagnoses of cannabis use disorder, alcohol dependence and an anxiety disorder; alcohol dependence without an anxiety disorder; anxiety disorders only (social phobia or panic disorder with or without agoraphobia); and normal controls with no life-time history of a DSM-III-R Axis I disorder. Based upon review by clinicians with expertise in substance abuse, probands were excluded from the study if there was evidence of significant organic mental impairment, schizoaffective disorder or schizophrenia. The normal controls were recruited using a random digit dialing procedure in which they were drawn from the same general population as those of the affected probands. The control group was subjected to the same protocol as the treatment groups.

Relatives

Inclusion criteria required that permission be obtained first from probands to contact living relatives, of whom 123 (57%) were interviewed directly, either in person or by telephone. The interview rates did not differ significantly between proband diagnostic groups, and all analyses controlled for the effect of interview status. Interviewed family members included all first-degree adult relatives (parents, children, siblings), as well as past and/or current spouses (individuals married to the proband or cohabitating for at least 6 years).

Overview of procedures

Diagnostic interview

Adult psychiatric disorders were established using the semi-structured Schedule for Affective Disorders and Schizophrenia (SADS), current and life-time versions [53], which was modified to include more detailed information on substance use including patterns and interrelationships of drug use. The SADS differs from other diagnostic instruments in that it uses a semi-structured format and is administered by experienced clinical interviewers, as opposed to the highly structured formats of the Composite International Diagnostic Interview (CIDI) or Diagnostic Interview Schedule (DIS) used by lay interviewers. Interviewers with experience in clinical psychiatry and/or substance abuse conducted the diagnostic interviews to establish diagnoses using DSM-III-R criteria. Reliability was established at the study initiation and was tested periodically throughout the study. Kappas derived from joint ratings of individual interviews were generally higher for substance abuse/dependence (0.72–0.94) than for anxiety or affective disorders (0.54–0.78) across the first three series of training sessions. Comparison of diagnoses obtained through direct face-to-face interview versus telephone interview showed high levels of agreement across all diagnostic categories [54].

Family history information

Family history information was obtained using a modified version of the Family History–Research Diagnostic Criteria [55] to obtain DSM-III-R diagnoses in adults and children. The modified version is more similar to instruments such as the Family Interview for Genetic Studies and enabled the collection of more extensive information on patterns and sequelae of drug and alcohol use as well as the extent of the informant’s knowledge on the index subject.

Diagnostic procedures

Diagnoses of mental and substance use disorders were based upon all available information, including the diagnostic interview, family history reports and medical records, using the best estimate method of diagnoses. Each drug of abuse was characterized by: age at onset, quantity, frequency, chronicity, substance of choice, number of symptoms and severity. Best-estimate diagnoses were made by clinicians with extensive experience in the evaluation and treatment of substance abuse, and who were made blind with respect to the diagnostic status of the probands when making best estimates of the relatives. Interview status was included as a covariate in the analyses because of the well-established underestimation of diagnoses in non-interviewed relatives. Direct interviews and family history are far more concordant for observable disorders such as drug abuse and behavior disorders than for less readily observable disorders such as mood and anxiety disorders.

Statistical analysis

Logistic regression analyses were used to estimate odds ratios (OR) and 95% confidence intervals (CI) for the differences in the proportion of affected versus unaffected first-degree relatives and spouses, controlling for proband sex, interview status (direct versus telephone interview), sample source (i.e. clinic versus community), relative sex and relative age. The association between DSM-III-R cannabis use disorders and the number of affected parents was assessed using the Armitage test for trend in proportions [56]. The association between the age-at-onset of cannabis use disorders between pairs of relatives was assessed using the statistic developed by Clayton & Oakes [57,58]. The Clayton–Oakes concordance measures the broad ‘correlation’ of two ages of onset through the so-called cross-ratio. The cross-ratio can be viewed as the OR of the two events occurring within a short time-interval under the assumption that this ratio is time-constant. In addition, we assessed the family aggregation, covariate effects and potential interaction between familial contribution and covariates using a frailty regression model [59,60]. In essence, a latent variable is introduced for each family of the study and is used to summarize the familial association. Conditional on this latent variable, cannabis use disorders are assumed to follow an ordinary logistic regression model. Thus, the marginal distribution of cannabis use disorders is described by a mixture logistic regression model. The presence and significance of the latent variable in the mixture logistic regression model is indicative of familial aggregation. The main advantage of this regression model is its flexibility to adjust for covariate effects or to examine the interactions between covariates and an underlying familial effect. We refer to Zhang & Merikangas [60] for a detailed description of the model.

RESULTS

Table 2 presents the concordance of frequency tables of cannabis abuse or dependence between probands and their first-degree relatives and spouses and adjusted ORs. The siblings of the probands who had cannabis abuse or dependence had a greater risk (OR 3.6, 95% CI: 2.0–6.5) of cannabis abuse or dependence than those of probands without a cannabis use disorder. Similarly, adult offspring of probands with a cannabis use disorder had a greater risk of developing a cannabis use disorder (OR 6.9, 95% CI: 1.4–33.5) compared to those relatives of probands without a cannabis use disorder. Probands with a cannabis use disorder were more than four times (OR 4.4, 95% CI: 1.9–10.1) more likely to have spouses with a cannabis use disorder than those without a history of a cannabis use disorder. The associations between the ages-at-onset of probands and siblings, and probands and spouses were of similar magnitude (4.07 for siblings and 4.12 for spouses), both of which were highly significant (P-values less than 0.0001).

Table 2.

Association of cannabis use disorders in probands, relatives and spouses.

	Siblings > 18	Offspring > 18	Adult relatives	Spouses
Cannabis use disorder in proband	n (% with cannabis use disorder)
Yes	28 (25.7)	4 (50.0)	32 (27.4)	15 (34.9)
No	42 (8.2)	15 (13.0)	57 (9.1)	22 (8.8)
Odds ratio (95% confidence intervals)a	3.6 (2.0–6.5)	6.9 (1.4–33.5)	4.5 (2.6–7.6)	4.4 (1.9–10.1)

^aAdjusted for sex and age of proband and relative.

Table 3 presents the results of the multivariate regression analysis using the frailty model that assesses the risks of cannabis use disorder among probands and relatives (latent familial factor). The ORs and 95% CIs from the multivariate model are reported for the effects of proband cannabis use disorder, sex and comorbid disorders on cannabis use disorders in relatives adjusting for the effects of age, sex, comorbid disorders and interview status of relatives. Effects and significance are reported for each of those risk factors.

Table 3.

Factors associated with cannabis use disorders in probands and relatives.

Variable	Estimate	SE	Odds ratio	95% CI	P-value
Cannabis (latent familial factor)	2.131	0.989	8.4	1.2–58.5	*
Probands
Sex (male versus female)	0.654	0.472	1.9	0.8–4.9	NS
Comorbidity (yes versus no)
Anxiety	−0.300	0.350	0.7	0.4–1.5	NS
Major depression	−0.059	0.329	0.9	0.5–1.7	NS
Alcohol dependence	0.948	0.372	2.6	1.2–5.4	**
Antisocial personality disorder	−0.813	0.599	0.4	0.1–1.4	NS
Relatives
Sex (male versus female)	−0.6	0.439	0.5	0.2–1.3	NS
Age (increase/year)	0.883	0.368	2.4	1.2–5.0	*
Comorbidity (yes versus no)
Anxiety	0.400	0.339	1.5	0.8–2.9	NS
Major depression	0.447	0.384	1.6	0.7–3.3	NS
Alcohol dependence	1.460	0.385	4.3	2.0–9.2	***
Antisocial personality disorder	1.308	0.535	3.7	1.3–10.6	**
Interview status (yes versus no)	−0.125	0.021	0.9	0.8–0.9	***
Sex of proband × sex of relative	−0.429	0.655	0.7	0.2–2.4	NS

Significance level: NS = not significant at 0.05;

^*P < 0.05;

^**P < 0.01;

^***P < 0.001. CI: confidence interval; SE: standard error.

There was a strong association between alcohol dependence, antisocial personality and cannabis use disorders in relatives, whereas there was no significant association between anxiety disorders or major depression and cannabis use disorders. Similarly, comorbid anxiety, depression and antisocial personality disorder in probands with a cannabis use disorder were not associated with cannabis use disorders in relatives. In contrast, the association between alcohol dependence in the probands and a cannabis use disorder in the relatives demonstrates that there may be common underlying risk factors for these two disorders. Finally, neither the sex of the proband nor the sex of the relative nor their interaction modified the familial association between substance use disorders in probands and relatives.

DISCUSSION

The primary aim of this study was to investigate the familial aggregation of cannabis use disorders among spouses and first-degree relatives of affected probands. The results show a significant degree of familial aggregation of cannabis use disorders among both siblings and adult offspring of probands with a cannabis use disorder. The increased rate of these disorders among siblings of affected probands supports previous research by Bierut et al. [27] and others [54,61], who showed high levels of specificity of familial transmission of cannabis use disorders within sibling pairs relative to controls, even after restricting the data to include only those pairs who have tried these illicit drugs. Similar to previous high-risk studies of substance abuse in the children of parents with substance use disorders [7,9,21,23,31,32,62], we found a significant elevation in life-time rates of cannabis abuse/dependence among adult offspring. This familial concordance may result from several factors including behavior modeling, shared environment, accessibility and genetic factors.

We also found a high concordance rate for cannabis use disorders between probands and their spouses: 47% of the male spouses and 25% of the female spouses of probands with a cannabis use disorder also had a cannabis use disorder. From an examination of patterns of spouse concordance before and after marriage to the probands, it could be observed that the age of onset of cannabis occurred prior to the marriage in 89% of the couples concordant for cannabis use disorders. While this may suggest that primary or secondary (social homogamy) assortative mating is implicated in spousal concordance, the average age of onset for cannabis use and the average age of marriage render it plausible that cohabitation may also influence cannabis use and potentially lead to more regular use, abuse and dependence. Moreover, the nearly identical concordance rates among siblings, who are biological relatives (OR 4.5), and spouses who are not biologically related (OR 4.4), highlights the probability that both environmental and genetic influences are implicated in the familial transmission of cannabis use disorders. That is, while the choice of a spouse who is concordant for cannabis may be understood as an environmental factor, it may increase both environmental determinants (e.g. modeling, availability of cannabis) as well as genetic similarities between spouses that in turn increase aggregation of cannabis use disorders within families. Irrespective of the mechanisms for spouse concordance, our findings suggest that future studies of the genetic transmission of drug use disorders incorporate assortative mating as it may distort segregation ratios and affect the distribution of affected and unaffected families in the population.

Our finding that familial concordance for cannabis abuse/dependence persisted after controlling for comorbid disorders in probands suggests that affective and other behavior disorders aggregate independently from these disorders. Similar conclusions were drawn from previous uncontrolled family studies [21,25,63], indicating that anxiety, affective and antisocial personality disorder could either be pre-morbid risk factors or consequences of cannabis use disorders. In contrast, our finding of cross-concordance between cannabis use disorders and alcohol dependence in probands and relatives suggests that shared familial factors underlie both these forms of substance dependence. This finding confirms previous observations from family and twin studies of alcohol-dependent probands [48,64–66], but extends the findings to cross-generational transmission. The lack of cross-generational transmission of cannabis abuse or dependence with other disorders such as anxiety suggests that their association at an individual level may be best explained by behavioral or psychological mechanisms [67].

To our knowledge, this is the first controlled family study that tests the inter-generational transmission of cannabis use disorders, independent of other substances. This information supplements the current knowledge derived largely from twin studies as well as from previous uncontrolled family studies of substance abuse and controlled family studies of cannabis use disorders in alcoholic probands. There are several unique design aspects of this study that enhance its contribution to the knowledge of cannabis use disorder etiology, including: selection of probands recruited from both community and treatment settings; direct interviews with relatives blind to probands’ disorders to ensure more accurate classification; inclusion of probands with alcohol dependence without a cannabis use disorder as a substance-disordered comparison group to enable evaluation of specificity; inclusion of a control group in which similar diagnostic and assessment methods were used; and incorporation of psychiatric comorbidity among probands and relatives in the design and analyses.

Concerning limitations of this study, the focus on cannabis abuse and dependence diagnoses may overlook a substantial percentage of individuals who experience cannabis-related problems and who are at risk for meeting diagnostic criteria for abuse or dependence in the future [68–70]. An additional consideration is that different generations may be more or less likely to use cannabis depending on changing societal views on this substance. In this regard, the prevalence of 12-month cannabis use in the United States has remained stable between 1990 and 2000 [71], with a modest increase in the prevalence of cannabis use disorders over this period (due possibly to increasing potency of the substance). Considerable variation has, none the less, been seen over time concerning the legal penalties for cannabis use since its placement under the jurisdiction of the Department of Justice following the Controlled Substances Act of 1970. These societal contexts, as well as changes in cannabis availability, may explain partially differences in rates observed between age cohorts, in particular for the parents of probands. Other limitations include the restriction of the sample to white residents of Connecticut, the lack of direct interviews for all relatives and statistical power concerns in light of a potentially large number of latent variables coupled with relatively small sample size. A final important consideration is the generalization of findings in light of the ‘two worlds’ of clinical and general population samples [72]. Although the sampling strategy recruited probands intentionally from both treatment and community settings to reduce biases that may affect familial aggregation [73], the comparability of these sample sources remains a subject of debate.

In conclusion, these findings add to the body of research that provides evidence not only for common familial factors underlying substance abuse in general, but also highlights the need to examine risk factors that may be specific to particular substances. Future research is crucial for elucidating the roles of genetic and environmental factors in the transmission of cannabis use disorders and their apparent familial association with alcoholism, as well as for identifying sources of heterogeneity in the etiology of cannabis abuse, particularly with respect to the role of comorbid psychiatric disorders and polysubstance abuse. Finally, the high spouse concordance for cannabis use disorders suggests that offspring of these probands are a particularly elevated risk for the development of these conditions themselves and should be targeted in future prevention and intervention studies.