Prevalence findings from a single location may not generalize to other regions

The levels of substance use in our results, similar to those found in a primary care sample with depression and an emergency services sample , indicate that a substantial number of patients were at risk. The results suggest that providers in psychiatric settings should conduct screening and offer treatment as needed. For example, brief motivational interventions could effectively supplement other psychiatric services and prevent escalation of alcohol and drug problems. The recent Screening, Brief Intervention, Referral, and Treatment initiative launched by the Substance Abuse and Mental Health Services Administration is actively promoting early intervention with non-dependent patients in primary care, mental health care, and other settings; online resources and multiple training opportunities are available for providers. Effective January 2008, codes approved by the U.S. Centers for Medicare and Medicaid Services allow reimbursement for screening and brief intervention. These policy initiatives recognize that early alcohol and drug use identification and treatment are important medical services that can improve multiple health outcomes. As a mechanism to facilitate identification, we found that the computerized system was acceptable to most patients, consistent with a prior study of computerized depression assessment . Alcohol and drug use questions also were acceptable , although we note that patients with serious alcohol and drug problems had previously been screened out. One limitation was that older adults appeared less willing or able to use the computer,vertical grow system but this may have been due in part to disabilities of patients seeking geropsychiatric services or to assumptions of reception staff regarding disabilities. Service issues for further investigation include staff training, optimal procedures to ensure confidentiality , software options for alcohol and drug screening, and computer-based interventions .

Potential benefits of computerized systems include the ability to collect more detailed information than is easily obtained by paper forms, greater validity for sensitive questions, and more time-efficient assessment. The records obtained from patients are useful for treatment planning at a program level and as a resource for clinical and services research. These findings need to be interpreted with the limitations of the study. Because patients completing the electronic intake were younger on average than those who did not complete it, our substance use findings are less representative of older adult patients.Our results may not generalize to psychiatry clinics that do not prescreen patients with serious alcohol and drug problems before intake. It would have been preferable to use a lower heavy-drinking cutoff for women than men because women are more sensitive to alcohol. Our use of the higher cutoff indicates that our findings regarding heavy drinking by women may be conservative. Although computerized measures are considered valid, under reporting of alcohol and drug use by patients would also make our prevalence rates conservative. Thus substance use in the sample may be even greater than our results indicate.An estimated 15.3 million adults in the United States met criteria for an alcohol use disorder in the past 12 months. Of those with alcohol use disorders, 2.3 million adults also met criteria for a drug use disorder with odds ratios estimated to be 7.4 for any drug use disorder, but 3.4 to 19.2 for specific drug use disorders . Both alcohol and drug use disorders are heritable, with approximately 50% of the variance attributable to heritable factors , although this estimate varies dramatically by substance , age and other characteristics, including co-morbid psychopathology . This heritable variation can be parsed into those genetic influences that are specific to each drug and importantly, those genetic factors that confer a general predisposition to alcohol and/or substance use disorders, and even other dis-inhibited behaviors . Two large twin studies have convincingly shown that a preponderance of the genetic factors influencing illicit drug use disorders overlap . Noticeably, when these models were extended to include alcohol use disorders, there was evidence for highly correlated genetic factors that individually influenced the covariation in alcohol and nicotine dependence as well as cannabis and cocaine dependence . The extent of genetic overlap was strong for some substances—for instance, 55% and 24% of the genetic variance in alcohol dependence was due to the licit and illicit drug factors, respectively, with the remainder being substance specific. In contrast, for nicotine dependence, 63% of the genetic variance was drug specific . Similar to the individual heritability of each substance, there is growing evidence that the heritable covariation across substances changes across development .

Irrespective of development and substance-specific variation, there is broad consensus that gene discovery efforts targeting aggregate genetic variation that indexes a shared liability to a variety of substance use disorders, as well as dis-inhibition, can be c , with one study showing evidence for genome wide pleiotropic effects across substance use disorders . There are multiple approaches, both phenotypic and genetic, to capture the commonality underlying alcohol and substance use disorders and the present study utilizes two straightforward phenotypic approaches. We opted for simple dependence-based phenotypic traits as they lend themselves to replication and future meta-analysis. First, we utilized a binary phenotype, with affection status defined as meeting dependence criteria for at least one substance , termed ANYDEP. Second, we used factor analysis to combine dependence criteria across substances into a continuous quantitative trait representing vulnerability to multiple substance dependence, termed QUANTDEP. This quantitative measure is heritable and has previously been used in genomic studies , the most recent of which utilized a similar expanded factorial measure of behavioral dis-inhibition to conduct genome-wide association and rare nonsynonymous variant analyses . These studies did not identify any single common or rare variant at a genome-wide significant level; however, the authors reported that 84% of the heritability in illicit drug use was explained by both common and rare variants. While the work of McGue and colleagues included multiple measures of nicotine use and dependence, we elected to exclude nicotine from these measures of general liability based on the work by Kendler and colleagues , which showed significant drug-specific genetic influences on nicotine dependence. In this study, we utilized data from 2322 subjects from 118 families of European-American descent ascertained for alcohol dependence liability to conduct genome-wide association analysis of a binary and a continuous index of general substance dependence liability. While some prior genome-wide efforts have utilized similar phenotypes in population samples of related individuals, the ascertainment strategy and extended family-based design in our study should increase our ability to detect genetic variation in this phenotype. First, there is substantial evidence that alcohol use disorders that co-aggregate with other substance use disorders may represent a more heritable form of addiction . Secondly, by modeling the strength of the phenotypic correlation across different degrees of genetic relatedness , we utilize data on all related individuals, even those not meeting criteria for diagnoses, allowing us to better explore the extent of co-aggregation of genetic risk across alcohol, cannabis, cocaine and opioid dependence.

Six sites participating in the Collaborative Study on the Genetics of Alcoholism recruited alcohol-dependent probands from in-patient and outpatient facilities. The probands and their family members were administered a poly-diagnostic interview, the Semi-Structured Assessment for the Genetics of Alcoholism . Individuals 17 years of age or younger were administered an adolescent version of the SSAGA. Institutional review boards at all sites approved the study. A subset of the most genetically informative families was selected for a family-based GWAS. This sample has been described in detail elsewhere but salient characteristics are presented here. Families were prioritized based on the number of family members with: available DNA who were also alcohol dependent; available DNA who also had electrophysiology data; and available DNA, regardless of other phenotypes. To reduce heterogeneity,only families consisting primarily of self-reported European-American ethnicity were included in the sample. The final sample was comprised of 118 large European-American families consisting of 2322 individuals with available DNA.Phenotype data for four substances were obtained from the SSAGA. Some individuals were assessed more than once,cannabis grow equipment in which case data from the SSAGA interview at which an individual reported the maximum number of DSM-IV criteria endorsements for the particular substance was used. Two phenotypes were used in the genetic analyses: ANYDEP, a binary aggregate substance dependence phenotype, and QUANTDEP, a quantitative substance dependence phenotype developed using factor analysis. For ANYDEP, individuals were considered affected if they met DSM-IV lifetime dependence criteria for any of the four substances, and unaffected if they did not meet DSM-IV dependence criteria for all four drugs. Individuals younger than 23 years old at their most recent interview who did not meet criteria for dependence on any of the four drugs were recoded to missing/unknown because they had not passed through the primary age of risk. Selection of this age cutoff was based on the median age of onset of alcohol, cannabis, cocaine and opioid/ heroin dependence in the White sub-sample of the US population-based National Epidemiologic Survey of Alcohol and Related Conditions . The median ages ranged from 18 to 22 years, supporting a cut-off of 23 years. In addition, those individuals with insufficient SSAGA to determine whether they were or were not dependent were also coded as unknown . QUANTDEP, the quantitative factor score, was constructed by conducting a confirmatory factor analysis of the seven DSM-IV lifetime dependence criteria for each of the four substances . As we were interested in those genetic underpinnings that were common to all dependence criteria across the four substances, we elected to use a single factor confirmatory model and did not conduct exploratory analyses, in addition to limiting the factor analysis to the dependence criteria to exclude abuse. All individuals with DSM-IV criteria data were utilized, regardless of age or substance use. The factor score from the resulting confirmatory analyses was utilized as the quantitative phenotype. Heritability was estimated for the two phenotypes using the polygenic option in SOLAR .

The correlation between the total number of DSM-IV criteria endorsed and QUANTDEP was estimated using the Pearson correlation coefficient. ANOVA was used to test if QUANTDEP differed according to the number of substance dependence diagnoses met. We also tested if the average QUANTDEP value differed across alcohol, cannabis, cocaine and opioid dependence diagnoses. Post hoc pairwise comparisons employed a Tukey correction for multiple testing.Genotyping for 2105 subjects in these 118 families was performed at the Genome Technology Access Center at Washington University School of Medicine in St. Louis using the Illumina Human OmniExpress array 12.VI . In addition, genotypes previously generated on the Illumina Human 1M-Duo BeadChip by the Center for Inherited Disease Research were included for 224 subjects from these families . Further details describing data cleaning can be found in Wetherill et al. . The final analytic sample included 2322 genotyped individuals. This yielded an average of 19.6 genotyped members per family. The Genome-Wide Association Analysis with Family Data package was utilized to analyze ANYDEP, implemented as a logistic regression model. Relatedness between family members was accounted for via generalized estimating equations. QUANTDEP was analyzed using a linear mixed effects model as implemented in the kinship library in R . This model in the kinship function allows for the covariance matrix to be completely specified for the random effects. The result is that each family has a different covariance pattern based on the kinship coefficients, to model the familial genetic random effects. Gender and birth cohort defined by year of birth were included as covariates in all analyses described above, including statistical models of association, to account for secular trends . As needed, genomic control was applied to correct for inflation. To reduce the scope of multiple testing, only genotyped single-nucleotide polymorphism were included in the initial analyses. After correcting for the final number of autosomal SNPs , the genome-wide significance threshold was P = 8.45 × 10−8 . In regions with significant association results, we analyzed imputed SNPs to further evaluate the evidence for association. SNPs were imputed to 1000 genomes using BEAGLE 3.3.1 as described in Wang et al. . Secondary analyses were performed for significant SNPs to test whether the observed genetic association could be attributed to dependence on a specific substance.