The study was carried out in the outpatient clinic of the division of substance abuse of the University Hospitals of Geneva. For 1 year, participation in the study was proposed systematically to each adult (age ≥ 18 years) outpatient who was treated for opioid or cocaine dependence. Patients were excluded if they had an acute psychotic or manic episode, had a severe major depressive episode, patients did not understand the French language, were unable to give informed consent, or were already in treatment for problems related to alcohol misuse at the time of inclusion.
The study protocol was approved by the Geneva Ethics Committee. All participants received written information about the study and gave written informed consent.
The participants were assessed with the following instruments:
Alcohol Use Disorders Identification Test (AUDIT)
The AUDIT  is a 10-item self-assessment questionnaire that presents good sensitivity and specificity for the screening of excessive alcohol use and dependence [26–28]. The first three questions of the AUDIT assess frequency and quantity of alcohol use. AUDIT has been translated and validated in the French language . According to the cutoff values in the French validation, three groups can be identified:
abstinent patients or occasional nonproblematic drinkers (score < 7 for men and < 6 for women); (b) excessive drinkers (7 ≤AUDIT score < 13 for men and 6 ≤AUDIT score < 13 for women); and (c) alcohol dependents (score > 13).
All screened patients received feedback that explained the meaning of their AUDIT score. All study participants were assessed with the AUDIT questionnaire at inclusion and those with excessive drinking or alcohol dependence were also assessed at 3 and 9 months. Diagnoses were established according to the criteria of the International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10)  by a resident and a senior psychiatrist.
Demographic and clinical characteristics collected at baseline
Patients identified as having excessive alcohol use or dependence (AUDIT scores ≥ 6 for women or ≥ 7 for men) were randomized to receive treatment as usual (control group) or treatment as usual plus BI (intervention group). The patients in both groups were already in treatment for opioid or cocaine dependence before study inclusion. The patients allocated to BI received this intervention 2 or 3 weeks after AUDIT screening (time to assign BI to the staff after randomization and to give the patient an appointment).
A BI was dispensed between 1 and 3 weeks after the screening with the AUDIT questionnaire by specifically trained staff. The form in which BI was dispensed was similar to that described elsewhere [14, 25] and was based on the following principles and actions:
Provide feedback to the patient about the result of the AUDIT questionnaire.
Identify risks and discuss consequences.
Display an emphatic and nonjudgmental attitude.
Solicit the patient's commitment.
Identify the goal: reduced drinking or abstinence.
Propose a decrease in alcohol consumption with a choice of personal strategies.
Emphasize personal responsibility for change and stimulate an attitude of change.
Give advice and encouragement.
Promote self-observation in the consumption of alcohol.
BI was provided by a multidisciplinary team (psychiatrists, psychologists, nurses, and social workers) in the Division of Substance Abuse of the University Hospitals of Geneva. BI took 16 min with a standard deviation of 4.7. Training was provided during two workshops (4 h) by an expert in the field. He provided the staff with guidelines and information about the principles of BI.
The control group received treatment as usual in addition to AUDIT and score feedback. Treatment as usual refers to outpatient pharmacological and psychosocial treatment in the Division of Substance Abuse of the University Hospitals of Geneva. The outpatient staff is a multidisciplinary team: a psychiatrist, general practitioner, psychologist, nurse, and social worker.
Maintenance treatment with methadone or heroin includes medical and psychiatric follow-up, primary health care, psychosocial interventions, and administration of opiate treatments in a clinical setting. Psychosocial treatment includes medical and psychiatric follow-up, primary health care, psychosocial interventions, and, if necessary, administration of pharmacotherapy in a clinical setting.
The outcomes were as follows: (a) the AUDIT scores; (b) the number of glasses of alcohol per week (1 glass: 10 g of alcohol; wine = 100 ml; beer = 250 ml; spirits = 25 ml); and (c) frequency of alcohol use (consumption rate). These outcomes were assessed at baseline and then at months 3 and 9.
Two hundred and fifty-four patients met the study inclusion criteria and accepted the invitation to participate. Of this number, and after an AUDIT screening conducted by a physician or a nurse, 112 patients were subsequently randomly assigned to intervention or control groups in a 1:1 ratio. The randomization scheme was drawn by a statistician, who used the Web site [http://www.randomizer.org/]. A random permuted block method was used, with blocks of 4 patients. The sequence was concealed from all investigators with numbered opaque sealed envelopes prepared by the statistician and handed over to the physician in charge of the study.
Statistical analysis was performed by using SPSS for Windows (version 18.0, IBM, Chicago, IL, USA). An initial exploratory analysis involved the calculation of proportions, means, and standard deviation to describe the baseline characteristics.
To analyze differences between participants who gave follow-up data and those who did not, we performed two logistic regressions. A binary dependent variable was generated, taking on a value of 0 if audit scores were missing at T3 and a value of 1 if audit scores were present. Type of drinkers (excessive drinkers vs. alcohol dependent) and treatment groups (treatment as usual vs. treatment as usual plus BI) served as independent variables. The same was done at T9. AUDIT internal consistency was explored by using the Cronbach α coefficient. This index varies between 0 and 1 and translates a greater degree of internal coherence if its value is close to 1. It is generally accepted that the internal consistency of an instrument is satisfactory when the value of the coefficient is equal to or above 0.70.
The evolution of AUDIT scores was analyzed by repeated measures analysis of variance (ANOVA), with treatment group (treatment as usual vs. treatment as usual plus BI), type of drinker (excessive drinker vs. alcohol dependent), and sex (male vs. female) as factors. The evolution of the quantity of alcohol consumed (number of glasses of alcohol per week) was also considered by ANOVA models, again with treatment group, type of drinker, and sex as factors. The variable that measured the quantity of alcohol consumed had to be log-transformed [x' = log(x+1)] beforehand, with the aim of making the distribution less skewed; two unlikely values, one in each group (200 glasses/week for one participant of the control group and 196 glasses/week for another of the BI group, respectively) were not considered in the analysis. In these ANOVA models, main effects, factor × time and factor × factor × time interactions, were paid due attention.
The consumption rate was estimated by using the first item of the AUDIT questionnaire (How often do you have a drink containing alcohol? Never (0), Monthly or less (1), Two to four times a month (2), Two to three times per week (3), Four or more times a week (4)). The first question of the AUDIT explores the frequency of alcohol consumption during the last year (never, once a month, 2 to 4 times per month, 2 or 3 times per week, 4 times or more per week).
We considered the decrease of consumption between the study time periods as a success (coded 1) and an increase or no change as a failure (coded 0). This method results in three binary outcomes (0, 1) for measures at baseline (T0), month 3 (T3), and month 9 (T9). Nonparametric Cochran's Q tests were carried out to assess whether the distribution of the values is the same for the three related dichotomous variables.
All analyses were done on a modified intention-to-treat basis. Missing data were handled by multiple imputation techniques in which scale variables were modeled with linear regression and categorical variables with logistic regression. Under the assumption that data are missing completely at random, pooled estimates were calculated and the complete data set could then be analyzed by standard methods. All statistical analyses were performed with a significant threshold of α = 0.05.