Men and women aged 26 to 65 years who were voluntarily seeking outpatient treatment for alcohol problems at three Helsinki municipal Alcohol-clinics (A-clinics). Patients with a history of heavy drinking (averaging five or more daily drinks for men and four or more daily drinks for women) for at least ten years, significant depression defined by the Beck Depression Inventory II (BDI-II > 16), and who were interested in voluntarily taking part in the study were recommended by their A-clinic doctor or social worker therapist to be interviewed and screened by the study physician. The patients were interviewed by the study doctor (psychiatrist LM) applying the Structured Clinical Interview for DSM- IV (SCID) and were required to meet the criteria of both alcohol dependence and MDD according to DSM-IV-TR. Abstinence was not required but encouraged. The time since the last prior inpatient detoxification had to be at least four weeks. In addition, the eligible patients had to be currently in a depressive episode lasting for more than two weeks. The exclusion criteria included other substance use dependence screened by urine test (amphetamine, benzodiazepines, cocaine, tetrahydrocannabinol and opiates, schizophrenia or other psychotic disorder, and bipolar I and II disorder, acute risk of suicide, pregnancy or breastfeeding, a severe untreated somatic problem, or a serious dysfunction of the liver (aspartate aminotransferase [AST] and alanine aminotransferase [ALT] > 200), and mental disability. Other medications prescribed by participants' physicians were allowed, with the exception of other antidepressants. All patients were Caucasian, and 55% were men. There were no significant differences between the groups in either their demographic characteristics or their initial alcohol and depressive measures 44 . The mean length of the present depressive period was 35 months. Current alcohol use was reported by 17 patients (43.6%) in the memantine group and 17 (42.5%) in the escitalopram group. The number of A-clinic visits (psychosocial counseling) during the study period was similar: in the memantine group 7.7 ± 8.8 (mean ± SD) and in the escitalopram group 7.1 ± 9.2. 44 .

All 58 subjects who completed the study attended all appointments and showed at least 80% compliance based on tablet counts. The average consumption (mg) of medication did not differ between the two medication groups: during the first 12 weeks, for memantine 17.4 ± 0.5 mg and for escitalopram 16.9 ± 0.6 mg; and during weeks 13–26, for memantine 17.4 ± 0.6 mg and for escitalopram 15.9 ± 0.8 mg.

The study was approved by the independent Hospital District of Helsinki and Uusimaa, Ethical Committee (permission 22/2004) and the Finnish National Agency of Medicine (KL# 87/2004). The study was conducted according the ICH Guidelines for Good Clinical Practice and the 1964 Declaration of Helsinki. The study was registered on the National Public Health study registry in March, 2005 (172–9), and the ClinicalTrials.gov Identifier (trial # NCT00368862). All patients had to be able to read and understand the patient information sheet and sign the informed consent. All participants were free to stop the study medication whenever they wanted. The patients were not paid or otherwise reimbursed for participation.

Eighty-nine patients were initially screened. A screening interview (SCID) was conducted to confirm the diagnoses of MDD and alcohol dependence. Patients completed questionnaires including the Obsessive-Compulsive Drinking Scale (OCDS 45 ) and the Alcohol Use Disorders Identification Test (AUDIT) 46 . AUDIT-QF 47 , and AUDIT-3 48 were used for a detailed drinking analysis. The recording of alcohol consumption during the 26-week treatment period was done with a personal drinking diary for all days, including abstinent days 49 .

All patients meeting the inclusion criteria were randomly assigned to the memantine or escitalopram group using a 1:1 ratio (n = 40 + 40). Eligible patients received orally either 20 mg/day escitalopram or 20 mg/day memantine. The starting dose was 5 mg/day for both drugs and was increased at weekly intervals by 5 mg/day to 20 mg/day. Patients were instructed to take the study medication in the morning. Patients were permitted to telephone the study physician at any time. If the patient did not appear at a scheduled visit, a new appointment was offered.

During the 26-week treatment period, the patients returned to the study site at weeks 1, 2, 4, 12 ± 2, and 26 ± 2 for data collection and for medication checking and dispensing. At each visit, the drinking diary and the study medication intake since the previous visit were recorded from the medication diary. The study medication was ensured by pill count from the returned blister-packs. Outcomes were recorded on specific weeks: OCDS (weeks 0, 4, 12 and 26); AUDIT (week 0, 12 and 26, the later ones modified to report the events in the previous month). Clinical laboratory tests (MCV, AST, ALT, CDT, and GGT) were taken at the beginning of the study and were repeated at weeks 4, 12, and 26, to ensure the safety of the medication. No breath or blood test for alcohol was performed, but if the patient was obviously intoxicated, a new appointment was offered. The study was monitored by an independent organization, Medikalla Oy, Medfiles, Turku.

All primary and secondary outcome statistical analysis was performed by an independent source (Medikalla Oy, MedFiles, Turku). All statistical evaluation utilized SAS Procedures in SAS^® system for Windows (Version 8.2), SAS-institute, Espoo, Finland.

Intent-to-treat sample, which included all randomized patients including two patients who discontinued early in the study and reported, taking no medication, were used in all tables and analyses. Descriptive statistics were calculated for all variables. Categorical variables were presented in frequencies tables (number of cases and percentages) by treatment. The numerical variables were tabulated by treatment. Baseline measures were analyzed by logistics regression or analysis of variance. All repeated dependent measures (drinking diary, OCDS, AUDIT, laboratory tests), were analyzed with analysis of variance for repeated measures (ANOVA) when treatment, time, and treatment * time interaction were in the model (PROC MIXED in SAS^®) and responses to the specific question (Has your alcohol use diminished during the study?) were analyzed by logistic regression (PROG LOGOSTIC in SAS^®).

Furthermore, predictors of treatment response by medication were analyzed with multiple linear regression analyses by adjusting for the baseline OCDS and AUDIT scores.

The baseline AUDIT and alcohol use histories are similar in both groups (Table 1). AUDIT scores decreased (Fig. 1) from baseline in both groups, from 27.4 ± 7.1 to 14.3 ± 9.9 in the memantine group and from 28.4 ± 6.4 to 17.6 ± 10.4 in the escitalopram group. The overall reduction was highly significant (F [2.77] = 48.42, p < .0001) in both groups combined. The treatment by time interaction was not significant (F [2.77] = 1.19, p = 0.31).

Alcohol consumption measured by the AUDIT QF (quantity-frequency) score was significantly reduced in both groups: in the memantine group from 6.2 ± 1.7 to 4.1 ± 2.5 and from 6.1 ± 1.7 to 4.3 ± 2.3 in the escitalopram group (F [2.77] = 23.53, p < .0001). The treatment by time interaction was not significant (F [2.77] = 1.58, p = 0.21). The number of heavy drinking days measured by the AUDIT-3 score was also diminished significantly in both groups: for the memantine group from 2.9 ± 1.1 to 1.8 ± 1.3 and from 3.1 ± 1.0 to 2.4 ± 1.3 for the escitalopram group (F [2.77] = 20.29, p > .0001). The treatment by time interaction was not significant (F [2.77] = 1.37, p = 0.27).

The number of abstinent days per week was high for both groups throughout the study. The treatment by time interaction in the number of abstinent days per week was not significant (F [2.74] = 0.07, p = 0.93) (Figure 2). The mean alcohol intake including abstinent days was 15.0 ± 2.6 g per day for the memantine group and 21.1 ± 3.6 g per day for the patients on escitalopram, with no significant difference between the groups (F [1.74] = 1.94, p = 0.17).

When questioned at the end of the intervention, 68.9% of the patients in the memantine group reported their alcohol use had decreased while 62.1% of the patients in the escitalopram group reported a decrease.

The OCDS total scores (Fig. 3) decreased in the memantine group from 18.8 ± 6.9 to 10.6 ± 7.2 and in the escitalopram group from 20.4 ± 4.9 to 12.8 ± 8.6. The overall reduction was highly significant (F [3.77] = 25.76, p < .0001) in both groups combined. The treatment by time interaction in the OCDS was not significant (F [3.77] = 0.69; p = 0.56).

In both groups, the mean serum concentrations of AST, ALT, GGT, and CDT were within normal limits, and there were no significant changes during the treatment period or any significant differences between the groups. The adverse events during the study are published in our previous article 44 . There was no significant difference in reporting adverse events between the medication groups.

In general, those patients who were abstinent at the beginning of treatment were more likely to complete the treatment than those who were still drinking at the beginning (χ ² = 6.51, df = 1, p = 0.011). This relationship was highly significant in the patients treated with memantine (χ ² = 7.25, df = 1, p = 0.007): 8 of the 11 who dropped out were among the 17 (47.1%) who were active drinkers at the baseline. The relationship was in the same direction in the escitalopram group but failed to reach statistical significance (χ ² = 0.901, df = 1, p = 0.343).

We tested by multiple linear regression analyses whether age at onset of depression and age at first alcohol intoxication predicted change during the six month treatment in the OCDS and the AUDIT scores. Age at onset of depression predicted differently change in the OCDS in the escitalopram group compared with the memantine group (R ² change for interaction term = 0.05; p for treatment by age at onset of depression interaction = 0.05). In the escitalopram group, earlier onset of depression predicted less change in the OCDS scores during the six months (B = -0.31, 95% CI = -0.53 to -0.09, p = 0.008) whereas in the memantine group no such association existed (B = -0.02, 95% CI = -0.22 to 0.17, p = 0.81). There were no differences between the medication groups in the association between age at onset of depression and change in the AUDIT (R ² change for interaction term = 0.02; p for interaction = 0.21) or in the association between age at first alcohol intoxication and change in the OCDS and the AUDIT (R ² change for interaction term < 0.022; p-values for interactions > 0.21). Neither did age at onset of depression predict change in the AUDIT (B = -0.16, 95% CI = -0.37 to 0.05, p = 0.12) or age at first alcohol intoxication predict change in the OCDS (B = -0.68, 95% CI = -1.41 to 0.05, p = 0.07) and the AUDIT (B = -0.24, 95% CI = -1.62 to 1.13, p = 0.72) in the subjects when both treatment groups were combined.