Temperament and character profiles are associated with depression outcome in psychiatric secondary care patients with harmful drinking
Full text
Temperament and character profiles are associated with depression outcome in psychiatric secondary care patients with harmful drinking Vesa Paavonena1, Kaisa Luotoa, Antero Lassilac, Esa Leinonena,b, Olli Kampmana,c a University of Tampere, Faculty of Medicine and Life Sciences, FI-33014 Tampere, Finland b Tampere University Hospital, Department of Psychiatry, FI-33014 Tampere, Finland c Seinäjoki Hospital District, Department of Psychiatry, Huhtalantie 53, 60220 Seinäjoki, Finland 1 Corresponding author: Vesa Paavonen; Address: University of Tampere, Faculty of Medicine and Life Sciences, PO Box 100, FI-33014 Tampere, Finland; Email: [email protected]; Tel: +358407389051; Fax: +358335516164 © 2018. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ This is the post print version of the article, which has been published in Comprehensive psychiatry 2018, 84, 26-31. The final publication is available via https://doi.org/10.1016/j.comppsych.2018.04.001
2 Abstract Background: Temperament and character profiles have been associated with depression outcome and alcohol abuse comorbidity in depressed patients. How harmful alcohol use modifies the effects of temperament and character on depression outcome is not well known. Knowledge of these associations could provide a method for enhancing more individualized treatment strategies for these patients. Methods: We screened 242 depressed patients with at least moderate level of depressive symptoms. The Alcohol Use Disorders Identification Test (AUDIT) was used for identifying patients with marked alcohol use ƉƌŽďůĞŵƐ;hWh/dшϭϭͿĨƚĞƌϲǁĞĞŬƐŽĨĂŶƚŝĚĞƉƌĞƐƐŝǀĞƚƌĞĂƚŵĞŶƚϭϳϯƉĂƚŝĞŶƚƐǁĞƌĞĂƐƐĞƐƐĞĚƵƐŝŶŐ the Montgomery-Åsberg Depression Rating Scale (MADRS), and the Temperament and Character Inventory (TCI-R). Outcome of depression (MADRS scores across three follow-up points at 6 weeks, 6 months and 24 months) was predicted with AUP, gender, and AUP x Gender and AUP x Time interactions together with temperament and character dimension scores in a linear mixed effects model. Results: Poorer outcome of depression (MADRS scores at 6 weeks, 6 months and 24 months) was predicted by AUP x Time interaction (p=0.0002) together with low Reward Dependence (p=0.003). Gender and all other temperament and character traits were non-significant predictors of the depression outcome in the mixed effects model. Conclusions: Possibly due to the modifying effect of alcohol use problems, high Reward Dependence was associated with better depression treatment outcome at 6 months. Harm Avoidance and Self-Directedness did not predict depression outcome when alcohol use problems were controlled. Key words: major depressive disorder; alcohol abuse; Reward Dependence; Self-Directedness; Harm Avoidance; TCI-R
3 1. Introduction2 It is well known that comorbidity of major depressive disorder (MDD) with substance use disorders (SUDs) can lead to impaired response to treatment, more chronic disease courses or recurrence of depressive episodes [1,2]. Because recurrent and chronic depression constitute a major burden of disease [1,2], more efficient treatment strategies and means for earlier identification of high-risk patients are needed. Assessment of individual temperament and character traits could provide a method for enhancing preventive and more individualized treatment strategies for these patients because certain traits and temperaments may predispose individuals to recurrence of depression, and are associated with different courses of depression and with substance use disorder comorbidity and drinking outcomes [3-6]. Individual personality could have pathoplastic effects on recovery from depression; i.e., differences in temperament or character could explain differences in the course of the illness [7]. How harmful alcohol use modifies the effects of temperament and character on depression outcome is not well known. The Temperament and Character inventory (TCI-R) is a 240-item questionnaire that collects information on human personality in the context of temperament and character. According to Cloninger’s psychobiological model, on which the TCI-R was based, temperament is divided into four dimensions: Novelty Seeking (NS), Harm Avoidance (HA), Reward Dependence (RD) and Persistence (P). Character is divided into three dimensions: Self-Directedness (SD), Cooperativeness (C) and Self-Transcendence (ST) [8]. These dimensions are thought to reflect combinations of different neurocognitive functions (e.g., memory or reward functions), and the biological basis of this model is supported by many findings [9,10]. According to Cloninger’s theory, temperament dimensions generally represent a stable part of personality, and only HA has shown clear state-dependent changes during depression [4,5,8]. However, high HA – and more particularly its sub-scores, anticipatory worry (HA1) and fatigability (HA4) – have also manifested as 2 Abbreviations: MDD = major depressive disorder, SUD = substance use disorder, NS = Novelty Seeking, HA = Harm Avoidance, RD = Reward Dependence, P = Persistence, SD = Self-Directedness, C = Cooperativeness, ST = SelfTranscendence, AUP = alcohol use problems, MADRS = Montgomery-Åsberg Depression Rating Scale, TCI-R = Temperament and Character Inventory
4 trait-like markers for risk of depression, i.e., index episodes, relapses or recurrent episodes and impaired treatment response [4,5]. High RD could be protective against depression in general population, but no associations have been reported with outcome of depression in clinically depressed patients [4,11-15]. Our earlier study of this patient sample suggested that RD is associated with depression treatment outcome in patients with alcohol use problems as change in RD was strongly associated with acute treatment response (0-6 weeks) to depression when alcohol use was taken into account [16]. High NS is a trait indisputably associated with risk of substance use disorders, more severe symptomatology and poorer outcome in SUD patients, and apparently experienced at a higher level in patients with dual diagnosis (concurrent SUD and mental illness) than in depressed patients [3,6,17-20]. Low P and high HA are associated with more severe alcohol dependence symptomatology [21-23]. Of the character traits, low SD is the trait most clearly predisposing to depression and recurrence of episodes, possibly because individuals with deficiencies in sub-traits such as self-acceptance, responsibility, goal-directedness associated with SD may be more prone to depression due to difficult situations encountered in their daily lives [24,25]. There is less evidence to suggest that low C is associated with the development of depression, whereas findings on the associations between ST and depression in different patient samples have been contradictory [4,15,26,27]. Low SD is also associated with more severe symptomatology and drinking outcomes in SUD patients [6,28,29]. In alcohol dependence character profile with high ST and low SD and C is associated with depression and anxiety [21]. In spite of a large body of knowledge of different associations separately between temperament and character traits and depression or substance use disorders, we found no follow-up studies addressing the associations between depression outcome and temperament, character and alcohol use. We investigated whether temperament and character trait scores (at 6 weeks) together with harmful alcohol use predict outcome of depression in follow-up from 6 weeks to 6 months and to 24 months in a clinically diverse sample of depressed patients. In light of earlier evidence we hypothesize that high HA and low SD and harmful alcohol use together explain poorer depression treatment outcome (measured as MADRS scores)
5 [4,5,25]. As harmful alcohol use had a modifying effect on both temperament and character dimensions during acute illness, we hypothesize that RD together with alcohol use is also associated with outcome of depression in the long-term follow-up (from 6 weeks to 6 and 24 months) [16]. High NS has been associated with more severe SUDs, and therefore we also hypothesized that this temperament trait might modify treatment outcome together with harmful alcohol use in depression [6]. 2. Methods 2.1 Participants In the period 2009 - 2013, 242 patients were screened for the study in the Finnish region of Southern Ostrobothnia (population 200,000). These patients were referred to psychiatric specialized care units (5 outpatients and 1 inpatient) due to depression, anxiety, self-destructiveness, insomnia or alcohol-related problems. To maximize clinical relevance, lenient inclusion criteria were used. Patients with at least moderate depressive symptomatŽůŽŐLJ;ĞĐŬĞƉƌĞƐƐŝŽŶ/ŶǀĞŶƚŽƌLJ/sĞƌƐŝŽŶϭƐĐŽƌĞшϭϳ[30]) were included in the study. Patients with organic brain disease or psychotic disorder (ICD-10 F2* diagnosis) were excluded. Their age range was 17–64 years (mean 38.8 years, SD±12.2). A more detailed description of the sample is presented in Tables 1a and 1b, and of the study setting elsewhere (see ClinicalTrials.gov Identifier NCT02520271, Ostrobothnia Depression Study [ODS], 2016). The study was approved by the local Human Subjects Review Committee, and patients gave their informed written consent. [Tables 1a and 1b] 2.2 Procedures Sociodemographic data were collected and clinical assessments conducted at screening (the Alcohol Use Disorders Identification Test (AUDIT) [31] and the BDI). There was some dropout before the baseline assessment using the Mini International Neuropsychiatric Interview 5.0 (MINI; [32]) and the MontgomeryÅsberg Depression Rating Scale (MADRS) [33] and 228 (94%) patients were assessed with MADRS at baseline. According to the MINI administered to 219 patients (data missing in 23 cases), 88.6% of the
6 patients had MDD, 4.1% dysthymic disorder, 5.5% anxiety disorder and 0.4% alcohol use disorder (AUD) as their main diagnosis, and 1.4% of the patients did not meet any of the diagnostic criteria. Twelve percent (12%) of the patients met the criteria for lifetime diagnosis of (hypo)manic episode. Sixty-three percent (63%) of patients with mood disorder as their main diagnosis had comorbid anxiety disorders, corresponding well to comorbidity proportions found in other samples in Finnish psychiatric secondary services [34]. Six patients (3%) with mood disorder as primary diagnosis had comorbid bulimia nervosa. Patients’ categorical personality disorder diagnoses were not assessed. In the total sample, 33.6% of the patients reported that this was their first episode of MDD. At baseline patients attended an appointment with a psychiatrist, where their medication was evaluated and changed if necessary. Antidepressant medication was prescribed to 206 patients (85%) with mean fluoxetine equivalent daily doses of 33.0 mg (SD±18.3). Of these, 82% had either an SSRI or SNRI as a primary antidepressant. Adherence to antidepressants was monitored during the first six weeks of the study using a paper and pencil diary (for more information see [16]). All patients received behavioral activation therapy with trained clinical staff. The median number of therapy sessions with patients was 6 (IQR = 3-11) with sessions taking place at 1 to 2-week intervals. The treatment of patients with alcohol use ƉƌŽďůĞŵƐ;hWh/dƐĐŽƌĞƐшϭϭͿǁĂƐĞŶŚĂŶĐĞĚǁŝƚŚŵŽƚŝǀĂƚŝŽŶĂůŝŶƚĞƌǀŝĞǁing (median number of sessions 4, IQR = 3-6) at the start of the treatment according to a specific treatment intervention procedure (see ClinicalTrials.gov Identifier NCT02520271, Ostrobothnia Depression Study [ODS], 2016 and [35]). The cut-off point for AUDIT was chosen because in Finnish clinical practice it indicates a significantly increased risk of harm due to alcohol [36,37]. In the present study setting the aim was to identify individuals with marked alcohol use problems with high specificity to obtain motivational interview as an add-on psychosocial treatment. After dropout at 6 weeks, 173 patients completed both the TCI-R and MADRS and were therefore eligible for inclusion in the main analysis of this study. Of these patients 61 (35%) showed marked alcohol use problems (AUP, h/dƐĐŽƌĞƐшϭϭͿƌŽƵŐŚůLJĐŽƌƌĞƐƉŽŶĚŝŶŐƚŽƚŚĞĐŽŵŽƌďŝĚŝƚLJƌĂƚŝŽŽĨϮϭ;ŽĨƐƵďƐƚĂŶĐĞƵƐĞ
7 disorders in depressed patients) observed in clinical samples where nearly one third of patients with major depressive disorder also have substance use disorders [38]. Of the 61 AUP patients MINI diagnosis had been assessed in 60 cases at baseline, 44 (73%) of which had been diagnosed with lifetime AUD and of the other 112 patients (with AUDIT scores < 11) MINI diagnosis had been assessed in 107 cases, only 3 (3%) of which had been diagnosed with lifetime AUD. Seven (16%) patients that had been diagnosed with AUD had also other SUD, and two (2%) patients without AUD were diagnosed with some other SUD. There was no exclusion of patients according to substance use. Alcohol use problems were more common in male patients: (males 69% vs. females 31%, ʖ2=31.5, p < 0.001; odds ratio (OR) = 6.6, 95% confidence intervals (CI) = 3.3–13.2). The follow-up included assessment of patients’ MADRS scores again at 6-month and 24month time-points. 2.3 Statistical methods Differences in continuous variables (AUDIT and MADRS scores, and TCI-R dimension scores) between dropouts and other patients were calculated with independent samples t-tests. Differences in nominal variables between dropouts and other patients and in the prevalence of AUP between genders were calculated with ʖ2 statistics. This study was conducted according to the intention-to-treat protocol and in cases of dropout MADRS scores were imputed in the follow-up according to the last observation carried forward (LOCF) method. This replaced missing values in MADRS follow-up with their last observed values at earlier follow-up time points. This replaced missing MADRS values in 32 (17%) cases at 6-month follow-up and in 93 (49%) cases at 24month follow-up. Exploratory analysis between LOCF MADRS scores and temperament or character scores was performed with Pearson’s correlation coefficients. A linear mixed effects model was used for the repeated measurement testing in the main analysis of the study. This model predicted LOCF MADRS scores from 6 weeks to 6 months and to 24 months with the scores of the seven temperament and character dimensions (NS, HA, RD, P, SD, C, and ST; at six weeks) used as explaining variables and was adjusted with AUP, gender, and AUP x gender and AUP x time
8 interactions. Individual-specific intercept and slope terms were used in the model. The -2 log likelihood information criteria was used in evaluating model fit and model with unstructured covariance structure was reported. Kenward-Roger adjustment of degrees of freedom was applied for estimates of fixed effects. The main analysis was performed with PROC MIXED, SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and all other analyses were performed with SPSS for Mac (version 24.0, IBM Inc., Armonk, New York, USA). 3. Results The dropout rate in the study was 65 (27%) at six weeks, 91 (38%) at six months, and 147 (61%) at 24 months, but no gender differences were seen at any assessment point. Of these patients, the clinical treatment had been concluded in co-operation with the patients in 10 (11%) cases during the first 6 months and in 28 (19%) cases before 24 months. The dropout analysis of the raw data revealed similar baseline MADRS scores in dropouts and other patients. Baseline AUDIT scores were higher in dropouts at both rating points (6 months: dropout 14.0±11.2 vs. other patients 8.9±8.6, p < 0.001; 24 months: 12.0±10.5 vs. 8.7±8.5, p = 0.006 for t-test). Dropouts also had lower Self-Directedness at baseline than did other patients in dropout analysis at 6 months: dropout 117.9±18.3 vs. other patients 124.2±18.2, p = 0.02 for t-test, and had trending but non-significant difference at 24 months (p = 0.09). In the LOCF data there were no statistically significant differences between dropouts and other patients’ 6-month LOCF MADRS scores in the 6-month dropout analysis (p = 0.2 for t-test). Dropouts had higher 24-month LOCF MADRS scores in the 24-month dropout analysis (dropout 12.9±8.9 vs. other patients 8.3±7.6, p < 0.001 for t-test). The main results emerging as predictors of depression treatment outcome (measured as MADRS scores) are presented in Table 2. Poorer outcome of depression was predicted by AUP x Time interaction together with low RD. The model resulted in steeper negative sloping of MADRS scores for non-AUP group when compared to AUP group x Time. This means that alcohol use problems associated with poorer outcome of depression in the follow-up from 6 weeks to 6 months and to 24 months. Low RD was the only temperament or character trait that was associated with depression outcome as a predictor of poorer outcome in the follow-up from 6 weeks to 6 and 24 months.
9 [Table 2] The distributions (mean±SD) for LOCF MADRS scores were: 1) 13.6±8.5 at 6 months and 2) 10.6±8.5 at 24 months. The mean±SD for temperament and character dimensions at six weeks were: NS 99.7±16.7; HA 113.4±18.9; RD 100.1±17.6; P 99.7±21.5; SD 125.4±17.3; C 131.9±18.0; ST 64.9±15.5. The Cronbach’s ɲ values for TCI-R dimensions have been reported elsewhere [16]. The MADRS scores, number of responders, remitted patients, and non-responders in follow-up are presented in Table 3. [Table 3] Pearson’s correlation coefficients between RD, HA and SD and LOCF MADRS scores were statistically significant for RD and MADRS scores at 6 months (r = -0.32, p < 0.001), HA and MADRS scores at 6 months (r = 0.18, p = 0.02), SD and MADRS scores at 6 months (r = -0.27, p < 0.001), HA and MADRS scores at 24 months (r = 0.26, p = 0.001), and SD and MADRS scores at 24 months (r = -0.32, p < 0.001). 4. Discussion The main hypotheses in this study were that temperament and character traits together with harmful alcohol use (assessed in the early stages of treatment) explain outcome of depression over a period of two years. The main finding in this study was that poorer outcome of depression was predicted by low Reward Dependence. Alcohol use problems were also associated with poorer outcome of depression in the followup from 6 weeks to 6 months and to 24 months. 4.1 Outcome of depression (from 6 weeks to 24 months) Poorer outcome of depression was predicted by low Reward Dependence in the linear mixed effects model adjusted with AUP, gender, AUP x gender and AUP x time. In contrast to some earlier studies our present and earlier findings suggest that Reward Dependence is associated with depression treatment outcome [4,16,26]. This difference in results could be explained by differences in patient samples, as patients with SUDs have been excluded in those earlier studies but not in this present one [4,16,26]. Our earlier and present results suggest that low RD predicts poorer outcome of depression particularly in depressed
16 [32] Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry 1998;59:22-33;quz. [33] Montgomery SA, Asberg M. A new depression scale designed to be sensitive to change. Br J Psychiatry 1979;134:382-389. [34] Melartin TK, Rytsala HJ, Leskela US, Lestela-Mielonen PS, Sokero TP, Isometsa ET. Current comorbidity of psychiatric disorders among DSM-IV major depressive disorder patients in psychiatric care in the Vantaa Depression Study. J Clin Psychiatry 2002;63:126-134. [35] Lindholm LH, Koivukangas A, Lassila A, Kampman O. Early assessment of implementing evidence-based brief therapy interventions among secondary service psychiatric therapists. Eval Program Plann 2015;52:182-188. [36] Luoto KE, Koivukangas A, Lassila A, Kampman O. Outcome of patients with dual diagnosis in secondary psychiatric care. Nord J Psychiatry 2016;70:470-476. [37] Babor TF, Higgins-Biddle JC, Saunders JB, Monteiro MG. Alcohol Use Disorders Identification Test, The: Guidelines for Use in Primary Health Care. 2nd ed. Geneva: Department of Mental Health and Substance Dependence, World Health Organization; 2001. [38] Davis LL, Frazier E, Husain MM, Warden D, Trivedi M, Fava M, et al. Substance use disorder comorbidity in major depressive disorder: a confirmatory analysis of the STAR*D cohort. Am J Addict 2006;15:278-285. [39] Koob GF, Le Moal M. Addiction and the brain antireward system. Annu Rev Psychol 2008;59:29-53. [40] Sellman JD, Mulder RT, Sullivan PF, Joyce PR. Low persistence predicts relapse in alcohol dependence following treatment. J Stud Alcohol 1997;58:257-263. [41] Kupfer DJ. Long-term treatment of depression. J Clin Psychiatry 1991;52:28-34. [42] Solomon DA, Keller MB, Leon AC, Mueller TI, Shea MT, Warshaw M, et al. Recovery from major depression. A 10-year prospective follow-up across multiple episodes. Arch Gen Psychiatry 1997;54:10011006. Ostrobothnia Depression Study (ODS). A Naturalistic Follow-up Study on Depression and Related Substance Use Disorders. https://clinicaltrials.gov/ct2/show/NCT02520271, (accessed: April 04, 2017)
Table 1a. Sociodemographic data on the patient sample Table 1b. Baseline scores on BDI and AUDIT by gender Men Women Total Baseline BDI mean (±SD) 28.5 (6.80) 27.5 (7.59) 27.9 (7.30) Baseline AUDIT mean (±SD) 15.1 (10.59) 7.9 (8.30) 10.7 (9.88) Abbreviations: BDI = Beck Depression Inventory; AUDIT = Alcohol Use Disorder Identification Test Men Women Total N % N % N % Total 94 38.8 148 61.2 242 100 Marital status Single 34 36.9 39 29.1 73 32.3 Married or cohabiting 39 42.4 71 53.0 110 48.7 Divorced 19 20.7 21 15.7 40 17.7 Widowed 0 0 3 2.2 3 1.3 Education Primary school 4 4.3 3 2.2 7 3.1 Comprehensive school 27 29.3 28 20.7 55 24.2 Tertiary education 10 10.9 25 18.5 35 15.4 Vocational school 33 35.9 54 40.0 87 38.3 Upper secondary education 10 10.9 7 5.2 17 7.5 Polytechnic or university 8 8.7 18 13.3 26 11.5 Work status before sick leave Employed 39 42.9 67 50.0 106 47.1 Unemployed 41 45.1 30 22.4 71 31.6 Housewife/husband 0 0 10 7.5 10 4.4 Pensioner 5 5.5 11 8.2 16 7.1 Student 6 6.6 16 11.9 22 9.8 Self-reported history of depression episode 62 67.4 90 65.7 152 66.4 First degree family history of depression 33 35.9 56 41.5 89 39.2 First degree family history of bipolar disorder 4 4.3 10 7.4 14 6.2
Table 2. Predictors of depression outcome (LOCF MADRS scores across the follow-up from 6 weeks to 6 months and to 24 months)a -2 x log-likelihood = 3530 for the SAS input code and output results a Results from the linear mixed effects model with temperament and character dimension scores, gender, AUP, and AUP x Gender and AUP x Time interactions as explanatory variables bB for the temperament and character variables effect on the dependent variable Abbreviations: LOCF = Last observation carried forward; NS = Novelty Seeking; HA = Harm Avoidance; RD = Reward Dependence; P = Persistence; SD = Self-Directedness; C = Cooperativeness; ST = SelfTranscendence; MADRS = Montgomery-Åsberg Depression Rating Scale; AUP = alcohol use problems Significant results are presented in bold face. Fixed effects LOCF MADRS scores Estimateb SE t p Intercept 20.07 10.04 2.00 0.049 NS at 6 weeks 0.04 0.03 1.16 0.25 HA at 6 weeks 0.06 0.04 1.70 0.09 RD at 6 weeks -0.11 0.04 -3.03 0.003 P at 6 weeks -0.03 0.03 -1.10 0.27 SD at 6 weeks -0.06 0.04 -1.59 0.11 C at 6 weeks 0.02 0.04 0.56 0.58 ST at 6 weeks 0.05 0.03 1.49 0.14 Male gender 0.24 1.76 0.14 0.23 AUP 3.67 1.64 2.25 0.12 AUP x Male gender -3.24 2.19 -1.48 0.14 AUP x Time -0.10 0.05 -1.98 0.0002 non-AUP x Time -0.33 0.06 -3.75 0.0002
Table 3. MADRS scores in raw data, LOCF data, and by patient subgroup and number of responders, patients in remission and non-responders at baseline and follow-up baseline 6 weeks 6 months 24 months Raw data MADRS scores (mean±SD) 23.2±6.7; n=228 16.9±8.0; n=188 13.1±8.7; n=156 8.3±7.6; n=95 Response* (n, %) 69, 44% 64, 67% Remission** (n, %) 49, 31% 50, 53% LOCF data MADRS scores, n=188 13.6±8.5 10.6±8.5 Response* (n, %) 80, 43% 112, 60% Remission** (n, %) 52, 28% 80, 43% MADRS scores of patient subgroups (Raw data scores): non-AUP (mean±SD) 22.9±6.7 (n=136) 16.7±8.0 (n=123) 12.8±8.7 (n=105) 6.7±6.2 (n=66) AUP (mean±SD) 23.7±6.7 (n=92) 17.2±8.1 (n=65) 13.7±8.7 (n=51) 12.0±9.2 (n=29) Female (mean±SD) 22.3±6.8 (n=137) 15.7±7.8 (n=112) 11.6±8.2 (n=97) 7.1±6.6 (n=64) Male (mean±SD) 24.6±6.3 (n=91) 18.6±8.0 (n=76) 15.6±9.1 (n=59) 10.8±8.9 (n=31) proportion of men (n, %) 91, 40 % 76, 40% 59 , 38% 31, 33 % proportion of AUP patients (n, %) 92, 40 % 65, 35% 51 , 33% 29, 31 % Number of patients with increase in symptoms from baseline (n, %) 23, 15 % 3, 3 % Abbreviations: AUP= Alcohol use problems; MADRS = Montgomery-Åsberg Depression Rating Scale; LOCF = last observation carried forward *at least 50% MADRS score decline from baseline; **MADRS scores < 8