9618_Antidepressant Treatment Of Major Depressive Disorder In Patients With Comorbid Alcohol Use Disorder

Yale University
Yale University
EliScholar – A Digital Platform for Scholarly Publishing at Yale
EliScholar – A Digital Platform for Scholarly Publishing at Yale
Yale Medicine Thesis Digital Library
School of Medicine
January 2020
Antidepressant Treatment Of Major Depressive Disorder In
Antidepressant Treatment Of Major Depressive Disorder In
Patients With Comorbid Alcohol Use Disorder: Two Meta-
Patients With Comorbid Alcohol Use Disorder: Two Meta-
Analyses Of Randomized Placebo-Controlled Trials
Analyses Of Randomized Placebo-Controlled Trials
Isaac Nathan Smullin Johnson
Follow this and additional works at: https://elischolar.library.yale.edu/ymtdl
Recommended Citation
Recommended Citation
Johnson, Isaac Nathan Smullin, “Antidepressant Treatment Of Major Depressive Disorder In Patients With
Comorbid Alcohol Use Disorder: Two Meta-Analyses Of Randomized Placebo-Controlled Trials” (2020).
Yale Medicine Thesis Digital Library. 3917.
https://elischolar.library.yale.edu/ymtdl/3917
This Open Access Thesis is brought to you for free and open access by the School of Medicine at EliScholar – A
Digital Platform for Scholarly Publishing at Yale. It has been accepted for inclusion in Yale Medicine Thesis Digital
Library by an authorized administrator of EliScholar – A Digital Platform for Scholarly Publishing at Yale. For more
information, please contact elischolar@yale.edu.
Antidepressant Treatment of Major Depressive Disorder in Patients with
Comorbid Alcohol Use Disorder: Two Meta-analyses of Randomized
Placebo-controlled Trials

A Thesis Submitted to the
Yale University School of Medicine
in Partial Fulfillment of the Requirements for the
Degree of Doctor of Medicine

by

Isaac Nathan Smullin Johnson
Yale School of Medicine Class of 2020

Table of Contents

Dedication and Acknowledgements………………………………………………………….3

Abstract……………………………………………………………………………………….4

Introduction……………………………………………………………………………………7

Methods for Aim 1……………………………………………………………………………11

Results for Aim 1…………………………………………………………………………….13

Methods for Aim 2……………………………………………………………………………19

Results for Aim 2…………………………………………………………………………….22

Discussion……………………………………………………………………………………27

Tables and Figures for Aim 1…………………………………………………………………32

Tables and Figures for Aim 2…………………………………………………………………44

References……………………………………………………………………………………56

Dedication and Acknowledgements:

I am thankful for the loving support that I have received from my mother Leslie
Bourne, my father Mark Johnson, and my brother Jacob Johnson.
In loving memory of my grandparents Samuel Smullin, Frances Smullin, Paul
Johnson, and Ruth Johnson.

I could not have asked for better mentorship than I have received throughout
medical school. I am tremendously grateful for the mentorship and guidance I
have received in life and in research from my thesis advisor Dr. Michael Bloch
and his wife Dr. Angeli Landeros-Weisenberger. They have been an ever-
present source of inspiration, support, advice, and humor over the course of my
5 years in medical school.

I am also grateful for the mentorship I have received from Dr. Robert
Rohrbaugh, Dr. Andrés Martin, Dr. James Leckman, Dr. Zheala Qayyum, Dr.
Brian Fuehrlein, Dr. Linda Mayes, Dr. Kirsten Wilkins, Dr. Karen Jubanyik, Dr.
Euripedes Miguel, Dr. Marcelo Hoexter, Dr. Yukiko Kano, Dr. Yu Hamamoto,
Dr. Emeric Bojarski, Dr. Eunice Yuen, Dr. João Paulo De Aquino and
numerous additional residents, fellows, and faculty who have inspired me with
their kindness and generosity.

My work is built upon the sacrifice of my family and my mentors.

Thesis advisor: Dr. Michael H. Bloch, Yale Child Study Center

Authors who contributed to this thesis:
Aim 1: Isaac N.S. Johnson, Bridget J. Shovestul,
Mark J. Niciu, Fenghua Li, and Michael H. Bloch
Aim 2: Jason I. Dailey, Bachaar Arnaout,
Isaac N.S. Johnson, Jessica A. Johnson, Megan McNivens,
and Michael H. Bloch
Research reported in this publication was supported by the National Institute on Alcohol Abuse and
Alcoholism of the National Institutes of Health under Award Number T35AA023760. The content is
solely the responsibility of the authors and does not necessarily represent the official views of the
National Institutes of Health. This publication was also made possible by the Yale School of Medicine
Medical Student Research Fellowship.

Abstract

Objective:
Aim 1: To examine the efficacy of antidepressant agents compared with placebo in reducing
depressive symptoms in subjects with comorbid Alcohol Use Disorders (AUD).
Aim 2: To examine the efficacy of antidepressant agents compared with placebo on measures
of alcohol consumption.
Data Sources:
Aim 1: PubMed was searched for randomized, placebo-controlled trials that examined the
efficacy of antidepressant medications for treating depression symptoms with comorbid
AUD.
Aim 2: Ovid MEDLINE (1946 to September 23, 2016) and CENTRAL (Issue 8, August
2016) were searched with no language limits for randomized placebo-controlled trials that
examined the effects of antidepressant medications on alcohol consumption.
Study Selection:
Aim 1: Trials were included if they: 1) were randomized, placebo-controlled clinical trials, 2)
examined the effects of an antidepressant medication for comorbid MDD and AUD, and 3)
reported depression outcomes.
Aim 2: Trials were included if they: 1) were randomized, placebo-controlled clinical trials, 2)
examined the effects of an antidepressant medication for comorbid MDD and AUD, and 3)
reported alcohol consumption outcomes.
Data Extraction:
Aim 1: Random effects meta-analysis was utilized to examine standardized mean difference
(SMD) in improvement of depressive symptoms and risk ratio for treatment response.
Stratified subgroup analysis was used to examine the moderating effects of type of
antidepressant medication and other trial characteristics.
Aim 2: We examined the effect of antidepressant treatment on four alcohol consumption
outcomes: (1) drinking days, (2) drinks per day, (3) hazardous drinking days, and (4)
abstinence rates. Our primary outcome was standardized mean difference for continuous
measures and risk ratio for dichotomous outcomes using random effects meta-analysis. We
also used stratified subgroup analysis to examine the moderating effects of type of
antidepressant medication and diagnostic indication.
Results:
Aim 1: Eighteen distinct trial arms involving 1,318 participants were included in this
systematic review and meta-analysis. In subjects with AUD, antidepressant medications
significantly decreased depression severity compared with placebo (SMD=0.33±0.10 (95%
Confidence Interval (CI): 0.14-0.51, k=18, z=3.4, p=0.001). Type of antidepressant
medication did not significantly affect the magnitude of depressive symptom improvement
compared with placebo (Test for subgroup differences χ2=2.15, df=2, p=0.34). TCAs
(SMD=0.51±0.19 (95% CI: 0.15-0.88, k=3, z=2.7, p=0.006) and SSRIs (SMD=0.22±0.12
(95% CI: -0.01-0.46, k=10, z=1.9, p=0.06) suggested similar benefits for depressive
symptoms in subjects with comorbid AUD. The use of concomitant psychotherapy (for either
depression or alcohol use) (Test for subgroup differences χ2=9.9, df=1, p=0.002) or
concomitant pharmacotherapy for AUD (Test for subgroup differences χ2=4.7, df=1, p=0.03)
was associated with a significantly smaller measured treatment benefit of antidepressant
agents.
Aim 2: Twenty-six trials involving 2,771 participants were included in this systematic review
and meta-analysis. Overall, antidepressant use was not associated with significant changes in
drinking outcomes (drinking days, drinks per day, abstinence rates, and hazardous drinking
days). When antidepressants were utilized to treat comorbid depression symptoms,
antidepressant treatment was associated with improved drinking outcomes on some (drinking
days and drinks per day) but not all measures (abstinence rates and hazardous drinking days).
When antidepressants were utilized primarily to treat symptoms of other disorders,
antidepressant treatment was associated with worsened drinking outcomes on some (drinking
days and drinks per day) but not all measures (abstinence rates and hazardous drinking days).
Class of antidepressant treatment did not significantly affect any drinking-related outcomes.
Conclusion:
Aim 1: Our meta-analysis suggests that antidepressant medications significantly decrease
depressive symptoms in participants with comorbid AUD. The magnitude of depressive
symptom improvement in subjects with comorbid AUD appears similar to that achieved in
MDD trials without comorbid substance use.
Aim 2: Antidepressant therapy results in improvement in some drinking outcomes when used
for comorbid depression, though it may worsen these outcomes in the absence of comorbid
depression. More research is needed on the impact of antidepressants on drinking outcomes,
including the potential moderating effects of age, genotype, and depression and anxiety
symptoms.

Introduction
Major Depressive Disorder (MDD) and Alcohol Use Disorder (AUD) are among the
most prevalent mental health conditions in adult populations. AUD is overrepresented in
adults with MDD compared with the general population, and depression is overrepresented in
patients with AUD.1 Recent genetic analysis supports a strong genetic overlap between MDD
and AUD.2 Patients with MDD and comorbid AUD tend to experience greater depression
severity, depressive symptoms at an earlier age of onset, increased suicidality and functional
impairment, higher rates of relapse and decreased likelihood of recovery from depressive
symptoms.3-7 In patients with AUD, comorbid depressive symptoms are associated with an
increased likelihood of treatment dropout and relapse.8-10
Pharmacotherapy with antidepressant medications is a first-line treatment for MDD.
In meta-analysis, antidepressant agents demonstrate a significant benefit compared with
placebo for the treatment of major depression with effect sizes of 0.3011 and 0.3712 reported
in the literature and a NNT of 6.13 Despite the high rate of comorbidity between AUD and
MDD, subjects who meet criteria for current or recent alcohol or other substance use
disorders are typically excluded from these pivotal randomized, placebo-controlled trials of
antidepressant medications. Thus, it is uncertain how well the results of positive
antidepressant trials in non-alcohol dependent patients will generalize to clinical MDD
populations, where patients often have comorbid AUD.14-20
Previous meta-analyses have found mixed results regarding the efficacy of
antidepressants in treating comorbid MDD and AUD. A 2004 meta-analysis found that
antidepressants have a “modest beneficial effect” in reducing depressive symptoms in
patients with a comorbid substance use disorder (not limited to AUD).21 A subsequent meta-
analysis suggested a similar effect when meta-analysis was confined to just trials involving
subjects with comorbid AUD and MDD. This meta-analysis further reported that Selective
Serotonin Reuptake Inhibitors (SSRIs), as a class, were not associated with an increased
likelihood of response in terms of depressive symptoms, compared to placebo.22
A more recent meta-analysis published in 2011 that similarly examined only
treatment response, demonstrated that antidepressant agents overall were more effective than
placebo at reducing depressive symptoms in patients with comorbid AUD. However, this
meta-analysis was not able to demonstrate that SSRIs, as a class, were effective in this
population and further suggested that they were less effective than other antidepressants.13
These previous meta-analyses examined only treatment response and not continuous
outcomes. Also, there are additional recent trials with second generation antidepressants that
have been published subsequent to these previous reviews.
Alcohol use disorder has a lifetime prevalence of 30.3% in the United States
according to results from the National Epidemiologic Survey on Alcohol and Related
Conditions (NESARC).23 Comorbid alcohol dependence and depression result in 44% more
healthcare costs compared to treating depression alone.24 It has been suggested that alcohol
use disorder may be causally linked to increased rates of depression,25 though genetic
variations in serotonin transporter (SERT) function have also been implicated in both
disorders.26 While second- and third-generation antidepressant medications remain the
mainstay for treating depression,27 they can also be used in many other psychiatric conditions
that are often comorbid with alcohol use disorder. For example, Americans with alcohol
dependence are three times more likely to have an anxiety disorder and more than five times
more likely to have nicotine dependence.23 Moreover, antidepressants were at one time the
most commonly prescribed medication for alcohol use disorder,28 though several reviews and
meta-analyses have questioned their efficacy for this indication.22, 29, 30 Unfortunately, there
has been some case report31 and clinical trial32-35 evidence that the SSRI antidepressants may
actually increase alcohol consumption in a subset of the population. This would suggest that
providers should consider a different class of antidepressants for patients prone to alcohol use
disorder. Of course, this assumes that classes of antidepressants other than SSRIs have
superior outcomes.
Two recent meta-analyses demonstrated the efficacy of antidepressants for treating
depressive disorders in patients with comorbid alcohol use disorder, but either did not report36
or reported only very limited data13 on alcohol consumption outcomes. Notably, a 2004 meta-
analysis21 studied antidepressant effects on both depression and substance use outcomes in
the treatment of depressive disorders with comorbid dependence on alcohol or illicit drugs,
and demonstrated improvement in substance use outcomes in the subset of studies in which
depressive symptoms improved. Similarly, a 2005 meta-analysis22 examined alcohol and
illicit drug outcomes in both studies of comorbid depression and studies without comorbid
depression, but only found a statistically significant effect on substance use outcomes for
first-generation antidepressants, in the treatment of comorbid depression and alcohol use
disorder.

Statement of Purpose
The goals of the current meta-analyses are to update previous meta-analyses, as well
as, to examine several unanswered questions regarding the use of antidepressant agents in
subjects with MDD and comorbid AUD. In Aim 1, we specifically sought to determine: (1)
What is the measured effect size and relative risk of response for subjects with MDD and AUD
treated with antidepressant agents compared with placebo?; (2) Do different medication classes
(TCA vs. SSRI) have the same measured benefit compared with placebo for subjects with
MDD and AUD?; (3) Does the use of concomitant psychotherapy, targeting either depression
or alcohol use, or concomitant pharmacotherapy for AUD moderate the benefits of
antidepressant agents in the treatment of MDD with comorbid AUD?; and (4) Does the
measured efficacy of antidepressant agents differ in trials where antidepressants are initiated
before or after alcohol detoxification is completed?
The systematic review and meta-analysis conducted as part of Aim 2 has the goal of
updating and expanding upon previous studies by systematically analyzing trials that study
the effects of second- and third-generation antidepressants on alcohol consumption outcomes,
regardless of indication for the medication. This aim attempts to better delineate the effects of
different antidepressant medication classes, specifically on alcohol consumption.

Author Contributions
IJ’s role in Aim 1 was in writing the manuscript, conceptual planning, identifying
articles for quantitative synthesis (meta-analysis) through assessment of articles by title,
abstract, and full-text based on inclusion and exclusion criteria, extraction and organization of
the data from each included article, creation of Table 1 and Figure 1, conducting a review of
the literature, and reviewing previous systematic reviews in this area. IJ’s role in Aim 2 was
in writing and organizing sections of the manuscript and conducting a review of the literature.
MHB supervised the conceptual planning and execution of both Aim 1 and Aim 2. He
also wrote and organized sections of both manuscripts and performed the statistical analyses.
JD’s role in Aim 2 was in writing the manuscript, conceptual planning, identifying
articles for quantitative synthesis (meta-analysis) through assessment of articles by title,
abstract, and full-text based on inclusion and exclusion criteria, extraction and organization of
data, conducting a review of the literature, and reviewing previous systematic reviews in this
area.
MN reviewed the manuscript for Aim 1 and BA reviewed the manuscript for Aim 2.
FL created all forest and funnel plot figures for both manuscripts. BS, JJ, and MM
contributed to writing sections of the manuscript. BS created Table 2 for Aim 1.
Aim 1: Effect of Antidepressants on Depression Outcomes
Methods:
Literature Search
We aimed to identify all randomized, placebo-controlled clinical trials of
antidepressants indicated for the treatment of either MDD or chronic dysthymic disorder in
participants with comorbid alcohol use disorder, that reported depression outcomes. PubMed
MEDLINE (1946 to 12/18/2017) was searched using the search strategy “Alcohol-Related
Disorders”[Mesh] AND “Depressive Disorder”[Mesh]) AND (“Antidepressive
Agents”[Mesh] OR “Antidepressive Agents”[Pharmacological Action]).” We further limited
the search using the clinical trials filter. There were no language limitations on the search.
We additionally examined the references of included trials and previous systematic reviews
in the area to identify additional citations.13, 21, 22, 37

Study Selection
Following the removal of duplicate citations, abstracts were independently screened
by two authors (IJ and BS) for full text review, according to the following inclusion criteria:
1) randomized, placebo-controlled clinical trial, 2) utilization of an antidepressant medication
to treat patients with AUD comorbid with either MDD or chronic dysthymic disorder, and 3)
trials where depression outcomes were reported. Following this screening, full text articles
were then reviewed by the same two authors according to the same inclusion criteria.
Disagreements in both screening and full text review phases were resolved by consensus
agreement in consultation with a senior reviewer (MHB). We excluded articles conducted in
adolescents and articles that included patients who did not meet criteria for a current
diagnosis of MDD or dysthymia. We included studies with concomitant use of naltrexone in
the treatment and placebo arms despite evidence in previous literature of the confounding
effects of this medication on alcohol consumption outcomes.38 This is due to the fact that the
present study focused exclusively on depression-related outcomes, and we conducted a
stratified subgroup analysis to examine the moderating effects of concomitant
pharmacotherapy.

Data Extraction
Custom designed Microsoft Excel spreadsheets were used to extract data. Data
extracted from the identified trials included: bibliographic information; indication for the
trial; antidepressant medication studied; maximum medication dose; duration of study; age
and gender of subjects; number of participants (n) in intention-to-treat sample and number
completing the trial; concomitant psychotherapeutic interventions and whether the
psychotherapy specifically targeted depression or alcohol use; concomitant
psychopharmacological interventions targeting alcohol use; and whether or not participants
were required to stop drinking prior to the start of the study. The primary outcomes extracted
for the meta-analysis included endpoint depressive symptom ratings and response in terms of
depressive symptoms, for both active and placebo arms of each study. If trial outcomes were
only reported in graphical form, a graph digitizer program called GetData39 was used to
extract the data. When no aggregate data was available, studies that reported data of multiple
treatment arms or by moderator subtype only were treated as separate studies for each study
moderator.

Statistical Analysis
Statistical analysis was performed using Comprehensive Meta-Analysis (version
3.0).40 Given the variation of depression rating scales reported in the included studies,
standardized mean difference was chosen as the summary statistic. For treatment response, a
dichotomous outcome, risk ratio was utilized as the summary statistic. Given the
heterogeneity in medications, trial design, and study outcomes, we chose to use a random-
effects model. Heterogeneity was assessed by calculating the Q statistic,41 and the I2. I2 is a
transformation of the Q statistic that indicates the proportion of observed variance that can be
attributed to heterogeneity rather than sampling error.42 Publication bias was assessed by
creating funnel plots for the outcome measures by plotting the effect size against standard
error for each included trial. In addition, publication bias was statistically tested by the Egger
test.43 We conducted stratified subgroup analyses to examine the effects of type of
antidepressant medication (SSRI, TCA, vs. other), whether concomitant psychotherapy was
administered and what it was indicated for, whether concomitant medication targeting AUD
was administered, and whether detoxification was performed before initiation of
antidepressant study medication. We used a mixed-effects meta-regression to examine the
effects of participant age and duration of treatment on the measured benefits of antidepressant
agents.

Results
Selection of Studies
Figure 1 depicts our procedure for selection of studies in accordance with the Preferred
Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.44 Our
initial search identified 75 citations, of which 16 clinical trials were eligible for inclusion in
this meta-analysis.
Table 1 describes the characteristics of the 18 distinct trial arms from 16 trials
included in this systematic review and meta-analysis, involving 1,318 participants.29, 38, 45-58
Ten trials examined SSRI medications, 3 trials examined TCA medications, and 5 studies
were of other antidepressants including mirtazapine, nefazodone, mianserin, and viloxazine.
Either MDD or chronic dysthymic disorder was an indication for antidepressant treatment in
each study. In 7 trials, subjects participated in a detoxification period prior to the initiation of
antidepressant therapy. In 12 trials, participants received concomitant psychotherapy, in
addition to treatment with the study medication. Concomitant psychotherapy targeted alcohol
use in 9 studies and additionally depression in 4 studies. Three trials utilized concomitant
pharmacotherapy, such as naltrexone, to target alcohol use. Eighteen trial arms reported
depression outcomes as continuous symptom improvement, while 12 trial arms included
response data in terms of depressive symptoms. The majority of included studies reported
depression outcomes with the Hamilton Rating Scale for Depression (HAM-D). However, the
Beck Depression Inventory (BDI), the Montgomery-Asberg Depression Rating Scale
(MADRS), and the Lehmann-Rockliff Depression Rating Scale were utilized by one study
each. One trial59 met our inclusion criteria, but did not contribute usable data to our study.
Table 2 examines the risk of bias in each included trial.

INSERT FIGURE 1 OF AIM 1 HERE
INSERT TABLE 1 AND 2 OF AIM 1 HERE

Depression Severity
In participants with AUD, antidepressant medications significantly decreased
depression severity, as compared with placebo (Standardized Mean Difference
(SMD)=0.33±0.10 (95% Confidence Interval (CI): 0.14-0.51, k=18, z=3.4, p=0.001). There
was significant heterogeneity between studies (χ2=42.3, df=17, p=0.001, I2=59.8%). Figure
2A depicts the effects of antidepressant use compared with placebo on depression severity
overall and stratified by medication type. Compared with placebo, antidepressant types did
not demonstrate significant differences in reduction of depression severity, when trials were
stratified by medication type (Test for subgroup differences χ2=2.15, df=2, p=0.34). Selective
Serotonin Reuptake Inhibitors (SSRIs) (SMD=0.22±0.12 (95% CI: -0.01-0.46, k=10, z=1.9,
p=0.06), Tricyclic Antidepressants (TCAs) (SMD=0.51±0.19 (95% CI: 0.15-0.88, k=3,
z=2.7, p=0.006), and other antidepressants (SMD=0.51±0.30 (95% CI: -0.07-1.09, k=5,
z=1.7, p=0.084) demonstrated similar improvements in depression severity compared with
placebo. Figure 3A demonstrates funnel plot asymmetry suggestive of publication bias
(Egger’s test p=0.015).

INSERT FIGURES 2 AND 3 OF AIM 1 HERE

In stratified subgroup analysis, trials in which subjects participated in a detoxification
period prior to initiating antidepressant treatment (SMD=0.54±0.17 (95% CI: 0.20-0.88, k=7,
z=3.1, p=0.002) demonstrated a similar measured benefit (Test for subgroup differences
χ2=2.7, df=1, p=0.10) of antidepressant treatment compared with trials in which a
detoxification period did not occur prior to initiation of an antidepressant medication
(SMD=0.20±0.12 (95% CI: -0.03-0.42, k=11, z=1.7, p=0.09).
Figure 4A depicts the effect of antidepressant treatment relative to placebo, stratified
by whether or not participants received concomitant psychotherapy during the course of
antidepressant treatment. In stratified subgroup analysis, trials in which participants did not
receive concomitant psychotherapy (SMD=0.90±0.25 (95% CI: 0.42-1.39, k=6, z=3.66,
p<0.001) demonstrated a significantly greater measured benefit of antidepressant treatment, compared with placebo, than in trials in which participants received concomitant psychotherapy (SMD=0.10±0.07 (95% CI: -0.03-0.23, k=12, z=1.46, p=0.15; Test for subgroup differences χ2=9.9, df=1, p=0.002). The measured benefit of antidepressant treatment, compared with placebo, was not significantly different when studies were stratified by whether participants did (SMD=0.24±0.13 (95% CI: -0.02-0.49, k=4, z=1.81, p=0.07) or did not (SMD=0.38±0.13 (95% CI: 0.12-0.63, k=14, z=2.85, p=0.004) receive psychotherapy for depression (Test for subgroup differences χ2=0.58, df=1, p=0.45). The measured benefit of antidepressant treatment was significantly different when studies were stratified by whether participants did (SMD=0.12±0.09 (95% CI:-0.05-0.30, k=9, z=1.37, p=0.17) or did not (SMD=0.57±0.18 (95% CI: 0.21-0.93, k=9, z=3.07, p=0.002) receive psychotherapy for alcohol use (Test for subgroup differences χ2=4.7, df=1, p=0.03). There was a significant difference in the measured benefit of antidepressant medications based on whether trials used concomitant pharmacotherapy, such as naltrexone, for alcohol use disorder (Test for subgroup differences χ2=5.1, df=1, p=0.024). Our analysis demonstrated a smaller measured benefit of antidepressant treatment when trials were stratified by whether concomitant pharmacotherapy was initiated to target alcohol use (SMD=-0.00±0.13 (95% CI: -0.26-0.26, k=3, z=-0.0, p=0.99) compared with when it was not initiated (SMD=0.40±0.12 (95% CI: 0.17-0.63, k=15, z=3.4, p=0.001) in the trial. INSERT FIGURE 4 OF AIM 1 HERE In meta-regression, participant age (β=-0.02±0.05, 95% CI: -0.11-0.08, k=16, z=-0.32, p=0.75) and duration of antidepressant treatment (β=0.001±0.02, 95% CI: -0.045-0.048, k=16, z=0.06, p=0.96) were not associated with a measured reduction in depression severity following treatment with antidepressants vs. placebo. Depression Response Rate Participants with AUD demonstrated a significantly greater likelihood of response to antidepressant medications than they did to placebo (Risk ratio (RR)=1.30 (95% Confidence Interval (CI): 1.07-1.58, k=12, z=2.6, p=0.009). There was significant heterogeneity between studies (χ2=33.3, df=11, p<0.001, I2=67%). Figure 2B depicts the response to antidepressant use compared with placebo overall and stratified by medication type. Compared with placebo, antidepressant types did not demonstrate a significant difference in response, when trials were stratified by medication type. However, there was a trend toward marginally increased response to TCAs (RR=1.97 (95% CI: 1.32-2.96, k=3, z=3.3, p=0.001), as opposed to SSRIs (RR=1.14 (95% CI: 0.91-1.43, k=7, z=1.1, p=0.27) and other antidepressants (RR=1.36 (95% CI: 0.29-6.35, k=2, z=0.39, p=0.7), when compared with placebo (Test for subgroup differences χ2=5.4, df=2, p=0.07). Figure 3B demonstrates funnel plot asymmetry suggestive of publication bias, despite lack of evidence of publication bias by Egger’s test (Egger’s test p=0.12). In stratified subgroup analysis, trials in which subjects participated in a detoxification period prior to initiating antidepressant treatment (RR=2.77 (95% CI: 1.12-6.84, k=6, z=2.2, p=0.028) demonstrated a similar likelihood of response to antidepressant treatment compared with trials in which a detoxification period did not occur prior to initiation of an antidepressant medication (RR=1.47 (95% CI: 0.71-3.02, k=6, z=1.0, p=0.30); Test for subgroup differences χ2=1.15, df=1, p=0.28). Figure 4B depicts the response to antidepressant treatment compared with placebo, stratified by whether or not participants received concomitant psychotherapy during the course of antidepressant treatment. In stratified subgroup analysis, trials in which participants did not receive concomitant psychotherapy (RR=1.94 (95% CI: 1.17-3.23, k=4, z=2.6, p=0.01) demonstrated a similar likelihood of response to antidepressant treatment vs. placebo, compared with trials in which participants received concomitant psychotherapy (RR=1.13 (95% CI: 0.89-1.43, k=8, z=1.0, p=0.31). There was, however, a trend toward marginally increased response to antidepressants vs. placebo among patients who did not receive concomitant psychotherapy (Test for subgroup differences χ2=3.6, df=1, p=0.06). The likelihood of response to antidepressant treatment, compared with placebo, was not significantly different when studies were stratified by whether participants did (RR=1.22 (95% CI: 0.95-1.56, k=4, z=1.54, p=0.12) or did not (RR=1.36 (95% CI: 0.95-1.94, k=8, z=1.68, p=0.09) receive psychotherapy for depression (Test for subgroup differences χ2=0.24, df=1, p=0.62) and also when they were stratified by whether participants did (RR=1.20 (95% CI:0.95-1.51, k=6, z=1.50, p=0.13) or did not (RR=1.47 (95% CI: 0.97-2.23, k=6, z=1.80, p=0.07) receive psychotherapy for alcohol use (Test for subgroup differences χ2=0.71, df=1, p=0.40). Similarly, there was no statistically significant difference in the likelihood of response to antidepressant treatment when studies were stratified by whether concomitant pharmacotherapy was initiated to target alcohol use (RR=1.00 (95% CI: 0.55- 1.80, k=2, z=0.0, p=0.99) or was not (RR=1.37 (95% CI: 1.05-1.79, k=10, z=2.3, p=0.02) in the trial (Test for subgroup differences χ2=0.93, df=1, p=0.34). In meta-regression, participant age (β=-0.039, 95% CI: -0.15-0.07, k=10, z=-0.73, p=0.47) and duration of antidepressant treatment (β=0.019, 95% CI: -0.03-0.07, k=10, z=0.77, p=0.44) were not associated with differences in the likelihood of depression response following treatment with antidepressants vs. placebo. Aim 2: Effect of Antidepressants on Alcohol Consumption Outcomes Methods: Literature Search We aimed to identify all randomized placebo-controlled clinical trials of antidepressants that reported alcohol consumption outcomes, regardless of the indication for antidepressant use. Literature search was conducted using Ovid MEDLINE (1946 to September 23, 2016) and the Cochrane Central Register of Controlled Trials (Issue 8, August 2016) with no language restrictions. Search terms used included a combination of alcohol use disorder, alcohol dependence, or alcoholism; and antidepressive agents, serotonin uptake inhibitors, serotonin and noradrenaline reuptake inhibitors, dopamine uptake inhibitors, bupropion, mirtazapine, fluoxetine, fluvoxamine, sertraline, paroxetine, citalopram, escitalopram, venlafaxine, desvenlafaxine, duloxetine, desipramine, imipramine, nefazodone, or viloxazine. Results were then limited to clinical trials. We obtained the primary articles associated with conference abstracts and secondary analyses resulting from our literature search where possible and attempted to identify additional studies via review of references and through consultation with two experts familiar with the published literature in this field. Study Selection Following removal of duplicates, abstracts were independently screened by two authors (JD and BA) for full-text review according to the follow inclusion criteria: 1) randomized placebo-controlled clinical trial, 2) of an antidepressant medication for any indication, 3) with alcohol consumption outcomes reported. Following this screening, full text articles were then reviewed by the same two authors, according to the same inclusion criteria. Disagreements in both screening and full text review phases were resolved by consensus agreement. We excluded articles conducted in an inpatient or human lab setting, and those conducted with adolescents. We also excluded studies with concomitant use of naltrexone in the treatment and placebo arms, due to evidence in previous literature of the confounding effects of this medication on alcohol consumption outcomes.38 Prior to data analysis, we also decided to exclude studies of first-generation antidepressants given their limited use in modern clinical practice and that few trials exist, and the fact that we had a sufficient number of articles utilizing second and third-generation antidepressants for statistical analysis. Data Extraction Data was extracted onto specially designed Microsoft Excel spreadsheets. Background data extracted from the identified trials included: bibliographic information; indication for the trial; antidepressant medication studied; maximum medication dose; duration of study; concomitant psychosocial interventions; age and gender of subjects; moderators of early-onset alcohol use disorder, family history, or genotype; number of participants (n) in intention-to-treat sample and of those completing the trial; and whether or not participants were required to stop drinking prior to the start of the study. Alcohol consumption outcome measures that were extracted included: number of drinks (drinks per drinking day, average drinks per day, or percent change of either of these variables), number or proportion of drinking days, number or proportion of hazardous drinking days, and number of abstinent subjects throughout or at the end of the study. Depression and anxiety outcomes were also extracted, where available. If trial outcomes were only reported in graphical form, a graph digitizer program called GetData39 was used to extract the data. Studies that reported data of multiple treatment arms or by moderator subtype only were treated as separate studies for each study moderator, when no aggregate data was available. In long-term studies, where data was reported at multiple time points, only data closest to a 12-week follow-up period was extracted, in order to maintain consistency with other included studies. Statistical Analysis Statistical analysis was performed using Comprehensive Meta-Analysis (version 3.0).60 Given the variation of alcohol consumption outcomes reported in the included studies for continuous outcomes -- number of drinking days, number of hazardous drinking days and drinks per day, standardized mean difference was chosen as the summary statistic. For abstinence, a dichotomous outcome, risk ratio was utilized as the summary statistic. Given the heterogeneity in medications, trial design and study outcomes, we chose to use a random- effects model. Heterogeneity was assessed by calculating the Q statistic41, and the I2, a transformation of Q that indicates the proportion of observed variance that can be attributed to heterogeneity, rather than sampling error.42 Publication bias was assessed by creating funnel plots for the outcome measures by plotting the effect size against standard error for each included trial. In addition, publication bias was statistically tested by the Egger test.43 We conducted stratified subgroup analyses to examine the effects of type of antidepressant medication (SSRI, SNRI vs other), indication for antidepressant treatment (depression vs. other) and whether detoxification was performed before initiation of the antidepressant study medication. Results Selection of Studies Figure 1 depicts our procedure for selection of studies in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.44 Our initial search identified 354 citations, of which 94 were reviewed in full-text for eligibility. Of the 94 studies eligible for review, 18 were found to be secondary analyses or otherwise duplicate reports of trials already assessed, 39 did not meet the inclusion criteria outlined above, and 11 met the exclusion criteria above. Reasons for not meeting inclusion criteria included the study not being a randomized placebo-controlled clinical trial (21 studies), not being a study of an antidepressant (2 studies), or not reporting results of one of the alcohol consumption outcomes outlined above (16 studies). We excluded studies of adolescents (2 studies) and those that included naltrexone in both the treatment and placebo arms (4 studies). After the literature search, but prior to data analysis we also excluded trials of first-generation antidepressants (5 studies), as discussed above. All categories are exclusive of each other, with excluded studies categorized per the order presented above. INSERT FIGURE 1 OF AIM 2 HERE Table 1 describes the characteristics of the 26 studies included in this systematic review, involving 2,771 participants.29, 32, 33, 48-50, 55, 57, 60-77 Seventeen studies were of an SSRI, 1 study was of an SNRI, and 8 studies were of other third-generation antidepressants including bupropion, mirtazapine, and nefazodone. Seven of the 26 included studies had a depressive disorder as a primary indication for the study and were limited to the medications: fluoxetine, sertraline, mirtazapine, and nefazodone. Participants were abstinent from alcohol at the start of the trial (antidepressant initiation) in 15 of the 26 studies, and all but 5 studies included some type of concomitant psychotherapy. There was considerable variation in which alcohol consumption outcomes were reported, with the most common outcome being proportion of drinking days per month (DD), followed by drinks per drinking day (DDD) or average drinks per day (ADD) when DDD was not available, abstinence either at the study’s end or throughout the study, and finally proportion of hazardous drinking days per month (HDD). Only 3 studies included all 4 outcome variables of interest, while 35% of the studies included at least 3 of these outcomes. INSERT TABLE 1 OF AIM 2 HERE Drinking Days Overall, the use of antidepressant medications was not associated with any difference in the number of drinking days compared to placebo (standardized mean difference (SMD) =0.05 ± 0.06 (95% Confidence Interval (CI): 0.06-0.16, k=24, z=0.9, p=0.39). There was modest but statistically significant heterogeneity between studies (Q-statistic=40.0, df=23, p=0.02, I2=26%). Figure 2A depicts the effects of antidepressant use compared to placebo on number of drinking days stratified by medication type. SSRI, SNRI and other antidepressant medications did not have significantly different effects on number of drinking days compared to placebo (Test for subgroup differences χ2=2.1, df=2, p=0.34). Figure 3A depicts the effects of antidepressant use compared to placebo on number of drinking days stratified by diagnostic indication (depression vs. other). Trials prescribing antidepressants to treat depression (SMD=-0.28 ± 0.12 (95% CI:-0.52-0.04, k=5, z=-2.3, p=0.02) demonstrated a greater reduction in number of drinking days with antidepressant treatment compared to those trials which utilized antidepressants for other indications (SMD=0.14 ± 0.07 (95% CI:0.01- 0.27, k=19, z=2.2, p=0.03; test for subgroup differences χ2=9.3, df=1, p=0.002). Antidepressants had no effect on number of drinking days regardless of whether trials started subjects on an antidepressant medication after a detoxication period or not (test for subgroup differences χ2=0.2, df=1, p=0.87). Antidepressant agents, compared to placebo, had no effect on number of drinking days in trials with a detoxification period (SMD=-0.06 ± 0.09 (95% CI:-0.11-0.23, k=15, z=-0.7, p=0.51) or without a detoxification period prior to initiating antidepressant treatment (SMD=-0.04 ± 0.11 (95% CI:-0.18-0.25, k=9, z=-0.3, p=0.75). There was no funnel plot asymmetry suggestive of publication bias and the Egger’s test was also not statistically significant (p=0.11). INSERT FIGURE 2 OF AIM 2 HERE INSERT FIGURE 3 OF AIM 2 HERE Drinks Per Day Overall, the use of antidepressant medications was not associated with any difference in the number of drinks per day compared to placebo (SMD=0.07 ± 0.06 (95% CI: -0.04- 0.18, k=17, z=1.27, p=0.21). There was modest but statistically significant heterogeneity between studies (Q-statistic=24.8, df= 16, I2=36%). Figure 2B depicts the effects of antidepressant use compared to placebo on number of drinks per day stratified by medication type. SSRI, SNRI and other antidepressant medications did not have significantly different effects on number of drinks per day compared to placebo (test for subgroup differences χ 2= 0.15, df=1, p=0.69). Figure 3B depicts the effects of antidepressant use compared to placebo on number of drinks per day stratified by diagnostic indication (depression vs. other). Trials prescribing antidepressants to treat depression (SMD=-0.45 ± 0.15 (95% CI: -074-(-)0.16, k=4, z=-3.02, p=0.003) demonstrated a greater reduction in number of drinks per day with antidepressant treatment compared to those trials which utilized antidepressants for other