Introduction

Substance use disorders (SUDs) and psychiatric disorders are linked to relapse, financial and social problems [1], and high medical costs [2]. SUDs co-occur in roughly 32% of people with mood disorders. 16.5% of people with mood disorders have an alcohol use disorder, and 18.5% have a SUD [3]. SUDs, mood disorders, suicide, and impulsive and self-destructive acts are linked. Suicide is common in depressed patients who use alcohol or other SUDs [3]. A 12-year prospective study found that comorbid alcohol or other SUDs reduced the recovery rate of generalized anxiety disorder (GAD) by about fivefold (risk ratio = 0.20, p = 0.01) and increased the likelihood of relapse by over threefold (risk ratio = 3.09, p = 0.05) [4]. The most common co-occurring disorders among combat veterans and civilians with post-traumatic stress disorder (PTSD) were alcohol use, depression, anxiety, conduct disorder, and non-alcohol substance abuse or dependence [5].

Chronic alcohol, nicotine, and other SUDs reduce serotonin (SERT) levels in the brain. Studies on SUDs’ neurochemical changes have shown SERT’s role from initial exposure to tolerance, withdrawal, abstinence, or relapse [6]. Most addictive drugs involve SERT. For example, in the prefrontal cortex, cannabinoid receptor 1 (cb1) stimulation reduces the effect of citalopram. Selective reuptake inhibitor (SSRI) blockade of the cb1 receptor can increase SERT levels in the prefrontal cortex [7]. Nicotine’s mechanism of action is partly related to SERT (5-HT2C) receptor agonists. 5-HT2C agonists reduce drugs’ discriminative and reinforcing effects in the midbrain [8]. Studies suggest that alcohol causes low cerebrospinal fluid 5-hydroxyindoleacetic acid and low plasma tryptophan. Different opioid receptor subtypes regulate serotonergic neurotransmission in the central nervous system (CNS). For example, mu- and delta-opioid receptor agonists increase 5-HT efflux in the dorsal raphe nucleus (DRN) [9]. However, kappa agonists reduce extracellular 5-HT in the DRN, median raphe nucleus, and forebrain [10]. Among stimulants, SERT modulates cocaine’s locomotor action [11]. Cocaine and amphetamine are controlled by the balance of dopamine (DA) and SERT release [12].

Tolerability, adherence to treatment, cost-effectiveness, general safety in overdose, and a more comprehensive range of therapeutic activities make SSRIs one of the best pharmacotherapy approaches for people with mental illnesses and addiction. In this context, research on their therapeutic benefits is limited and inconclusive.

The term “antidepressants” was used in frequentist meta-analyses to look at how well they worked to treat depression in people with SUDs. A few of these studies focused on SSRIs’ efficacy. Torrens et al. [13] concluded that SSRIs have no clinical advantages over tricyclics in SUDs. Tricyclics and nefazodone were recommended by Iovieno et al. [14], but not SSRIs. Stokes et al. [15] found no difference in treatment effects between SSRIs and non-SSRIs in individuals with depression and SUDs. Two more studies supported SSRIs. Nunes and Levin [16] suggested that antidepressants exert a modest beneficial effect on individuals with depression and SUDs. A study published in 2021 found that SSRIs may be associated with a reduction in the risk of substance misuse compared to 1 month before treatment initiation [17]. There is conflicting data regarding improving clinical outcomes when the SERT system is modulated in patients with a psychiatric disorder and comorbid SUDs. Further investigation is essential to clarify the extent of the evidence and guide therapy. Therefore, a meta-analysis of randomized controlled trials (RCTs) was undertaken using a Bayesian approach to explore SSRIs’ efficacy in treating depression, anxiety, PTSD, and substance use in individuals with SUDs.

Methods

Criteria for considering studies

This meta-analysis follows PRISMA guidelines [18] and is registered under PROSPERO registration (CRD42020164944). RCTs comparing an SSRI to a placebo or an active drug in various clinical settings were included if they involved participants of various ages, genders, and races. Open trials and laboratory experiments were excluded. The outcome of interest was efficacy as measured by SSRIs’ ability to treat depression, anxiety, and PTSD (symptom reduction on clinician rating or self-report scales) and facilitate drug abstinence, manage craving, and reduce drug use.

Identification of studies

We searched the following electronic databases: Google Scholar, PubMed, Scopus, OVID MEDLINE, and Academic Search Complete from their inception to August 6, 2021. We identified all published RCTs in any language using medical subject headings and terms such as substance use, addictive drugs, drug use, SSRIs, depression, anxiety, craving, abstinence, and SSRIs and SUDs. Here is an example of a Uniform Resource Locator (URL) search strategy: https://PubMed.ncbi.nlm.nih.gov/?term=Selective+serotonin+reuptake+inhibitors+%28SSRI%29+depression+and+other+affective+disorders+in+patients+with+substance+use+disorder.

Data collection

Two independent assessors (DF and PM) screened titles and extracted data, while a third author (AJ) resolved conflicts between DF and PM. We extracted general information such as publication status, title, authors’ names, source, country, year of publication, duplicate publication, and trial characteristics.

Bias risk assessment

We assessed articles’ quality and risk of bias using the Revised Cochrane RCT risk of bias tool [19], focusing on randomization, deviations from intended interventions, missing outcome data, outcome measurement, and reporting results.

Data synthesis

We summarized pooled results as the standardized difference between two means, Cohen’s d (“small, d = .2,” “medium, d = .5,” and “large, d = .8”) [20]. Because the random-effects model assumes that between-study heterogeneity causes the actual effect sizes of studies to differ, we, therefore, included an estimated standard deviation of underlying effects across studies, tau (τ) [21].

Frequentist meta-analysis uses a fixed state of nature to interpret a parameter of interest, while Bayesian analysis allows an unknown quantity to have a probability distribution [22]. We set prior d to 0, width parameter to 0.707, truncated Cauchy because we were 50% confident that the effect size resided somewhere between d = 0 and d = 0 0.707 [23]. We compared how well the alternative hypothesis (HA= SSRIs) or the null hypothesis (H0 = control) predicted the data in the following relationship: Prior d of HA × BF (Bayes factor) = Posterior d of HA, where BF = [Probability (data, given HA)]/[Probability (data, given H0)] postulated the likelihood of evidence of a treatment effect in place of the p-value in frequentist meta-analysis. For this analysis, a BF = 3 = posterior probability (d > 0) = 75%. The larger numbers are essentially non-interpretable; they were stated as BF > 99. The modeling procedure was as follows: adaptation: 1000, burn-in: 5000, number of iterations: 1000, and number of chains: 3. Subgroup analysis was carried out on SSRIs to tease apart the one with the most robust performance. Meta-regression was undertaken to determine whether attrition rate (5–64%), SSRI dosage (citalopram 20–100; escitalopram 10–30; fluoxetine 20–80; fluvoxamine 100–300; paroxetine 20–60; sertraline 20–200), or treatment duration (4–52 weeks) influenced SSRI's efficacy over placebo. Meta-analysis was conducted in Jeffreys’s Amazing Statistics Program, JASP Team (2020). JASP (Version 0.14.1) [Computer software] (University of Amsterdam Nieuwe Achtergracht 129B, Amsterdam, The Netherlands).

Results

The search found 2,921 articles (Fig. 1). Two thousand six hundred sixty of these were deemed irrelevant. Two hundred sixty-one articles were eligible for screening, with 153 being rejected. One hundred eight publications were suitable for data collection, but 38 did not provide data appropriate for analysis. The meta-analysis included 70 of them. One article was omitted due to data duplication [88]. Two articles did not provide statistical data suitable for the analysis [89, 90], and three articles used data from previous RCTs [91,92,93]. Sixty-four RCTs [24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87] were included in the Bayesian meta-analysis. Sixty-one of the trials compared an SSRI with a placebo, and three of them used another medication as a control [78, 86, 87]. The investigation used a random sample of 6128 participants (Table 1) (Supp Table 2–15).

Fig. 1
figure 1

Flow diagram of included studies

Full size image
Table 1 Characteristics of included studies
Full size table

More than 80% of participants were male, and over 45% were Caucasian. Ages ranged from 12 to 75. Forty-four of the sixty-three trials were conducted in the USA. Around 30% were university-based, and 17% were outpatient. The treatment lasts from 4 to 52 weeks. We found fluoxetine, sertraline, paroxetine, citalopram, escitalopram, and fluvoxamine data. Fluoxetine (25) and sertraline (17) were the most studied. Twenty-nine trials had an attrition rate of 20% or more. Twelve trials had missing outcome data, while fifteen were deemed at risk of bias in the selection of the reported result.

Depression, anxiety, and PTSD

Carried out with a placebo arm, a reduction in depressive symptoms (d = 0.353, BF > 99) was demonstrated for fluoxetine, sertraline, paroxetine, citalopram, escitalopram, and fluvoxamine in opioid, alcohol, cocaine, marijuana, and nicotine use disorders. Fluoxetine showed the highest antidepressant effect (Supp Fig. 11). Paroxetine, citalopram, and sertraline combined with naltrexone produced no antidepressant effect (Supp Fig. 3). Citalopram, escitalopram, fluoxetine, and sertraline reduced the severity of GAD symptoms (d = 0.345, BF = 4.236) in alcohol, cocaine, marijuana, and nicotine use disorders. Despite fluoxetine’s advantage, no SSRI outperformed the others. Paroxetine lessened social anxiety symptoms (d = 0.687, BF = 8.726) in alcohol use disorder. Sertraline’s efficacy was inconclusive for PTSD symptoms in alcohol use disorder (Supp Fig. 6). Attrition rate, SSRI dosage, or treatment duration did not show any meta-regression change of SSRI’s response over placebo for depression (n = 29) (β = 0.308, BF = 0.568), GAD (n = 5) (β = 0.389, BF = 0.742), social anxiety (n = 4) (β = 0.988, BF = 1), and PTSD (n = 2) (β = 0.477, BF = 1).

Abstinence, craving, and substance use reduction

Carried out with a placebo arm, citalopram, escitalopram, sertraline, fluvoxamine, fluoxetine, and paroxetine improved abstinence for opioid, alcohol, cocaine, cannabis, or nicotine use (d = 0.325, BF > 99). On subgroup analysis, sertraline fared better (Supp Fig. 12). A reduction in the severity of craving was noticeable for fluoxetine, citalopram, sertraline, fluvoxamine, and paroxetine in alcohol, cocaine, or nicotine use (d = 0.513, BF = 12.331). Fluoxetine showed the highest anti-craving efficacy. Fluoxetine, sertraline, paroxetine, or citalopram significantly reduced alcohol use (d = 0.438, BF > 99). Fluoxetine had the superior outcome among them. Fluoxetine, citalopram, and sertraline reduced cocaine use (d = 0.255, BF = 3.87). SSRIs had no effect on reducing opioid, nicotine, or cannabis use (Supp Fig. 13–15). Attrition rate, SSRI dosage, or treatment duration did not show any meta-regression change for abstinence (n = 22) (β = 0.401, BF = 0.192), craving (n = 8) (β = 0.489, BF = 0.381), or alcohol use reduction (n = 21) (β = 0.596, BF = 0.350). However, based on SSRI dosage, there was a trend that SSRIs had no effect in reducing cocaine use (n = 10) (β = 0.139, BF = 2,012).

Discussion

SSRIs effectively treat depression, GAD, and social anxiety in individuals with SUDs. SSRIs can also help with alcohol abstinence and craving, as well as the reduction of alcohol and cocaine use. Fluoxetine performs better than the other SSRIs. The findings of this meta-analysis agree with those of Nunes and Levin (2004) [16], indicating a beneficial, albeit modest, effect of antidepressants in the treatment of depression in patients with SUDs. The effect size in Nunes’s study was 0.20. Our investigation found 0.3, with a BF > 99. This analysis agrees with Virtanen et al. 2022 [17] and supports the author’s finding of an association between SSRIs and substance risk reduction. This investigation, however, contradicts Iovieno et al.’s conclusions [14], which indicated that SSRIs failed to treat depression in alcohol use disorder.

SSRIs do not reduce opioid, nicotine, or cannabis consumption. An unfavorable outcome resulted from the intricate interactions between SERT and the three psychoactive substances. Mu, delta, and kappa opioid receptors differently influence SERT neurotransmission in the CNS [10]. It is difficult to predict how 5-HT receptor subtypes respond to this complex interface. Similarly, subtle and varied responses to nicotine with different 5-HT receptor subtypes have been reported [6]. The cannabis system has a specific colocalization pattern with DA and SERT receptors. For example, in low-CB1-expressing cells (principal projecting neurons), CB1 colocalizes with DA receptors, D1, D2, and 5-HT1B, while in high-CB1-expressing cells, primarily gamma-aminobutyric acid (GABAergic cells), CB1 colocalizes with 5-HT3 [94]. These reactions are convoluted and unpredictable since SERT subtypes influence or target cannabinoid receptors. Albeit a small number of RCTs, no prediction can be made about SSRIs’ capacity to reduce PTSD symptoms. Unexpected PTSD symptoms may modulate the benefits of SSRIs. Targeting numerous receptors at once may help treat this complex illness. 5-HT1B receptors are essential because they control fear and anxiety [95].

More than 20% of the overall trials had elevated attrition rates or did not provide data on attrition, which raises critical questions about their study design, confounders, or other issues inherent to the population studied. Nevertheless, judging by the review authors, all the trials underwent a controlled process, and there is no deviation from ethical or scientific standards. A behavioral intervention was undertaken in more than 50% of the trials. Although their purpose was not therapeutically oriented, their impact on the expected outcomes remains unclear.

There are limitations inherent to the design of this investigation. The protocol included only RCTs, which already overemphasized potential beneficial effects, a priori, in addition to publication bias. Also, in a small number of trials, their study design was not explicitly geared toward understanding the benefit of SSRIs in SUDs. Still, the authors could extract valuable data that fit into the Bayesian model. Furthermore, this analysis is not devoid of between-study statistical heterogeneity. A tentative explanation of heterogeneity veered attention to issues related to the use of distinctive and heterogeneous scales, nonspecific outcome measures from some trials, bias in reporting results, or low sample size. Thus, there is a need for a more homogeneous method to measure the treatment effect. Increasing power seems more difficult due to the problematic nature of treating addiction with high recidivism.

SUD is linked to antidepressant resistance [96]. The symptom overlaps between GAD and substance use, or withdrawal complicates GAD diagnosis. Clinicians are hesitant to prescribe SSRIs to substance-abusing patients due to misdiagnosis and poor treatment response [97]. Around half individuals with SUDs have PTSD. They had worse results [98]. The best treatment for them is unknown. There are no therapeutic guidelines for depression, anxiety, or PTSD unique in SUDs. What works for depression, anxiety, or PTSD in the absence of SUDs may or may not work in the presence of SUDs. This study analyzes these limitations since SUDs and mental diseases should be treated in concert, not separately. There are a lot of possible outcomes that are not covered in this meta-analysis, but they still need to be investigated. These include decreased risk behaviors; social, economic, and family functioning; hospitalizations; and suicide rates.

Research may focus on the interaction between SERT and other neurotransmitters such as GABA, N-methyl-d-aspartate, cannabinoids, opioids, or norepinephrine to better and perhaps mitigate dropouts and facilitate retention in treatment for SUDs and to better understand the neurobiology of SUDs comorbid with psychiatric disorders. Overall, fluoxetine seems to be a promising candidate. One might have to weigh risks and benefits before prescribing an SSRI constantly. One might preemptively attempt to mitigate potential adverse events such as QT prolongation, serotonin syndrome, bleeding, and suicide risk.

Conclusions

SSRIs appear promising as pharmacological interventions in treating depression and anxiety in individuals with SUDs and maintaining abstinence, facilitating, and reducing substance use. There was no effect of attrition rate, SSRI dosage, or treatment length on SSRI's efficacy. This is a preliminary work that may guide treatments in diverse clinical settings. Further research could help determine how much SSRIs should be taken, how well they work, and how safe they are for individuals with SUDs and psychiatric disorders.