The effect of probiotics on depressive symptoms: A meta-analysis of randomized controlled trails

Increasing evidence indicates bidirectional communication occurs between the gut and the brain via the microbiomegut-brain axis. Accordingly, several studies point to a positive effect of probiotics on depressive symptomatology. Our study aimed to provide an updated meta-analysis on the subject. We conducted a systematic search for randomized clinical trials in multiple databases. Data analysis was performed using the Review Manager 5.4. We reviewed 26 studies, totaling 2,170 participants. The use of probiotics had a significant effect in reducing depressive symptoms (SMD: 0.22; 95% CI: 0.30 to 0.13; p < 0.0001) compared to placebo. Subgroup analyses showed that probiotics influenced outcomes in participants with depression (SMD: 0.67; 95% CI: 0.93 to 0.41; p < 0.00001) and with other clinical diagnoses (SMD: 0.23; 95% CI: 0.36 to 0.09; p = 0.0008) but not healthy participants (SMD: 0.11; 95% CI: 0.23 to 0.02; p = 0.100). The use of probiotics can be a promising approach in improving the treatment of depression. It is important to conduct further studies to better assess the efficacy of probiotic as an adjuvant treatment to antidepressant medications, as well as their most effective dosages.


Introduction
The human intestine is colonized by approximately 3.8 x 10 ^ 13 microorganisms, a number larger than that of actual human cells (Sender et al., 2016). These microorganisms can communicate with the central nervous system (CNS) through several mechanisms. The gut microbes produce γ-aminobutyric acid (GABA), serotonin and catecholamines. These substances work as local neurotransmitters, modulating exocrine secretion, local motility and vagal afferences, consequently can indirectly stimulates the CNS. Gut microbiota can still produce metabolites such as short-chain fatty acids (SCFA) which can cross the blood-brain barrier and influence neuron's metabolism. Microbes can elicit the production of pro-inflammatory cytokines that influence microglia activation and stimulate the brain through vagus nerve. The CNS, in turn, can modulate the release of substances, intestinal motility and permeability. This set of bidirectional communication is called the microbiota-gut-brain axis (Sherwin et al., 2016).
The intestinal microbiota manipulation in animal models results in visible behavioral changes (Reis et al., 2018). Clinical evidence associates abnormalities of the normal microbiota (dysbiosis) with the appearance of neurological and psychiatric diseases (Kong et al., 2019;Jiang et al., 2015), thus encouraging research that uses microorganisms (probiotics) in the prevention and treatment of these diseases.
In recent years, the clinical use of probiotics has expanded. Meta-analyses have confirmed its effectiveness in the treatment of irritable bowel syndrome (IBS) and ulcerative colitis although the strength of evidence is low and probiotic supplement is not yet broadly recommended (Niu et al., 2020;Astó et al., 2019;Guo et al. 2019).
Besides, new evidence reveals that probiotics have effects on psychiatric symptoms (Sherwin et al., 2016). For example, studies have shown that the consumption of probiotics significantly reduced the perception of depressed mood (Steenbergen et al., 2015), reduced depression scores,10 and improved mental health measures Research, Society andDevelopment, v. 10, n. 9, e54510918359, 2021 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v10i9.18359 3 in healthy adults (Mohammadi et al., 2016). This opens the possibility that probiotics could be used to treat and prevent psychiatric illnesses and relieve psychological stress.
According to the World Health Organization (WHO, 2017), psychological stress affects 90% of the world's population and mental disorders are among the most disabling diseases. Major depressive disorder (MDD) is one of the most prevalent psychiatric illnesses. It is characterized by prolonged lowering of mood and may be accompanied by changes in memory, attention, sleep, motivation, muscle tension, and autonomic functioning (Kupfer et al. 2014). Psychiatric diseases have a great social and economic impact, being associated with low school performance and lower productivity at work, a greater number of traffic accidents, drug abuse, and suicide (WHO, 2001).
The pathophysiology of psychic diseases is complex and multifactorial, involving several neural, inflammatory, and metabolic pathways, which makes prevention and treatment difficult. Current therapeutic approaches to depression are not satisfactory, since the average response to the most used drugs is around 50% (McCormack et al., 2018). Clinical evidence suggests an association between psychic and gastrointestinal symptoms (Ma et al., 2019;Stasi et al., 2019). Abnormalities in the intestinal microbiota, common in psychiatric patients, may compose the pathophysiology of mental illness and/or aggravate the psychic symptoms of individuals in this condition (Petra et al., 2015;Myers et al., 2009).
Depression can be screened and diagnosed using assessment instruments. These instruments are validated scales of either interview or self-reported measurements that quantify depressive symptoms and determine treatment outcomes. Some notorious examples are Beck Depression Inventory (BDI) and Hamilton Depression Rating Scales (HAM-D) (https://www.apa.org/depression-guideline/assessment) The relevance of mental illness in our society, the limitations of usual treatment, and the large availability of probiotics in the market justify the review of the evidence obtained in randomized controlled trials (RCT) on the effect of probiotics on depressive symptoms. This study aimed to examine the effect of probiotics on depression scales and to separately analyze subgroups of participants.

Methods
The literature search was carried out according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Moher et al 2009 Research, Society andDevelopment, v. 10, n. 9, e54510918359, 2021 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v10i9.18359 4 There was no language restriction. An initial screening of titles and abstracts identified potential studies, after which an analysis of complete articles was conducted to definitively include or exclude the study based on the eligibility criteria described below. This process was carried out independently by two of the authors (VP and MP). Any disagreement was resolved through discussion and consensus. In cases where consensus could not be reached, it was planned that a third author (CG) would make the final decision.

Eligibility Criteria and Study Selection
The inclusion criteria were the following: (1) RCT design comparing probiotic and placebo; (2) Studies in healthy humans or with a diagnosed psychiatric illness, such as depression or anxiety; (3) Any formulation and administration of probiotics that report their values of colony-forming units (CFU), species of the product, dosages, and duration of the intervention; (4) Use of validated scales for symptoms of depression; (5) Studies containing pre-and postintervention results in the form of means and standard deviations, or that could be converted mathematically into these two measures.
Studies that used probiotics in addition to some other nutrient or adjuvant, such as vitamin D or prebiotics, were included. Studies were also included in which the volunteers had other diseases such as fibromyalgia (Roman et al., 2018) and polycystic ovary syndrome (Jamilian et al., 2018;Ostadmohammadi et al. 2019).
Studies with the respective criteria were excluded from the meta-analysis: (1) Use of probiotics to treat gastrointestinal diseases (IBS, ulcerative colitis, antibiotic-associated diarrhea), considering that probiotics may have a positive effect on the symptoms of disease and secondarily impact mood and depression scales; (2) They did not provide clear or calculable data before and after intervention or incomplete data; (3) Inclusion of children or adolescents (≤ 18 years); (4) Studies exclusively reported anxiety measures; (5) Instruments related to stress, quality of sleep, quality of life, general health condition, and intestinal symptoms were excluded from the metaanalysis to reduce the heterogeneity of the studies.
The search methodology was outlined in a flowchart and described in Figure 1.

Data Extraction
The relevant data were extracted using the Google Forms tool and later transferred to an Excel spreadsheet. A standard table was developed by author VP, MP and CG following the Cochrane Handbook for Systematic Reviews of Interventions.
For the articles that made their results available in standard error of mean (SEM), we transformed them into standard deviations (SD) using the mathematical formula SEM = SD / √n, for which "n" represents the size of the sample studied.
Dropout values were compared between the placebo and probiotic groups to assess the impact of adverse effects on the study segment.

Risk of Bias Assessment
The risk of bias was measured using the Cochrane Risk of Bias Assessment Tool (Higgins et al., 2011) and is described in Figure 2.

Statistical Analysis
Each study was conducted separately by two authors and analyzed using Revman 5.4 software. The treatment effect was calculated using standardized mean difference (SMD) between the value of the first postintervention measure in the probiotic group and the value in the placebo group for the depression symptom scales.
This methodology was used both for the total analysis and for the analysis of subgroups.
Heterogeneity between studies was assessed using the chi-square (Χ2) test and Ι2 statistic. Great heterogeneity between articles is indicated when p < 0.05 for Χ2 or Ι2 > 50%. Possible publication bias was investigated using a funnel plot (Egger et al., 1997).
We proceeded with the analysis of three subgroups: 1) MDD patients; 2) participants with clinical diagnoses other than depression; 3) healthy participants.
A t-test for independent samples was performed to investigate the difference between mean dropout rates of the placebo and probiotic groups.

Included Studies
From a total of 2,185 studies, we analyzed 26 double-blinded, randomized, placebo-controlled trials. Figure 1 shows the identification, screening, selection criteria, and inclusion process.
Twelve studies investigated the effect of probiotics in healthy patients (Chung et al., 2014;Lee et al., 2021;Messaoudi et al., 2011;Mohammadi et al., 2016;Nishida et al., 2019;Nishihira et al., 2014;Östlund-Lagerström et al., 2016;Romijn et al., 2017;Siegel & Conklin, 2020;Smith-ryan et al., n.d.;Steenbergen et al., 2015;Venkataraman et al., 2021). In terms of clinical diagnoses, five studies used a sample of patients diagnosed with MDD (Akkasheh et al., 2016;Chahwan et al., 2019;Ghorbani et al., 2018;Kazemi et al., 2019;Reininghaus et al., 2020), two studies included women with polycystic ovarian syndrome (POS) (Jamilian et al., 2018;Ostadmohammadi et al., 2019) two studies integrated patients with type 2 diabetes mellitus and coronary heart disease (Raygan et al., 2018(Raygan et al., , 2019, two studies included a sample of pregnant women (Dawe et al., 2020;Slykerman et al., 2017), one study comprised patients with myocardial infarction who underwent percutaneous coronary intervention, (Moludi et al., 2019) one article included patients with coronary artery disease (CAD) (Moludi et al., 2021) and one study included patients with fibromyalgia, (Roman et al., 2018). The former nine articles made up the "other clinical diagnoses" subgroup.  Figure 2A summarizes the risk of bias for each study. Figure 2B presents the risk of bias of all RCTs included in the study using a distribution chart of the reviewers' judgments. Eight studies showed a high risk of bias for a single domain, but the overall risk of bias was low.

Probiotics and Depression
Main characteristics of the 26 RCTs are summarized in Table 1, which contains summary information about the date of publication, country of study, general condition of participants, number of patients recruited, and the average age for each intervention group: probiotic (PRO) or placebo (PLA). Further, the gender of participants, species, and strain of the probiotic, dosage in CFU total daily, presentation of the probiotic, duration of the intervention, and results of interest to this meta-analysis are also summarized.

Subgroup Analysis of Probiotic Efficacy Based on Clinical Condition
Assessing the efficacy of a probiotic over a placebo for the subgroup of patients diagnosed with MDD involved five studies, showing a significant impact on reducing depressive symptoms in the first postintervention measure (SMD: -0.67; 95% CI: -0.93 to -0.41; p < 0.00001; I 2 = 74%). For participants with other clinical diagnoses, a significant improvement was also found compared to a placebo (SMD: -0.23; 95% CI: -0.36 to -0.09; p = 0.0008; I 2 = 32%). As for healthy participants, treatment with probiotics did not have a significant effect (SMD: -0.11; 95% CI: -0.23 to 0.02; p = 0.100; I 2 = 84%). The results are summarized in Figure 3. The general analysis of all studies obtained I 2 = 78%, considerable heterogeneity, while the analyses of the subgroups showed I 2 ≥ 32% -84%, moderate to substantial heterogeneity.

Publication Bias and Sensitivity Analyses
Regarding the funnel plot (Figure 4), the visual inspection of the figure showed an asymmetric distribution of the studies. This suggests a possible bias of publishing that favors studies reporting a positive effect of probiotics.
The results of the sensitivity analysis did not show significant changes between groups. This test was performed excluding each study individually, including those most vulnerable to bias.

Adverse Effect of Probiotics
The number of volunteers who discontinued treatment was slightly greater in the probiotic group when considering all 26 studies. The independent student's t-test showed no significant difference between the means of dropping out of the placebo and probiotic groups (p = 0.791). Research, Society and Development, v. 10, n. 9, e54510918359, 2021 (CC BY 4.  L. acidophilus (2 x 10 9 ), L. reuteri (2 x 10 9 ), L.

Discussion
The present meta-analysis suggests that probiotics significantly improved depressive symptoms compared to a placebo. These results differ from those of Ng et al's meta-analysis (Ng et al., 2018), that, in 2018, included 10 articles. In comparison, we included sixteen more studies and increased the number of participants from 1,349 to 2,170. Similar to antidepressants, probiotics have a greater effect on those diagnosed with MDD. The usually mild depressive symptomatology found in non-depressed individuals may lead to greater dispersion, making it difficult to analyze the effect of interventions, such as probiotics (Goh et al., 2019). Another reason is the commonly found abnormality of the intestinal microbiota in patients with diagnosed depression, which may be a relevant pathophysiological mechanism modified by probiotic use.5 What may explain the effect of probiotics on participants with other clinical conditions is that they tend to have a more depressive baseline mood and a higher prevalence of depression compared to healthy participants (Li et al., 2019).
It is worth mentioning that, due to the low number of studies, it is difficult to quantify the additional effect of the probiotic in the treatment of depression. Of the five articles with patients diagnosed with MDD eligible for this meta-analysis, only one specified the conventional antidepressant drug used both in the placebo and exposed groups (Ghorbani et al., 2018).
Several mechanisms are suggested to explain the impact of probiotics on mood and are summarized in what is known as the microbiota-gut-brain axis. Preclinical methodology, in comparison to clinical studies, is better suited for studying pathophysiological mechanisms and thus clarifying this axis. Nonetheless, despite optimistic findings in preclinical studies, the use of probiotics in human participants remains a challenge. Models for to assessing changes in behavior and mood in non-human subjects are limited, and their interpretations do not allow large generalizations for humans (Harro, 2019). Animal testing is based on indirect inferences of the animal's mood from objective behavioral data, while human scales are subjective, based on self-rated scales (Harro, 2019). Differences in intestinal microbiota, genetics, and neurological complexity between rodents and humans make the translation of preclinical studies even more difficult for clinical applications (Kelly et al., 2017).
Still, it is known that the vagus nerve has a crucial effect on communication between the intestine and the brain. The administration of probiotics in animals caused behavioral changes related to fear and anxiety, both avoided through the vagotomy prior to the administration of the probiotic (Bercik et al., 2011;Marvel et al., 2004). Short-chain fatty acids, a product of bacterial metabolism, can directly affect animal behavior (DeCastro et al., 2005). Evidence suggests that the intestinal microbiota produces important neurotransmitters or their precursors that, acting locally, can influence the brain through vagus afferences: tryptophans, serotonin, and GABA. Treatment with probiotics can modulate the production of these substances in rodents and humans (Barrett et al., 2012;Desbonnet et al., 2008).
The hypothalamic-pituitary-adrenal (HPA) axis and inflammatory mechanisms have great relevance in the pathophysiology of mood disorders (Foster & McVey Neufeld, 2013). The use of probiotics reduced the hyperactivity of the HPA axis (Liu et al., 2016), and changes in the intestinal microbiota are related to the proinflammatory state in the brain of rodents and humans (Daulatzai, 2014;Huuskonen et al., 2004). A survey conducted by Mohammadi et al. (2016) demonstrated that the use of probiotics has been able to significantly reduce circulating cortisol levels in humans.
Unlike previous meta-analyses on the topic, asymmetry of the funnel plot was found in this review.
Considering the inexistence of inadequate analyses among the RCTs used, as well as the correct choice of the measure of the effect, the pattern found by us suggests publication bias. This indicates a preference for publications in which probiotics demonstrate significant effects. However, since the included RCTs are good methodological quality, this risk of bias is lower.
There is no consensus about strains, the dose in CFU, duration, adjuvants, and frequency of use of probiotics that should be used in humans. The differences in these dosage variables and the depressive mood scales explain the sample's considerable heterogeneity and hence may have interfered with the funnel plot asymmetry. Moreover, most studies showing significative beneficial effect of probiotics came from Iran and all studies outside Iran except one from New Zealand and other from India show no beneficial effect. It seems to be a fundamental limitation to generalize the present outcome to other ethnic groups.
Probiotics are notoriously safe treatments with mild and infrequent adverse effects (Niu, 2020). The analyzed studies sporadically indicated nausea (Slykerman et al., 2017), diarrhea, constipation (Östlund-Lagerström et al., 2016), and abdominal discomfort (Roman et al., 2018). Symptoms also present in the placebo group, which may indicate the effect of adjuvant substances (e.g., corn starch, fructooligosaccharides, maltodextrin). This meta-analysis also suggests no significant difference in the dropout rate between the placebo and exposed groups.
One of the most important limitations of this meta-analysis is the lack of uniformity between the designs of the included RCTs. First, the studies involved participants with different diagnoses, with only five involving patients with MDD. Allied to this, a large difference in the methods of therapeutic intervention and the scales to access the results was seen. This contributes to increased heterogeneity and, therefore, limits the power of generalization of the results. Thus, its reduction should be a target in the next meta-analyses. Another limiting fact is that the available studies do not assess the persistence of the effect of the probiotics by long-term follow-up measures. In most studies, only the first postintervention measure, performed immediately after the end of treatment, was presented.
For these reasons, we recommend an elaboration of studies that evaluate the effect of probiotics in patients diagnosed with MDD, with more participants and multiple scales for depressed mood, valuing those that are repeated in the studies. Future research could also ascertain the best dosage of probiotics and compare doses, duration, frequency, formulation, strains, and the number of species as well as the persistence of the effect through follow-up.

Conclusion
This meta-analysis with 26 RCTs found that probiotic supplementation had a significant effect in reducing depressive symptoms. A subgroup analysis showed a significant reduction in participants with MDD and other diagnosed clinical conditions, but not in healthy volunteers. The methodological differences between the studies and the low number of RCTs involving participants with depression decrease the ability to generalize this meta-analysis. Therefore, we suggest more clinical trials involving a larger number of participants diagnosed with MDD who use probiotics as an adjunct to antidepressants with a focus on using recurrent scales in studies and finding the best dosage.