Gut bacteria imbalance linked to brain-gut disruptions in psychiatric disorders

Our inclusion criteria were articles that investigated the gut microbiome and the effects on the development and/or prognosis of psychiatric disorders in patients diagnosed by a healthcare professional using DSM-V and DSM-IV criteria, as well as scales approved by the DSM. Only primary studies (cross-sectional, cohort, case–control, case series, and clinical trials) in both English and Spanish were included in this systematic review. There were no limitations related to the year of publication, age, gender, or ethnicity of the participants. The exclusion criteria included neuropsychiatric disorders caused by an identifiable organic lesion (i.e., tumour, ischaemia, etc.); animal and in-vitro studies were also excluded. Studies that considered self-reporting or the use of a diagnostic tool other than the DSM-V or those approved by the DSM-V, were not considered.

We queried the most relevant biomedical databases for our study focus, including MEDLINE (PubMed), Scopus, CENTRAL, and PsycINFO until February 13, 2022. Searches were supplemented by manual retrieval of any additional articles meeting eligibility criteria found in the references of selected articles. The key terms used in the search were gastrointestinal microbiome, mental disorders, psychiatric disorders, microbiome-gut–brain axis, gut–brain axis, neuropathogenesis, immune system, dysbiosis, and their variants; the complete search strategy used in each database can be found in the. Supplementary materials

Two blinded authors independently reviewed the titles and abstracts of all the articles, after deduplication, against the aforementioned eligibility criteria; if any discrepancies were identified, a third author weighed in until mutual consensus was achieved. Afterwards, in a similar fashion, the remaining articles were further assessed by reading the full-text. Articles that successfully passed the process were then scrutinized for relevant information that was recollected into an Excel spreadsheet. Studies using ICD-10 or ICD-11 diagnostic criteria were included, as these frameworks are internationally harmonized with DSM classifications for the psychiatric disorders considered. In addition, disorder-specific validated diagnostic instruments (e.g., ADOS, ADI-R, K-SADS) were included when they operationalize DSM- or ICD-based diagnoses. Studies relying solely on self-report measures or non-validated screening tools were excluded. No analytical weighting was applied based on diagnostic framework, and all included studies were synthesized descriptively.

Data was collected individually by the reviewers and any discrepancy was solved by discussion and mutual consensus. Data was extracted in an excel spreadsheet that contained the following variables: article identifying information (i.e., DOI, authors, year), sample size, study design, number, gender, and age of participants, prior psychiatric diagnosis, tool used for diagnosis, microbiome assessment, bacteria identified and categorized by phylum, family, and genus, change in the bacteria identified, and correlations with the psychiatric condition.

We assessed risk of bias in the included studies using NIH’s Study Quality Assessment Tools—namely the Quality Assessment of Controlled Intervention Studies Tool, the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies, the Quality Assessment of Case–Control Studies tool, and the Quality Assessment Tool for Case Series Studies—for evaluation of experimental, cohort, cross-sectional, case–control, and case series studies, respectively (NHLBI and NIH, 2025). Two authors independently applied the respective tool to each included study and recorded the answers for every question. There were three possible answers: yes, no, or other (cannot determine or not applicable). We then calculated a percentage for every study based on the number of yes out of the total number of questions. We classified every study into three categories: minimally low risk if the percentage of “yes” was 80% or higher, moderately low risk if the percentage was between 50 and 79%, and high risk if the percentage was less than 50%. Any disagreements were resolved by a third author who used the same methodology.

To quantify the effect measure we calculated the relationship between the sample size and the change proportion in the microbiota for each mental disorder, considering the total amount of participants within each category of mental disorder diagnosis, and the bacteria presented in each case. We obtained an increase or decrease percentage value for each bacterium. The distribution of mental disorder categories was made as follows: (1) Schizophrenia; (2) autism spectrum disorder; (3) mood disorders including major depressive disorder (MDD), generalized anxiety disorder (GAD), and bipolar disorder (BPD); (4) attention deficit hyperactivity disorder (ADHD); (5) eating disorders including anorexia nervosa and binge eating disorder; and, (6) other mental disorders such as hypoactive sexual disorder, sleep disorders [idiopathic rapid eye movement (REM) sleep behavior disorder (iRBD)] and post-traumatic stress disorder (PTSD).

All reported outcomes were arranged in tables that highlighted the detailed differences in the gut microbiome at the phylum, family, and genus levels between the case subjects and the control groups. The results from each study were summarized by categorizing changes in the relative abundance (percentage), absolute abundance (counts), or diversity of each microorganism as increased, decreased, or unchanged. To summarize microbiome alterations across heterogeneous studies, we calculated descriptive percentages reflecting the proportion of participants within each psychiatric disorder category for whom a given taxon was reported as increased or decreased relative to controls. These values were derived by aggregating participant counts from individual studies reporting a directional change in a specific taxon and expressing this number as a percentage of the total number of participants within that diagnostic category. Importantly, these percentages do not represent pooled prevalence estimates, effect sizes, or meta-analytic measures, nor do they reflect the proportion of studies reporting a given change. Rather, they provide a descriptive synthesis intended to illustrate the relative consistency and recurrence of reported microbial alterations across the included literature. Accordingly, these percentages should not be interpreted as true prevalence rates or measures of association strength.

Our search yielded a total of 3,512 articles, after duplicate removal, of which 80 articles were finally selected for our review, providing a total of 2,691 participants (Aarts et al., 2017; Aizawa et al., 2016; Armougom et al., 2009; Averina et al., 2020; Bojović et al., 2020; Borgo et al., 2017; Cao et al., 2021; Carissimi et al., 2019; Chen et al., 2018, 2019, 2020; Chung et al., 2019; Coello et al., 2021; Angelis et al., 2013; Dong et al., 2020; Evans et al., 2017; Finegold et al., 2010; Gondalia et al., 2012; Grimaldi et al., 2018; Hanachi et al., 2019; Hemmings et al., 2017; Hua et al., 2020; Huang et al., 2018; Inoue et al., 2016; Pärtty et al., 2015; Jiang et al., 2015; Jiang H. -Y. et al., 2018; Jiang H. et al., 2018; Kang et al., 2013, 2017, 2018, 2019; Kleiman et al., 2015; Kong et al., 2019; Kushak et al., 2017; Lai et al., 2021; Leyrolle et al., 2021; Li et al., 2019, 2020, 2021; Lin et al., 2017; Liu B. et al., 2019; Liu et al., 2017; Liu S. et al., 2019; Lu et al., 2019; Luna et al., 2017; Ma et al., 2019, 2020; Mason et al., 2020; Morita et al., 2015; Naseribafrouei et al., 2014; Nguyen et al., 2019, 2021; Nishiwaki et al., 2020a; Painold et al., 2019; Pan et al., 2020; Pełka-Wysiecka et al., 2019; Plaza-Díaz et al., 2019; Prehn-Kristensen et al., 2018; Rong et al., 2019; Shaaban et al., 2018; Shen et al., 2018; Son et al., 2015; Strati et al., 2017; Sun et al., 2019; Tomova et al., 2020; Vinberg et al., 2019; Wan et al., 2020; Wang et al., 2019a,b; Williams et al., 2011, 2012; Yuan et al., 2018; Zhai et al., 2019; Zhang et al., 2018; Zhang M. et al., 2020; Zhang X. et al., 2020; Zou et al., 2021; Zurita et al., 2020; Pulikkan et al., 2018). The complete screening process can be found on Figure 1. Our review considered a variety of diagnoses including psychotic disorders (schizophrenia and first psychotic episode), autism spectrum disorder, mood disorders (major depressive disorder and bipolar disorder), generalized anxiety disorder, eating disorders (anorexia nervosa, food addiction), post-traumatic stress disorder (PTSD), and sleep disorders.

The diagnostic tools used across the studies included the Diagnostic and Statistical Manual of Mental Disorders (DSM), fifth and fourth editions (DSM-IV, DSM-5), International Classification of Diseases (ICD), tenth and eleventh revisions (ICD-10 and ICD-11), Autism Diagnostic Interview–Revised (ADI-R), Autism Diagnostic Observation Schedule (ADOS), ADOS-2, Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS), the Hamilton Depression Rating Scale, Autism Treatment Evaluation Checklist (ATEC), Pervasive Developmental Disorder Behavior Inventory (PDD-BI), and the Classification of Sleep Disorders Criteria-Third Edition (Aarts et al., 2017; Aizawa et al., 2016; Armougom et al., 2009; Averina et al., 2020; Bojović et al., 2020; Borgo et al., 2017; Cao et al., 2021; Carissimi et al., 2019; Chen et al., 2018, 2019, 2020; Chung et al., 2019; Coello et al., 2021; Angelis et al., 2013; Dong et al., 2020; Evans et al., 2017; Finegold et al., 2010; Gondalia et al., 2012; Grimaldi et al., 2018; Hanachi et al., 2019; Hemmings et al., 2017; Hua et al., 2020; Huang et al., 2018; Inoue et al., 2016; Pärtty et al., 2015; Jiang et al., 2015; Jiang H. -Y. et al., 2018; Jiang H. et al., 2018; Kang et al., 2013, 2017, 2018, 2019; Kleiman et al., 2015; Kong et al., 2019; Kushak et al., 2017; Lai et al., 2021; Leyrolle et al., 2021; Li et al., 2019, 2020, 2021; Lin et al., 2017; Liu B. et al., 2019; Liu et al., 2017; Liu S. et al., 2019; Lu et al., 2019; Luna et al., 2017; Ma et al., 2019, 2020; Mason et al., 2020; Morita et al., 2015; Naseribafrouei et al., 2014; Nguyen et al., 2019, 2021; Nishiwaki et al., 2020a; Painold et al., 2019; Pan et al., 2020; Pełka-Wysiecka et al., 2019; Plaza-Díaz et al., 2019; Prehn-Kristensen et al., 2018; Rong et al., 2019; Shaaban et al., 2018; Shen et al., 2018; Son et al., 2015; Strati et al., 2017; Sun et al., 2019; Tomova et al., 2020; Vinberg et al., 2019; Wan et al., 2020; Wang et al., 2019a,b; Williams et al., 2011, 2012; Yuan et al., 2018; Zhai et al., 2019; Zhang et al., 2018; Zhang M. et al., 2020; Zhang X. et al., 2020; Zou et al., 2021; Zurita et al., 2020; Pulikkan et al., 2018).

Table 1 showcases the calculated risk of bias, and related scores, that resulted from applying the NIH’s Study Quality Assessment Tools (NHLBI and NIH, 2025).

Overall, the methodological quality of the included studies was moderate. Although most studies met criteria for classification as having a moderately low risk of bias according to NIH assessment tools, many scores clustered near the lower bound of this category (approximately 50–60%), indicating relevant methodological limitations. A smaller subset of studies was classified as high risk of bias (Aizawa et al., 2016; Chen et al., 2018; Dong et al., 2020; Huang et al., 2018; Inoue et al., 2016; Jiang H. et al., 2018; Kang et al., 2013, 2018; Lin et al., 2017) and only study was considered to have minimally low risk of bias (Pärtty et al., 2015). These findings suggest that the current evidence base should be interpreted with caution.

Figure 2, Table 2 showcases a summary of the relevant changes in IM presented as percentage of affected patients and common patterns of change or consequences of the IM change. Table 3 showcases the potential diagnostic implication of taxa alteration in relevant psychiatric conditions included.

The findings of this review reveal reproducible, disorder-specific patterns of gut microbial dysbiosis that may have diagnostic utility in psychiatry. Notably, decreased Firmicutes in ASD, increased Enterobacteriaceae in schizophrenia, and complete loss of Akkermansia in eating disorders highlight taxa that may distinguish clinical phenotypes. Although larger absolute reductions in Firmicutes were observed in conditions such as anorexia nervosa and insomnia, these changes are likely driven by acute nutritional deprivation or disorder-specific physiological states rather than stable, disorder-specific microbial signatures. In contrast, the reduction of Firmicutes in autism spectrum disorder has been consistently reported across multiple independent cohorts and age groups, supporting its potential relevance as a reproducible and diagnostically informative feature rather than a secondary metabolic consequence. These microbial signatures offer promise as non-invasive, adjunctive tools for early or differential diagnosis. Certainly, the application of microbial panels, comprising combinations of discriminative taxa, has already been explored in pilot studies. For instance, classifiers trained on faecal microbiota data have achieved moderate-to-high diagnostic accuracy in distinguishing patients with major depressive disorder, schizophrenia, and ASD from healthy controls. Machine learning-based algorithms, using genus- or family-level microbiota features, have shown preliminary success in differentiating disease subtypes and predicting symptom severity. These computational approaches represent a scalable framework for microbiome-based diagnostic development. Furthermore, biomarker assays targeting microbial metabolites or microbial DNA are being developed as potential clinical tests. For example, altered levels of short-chain fatty acids, trimethylamine-N-oxide (TMAO), or lipopolysaccharide-producing bacteria may reflect underlying pathophysiology and could be incorporated into multimodal diagnostic workflows.

However, current evidence remains preliminary. Variation in study design, sampling techniques, and bioinformatics pipelines limits generalisability. For microbiome signatures to become clinically actionable diagnostic tools, prospective longitudinal studies with standardised methods and independent validation cohorts are needed. Nevertheless, our findings provide a foundation for the development of potential microbiome-informed diagnostic frameworks in psychiatry. However, it is important to distinguish between disorder-specific microbial signatures and transdiagnostic patterns shared across multiple psychiatric conditions. Several taxa highlighted in this review, such as alterations in the Firmicutes–Bacteroidetes ratio or changes in Christensenellaceae, appear across multiple diagnostic categories and are therefore more likely to reflect shared metabolic, inflammatory, or lifestyle-related factors rather than disorder-specific markers. In contrast, certain microbial alterations, such as the complete loss of Akkermansia in binge eating disorder or reproducible reductions in Firmicutes in autism spectrum disorder, may represent more diagnosis-informative candidate signatures if properly replicated in future studies.

Across disorders, dysbiosis appears to be characterized by a general imbalance in the phyla Firmicutes, Bacteroidetes, and Actinobacteria, along with disorder-specific fluctuations at the family and genus levels. These shifts underscore a shared microbial signature associated with psychiatric symptomatology, while also pointing to unique microbial alterations per disorder, suggesting a multifactorial and diagnosis-specific microbiota-psychopathology relationship.

Across disorders, dysbiosis was most consistently characterized by a disruption of the Firmicutes–Bacteroidetes balance rather than a uniform directional change (Aarts et al., 2017; Aizawa et al., 2016; Armougom et al., 2009; Averina et al., 2020; Bojović et al., 2020; Borgo et al., 2017; Cao et al., 2021; Carissimi et al., 2019; Chen et al., 2018, 2019, 2020; Chung et al., 2019; Coello et al., 2021; Angelis et al., 2013; Dong et al., 2020; Evans et al., 2017; Finegold et al., 2010; Gondalia et al., 2012; Grimaldi et al., 2018; Hanachi et al., 2019; Hemmings et al., 2017; Hua et al., 2020; Huang et al., 2018; Inoue et al., 2016; Pärtty et al., 2015; Jiang et al., 2015; Jiang H. -Y. et al., 2018; Jiang H. et al., 2018; Kang et al., 2013, 2017, 2018, 2019; Kleiman et al., 2015; Kong et al., 2019; Kushak et al., 2017; Lai et al., 2021; Leyrolle et al., 2021; Li et al., 2019, 2021, 2020; Lin et al., 2017; Liu B. et al., 2019; Liu et al., 2017; Liu S. et al., 2019; Lu et al., 2019; Luna et al., 2017; Ma et al., 2019, 2020; Mason et al., 2020; Morita et al., 2015; Naseribafrouei et al., 2014; Nguyen et al., 2019, 2021; Nishiwaki et al., 2020a; Painold et al., 2019; Pan et al., 2020; Pełka-Wysiecka et al., 2019; Plaza-Díaz et al., 2019; Prehn-Kristensen et al., 2018; Rong et al., 2019; Shaaban et al., 2018; Shen et al., 2018; Son et al., 2015; Strati et al., 2017; Sun et al., 2019; Tomova et al., 2020; Vinberg et al., 2019; Wan et al., 2020; Wang et al., 2019a,b; Williams et al., 2011, 2012; Yuan et al., 2018; Zhai et al., 2019; Zhang et al., 2018; Zhang M. et al., 2020; Zhang X. et al., 2020; Zou et al., 2021; Zurita et al., 2020; Pulikkan et al., 2018). While the magnitude and direction of Firmicutes and Bacteroidetes alterations varied by diagnosis, this recurrent imbalance suggests shared perturbations in core metabolic and immunomodulatory pathways across psychiatric conditions. These findings are consistent with the disorder-specific patterns reported in the Results section and Table 2, where both increases and decreases in Firmicutes and Bacteroidetes were observed depending on the psychiatric phenotype. Firmicutes are generally associated with butyrate production and intestinal barrier integrity, while Bacteroidetes play roles in polysaccharide digestion and immune regulation (Qin et al., 2010; Rinninella et al., 2019); disruption in this ratio has been previously linked with neuroinflammatory processes and altered neurotransmitter production (Kelly et al., 2016; Dinan and Cryan, 2015). Additionally, the increase in Actinobacteria, particularly Bifidobacteriaceae, in disorders such as ASD and schizophrenia, may represent a compensatory response or reflect underlying dietary and metabolic adaptations (Strati et al., 2017; Zhang et al., 2018; O’Mahony et al., 2009).

When looking at specific psychiatric disorders, patients with ASD consistently exhibited reductions in Firmicutes and Bacteroidetes, and increases in Actinobacteria, Eggerthellaceae, and Bifidobacteriaceae. Genus-level increases in Faecalibacterium, Prevotella, and Parabacteroides, coupled with marked reductions in Clostridium and Sutterella, align with studies linking ASD to microbial pathways involved in GABA and serotonin metabolism (Finegold et al., 2010; Luna et al., 2017). Furthermore, reduced microbial diversity and altered short-chain fatty acid (SCF) production, particularly butyrate, may influence neurodevelopmental trajectories and social behavior through epigenetic and neuroimmune pathways (Sharon et al., 2019).

In major depressive disorder and bipolar disorder, shifts in Christensenellaceae (↑ 18%) and Ruminococcaceae (↓ 2%) were salient. These families have been implicated in gut permeability and pro-inflammatory cytokine production, mechanisms increasingly recognized in depression’s pathophysiology (Mason et al., 2020; Zheng et al., 2016). Elevated levels of Faecalibacterium and Flavonifractor suggest a compensatory attempt at restoring gut-brain homeostasis; however, reductions in Clostridium XI (↓ 6.01%) may indicate a persistent dysbiotic state unable to support proper neuromodulatory signaling (Painold et al., 2019; Evans et al., 2017).

Individuals with schizophrenia exhibited elevated Enterobacteriaceae and Christensenellaceae, along with reductions in Turicibacteraceae and Pasteurellaceae. These alterations points toward the possibility of increased gut permeability, systemic inflammation, and microbial-derived metabolites such as D-lactic acid and lipopolysaccharides, which have been shown to breach the blood–brain barrier and modulate microglial activation (Pełka-Wysiecka et al., 2019; Fung et al., 2017; McGuinness et al., 2022). Genera such as Succinivibrio and Clostridium were frequently increased, echoing earlier reports of their role in dopaminergic modulation (Xu et al., 2020).

Anorexia nervosa was characterized by sharp declines in Lactobacillus, Clostridium, and Roseburia, reflecting a state of nutrient deficiency and intestinal inflammation (Armougom et al., 2009; Hanachi et al., 2019; Di Lodovico et al., 2021). In contrast, Methanobrevibacter was consistently elevated, suggesting altered fermentation and methane metabolism, possibly contributing to constipation and delayed intestinal transit (Xu et al., 2022). Binge eating disorder revealed a striking 100% decrease in Akkermansia, a mucin-degrading genus critical for gut barrier function and metabolic health (Dao et al., 2016).

ADHD presented a unique profile of Firmicutes elevation and Bacteroidetes reduction. This mirrors prior work suggesting that Firmicutes dominance may modulate impulsivity and hyperactivity through altered SCFA signalling and dopamine turnover (Prehn-Kristensen et al., 2018). Additionally, changes in Bacteroidetes, noted for their role in polysaccharide degradation, may affect systemic energy availability and inflammatory tone (Wang et al., 2020).

Finally, the reviewed studies on PTSD and sleep disorders pointed to phylum-level losses in Verrucomicrobia, Lentisphaerae, and Actinobacteria, and genus-level gains in Akkermansia (RBD) and Bacteroides (insomnia). These changes may reflect altered circadian rhythms and stress-related neuroendocrine dysfunctions influencing microbial ecology (Nishiwaki et al., 2020b; Lin et al., 2024).

While causality remains elusive, the associations presented in our review support the growing rationale for the link between the microbiota and psychiatric disorders and the future potential of therapies directed at microbiota in psychiatry and the use of these dysbiosis patterns as potential biomarkers; something that still requires further high-quality investigation, with standardized protocols and different types of populations. Interventions such as probiotic supplementation, prebiotics, and faecal microbiota transplantation are currently under investigation, particularly in depression and ASD (Wallace and Milev, 2017; Liu R. T. et al., 2019; Slykerman et al., 2018). Early trials suggest improvements in mood, social behavior, and inflammation markers, though heterogeneity in strain specificity, delivery method, and patient phenotype remains a challenge (Kumar et al., 2024). The existing literature, while promising, underscores the necessity for further, more nuanced research to unravel the complexities of the microbiome’s contributions to mental health. This includes elucidating the mechanisms through which the microbiome influences brain function and mental well-being, understanding how interventions can be tailor-made to harness the microbiome for mental health benefits, and identifying specific microbial signatures that could predict response to treatment or the course of mental illnesses.

Advancing the diagnostic utility of gut microbiome research in psychiatry requires a systematic shift toward biomarker development, validation, and clinical translation. While current findings reveal disorder-specific microbial alterations, such as ↓ Firmicutes in ASD or ↑ Enterobacteriaceae in schizophrenia, these must now be leveraged to build clinically useful tools. First, there is a need for large-scale, longitudinal studies with harmonized protocols for sample collection, sequencing, and analysis. Diagnostic performance metrics (e.g., sensitivity, specificity, AUC) of candidate microbial biomarkers should be reported consistently. Multi-site cohorts with diverse geographic and ethnic representation will be essential to confirm generalisability. Second, the development of diagnostic classifiers based on machine learning algorithms trained on microbiota profiles represents a promising avenue. Early pilot studies suggest that microbial data alone, or in combination with symptom ratings and metabolomics, may predict psychiatric diagnoses with moderate-to-high accuracy. Future work should prioritise the refinement and external validation of these models. Third, integration with multi-omics platforms (metabolomics, proteomics, and host genomics) could enhance diagnostic resolution. For instance, combining gut microbiota data with microbial-derived metabolites such as short-chain fatty acids or neurotransmitter precursors may yield composite biomarkers with greater discriminative power. Finally, regulatory, ethical, and practical considerations must be addressed. Establishing standardised thresholds, reproducible pipelines, and clinical-grade assays will be crucial for transitioning microbiome-based diagnostics from research to psychiatric practice. Interdisciplinary collaboration across psychiatry, microbiology, data science, and regulatory science will be essential to realise this potential.

Interpreting the results of this systematic review should be accompanied by understanding its potential limitations. First, the predominance of cross-sectional designs in the included research restricts the capacity to properly establish causal links between gut microbiota changes and psychiatric diseases. Still rare are longitudinal and interventional investigations, which limit understanding of the temporal dynamics of microbiome changes in respect to disease development, progression, and treatment response. The moderate overall methodological quality of the included studies further constrains diagnostic inference. Given that many studies exhibited limitations in participant selection, confounder control, and microbiome assessment, proposed microbial signatures should be considered exploratory and hypothesis-generating rather than diagnostically actionable. This underscores the need for higher-quality, prospectively designed studies before microbiome-based diagnostic tools can be reliably developed. Additionally, the percentage-based summaries reported in this review are descriptive and intended to convey the relative recurrence of microbial alterations across heterogeneous studies. They do not constitute prevalence estimates or effect sizes and should be interpreted cautiously, particularly given variation in study design, sequencing methods, and reporting practices.

Second, substantial methodological heterogeneity is present across included studies. Sequencing approaches varied widely, with some studies employing 16S rRNA gene sequencing and others using shotgun metagenomic sequencing, resulting in differences in taxonomic resolution, sensitivity, and functional inference. In addition, reporting was inconsistent across taxonomic levels, with some studies providing phylum- or family-level data only, while others reported genus- or species-level alterations. This heterogeneity limits direct comparability of microbial findings across studies. Beyond technical variability, multiple clinical and environmental confounders were insufficiently controlled across studies. Psychotropic medication exposure, particularly antipsychotics and antidepressants, is known to independently alter gut microbiota composition and was variably reported or adjusted for. Similarly, body mass index, physical activity, lifestyle factors, and geographic and cultural background differed markedly across cohorts and were often incompletely documented. These variables are well-established determinants of gut microbial structure and likely contributed to inter-study heterogeneity and inconsistent findings. Given this degree of methodological, clinical, and environmental heterogeneity, a formal quantitative meta-analysis was not feasible. The lack of standardized sequencing platforms, inconsistent taxonomic reporting, variable outcome definitions, and insufficient adjustment for key confounders would have rendered pooled effect estimates unreliable and potentially misleading. Consequently, we adopted a structured descriptive synthesis to summarize recurring patterns of microbial alteration while avoiding overinterpretation of heterogeneous data.

Third, neither consistently controlled for nor reported dietary habits, drug use (especially psychiatric medication and antibiotics), nor concomitant medical disorders. The lack of standardizing in these variables presents possible uncontrolled confounding factors that might have affected the detected microbial signatures given their recognized effect on gut microbiota composition.

Fourth, psychiatric diagnosis itself remains clinically diverse despite the inclusion of only studies using DSM-IV, DSM-V, or comparable validated diagnostic methods; symptom severity, disease duration, and treatment history were often insufficiently documented; this lack of detail restricts the discovery of microbiome patterns linked with particular subtypes or phases of psychiatric disease. Certainly, dietary intake represents a major unaddressed confounding factor in the current literature on the gut-brain axis. Gut microbiome composition is highly sensitive to dietary patterns, including macronutrient distribution, fiber intake, consumption of ultra-processed foods, and intake of fermented products. Most studies included in this review did not systematically assess or control for participants’ dietary habits, which likely contributes substantially to inter-study heterogeneity and inconsistent microbial findings across psychiatric disorders. Consequently, some reported microbial alterations may reflect dietary effects rather than disorder-specific pathophysiology. Future studies should incorporate standardized dietary assessments, such as validated food frequency questionnaires, 24-h dietary recalls, or dietary pattern indices, and consider dietary stratification or adjustment in statistical models. Controlled feeding designs or run-in dietary standardization periods may further reduce confounding. Integrating dietary data with microbiome, metabolomic, and clinical phenotyping will be essential to disentangle diet-driven microbial variation from psychiatric disease-related signatures.

Lastly, few research involved different cohorts, therefore restricting the generalizability of results; most studies were carried out in particular regional or ethnic communities. The microbiome is clearly shaped by host genes, environment, and sociocultural elements, so, future research should aim to include more general demographic and geographic representation.

Taken together, these constraints draw attention to the need of standardized methods, bigger and more varied cohorts, and longitudinal multi-omics techniques to better grasp the function of the gut microbiome in psychiatric diseases and its potential for diagnostic and therapeutic innovation.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/. Supplementary material