Gut microbiota alterations in Alzheimer's disease and mild cognitive impairment: A systematic review and meta-analysis

Abstract

Background

Altered gut microbiota has been implicated in Alzheimer's disease (AD) and mild cognitive impairment (MCI), but findings across human studies are inconsistent.

Objective

To synthesize observational evidence on gut microbiota differences in older adults with AD or MCI compared with cognitively normal (CN) controls, and to assess the interpretive value of reported microbiome measures across disease stages.

Methods

We searched PubMed/MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Library for observational studies published from 1 January 2012 to 10 December 2025 reporting fecal microbiota profiles in AD, MCI, and CN adults (mean age ≥60). Random-effects meta-analysis was used for α-diversity; β-diversity and taxonomic changes were summarized narratively.

Results

Twenty-three studies were included (AD = 698, MCI = 485, CN = 1060). In AD versus CN, pooled α-diversity estimates showed no robust statistically significant differences (Shannon SMD = – 0.23, 95% CI: – 0.57 to 0.11; Chao1 SMD = –0.36, 95% CI: −0.74 to 0.02; ACE SMD = –0.38, 95% CI: −0.88 to 0.11). In MCI versus CN, differences were small and non-significant (Shannon SMD = –0.01, 95% CI: −0.18 to 0.15; Chao1 SMD = –0.17, 95% CI: −0.37 to 0.02). β-diversity and taxonomic findings more often suggested community-structure disruption and altered microbial composition in AD, whereas MCI findings were less consistent.

Conclusions

α-diversity is not supported as a useful biomarker for distinguishing AD or MCI from CN aging. Community-structure and taxonomic patterns may be more informative, but heterogeneity limits interpretation. Future studies should use standardized, function-oriented, and biomarker-informed approaches to clarify AD-continuum microbiome changes.

Keywords

α-diversity Alzheimer's disease β-diversity gut microbiota microbiome mild cognitive impairment

Introduction

Alzheimer's disease (AD) is the most common cause of dementia, and mild cognitive impairment (MCI) often represents an intermediate clinical stage on the continuum of cognitive decline. With an aging population, the number of people living with cognitive impairment (CI) continues to rise, reinforcing the need to clarify mechanisms and identify earlier risk indicators that are feasible to study in older adults.^1,2

Neuropathologically, AD is defined by extracellular amyloid-β (Aβ) plaques and intraneuronal neurofibrillary tangles composed of hyperphosphorylated tau, with stereotyped spread across brain regions as the disease progresses.^3–5 Pathogenesis is increasingly viewed as the product of interacting processes including Aβ dyshomeostasis, tau-mediated cytoskeletal disruption, neuroinflammation, oxidative stress, synaptic dysfunction, and vascular/blood–brain barrier (BBB) impairment that together drive neurodegeneration and clinical decline.^3,5,6 Contemporary research further emphasizes that biological changes may precede symptoms by years, highlighting a gap in which peripheral signals could complement existing biomarker strategies.⁵

Among these mechanisms, neuroinflammation has moved from a downstream epiphenomenon to a core contributor to AD progression. Activated microglia and innate immune signaling can influence Aβ clearance, synaptic remodeling, and tau phosphorylation, and may help shape the transition from prodromal impairment to dementia.⁶ Similarly, BBB dysfunction and neurovascular unit impairment may amplify central vulnerability by facilitating entry of peripheral inflammatory mediators and microbial products into the brain.^6,7 These immune-vascular interfaces provide a mechanistic bridge to peripheral systems that are amenable to measurement and, potentially, intervention.

In this context, the microbiota-gut-brain axis offers a biologically plausible route linking peripheral ecology to central neurodegeneration. Gut microbes generate metabolites, particularly short-chain fatty acids (SCFAs), and immune-active components that can regulate intestinal barrier function, BBB permeability, and neuroimmune tone.^8–11 Experimental studies indicate that microbiota-derived signals shape microglial maturation and responsiveness, and SCFAs can modulate microglial states and promote Aβ plaque deposition in transgenic models, supporting a potential causal pathway from gut ecology to amyloid pathology.^10,12 Conversely, dysbiosis-associated increases in pro-inflammatory taxa may contribute to endotoxemia; gram-negative bacterial molecules and lipopolysaccharide have been reported in association with AD neuropathology and may trigger microglial activation and downstream inflammatory cascades relevant to Aβ/tau biology.^13,14

Consistent with this mechanistic plausibility, human observational studies have reported gut microbiome alterations in AD and MCI, including shifts in community composition, reduced abundance of putative butyrate-producing taxa, and enrichment of taxa linked to pro-inflammatory profiles.^15–18 However, results remain heterogeneous across cohorts, likely reflecting differences in diagnostic criteria, comorbidity and medication exposure, diet and geography, sequencing regions/platforms, and analytic pipelines.^15–17 Importantly, many reports either combine AD and MCI or do not present stage-stratified quantitative summaries, limiting inference about when and how microbiome changes emerge along the cognitive impairment trajectory.

Two evidence gaps require attention in the field of clinical neurology. First, it remains unclear whether gut microbiome features can reliably distinguish AD from MCI at the level of within-sample diversity (α-diversity), between-sample community structure (β-diversity), and reproducible taxonomic shifts.^15–18 Second, despite the centrality of Aβ and tau biology, microbiome studies rarely integrate core AD biomarkers (e.g., cerebrospinal fluid (CSF) Aβ/tau or amyloid imaging) or assess whether microbial patterns track biomarker-defined disease biology, constraining mechanistic interpretation and clinical translation.^5,18

To address these gaps, we conducted a systematic review and meta-analysis of human observational studies in older adults examining gut microbiota in clinically diagnosed AD or MCI compared with cognitively normal controls. We quantitatively synthesized α-diversity indices, summarized β-diversity findings, and compared taxonomic patterns separately for AD and MCI where possible. By focusing on both reproducible signals and evidentiary limitations, this review aims to clarify the current interpretive value, and current limits, of gut microbiota measures in AD and MCI.^2,5

Methods

A research protocol was developed in advance which specified the research questions, inclusion criteria, findings and methods of analysis. The protocol was not prospectively registered in PROSPERO or other registries.¹⁹ This systematic review was reported in accordance with the PRISMA 2020 statement.²⁰ The full electronic search strategies for all databases are provided in Supplemental Table 1.

Search strategy

We searched PubMed/MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Library for studies published from 1 January 2012 to 10 December 2025. The search combined controlled vocabulary (e.g., MeSH and Emtree) and free-text terms for (1) cognitive impairment and dementia (e.g., Alzheimer*, mild cognitive impairment, dementia, cognitive decline, subjective cognitive decline) and (2) gut microbiota (e.g., gut microbiome, gut microbiota, intestinal microbiota, gut flora, and related synonyms). Boolean operators were applied using OR within concepts, AND between concepts. Reference lists of included studies and relevant reviews were screened to identify additional eligible reports.

Eligibility criteria

The research question was framed using an adapted PEOS approach (Population, Exposure, Outcome, Study design). We included observational studies (case-control, cross-sectional, or cohort) that compared older adults with AD or MCI against cognitively normal (CN) controls and reported fecal gut microbiota profiles measured by high-throughput sequencing. Eligible studies required a mean participant age of 60 years or older.

Population and cognitive status definitions

(1) AD: diagnosis based on established clinical criteria (e.g., NINCDS-ADRDA or NIA-AA recommendations, or DSM/ICD based clinical diagnosis as reported by the original study).²¹ (2) MCI: diagnosis based on established criteria (e.g., Petersen criteria or NIA-AA recommendations for MCI due to AD).^22,23 (3) CN: explicitly described as having no cognitive impairment, with confirmation by clinical assessment and/or validated instruments (e.g., Montreal Cognitive Assessment, Mini-Mental State Examination, Clinical Dementia Rating) as reported by the original study.

Exposure

Eligible studies had to report at least one of the following: (1) α-diversity indices (e.g., Shannon, Chao1, ACE, Simpson or inverse Simpson), (2) β-diversity metrics (e.g., Bray-Curtis, Jaccard, weighted or unweighted UniFrac, or Aitchison distance), or (3) relative abundance of specific taxa at the phylum, family, or genus level.

Outcome

The primary outcomes were differences in α-diversity between AD or MCI and CN controls. Secondary outcomes included β-diversity patterns (community composition differences) and taxonomic composition differences (relative abundance shifts) between groups.

Study design

We excluded intervention trials, animal studies, reviews, protocols, conference abstracts, and studies without extractable quantitative data for the prespecified outcomes. Studies were excluded when participants had recent antibiotic exposure or other major microbiome-altering treatments around sampling (as defined by the original study) or when major clinical conditions likely to confound gut microbiota and cognition were clearly present and not addressed analytically (as reported by the original study).

Study selection

All records were exported to EndNote (Version 21) for reference management and deduplication. Two reviewers independently screened titles and abstracts and then evaluated full texts against the eligibility criteria. Disagreements were resolved by discussion, with adjudication by a third reviewer when needed.

Data extraction

Two reviewers independently extracted data using a standardized form. Extracted items included study characteristics (first author, year, country or region, design), participant characteristics (sample size, age, sex distribution, diagnostic criteria), and microbiome methods (sample type, sequencing strategy, targeted 16S rRNA region where applicable, and bioinformatic pipeline features such as OTU-based versus ASV-based processing). Detailed data are provided in Supplemental Tables 2 and 3.

For α-diversity outcomes, we preferentially extracted means and standard deviations (SDs) for Shannon, Chao1, ACE, Simpson and inverse Simpson indices. If outcomes were reported as medians with interquartile ranges (IQRs) or ranges, we converted them to means and SDs using the method of Wan et al.²⁴ When necessary, SDs were derived from standard errors, confidence intervals, or p-values using standard formulas. Inverse Simpson was synthesized separately when directly extractable data were available, whereas studies reporting a metric labeled as Simpson without consistent specification of the mathematical formulation were summarized narratively. Exploratory subgroup analyses were also prespecified for selected study-level characteristics, including region and sequencing/processing strategy, to assess potential contributors to between-study heterogeneity in α-diversity estimates.

For β-diversity, we extracted the distance metric, ordination method, and global test (e.g., PERMANOVA or ANOSIM), together with reported statistics and study conclusions. For taxonomic composition, we extracted the direction of between-group differences (enriched or depleted) at phylum, family, and genus levels and any available statistics (e.g., LEfSe LDA scores).

Biomarker availability statement: We prespecified the extraction of AD biomarkers, including cerebrospinal fluid Aβ and tau, when available. All included studies were assessed for the availability of these biomarker data, but no eligible study provided extractable quantitative biomarker results (e.g., concentrations or subtype ratios) suitable for synthesis. Therefore, no biomarker-related extraction or analyses were performed.

Quality assessment

Methodological quality and risk of bias were assessed using the Newcastle-Ottawa Scale (NOS) for nonrandomized studies.²⁵ For cross-sectional studies, we applied a prespecified NOS adaptation with comparable domains (selection, comparability, and outcome or exposure). Two reviewers performed assessments independently, and discrepancies were resolved by consensus. NOS ratings were used to support interpretation and planned sensitivity analyses rather than as an exclusion criterion.

Data synthesis and statistical analysis

Meta-analyses were conducted in R (Version 4.5.1) using the meta package (Version 8.2–1).²⁶ For each α-diversity index, standardized mean differences (SMDs) with 95% confidence intervals (CIs) were calculated to compare AD versus CN and MCI versus CN. Random-effects models were used as the primary approach to account for anticipated clinical and methodological heterogeneity, with between-study variance (tau-squared) estimated via restricted maximum likelihood.

To provide conservative inference under random-effects models, we applied the Hartung-Knapp-Sidik-Jonkman (HKSJ) method to construct 95% CIs for pooled effects.²⁷ In meta, this was implemented using rma.uni with REML tau-squared estimation and Knapp-Hartung type inference (e.g., test = “knha”). Statistical heterogeneity was assessed using Cochran's Q and quantified using the I-squared statistic.²⁸

Sensitivity and exploratory analyses

We performed leave-one-out analyses to evaluate whether pooled estimates were driven by any single study. Fixed-effect inverse-variance models were considered as a sensitivity analysis when heterogeneity was negligible, to assess robustness to model choice.²⁹ A pooled cognitive impairment analysis combining AD and MCI was conducted as an exploratory supplementary analysis to provide an overall estimate across the cognitive impairment continuum, recognizing that stage mixing may increase heterogeneity and should not be interpreted as a stage-specific effect.

Reporting bias

When at least 10 studies contributed to an outcome, small-study effects were assessed using funnel plots and Egger's regression test.³⁰ As an exploratory adjustment, we applied the trim-and-fill method to estimate the potential influence of missing studies on pooled effects.³¹ All statistical tests were two-sided with p < 0.05 considered statistically significant.

Qualitative synthesis of β-diversity and taxonomic composition

Because β-diversity and taxonomic composition outcomes were reported heterogeneously in terms of distance metrics, ordination, differential abundance pipelines, and statistical reporting, pooled quantitative synthesis was generally not appropriate. We therefore conducted a structured narrative synthesis. For each comparison (AD versus CN; MCI versus CN), we extracted the comparison groups and sample sizes, distance metric, ordination method, global test (e.g., PERMANOVA or ANOSIM), reported statistics, whether covariate adjustment was incorporated in the analysis when stated, and the main finding as reported by the original study authors.

Given that sequencing-derived relative abundance data are compositional, we interpreted differential abundance findings cautiously and prioritized results supported by appropriate compositional or multivariable analyses when reported.³²

Results

Study selection progress

The searches identified 19,739 records. After deduplication and title/abstract screening, 63 full texts were assessed for eligibility. Twenty-three studies met the inclusion criteria and were included in the systematic review and meta-analysis. The screening and selection process is summarized in the PRISMA 2020 flow diagram (Figure 1).

Figure 1.

PRISMA flow diagram illustrating the screening and selection process of studies.

Study characteristics

The 23 observational studies were published between 2017 and 2025. Most were conducted in China (12/23), with additional studies from the USA and other regions. Eighteen studies were case-control and five were cross-sectional (Table 1).

Table 1.

Characteristics of included studies.

Study	Country	Study types	Experimental group				Control group			NOS
Study	Country	Study types		N	Mean age (SD)	Female	N	Mean age (SD)	Female
Vogt 2017¹⁵	USA	Case-control	AD	25	71.3 (7.3)	68%	25	69.3 (7.5)	72%	8
Zhuang 2018¹⁶	China	Case-control	AD	43	70.12 (8.78)	46.5%	43	69.72 (9.24)	46.5%	8
Haran 2019³³	USA	Case-control	AD	24	84.7 (8.1)	83.3%	51	83 (10.2)	84.3%	7
Li 2019³⁴	China	Case-control	AD	30	66.3 (5.1)	50.0%	30	63.9 (5.1)	56.7%	8
Li 2019³⁴	China	Case-control	MCI	30	65.4 (7.6)	60.0%	30	63.9 (5.1)	56.7%	8
Liu 2019¹⁷	China	Case-control	AD	33	74.85(11.37)	42.4%	32	76.88 (9.35)	50.0%	8
Liu 2019¹⁷	China	Case-control	MCI	32	70.53 (11.0)	56.3%	32	76.88 (9.35)	50.0%	8
Xi 2021³⁵	China	Case-control	AD	21	76.2 (9.9)	38.10%	44	78.4 (6.6)	54.6%	8
Ueda 2021³⁶	Japan	Case-control	AD	7	83.0 (9.3)	71.40%	21	72.7 (5.3)	61.9%	7
Ueda 2021³⁶	Japan	Case-control	MCI	15	79.8 (6.1)	60%	21	72.7 (5.3)	61.9%	7
Zhou 2021³⁷	China	Case-control	AD	60	72.82 (7.25)	60%	32	71.06 (5.92)	56.3%	8
Ling 2021³⁸	China	Case-control	AD	100	74.14 (9.21)	57%	71	73.11 (7.75)	50.7%	8
Hou 2021³⁹	China	Case-control	AD	30	71.9 (6.9)	43.3%	47	71.1 (6.7)	53.2%	8
Guo 2021⁴⁰	China	Case-control	AD	18	63.5 (4.7)	88.9%	18	64.2 (4.7)	77.8%	8
Guo 2021⁴⁰	China	Case-control	MCI	20	64.5 (4.5)	80.0%	18	64.2 (4.7)	77.8%	8
Duan 2021⁴¹	China	Case-control	MCI	18	71.22 (3.98)	88.9%	36	70.78 (4.41)	72.2%	7
Liu 2021⁴²	China	Case-control	MCI	20	68.8 (11.2)	40.0%	22	72.7 (8.05)	59.1%	8
Pan 2021⁴³	China	Case-control	MCI	22	71.45 (8.03)	63.6%	26	67.31 (5.27)	73.1%	8
Yıldırım 2022⁴⁴	Turkey	Case-control	AD	47	71.4 (5.1)	48.9%	51	67 (5.3)	45.0%	8
Yıldırım 2022⁴⁴	Turkey	Case-control	MCI	27	69.2 (6.4)	40.0%	51	67 (5.3)	45.0%	8
Pei 2023⁴⁵	China	Cross-sectional	AD	45	71.67 (8.33)	64.4%	35	68.51 (8.19)	54.3%	7
Pei 2023⁴⁵	China	Cross-sectional	MCI	38	69.08 (7.16)	60.5%	35	68.51 (8.19)	54.3%	7
Klee 2024⁴⁶	Luxembourg	Cross-sectional	MCI	58	67.6 (9)	25.9%	200	63.76 (7.84)	43.5%	8
Cabrera 2025⁴⁷	Spain	Cross-sectional	AD	31	74.61 (5.60)	48.4%	38	68.21 (5.99)	50.0%	7
Cabrera 2025⁴⁷	Spain	Cross-sectional	MCI	30	72.13 (5.18)	43.3%	38	68.21 (5.99)	50.0%	7
Hosseinkhani 2025⁴⁸	Netherlands, Switzerland and Italy	Cross-sectional	AD	17	69 (8)	47.1%	86	63 (10)	66.3%	7
Hosseinkhani 2025⁴⁸	Netherlands, Switzerland and Italy	Cross-sectional	MCI	111	71 (8)	45.9%	86	63 (10)	66.3%	7
Lwere 2025⁴⁹	Uganda	Case-control	AD	77	76.0 (70,83)^a	81.8%	13	70.0 (69,71)^a	69.2%	8
Lwere 2025⁴⁹	Uganda	Case-control	MCI	14	76.5 (68,83)^a	78.6%	13	70.0 (69,71)^a	69.2%	8
Rout 2025⁵⁰	India	Case-control	AD	40	66.2 (60–81)^b	27.5%	40	69.3 (60–85)^b	40.0%	8
Rouskas 2025⁵¹	Greece	Cross-sectional	MCI	50	68 (9.8)	62.0%	49	62.7 (8.6)	69.4%	8
Sepúlveda-Rivera 2025⁵²	USA	Case-control	AD	50	55–65 (6%) 66–76 (38%) >77 (56%)^c	70.0%	50	55–65 (32%) 66–76 (54%) >77 (14%)^c	72.0%	8

AD: Alzheimer's disease; MCI: mild cognitive impairment; CN: cognitively normal; NOS: Newcastle-Ottawa Scale. Age is reported as mean (SD) unless otherwise stated.

Age Median (Interquartile Range).

Continuous Mean (Range).

Age Range.

Across studies, there were 698 participants with AD, 485 with MCI, and 1060 CN controls. Individual study sample sizes ranged from 7 to 100 AD cases, 14 to 111 MCI cases, and 13 to 200 controls. Participant ages and sex distributions were broadly comparable between case and control groups when reported. Detailed information relevant to between-study comparability, including clinical/diagnostic characteristics and microbiome laboratory and analytic pipeline features, is provided in Supplemental Tables 2 and 3.

Study quality was generally high. NOS scores ranged from 7 to 8, with common limitations relating to comparability, including incomplete adjustment for key confounders, and heterogeneity in exposure and outcome ascertainment. Detailed item-level assessments are provided in Supplemental Table 4.

AD pathological biomarker data availability. All included studies were evaluated for the availability of CSF Aβ and tau biomarkers. However, none of the eligible studies reported extractable quantitative biomarker data (e.g., concentrations or subtype ratios). Therefore, no biomarker-related extraction or synthesis was performed in this review.

Meta-analysis of gut microbiota α-diversity

Mild cognitive impairment versus CN

For Shannon diversity, no difference was observed between MCI and CN (SMD = −0.01, 95% CI: −0.18 to 0.15; p = 0.296; I² = 17.0%). For Chao1 richness, MCI did not differ significantly from CN (SMD = −0.17, 95% CI: −0.37 to 0.02; p = 0.172; I² = 37.4%). Data were insufficient for a robust ACE meta-analysis in MCI (Figure 2).

Figure 2.

Forest plots of α-diversity in mild cognitive impairment versus cognitively normal controls: (A) Shannon index, (B) Chao1 index.

Alzheimer's disease versus CN

No significant difference was found in Shannon diversity (SMD = −0.23, 95% CI: −0.57 to 0.11), with substantial heterogeneity (I² = 61.7%, p = 0.008). For Chao1 and ACE, the results were −0.36 (95% CI: −0.74 to 0.02; I² = 50.7%) and −0.38 (95% CI: −0.88 to 0.11; I² = 61.5%), respectively. Overall, richness-based indices tended lower in AD. However, all pooled estimates were imprecise, and their confidence intervals crossed the null. No statistically significant between-group differences were observed (Figure 3).

Figure 3.

Forest plots of α-diversity in Alzheimer's disease versus cognitively normal controls: (A) Shannon index, (B) Chao1 index, (C) ACE index.

Exploratory pooled cognitive impairment (AD and MCI) versus CN

As an exploratory supplemental analysis, studies of AD and MCI were pooled to estimate an overall CI versus CN difference. Shannon diversity showed no statistically significant difference between CI and CN (SMD = −0.11, 95% CI: −0.30 to 0.09), with moderate heterogeneity (I² = 51.7%, p = 0.007). In contrast, Chao1 richness was lower in CI than CN (SMD = −0.27, 95% CI: −0.50 to −0.04), with moderate heterogeneity (I² = 46.5%, p = 0.044). Overall, across the CI spectrum, richness-based indices suggested a modest reduction in estimated species richness, whereas the Shannon index did not show a consistent difference (Figure 4).

Figure 4.

Forest plots of α-diversity in cognitive impairment versus cognitively normal controls: (A) Shannon index, (B) Chao1 index.

Evenness-related analysis

Several studies reported a metric labeled as Simpson; however, its mathematical formulation was not consistently specified and reporting formats were heterogeneous. These findings were summarized narratively rather than quantitatively pooled. In AD, Liu 2019 and Ling 2021 showed numerically higher Simpson values in controls than in AD, whereas Zhou 2021 showed a small difference in the opposite direction. In MCI, Liu 2019 and Liu 2021 showed higher Simpson values in MCI than in controls, while Pan 2021 showed no apparent difference. Overall, these Simpson-based findings did not support a consistent stage-specific pattern. By contrast, inverse Simpson index was explicitly reported in two MCI studies, and exploratory supplementary meta-analysis showed no significant difference between MCI and cognitively normal controls (Supplemental Figure 1).

Subgroup and sensitivity analyses

Prespecified subgroup analyses (publication period, sample size, geographic region, and sequencing strategy) did not materially change the direction of pooled estimates. These findings suggest that study context and microbiome processing strategy may have contributed to variability across studies, while also indicating that no single prespecified factor fully explained the observed inconsistency. Full subgroup results are provided in Supplemental Figures 2 and 3. Leave-one-out analyses were undertaken to assess whether pooled estimates were driven by any single study. Across outcomes, results were broadly stable, and no individual study qualitatively altered the direction of the main pooled estimates. Leave-one-out diagnostics are reported in Supplemental Figures 4 and 5.

Small-study effects and publication bias

Funnel plots were inspected where sufficient studies were available. Egger's regression test was only considered when at least 10 studies contributed to a pooled outcome. For several outcomes, the number of contributing studies was limited, reducing the reliability of formal tests. Overall, there was no strong indication of pronounced small-study effects on visual inspection, but publication bias cannot be ruled out. Funnel plots are reported in Supplemental Figure 6.

Descriptive systematic review of β-diversity results

β-diversity was assessed using a range of distance metrics and ordination approaches, most commonly principal coordinates analysis (PCoA) with Bray-Curtis dissimilarity and/or UniFrac distances. Global between-group differences were typically tested using PERMANOVA. Table 2 summarizes group sample sizes, ordination, β-diversity metric, formal statistical testing, covariate adjustment, and the original study authors’ main findings. Overall, AD versus CN comparisons more often showed global separation than comparisons involving MCI.

Table 2.

Overview of β-diversity analyses and main findings.

Study	Compare groups (N)	Ordination	β-diversity	Statistical testing	Covariate adjustment	Main finding
Vogt 2017¹⁵	AD 25 vs. CN 25	NMDS	Bray-Curtis dissimilarity and UniFrac analysis (weighted and unweighted)	PERMANOVA, Bray-Curtis: F = 2.87 (p < 0.001), weighted UniFrac: F = 3.84 (p < 0.001); unweighted UniFrac: F = 2.60 (p < 0.005)	NR	Compositional differences in the microbiome between AD and Control groups.
Zhuang 2018¹⁶	AD 43 vs. CN 43	PLS-DA	Weighted UniFrac	NR	NR	Fecal microbial communities separated between AD and control group on family, genus, species and OTU level, suggesting a unique gut microbiota in AD
Haran 2019³³	AD 24 vs. CN 51	t-SNE	Jaccard	PERMANOVA (p = 0.001)	NR	Elders with AD cluster away from those without dementia.
Liu 2019¹⁷	AD 33 vs. CN 32	PCoA	Bray-Curtis, weighted UniFrac, and unweighted UniFrac	PERMANOVA for Bray-Curtis R = 0.182 (p = 0.017); weighted UniFrac R = 0.184 (p = 0.067); unweighted UniFrac R = 0.134 (p = 0.183)	NR	Significant compositional differences between AD and CN groups in terms of Bray-Curtis but not weighted or unweighted UniFrac.
	AD 33 vs. MCI 32	PCoA	Bray-Curtis, weighted UniFrac, and unweighted UniFrac	PERMANOVA for Bray-Curtis R = 0.197 (p = 0.005); weighted UniFrac R = 0.274 (p = 0.003); unweighted UniFrac R = 0.148 (p = 0.044)	NR	Significant compositional differences between AD and MCI groups.
	MCI 32 vs. CN 32	PCoA	Bray-Curtis, weighted UniFrac, and unweighted UniFrac	PERMANOVA for Bray-Curtis R = 0.176 (p = 0.012); weighted UniFrac R = 0.2226 (p = 0.01); unweighted UniFrac R = 0.138 (p = 0.138)	NR	Significant compositional differences between MCI and CN groups in terms of Bray-Curtis and the weighted but not unweighted UniFrac.
Li 2019³⁴	AD 30 vs. MCI 30 vs. CN 30	PCA	Weighted UniFrac	ANOSIM R = 0.345 (p = 0.001)	NR	Significant differences among the three groups; no significant difference between AD and MCI.
Li 2019³⁴	AD 30 vs. MCI 30 vs. CN 30	PCA	Unweighted UniFrac	ANOSIM R = 0.22 (p = 0.001)	NR	Significant differences among the three groups; no significant difference between AD and MCI.
Xi 2021³⁵	AD 21 vs. CN 44	PCoA	Bray-Curtis	PERMANOVA, R²= 0.025 (P = 0.027)	NR	Significant differences between the AD group and the CN group.
Zhou 2021³⁷	AD 60 vs. CN 32	PCoA	Unweighted, and weighted UniFrac	PERMANOVA, unweighted UniFrac: AD versus HC: R² = 0.029 (p < 0.001); weighted UniFrac: AD versus HC: R² = 0.026 (p = 0.026)	NR	Significant differences between the AD group and the CN group.
Ling 2021³⁸	AD 100 vs. CN 71	PCoA	Jaccard, Bray-Curtis, weighted and unweighted UniFrac	PERMANOVA (p < 0.01)	NR	AD and CN were divided into different clusters.
Hou 2021³⁹	AD 30 vs. CN 47	PCoA	Bray-Curtis	ANOSIM R² = 0.028 (p = 0.039)	NR	Slight difference in gut microbial composition between groups.
			Weighted UniFrac	ANOSIM R² = 0.025 (p = 0.065)	NR	Hardly revealed distinguishable bacterial microbiota of AD subjects compared to CN.
			Unweighted UniFrac	ANOSIM R² = 0.023 (p = 0.233)	NR	Hardly revealed distinguishable bacterial microbiota of AD subjects compared to CN.
Guo 2021⁴⁰	AD 18 vs. CN 18	PCA	Bray curitis	p = 0.016	NR	Significant difference between AD, MCI and CN.
	MCI 20 vs. CN 18		Bray curitis	p = 0.017	NR
	AD 18 vs. CN 18		Weighted UniFrac	p < 0.001	NR
	MCI 20 vs. CN 18		Weighted UniFrac	p < 0.001	NR
Ueda 2021³⁶	AD 21 vs. MCI 15 vs. CN 7	PCoA	Bray-Curtis	PERMANOVA between MCI and CN groups, R²= 0.0465 (p = 0.0968); between AD and CN groups R² = 0.0534 (p = 0.1423)	NR	Genus's composition had a tendency to differ between the healthy and the MCI group but not between the healthy and the AD group.
Pan 2021⁴³	MCI 22 vs. CN 26	PCoA, NMDS	Weighted UniFrac	PERMANOVA (p = 0.048)	NR	Gut microbiota profiles of the MCI cases clustered apart from CN.
Duan 2021⁴¹	MCI 18 vs. CN 36	PCA, PCoA	Bray-Curtis	PCA ANOSIM R² = 0.0375 (p = 0.026); PCoA ANOSIM R² = 0.0416 (p = 0.004)	NR	Clearly different among the three groups.
Yıldırım 2022⁴⁴	AD 47 vs. MCI 27 vs. CN 51	PCoA	Bray-Curtis and Jaccard indices	PERMANOVA (p = 0.0001)	Age and education considered in analyses; broader downstream models included additional covariates	Highly significant separation of the three groups.
Yıldırım 2022⁴⁴	AD 47 vs. MCI 27 vs. CN 51	PCA	Aitchison	PERMANOVA (p = 0.02)	Age and education considered in analyses; broader downstream models included additional covariates	Three groups are distinct.
Pei 2023⁴⁵	AD 45 vs. MCI 38 vs. CN 35	PCA	Bray-Curtis	PCA, p = 0.034; Bray Curtis, p = 0.019	NR	There were significant differences among the three groups.
Lwere 2025⁴⁹	AD 77 vs. MCI 14 vs. CN 13	PCoA, CAP	Weighted UniFrac distances	PERMANOVA	Age and BMI assessed in downstream analyses	Distinct clustering patterns among AD, MCI, and control groups.
Lwere 2025⁴⁹	AD 77 vs. MCI 14 vs. CN 13	PCoA, CAP	Bray-Curtis	PERMANOVA	Age and BMI assessed in downstream analyses	With a pronounced distinction between the AD and CN, and MCI partially overlapped with both AD and Control groups.
Rout 2025⁵⁰	AD 40 vs. CN 40	PCoA	Bray-Curtis	PERMANOVA (p = 0.009)	NR	There were significant differences between the CN and the AD group.
Sepúlveda-Rivera 2025⁵²	AD 50 vs. CN 50	NMDS	Bray-Curtis	PERMANOVA (p = 0.034), ANOSIM (p = 0.061) PERMDISP	Main multivariable taxonomic models adjusted for BMI, age range, and gender; β-diversity global tests were run in the QIIME2 workflow	There was no statistically significant difference in species composition between the two groups.
Hosseinkhani 2025⁴⁸	AD 17 vs. MCI 111 vs. CN 86	PCoA	Weighted UniFrac	NR	NR	No significant compositional differences between groups.
Cabrera 2025⁴⁷	AD 31 vs. MCI 30 vs. CN 38	PCoA, NMDS	Bray–Curtis	PERMANOVA (F = 1.29, R² = 0.026, P = 0.219); ANOSIM (R = 0.037, P = 0.059)	Age and education adjusted in PERMANOVA; broader models included sex, BMI, APOE ε4, and medication use	No significant differences across groups.
Rouskas 2025⁵¹	MCI 50 vs. CN 49	PCoA	Bray–Curtis	PERMANOVA (R² = 0.013, P = 0.054)	PERMANOVA accounted for selected covariates; sequencing platform also showed technical effects	No differences between healthy controls and individuals with MCI.

AD: Alzheimer's disease; aMCI: amnestic mild cognitive impairment; CN: cognitively normal; MCI: mild cognitive impairment; SCD: subjective cognitive decline; NR: not reported; PCoA: principal coordinates analysis; NMDS: non-metric multidimensional scaling; PLS-DA: partial least squares-discriminant analysis; PERMANOVA: permutational multivariate analysis of variance; ANOSIM: analysis of similarities.

Main finding refers to the interpretation reported by the original study authors rather than the conclusions of the present systematic review. NR indicates that the relevant information was not reported in the source article.

Across studies reporting multiple metrics, a tentative pattern emerged. In both AD and MCI comparisons, significant separation was more often observed with Bray–Curtis and, in several cohorts, weighted UniFrac, whereas unweighted UniFrac and other presence/absence-oriented metrics were less consistently discriminatory. This pattern suggests that reported between-group differences more often reflected shifts in the relative abundance of shared taxa, and in some datasets phylogenetically structured community differences, rather than a uniform signal driven by the presence or absence of rare taxa alone. However, this interpretation remains provisional because metric selection, pairwise versus multi-group testing, preprocessing choices, and covariate handling varied substantially across studies. Accordingly, β-diversity findings were interpreted narratively rather than quantitatively pooled.

Taxonomic composition at phylum, family, and genus levels

Taxonomic findings were summarized separately for AD and MCI to maintain stage-specific interpretation (Figure 5A, B). In AD versus CN, differences involved multiple phyla, particularly Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria, with recurrent alterations also reported at the family and genus levels in taxa related to Lachnospiraceae, Ruminococcaceae, Enterobacteriaceae, Bifidobacteriaceae, Faecalibacterium, Roseburia, Bacteroides, Blautia, and Bifidobacterium.^{15,17,34,35,37,38,47,50} In MCI versus CN, taxonomic findings were fewer and less coherent, with reported shifts involving major phyla and selected family/genus-level taxa, but without a stable directional pattern across studies.^{17,34,36,40,43,45,47,49} Overall, AD showed a broader and denser pattern of reported taxonomic alterations, whereas MCI displayed a sparser and more heterogeneous profile. These results support phenotype-stratified interpretation rather than pooled cognitive-impairment summarization. Given substantial variation in sequencing and analytical methods, together with limited reproducibility across studies, these taxonomic findings should be regarded as hypothesis-generating rather than reproducible stage-specific biomarkers.

Figure 5.

Stage-stratified heatmaps of reported gut microbiota alterations at the phylum, family, and genus levels. (A) Alzheimer's disease versus cognitively normal controls. (B) Mild cognitive impairment versus cognitively normal controls.

Discussion

This review shows that current gut microbiota measures have limited interpretive value across the AD continuum. Pooled α-diversity results were weak and inconsistent. In AD, richness-related indices were generally lower than in cognitively normal controls, but estimates were imprecise and confidence intervals generally crossed the null. In MCI, pooled α-diversity differences were small and not statistically significant. By contrast, β-diversity and some taxonomic findings appeared more informative, particularly when AD and MCI were interpreted separately. These results support stage-specific interpretation. They also indicate that α-diversity alone is unlikely to provide a useful standalone microbiome biomarker.

The findings should be interpreted cautiously because heterogeneity was substantial. Study populations, diagnostic criteria, exclusion rules, comorbidity control, medication reporting, sequencing platform, targeted 16S region, OTU or ASV processing, and downstream statistical methods varied across studies. These differences likely contribute to inconsistent α-diversity estimates and limited reproducibility of taxonomic findings. Exploratory subgroup analyses did not materially change the direction of pooled estimates. Several important sources of heterogeneity, particularly diet, medication burden, and comorbidity burden, were reported inconsistently and could not be formally modeled across the full evidence base.

A recurring pattern was lower α-diversity or richness in AD compared with CN, whereas results for MCI were less consistent. This stage contrast is plausible. MCI is clinically heterogeneous and includes individuals with multiple etiologies and trajectories. By contrast, clinically diagnosed AD is more likely to reflect advanced neurodegenerative and systemic changes that could influence gut ecology. At the same time, α-diversity indices are sensitive to sequencing depth, 16S region choice, preprocessing and rarefaction, which limits comparability across studies.⁵³ Taxonomic findings were also variable across cohorts. Several studies reported differences in taxa linked to SCFA production and immune regulation in AD.^15,16 Since 16S amplicon data are compositional, these findings should not be over-interpreted as direct functional deficits. The current taxonomic results are better viewed as hypothesis-generating patterns. Their clinical applicability remains limited because reproducibility across cohorts is still insufficient.^9,54

Experimental work suggests plausible links between the microbiota and AD-related processes. These include microbial metabolites, BBB integrity, microglial regulation, systemic inflammation, and bacterial products such as lipopolysaccharide.^{8,10,12,13,18} However, human evidence remains insufficient for mechanistic inference. It is still unclear whether dysbiosis is a driver, a consequence, or an epiphenomenon of neurodegeneration and aging-related comorbidities.^9,54

Another gap is the limited ability to separate MCI-specific microbial patterns from those observed in established AD. This reflects both statistical and clinical heterogeneity. MCI definitions, cognitive testing and comorbidity profiles vary widely across settings. Large, deeply phenotyped cohorts will be needed to test whether microbiome signatures track clinical progression and whether they add predictive value beyond established risk factors. Improved reporting and harmonized study design will also be important for improving comparability across cohorts. The STORMS checklist provides a practical reporting framework for future human microbiome studies.⁵³

Strengths and limitations

This review has several strengths. It followed a prespecified protocol without deviations and used established reporting standards for systematic reviews and observational evidence.^20,55 Conservative random-effects inference was applied with the Hartung-Knapp-Sidik-Jonkman approach.²⁷ Analyses were conducted with transparent meta-analytic tools. AD and MCI were also examined separately, which is important because progression and underlying pathology may differ across stages.

Several limitations should be noted. Most included studies were observational and often cross-sectional, which preclude causal inference and limits conclusions about directionality. Residual confounding by diet, metabolic health, frailty, medication exposure and other lifestyle factors is likely. Outcome reporting was inconsistent, which restricted quantitative pooling for several endpoints, particularly β-diversity and multivariate ordination results.

Integration with core AD biomarkers was limited. Although some studies have linked gut microbiota to amyloid- or tau-related markers, biomarker-informed phenotyping remains limited and inconsistent.^56,57 No extractable quantitative Aβ or tau data were available for synthesis in the present review. The findings are therefore more informative for clinically diagnosed AD and MCI than for biologically defined AD and cannot yet be reliably interpreted within an AT(N)-based framework.⁵

Implications and future directions

Progress toward mechanistic inference and translation will require designs that better address confounding and reverse causation, including prospective cohorts and, where appropriate, intervention studies. Priority areas include harmonized clinical phenotyping across the AD continuum, biomarker-informed staging, and functional readouts such as metabolomics and inflammatory profiling.^{8–10,12,13,18,54} Current evidence supports continued investigation of the microbiota–gut–brain axis in cognitive aging, but immediate clinical applicability remains limited. Future progress will require longitudinal, standardized, and biomarker-informed studies that integrate microbiome profiling with functional readouts, including metabolomic and inflammatory measures.^53,54

Conclusion

This systematic review and meta-analysis indicate that gut microbiota alterations are present in older adults with cognitive impairment, but their current interpretive value remains limited. α-diversity results did not provide a robust signal for distinguishing AD or MCI from cognitively normal aging. β-diversity more often suggested community-structure disruption in AD, whereas findings in MCI were less consistent. At the taxonomic level, AD more often showed depletion of putatively beneficial taxa such as Faecalibacterium and Roseburia and enrichment of potentially harmful taxa such as Escherichia, although reproducibility remained limited. Overall, current findings are more useful for hypothesis generation than for direct clinical application. Future research should prioritize longitudinal, biomarker-informed staging, and function-oriented design to clarify microbiome changes across the AD continuum.

Supplemental Material

sj-docx-1-alz-10.1177_13872877261465913 - Supplemental material for Gut microbiota alterations in Alzheimer's disease and mild cognitive impairment: A systematic review and meta-analysis

Supplemental material, sj-docx-1-alz-10.1177_13872877261465913 for Gut microbiota alterations in Alzheimer's disease and mild cognitive impairment: A systematic review and meta-analysis by Yujia Liu, Xiuru Wang, Weilin Huang, Rui Li, Hanxiang Zhang, Songhaer Jiakesileke, Ping Zhang, Jindong Ding Petersen and Wenting Cao in Journal of Alzheimer's Disease

Footnotes

Acknowledgements

The authors have no acknowledgments to report.

ORCID iDs

Yujia Liu

Weilin Huang

Hanxiang Zhang

Ethical considerations

Not applicable

Consent to participate

Not applicable

Consent for publication

Not applicable

Author contribution(s)

Yujia Liu: Conceptualization; Formal analysis; Methodology; Writing – original draft.

Xiuru Wang: Conceptualization; Software; Writing – original draft.

Weilin Huang: Formal analysis; Methodology; Writing – original draft.

Rui Li: Formal analysis; Resources; Writing – original draft.

Hanxiang Zhang: Data curation; Writing – original draft.

Songhaer Jiakesileke: Data curation; Resources; Writing – original draft.

Ping Zhang: Funding acquisition; Supervision; Writing – review & editing.

Jindong Ding Petersen: Funding acquisition; Supervision; Writing – review & editing.

Wenting Cao: Funding acquisition; Supervision; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Natural Science Foundation of China (Grant No. 82360673), the Natural Science Foundation of Hainan Province (Grant No. 824RC518 and No. 821RC583), the Academic Enhancement Program of Hainan Medical University (Grant No. XSTS2025003 and No. XSTS2025130).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data used in the current study are available from the corresponding author on reasonable request.

Supplemental material

Supplemental material for this article is available online.

References

Nichols

Steinmetz

Vollset

, et al. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the global burden of disease study 2019. Lancet Public Health 2022; 7: e105–e125.

Livingston

Huntley

Liu

, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet Standing Commission. Lancet 2024; 404: 572–628.

Hardy

Selkoe

. The amyloid hypothesis of Alzheimer’s disease: progress and problems on the road to therapeutics. Science 2002; 297: 353–356.

Braak

. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 1991; 82: 239–259.

Jack

Bennett

Blennow

, et al. NIA-AA research framework: toward a biological definition of Alzheimer’s disease. Alzheimers Dement 2018; 14: 535–562.

Heneka

Carson

El Khoury

, et al. Neuroinflammation in Alzheimer’s disease. Lancet Neurol 2015; 14: 388–405.

De Strooper

Karran

. The cellular phase of Alzheimer’s disease. Cell 2016; 164: 603–615.

Braniste

Al-Asmakh

Kowal

, et al. The gut microbiota influences blood-brain barrier permeability in mice. Sci Transl Med 2014; 6: 263ra158.

Cryan

Dinan

. Mind-altering microorganisms: the impact of the gut microbiota on brain and behaviour. Nat Rev Neurosci 2012; 13: 701–712.

10.

Erny

de Angelis

ALH

Jaitin

, et al. Host microbiota constantly control maturation and function of microglia in the CNS. Nat Neurosci 2015; 18: 965–977.

11.

Loh

Mak

Tan

LKS

, et al. Microbiota–gut–brain axis and its therapeutic applications in neurodegenerative diseases. Signal Transduct Target Ther 2024; 9: 37.

12.

Colombo

Sadler

Llovera

, et al. Microbiota-derived short chain fatty acids modulate microglia and promote Aβ plaque deposition. eLife 2021; 10: e59826.

13.

Zhan

Stamova

Jin

L-W

, et al. Gram-negative bacterial molecules associate with Alzheimer disease pathology. Neurology 2016; 87: 2324–2332.

14.

Zhan

Stamova

Sharp

. Lipopolysaccharide associates with amyloid plaques, neurons and oligodendrocytes in Alzheimer’s disease brain: a review. Front Aging Neurosci 2018; 10: 42.

15.

Vogt

Kerby

Dill-McFarland

, et al. Gut microbiome alterations in Alzheimer’s disease. Sci Rep 2017; 7: 13537.

16.

Zhuang

Z-Q

Shen

L-L

W-W

, et al. Gut microbiota is altered in patients with Alzheimer’s disease. J Alzheimers Dis 2018; 63: 1337–1346.

17.

Liu

Peng

, et al. Altered microbiomes distinguish Alzheimer’s disease from amnestic mild cognitive impairment and health in a Chinese cohort. Brain Behav Immun 2019; 80: 633–643.

18.

Cattaneo

Cattane

Galluzzi

, et al. Association of brain amyloidosis with pro-inflammatory gut bacterial taxa and peripheral inflammation markers in cognitively impaired elderly. Neurobiol Aging 2017; 49: 60–68.

19.

University of York. PROSPERO: International prospective register of systematic reviews, https://www.crd.york.ac.uk/prospero/ (accessed 4 February 2026).

20.

Page

McKenzie

Bossuyt

, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Int J Surg 2021; 88: 105906.

21.

McKhann

Knopman

Chertkow

, et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 263–269.

22.

Petersen

Smith

Waring

, et al. Mild cognitive impairment: clinical characterization and outcome. Arch Neurol 1999; 56: 303–308.

23.

Albert

DeKosky

Dickson

, et al. The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 2011; 7: 270–279.

24.

Wan

Wang

Liu

, et al. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol 2014; 14: 135.

25.

Stang

. Critical evaluation of the Newcastle-Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses. Eur J Epidemiol 2010; 25: 603–605.

26.

R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, 2025.

27.

IntHout

Ioannidis

Borm

. The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method. BMC Med Res Methodol 2014; 14: 25.

28.

Higgins

JPT

Thompson

Deeks

, et al. Measuring inconsistency in meta-analyses. Br Med J 2003; 327: 557–560.

29.

DerSimonian

Laird

. Meta-analysis in clinical trials. Control Clin Trials 1986; 7: 177–188.

30.

Egger

Smith

Schneider

, et al. Bias in meta-analysis detected by a simple, graphical test. Br Med J 1997; 315: 629–634.

31.

Duval

Tweedie

. Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics 2000; 56: 455–463.

32.

Gloor

Macklaim

Pawlowsky-Glahn

, et al. Microbiome datasets are compositional: and this is not optional. Front Microbiol 2017; 8: 2224.

33.

Haran

Bhattarai

Foley

, et al. Alzheimer’s disease microbiome is associated with dysregulation of the anti-inflammatory P-glycoprotein pathway. mBio 2019; 10: e00632–19.

34.

, et al. Mild cognitive impairment has similar alterations as Alzheimer’s disease in gut microbiota. Alzheimers Dement 2019; 15: 1357–1366.

35.

Ding

Zhu

, et al. Disturbed microbial ecology in Alzheimer’s disease: evidence from the gut microbiota and fecal metabolome. BMC Microbiol 2021; 21: 226.

36.

Ueda

Shinkai

Shiroma

, et al. Identification of Faecalibacterium prausnitzii strains for gut microbiome-based intervention in Alzheimer’s-type dementia. Cell Rep Med 2021; 2: 100398.

37.

Zhou

Wang

Quan

, et al. Gut microbiota changes and their correlation with cognitive and neuropsychiatric symptoms in Alzheimer’s disease. J Alzheimers Dis 2021; 81: 583–595.

38.

Ling

Zhu

Yan

, et al. Structural and functional dysbiosis of fecal microbiota in Chinese patients with Alzheimer’s disease. Front Cell Dev Biol 2021; 8: 634069.

39.

Hou

Ran

, et al. APOE-ε4 carrier status and gut microbiota dysbiosis in patients with Alzheimer disease. Front Neurosci 2021; 15: 619051.

40.

Guo

Peng

Huang

, et al. Gut microbiome features of Chinese patients newly diagnosed with Alzheimer’s disease or mild cognitive impairment. J Alzheimers Dis 2021; 80: 299–310.

41.

Duan

Liu

, et al. Preoperative microbiomes and intestinal barrier function can differentiate prodromal Alzheimer’s disease from normal neurocognition in elderly patients scheduled to undergo orthopedic surgery. Front Cell Infect Microbiol 2021; 11: 592842.

42.

Liu

Jia

Chen

, et al. Gut microbiota interacts with intrinsic brain activity of patients with amnestic mild cognitive impairment. CNS Neurosci Ther 2021; 27: 163–173.

43.

Pan

Y-Q

Guo

, et al. Elderly patients with mild cognitive impairment exhibit altered gut microbiota profiles. J Immunol Res 2021; 2021: 5578958.

44.

Yıldırım

Nalbantoğlu

ÖU

Bayraktar

, et al. Stratification of the gut microbiota composition landscape across the Alzheimer’s disease continuum in a Turkish cohort. mSystems 2022; 7: e0000422.

45.

Pei

, et al. Gut microbiota and intestinal barrier function in subjects with cognitive impairments: a cross-sectional study. Front Aging Neurosci 2023; 15: 1174599.

46.

Klee

Aho

VTE

May

, et al. Education as risk factor of mild cognitive impairment: the link to the gut microbiome. J Prev Alzheimers Dis 2024; 11: 759–768.

47.

Cabrera

Carrión

Mateo

, et al. Gut microbiota characterization in ageing, mild cognitive impairment, and Alzheimer’s disease in the context of Mediterranean lifestyle in a Spanish population. Alzheimers Res Ther 2025; 17: 211.

48.

Hosseinkhani

Chevalier

Marizzoni

, et al. Plasma and feces multiomics unveil cognition-associated perturbations of chronic inflammatory pathways of the gut-microbiota–brain axis. Alzheimers Dement 2025; 21: e70844.

49.

Lwere

Muwonge

Sendagire

, et al. Characterization of the gut microbiome in Alzheimer disease and mild cognitive impairment among older adults in Uganda: a case–control study. Medicine (Baltimore) 2025; 104: e42100.

50.

Rout

Dash

Satish

, et al. Exploring the role of acylated ghrelin and gut microbiome in delineating cognitive health in the elderly. Aging (Albany NY) 2025; 17: 393–407.

51.

Rouskas

Mamalaki

Ntanasi

, et al. Gut microbiome alterations in mild cognitive impairment: findings from the ALBION Greek cohort. Microorganisms 2025; 13: 2112.

52.

Sepúlveda-Rivera

Olivieri-Henry

Morales-González

, et al. Gut microbiota distinguishes aging Hispanics with Alzheimer's disease: associations with cognitive impairment and severity. Sci Rep 2025; 15: 28505.

53.

Mirzayi C, Renson A, Genomic Standards Consortium, et al. Reporting guidelines for human microbiome research: the STORMS checklist. Nat Med 2021; 27: 1885–1892.

54.

Jiang

Huang

, et al. The gut microbiota and Alzheimer’s disease. J Alzheimers Dis 2017; 58: 1–15.

55.

Stroup

. Meta-analysis of observational studies in epidemiology: a proposal for reporting. JAMA 2000; 283: 2008.

56.

Verhaar

BJH

Hendriksen

HMA

de Leeuw

, et al. Gut microbiota composition is related to AD pathology. Front Immunol 2021; 12: 794519.

57.

Ferreiro

Choi

Ryou

, et al. Gut microbiome composition may be an indicator of preclinical Alzheimer’s disease. Sci Transl Med 2023; 15: eabo2984.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

2.68 MB