Introduction

Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by the degeneration of dopaminergic neurons in the substantia nigra and the accumulation of pathological α-synuclein (α-syn) aggregates.1 Dopamine replacement therapy, particularly levodopa, remains the cornerstone of symptom management. However, it primarily offers transient relief and fails to alter disease progression.2,3 Over time, many patients develop motor fluctuations and inadequate control of nonmotor symptoms,4 underscoring the urgent need for novel therapeutic strategies beyond conventional pharmacotherapy. Fecal microbiota transplantation (FMT), the transfer of stool from healthy donors to patients, has been an established treatment for recurrent Clostridium difficile infection since 2013.5,6 In recent years, interest in FMT has expanded to include neurodegenerative diseases because of its ability to modulate the gut microbiome, reduce inflammation, and influence neuroimmune signaling. Intriguingly, the gut-brain axis has emerged as a critical component of PD pathogenesis. Experimental models suggest that misfolded α-syn may originate in the gut and ascend to the brain via the vagus nerve,7,8 supporting the hypothesis that gastrointestinal pathology may initiate or contribute to central neurodegeneration. This hypothesis is further supported by clinical observations. Gastrointestinal symptoms, particularly constipation, often precede motor symptoms by years and may signal prodromal neurodegeneration.9 Moreover, PD patients frequently exhibit intestinal barrier dysfunction, inflammation, and dysbiosis,10,11 reinforcing the notion that gut microbiota imbalances contribute to disease onset and progression.12

Early-phase clinical studies have reported that FMT can alleviate constipation in PD patients and, in some cases, also improve motor symptoms.13 However, findings from randomized controlled trials have been inconsistent. For example, Bruggeman et al. observed significant motor improvement following a single nasojejunal donor FMT (dFMT) in early-stage PD patients,14 while Scheperjans et al. reported no motor benefit from a single colonoscopic dose of anaerobically prepared FMT,15 and DuPont et al. found only temporary effects from repeated oral FMT capsules.16 Such discrepancies likely stem from differences in delivery routes, microbial preparation, dosing schedules, and patient selection.

Notably, most prior trials enrolled patients already receiving dopaminergic therapy, complicating interpretation due to potential drug-microbiota interactions and placebo effects. Although recent trials, including those by Bruggeman et al.,14 Scheperjans et al.,15 DuPont et al.,16 and Cheng et al.17 have explored the feasibility of FMT in PD, variations in trial design and outcome measures limit their generalizability to the drug-naïve population. Furthermore, no consensus has been reached regarding the optimal FMT delivery strategy for neurological conditions.

To address these critical gaps, particularly the efficacy, safety, and feasibility of repeated FMT in drug-naïve PD, we conducted a phase 2, randomized, double-blind, placebo-controlled trial using colonic transendoscopic enteral tubing (TET).18 This trial evaluated whether repeated cycles of dFMT, compared with autologous FMT (aFMT), using the individual’s own stool, could improve motor and nonmotor symptoms, modulate the gut microbiota composition, and ameliorate intestinal pathology in patients with de novo PD. In addition to assessing clinical outcomes, we performed a multimodal assessment including gut microbiota profiling, fecal dopamine metabolite analysis, and colonic histopathology to thoroughly investigate the therapeutic impact of FMT in treatment-naïve PD patients.

Results

Participant characteristics

A total of 212 patients were screened, of whom 72 met the eligibility criteria and were randomized equally to receive either dFMT (n = 36) or aFMT (n = 36) (Fig. 1). Based on established UPDRS III criteria (mild: 10–25; moderate: 26−40; severe: > 40),19 our cohort primarily comprised patients with moderate impairment (aFMT: 61.1%; dFMT: 55.6%) and mild impairment (aFMT: 30.6%; dFMT: 36.1%). The baseline disease severity distribution was well balanced between the two groups (Table 1).

Fig. 1: CONSORT flow diagram of patient enrollment and allocation.
Fig. 1: CONSORT flow diagram of patient enrollment and allocation.
Full size image

Flowchart depicting the screening, randomization, intervention, and follow-up phases of the FMT trial in de novo Parkinson’s disease patients. wk, week

Table 1 Demographics and Baseline Characteristics
Full size table

Three participants from each group withdrew before completing the study, resulting in a final cohort of 66 patients (33 in each group) for the intention-to-treat (ITT) and safety analyses. The overall dropout rate was 8.3% (Fig. 1). The baseline demographic and clinical characteristics were well balanced between the groups, with no significant differences in age, sex, Hoehn and Yahr (H&Y) stage, UPDRS scores, Mini-Mental State Examination (MMSE) scores, Montreal Cognitive Assessment (MoCA) scores, or gut microbiota indices (Table 1).

Primary outcomes

The primary outcome was the change in the UPDRS III (motor function) score from baseline to week 35, which showed a mean change of -3.8 points (standard deviation (SD), 1.2) in the dFMT group but minimal change (0.1 points; SD, 1.1) in the aFMT group. The adjusted between-group difference was −3.9 points (95% CI: −5.8 to −2.0; p = 0.0001), favoring dFMT. A clinically meaningful improvement of ≥ 3.25 points (minimal clinically important difference, MCID threshold)20,21 was achieved by 45.5% of patients in the dFMT group compared with 21.2% in the aFMT group (χ² = 4.36, p = 0.0367) (Fig. 2a and Table 2). Individual changes in UPDRS III scores for each participant are graphically depicted in Fig. 3.

Fig. 2: Clinical and functional outcomes of FMT in de novo PD patients.
Fig. 2: Clinical and functional outcomes of FMT in de novo PD patients.
Full size image

a UPDRS III (motor examination). b UPDRS II (motor experiences of daily living). c UPDRS I (nonmotor experiences of daily living). d Constipation Scoring System (CSS). e Patient Assessment of Constipation-Quality of Life (PAC-QoL), normalized to a 0–4 scale (raw score/28). f Gastrointestinal Symptom Rating Scale (GSRS). g SCOPA-AUT (autonomic symptoms). h Mini-Mental State Examination (MMSE). i Montreal Cognitive Assessment (MoCA). j 39-item Parkinson’s Disease Questionnaire (PDQ-39). Data are presented as the mean ± standard deviation; n = 33 per group. aFMT, autologous fecal microbiota transplantation; dFMT, donor fecal microbiota transplantation; w, week. Between-group differences were assessed at baseline and at 11, 23, and 35 weeks. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Fig. 3: Individual trajectories of motor symptom change (UPDRS III).
Fig. 3: Individual trajectories of motor symptom change (UPDRS III).
Full size image

Parallel line plots illustrating individual changes in UPDRS III scores from baseline to week 35 for each participant. Lines represent motor symptom improvement or worsening. n = 33 per group. aFMT, autologous fecal microbiota transplantation; dFMT, donor fecal microbiota transplantation

Table 2 Primary and Secondary outcomes
Full size table

Secondary outcomes

Motor and daily living functions

The UPDRS II (activities of daily living) scores improved significantly in the dFMT group at weeks 11 and 23 (week 11 difference: −0.5, 95% CI: −0.9 to −0.1, p = 0.0255; week 23 difference: −0.5, 95% CI: −0.9 to −0.1, p = 0.0184), but the between-group difference was not sustained at week 35 (Fig. 2b and Table 2). No significant differences were observed for UPDRS I (nonmotor aspects of daily living) at any time point (Fig. 2c and Table 2). The secondary efficacy outcomes are summarized in Table 2.

Constipation and gastrointestinal symptoms

Constipation severity, as measured by the Constipation Scoring System (CSS) score, was significantly improved in the dFMT group. The mean CSS score decreased by −6.5 at week 35 compared to −0.7 in the aFMT group (adjusted difference: −5.8, 95% CI: −7.0 to −4.6; p < 0.0001) (Fig. 2d and Table 2). The mean Patient Assessment of Constipation-Quality of Life (PAC-QoL) score decreased by −0.3 at week 35 compared to −0.1 in the aFMT group (adjusted difference: −0.3, 95% CI: −0.5 to −0.2; p < 0.0001) (Fig. 2e and Table 2). This finding aligns with multinational validation data showing that PAC-QoL effect sizes > 0.7 (moderate-to-large) indicate clinically relevant change, and our dFMT group’s PAC-QoL effect size (1.23) meets this threshold. Improvements were also noted in the Gastrointestinal Symptom Rating Scale (GSRS) (all p < 0.05) scores across all assessment points (Fig. 2f and Table 2), indicating a clinically significant relief of gastrointestinal symptoms. The dFMT group achieved raw GSRS decreases of 3.5 (week 11), 3.1 (week 23), and 2.7 (week 35), all exceeding the MCID lower bound (1.5 points). To explore the gut-brain relationship, we analyzed the correlation between motor function improvement (ΔUPDRS III) and gastrointestinal symptom relief (CSS and GSRS), revealing partial evidence of a relationship, particularly during the earlier phase of follow-up. A moderate, statistically significant positive correlation was observed between the improvement in UPDRS III motor scores and the improvement in CSS scores at both week 11 (r = 0.5405, p = 0.0012) and week 23 (r = 0.4481, p = 0.0089) (Supplementary Fig. 1a–c), suggesting a temporal association between motor and bowel function recovery. We also examined correlations with the GSRS. While changes in the general GSRS score did not show significant correlation with motor symptom changes at any time point (Supplementary Fig. 1a–c), the GSRS lower digestive tract subscale (GSRS-LDT) demonstrated a significant association. Specifically, improvements in GSRS-LDT scores were significantly correlated with motor improvements at week 11 (r = 0.4861, p = 0.0041) and week 35 (r = 0.4509, p = 0.0085) but not at week 23 (Supplementary Fig. 1d–f). These findings indicate that improvements in lower gastrointestinal symptoms are temporally aligned with both early and sustained motor recovery following dFMT.

Autonomic and cognitive function

Scales for Outcomes in Parkinson’s Disease-Autonomic (SCOPA-AUT) scores improved significantly in the dFMT group at week 11 (p = 0.0001) but not at weeks 23 and 35 (both p > 0.05; Fig. 2g and Table 2). The MMSE and MoCA scores remained stable in both groups, with no significant differences at any time point, confirming cognitive safety (Fig. 2h, i, and Table 2). Additionally, three months of FMT treatment led to improved 39-item Parkinson’s Disease Questionnaire (PDQ-39) subscale scores for mobility, activities of daily living, emotional well-being, stigma, and bodily discomfort in the dFMT group, with no changes observed in the PDQ-39 subscale scores in the aFMT group (Fig. 2j and Table 2).

Microbiota composition

Microbiota analyses revealed modest increases in alpha diversity in the dFMT group, but these changes were not statistically significant (Supplementary Fig. 2a). Beta diversity analysis revealed a shift in microbial composition in the dFMT group at weeks 11 and 23 (PERMANOVA, week 11, p = 0.001; week 23, p = 0.001) (Supplementary Fig. 2b), with post-FMT profiles becoming more similar to donor profiles. Escherichia-Shigella abundance was significantly reduced in the dFMT group (Supplementary Fig. 3a and Supplementary Table 1) and correlated with improved motor and gastrointestinal outcomes (Supplementary Fig. 2c).

Biochemical and histological findings

Stool dopamine and 3,4-dihydroxyphenylacetic acid (DOPAC) concentrations were significantly greater in the dFMT group at week 35 (dFMT vs. aFMT, dopamine: 390.76 ± 361.33 vs. 115.30 ± 97.85 ng/mL; p = 0.0002; DOPAC: 576.24 ± 284.60 vs. 339.62 ± 274.66 ng/mL; p = 0.0049) (Supplementary Fig. 4a-c and Table 2). Additionally, hematoxylin and eosin staining showed that the colonic pathological score in the dFMT group was significantly decreased at week 35 (Supplementary Fig. 5a). Immunohistochemical analysis revealed reduced immunoreactivity of phosphorylated α-syn (pα-syn) at week 35 (dFMT vs. aFMT, 11.10 ± 3.30% vs. 14.71 ± 4.83%, p = 0.0301) (Supplementary Fig. 5b) and immunofluorescence staining showed increased E-cadherin expression (dFMT vs. aFMT, 49.66 ± 6.98 vs. 27.98 ± 3.42, p < 0.0001) in colonic biopsies from the dFMT group at week 35, which was consistent with improved intestinal barrier integrity (Supplementary Fig. 5c). Furthermore, Spearman correlation analysis revealed a positive relationship between the change in pα-syn and the alteration in the abundance of Escherichia-Shigella (r = 0.3775, p = 0.0277), a negative correlation between the change in Escherichia-Shigella and the change in dopamine (r = –0.3539, p = 0.0036), and a negative correlation between the change in Escherichia-Shigella and the change in DOPAC (r = –0.2583, p = 0.0363) (Supplementary Fig. 3b–d). These findings suggest a potential link between changes in the composition of the gut microbiota, particularly Escherichia-Shigella, and a reduction in pathological α-syn aggregates.

Safety

FMT was well tolerated in both groups. All adverse events were mild, including abdominal discomfort, bloating, and transient diarrhea. There were no serious adverse events. Six patients (3 in each group) discontinued due to mild gastrointestinal symptoms during the first cycle. The potential symptoms relating to discomfort included diarrhea (n = 1 [2.8%] in the dFMT group and n = 1 [2.8%] in the aFMT group), nausea (n = 1 [2.8%] in the aFMT group), abdominal pain (n = 1 [2.8%] in the dFMT group and n = 1 [2.8%] in the aFMT group) and abdominal bloating (n = 1 [2.8%] in the dFMT group). For the 6 patients above, these adverse events were reasons to discontinue their intervention. Overall, there were no serious adverse events observed in this trial (Table 3).

Table 3 Adverse Events by Treatment Groupa
Full size table

Discussion

This randomized, double-blind, placebo-controlled trial demonstrated that repeated dFMT confers clinically meaningful and statistically significant symptomatic benefits in patients with de novo PD. Compared with those receiving aFMT, those receiving dFMT experienced greater improvements in motor symptoms, constipation severity, and gastrointestinal-related quality of life over the 35-week follow-up. Together, these results directly address the therapeutic question posed in the introduction and reinforce the growing body of evidence supporting gut microbiota modulation as a viable disease-modifying avenue in PD.

Mounting evidence has implicated gut dysbiosis in PD pathogenesis.10,11,12 Building on earlier clinical studies,13,14,15,16 our findings extend prior work by implementing a rigorous, repeated dFMT regimen in drug-naïve patients. The drug-naïve status of the participants minimized the confounding effects of antiparkinsonian medications. Although the motor improvement (mean change of –3.8 points in the UPDRS III in the dFMT group; Table 2) was modest compared with dopaminergic therapy,2,3 it exceeded the MCID threshold established in PD trials.22,23 The MCID for the UPDRS III score varies with disease stage and treatment context. Specifically, thresholds are higher (~ 5 points) for advanced or medicated PD populations,24,25 reflecting the larger change required for symptomatic benefit in patients with severe or fluctuating deficits. In contrast, early-stage or drug-naïve cohorts show smaller MCIDs (2.5 to 3.25 points20,21) owing to milder baseline deficits. Our findings align with this early-stage PD reference range. The dFMT group showed a mean 3.8-point reduction in UPDRS III scores (adjusted between-group difference, –3.9 points; Table 2), exceeding the established 2.5 to 3.25 points MCID interval for this cohort. Responder analysis further supported this clinical relevance, with 45.5% of dFMT participants achieving ≥ 3.25-point improvement compared with 21.2% in the aFMT group (p = 0.0367).

As dFMT is a microbiota-modulating rather than dopaminergic intervention, the magnitude of motor improvement is expectedly modest relative to dopamine replacement therapies. Nonetheless, the sustained and biologically coherent effects observed at week 35, accompanied by reduced α-syn aggregation and neuroinflammation, strongly support a genuine therapeutic benefit mediated via gut-brain axis modulation rather than symptomatic dopaminergic compensation. Beyond motor function, we observed substantial and durable improvements in nonmotor symptoms, particularly constipation and gastrointestinal quality of life, as evidenced by the CSS, PAC-QoL, and GSRS scores. These effects emerged as early as week 11 and were maintained through week 35. While improved gastrointestinal function may contribute to better overall well-being and indirectly influence motor symptoms, the causal relationships remain to be fully elucidated. These observations are concordant with our recent study in progressive supranuclear palsy–Richardson’s syndrome, in which intermittent FMT improved both motor and nonmotor symptoms and was well tolerated.26

Mechanistically, our data revealed a marked posttreatment reduction in Escherichia-Shigella abundance, a genus pattern frequently associated with PD-related gut dysbiosis.27 This microbial shift was associated with improved clinical outcomes and a significant reduction in colonic pα-syn and increased E-cadherin expression, findings that support the gut-brain axis hypothesis in PD.7,28 Reduced mucosal α-syn likely indicates alleviation of local inflammation.29,30 In carriers of LRRK2 risk variants (e.g., R1628P, G2385R), Escherichia coli can induce α-syn pathology via extracellular-vesicle–associated curli and impaired autophagic clearance. In animal models, this gene–environment interaction promoted gut-to-brain α-syn propagation and motor deficits, both of which were alleviated by FMT,31 supporting microbiota-targeted interventions in genetically susceptible PD populations. Furthermore, elevated fecal levels of dopamine and DOPAC in the dFMT group suggest an influence of the microbiota on host neurotransmitter metabolism. Although these metabolites do not cross the blood-brain barrier and must be interpreted as peripheral markers, they may reflect gut-microbiome interactions relevant to central nervous system (CNS) function. Future studies incorporating metagenomics and metabolomics are warranted to elucidate underlying functional pathways and host-microbe dynamics.32 Our use of repeated TET infusions was guided by mechanistic considerations relevant to chronic neurodegenerative disease. Precise delivery to the pathogenic origin is critical: accumulating evidence implicates the distal gut as the initial site of α-syn pathology and its subsequent propagation to the CNS.33 The TET approach uniquely allows for direct and repeated deposition of a large volume of donor microbiota precisely into the distal small bowel and proximal colon, maximizing therapeutic action at the putative origin of PD pathology.34,35 Durable microbial engraftment is critical for sustained efficacy in chronic neurodegenerative disease. A single bolus was unlikely to ensure stable colonization; therefore, repeated multiday TET infusions were designed to maintain engraftment and healthy host-microbe interactions. Although TET requires endoscopic expertise and is not scalable for general use, our phase 2 approach prioritized biological certainty, establishing proof of efficacy that can inform the development of less invasive delivery methods.

This study features a comprehensive clinical, microbial, and histological assessment and demonstrates a minimal placebo response, validating the rigor of the procedure-matched, randomized double-blind design. Use of aFMT ensured that all participants underwent identical, intensive procedures, effectively eliminating expectancy bias common in single-blind trials. Furthermore, selection of a de novo, drug-naïve PD cohort removed confounding from dopaminergic medications, and the clinical stability of the aFMT group mirrored the natural history of early PD, confirming a true baseline. Together with convergent evidence in the dFMT group, where clinical improvements correlated with significant microbial and neurochemical changes (e.g., reduced colonic pα-syn), these findings strongly support the biological efficacy of donor microbiota intervention. Overall, the study’s strengths include a rigorous randomized double-blind design, enrollment of a drug-naïve population, multimodal outcome assessment, and a favorable safety profile, with only mild and self-limiting gastrointestinal adverse events observed. Several limitations warrant consideration. The single-center design and modest sample size may limit generalizability, and future multicenter trials with larger cohorts (n > 200) are needed. Microbiota profiling relied primarily on 16S rDNA sequencing, which lacks functional resolution. Parallel metagenomic analysis revealed clear group separation at week 35 (Supplementary Fig. 6a), with differential taxa associated with clinical outcomes (Supplementary Fig. 6b, c), warranting further validation. Antibiotic pretreatment may have independently altered gut microbiota, potentially confounding observed effects36; antibiotic-free preparation protocols should be explored. Finally, constipation assessment did not include complete spontaneous bowel movement frequency, an important clinical endpoint in PD. In summary, this phase 2 trial provides evidence that repeated dFMT improves both motor and gastrointestinal symptoms in drug-naïve PD, supporting microbiota modulation as a promising therapeutic strategy. These findings add to the expanding literature implicating the gut microbiome in neurodegenerative disease pathophysiology.32,37 Future studies should validate these results in larger, multicenter cohorts, integrate multi-omics profiling to elucidate functional mechanisms, refine delivery approaches, and identify microbial biomarkers to enable precision microbiome-based interventions in PD.

Materials and methods

Study design

This was a 35-week, double-blind, randomized, placebo-controlled, single-center phase 2 trial of FMT in patients with de novo PD (i.e., drug-naïve, not previously treated with dopaminergic therapy). The trial was conducted at the First Affiliated Hospital of Zhengzhou University. Participants were enrolled between September 2022 and January 2023 and followed until October 2023. The study was approved by the Institutional Ethics Committee (approval number 2021-KY-0385-002) and registered in the Chinese Clinical Trial Registry (ChiCTR2200064151; https://www.chictr.org.cn/bin/userProject). Written informed consent was obtained from all participants prior to enrollment.

Participants

Eligible patients were aged 31–71 years, were newly diagnosed with PD on the basis of the UK Parkinson’s Disease Society Brain Bank criteria, and had not received dopaminergic therapy. Additional inclusion criteria included H&Y stage 1.5–3.0 and a UPDRS III score of 15–67. The key exclusion criteria included a diagnosis of dementia (MMSE < 25), major depression or psychosis, significant gastrointestinal disease, recent abdominal or anorectal surgery, antibiotic or probiotic use within the past 3 months, contraindications for colonoscopy, and serious comorbidities (more detailed exclusion criteria are provided in the Supplementary Material).

Donor selection and stool processing

Healthy stool donors underwent extensive screening, including medical history, physical examination, and laboratory tests for infectious and metabolic diseases38,39 (Supplementary Material). Among 144 candidates, 15 (10.4%) were deemed eligible. Stool from three eligible donors was pooled to form each batch, with each batch containing 200 g of stool mixed with 800 mL of 0.9% saline and 80 mL of 100% glycerol. The mixtures were homogenized, aliquoted, and stored at –80 °C. Autologous stool samples were processed via the same protocol.

Randomization and interventions

The participants were randomized 1:1 to receive either dFMT or aFMT via a computer-generated randomization sequence. Allocation was concealed via sealed envelopes and managed by nursing staff not involved in FMT procedures or outcome assessments. Both the participants and the investigators were blinded to the group assignments.

All participants received a 7-day course of oral ciprofloxacin (500 mg twice daily) and metronidazole (500 mg thrice daily) before the first FMT cycle. After bowel preparation, a TET device was placed into the ileocecal region via colonoscopy.36 FMT was administered for 7 consecutive days per cycle (200 mL on days 1–3; 50 mL on days 4–7) and was repeated every 28 days for a total of three cycles (Supplementary Fig. 7).

  • dFMT: Stools from screened healthy donors (pooled in batches of three) were processed with 0.9% saline and 100% glycerol, frozen at −80 °C, and thawed before administration.

  • aFMT: Each patient’s own stool was collected and processed identically for use in the placebo arm.

Participants remained in the prone position for 2 h post-infusion.

Outcomes

The primary outcome was the change in the UPDRS III score from baseline to week 35. The secondary outcomes included changes in the UPDRS I and II, CSS, GSRS, SCOPA-AUT, PDQ-39, PAC-QoL (normalized to a 0–4 scale (raw score/28)), MMSE, and MoCA scores.

Biochemical and histological assessments

Stool neurotransmitter (dopamine, DOPAC) levels were measured via standard ELISA kits. Intestinal biopsies were collected during colonoscopy at baseline and at follow-up time points. Immunohistochemistry and immunofluorescence staining were performed to detect the expression of pα-syn and E-cadherin, respectively. Scoring was conducted by blinded pathologists using predefined criteria.

Sample size calculation

Sample size determination was based on detecting a 3.25-point difference in UPDRS III (SD = 5),22 with 80% power at a two-sided α = 0.05. This yielded a required sample of 36 participants per group based on a 15% anticipated dropout rate. Calculations were performed using an online tool (http://riskcalc.org:3838/samplesize/).

Microbiome analysis

Stool samples were collected at baseline and at weeks 11, 23, and 35. Microbial DNA was extracted using the cetyltrimethylammonium bromide and sodium dodecyl sulfate (CTAB/SDS) method according to the manufacturers’ instructions. The V3–V4 hypervariable regions of the 16S rRNA gene were amplified using standard primers (V3 forward: 5’-TACGGRAGGCAGCAG-3’; V4 reverse: 5’-CTACCNGGGTATCTAAT-3’) and sequenced following an optimized and standardized amplicon library preparation protocol. Sequence data were processed using QIIME2. Denoising was performed using the DADA2 algorithm (or deblur, where applicable) to generate amplicon sequence variants (ASVs). Taxonomic classification of representative sequences from the kingdom to genus level was conducted using the q2-feature-classifier plugin against the Silva database (http://www.arb-silva.de). Microbial alpha diversity indices, including observed ASVs, Chao1, Shannon, and Simpson indices, were calculated using QIIME2. Beta diversity was assessed using Bray‒Curtis distance matrices generated with the vegan package in R (version 4.3.3). Principal coordinate analysis (PCoA) was applied to visualize differences in microbial community composition, and the statistical significance of beta diversity differences between groups was evaluated using permutational multivariate analysis of variance (PERMANOVA) with the adonis function in the vegan package.

Detailed microbiome analysis procedures are provided in the Supplementary Material.

Statistical analysis

The data were analyzed on an ITT basis. Continuous variables were analyzed via independent t test or ANCOVA adjusted for baseline values. Repeated-measures mixed models were applied for longitudinal outcomes. Categorical variables were assessed using the chi-square test or Fisher’s exact test, as appropriate. Spearman correlation analyses between changes in microbial taxon abundance and clinical or histological outcomes were adjusted for multiple comparisons using the Benjamini‒Hochberg false discovery rate (FDR) correction. All statistical analyses were two-sided, with an adjusted p < 0.05 considered statistically significant.