Obesity and its associated metabolic disorders, including metabolic syndrome, type 2 diabetes (T2DM), and metabolic dysfunction-associated steatohepatitis (MASLD)/non-alcoholic fatty liver disease (NAFLD), have emerged as one of the most severe global health epidemics of the 21st century. The impact of these diseases is vast: the prevalence of metabolic syndrome now rivals that of obesity and diabetes, affecting up to 45% of the global population (1); T2DM affects hundreds of millions worldwide, with cases projected to double by 2030 (2); and NAFLD/MASLD has become the most common chronic liver disease globally, with prevalence soaring to 71–92% among obese individuals (3). These conditions not only substantially increase morbidity and mortality risks but may also progress to cirrhosis, hepatocellular carcinoma, or cardiovascular disease, driving sustained healthcare expenditures (4). However, conventional lifestyle interventions and pharmacological treatments face limitations in efficacy, cost-effectiveness, or long-term adherence. Consequently, exploring innovative therapeutic approaches has become an urgent public health imperative (5).

In recent years, the gut microbiota has been recognized as a key regulator of host metabolism and energy homeostasis. Increasing evidence indicates that dysbiosis represents one of the core pathophysiological mechanisms underlying obesity, insulin resistance, and related metabolic disorders (6–8). Consequently, modulating the gut microbiota to restore its healthy balance is viewed as a highly promising new therapeutic target (9). As a potent intervention capable of reconstructing the gut microbiome, FMT—involving the transplantation of gut microbiota from healthy donors—has demonstrated remarkable success in treating conditions such as recurrent Clostridioides difficile infection (10, 11). This success has sparked significant interest in its potential application for treating metabolic diseases.

However, as FMT transitions from animal models to clinical applications, its actual efficacy in treating metabolic diseases remains highly controversial. Pioneering trials in this field initially reported that FMT significantly improved insulin sensitivity in patients with metabolic syndrome, driving the advancement of FMT as a treatment for metabolic disorders (12, 13). Yet these positive signals have not been consistently replicated in subsequent studies. Conversely, numerous subsequent randomized controlled trials (RCTs) evaluating key metabolic markers have yielded a multitude of conflicting conclusions, reporting either no effect from FMT or even slight deterioration (14–16). This widespread heterogeneity in outcomes likely stems from substantial variations in donor selection, intervention protocols, administration routes, and baseline characteristics of participants. These differences make it challenging for clinicians and researchers to reliably assess the true efficacy of FMT. Consequently, the contradictory nature of the existing evidence significantly limits the objective evaluation of its clinical value.

Therefore, given the current inconsistencies in clinical evidence and the limitations of individual studies, a comprehensive systematic review and meta-analysis are warranted. This study aims to objectively evaluate the true impact of FMT on key metabolic indicators in obese and metabolic disease populations by integrating data from existing relevant RCTs to provide a more precise and robust effect size estimate. The findings will provide critical evidence-based medical support for the clinical application of FMT in this field and future research directions.

This systematic review and meta-analysis were conducted in accordance with the PRISMA statement (17, 18). The study protocol has been registered with PROSPERO (registration number: CRD420251172011).

To our knowledge, this study represents one of the most recent and comprehensive systematic reviews and meta-analyses examining the efficacy of FMT in treating overweight or obese adults with metabolic disorders. In the primary analysis, our pooled results showed that allogeneic FMT intervention did not yield statistically significant improvements in primary outcome measures (BMI, HOMA-IR, HbA1c) or secondary lipid parameters compared to placebo or autologous FMT controls.

However, the evidence base for this overall negative conclusion remains weak. In our sensitivity analyses, outcomes for BMI exhibited high sensitivity to statistical estimation methods: adjusting the correlation coefficient (Corr) from 0.5 to 0.75 for estimating the standard deviation of change transformed FMT’s effect into statistical significance (p = 0.010) (Supplementary Table S3). The leave-one-out analysis (Supplementary Table S2) for BMI also demonstrated this; the primary pooled result was non-significant (p = 0.070), whereas the pooled result excluding the study by Gómez-Pérez et al. (21) was statistically significant (p = 0.050). This instability highlights the complexity of the current evidence base and suggests that the potential efficacy of FMT may be obscured by heterogeneity in study design and statistical methods.

Nevertheless, statistical significance does not equate to clinical relevance, and the magnitude of the observed effects warrants critical scrutiny. The pooled BMI reduction of 0.65 kg/m², while favoring intervention, remains below the 5–10% weight loss threshold typically required to improve cardiometabolic health (28). Similarly, the decrease in HOMA-IR (-0.64) offers limited benefit for severe insulin resistance, where normalization necessitates more substantial shifts. These findings imply that FMT’s current efficacy as a standalone monotherapy is modest compared to standard interventions. Thus, it may be more clinically viable as an adjunctive strategy rather than a primary treatment.

This finding of non-significant effects aligns with trends reported in existing systematic reviews in this field. Recent meta-analyses targeting metabolic syndrome have observed no significant improvement in HOMA-IR, lipid profiles, or BMI following FMT (29, 30). This suggests that under the current research paradigm, the overall efficacy of FMT in improving systemic insulin resistance remains to be established. However, the therapeutic efficacy of FMT in specific gastrointestinal disorders, such as ulcerative colitis (UC) and irritable bowel syndrome (IBS), is widely recognized (31, 32). This disparity in efficacy across indications may reflect differing intervention challenges: correcting systemic disorders like metabolic syndrome—involving multiple organs with complex pathophysiological networks and potentially more stable dysbiosis states—may present higher intervention barriers compared to localized intestinal inflammatory diseases such as UC or IBS.

1) First is the intensity and duration of the intervention. Our analysis indicates that the administration regimen (single vs. multiple doses) is significantly associated with changes in HbA1c (p = 0.040), with multiple-dose regimens linked to a greater trend toward HbA1c reduction (β = -0.56). This suggests that single-dose FMT interventions may struggle to sustain long-term effects in metabolic disease studies. This perspective is supported by the original research: Ghorban et al. (24) noted in their discussion that FMT’s positive effects diminished at 3 months, indicating that repeated transplants may be necessary to maintain the potential benefits. Similarly, follow-up duration was identified as a factor influencing HbA1c changes (p = 0.014). Craven et al. (15) also noted in exploratory analyses that FMT may require an extended time to potentially improve glucose and lipid metabolism disorders.

2) Second, differences in patient baseline characteristics. Meta-regression analysis revealed that patients’ baseline metabolic status (i.e., whether they were diabetic) was one factor influencing HbA1c changes (p = 0.016). A potential “floor effect” may exist in this study, where subjects had relatively well-controlled baseline metabolism, limiting the scope for FMT to produce significant improvements. For example, Yu et al. (16) and Gómez-Pérez et al. (21) both noted in their discussions that their subjects had only mild to moderate insulin resistance or well-controlled baseline metabolism. In contrast, Mocanu et al. (27) found that baseline hyperinsulinemia was an independent predictor of improved insulin resistance after FMT. Yu et al. (16) also observed, in their exploratory analysis, that FMT showed potential improvements in total cholesterol and HbA1c only in the subgroup with “low baseline microbial diversity.” Furthermore, Ng et al. (25) and Mocanu et al. (27) noted that FMT may require adjunctive measures to enhance efficacy, such as combined dietary interventions or high-fermentable fiber supplementation. These differences in study design collectively suggest that pooling data from studies with varying intervention frequencies, observation periods, participant baselines, and adjunctive measures may dilute or obscure positive effects present in specific subgroups.

3) Beyond factors quantified through meta-regression analysis, donor-related heterogeneity remains a key contributor to inconsistent FMT efficacy. Currently, included studies predominantly employ “healthy lean donors” as selection criteria, yet this functional definition exhibits considerable breadth. For instance, among the studies included in this meta-analysis, Witjes et al. (26) restricted donors to individuals following a plant-based diet, illustrating the lack of operational consistency in the “healthy” criterion. In recent years, the concept of “super-donors” has gained attention, suggesting that fecal microbiota from certain specific donors may yield superior clinical outcomes compared to others (33, 34). This variability is reflected in the included studies: Groenewegen et al. (20) observed that the downward trend in triglyceride levels was primarily driven by a single donor; Yu et al. (16) also reported significant “graft variability, “ where microbial communities from different donors exhibited differing colonization efficiency and persistence in recipients, potentially directly contributing to inconsistent metabolic responses. Therefore, combining data from donors with diverse functional backgrounds may result in potential positive effects being “averaged out” or diluted. Mocanu et al. (27) proactively assessed donor allocation across groups in their design to deliberately exclude interference from “super donor” phenomena, confirming that donor selection is a recognized key variable influencing FMT efficacy. Yu et al. (16) also recommended that future studies incorporate “donor-recipient pre-selection.”

From a mechanistic perspective, this variation in functional potential has been revealed in recent studies. Clinical response correlates with the successful engraftment of specific functional microbiota. For instance, one study linked weight loss in responders to the successful engraftment of Phascolarctobacterium genus and Acidaminococcaceae family members (both involved in short-chain fatty acid SCFA production) from donor sources (35). Moreover, the functional potential of donors extends beyond the production of beneficial substances like SCFAs. It also encompasses their ability to modulate bile acid profiles (e.g., converting primary bile acids into secondary bile acids, DCA and LCA) or clear harmful metabolites (e.g., ethanolamine), which are key signaling molecules regulating host glucose and lipid metabolism (36, 37). The donor’s own systemic metabolic phenotype has been found to potentially transfer to the recipient. A study by de Groot et al. (38) demonstrated that fecal transplantation from a metabolically impaired obese donor (METS-D) significantly worsened peripheral insulin sensitivity in recipients (p < 0.01). This finding was accompanied by deterioration in the recipient’s metabolome, notably elevated levels of secondary bile acids LCA and DCA. These findings suggest that an effective donor may require a functionally active microbiota-metabolite complex capable of correcting recipient-specific defects. Therefore, current donor screening criteria, which focus solely on excluding infectious pathogens, may be insufficient to ensure the efficacy of FMT in metabolic diseases (39).

4) Regarding administration routes, there is no clear consensus on an optimal approach. Among the trials included in this study, upper gastrointestinal routes (including oral capsules, nasojejunal tubes, and endoscopic infusion) were the primary methods, while the colonic route was less frequently employed. Different administration routes not only determine the initial colonization site and survival rate of the microbiota within the gut but may also introduce confounding factors. For example, Yu et al. (16) speculated that their oral capsule approach (without antibiotic pretreatment or bowel preparation) might be one reason for the lower graft efficiency and therapeutic differences compared to the endoscopic administration used in the Dutch study (12). da Ponte Neto et al. (22), who employed the colonoscopic route, noted that this approach “may not be optimal.” Craven et al. (15) further pointed out that even with duodenal administration, subsequent analysis of stool samples alone may not fully capture the actual microbial changes in the small intestine or proximal colon.

5) Key disagreements also exist regarding control group design. Autologous FMT and inert placebo are the two primary control forms. Autologous FMT is widely regarded as a stricter placebo because it more closely mimics the full experience of the transplant process. Although our subgroup analysis failed to detect statistically significant differences between control types due to the small number of studies within each group, this does not imply that such differences do not exist. The choice of placebo type may still influence trial outcomes (40).

First, the certainty of evidence for this analysis was rated as “Low” according to the GRADE approach. The downgrading primarily stemmed from two aspects: (1) Risk of Bias: Among the 11 included RCTs, 9 (81.8%) were rated as having “some concerns.” These concerns were largely focused on the randomization process and deviations from intended interventions, potentially reflecting limitations in random sequence generation or suboptimal blinding efficacy, which may compromise the internal validity of individual trials. (2) Imprecision: The 95% confidence intervals for the pooled effect sizes of all primary outcomes spanned the null value, indicating that the current pooled data are insufficient to yield precise effect estimates.

Second, the robustness of this study’s primary conclusions is constrained by sensitivity to statistical assumptions. The negative result for BMI (p = 0.070) in the main analysis (based on Corr = 0.5) became statistically significant (p = 0.010) in the sensitivity analysis when the correlation coefficient was adjusted to 0.75 (Supplementary Table S3). The leave-one-out analysis (Supplementary Table S2) for BMI also demonstrated this; the primary pooled result was non-significant (p = 0.070), whereas the pooled result excluding the study by Gómez-Pérez et al. (21) was statistically significant (p = 0.050). This high sensitivity to statistical assumptions indicates that the negative conclusions of the primary analysis should not be interpreted as definitive evidence of FMT ineffectiveness, but rather suggest that current evidence relies heavily on statistical estimation methods.

Third, incomplete outcome reporting across the included RCTs restricted the data available for the pooled analysis. Consequently, effective sample sizes for the primary outcome (BMI) and secondary parameters (e.g., triglycerides) were reduced to only 5 and 4 trials, respectively. This likely left these specific analyses underpowered, potentially preventing the observed trends from reaching statistical significance. Therefore, future trials must prioritize the comprehensive reporting of these metabolic endpoints to allow for more robust conclusions.

Fourth, the variability of control interventions complicates the interpretation of our findings. The mix of autologous FMT, inert placebos, and active drugs (e.g., metformin) makes it challenging to assess the true efficacy of FMT. Notably, active comparators can improve outcomes in the control arm, thereby narrowing the performance gap between groups and potentially masking the intervention’s benefit. Future research should aim for consistent control protocols to ensure more reliable comparisons.

Finally, as fewer than 10 studies were included for each outcome measure, funnel plots or Egger’s test could not be used to quantitatively assess publication bias. Consequently, the potential impact of publication bias on the pooled effect size cannot be ruled out.

Based on current evidence of “low” certainty, this systematic review and meta-analysis indicate that FMT did not show overall statistically significant improvement in key metabolic indicators for overweight or obese patients with metabolic diseases in the primary analysis. However, this negative conclusion exhibits instability in sensitivity analyses (i.e., insufficient robustness) and is likely highly influenced by methodological heterogeneity (e.g., administration protocols) and donor functional variability. Therefore, the true effectiveness of FMT in this setting remains uncertain. Future trials should focus on precision medicine, standardizing donor selection, and optimizing administration protocols.