Faecal microbiota transplantation for treating acute lower gastrointestinal graft‐versus‐host disease after allogeneic haematopoietic stem cell transplantation

Objectives

This is a protocol for a Cochrane Review (intervention). The objectives are as follows:

To evaluate the benefits and harms of faecal microbiota transplantation for treating steroid‐refractory, acute lower gastrointestinal graft‐versus‐host disease in recipients of allogeneic haematopoietic stem cell transplantation, compared with standard care, observation, placebo, or other comparator interventions.

Background

Description of the condition

Allogeneic haematopoietic stem cell transplant (allo‐HSCT) is a procedure where a patient receives healthy blood‐forming stem cells from a genetically matched donor to replace their own diseased or damaged bone marrow. It is a potentially curative treatment for a range of malignant and non‐malignant haematologic disorders (conditions that affect the blood, bone marrow, and lymphatic systems), and is mainly used to cure severe blood cancers, immune disorders, and bone marrow diseases.

Acute graft‐versus‐host disease (aGVHD) is one of the most serious complications of allo‐HSCT [1]. It is an immune‐mediated complication in which immunocompetent donor T lymphocytes (i.e. mature, fully functional T cells capable of recognising and mounting an immune response against foreign antigens) recognise recipient tissues as foreign and mount an inflammatory attack [1]. aGVHD typically occurs within the first 100 days after stem cell transplantation. The gastrointestinal tract is a principal target organ in people with aGVHD, and lower gastrointestinal tract involvement is associated with diarrhoea, abdominal pain, gastrointestinal bleeding, and inferior survival outcomes [2].

Diagnosis of aGVHD is based on clinical features supported by histopathological assessment where appropriate, and severity is graded using established criteria such as the Glucksberg criteria [3]. First‐line treatment of aGVHD is steroids. Steroid‐refractory aGVHD is commonly defined as progression of graft‐versus‐host disease (GVHD) after 3 to 5 days of systemic corticosteroid therapy, or failure to achieve at least a partial response after 7 to 14 days of treatment. It is associated with poor prognosis and limited effective second‐line treatment options [4]. Despite advances in transplant practices and GVHD prophylaxis, clinically significant aGVHD requiring systemic therapy remains a major cause of morbidity and mortality following allo‐HSCT, highlighting the need for effective therapeutic strategies [5].

The incidence of aGVHD has reduced over time. In a large longitudinal cohort study spanning from 1976 to 2020, the cumulative incidence of grade II to IV aGVHD decreased from 47% before 2000, to 24% between 2000 and 2010, and to 16% after 2010, while grade III to IV aGVHD declined from 13% to approximately 4% in the modern era (2011 to 2020) [6]. However, contemporary multicentre data demonstrate that a substantial proportion of allo‐HSCT recipients continue to develop clinically significant aGVHD requiring systemic therapy [7]. Incidence varies according to donor type, conditioning intensity, graft source, and GVHD prophylaxis strategies [5, 8].

In the allo‐HSCT setting, conditioning refers to pre‐transplant chemotherapy or radiotherapy, or both, administered to eradicate residual disease, suppress recipient immunity, and prepare the bone marrow niche for donor engraftment. Conditioning regimens cause significant damage to the membrane that lines various cavities in the body and covers the surface of internal organs (mucosal injury) and disruption of the intestinal barrier. This can contribute to profound changes in the community of microorganisms that live in the gut (gut microbiota), which includes bacteria, viruses, and fungi, and subsequent imbalance in the gut microbiota (dysbiosis) and predisposition to aGVHD [9, 10].

Description of the intervention and how it might work

The intestinal microbiota plays an important role in maintaining intestinal barrier integrity, and regulating mucosal and systemic immune responses. In the allogeneic transplant setting, the gut microbiota is profoundly disrupted. This disruption is characterised by reduced microbial diversity (alpha diversity, referring to the richness and evenness of microbial species within an individual) and loss of protective commensal organisms, including butyrate‐producing bacteria that support epithelial integrity and immune tolerance. Such dysbiosis has been consistently associated with the development, severity, and outcomes of acute gastrointestinal GVHD [11, 12].

Faecal microbiota transplant (FMT) involves the administration of a processed stool sample from a healthy donor to a recipient. The suggested mechanism of action is the establishment of a new gut microbiota, restoring microbial diversity, improving regulatory T cell (T reg) function (a specialised immune cell that helps prevent the immune system from attacking the body's own tissues), and helping to restore the gut epithelial barrier [13]. In the context of allo‐HSCT, FMT has been investigated primarily in patients with acute lower gastrointestinal GVHD. The intervention has been used in both inpatient and outpatient transplant settings, most often as a second‐line or later‐line therapy following failure of standard immunosuppressive treatments.

In current practice, ruxolitinib remains standard second‐line therapy for steroid‐refractory patients. However, response failure and infectious toxicity are concerning, requiring investigation of FMT as an alternate strategy in such patients [14].

FMT may be administered via multiple routes, including lower gastrointestinal delivery (e.g. colonoscopy or enema), upper gastrointestinal delivery (e.g. nasogastric or nasojejunal tube), or oral administration using encapsulated preparations, depending on clinical context and study design. Despite growing interest in FMT for acute gastrointestinal GVHD, important safety concerns remain. These include the risk of infections, viral reactivation, and uncertainty regarding long‐term safety. Strict donor screening, standardisation of FMT preparation and administration, and careful post‐procedure monitoring are therefore essential, yet remain inconsistently reported across studies [15, 16, 17, 18].

We anticipate that the effect of FMT on acute lower gastrointestinal GVHD may vary due to other factors. These include age, route of administration (which may affect delivery to the colon and microbial exposure), frequency and formulation (which may influence the dose and durability of microbiome restoration), and timing of FMT relative to GVHD onset or steroid initiation (as dysbiosis and mucosal injury evolve rapidly after allo‐HSCT). Baseline GVHD severity may modify response because it reflects the extent of epithelial injury and inflammatory burden. In addition, co‐interventions (particularly concurrent immunosuppression) may influence both efficacy and harms and may contribute to between‐study differences.

The use and outcomes of FMT may vary across specific patient populations. Paediatric and adult allo‐HSCT recipients differ in baseline microbiome composition, immune reconstitution, and susceptibility to infectious complications, which may influence both the efficacy and safety of FMT. In addition, differences in access to donor screening and microbiome‐based therapies in resource‐limited settings may affect intervention delivery and outcomes. These considerations provide a clinical rationale for the planned subgroup analyses in this review.

Why it is important to do this review

Acute gastrointestinal GVHD is associated with high morbidity, mortality, and limited effective treatment options beyond first‐line steroid therapy. Although ruxolitinib is the standard second‐line treatment, incomplete response and high rates of infectious complications result in ongoing uncertainty regarding optimal second‐line and subsequent treatment strategies for this high‐risk population.

FMT has emerged as a potential therapeutic approach. Preliminary studies and early‐phase trials have reported encouraging results [19, 20, 21]. However, the evidence base remains heterogeneous, with important uncertainties regarding patient selection, intervention regimens, routes of administration, comparative effectiveness, and short‐term and long‐term safety in severely immunocompromised recipients [22].

A 2023 systematic review and meta‐analysis evaluated the safety and efficacy of FMT in people with acute gastrointestinal GVHD. However, it was limited by the small number of available studies, inclusion of predominantly early observational data, and restricted exploration of methodological quality, risk of bias, and sources of heterogeneity [23]. Since its publication, additional studies have emerged, and ongoing debate remains regarding the clinical role of FMT in this setting. An updated and methodologically rigorous systematic review is therefore required to address persistent uncertainties, evaluate the evolving evidence base, and inform clinical practice, guideline development, and future research.

This Cochrane review aims to systematically assess the benefits and harms of FMT for people with acute gastrointestinal GVHD, identify gaps in the current evidence, explore sources of heterogeneity, and provide a transparent and reliable summary of the available data to support evidence‐based decision‐making.

Objectives

To evaluate the benefits and harms of faecal microbiota transplantation for treating steroid‐refractory, acute lower gastrointestinal graft‐versus‐host disease in recipients of allogeneic haematopoietic stem cell transplantation, compared with standard care, observation, placebo, or other comparator interventions.

Methods

We will follow the Methodological Expectations for Cochrane Intervention Reviews (MECIR) [24], and PRISMA 2020 reporting guideline [25]. Key abbreviations and technical terms are defined in the glossary in Table.

1. Glossary.

Term	Definition
Acute graft‐versus‐host disease (aGVHD)	An immune‐mediated complication of allogeneic haematopoietic stem cell transplantation in which donor immune cells attack recipient tissues, commonly affecting the skin, liver and gastrointestinal tract.
Adverse event (AE)	Any untoward medical occurrence in a participant receiving an intervention, whether or not it is causally related to that intervention.
Allogeneic haematopoietic stem cell transplantation (allo‐HSCT)	Transplantation in which a patient receives blood‐forming stem cells from a donor (rather than their own cells).
Alpha diversity	A measure of the diversity of microorganisms within a single biological sample, reflecting the richness and evenness of taxa.
Butyrate‐producing bacteria	Gut microorganisms that produce butyrate, a short‐chain fatty acid generated through microbial fermentation.
Clostridioides difficileinfection	Infection caused by Clostridioides difficile, typically associated with antibiotic exposure and characterised by diarrhoea and colitis.
Cochran's Q	A Chi2 statistic used to test whether observed differences between study results are compatible with chance alone.
Complete response (CR)	Complete resolution of disease manifestations according to the response criteria used in a study.
Glucksberg criteria	A commonly used grading system for acute graft‐versus‐host disease based on organ involvement and severity.
Intracluster correlation coefficient (ICC)	A measure of the degree of similarity of outcomes among participants within the same cluster in a cluster‐randomised trial.
I‐squared (I²)	A statistic describing the percentage of total variation across study effect estimates that is due to heterogeneity rather than chance.
Last observation carried forward (LOCF)	A method of handling missing data in which the participant's last observed value is used to replace missing subsequent values.
Mantel‐Haenszel method	A statistical method commonly used to pool dichotomous outcomes in meta‐analysis.
Overall response rate (ORR)	The proportion of participants achieving either a complete or partial response, in accordance with the response criteria used in a study.
PRISMA flow diagram	A standard diagram used to document the flow of records through identification, screening, eligibility assessment and inclusion in a systematic review.
Rayyan	A web‐based software platform used to support screening and collaboration in systematic reviews.
Review Manager (RevMan)	Cochrane's software used for analysing systematic reviews.
Serious adverse event (SAE)	An adverse event that results in death, is life‐threatening, requires or prolongs hospitalisation, results in persistent disability or incapacity, causes a congenital anomaly, or is considered medically important.
Short‐chain fatty acids	Fatty acids with fewer than six carbon atoms produced mainly by microbial fermentation in the gut.
Steroid‐refractory acute graft‐versus‐host disease	Acute graft‐versus‐host disease that progresses despite systemic corticosteroid therapy or fails to achieve adequate clinical response within the expected time frame as specified by study definitions.
Synthesis Without Meta‐analysis (SWiM)	Guidance for reporting synthesis methods transparently when meta‐analysis is not appropriate or feasible.

Criteria for considering studies for this review

Types of studies

Study design

Inclusion criteria

We will include randomised controlled trials (RCTs), including parallel‐group, cross‐over, and cluster‐randomised trials, as they are the gold standard in clinical research due to their ability to minimise bias and establish causality. Although less rigorous, we will include quasi‐RCTs as they may still provide valuable insights when RCTs are not feasible. A quasi‐RCT is defined as a study where participants are allocated to intervention groups using a systematic, predictable rule that is not truly random (e.g. alternation/rotation, date of birth), making assignment deterministic and difficult to conceal.

Cross‐over RCTs will be eligible provided that appropriate washout periods are used and that outcomes are measured in a manner that minimises carry‐over effects. We will interpret cross‐over RCTs cautiously, recognising the potential for residual or irreversible treatment effects and the limited applicability of cross‐over designs in the context of acute GVHD.

We will include cluster‐RCTs if they meet appropriate methodological criteria, including proper adjustment for clustering in the analysis (e.g. accounting for the intra‐cluster correlation coefficient) in accordance with guidance in Chapter 23 of the Cochrane Handbook for Systematic Reviews of Interventions [26]. Cluster‐RCTs may be necessary in certain clinical settings, e.g. FMT trials in hospitalised participants and, when appropriately designed, provide strong evidence comparable to individual‐level RCTs.

In addition, we will include comparative non‐randomised studies of interventions (NRSIs), as defined in Chapter 24 of the Cochrane Handbook [27]. Eligible NRSIs will include prospective or retrospective cohort studies and case‐control studies that compare outcomes between participants receiving FMT and a clearly defined comparator group (e.g. placebo or standard of care). We will include eligible NRSIs because RCTs evaluating FMT for steroid‐refractory acute lower gastrointestinal GVHD are scarce, given the rarity of the condition, the clinical complexity of the patient population, and the practical and ethical challenges of conducting RCTs in severely immunocompromised patients with limited treatment options. In this context, well‐designed comparative NRSIs represent an important source of evidence and their inclusion is consistent with the guidance in Chapter 24 of the Cochrane Handbook [27].

Exclusion criteria

We will exclude non‐comparative designs, including single‐arm studies, case series, case reports, and uncontrolled before‐after studies.

Study characteristics

Language

We will not apply language restrictions. Where non‐English studies appear eligible, we will seek translated abstracts or full texts where feasible. We will list studies that we cannot adequately assess due to language barriers as ‘studies awaiting classification’.

Year of publication

Studies published from 2016 onwards will be included. We have set this restriction because FMT was initially used in clinical settings as a therapeutic intervention for GVHD after 2016. Before 2016, FMT was not used in a systematic manner for these indications.

Report status

We will include published reports, including peer‐reviewed journal articles and conference abstracts. We will do this as peer‐reviewed publications have undergone rigorous scientific review. In cases where high‐quality evidence from a study is only available in abstract form (e.g. conference proceedings), we will still include the study to avoid excluding potentially relevant data provided they meet other eligibility criteria.

Due to the heterogeneity of techniques, FMT source, and administration methodology, we will exclude unpublished reports or non‐peer‐reviewed sources (internal reports, pre‐prints, dissertations). They carry a higher risk of bias and have not undergone rigorous scientific scrutiny.

Types of participants

Inclusion criteria

• Population: children and adults (all sexes) who have undergone allo‐HSCT and have acute lower gastrointestinal aGVHD, defined by clinical features consistent with lower gastrointestinal involvement and graded/staged using standard criteria i.e. MAGIC or modified Glucksberg [3, 28], with endoscopic or histopathological confirmation, or both, where reported.

• Subset of participants: we will only include studies with mixed populations (e.g. different GVHD organs, steroid‐sensitive and steroid‐refractory disease) if we can extract separately and directly analyse outcome data for participants with steroid‐refractory acute lower gastrointestinal aGVHD. In case(s) where we cannot extract subgroup data, we will only include studies where ≥ 80% of participants meet the criteria for steroid‐refractory acute lower gastrointestinal aGVHD. We will explore this further in sensitivity analyses.

• Clinical subgroup: we will define steroid‐refractory acute lower gut GVHD as any of the following:

  • progression of aGVHD after at least three days of systemic steroid therapy at a dose equivalent to > 2 mg/kg; or

  • failure to achieve at least partial response by at least > seven days of systemic corticosteroid therapy at a dose equivalent to > 2 mg/kg; or

  • recurrence or worsening of aGVHD during steroid taper.

• Donor/transplant types: any allogeneic donor type (matched related, matched unrelated, haploidentical, cord blood).

• Comorbidities/concomitant therapy: studies including participants with co‐morbidities or treatment associated with GVHD and post‐transplant status will be eligible provided the main objective of the study is steroid‐refractory acute gastrointestinal GVHD and FMT is the main intervention. This is to ensure the inclusion of important and relevant data, as post‐transplant participants commonly have other co‐morbidities (e.g. active infections) and other prior and concomitant treatment for relevant or other conditions (e.g. immunosuppressive therapy).

Exclusion criteria

• Autologous HSCT (transplantation using the patient's own stem cells, as opposed to cells from a donor) recipients, including GVHD‐like syndromes after autologous transplantation

• Studies in which co‐interventions (e.g. corticosteroids, immunosuppressants, antibiotics, supportive care) are not applied consistently across study arms such that attribution of effects to FMT is not possible. Furthermore, we will exclude studies where FMT is initiated concurrently with new second‐line or third‐line systemic immunosuppressive agents (such that we cannot distinguish the independent effects of FMT).

• Mixed‐population studies where we cannot extract data specific to steroid‐refractory acute lower gastrointestinal aGVHD and where < 80% of the participants meet the subgroup criteria

• Studies evaluating autologous FMT (recipient‐derived stool), as the participant's faecal microbiome is unlikely to be diverse and healthy at baseline due to disease and prior treatment

Types of interventions

Allogeneic FMT (faecal microbiota transplantation using stool donated by a healthy third‐party donor) administered for treating acute gastrointestinal GVHD will be eligible for inclusion, regardless of formulation, route of administration, dose, frequency, or duration.

Studies should report sufficient detail on FMT characteristics, including dose (e.g. amount of stool or microbial preparation), frequency and number of administrations, route of delivery (e.g. nasoenteric, colonoscopic, enema, or oral capsule), and duration of treatment. We will extract these characteristics systematically, as they may influence treatment effects and contribute to clinical heterogeneity.

Co‐interventions

We will include studies that allow co‐interventions (e.g. corticosteroids, immunosuppressive agents, supportive care, and antibiotics) if the co‐interventions are consistently applied across both intervention and control groups. This is to avoid confounding effects. This is allowed as participants with GVHD often receive these concomitant interventions for both GVHD and unrelated conditions associated with post‐transplant status.

Comparisons

For people with steroid‐refractory acute lower gastrointestinal GVHD, we will compare the following:

• FMT plus standard care versus standard care alone;

• FMT versus placebo;

• FMT versus observation;

• FMT versus other active comparator interventions.

Outcome measures

Since there are no established core outcomes for our review question, we will synthesise the most commonly reported and clinically relevant outcomes. We will measure outcomes as described below.

Critical outcomes

We will assess the following critical and important outcomes at pre‐specified time points.

• Overall response rate (ORR) at day 28, defined as the proportion of participants achieving either a complete or partial response of acute lower gastrointestinal GVHD, assessed using validated response criteria (e.g. MAGIC Consortium criteria [28], Glucksberg criteria [3], or IBMTR criteria [29]

• Non‐relapse mortality, defined as death without prior relapse of the underlying disease

• Serious adverse events (SAEs) related to FMT, including life‐threatening events, sepsis, procedure‐related complications, or events resulting in death or prolonged hospitalisation

Important outcomes

• Complete response at day 28, defined as complete resolution of acute lower gastrointestinal GVHD

• Infectious complications, including bacterial, viral, and fungal infections occurring after FMT. We will define infectious complications as clinically and microbiologically new or worsening infections, or undiagnosed infections that arise as a direct or indirect result of acute gastrointestinal GVHD among participants within 30 days of FMT administration, including bacteraemia from gut‐derived organisms, Clostridioides difficile infection (CDI) (both new onset and recurrent), viral and fungal infections, and septicaemia unrelated to other sources of infection. We will not consider pre‐existing infections related to FMT administration.

• Non‐infectious adverse events related to FMT, including gastrointestinal bleeding, electrolyte disturbances, abdominal pain, bloating, nausea, and procedure‐related reactions

• Overall survival will include the proportion of participants alive at specific time points (e.g. 1, 3, 6, and 12 months) after receiving FMT. This is because steroid‐refractory acute GVHD is associated with around 75% mortality.

Additional outcomes

• Additional outcomes (where reported)

• Microbiota‐related outcomes, including measures of microbial diversity (e.g. alpha diversity), relative abundance of key taxa (e.g. Firmicutes/Bacteroidetes), and recovery of obligate anaerobes

• Patient‐reported outcomes, including quality of life and tolerability of the intervention, length of hospital stay assessed using validated instruments where available

We will analyse these outcomes narratively if quantitative synthesis is not feasible.

Outcome assessment timing

We will assess outcomes at prespecified, clinically relevant, fixed time points. For response outcomes (ORR and complete response), the primary time point of interest will be day 28 post‐intervention, consistent with MAGIC Consortium standards for acute GVHD assessment [28]. If outcomes are not reported exactly at day 28, we will extract data from the closest reported time point within the acute phase (up to day 56 post‐intervention), prioritising time points nearest to day 28.

Search methods for identification of studies

Electronic searches

We will develop the search strategy in collaboration with an Information Specialist (Ina Monsef) to identify published, unpublished, and ongoing studies, with no restrictions on language or date of publication. The strategy will be designed to maximise sensitivity while maintaining clinical relevance for studies evaluating faecal microbiota‐based interventions in acute gastrointestinal GVHD, including steroid‐refractory or steroid‐resistant disease. The search will combine controlled vocabulary (e.g. MeSH and Emtree terms) and free‐text terms relating to the following.

• Population: acute GVHD, gastrointestinal GVHD, lower gastrointestinal GVHD, steroid‐refractory GVHD, steroid‐resistant GVHD

• Intervention: faecal microbiota transplantation, faecal microbiota transfer, faecal microbiota transplantation/transfer, FMT

Database‐specific methodological filters will be applied to identify eligible study designs, where appropriate. For MEDLINE, we will use the Cochrane Highly Sensitive Search Strategy for identifying RCTs (sensitivity‐maximising version, 2023 revision) [30], a search filter for Phase 3 studies [31], and NRSIs [32]. For NRSIs, we will apply validated and database‐specific methodological filters for controlled non‐randomised studies where appropriate, with careful consideration to minimise the risk of excluding relevant studies. Our choice and adaptation of filters will be guided by the guidance in Chapter 4 of the Cochrane Handbook for Systematic Reviews of Interventions [30]. No language or publication status restrictions will be applied. We will provide the full search strategies for each database in the appendices.

The detailed MEDLINE search strategy is provided in Supplementary material 1. This will serve as the base strategy for translation to other databases. We will search the following sources from inception to the date of the final search:

• Cochrane Central Register of Controlled Trials (CENTRAL; current issue), via the Cochrane Library

• Ovid MEDLINE ALL

• Ovid Embase

• ClinicalTrials.gov (www.clinicaltrials.gov)

• World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (apps.who.int/trialsearch/)

We will update the final literature search immediately prior to review publication. For future updates of the Cochrane review, we will establish search surveillance. We will not apply any language restrictions.

Searching other resources

We will search the following sources.

• Grey literature sources, including Google Scholar

• Conference proceedings and abstracts from the:

  • American Society of Hematology (ASH)

  • European Society for Blood and Marrow Transplantation (EBMT)

  • American Society for Transplantation and Cellular Therapy (ASTCT)

• Before publication, we will check included studies for errata or retractions using PubMed and the Retraction Watch database.

• In accordance with the guidance in Chapter 4 of the Cochrane Handbook for Systematic Reviews of Interventions [30], we will also screen the reference lists of all included studies and relevant systematic reviews identified during the search process to identify additional eligible studies.

Data collection and analysis

Selection of studies

We will upload all retrieved records to Rayyan.ai [33], and will identify and remove duplicate records using Rayyan’s inbuilt de‐duplication function. We will use Rayyan to support the screening process. Two review authors (KaDA and MK) will independently screen titles and abstracts, followed by full‐text assessment of potentially eligible studies. Disagreements will be resolved by discussion, with arbitration by a third review author (RI) if necessary.

We will collate multiple reports of the same study and treat as a single study. The study selection process will be documented narratively and presented using a PRISMA flow diagram. We will include studies reported as full texts or conference abstracts if they provide sufficient information for data extraction and analysis. Where data are incomplete or unclear, we will attempt to contact study authors or identify corresponding full publications. If we are unable to obtain sufficient information, we will list such studies as ‘studies awaiting assessment’. We will document all attempts to contact study authors.

We will screen titles and abstracts irrespective of language. For potentially eligible studies published in languages other than English, we will seek translation support through Cochrane Engage, Google Translate, or institutional resources. If translation is not immediately feasible, we will classify such studies as ‘studies awaiting classification’. We will screen for duplicate reports by comparing participant numbers and study characteristics across publications. Where we identify discrepancies, we will seek clarification from the published report or will contact study authors if necessary.

Data extraction and management

Two review authors (MK, MM) will independently perform data extraction. Each review author will extract data using a standardised data extraction form developed specifically for this Cochrane review. The review authors will extract data extraction using the Rayyan data extraction module [33] if available; otherwise, a standardised Excel or REDCap‐compatible form will be used. We will pilot the data extraction form on a sample of three included studies to ensure clarity, comprehensiveness, and consistency in data collection. The review authors will resolve any disagreements through discussion, and a third review author (RI) will arbitrate if necessary.

We will extract the following information from each included study.

• Study information, including first author, year of publication, and reported conflicts of interest

• Study design, duration, and setting

• Participant characteristics, including age, sex, GVHD stage and grade, treatment allocation (intervention/comparator), and concomitant immunosuppressive therapies

• Intervention details (FMT), including formulation, dose, route, frequency, and duration

• Outcomes as defined in the review protocol

• Methods of outcome measurement (e.g. Glucksberg criteria for GVHD [3]) at baseline and after intervention

• Time points of outcome assessment

• Outcome metrics and summary statistics used to characterise participant results (e.g. changes in GVHD grade reported as continuous data)

• Methods of aggregation (e.g. mean and standard deviation (SD), proportion of participants experiencing an event)

• Adverse effects and infectious complications, including serious and severe adverse events

• Overall survival, where reported

We will summarise the characteristics of included studies in a ‘Characteristics of included studies’ table, and will provide an ‘Overview of Synthesis and Included Studies’ (OSIS) table.

Risk of bias assessment in included studies

Two review authors (GUS, AR) will independently assess the risk of bias for included studies in accordance with the guidance in Chapter 8 and 25 of the Cochrane Handbook for Systematic Reviews of Interventions [34, 35]. The review author will resolve any disagreements through discussion, and if necessary, by consulting a third review author (RI).

Randomised studies

For RCTs, we will use the Risk of Bias 2 (RoB 2) tool [36]. We will assess the risk of bias at the outcome level, rather than at the study level, and we will complete RoB 2 assessments for all critical and important outcomes included in the summary of findings table(s) at the primary time point of day 28 post‐intervention, or the closest available time point up to day 56 where day 28 data are not reported.

For each outcome, we will assess the risk of bias for the specific result used in the synthesis, following the guidance regarding selecting results for RoB 2 assessments in Chapter 8 of the Cochrane Handbook [34]. We will categorise judgements as either ‘low risk of bias’, ‘some concerns’, or ‘high risk of bias’ across the following domains.

• Bias arising from the randomisation process

• Bias due to deviations from intended interventions

• Bias due to missing outcome data

• Bias in measurement of the outcome

• Bias in selection of the reported result

For cluster‐randomised trials and cross‐over trials, we will use the appropriate RoB 2 extensions for these designs, which allow assessment of additional design‐specific issues such as recruitment bias and incorrect analysis in cluster trials, and carry‐over effects in cross‐over trials.

For randomised studies, we will assess the effect of assignment to intervention (intention‐to‐treat effect) for critical and important outcomes, regardless of adherence to the assigned intervention.

NRSIs

For comparative NRSIs, we will use the ROBINS‐I tool [37], adapting it as necessary based on study design characteristics, in line with the guidance in Chapter 25 of the Cochrane Handbook [35]. The effect of interest for NRSIs will be the effect of assignment to intervention, recognising that treatment changes and co‐interventions may occur over time.

ROBINS‐I assessments will consider the following domains.

• Bias due to confounding

• Bias in selection of participants into the study

• Bias in classification of interventions

• Bias due to deviations from intended interventions

• Bias due to missing data

• Bias in measurement of outcomes

• Bias in selection of the reported result

Prespecified key confounders for NRSIs will include baseline severity and extent of acute GVHD, timing of FMT in relation to GVHD onset and steroid initiation, concurrent immunosuppressive and antimicrobial therapies, conditioning intensity, donor and graft characteristics, GVHD prophylaxis strategy, and key FMT protocol characteristics (dose, frequency, route, formulation, and timing of administration). Where available, we will prioritise appropriately adjusted effect estimates in the synthesis.

Bias due to missing evidence

We will assess the risk of bias due to missing results at the synthesis level using the Risk of Bias due to Missing Evidence (ROB‐ME) tool, including from studies that use FMT but do not report GVHD outcomes using validated response criteria [38]. ROB‐ME will be applied to each meta‐analysis and summary of findings table to evaluate the potential impact of unavailable results, selective non‐reporting, and publication bias.

If uncertainties arise regarding reported methods or outcomes, we will attempt to contact study authors for clarification. We will document all contact attempts and responses.

Review authors who have been involved in the conduct, analysis, or publication of an included study will not participate in study eligibility decisions, data extraction, risk of bias assessment, or GRADE evaluations for that study, in accordance with the Cochrane conflict of interest policy [39].

Measures of treatment effect

Effect measures for synthesis

We will present effect estimates using both relative measures (e.g. risk ratios (RRs), odds ratios) and absolute measures, including absolute risk reduction (ARR) and number needed to treat for an additional beneficial outcome (NNTB) or number needed to treat for an additional harmful outcome (NNTH), where data permit in accordance with Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions [40]. We will summarise the results in tables, including summary of findings tables, to enhance clinical interpretability and support informed decision‐making.

For dichotomous outcomes, we will calculate RRs with 95% confidence intervals (CIs). Where event rates and data permit, we will also present absolute effects, including risk differences, NNTB, and NNTH, to enhance clinical interpretability, using an appropriate assumed baseline risk from the comparator group.

For continuous outcomes, we will use the mean difference (MD) with 95% CIs when outcomes are measured using the same scale across studies. When different measurement scales are used for the same construct, we will use the standardised mean difference (SMD) with 95% CIs.

We will interpret effect estimates in relation to clinical context, and where possible, using absolute effects and baseline risk. For outcomes analysed using SMDs, we will use Cohen's rule of thumb to help interpretation (0.2 SD = small effect, 0.5 SD = moderate effect, 0.8 SD = large effect). We recognise that these thresholds are merely a guide and clinical importance is outcome‐specific, in accordance with the guidance in Chapter 15 of the Cochrane Handbook [41].

Where absolute effects are presented, we will calculate them by applying relative effect estimates to an assumed baseline risk in the comparator group derived from included studies or relevant external data.

Unit of analysis issues

The primary unit of analysis will be the study, using aggregated published data. We will not routinely request or analyse individual participant data. If individual‐level data are provided by study authors, we will pre‐process these data and summarise into aggregate measures before inclusion in analyses. We will not pool them directly with published individual‐level data.

Multiple intervention groups (multi‐arm studies)

In studies with more than two arms, including multiple intervention arms (e.g. different doses or formulations of FMT) or multiple comparator arms (e.g. FMT versus placebo and FMT versus standard of care), we will take care to avoid unit‐of‐analysis errors and double counting of participants.

Where a study includes one shared intervention arm and multiple relevant comparator arms, we will not treat each comparison as independent within the same meta‐analysis. In accordance with guidance from Chapter 23 of the Cochrane Handbook [26], we will combine comparator groups where this is clinically appropriate. If combining groups is not appropriate, we will split the shared intervention or comparator arm by dividing the sample size (and number of events for dichotomous outcomes) equally across comparisons so that each participant contributes only once to each meta‐analysis.

For studies with multiple FMT intervention arms, we will combine intervention arms when clinically and methodologically appropriate. If combining is not appropriate, we will select a single, most relevant comparison a priori or split the comparator group accordingly.

Cluster‐randomised trials

For cluster‐randomised trials, we will account for clustering in the analysis. Where studies report effect estimates that have not been adjusted for clustering, we will adjust the effective sample size using the design effect, based on an intracluster correlation coefficient (ICC), in accordance with Chapter 23 of the Cochrane Handbook [26], to avoid artificially inflated precision. If an ICC is not reported, we will use ICC estimates from similar studies, report the source and justification of any external ICC, and conduct sensitivity analyses using alternative plausible ICC values to assess the robustness of findings.

Cross‐over trials

For cross‐over trials, we will address unit‐of‐analysis issues in accordance with Chapter 23 of the Cochrane Handbook [26]. To avoid potential carry‐over and washout effects, we will extract and use first‐period data only, treating these data as equivalent to results from a parallel‐group RCT for the purpose of synthesis.

Where first‐period data are not reported or cannot be isolated, we will not include the cross‐over study in quantitative synthesis and will instead describe it narratively. We will not use random‐effects or mixed‐effects within‐participant modelling for cross‐over trials, as we do not anticipate that individual participant data will be available. Due to the anticipated lack of reported correlation coefficients and the unavailability of individual patient data, more efficient analysis methods for cross‐over trials will not be possible. We will include first‐period data from cross‐over trials in the same meta‐analysis as parallel RCTs if they provide comparable outcome data at the prespecified time point of interest.

Repeated outcome measurements (multiple time points)

When studies report outcomes at multiple time points, we will extract data at pre‐specified, clinically relevant fixed time points, as defined in the Outcome measures section. The primary time point of interest for response outcomes will be day 28 post‐intervention, and we will analyse later time points (e.g. 3 months, 6 months) separately where sufficient data are available. We will not use mixed‐effects or other models to analyse repeated measures over time, as such models require correlation parameters (or individual participant data) that are not anticipated to be available from study reports.

Events that may recur

Some outcomes (e.g. infections or GVHD flares) may occur more than once in the same participant during follow‐up. In line with the guidance in Chapter 6 of the Cochrane Handbook [42], we will avoid double‐counting participants by prioritising the analysis of the number or proportion of participants experiencing at least one event during the follow‐up period. We will not use recurrent‐event models for meta‐analysis, as these require individual participant data and are not feasible with aggregate study‐level data. If studies only report total event counts without the number of participants affected, we will analyse these data separately and interpret them with caution due to the risk of unit‐of‐analysis errors.

Non‐randomised studies

For NRSIs, we will ensure that effect estimates are derived from the analytical population as defined by the study authors. Consideration of matching or weighting approaches used in these studies will inform risk‐of‐bias assessment and interpretation of results, rather than unit‐of‐analysis handling. For NRSIs, we will prioritise the most appropriately adjusted effect estimates to minimise confounding bias. We will extract details of all adjusted and unadjusted effect estimates reported, including the variables adjusted for in each study, and clearly indicate which estimates are adjusted. This approach aligns with guidance in Chapter 8 of the Cochrane Handbook [34].

Dealing with missing data

We will deal with missing data according to the recommendations in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions [40].

Assessment and contact with study investigators

For each included study, we will document the levels, patterns, and reported reasons for missing outcome data. Where crucial data for analysis (e.g. SDs, number of events, details of adjusted analyses) are missing or unclear, we attempt to contact the original study authors to request the necessary information.

Data imputation for meta‐analysis

Missing summary statistics

If measures of variance (e.g. SDs) are missing, and we cannot obtain them from the study authors, we will impute them using recommended methods (e.g. from other studies in the same meta‐analysis, using a coefficient of variation, or from standard errors, CIs, or P values as described in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions [40].

Missing participant data

We will primarily analyse the data as reported by the trial authors (e.g. intention‐to‐treat analyses). We will not use simple imputation methods, such as last observation carried forward (LOCF), at the review level. If the reported analysis is likely to be biased due to high or differential attrition, this will be reflected in the risk of bias assessment.

Sensitivity analysis

To assess the robustness of our conclusions to assumptions about missing participant data, we will perform sensitivity analyses for critical outcomes where missingness is substantial. This may involve comparing the primary analysis to plausible best‐case and worst‐case scenario imputations for missing participants.

Reporting bias assessment

We will apply the methods outlined in Chapter 13 of the Cochrane Handbook [43]. We will evaluate the risk of bias using the RoB 2 tool for RCTs. We will also assess reporting biases using graphical methods (such as funnel plots) for comparison of treatment effects across studies, particularly in cases where meta‐analysis is performed. If funnel plots suggest asymmetry on visual inspection, we will incorporate findings into our certainty of evidence, leading to a downgrade in the certainty of the evidence. If fewer than 10 studies are eligible for meta‐analysis, we will note our inability to rule out publication bias or small study effect. We will assess risk of bias due to missing results at the synthesis level using the Risk of Bias due to Missing Evidence (ROB‐ME) tool, as outlined in Chapter 13 of the Cochrane Handbook [38, 43]. ROB‐ME will be used to evaluate the potential impact of unavailable results, selective non‐reporting of outcomes, and publication bias on each meta‐analysis and corresponding summary of findings table. This assessment will complement, but not replace, study‐level risk of bias assessments conducted using appropriate tools.

For NRSIs, we will consider differences in FMT characteristics (including dose, frequency, duration, route, formulation, and timing of administration), as well as co‐interventions such as immunosuppressive therapy and baseline disease severity, as potential confounders. We will assess them within the confounding domain of the risk of bias assessment, in accordance with Chapter 25 of the Cochrane Handbook [35].

Synthesis methods

We will perform meta‐analyses of critical and important outcomes for the following comparisons, where data permit using RevMan [44].

• FMT plus standard care versus standard care alone

• FMT versus placebo

• FMT versus observation

• FMT versus other active comparator interventions used for the treatment of steroid‐refractory acute lower gastrointestinal GVHD

Quantitative synthesis will be undertaken only where studies are sufficiently similar in terms of populations, interventions, comparisons, and outcome definitions. We will conduct meta‐analyses using standard methods described in Chapter 10 of the Cochrane Handbook [40]. Use of a fixed‐effect model will be reserved for sensitivity analyses, as described in the Sensitivity analysis section.

We will use a random‐effects model, as we expect there will be clinical diversity. The Restricted Maximum Likelihood (REML) estimator will be used to estimate between‐trial variance [40]. The Hartung‐Knapp‐Sidik‐Jonkman method will be used to calculate a CI for the meta‐analysis effect estimate when there are at least 3 studies, and the estimate of heterogeneity is greater than zero [45]. In other scenarios (i.e. in pooled analyses of two studies, or where the estimate of heterogeneity is equal to zero) we will use the Wald‐type method [40].

Where meta‐analysis is not appropriate due to substantial clinical or methodological heterogeneity, or insufficient data, we will undertake synthesis without meta‐analysis (SWiM) [46], presenting results narratively and supported by structured tabulations and graphical displays, in accordance with Cochrane guidance. We will use forest plots for pooled analyses, and will use funnel plots to explore reporting bias where appropriate.

Handling of NRSIs

We will not combine RCTs and NRSIs in the same meta‐analysis. Separate syntheses will be conducted by study design, in accordance with Chapter 24 of the Cochrane Handbook [27].

Where sufficient and clinically homogeneous NRSIs are available, we will perform separate meta‐analyses using adjusted effect estimates, prioritising those accounting for key pre‐specified confounders (FMT characteristics, concomitant immunosuppression, and baseline GVHD severity). Where quantitative synthesis is not appropriate, we will summarise the findings from NRSIs using a structured narrative synthesis. We will assess the certainty of the evidence from NRSIs separately from RCT evidence within the GRADE framework.

Investigation of heterogeneity and subgroup analysis

Two review authors (MK, MM) will independently assess clinical and methodological heterogeneity across included studies, considering differences in study populations, interventions, comparators, and outcome definitions. We will undertake meta‐analysis only when both review authors agree that studies are sufficiently similar to justify quantitative synthesis.

We will assess statistical heterogeneity using the I² statistic, the Chi² test, and visual inspection of forest plots. Interpretation of heterogeneity will not rely on fixed thresholds for the I² statistic. Instead, we will assess the importance of inconsistency in context, taking into account the magnitude and direction of effect estimates, their precision, and the degree of clinical and methodological diversity across studies, in accordance with the guidance in Chapter 10 of the Cochrane Handbook [40].

Where we observe heterogeneity, we will explore potential sources through pre‐specified subgroup analyses, where sufficient data are available and where there is a clear clinical rationale. Where we undertake subgroup analyses, we will use the formal test for subgroup differences implemented in RevMan to assess whether effect estimates differ between subgroups beyond what would be expected by chance, in accordance with the guidance in Chapter 10 of the Cochrane Handbook [40]. We will undertake subgroup analyses cautiously and interpret them as exploratory, recognising the risks of confounding and multiple comparisons. Where subgroup analyses are not feasible, we will explore these factors narratively as potential sources of heterogeneity.

Pre‐specified subgroup analyses will include the following factors, where data permit. These factors are believed to impact the effects of FMT, as mentioned in the Background section.

• Route of FMT administration (upper gastrointestinal, lower gastrointestinal, oral capsule)

• Timing of FMT in relation to acute GVHD onset or steroid initiation

• Baseline severity of acute GVHD

• Age (adults versus paediatric population)

Sensitivity analysis

Where we perform meta‐analysis, we will repeat the primary random‐effects meta‐analysis (REML estimator [47]; Hartung‐Knapp‐Sidik‐Jonkman CIs when ≥ three studies [45, 48], and τ² > 0; Wald‐type method otherwise [40]) to assess the robustness of the review findings to key methodological decisions. We will repeat the analysis using a fixed‐effect model, applying the Mantel‐Haenszel method for dichotomous outcomes and the inverse variance method for continuous outcomes and an alternative estimator for between‐study variance (e.g. DerSimonian‐Laird) or an alternative method, or both, for CIs as appropriate.

First, to evaluate the influence of risk of bias on pooled estimates, we will repeat analyses excluding studies we judge to be at high risk of bias, based on the RoB 2 tool for randomised trials and the ROBINS‐I tool for NRSIs [36, 37]. Second, where we consider studies to be borderline in terms of eligibility criteria or data reporting, we will perform sensitivity analyses by including and excluding these studies to examine their impact on the direction and magnitude of effect estimates. Examples of borderline cases may include studies with unclear or non‐standard definitions of GVHD response, incomplete reporting of key FMT characteristics (such as route, dose, or timing), or imprecise alignment of outcome assessment with the prespecified time point of interest.

We will interpret sensitivity analyses through informal comparison of effect estimates and CIs between the primary and sensitivity analyses. No formal statistical tests will be used to compare sensitivity analyses. We will not use changes in P values alone to judge robustness, as loss of statistical significance may reflect reduced precision rather than true differences in effect.

We will report the results of sensitivity analyses narratively in the text.

Certainty of the evidence assessment

Two review authors will independently assess the certainty of the evidence assessments using the GRADE approach, as outlined in the GRADE Handbook [49], and Chapter 14 of the Cochrane Handbook [50], for all critical and important outcomes as pre‐specified in the Outcome measures section. This will be performed regardless of whether data are available, in order to transparently highlight gaps in the evidence base where no studies have collected data on outcomes considered of critical importance to the review question. The review authors will resolve any disagreements in judgments across GRADE domains (risk of bias, inconsistency, indirectness, imprecision, and publication bias) through discussion. If consensus cannot be reached, a third review author will adjudicate. We will document the final GRADE judgments and justifications transparently in the summary of findings tables, in accordance with guidance from Chapter 14 of the Cochrane Handbook [50], and MECIR standards [24]. We will prepare the summary of findings tables to present the certainty of evidence for a single, pre‐specified comparison per table, in accordance with Cochrane guidance.

We will prepare separate summary of findings tables for different comparator types (e.g. FMT versus placebo, FMT versus standard of care). Each table will include only one comparison.

Where multi‐arm trials contribute to more than one comparison, we will handle shared intervention arms in accordance with the Unit of analysis issues section (e.g. by combining comparator groups or splitting the shared arm), to ensure that participants are not double‐counted within any single meta‐analysis. We will assess the certainty of the evidence as follows.

• High certainty: we are very confident that the true effect lies close to the estimate of the effect.

• Moderate certainty: the true effect is likely to be close to the estimate, but there is a possibility that it is substantially different.

• Low certainty: the true effect may be substantially different from the estimate of the effect.

• Very low certainty: we have very little confidence in the effect estimate.

Interpretation of effect sizes

To support interpretation of findings, we will use pre‐specified thresholds informed by methodological guidance and commonly used conventions, recognising that clinical importance depends on context.

For dichotomous outcomes (RRs):

• small or trivial effect: RR 0.90 to 1.11;

• moderate effect: RR 0.75 to 0.90 or 1.11 to 1.33;

• large effect: RR < 0.75 or > 1.33.

These thresholds are consistent with the guidance in Chapter 15 of the Cochrane Handbook [41], and reflect differences commonly considered clinically important in serious conditions.

For continuous outcomes (standardised mean differences):

• small effect: 0.2;

• moderate effect: 0.5;

• large effect: 0.8.

These thresholds are based on Cohen’s conventional benchmarks and are widely used in Cochrane reviews. Interpretation will also consider the clinical relevance of each outcome; for example, even small effects on survival or remission may be clinically meaningful in steroid‐refractory acute lower gastrointestinal GVHD.

We will transparently document the reasons for downgrading the certainty of the evidence using footnotes and explanatory comments in the summary of findings tables. The wording of effect estimates and conclusions will follow recommended phrasing, as outlined in Chapter 15 of the Cochrane Handbook [41]. The summary of findings table(s) will present the certainty of evidence for each critical and important outcome associated with FMT for acute gut GVHD, with outcome terminology consistent across the Outcomes, Results, and Certainty of the evidence sections.

Equity considerations

This Cochrane review has a focused clinical scope, and we expect to include a limited number of eligible RCTs. We have not listed any exclusion criteria based on age or sex (see Types of participants section). However, we recognise that the populations represented in studies of FMT for acute lower gastrointestinal GVHD may not fully reflect all groups undergoing allo‐HSCT in routine practice.

In particular, available studies are anticipated to predominantly include adult recipients treated at specialised transplant centres, with limited representation of paediatric patients, older adults, or individuals with significant comorbidities. In addition, sex‐specific analyses are unlikely to be reported consistently, which may limit assessment of differential effects by sex.

We will consider these factors when we interpret the applicability and equity implications of the review findings. Where data permit, we will describe the age and sex distribution of included participants. Also, we will highlight gaps in evidence relevant to under‐represented populations, in accordance with guidance on equity considerations in Chapter 16 of the Cochrane Handbook [51].

Consumer involvement

We did not undertake any formal consumer or patient involvement (such as participation in advisory groups or workshops) during the development of this Cochrane protocol. This reflects the highly specialised clinical context of the review and practical constraints related to scope and timelines.

Nevertheless, we have designed the review to address outcomes of direct relevance to patients undergoing allo‐HSCT, including survival, treatment response, adverse events (particularly infectious complications), and healthcare utilisation. Consideration of these outcomes aims to enhance the applicability and usefulness of the review findings for patients, caregivers, and clinicians, in accordance with guidance on consumer involvement in Chapter 1 of the Cochrane Handbook [52].

Supporting Information

Supplementary materials are available with the online version of this article: 10.1002/14651858.CD016225.

Supplementary materials are published alongside the article and contain additional data and information that support or enhance the article. Supplementary materials may not be subject to the same editorial scrutiny as the content of the article and Cochrane has not copyedited, typeset or proofread these materials. The material in these sections has been supplied by the author(s) for publication under a Licence for Publication and the author(s) are solely responsible for the material. Cochrane accordingly gives no representations or warranties of any kind in relation to, and accepts no liability for any reliance on or use of, such material.

Supplementary material 1 Search strategies

New

Additional information

Contributions of authors

Raheel Iftikhar: conceptualisation, supervision, writing – review & editing

Maryam Khan: methodology, project administration, writing – original draft, writing – review & editing

Mahnoor Mahnoor: methodology, project administration, writing – original draft, writing – review & editing

Kashaf ad Duja Awais: methodology; project administration, writing – original draft, writing – review & editing

Amina Risalat: investigation, writing – review & editing

Mehwish Gilani: methodology, writing – review & editing

Ghassan Umair Shamshad: investigation, writing – review & editing

Sakina Sadiq Malik: supervision, writing – review & editing

Nicole Skoetz: methodology, supervision, writing – review & editing

Ina Monsef: investigation, methodology, writing – review & editing

Shahrukh Hashmi: supervision, writing – review & editing

Declarations of interest

RI has declared that they have no conflicts of interest.

MK has declared that they have no conflicts of interest.

MM has declared that they have no conflicts of interest.

KaDA has declared that they have no conflicts of interest.

AR has declared that they have no conflicts of interest.

MG has declared that they have no conflicts of interest.

GUS has declared that they have no conflicts of interest.

SSM has declared that they have no conflicts of interest.

NS is a Cochrane editor. She was not involved in the editorial or peer review process of this review.

IM is a Cochrane editor. She was not involved in the editorial or peer review process of this review.

SH has declared that they have no conflicts of interest.

Sources of support

Internal sources

• No internal sources of support, Other

  No internal sources of support

External sources

• No external sources of support, Other

  No external sources of support

Registration and protocol

Cochrane approved the proposal for this review on 05 August 2024.

Data, code and other materials

Data sharing is not applicable to this article as it is a protocol, so no datasets were generated or analysed.

Disclosure of artificial intelligence use

The review author team used artificial intelligence tools to support language editing, formatting, and structuring of the protocol. The authors take full responsibility for the content, accuracy, and interpretation of the work.