Efficacy and Outcomes of Faecal Microbiota Transplantation for Recurrent Clostridioides difficile Infection in Children with Inflammatory Bowel Disease

Abstract

Background

Children with inflammatory bowel disease [IBD] are disproportionally affected by recurrent Clostridioides difficile infection [rCDI]. Although faecal microbiota transplantation [FMT] has been used with good efficacy in adults with IBD, little is known about outcomes associated with FMT in paediatric IBD.

Methods

We performed a retrospective review of FMT at 20 paediatric centres in the USA from March 2012 to March 2020. Children with and without IBD were compared with determined differences in the efficacy of FMT for rCDI. In addition, children with IBD with and without a successful outcome were compared with determined predictors of success. Safety data and IBD-specific outcomes were obtained.

Results

A total of 396 paediatric patients, including 148 with IBD, were included. Children with IBD were no less likely to have a successful first FMT then the non-IBD affected cohort [76% vs 81%, p = 0.17]. Among children with IBD, patients were more likely to have a successful FMT if they received FMT with fresh stool [p = 0.03], were without diarrhoea prior to FMT [p = 0.03], or had a shorter time from rCDI diagnosis until FMT [p = 0.04]. Children with a failed FMT were more likely to have clinically active IBD post-FMT [p = 0.002] and 19 [13%] patients had an IBD-related hospitalisation in the 3-month follow-up.

Conclusions

Based on the findings from this large US multicentre cohort, the efficacy of FMT for the treatment of rCDI did not differ in children with IBD. Failed FMT among children with IBD was possibly related to the presence of clinically active IBD.

1. Introduction

Rates of Clostridioides difficile infection [CDI] have increased in the past two decades in children and adults.¹ Paediatric patients with inflammatory bowel disease [IBD] are disproportionately affected, with CDI rates over 12 times their non-IBD peers.² They are also at increased risk of severe and recurrent CDI [rCDI] and more likely to have a severe course of their underlying IBD.^3–5

Faecal microbiota transplantation [FMT], although considered investigation by the U.S. Food and Drug Administration, is a highly successful treatment for rCDI with efficacy of 81% after a single FMT in our previously published cohort of pediatric patients.⁶ However, efficacy concerns remain for FMT in patients with IBD. Data on FMT efficacy for the treatment of rCDI in adult patients with IBD are mixed, with some studies demonstrating lower efficacy in IBD patients when compared with a non-IBD cohort⁷ and others without efficacy differences.⁸ A recent meta-analysis identified the presence of IBD as a predictor of FMT failure when used for the treatment of rCDI.⁹ In addition, flares are well described in IBD patients after FMT, with rates of 13–26% reported,^7,10,11 suggesting safety concerns that warrant additional attention. Data on the efficacy and safety of FMT for the treatment of rCDI in paediatric patients with IBD remain limited to case reports and case series.^12,13 Here, we report a retrospective multicentre study to compare the efficacy rates of FMT for the treatment of rCDI in paediatric patients with and without IBD, with additional evaluation of safety and IBD-specific outcomes in a large paediatric IBD cohort.

2. Methods

2.1. Setting and participants

This multicentre retrospective study includes paediatric and young adult patients [≤21 years of age] who underwent FMT at 20 paediatric centres across the USA for a diagnosis of rCDI, from March 2012 to March 2020. Centres were recruited through the North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition [NASPGHAN] FMT Special Interest Group and included Boston Children’s Hospital [coordinating centre], Vanderbilt University Medical Center, Children’s Hospital of Philadelphia, Mayo Clinic, Connecticut Children’s Hospital, Primary Children’s Hospital/University of Utah Pediatrics, Johns Hopkins Children’s Center, Pediatric Specialists of Virginia, Seattle Children’s Hospital, Children’s Healthcare of Atlanta, Children’s Wisconsin, Children’s Mercy Kansas City, Children’s Hospital of Colorado, MassGeneral Hospital for Children, Baylor University, University of Chicago, New York Presbyterian Komansky Children’s Hospital, Nationwide Children’s Hospital, Cedars Sinai Medical Center, and the University of Southern California.

2.2. Data collection

Study data, including demographics, FMT practices, CDI outcomes, IBD characteristics, and post-FMT complications, were collected and managed using REDCap [Research Electronic Data Capture] tools hosted at Vanderbilt University Medical Center.¹⁴ Data were imported into REDCap from the review of clinical records by individual institutions.

The primary study aim was to determine the success rate of FMT in paediatric and young adult patients [≤21 years old, hereby referred to as children] with a diagnosis of IBD,= undergoing FMT for rCDI, and to compare this with a non-IBD affected cohort. Recurrent CDI as an FMT indication was defined as an episode of CDI occurring within 8 weeks of a previous infection per standard definitions.¹ The number and treatment of recurrences prior to undergoing FMT were captured. Successful FMT was defined as no recurrence of CDI within 3 months following FMT. Recurrence of CDI required both a return of diarrhoea and positive testing for C. difficile based on the accepted laboratory testing at a given study site (e.g., nucleic acid amplification test [NAAT] alone or two-step algorithm including stool toxin test). Routine collection of stools post-FMT was not undertaken. Patients with less than 3 months of follow-up, hence undefined recurrence status, were excluded from this analysis. Patients who had FMT performed for a non-rCDI indication and those ≥21 years of age at the time of FMT were also excluded.

The secondary study aim focused on the paediatric IBD cohort only, with a goal to identify factors associated with a successful FMT among patients with a diagnosis of IBD [Aim 2a]. Additional data were collected on IBD-related outcomes after FMT [Aim 2b]. Adverse events [AEs] were not fully evaluated due to the retrospective nature of this study. Severe adverse events [SAEs] were collected in the IBD cohort and included death, life-threatening events, hospitalisations, or other important medical events.

2.3. Statistical methods

Data are presented as frequency [percentage] if categorical, and median [interquartile range; IQR] if continuous. Unadjusted comparison of patient characteristics in those with versus without IBD [Aim 1], as well as successful FMT versus unsuccessful FMT among IBD patients [Aim 2], was made by logistic regression using Firth’s penalised maximum likelihood.¹⁵

Adjusted comparisons of successful versus unsuccessful FMT was explored with two models: one to compare children with and without IBD after adjusting for potential independent predictors; and another to explore predictors of response among children with IBD. Due to the presence of missing data among the independent predictors, multiple imputation with fully conditional specification for arbitrary missing data pattern was used. Candidate predictors included variables listed in Tables 1 and 2, except for extent, severity, location, and behaviour of IBD, which are phenotype-specific and would lead to structural zeros [e.g., extent of ulcerative colitis would not be applicable for patients with Crohn’s disease]. Backward elimination with retention-threshold p <0.05 was used to identify independent predictors of response. Logistic regression with year as a continuous variable was used to determine if there was a change in FMT success in children with IBD over the study period. All tests of significance were two-sided with alpha = 0.05. Statistical analysis was performed with SAS^® version 9.4 [Cary, NC].

Table 1.

Unadjusted comparison of faecal microbiota transplantation [FMT] for Clostridioides difficile infection among paediatric subjects with [N = 148] and without [N = 248] inflammatory bowel disease [IBD].

Group [reduction in sample size shown in parentheses]	IBD [n = 148]	Non-IBD [n = 248]	OR [95% CI]	p  a
Age at FMT, median [IQR]	14 [9, 16]	5 [2, 13]	1.18 [1.13 – 1.23]	<0.0001
Female [vs male] sex	64 [43%]	122 [49%]	0.79 [0.52 – 1.19]	0.25
Caucasian [vs all other] race	134 [90%]	219 [88%]	1.25 [0.64 – 2.44]	0.52
Asian	1 [1%]	6 [2%]
Black	10 [7%]	7 [3%]
Native American	0 [0%]	1 [1%]
Hawaiian or Pacific Islander	0 [0%]	0 [0%]
Unknown race	3 [2%]	18 [7%]
Type of delivery [non-IBD, n = 244]				0.09
Sigmoidoscopy/colonoscopy	105 [71%]	181 [74%]	0.37 [0.15 – 0.91]
Upper delivery	30 [20%]	55 [23%]	0.35 [0.13 – 0.93]
Capsule	13 [9%]	8 [3%]	Reference
Fresh [vs frozen] stool [IBD n = 147]	36 [24%]	107 [43%]	0.43 [0.27 – 0.68]	0.0003
Stool source [IBD n = 147; non-IBD n = 247]				0.0006
Local stool bank	25 [17%]	35 [14%]	2.18 [1.15 – 4.12]
Commercial stool bank	87 [59%]	105 [43%]	2.51 [1.56 – 4.04]
Patient-selected	36 [24%]	107 [43%]	Reference
Bowel lavage prior to FMT [IBD n = 146; non-IBD n = 241]	116 [79%]	208 [86%]	0.61 [0.36 – 1.06]	0.08
Loperamide use post-FMT [IBD n = 147; non-IBD n = 244]	50 [34%]	90 [37%]	0.88 [0.58 – 1.36]	0.57
Diarrhoea at time of FMT [prior to colonoscopy preparation] [IBD n = 134; non-IBD n = 209]	82 [61%]	119 [57%]	1.19 [0.76 – 1.85]	0.44
Number of CDI episodes prior to FMT, median [IQR] [IBD n = 143; non-IBD n = 241]	3 [3, 4]	4 [3, 4]	0.98 [0.87 – 1.10]	0.73
Months from CDI diagnosis until FMT [IBD n = 142, non-IBD n = 233]	8 [5, 18]	7 [4, 12]	1.01 [1.00 – 1.03]	0.06

OR, odds ratio; CI, confidence interval; IQR, interquartile range.

^a  p-value from unadjusted logistic regression.

Table 2.

Unadjusted comparison of clinical predictors of response to faecal microbiota transplantation [FMT] for Clostridioides difficile infection among paediatric subjects with inflammatory bowel disease [IBD] [N = 148].

Group [reduction in sample size shown in parentheses]	No response [n = 36]	Response [n = 112]	OR [95% CI]	p  a
Age at FMT, median [IQR]	13 [8, 16]	14 [10, 16]	1.06 [0.97 – 1.15]	0.19
Female sex	15 [42%]	49 [44%]	1.08 [0.51 – 2.30]	0.84
Caucasian race	32 [89%]	102 [91%]	1.35 [0.40 – 4.53]	0.63
IBD type				0.74
Crohn’s disease	15 [42%]	51 [46%]	1.79 [0.40 – 7.89]
Ulcerative colitis [UC]	18 [50%]	55 [49%]	1.62 [0.37 – 7.02]
IBD unclassified	3 [8%]	6 [5%]	Reference
Extent of UC [no response n = 18, response n = 50]				0.73
Ulcerative proctitis [ref: pancolitis]	1 [6%]	1 [2%]	0.42 [0.03 – 7.24]
Left-sided UC [ref: pancolitis]	2 [11%]	11 [22%]	1.95 [0.42 – 9.12]
Extensive [ref: pancolitis]	0 [0%]	2 [4%]	2.12 [0.05 – 91.8]
Pancolitis	15 [83%]	36 [72%]	Reference
Ever severe UC [no response n = 17, response n = 50]	12 [71%]	25 [50%]	0.44 [0.14 – 1.41]	0.17
Crohn’s disease location [no response n = 18, response n = 57]
Distal 1/3 ileum/limited caecal disease	1 [6%]	5 [9%]	1.22 [0.16 – 9.26]	0.85
Colonic	8 [44%]	19 [33%]	0.63 [0.21 – 1.83]	0.39
Ileocolonic	9 [50%]	30 [53%]	1.11 [0.39 – 3.17]	0.85
Upper proximal to ligament of Treitz	5 [28%]	15 [26%]	0.90 [0.28 – 2.89]	0.85
Upper proximal to ligament of Treitz and proximal to distal 1/3 ileum	1 [6%]	1 [2%]	0.31 [0.02 – 5.21]	0.42
Crohn’s behaviour [no response n = 15, response n = 51]
Non-stricturing, non-penetrating [ref: penetrating, structuring, perianal]	11 [73%]	42 [82%]	1.75 [0.46 – 6.66]	0.41
Therapy for IBD
5-aminosalicylic acid	22 [61%]	64 [57%]	0.86 [0.40 – 1.84]	0.69
Immunomodulator	17 [47%]	51 [46%]	0.93 [0.44 – 1.97]	0.86
Biologic	19 [53%]	53 [47%]	0.81 [0.38 – 1.71]	0.57
Prednisone	20 [56%]	44 [39%]	0.52 [0.25 – 1.11]	0.09
Topical steroid	5 [14%]	11 [10%]	0.65 [0.21 – 1.99]	0.45
Antibiotic	8 [22%]	19 [17%]	0.70 [0.28 – 1.76]	0.45
Enteral therapy	3 [8%]	7 [6%]	0.68 [0.17 – 2.72]	0.59
Disease activity within 1 month prior to FMT
Clinical remission	4 [11%]	20 [18%]	1.60 [0.52 – 4.89]	0.41
Clinically active [gastrointestinal symptoms]	32 [89%]	83 [74%]	0.39 [0.13 – 1.16]	0.09
Biologic remission [no laboratory evidence of inflammation]	3 [8%]	11 [10%]	1.08 [0.30 – 3.97]	0.90
Biologic disease activity	12 [33%]	50 [45%]	1.58 [0.73 – 3.46]	0.25
History of bowel surgery	5 [14%]	12 [11%]	0.71 [0.24 – 2.16]	0.55
Hospitalisation within 1 year prior to FMT	11 [31%]	28 [25%]	0.75 [0.33 – 1.71]	0.49

OR, odds ratio; CI, confidence interval; IQR, interquartile range.

^a  p-value from unadjusted logistic regression.

2.4. Ethical considerations

The institutional review boards of all institutions approved the study.

3. Results

There were 461 paediatric patients who received FMT for rCDI, enrolled across 20 centres, from March 2012 to March 2020. Patients were excluded if they received FMT for a non-rCDI indication [n = 23, 17 with IBD], were ≥21 years of age [n = 2, 0 with IBD], or they had insufficient follow-up to determine recurrence status [n = 40, 8 with IBD]. The final cohort numbers for analysis included 248 non-IBD patients and 148 with IBD [Figure 1].

Figure 1.

Schematic of patient selection.

Children with and without a diagnosis of IBD who underwent FMT for rCDI were compared. Children with IBD were older than non-IBD treated patients [median age 14 years [IQR 9, 16] vs 5 years [2, 13], p <0.001] and less likely to receive FMT with a fresh, donor- identified sample [24% vs 43%, p <0.001] [Table 1]. Children with IBD were no less likely to have a successful first FMT than their non-IBD counterparts [112/148, 76% versus 202/248, 81%; p = 0.17] on unadjusted analysis. This finding was unchanged after adjustment for independent predictors of FMT success in the logistic regression model.

Among children with IBD, IBD type was classified as Crohn’s disease in 66 [45%] patients, as ulcerative colitis in 73 [49%] patients, and was unclassified [IBD-U] in nine [6%] patients. IBD was clinically active in most children, based on physician assessment of gastrointestinal symptoms in the month prior to FMT [n = 124, 84%], and 39 [26%] had been hospitalised, for any cause, in the year prior to FMT. The median [IQR] number of CDI episodes prior to undergoing FMT was 3 [3, 4]. Treatment of CDI prior to undergoing FMT included: metronidazole [n = 112, 76%], a standard course of oral vancomycin [n = 134, 91%], oral vancomycin pulse or taper [n = 102, 69%], fidaxomicin [n = 14, 9%], rifaximin [n = 10, 7%], and nitazoxanide [n = 6, 4%].

In children with IBD, FMT was performed by colonoscopy/sigmoidoscopy in 105 [71%], upper gastrointestinal tract delivery via esophagogastroduodenoscopy [EGD], nasogastric, nasoduodenal, or nasojejunal tube in 30 [20%] and capsule in 13 [9%] [Table 1]. The source of donor material was commercial stool bank in 87 [59%], local stool bank in 25 [17%], and patient-selected stool in 36 [24%]. A bowel lavage prior to FMT was performed in 116 [79%] patients and loperamide was used after FMT in 50 [34%] patients. Patient-selected material was not frozen prior to FMT. Donor material from commercial stool banks was provided by OpenBiome and storage and administration protocols followed.¹⁶ Protocols from local stool banks differed slightly but all were frozen at -80° and thawed and administered within 6 h of FMT delivery.

There was no significant change in FMT success in children with IBD over the study period based on logistic regression modelling [Figure 2]. In the 36 [24%] patients with IBD who had a failed FMT, additional CDI treatments included the following: vancomycin [n = 24], metronidazole [n = 3], fidaxomicin [n = 3], and repeat FMT [n = 13]. In patients who underwent a repeat FMT, 7/13 [54%] were successful.

Figure 2.

Success of faecal wmicrobiota transplantation in children with inflammatory bowel disease did not substantially change over the study period; p-value calculated from logistic regression with year as a continuous variable.

Children with IBD with and without a treatment failure were compared on clinical variables [Table 2]. The proportion of FMT failures did not differ by type of IBD, IBD disease activity, or IBD treatment prior to FMT. Among patients who had a treatment failure, there was a higher percentage who had clinically active IBD in the month prior to FMT than those with a successful FMT [89% versus 74%, p = 0.09], although this was not statistically significant.

Children with IBD were then compared on FMT procedure-related characteristics [see Table 3 for unadjusted comparisons]. Backwards elimination logistic regression determined three independent predictors of response to FMT. Children with IBD who received fresh stool from a patient-directed donor were more likely to have a successful response to FMT than those who received FMT with thawed, previously frozen, stool {odds ratio (OR) 3.73 (95% confodence interval [CI] 1.12 – 12.42); p = 0.03}, as were children without [vs with] diarrhoea at the time of FMT [after C. difficile-targeted antibiotic therapy and prior to FMT cleanout preparation] (OR = 1.67 [95% CI 1.04 – 2.67]; p = 0.03). Shorter time from initial CDI diagnosis until FMT was also independently predictive of a successful response (O = 0.97 [95% CI 0.95 – 0.99]; p = 0.04) in children with IBD.

Table 3.

Unadjusted comparison of procedure-related predictors of response to faecal microbiota transplantation [FMT] for Clostridioides difficile infection [CDI] among paediatric subjects with inflammatory bowel disease [IBD] [N = 148].

Group [reduction in sample size shown in parentheses]	No response [n = 36]	Response [n = 112]	OR [95% CI]	p  a
Type of delivery				0.18
Sigmoidoscopy/colonoscopy	23 [64%]	82 [73%]	1.12 [0.43 – 2.91]
Capsule	6 [17%]	7 [6%]	0.37 [0.09 – 1.45]
Upper delivery	7 [19%]	23 [21%]	Reference
Frozen stool [vs fresh] [n = 147]	33 [92%]	78 [70%]	0.24 [0.07 – 0.80]	0.02
Local stool bank	10 [28%]	15 [13%]	0.40 [0.16 – 0.99]	0.02
Commercial stool bank	23 [64%]	64 [57%]	0.76 [0.35 – 1.65]	0.45
Patient-selected stool	2 [6%]	34 [30%]	6.06 [1.55 – 23.68]	0.01
Bowel lavage prior to FMT [response n = 110]	26 [72%]	90 [82%]	1.75 [0.73 – 4.18]	0.21
Loperamide use post-FMT [no response n = 35]	9 [26%]	41 [37%]	1.62 [0.70 – 3.76]	0.26
Diarrhoea at time of FMT [prior to colonoscopy preparation] [no response n = 34; response n = 100]	27 [79%]	55 [55%]	0.33 [0.14 – 0.82]	0.02
Abnormal gross appearance of colonoscopy [no response n = 22; response n = 80]	15 [68%]	58 [73%]	1.26 [0.46 – 3.46]	0.66
Number of CDI episodes prior to FMT, median [IQR] [no response n = 33, response n = 110]	4 [3, 5]	3 [3, 4]	0.86 [0.71 – 1.03]	0.09
Months from CDI diagnosis until FMT [no response n = 31; response n = 111]	15 [7, 26]	8 [5, 14]	0.97 [0.95 – 0.99]	0.02

OR, odds ratio; CI, confidence interval; IQR, interquartile range.

^a  pvalue from unadjusted logistic regression.

The safety of FMT in children with IBD and IBD-specific outcomes were also assessed. There were 29 SAEs among 27 subjects in the 3 months following FMT, including 20 hospitalisations [19 related to IBD and one related to pancreatitis] and nine IBD-related surgeries. There were no deaths during the follow-up period. An additional 44 children [30%] were started on a new medication for IBD in the 3-month follow-up period. These safety and outcome variables did not differ in paediatric IBD patients with or without a successful FMT [Table 4].

Table 4.

Unadjusted comparison of post-faecal microbiota transplantation [FMT] variables among paediatric subjects with inflammatory bowel disease [IBD].

Group [reduction in sample size show in parentheses]	No Response [n = 36]	Response [n = 112]	OR [95% CI]	p*
Hospitalisation related to IBD within 3 months post-FMT	7 [19%]	12 [11%]	0.49 [0.18 – 1.35]	0.17
IBD-related surgery within 3 months post-FMT	1 [3%]	8 [7%]	1.92 [0.30 – 12.47]	0.49
New medication within 3 months post-FMT	14 [39%]	43 [38%]	0.97 [0.45 – 2.09]	0.94
Clinically active IBD within 3 months post-FMT	31 [86%]	61 [54%]	0.21 [0.08 – 0.56]	0.002
IBD clinical response 3 months post-FMT [no response n = 32, response n = 103]				0.07
Improved	7 [22%]	44 [43%]	1.79 [0.55 – 5.88]
No change	19 [59%]	38 [37%]	0.60 [0.21 – 1.70]
Worsened	6 [19%]	21 [20%]	Reference

OR, odds ratio; CI, confidence interval.

^a  p-value from unadjusted logistic regression.

In the 3 months after FMT and based on treating provider assessment, 20% of patients had worsening of their IBD clinical course, 42% had no change, and 38% had improvement in their IBD status. The patients who had worsening of their IBD following FMT did not differ by type of IBD, IBD activity, or IBD treatment when compared with those who had no change or improvement in their IBD status. In the 3 months after FMT, 92 [62%] children had clinically active IBD and 56 [38%] were in clinical remission. A higher proportion of children with an FMT failure had clinically active IBD after FMT than those with a successful FMT (31 [86%] vs 61 [54%], p = 0.002) [Table 4].

4. Discussion

This is the largest multicentre study in paediatric IBD patients to evaluate the efficacy, safety, and IBD-specific outcomes associated with the use of FMT for the treatment of rCDI. For Aim 1 of our study, to compare paediatric IBD patients with a non-IBD affected cohort, we found that children with IBD had good efficacy with a single FMT for the treatment of rCDI [76%], and they were no less likely to have a successful single FMT than their non-IBD counterparts [112/148, 76% versus 202/248, 81%; p = 0.17], even when adjusted for additional covariates in the model. A prior systematic review of adults with IBD identified an initial cure rate of 81%,¹¹ similar to what we identified in our paediatric study, although there remains significant discordance in findings between other studies. Most recently, an open-label, prospective cohort study of FMT for CDI in adults with IBD demonstrated an FMT failure of only 10.2%,¹⁷ which is lower than reported in prior retrospective studies.^7,10 The reasons for these discordant results are unclear, but may reflect differences in patient selection or treatment protocols.

Additional studies have attempted to identify differences in the microbial communities that may occur after FMT for the treatment of CDI in patients with and without IBD, also with discordant results. In adult patients treated with FMT for a diagnosis of CDI, Hirten et al. observed significant improvement in alpha diversity, which was maintained through the 12-month follow-up period, in both IBD patients and non-IBD treated controls.¹⁸ However, Khanna et al. demonstrated that after FMT, patients with IBD demonstrated a decreased growth of new taxa and increased number of CDI relapses in the 24 months following FMT when compared with non-IBD controls.¹⁹ In a single-centre paediatric study of FMT for rCDI that included eight children, five of whom had IBD, bacterial diversity increased in all patients after FMT, on initial analysis. However, at 6-month follow-up, bacterial diversity in those with IBD returned to pre-FMT baseline.²⁰ Our study did not include microbiota analysis, but this should be included in future studies as it may help elucidate mechanistic data and inform the successful use of FMT in children with IBD.

In our study, paediatric patients with IBD who received FMT with fresh versus thawed [previously frozen] stool were more likely to have a successful outcome [OR = 3.73; p = 0.03]. We identified this predictor on our prior paediatric FMT study as well,⁶ although an adult randomised clinical trial of fresh versus frozen FMT product for the treatment of CDI did not identify efficacy differences.²¹ Another recent retrospective study of 50 adult patients with CDI did identify superior efficacy when patients received FMT via a higher-volume fresh faecal filtrate when compared with those receiving FMT with a lower-volume frozen faecal preparation. It is unknown if alterations in efficacy were related to faecal preparation or volume, as both differed per study design.²² We previously hypothesised that alterations in the microbiome or metabolome that occur during a freeze-thaw cycle may make frozen donor stool less appropriate for paediatric patients, but additional study is warranted.⁶ In addition, it is possible that some patients receiving FMT from a patient-directed donor may have received stool from a paediatric source, which would not have occurred in the setting of a commercial or local stool bank. The effect of an age-matched donor on paediatric FMT is not fully established, but a recent study demonstrated improved efficacy of FMT for the treatment of IBD when a smaller difference existed between donor and recipient age [0–10 years difference versus ≥11 years difference, p = 0.003].²³

Children in our study were less likely to have a successful FMT if they had diarrhoea at the time of FMT [after completing CDI antibiotic therapy but prior to undergoing colonoscopy preparation] or if they had clinically active IBD after FMT. This may suggest that the presence of clinically active IBD is a risk factor for FMT failure, although the presence of endoscopic findings at the time of colonoscopy [for those who had a colonoscopy] was not predictive of FMT outcome. IBD treatment escalation before pursuing FMT in children may be warranted and may improve FMT efficacy. An alternative explanation for these findings may be related to increased C. difficile testing that occurs in the setting of active IBD and high rates of C. difficile colonisation in children with IBD.²⁴

Issues around colonisation affect the diagnostic yield of current testing options, and the appropriate diagnostic test for CDI, particularly in patients with additional comorbidities, remains uncertain. A limitation of our study is that we did not obtain the type of C. difficile diagnostic testing used, which may have varied between centres and possibly changed during the study period. As NAAT testing was approved by the US FDA in 2009, and was widely adopted as the preferred laboratory diagnostic tool,²⁵ we anticipate that the majority of diagnostic testing performed during our study was NAAT. The 2017 Clinical Practice Guidelines for Clostridium Difficile Infection in Adults and Children [published in 2018] recommended the use of a stool toxin test as part of a multistep algorithm.¹ Some centres may have adopted this approach during the remaining 2 years of the study, which may have altered CDI diagnosis and treatment. However, the best test for CDI in children, particularly those with comorbidities, remains a research priority. Parnell et al. identified that, in a cohort of 88 children including 17 with IBD, a two-step testing algorithm involving NAAT-based testing plus enzyme immunoassay was unable to differentiate between symptomatic and colonised children.²⁶ In addition, the recent American College of Gastroenterology C. difficile guidelines note that, due to limitations in all current testing options, the decision to diagnose and treat CDI remains a clinical one.²⁷

Shorter time from rCDI diagnosis until FMT was also independently predictive of response [OR = 0.97; p = 0.04] in our paediatric IBD cohort. The optimal timing of FMT has not been well established, but most guidelines recommend the use of FMT only after at least 3 episodes of CDI.¹ Interestingly, Hocquart et al. demonstrated decreased mortality in hospitalised adult patients who underwent FMT for the initial treatment of CDI when compared with those in the non-FMT group.²⁸ In addition, Varier et al. demonstrated that FMT was a cost-saving intervention when compared with vancomycin.²⁹ FMT may be more successful when used earlier in the treatment of severe and recurrent CDI, and this needs to be considered in future treatment algorithms.

In our paediatric IBD cohort, hospitalisations related to IBD occurred in 19 [13%] patients in the 3 months after FMT. Unfortunately, it is not possible to ascertain whether these hospitalisations were attributable to the FMT, the patient’s underlying IBD, the treatment of a failed FMT with antibiotics, or a combination of factors. Similarly, it is difficult to determine whether FMT predisposes to exacerbation of IBD, or, conversely, that there is increased rCDI in patients with more severe IBD, a group already at higher risk for IBD-related complications including hospitalisations. Notably however, 26% of children in our study had been hospitalised in the 1 year prior to FMT, which may suggest a higher severity of IBD in our paediatric cohort. A prior meta-analysis identified IBD flares in 22.7% of adult patients undergoing FMT for any indication, although there was significant variability among studies.³⁰

Additional limitations of the current study include its retrospective nature and lack of microbiome analysis. In addition, centre-specific differences in FMT protocols may not have been fully measured by the collected variables and account for changes in efficacy and safety. Related to the retrospective study design, post-FMT IBD activity assessments were made at any point within the 3-month follow-up period, as timing of IBD-related follow-up between centres was variable. More appropriate assessments could be made at standard intervals in the setting of a prospective study. In addition, there continue to be significant difficulties diagnosing CDI in patients with IBD related to high rates of colonisation, with rates of 11–17% reported.^24,31 Various testing algorithms have been applied to improve diagnostic yield,¹ but none validated in children. We allowed C. difficile diagnostic approaches per institutional guidelines for this study, as recommendations changed during the 8-year study period. Future studies should focus on improving diagnostic algorithms in children, particularly those with comorbidities.

In conclusion, in our large US multicentre cohort study of pediatric patients undergoing FMT for the treatment of rCDI, we identified no differences in first FMT success in children with IBD when compared with children without IBD. Children with IBD were more likely to have a successful FMT if they received fresh versus previously frozen stool, were without diarrhoea prior to FMT, or had a shorter time from CDI diagnosis until FMT. In the 3 months following FMT, a higher proportion of children with an FMT failure had clinically active IBD than those with a successful FMT, and a small proportion were admitted for an IBD flare. The presence of active IBD may warrant treatment escalation prior to FMT to improve efficacy in children, although additional prospective studies are warranted.

The data underlying this article cannot be shared publicly due to the privacy of individuals who participated in the study. The data will be shared on reasonable request to the corresponding author through the creation of an institutional agreement.

Funding

This work was supported by Cures Within Reach [PI: Kahn], a National Institute of Allergy and Infectious Diseases K23 award [No. 1K23AI156132-01 to MRN], a National Institute of Child Health and Human Development K23 award [No. K23HD099240 to SKH], and a National Institutes of Health/National Center for Advancing Translational Sciences Grant Support [No. UL1 TR000445] for REDCap [Vanderbilt University].

Conflict of Interest

The authors have no financial conflicts of interest relevant to this publication to report.

Authors Contributions

MRN: assisted in study concept and design, created data collection form, oversaw acquisition of data, assisted with data analysis and interpretation of the data, drafted the manuscript and approved the manuscript as submitted. PDM: assisted in study concept and design, performed the statistical analysis and assisted in interpretation of the data, performed critical revision of the manuscript for important intellectual content, and approved the manuscript as submitted. EA, SB, ZD, MD, MD, JG, AG, SKH, MKJ, JLK, RK, JK, MK, SK, EDK, JL, JDL, SM, MOH, TP, KQ, SS, ABS, DLS, MW, SW, EFD: assisted in study concept and design, imported data, performed critical revision of the manuscript for important intellectual content and approved the manuscript as submitted. SAK: assisted in study concept and design, recruited study sites, obtained study funding, oversaw acquisition of data, participated in interpretation of the data, performed critical revision of the manuscript for important intellectual content, and approved the final manuscript as submitted.

Preliminary data were presented at Digestive Diseases Week, May 2018, Washington DC.