Low Cure Rates in Controlled Trials of Fecal Microbiota Transplantation for Recurrent Clostridium difficile Infection: A Systematic Review and Meta-analysis

Abstract

Background

Fecal microbiota transplantation (FMT) is highly effective for treating recurrent Clostridium difficile infection (CDI) in observational studies (>90%), but cure rates in clinical trials are lower. We performed a systematic review and meta-analysis to assess the efficacy of FMT for recurrent CDI in open-label studies and clinical trials .

Methods

A systematic search from January 1978 to March 2017 was performed to include clinical trials of FMT for CDI. We analyzed CDI resolution by calculating weighted pooled rates (WPRs).

Results

Thirteen trials were included, comprising 610 patients with CDI treated with single FMT. Overall, 439 patients had clinical cure (WPR, 76.1%; 95% confidence interval (CI), 66.4%–85.7%). There was significant heterogeneity among studies (I² = 91.35%). Cure rates were lower in randomized trials (139/216 patients; WPR, 67.7%; 95% CI, 54.2%–81.3%) than in open-label studies (300/394 patients; WPR, 82.7%; 71.1%–94.3%) (P < .001). Subgroup analysis by FMT delivery modality showed lower cure rates with enema than colonoscopy (WPR, 66.3% vs 87.4%; P < .001) but no difference between colonoscopy and oral delivery (WPR, 87.4% vs 81.4%; P = .17). Lower rates were seen for studies including both recurrent and refractory CDI than for those including only recurrent CDI (WPR, 63.9% vs 79%; P < .001).

Conclusions

FMT was associated with lower cure rates in randomized trials than in open-label and in observational studies. Colonoscopy and oral route are more effective than enema for stool delivery. The efficacy also seems to be higher for recurrent than for refractory CDI.

clinical cure, Clostridium difficile infection, controlled trials, fecal microbiota transplantation, meta-analysis

Clostridium difficile infection (CDI) is a common cause of nosocomial and community-acquired diarrhea and has an incidence of 453000 cases annually in the United States [1]. Additional healthcare costs for an episode of CDI are estimated at $21448 for each hospitalization [2]. The efficacy of standard antibiotic therapy for a primary CDI episode is between 58% and 78% and decreases substantially after the first recurrence or multiple recurrences [3, 4]. In addition, antibiotics lead to further disruption of the intestinal flora and increased susceptibility to recurrent CDI. The rates of recurrent CDI are increasing despite stabilization in the incidence of primary CDI [5, 6].

Fecal microbiota transplantation (FMT) is used to treat recurrent CDI through restoration of normal gut microbiota by infusion of healthy donor stool. The efficacy of FMT has been shown to be >85% in observational studies [7], but the efficacy described in clinical trials has been variable. A trial reported in 2016 demonstrated that FMT was more effective than placebo (90% vs 62.5%) [8]. In contrast, another randomized controlled trial (RCT) of 30 patients demonstrated that FMT via enema was associated with a resolution rate of 43.8%, compared with resolution in 53.8% in a vancomycin taper group [9]. These reports suggest a difference in apparent clinical efficacy between real-life settings and clinical trials, as has been reported for other conditions such as inflammatory bowel disease [10]. Given these discrepancies, we performed a systematic review and meta-analysis to evaluate the efficacy of FMT in RCTs compared with that in open-label studies.

METHODS

We used the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines to conduct this meta-analysis [11].

Selection Criteria

The studies considered in this meta-analysis were clinical trials (open label, randomized trials with no control group, and placebo controlled) that included a study population of patients with recurrent or refractory CDI who were treated with FMT via any delivery modality. We excluded observational studies and studies that did not evaluate clinical resolution of CDI symptoms as an outcome.

Data Sources and Search Strategy

We conduced a comprehensive search of several electronic databases from January 1978 through March 2017. These databases included Ovid MEDLINE In-Process & Other Non-Indexed Citations, Ovid MEDLINE, Ovid Embase, Ovid Cochrane Central Register of Controlled Trials, Ovid Cochrane Database of Systematic Reviews, Web of Science, and Scopus, along with abstracts from major gastroenterology and infectious diseases conferences. The search strategy was designed and conducted independently by Mayo Clinic library staff and 2 study investigators (R. T. and S. K.). A controlled vocabulary supplemented with keywords was used to search for studies that used FMT for CDI. The following were the main keywords used in the search: Clostridium difficile, C diff, C difficile, Clostridium difficile infection, CDI, Clostridium difficile–associated diarrhea, or CDAD, AND faecal or faeces or fecal or feces or stool or microbiota, with infusion or transplant or transfer or instill or reconstitute or donor or bacteriotherapy. The search was limited to English-language publications. The detailed search strategy is shown in the Supplementary Material.

Two authors (R. T. and S. K.) independently reviewed the titles and abstracts of the identified studies, and those that did not answer the research question of interest were excluded. The remaining articles were reviewed to determine inclusion criteria fulfillment. The reference lists of articles with information on the topic were also reviewed for additional pertinent studies.

Data Abstraction

Data were independently abstracted to a predetermined collection form by 2 investigators (R. T. and S. K.). Data collected for each study included study setting and design, year of publication, location, number of patients, patient characteristics, indication for FMT, FMT modality, FMT donor type, duration of follow-up, and outcomes. Conflicts in data abstraction were resolved by consensus, referring to the original article.

Risk of Bias

The Cochrane Collaboration risk of bias tool was used to assess the methodologic quality of the included trials [12]. Risk-of-bias items were assessed, which included methods used to generate the randomization schedule and conceal allocation, blinding, completeness of outcome data, and evidence of selective outcome reporting (Supplementary Table 1).

Outcomes Assessed

Our primary analysis focused on assessing clinical resolution rates for recurrent CDI after initial FMT in a controlled setting. In our primary analysis, we did not include patients treated with multiple FMTs after clinical failure with initial FMT. We performed secondary analyses to calculate overall clinical cure rates by including patients for whom initial FMT failed and who were treated with multiple FMTs.

Statistical Analyses

Our primary outcome of the pooled analysis was clinical cure rates. The random-effects model described by DerSimonian and Laird [13] was used to calculate the weighted pooled rate (WPR). We calculated WPRs with corresponding 95% confidence interval (CIs) for the overall analysis as well as subgroup analyses. Data were weighted on the basis of sample size in each trial to calculate WPR. We assessed heterogeneity within groups with the I² statistic, which estimates the proportion of total variation across studies that is due to heterogeneity in study patients, design, or interventions rather than chance; I² values >50% suggest substantial heterogeneity [14]. All P values reported are 2 tailed. For all tests (except for heterogeneity), a P value <.05 was considered statistically significant. Calculations were performed and graphs constructed using Open MetaAnalyst software (version 10.10, developed by Tufts Evidence-based Practice Center under contract with the US Agency for Healthcare Research and Quality) [15]. Differences in clinical resolution rates between the subgroups were compared by calculating weighted proportion difference, using MedCalc software (version 17.4.4, MedCalc Software, Ostend, Belgium) which uses an “N − 1” χ² test [16].

A priori–defined sensitivity analyses included mode of FMT delivery and type of stool used for FMT. To explore the causes of heterogeneity and lower-than-expected resolution rates, we also performed subgroup analyses based on the type of donor used and the indication for FMT.

RESULTS

Search Results

The described search strategy retrieved 1056 unique studies; titles and abstracts were screened and relevant articles were obtained. Of the 44 potentially relevant articles, 31 were excluded for various reasons, leaving 13 studies that were included in this meta-analysis (Figure 1). Risk-of-bias items for the included studies are shown in Supplementary Table 1. Random sequence generation was used in all RCTs, and blinding of participants and personnel was done in 3 of 6. All the open-label trials and trials without a non-FMT group had a moderate risk of bias owing to lack of random sequence generation and blinding, given the lack of a control group.

Figure 1.

Flow diagram of included studies.

Characteristics of Included Studies

The characteristics of the 13 included studies are shown in Table 1. Eight studies were performed in the Unites States, 3 in Canada, 1 in the Netherlands, and 1 in Italy. The earliest study recruitment period began in January 2008, and the latest ended in April 2016. Ten studies focused on FMT for recurrent CDI, and 3 studies included both patients with recurrent and patients with refractory CDI [17–19]. Eleven studies used unrelated stool donors, 1 included both related and unrelated donors [8], and 1 included only related donors [9]. The various delivery modalities included colonoscopy, enema, oral capsules, nasoduodenal infusion, or nasogastric infusion [17]. Fresh stool was used for all patients in 3 studies and for some of the patients in 2 studies [19, 20], and frozen stool was used in the remaining studies. Follow-up after FMT ranged from 8 to 17 weeks.

Table 1.

Characteristics of Included Studies

Study (Year)	Study Design	Sample Size, No. of Patients	Location	Indication for FMT	Type of CDI Testing	Time Period	Stool Delivery Modality	Stool Type	Stool Volume	Donor Type	Follow-up, wk
van Nood et al [21] (2013)	RCT with placebo arm	42 (FMT group: 16)	The Netherlands	Recurrent CDI	PCR	January 2008 to April 2010	Nasoduodenal tube	Fresh	500 mL	Anon	10
Cammarota et al [22] (2015)	RCT with placebo arm	39 (FMT group: 20)	Italy	Recurrent CDI	NA	July 2013 to June 2014	Colonoscopy	Fresh	500 mL	Anon	10
Kelly et al [8] (2016)	RCT with placebo arm	46 (FMT group: 22)	United States	Recurrent CDI	PCR	November 2012 to March 2015	Colonoscopy	Fresh	100 g/500 mL	Both	8
Dubberke et al [23] (2016)	RCT with placebo arm	127 (FMT group: 83)	United States	Recurrent CDI	NA	December 2014 to November 2015	Enema	Microbiota based product	NA	Anon	8
Hota et al [9] (2017)	RCT with placebo arm	30 (FMT group: 16)	Canada	Recurrent CDI	EIA or PCR	NA	Enema	Fresh	50 g/500 mL	Known	17.14
SER-109 [24] (2016)	RCT with placebo arm	89 (FMT group: 59)	United States	Recurrent CDI	NA	NA	Oral	Frozen	NA	Anon	8
Youngster et al [17] (2014)	Open-label trial	20	United States	Recurrent + refractory CDI	EIA	NA	Nasogastric tube (n = 10) or colonoscopy (n = 10)	Frozen	41 g/340 mL	Anon	8
Youngster et al [18] (2014)	Open-label trial	20	United States	Recurrent + refractory CDI	EIA	August 2013 to June 2014	Oral	Frozen	30 capsules (48 g of stool)	Anon	8
Kao et al [25] (2015)	RCT without non-FMT group	43	Canada	Recurrent CDI	EIA or PCR	NA	Oral (n = 22) or colonoscopy (n = 21)	Frozen	100 g/400 mL	Anon	NA
Orenstein et al [26] (2016)	Open-label trial	31	United States	Recurrent CDI	NA	August–December 2013	Enema	Frozen	50 g/150 mL	Anon	8
Lee et al [19] (2016)	RCT without non-FMT group	178a	Canada	Recurrent and refractory CDI	EIA or PCR	July 2012 to September 2014	Enema	Frozen or (n = 91) or fresh (n = 87)	100 g/300 mL	Anon	13
Khanna et al [27] (2016)	Open-label trial	30	United States	Recurrent CDI	NA	NA	Oral	Frozen	150 g/30 capsules	Anon	8
Jiang et al [20] (2017)	RCT without non-FMT group	72	United States	Recurrent CDI	EIA	September 2013 to April 2016	Colonoscopy	Fresh (n = 25) Frozen (n = 24) or lyophilized (n = 23)	50 g	Anon	8

Abbreviations: Anon, anonymous; CDI, Clostridium difficile infection; EIA, enzyme immunoassay; FMT, fecal microbiota transplantation; NA, not available; PCR, polymerase chain reaction; RCT, randomized controlled trial.

^aPer-protocol population was included in the analysis.

Outcomes reported in various trials included clinical cure (defined as absence of watery diarrhea only or diarrhea with negative CDI test results) only, or resolution with no further recurrence, or reduced risk of recurrence (Supplementary Table 2). The definitions of these outcomes varied among the studies. The mean age of patients in the studies varied from 51.5 to 73 years (Supplementary Table 2). Five studies reported the Charlson Comorbidity Index (range, 1–5.3). Patients with inflammatory bowel diseases were excluded in 4 studies. Most of the trials included patients with positive results on polymerase chain reaction (PCR) and enzyme immunoassay (EIA). Three trials included patients with EIA only, and 2 included patients with PCR only (Table 1). No serious adverse events were reported in any trial.

Clinical Cure

Overall Clinical Cure With Single FMT

Together, the 13 studies reported a total of 610 patients treated with FMT, of which 439 experienced clinical cure with a single FMT (WPR, 76.1%; 95% CI, 66.4%–85.7%). There was significant heterogeneity among studies, with an I² of 91.35% (Figure 2).

Figure 2.

Analysis of all included studies. Forest plot shows weighted clinical resolution rates of fecal microbiota transplantation in patients with Clostridium difficile infection. Ev/Trt indicates number of patients with cure/number of patients treated. Abbreviation: CI, confidence interval.

Overall Clinical Cure With Multiple FMTs

We performed an analysis to include patients for whom the first FMT failed and who had clinical cure with >1 FMT, via colonoscopy, oral capsules, or enema. The overall clinical cure rate (WPR) with multiple FMTs, including the patients with failure of initial FMT, was 89.0% (95% CI, 84.1%–94.0%) (Figure 3).

Figure 3.

Analysis of all studies including patients with Clostridium difficile infection (CDI) for whom the first fecal microbiota transplant (FMT) failed but who had clinical cure with >1 FMT. Forest plot shows weighted clinical resolution rates of FMT in patients with CDI. Ev/Trt indicates number of patients with cure/number of patients treated. Abbreviation: CI, confidence interval.

Clinical Cure in Controlled Trials With Control Arm

Among 6 trials with a control arm (a placebo or antibiotic comparator group) [8, 9, 21–24], 216 patients were treated with FMT, and 155 were assigned to placebo or standard-of-care antibiotics. Of the patients treated with FMT, 139 experienced clinical cure (WPR, 67.7%; 95% CI, 54.2%–81.3%) (Figure 4A), compared with 68 patients in the control arm having clinical resolution (WPR, 43.4%; 31.8%–51.1%).

Figure 4.

Analysis by study design. Forest plots show weighted clinical resolution rates of fecal microbiota transplantation (FMT) in patients with Clostridium difficile infection (CDI) in trials with a non-FMT comparator group (A) and in open-label trials (B). Ev/Trt indicates number of patients with cure/number of patients treated. Abbreviation: CI, confidence interval.

Clinical Cure Rate in Trials With No Comparator Group

Among the 7 trials with no control group [17–20,25–27], 394 patients were treated with FMT, with symptom resolution in 300 (WPR, 82.7%; 95% CI, 71.1%–94.3%) (Figure 4B). Comparison of cure rates between RCTs and open-label trials revealed a lower cure rate in RCTs (WPR, 67.7% [95% CI, 54.2%–81.3%] vs 82.7% [71.1%–94.3%]; P < .001).

Subgroup Analyses

To explore the possible reasons for the lower resolution rates in RCTs, we performed several subgroup analyses (Table 2).

Table 2.

Subgroup Analyses

Subgroups	WPR, % (95% CI)	Proportion Difference in WPR, % (95% CI)	P Value
Delivery modality
Colonoscopy	87.4 (79.7–95.2)	Colonoscopy vs enema: 21.1 (12.71–28.54);	<.001
Enema	66.3 (52.7–79.9)
Oral	81.5 (64.5–98.5)	Colonoscopy vs enema: 5.9 (−3.13 to 15.05)	.17
Stool type
Fresh	80.2 (63.3–97.1)	3.2 (−4.69 to 10.42)	.40
Frozen	77.0 (65.3–88.7)
CDI type
Recurrent	79.0 (69.1–88.8)	15.1 (7.36–22.91)	<.001
Recurrent and refractory	63.9 (57.5–70.3)
Donor type
Anonymous	76.9 (66.5–87.2)	8 (−6.51 to 25.40)	.26
Known and anonymous	68.6 (22.4–114.7)
Type of CDI testing
EIA	78.6 (65.2–92)	2.1 (−7.55– to 0.58)	.65
Both EIA/PCR or PCR alone	76.5 (56.9–96.1)

Abbreviations: CDI, Clostridium difficile infection; CI, confidence interval; EIA, enzyme immunoassay; PCR, polymerase chain reaction; WPR, weighted pooled rate.

Recurrent Versus Refractory CDI

Studies that used FMT only for recurrent CDI reported higher resolution rates than studies that used FMT for both recurrent and refractory CDI (WPR, 79% vs 63.9%; P < .001). This suggests lower efficacy for FMT in patients who have disease refractory to standard-of-care antibiotics.

Delivery Modality

The resolution rate was lower after FMT delivered by enema than by colonoscopy (WPR, 66.3% vs 87.4%; P < .001) but did not differ between colonoscopy and oral delivery routes (87.4% vs 81.5%; P = .17).

Fresh Versus Frozen Donor Stool

There were no differences in resolution rate between use of fresh versus frozen stool (WPR, 80.2% vs 77.0%; P = .40).

Stool Donor Type

There were no differences in resolution rate between use of stool from unrelated or both known and unrelated stool donors (WPR, 76.9% vs 68.6%; P = .26).

Type of CDI Testing

Of all the trials, 3 included patients with CID who had undergone EIA testing. There was no difference in the resolution rates between trials that used EIA versus both EIA and PCR or PCR alone (WPR, 78.6% vs 76.5%; P = .65).

DISCUSSION

In the current study, we demonstrate that the efficacy of FMT for CDI was much lower in RCTs than in observational studies. A previous meta-analysis described a primary cure rate of 89% with FMT, but no clinical trials were included in that study [7]. Another systematic review described data from 21 case studies, with a clinical resolution of 85% in patients treated with FMT for recurrent CDI [28].

The differences in the estimated effects of treatment between RCTs and observational studies may be due to application of strict inclusion and exclusion criteria in the controlled setting, which leads to inclusion of a small proportion of patients from daily clinical practice and limits the generalizability of results to patients seen in clinical practice. Study populations in RCTs might not be a true reflection of the patients seen in the real world, whereas results from observational studies could be attributed to a broader range of patients. It is also possible that the lower resolution rates in the clinical trials may be due to differences in reported primary outcomes (clinical resolution vs no recurrence) and strict definitions of clinical resolution and recurrence compared with those in real-life clinical settings, as seen in inflammatory bowel disease management [10].

The different modes of delivery of FMT and use of FMT for refractory CDI in several trials may also have affected the resolution rates. In addition, differences in stool processing, storage, and the amount of stool administered may also lead to differences in resolution rates. Another explanation for the difference could be that strict monitoring and appropriate data collection in trials might have overestimated the recurrence rates by identifying false-positive cases or carriers with sporadic diarrhea rather than persons with true infection.

We performed secondary analyses to include patients who had failure of initial FMT but who achieved cure after multiple FMTs. Of interest, the resolution rates with multiple FMTs were similar to those reported in observational studies.

Subgroup analysis on the basis of FMT delivery mode indicated that colonoscopy was 21% more effective than enema. However, there was no difference between resolution rates for FMT delivered via the oral route or colonoscopy. Several studies have suggested that FMT delivered via the lower gastrointestinal tract (colonoscopy and enema) is more effective than that delivered via the upper tract (nasogastric or nasojejunal tube, gastroscopy, or gastrostomy tube) [7, 29, 30]. It can be postulated that the presence of gastric acid may lead to lower viability of the transplanted bacteria and lower efficacy.

One systematic review described FMT by enema and colonoscopy as more effective than FMT delivered via gastroscopy or nasogastric tube (resolution rate, >85% vs 76.4%), with similar resolution rates (>85%) for both enema and colonoscopy [31]. However, our analysis showed superior efficacy for FMT delivered by colonoscopy, compared with delivery by enema. Theoretically, colonoscopy might be better than enemas because it can infuse stool throughout the entire colon, whereas enemas deliver material only to the left side of the colon. Furthermore, multiple enemas might be required, compared with a single colonoscopy to achieve resolution [32]. Enemas have been done without bowel preparation, which also may affect the efficacy rate. Our analysis shows that the oral route might be as effective as colonoscopy and could be easily administered without any need for a gastrointestinal procedure. It can be postulated that formulations of oral products might be protective against stomach acid, which may lead to equal efficacy compared with colonoscopy. An RCT comparing instillation via colonoscopy versus enema versus oral capsules is needed to evaluate the best modality for FMT.

Subgroup analysis based on the use of fresh versus frozen stool and use of related or unrelated donors demonstrated no significant differences in resolution rates. The use of frozen stool is more practical, logistically, than the use fresh stool. In addition, freezing stool after processing sometimes allows for dividing the donated stool, so that it could be used for >1 patient. This would decrease the number and frequency of donor screenings and potentially also the costs and operational difficulties. In addition, subgroup analysis based on the use of type of CDI testing (EIA vs PCR) demonstrated no significant difference in resolution rates.

A few trials included in our meta-analysis used FMT for both recurrent and refractory CDI. Hence, we performed subgroup analysis separating the studies that used FMT for recurrent CDI alone and studies that included patients with recurrent or refractory CDI. Our analysis showed that FMT was 15.1% more effective for treatment of recurrent CDI alone. This analysis also partly explained the lower cure rates seen in the overall analysis of RCTs. One previous systematic review reported the efficacy of FMT for refractory CDI to be 55%, compared with 85% for recurrent CDI [28]. In our meta-analysis, none of the trials recruited only patients with refractory CDI; hence, we could not compare the efficacy of FMT for refractory CDI alone.

The strengths of our study include a comprehensive literature review with a large patient population from clinical trials. Our study also has several limitations. A 2017 meta-analysis demonstrated that the reporting of FMT preparation and other aspects of FMT is low and heterogeneous [33]. The individual trials included in our meta-analysis varied in several ways, including trial design, patient population, definitions of recurrent CDI and resolution of CDI, tests used to diagnose recurrent CDI episodes, route of FMT delivery, use of fresh versus frozen stool, quantity of stool transplanted, and use of FMT for recurrent or refractory CDI. These different aspects led to substantial heterogeneity, which limits the generalizability of our results. The information regarding preparation of stool was only available in a few clinical trials. Given the different preparations used in each trial, we were unable to perform an analysis based on stool preparation. Another factor that affects FMT outcomes is concomitant antibiotic exposure. These factors were not uniformly studied or not reported. Thus, more uniform reporting of FMT trials should be promoted.

In conclusion, single FMT in an RCT setting may be associated with a lower cure rate than that in open-label series. Further studies with clear definitions of recurrence and response and use of less strict patient exclusion criteria might help explore the reasons for lower resolution rates in RCT settings. In particular, the definition of recurrent CDI should be limited to diarrheal stool ≥3 times per day for ≥2 days along with a positive stool test for CDI, to decrease overestimation of recurrent CDI. Large follow-up registries are needed to determine the true efficacy and safety profile of FMT for recurrent CDI.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Presented in part: American College of Gastroenterology Annual Scientific Conference, Orlando, Florida, 13–18 October 2017.

Notes

Author contributions. R.T. has contributed to study conception, data abstraction, data analysis and interpretation and manuscript writing. D.S.P. and M.G.B. has contributed to study conception, data interpretation and manuscript writing. S. K. had full access to all the data in the study. Takes responsibility for the integrity of the data and accuracy of the data analysis, and takes responsibility for the integrity of the work as a whole, from inception to published article.

Potential conflicts of interest. D. S. P. has served as a consultant for Assembly Biosciences, Merck, Seres Therapeutics, C3Jain Therapeutics, Nestlé, and Salix Pharmaceuticals. S. K. has served as a consultant for Rebiotix, Assembly Biosciences, and Summit Pharmaceuticals International and reports personal fees from Facile Therapeutics, Merck, Premier, Probio Tech, and Shire, outside the submitted work. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.