An open-label, non-randomized phase I/II clinical study was conducted at the First Affiliated Hospital of Soochow University. Protocols and other trial-related procedures were approved by the Institutional Review Board of the hospital. Written informed consent was obtained from all the patients. The FMT and control groups (without FMT) were set according to the patients’ decision after introducing the possible benefits and disadvantages of fecal bacteria transplantation. FMT was performed after steroid-refractory GI-GvHD had been diagnosed. All the patients received a second-line immunosuppressant treatment. Only the FMT group received FMT. The Center for International Blood and Marrow Transplant Research (CIBMTR) criteria were used to assess the grades of GI-GvHD (3, 21, 22). The criteria for diagnosing steroid-refractory gut GvHD were described previously (3). Patients with uncontrollable infection, irreversible organ failure, and other abnormal conditions that might interfere with the evaluation were excluded from the study (online supplement protocol).

The study was registered with ClinicalTrials.gov as #NCT03148743.

The fecal materials were handled in sealed, fully automatic machines (GenFMTer, Nanjing, China). The fecal microbiota samples collected from four healthy donors (two women aged 23 years and two men aged 20 years) were conserved at −80°C with glycerine (online supplement protocol). As these patients did not tolerate gastroscopy or enteroscopy, 40–50 mL of frozen fecal microbiota was suspended in 150–200 mL of warm normal saline and delivered into the intestine of the recipients through a nasojejunal or gastric tube after diagnosing grade IV steroid-refractory GI-GvHD (3). If there was no improvement, FMT was repeated in the following week.

The primary outcomes were event-free survival time (EFS) and overall survival (OS) on day 90 after the diagnosis of steroid-refractory GI-GvHD; EFS and OS after steroid-refractory GI-GvHD were recorded until November 1, 2018.

Secondary outcomes were clinical remission or partial remission on days 14, 21, and 28 after the diagnosis of steroid-refractory GI-GvHD.

The efficacy of FMT was evaluated according to the severity of symptoms, such as abdominal pain, diarrhea (frequency and volume), and bloody purulent stool within 14 and 21 days after FMT. The abdominal pain scores were assigned as follows: occasional pain (0.5), mild pain (1), moderate pain (2), severe pain without intervention (3), and severe pain (4). Clinical remission was defined as a condition in which diarrhea and intestinal spasms and/or bleeding disappeared or stool volume decreased by ≥500 mL on average within 3 days. Clinical improvement was defined as a condition in which the stool volume decreased by <500 mL, or the abdominal pain value and bleeding were relieved. EFS was defined as the period during the follow-up after the first FMT with no progression of GI-GvHD, no death, no GvHD involvement of other organs, and no new infection with cytomegalovirus (CMV) or Epstein–Barr virus (EBV) (3). OS referred to the period from the diagnosis of steroid-refractory GI-GvHD to November 1, 2018. All deaths in this period, including those related to relapse and those due to other causes, were included in the statistical analysis.

For each patient, safety was evaluated according to adverse events (including death or drop-out) during FMT and the follow-up period.

Fecal samples were stored at −80°C until DNA extraction. After DNA extraction, bacterial 16S rDNA was successfully detected in all of the samples by polymerase chain reaction (PCR) using general bacterial primers (16S V4-V5): 515F:5’-GTGCCAGCMGCCGCGGTAA-3’; 926R: 5’-CCGTCAATTCMTTTGA -GTTT-3’. After purification, the pooled libraries were sequenced by 2×300 bp paired-end sequencing on the MiSeq platform (Tiny Gene, Shanghai, China) using the MiSeq v3 Reagent Kit (Illumina). Mothur, UPARSE, and R software were used to analyze the 16S sequencing data. The composition of fecal bacteria was analyzed at the phylum level. Moreover, the Shannon diversity index was used to depict the diversity of the microbiota (online supplement methods) (3, 23).

The statistical software SPSS 16.0 (SPSS, Inc., IL, USA) was used to construct actuarial rate curves, to calculate log-rank hazard ratios (HRs) and Fisher’s exact tests, and to perform significance and risk determinations. Cochran’s and Mandel-Haenszel statistical methods were used to examine the differences between the groups. The ‘survival’ package in R statistical software (Vienna, Austria) was used for the permutation tests. Significance was determined using Cox proportional-hazards models with time-varying covariates.

A total of 55 patients with steroid-refractory GI-GvHD were enrolled in this study. C. difficile infection was not observed in any of the patients, and they did not respond to methylprednisolone (mPSL) at ≥2 mg/kg per day. Immunosuppressants as a second-line therapy were given to all the patients. Eight patients were excluded: Four patients were reluctant to participate in the study, while another four patients failed to meet the inclusion criteria (one for primary disease recurrence, two for combined thrombotic microangiopathy (TMA), and one for combined CMV before FMT). Of 26 patients in the FMT group, the data of three patients with < grade IV GI-GvHD were not selected for statistical analysis. Of the 21 patients in the control group, the data of three patients (one with missed follow-up and two with < grade IV GI-GvHD) were not used for statistical analysis ( Figure 1 ).

The patients’ characteristics were shown in Table 1 and Supplementary Table 1 . Immunosuppressive drugs used were shown in Supplementary Table 2 . The median age of the 23 patients in the FMT group was 30 years (range, 13–55 years). The male-to-female ratio was 16/7. The median stool volume was 660 mL/day (range, 360–2,080 mL/day). The median stool frequency was 6 times/day (range, 3–21 times/day). The median abdominal pain score was 3 (range, 1–4) ( Table 2 and Supplementary Table 3 ). For the 18 patients in the control group, the median age was 31.5 years (range, 13–59 years) (vs. FMT group, p>0.05). The male-to-female ratio was 7/11 (vs. FMT group, p>0.05). The median stool volume was 520 mL/day (range, 250–1,400 mL/day) (vs. FMT group, p<0.05). The median stool frequency was 5 times/day (range, 3–20 times/day) (vs. FMT group, p>0.05). The median abdominal pain score was 2 (range, 0–4) (vs. FMT group, p>0.05) ( Table 2 and Supplementary Table 3 ). No differences were observed in the occurrence of hematologic disease, stem cells donor gender match, or stem cell donor relationship between the two groups ( Table 1 and Supplementary Table 1 ). In the FMT group, 11, 9, 2, and 1 patients underwent 2, 1, 3, and 6 FMT sessions, respectively ( Supplementary Table 4 ).

Cox regression analysis showed that immunosuppressants did not affect the outcomes of the two groups ( Supplementary Table 2 ).

On day 14 (the day after the diagnosis of grade IV steroid-refractory GI-GvHD), 12 patients (52.2%) in the FMT group and none in the control group achieved clinical remission according to modified intention-to-treat analysis (p<0.05). Meanwhile, 19 patients (82.6%) in the FMT group and seven patients (39%) in the control group showed an effective response (clinical remission + partial remission) (RR=7.46; 95% CI, 1.78–31.4; p=0.006). Three patients (13.0%) in the FMT group and one patient (5.5%) in the control group died (RR=2.55; 95% CI, 0.24–26.84; p=0.436) ( Figure 2 , Table 2 , and Supplementary Table 3 ). No relapse of GI-GvHD was recorded in any of the patients at this time point.

On day 21, clinical assessment showed that clinical remission was significantly more obvious in the FMT group than in the control group (13 [56.5%] of 23 patients vs. 3 [16%] of 18 patients; RR 9.36; 95% CI, 2.076–42.34; p=0.004), but the clinical response did not differ (16 [69.5%] of 23 patients vs. 9 [50%] of 18 patients; RR=2.83; 95% CI, 0.76–10.52; p=0.120). Three patients died in the FMT group, while two patients died in the control group (3 [13.0%] of 23 patients vs. 2 [11%] of 18 patients; RR=1.35; 95% CI, 0.20–9.02; p=0.57). Two patients in the FMT and control groups each showed GI-GvHD relapse, with no significant differences (2 [8.6%] of 23 patients vs. 2 [11%] of 18 patients; RR=0.857; 95% CI, 0.11–6.72; p=0.883) ( Table 2 and Supplementary Table 3 ). On day 28, the rate of clinical remission and effective response were higher in the FMT group than in the control group ( Supplementary Table 5 ).

Within 90 days of follow-up, no significant difference was observed in EFS between the two groups (HR=1.8; 95% CI, 0.77–4.3; p=0.174) ( Figure 3A ). The FMT group showed better OS (HR=4.2; 95% CI, 1.1–16.0; p=0.031) ( Figure 3B and Supplementary Table 6 ). At the end of the study, the median survival time was >539 days in the FMT group and 107 days in the control group (HR=3.51; 95% CI, 1.21–10.17; p=0.021). Both the EFS (HR=2.3; 95% CI, 0.99–5.4; p=0.08) and OS (HR= 4.4; 95% CI, 1.5–13.04; p=0.008) were higher in the FMT group than in the control group during the follow-up period ( Figures 3C, D and Supplementary Table 7 ).

Only one patient experienced thrombocytopenia after FMT, and one patient developed a cardiac event on the third day after FMT. It cannot be completely excluded that the occurrence of these two events was associated with FMT. No other severe adverse events were observed in the FMT group during 7 days of follow-up after FMT. Other common adverse events included an incomplete ileus in one patient, fever in one patient, vomiting and low fever in two patients, and grade-3 rash in two patients; these patients underwent symptomatic treatment.

Overall, the mortality rate was low in the FMT group (HR=3.97; 95% CI, 1.34–11.75; p=0.013). No significant differences were observed between the two groups in the occurrence of hemorrhagic cystitis (p>0.05), bacterial and fungal infection (p>0.05), CMV and EBV infection (p>0.05), septicemia (p>0.05), TMA (thrombotic microangiopathy) (p>0.05), cardiac events (p>0.05), thrombocytopenia (p>0.05), or epilepsy (p>0.05) ( Table 3 and Supplementary Table 8 ).

Given the severity and emergency of steroid-refractory grade IV GI-GvHD, only 10 patients provided fecal samples at baseline and at week 1 after FMT. Available fecal samples were subjected to microbiota analyses (N=10). Compared with that of the donors, the diversity of fecal microbiota in the fecal samples of the patients was lower before FMT ( Figure 4A ). The relative abundance of Proteobacteria increased, while that of Firmicutes decreased at the phylum level in the microbiota of the patients before FMT ( Figure 4B ). However, there was no significant difference before and after FMT because of the large variations between the patients. After week 1, the composition of microbiota was reconstructed in FMT patients and showed a trend back to normal ( Figures 4C, D ). Bacterial diversity improved at week 1 after FMT in half of the patients (5/10) ( Supplementary Table 4 and Supplementary Figure S1A ). Similar to the results of our prior study, the ratio of Firmicutes to Proteobacteria was restored (7/10), the relative abundance of Proteobacteria decreased (9/10), and the relative abundance of Firmicutes increased (6/10) after FMT ( Supplementary Figure S1B ). The relative abundance of Bacteroidetes increased (7/10) in the fecal microbiota of patients with steroid-refractory GI-GvHD ( Supplementary Table 4 and Supplementary Figure S1C ) after FMT.