Introduction

Crohn’s disease (CD) is an immune-mediated chronic recurrent, systemic disease characterized by gastrointestinal inflammation. Perianal fistula is a common complication of CD, and is estimated to affect up to 26–28% of patients in the first two decades after diagnosis [1, 2]. The treatment of pCD includes conventional medical treatment strategies (e.g., antibiotics and immunomodulators) and anti-tumour necrosis factor agents (anti-TNF), or surgical treatments. However, up to 60–70% of patients relapse after stopping medical treatment, and only a few patients achieve long-term remission [3]. Surgical treatment is associated with problems of poor healing and a high rate of recurrence. In addition, impaired sphincter muscle function following surgery causes postoperative anal incontinence in about 10–35% patients. In some severe cases, fecal diversion may be used to manage CD [4], but recurrence is high with restoration of continuity, and up to 20–40% of these patients will eventually undergo a proctectomy [5]. Therefore, management of refractory pCD and the realization of long-term healing have remained a challenging problem.

In recent years, tremendous progress has been made in the field of regenerative medicine and stem-cell biology. Stem-cell therapy has become a hot research topic. Stem cells are a unique group of undifferentiated cells that have capacity of self-renewal and they can be broadly categorized as embryonic or adult-derived stem cells. Among adult stem cells, the best defined cells are the hematopoietic stem cells (HSCs), MSCs, and intestinal stem cells. Given that MSCs are able to down-regulate immune responses, anti-inflammatory properties, and promote tissue healing [6,7,8,9,10], they are most commonly used to treat pCD. MSCs can achieve the reconstruction of intestinal immunity, to avoid immune-mediated intestinal inflammation, and may achieve long-term healing of pCD, significantly improving the quality of life of patients. Therefore, MSCs may serve as an alternative treatment for refractory pCD. Multiple studies have been conducted to assess efficacy and safety of MSC therapy in pCD, but the outcomes still remain controversial. We performed a systematic review and meta-analysis of the literature to determine the efficacy and safety of MSC therapy.

Materials and methods

We followed the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines [11].

Literature search

To identify relevant studies, a comprehensive electronic search of Pubmed, Cochrane Library, Embase, and CNKI was performed from their inception to October, 2018. Combinations of the following search terms were used: (all fields): (“inflammatory bowel disease” OR “Crohn’s disease” OR “Crohn disease”) and (“mesenchymal stem/stromal cells” OR “stem cell” OR “stromal cell”) and (“perianal fistula” OR “Crohn’s perianal fistula”) and any appropriate abbreviations.

Study selection

Inclusion criteria were: (1) human studies; (2) randomized controlled trials (RCTs), non-randomized experimental studies, and so on; (3) established diagnosis of CD perianal fistula by accepted criteria, with no limits on the age and gender of patients; (4) MSCs were used for treatment of pCD; (5) interventions: local therapy with MSCs for pCD; (6) efficacy and/or safety were reported; and (7) the primary endpoint was clinical efficacy (fistula healing, clinical response) of MSCs for the treatment of pCD, and the secondary endpoint was safety. Exclusion criteria were: (1) non-human studies; (2) use of total-body irradiation or myeloablative regimens; (3) not belonging to pCD (e.g., rectovaginal fistula or enterocutaneous fistulas); (4) and efficacy of MSCs as systemic infusion for CD.

Data extraction and quality assessment

One reviewer (Z.H) extracted data through a standardized data collection form, and then, another reviewer (Z.L) checked the data for accuracy. Any inconsistent results were handled by discussion. These forms included: (1) author; (2) journal; (3) study characteristics (year, location, and type of study); (4) patient demographics (number of patients, age); (5) number of healthy controls (if any); (6) type and source of MSCs; (7) efficacy outcome; (8) dosage and modalities of administration; (9) the use of concomitant therapy with anti-TNF; (10) follow-up time and recurrence; (11) refractory disease (defined as patients who had been unresponsive to or unsuitable for all the previous medical treatment including anti-TNF or unsuccessfully treated by surgery; and (12) adverse events (AEs) (defined as an untoward medical occurrence associated with the use of MSCs whether related to the MSCs or not).

Assessment of the quality of studies was performed using the Cochrane risk of bias tool [12].

Statistical analysis

The meta-analysis was performed using the metaprop command of the meta package in R (version 3.4.2) [13] and RevMan5.3 (Cochrane Collaboration, Oxford, UK) to assess the efficacy (clinical healing and response) and safety (AEs, studies with a comparison arm were included in the meta-analysis). OR and related 95% CIs were calculated to compare MSCs with control cohorts. Log transformation was used to calculate pooled proportions under the fixed and random-effects model [14]. The Chi-square test (P < 0.05 indicating statistical significance) and the I2 statistic (an I2 of ≥ 50%, < 50% indicating substantial and low heterogeneity, respectively) were used to qualitatively and quantitatively evaluate statistical heterogeneity, respectively. Subgroup data were analyzed using the Chi-square test at the subgroup analysis, and P < 0.05 was considered statistically significant.

Results

Literature search

The initial literature retrieval identified 2644 citations. After reading through the title and abstract to exclude the articles, these not meet the inclusion criteria. Of the final 13 studies selected for inclusion in the review [15,16,17,18,19,20,21,22,23,24,25,26,27]. Five studies were RCTs [16, 18, 23, 24, 27], and eight studies were non -randomized experimental studies [15, 17, 19,20,21,22, 25, 26]. The detailed literature screening process is shown in Fig. 1. Eleven studies used adipose-derived MSCs [15, 16, 18,19,20,21,22, 24,25,26,27], and eight studies used autologous MSCs [15,16,17,18, 20,21,22, 26]. Five studies used allogeneic MSCs [19, 23,24,25, 27] and two studies used bone-marrow-derived MSCs [17, 23]. In seven studies, an-TNF had been stopped since enrolment [15,16,17, 19,20,21, 25], and six studies maintained stable dose of anti-TNF during the entire study period [18, 22,23,24, 26, 27]. In four studies, the dosage of MSCs was determined based on the fistula size [20,21,22, 25], which was estimated from the diameter and length of the fistula. Eight studies used the fixed dose of MSCs [15,16,17,18,19, 23, 24, 27]. One study used MSC-coated matrix fistula plugs [26]. Study demographics and clinical characteristics are summarized in Table 1.

Fig. 1
figure 1

Flow diagram of included and excluded studies in this meta-analysis

Full size image
Table 1 MSC therapy for CD patients with perianal fistulas
Full size table

Comparison of MSCs vs control (local therapy)

Five RCTs [16, 18, 23, 24, 27] reported raw data on fistulas healing of patients with MSCs. Using RevMan for meta-analysis, in a fixed-effects model, MSCs group were more effective for fistulas healing than placebo with an OR of 2.05 (95% CI 1.41, 3.00), with no heterogeneity between the studies (I2 = 0%) (Fig. 2). Pelvic MRI is a non-invasive, highly accurate examination for the diagnosis and classification of fistula and, therefore, considered to be the imaging gold standard for CD fistulas [28]. It also facilitates early detection of fistula recurrence. Four RCTs using MRI to evaluate the healing of fistulas were pooled [18, 23, 24, 27], yielding an OR of 1.95 (95% CI 1.33, 2.87) for achieving fistulas healing in patients who received MSCs vs control (Fig. 3), no heterogeneity (I2 = 0%).

Fig. 2
figure 2

Efficacy of MSCs and placebo treatment in pCD

Full size image
Fig. 3
figure 3

MRI + reepithelialization of external opening of fistula in evaluation of fistula healing

Full size image

Efficacy of MSCs for pCD (local therapy)

Thirteen studies reported raw data on clinical healing of patients after MSCs therapy with substantial heterogeneity (I2 = 67%), and the pooled HR was 63% (95% CI 0.53, 0.74) (Fig. 4). Eleven studies [15,16,17, 19,20,21,22,23,24,25, 27] reported raw data on clinical response of patients treated with MSCs. A random-effects pooled rate of clinical response was 30% (95% CI 0.18, 0.48) (Fig. 5). When MRI was used to evaluate the healing of fistulas, the random-effects pooled rate of healing was 56% (95% CI 0.46, 0.69) (8 studies [17,18,19, 23,24,25,26,27]) (Fig. 6).

Fig. 4
figure 4

Forest plot of studies evaluating healing of pCD after local therapy with MSCs

Full size image
Fig. 5
figure 5

Forest plot of studies evaluating clinical response of pCD after local therapy with MSCs

Full size image
Fig. 6
figure 6

Forest plot of studies evaluating healing of pCD after local therapy with MSCs (use of MRI)

Full size image

Subgroup analysis of efficacy of MSCs for pCD (local therapy)

In a fixed-effects model, autologous MSCs [15,16,17,18, 20,21,22, 26] were more effective for the healing of perianal fistulas than allogeneic MSCs [19, 23,24,25, 27] (79% vs 52%), with a significantly statistical difference (P < 0.05) (Fig. 7). Two studies [17, 23] reported raw data on the outcome of patients with bone-marrow-derived MSCs. The pooled rate of clinical healing was 57% compared to 64% for patients who underwent adipose-derived MSCs, with no statistically significant differences (P > 0.05) (data not shown). Six studies [18, 22,23,24, 26, 27] reported raw data on the healing of pCD that maintained stable dose of anti-TNF during the entire study period, the pooled rate of clinical healing was 63% compared to 61% (in 7 studies [15,16,17, 19,20,21, 25]) of patients who had been stopped an-TNF after enrolment, with no statistically significant differences (P > 0.05) (data not shown). In a fixed-effects model, the dosage of MSCs [20,21,22, 25] determined based on the fistula size were more effective for the healing of perianal fistulas than fixed-dose of MSCs [15,16,17,18,19, 23, 24, 27] (80% vs 55%), with a significantly statistical difference (P < 0.05) (Fig. 8).

Fig. 7
figure 7

Clinical healing (allogeneic or autologous MSCs)

Full size image
Fig. 8
figure 8

Clinical healing (different MSC injection modalities)

Full size image

Safety of MSCs for pCD (local therapy)

The included studies demonstrate a good safety and tolerability profile of MSC treatment for pCD. In our study, safety is assessed through AEs analysis: there were no significant increases in AEs (OR = 1.14; 95% CI 0.75, 1.74; P = 0.54) when MSC and non-MSC cohorts were compared (Fig. 9). No MSC-related AE has been reported so far. Common AEs such as anal pain, anal bleeding, fever, abdominal pain, diarrhea, and perianal abscess are mostly transient and rarely result in patients needing hospitalization or in disability. In addition, these AEs are associated with local injection procedures of MSCs (e.g., curettage of a fistula and closure of an internal opening) or the progression of the disease itself rather than the MSCs themselves. Complications often associated with surgery such as fecal incontinence have not been reported in MSCs’ studies. Considering their immunosuppressive properties, it is not clear whether MSCs increase the risks of malignancy at present. Some animal studies suggested that MSCs could promote tumour growth [29, 30], but this has never been confirmed by clinical studies. One patient treated with 1 × 107 MSCs developed an adenocarcinoma of the cecum with peritoneal carcinomatosis > 15 months after the surgical intervention [23]. Baseline and week-12 endoscopy of the rectum revealed no abnormalities. In addition, the last endoscopy of the entire colon and the biopsies taken at that time was completely normal. Moreover, given that MSCs have a very limited lifespan in the body and the patient had a family history of colon cancer, we think that the occurrence of the tumour is associated with genetic factors rather than with MSCs. One patient in the study by de la Portilla et al. [19] was diagnosed with uterine leiomyomata during follow-up. It should be taken into account that uterine fibroids are noncancerous growths of the uterus that often appear during childbearing years and are not associated with an increased risk of uterine cancer In view of these considerations, this study did not reveal any specific signs of carcinogenicity. A follow-up of up to 4 years suggests that MSCs have no carcinogenic risk [18]. However, we should consider that in this study, the number of patients is small, and there is no standard on the evaluation of safety. Therefore, in the future, we need more patients with long-term follow-up, to thoroughly evaluate the safety of MSC treatment for pCD.

Fig. 9
figure 9

Pooled OR of adverse events when MSCs were compared with non-MSCs

Full size image

Discussion

In this meta-analysis, we used different softwares to analyze the efficacy of MSCs for pCD. However, there are several limitations: (1) studies used different types and sources of MSCs including adipose tissue, bone marrow from autologous as well as allogeneic; (2) the timepoint of fistula healing fluctuates significantly (8 weeks–40 months); (3) different definitions of fistula healing, only eight studies used pelvic MRI, and the rest defined healing as reepithelialization of external opening; (4) different dosages and modalities of administration; (5) exclusion of younger patients; and (6) through the Cochrane risk of bias tool, we can see that the eight non-randomized experimental studies included are studies with poor stability and quality.

Now, as stem-cell-based treatments are gaining a strong foothold in fistula management worldwide, more clinical studies are being performed to further ascertain the efficacy and safety of MSC therapy for pCD. Standardization is crucial to assess the efficacy of current and future local treatment strategies. Therefore, with the goal of optimizing this emerging therapy, we need to propose a standardized definition of AEs, and need more objective definitions of the healing of pCD (e.g., endoscopy, C-reactive protein, fecal calprotectin, and imaging including MRI) in future studies.

Several meta-analysis have shown that MSCs are effective and safe for the treatment of pCD [31,32,33,34,35]. However, the results of these studies are still quite different, the explanation for this result may be due to the different sources, different differentiation and regeneration capacity of MSCs and different clinical trials for the isolation of MSCs, different culture programs, and the patient’s age, severity of the disease and other factors may also affect the efficacy of MSCs. In our review, autologous, MSCs have shown more promising results than allogeneic MSCs in pCD fistulas. Moreover, the use of autologous MSCs is more acceptable and there is no ethical conflict. However, the possibility of disease-related effects on autologous MSCs must be taken into account. In addition, autologous MSCs are not immediately available upon request, because isolation and expansion of MSCs to a sufficient numbers of cells require weeks that may result in treatment delay. In addition, due to the inflamed intestinal mucosa, patients’ nutritional status can be poor and patients are frequently very thin, so it is often difficult to gain access to an adequate amount of MSCs. Allogeneic MSCs, easily obtained from a healthy donor, is a product accessible to more patients. Thus, an easily available treatment can be rapidly administered from a completely validated cell bank and provides an affordable therapy to large numbers of patients. In some trials, MSCs from young healthy donors that may have better immunomodulatory and regenerative potential [36, 37] may be selected to avoid the effects of the underlying autoimmune disease. Therefore, our study shows that the efficacy of autologous MSCs is better than that of allogeneic MSCs, but we need to consider the timeliness and cost-effectiveness of treatment to optimize the treatment. There were no statistically significant differences in the HR in our study when the origin of MSCs used (adipose-derived vs bone-marrow-derived) was considered. It should be noted that human adipose tissue is plentiful and easily accessible, and contains a large number of stem cells compared to other tissues in the body. Obtaining adipose tissue minimizes side effects on donors (regardless whether patient or healthy donor). To obtain bone-marrow MSCs in some special donors such as patients with a history of myocardial infarction is dangerous. More and more data are becoming available on the differences between bone-marrow- and adipose tissue-derived MSCs, including on their immunomodulatory properties. One study [37] suggests that adipose tissue-derived MSCs might be superior to bone-marrow MSCs in suppressing immune responses in vitro. Therefore, adipose tissue-derived MSCs can be considered a priority in the future. There were only two studies [17, 23] used bone-marrow-derived MSCs, which was a limitation of our study. More RCTs should compare their efficacy and safety. MSCs were used alongside with anti-TNF in some clinical studies, and they share common targets. Their interaction should be studied carefully. In our review, combined anti-TNF with MSCs had no synergistic effect on the healing of fistulas. Most patients included in our study were refractory to anti-TNF. Treatment of anti-TNF does not seem to affect the long-term efficacy of MSCs for refractory patients, and the long-term application of anti-TNF caused by economic problems and AEs. The exposure of MSCs to physiological concentrations of anti-TNF agents does not affect their survival and their inhibitory capacities on peripheral blood mononuclear cell proliferation [38]. However, whether there is a synergistic effect in combination with MSCs in patients with a response to anti-TNF remains to be determined by further studies.

There are still many unsolved questions concerning the modalities of administration and dosage. In our review, there were two administration methods. One was locally injected MSCs; the fistula tract was thoroughly curetted and irrigated under anesthesia; and the internal opening was closed MSCs which were injected into the submucosa surrounding the internal opening and the fistula track wall. The other method was using MSC-coated matrix fistula plugs. In this study, the volume of injected MSCs proportional to fistula size improved the HR compared with fixed-dose MSCs. CD fistulas usually involve multiple trajectories over a wide range of regions, so treatment of the fistula with a fixed small dose of cells may not be sufficient. We believe that cell number in proportion to fistula size can improve the efficacy, because cells can be injected evenly into the fistula. Therefore, our study provides guidance for the modalities of administration of MSCs. Due to the limitations of studies, it is difficult to provide recommendations on the optimal dose. However, Molendijk et al. [23] reported that local therapy with 3 × 107 MSCs showed superior fistulas healing compared with 9 × 107 MSCs treatment (80% vs 20%. The authors argued that higher cell concentrations could result in a lower survival rate and/or cell function and that a larger number of cells could be immunogenic resulting in increased clearance or deactivation of the cells. This provides a guidance in choosing the optimal dosage of MSCs for the treatment of pCD.

Studies included in our meta-analysis used variable definitions of fistulas healing. We showed that when pelvic MRI was used to evaluate fistula healing, the HR fell. We think that healed fistula was defined as complete closure evident on clinical examination and reepithelialization of the external opening of the fistula was not objective and accurate, as it cannot replace the inside of the entire fistula. Pelvic MRI is considered be the imaging gold standard for CD fistula, but MRI is insensitive to reduced fistula caliber. Therefore, we need improved instruments for evaluation of perianal CD fistula healing. Parks’ classification of the anatomy of perianal fistulas: (1) intersphincteric; (2) transsphincteric; (3) suprasphincteric; (4) extrasphincteric; and (5) not involving the sphincter complex is a clinically useful classification of perianal fistulas in CD and can help the treating physician to determine optimal management strategy disease. However, due to the limitations of our study, differences in classification of pCD were not analyzed in detail and compared with the differences in response to MSC treatment. RCTs are needed to clarify the mechanism of MSCs for treating CD fistulas by directly comparing the therapeutic effects between MSCs and immunomodulators, biologics, and surgical procedures. Comprehensive consideration of factors that affect the healing of fistulas (such as gender, age, diameter/length of fistula, type of fistula, CD duration, CDAI activity, and other factors), and the selection of appropriate MSCs (type, source, dose, frequency of injection, and mode of delivery) for patients receive the optimal treatment must be further investigated.

Assessment of study quality

There was overall high risk of performance bias, detection bias, and reporting bias in the studies included in the review (Fig. 10a, b).

Fig. 10
figure 10

a Overall quality of the included studies assessed by Cochrane risk of bias assessment tool. b Risk of bias within studies assessed by Cochrane risk of bias assessment tool

Full size image

Conclusions

The results of this study are encouraging, and MSCs appear to offer a safe and effective treatment approach for pCD fistulas. Based on subgroup analysis, autologous, adipose MSCs had relatively higher rates of clinical healing. MSC dosage in proportion to fistula size also had higher rates of clinical healing. The efficacy of MSCs does not seem to be influenced by prior unsuccessful treatment with anti-TNF and may be an attractive option for patients failing this treatment. However, these findings should be interpreted with much caution owing to the limited number of studies and potential biases, More RCTs are needed to support our conclusions.