We identified 364 references by electronic and manual search. Of these, 92 full-text studies were evaluated according to the eligibility criteria. Finally, 51 studies were included, 16 related to the administration of allogeneic MSC for the prophylaxis of GvHD and 35 for the treatment of aGvHD and/or cGVHD (Fig. 1).
Mesenchymal stromal cells for the prevention of GvHD in patients undergoing HSCT
Characteristics of the included prophylaxis-related studies are reported in Table S2 (see Additional file, pp. 31–32). These studies comprised 654 patients, 298 patients who received only HSCT (control group) and 356 patients infused with MSC (MSC group) for GvHD prevention. Included studies comprised data from children (< 18 years old, n = 8 studies), adults (≥ 18 years old, n = 4 studies) and from both adults and children (n = 4 studies). Five of the included reports did not include the control group. Studies varied highly in terms of MSC doses (0.03–10.12 × 106 MSC/kg). The MSC sources were bone marrow (BM, n = 12 studies) and umbilical cord (UC, n = 4 studies), the majority of them from a different donor than HSCT. For most studies, a single dose of MSC was administered on the same day than HSCT while in 4 studies, a second dose of MSC was also administered on day + 2 [25], + 14 [26, 27] or + 21 [28] after HSCT. In two studies, the single administration of MSC was carried out after HSCT, on day + 28 (19–54) [29] or after more than 4 months after transplantation [30]. Any of the studies included to evaluate the effect of MSC infusion for the prophylaxis of GvHD showed low risk for the six domains of RoBANS (Figure S1, see Additional file, pp. 4). All the studies had not blinded the outcome, except in the double-blinded study reported by Gao et al. [30], which presented an unclear risk for detection bias since the blinding of outcome assessors was not stated. Nine studies showed low risk in all other RoBANS criteria.
At last follow-up, 73% (95% CI, 0.67–0.79, I2 = 42%) and 59% (95% CI, 0.52–0.66, I2 = 35%) of the patients from the MSC (Figure S2A) and control (Figure S2B) groups were alive, respectively (see Additional file, pp. 5). OS did not differ between children and adults for the control (P = 0.28) and MSC (P = 0.33) groups. Funnel plots show OS distribution for individual studies from the control (Figure S3A) and MSC (Figure S3B) groups (see Additional file, pp. 6). Egger test showed no evidence of publication bias for the MSC (P = 0.259) and control (P = 0.100) groups. However, Egger test showed some degree of publication bias on subgroup analyses. It was not possible to test the publication bias for the subgroup adults in the MSC group due to the low number of studies. The dichotomous analysis of 11 studies containing data related to OS from both control (n = 298 patients) and MSC (n = 213 patients) groups showed that the administration of MSC and HSCT was associated with a 17% increased OS (95% CI, 1.02–1.33, I2 = 0%, Fig. 2a).
Data related to the incidence of aGvHD in the control (n = 235 patients) and MSC (n = 150 patients) groups were collected from 10 studies. Patients from the MSC group tended to suffer less aGvHD compared with the control group (RR = 0.84; 95% CI, 0.66–1.07, I2 = 0%, Figure S4, pp. 7). The incidence of severe aGvHD (grades III–IV) was not significantly different between treatments (RR = 0.53; 95% CI, 0.20–1.42, I2 = 0%; Figure S5, pp. 8). Although most studies report together grades III and IV, the analysis of four studies comprising 144 patients showed that the administration of MSC was associated with a lower incidence of grade IV aGvHD, compared with the control group (RR = 0.22; 95% CI, 0.06–0.81, Fig. 2b). In addition, the analysis of nine studies (MSC, n = 148 patients; control, n = 236 patients) showed that the infusion of MSC was associated with a reduced cGvHD incidence (RR = 0.64; 95% CI, 0.47–0.88, I2 = 0%, Fig. 3a) and a trend to a lower incidence of extensive cGvHD (RR = 0.50; 95% CI, 0.25–1.01, P = 0.05, Fig. 3b). The risk to suffer acute and chronic GvHD did not differ between adults and children. For the MSC group, OS and GvHD incidence were not correlated with MSC dose (Figure S6, pp. 9).
Mesenchymal stromal cells for the treatment of steroid-refractory GvHD
Next, we analyzed the outcome of patients who were infused with MSC for the treatment of aGvHD (n = 943 patients) or cGvHD (n = 76 patients). The characteristics of the 35 studies reporting data for the treatment of aGvHD and/or cGvHD are summarized in Table S3 (see Additional file, pp. 33–36). Only six of these studies reported data of interest for control patients (aGvHD, n = 182 patients; cGvHD, n = 14 patients; Table S4, pp. 37). Included studies reported data from children (n = 10 studies), adults (n = 22 studies), and both children and adults (n = 8 studies). Studies varied greatly in terms of MSC dose (aGvHD, 0.22–6.81; cGvHD, 0.6–2.28 × 106 MSC/kg), dose number (aGvHD, 1–10; cGvHD, 1–11), and time for first infusion after HSCT (aGvHD, + 2.5 to + 124.7 days; cGvHD, + 64.1 days to + 45.1 months). The sources of the infused MSC were BM (94.90%), adipose tissue (AD, 2.94%), or BM and AD (2.16%), most of them from a different donor than the HSCT. Table S5 shows the number of patients with aGvHD for individual studies which report the grade and organs implicated (see Additional file, pp. 38–40). Most of the included patients suffered grade III–IV (82.99%). Since Dalowski et al. [31] included data from von Bonin et al. [32], data from this last study were included only when it considered subgroups or outcomes that were not reported by Dalowski et al. Only one of the included studies to analyze the effect of the administration of MSC for the treatment of GvHD showed low risk for the six domains of RoBANS. Eighteen studies showed low risk for five RoBANS criteria (Figure S7, pp. 10).
The analysis of 26 studies showed that 50% (95% CI, 0.41–0.59, I2 = 88%) of aGvHD patients treated with MSC (n = 878 patients) were alive at last follow-up (Fig. 4a). On the other hand, the analysis of five studies (n = 182 patients) showed that only 25% (95% CI, 0.11–0.39, I2 = 81%) of the patients from the control group were alive at the last follow-up (Figure S8A, pp. 11). The OS for the different treatment groups was significantly different (P = 0.0005). The analysis of five studies containing data for the OS of both control and MSC groups showed a trend to a higher OS for patients infused with MSC (RR = 1.33; 95% CI, 0.84–2.10, I2 = 51%, Figure S8B, pp. 11). Regarding the MSC group, the OS did not differ between children and adults (P = 0.29), with respect to aGvHD grade (P = 0.73, Figure S9), the number of organs involved (P = 0.50, Figure S10), or organ-specific involvement (P = 0.37, Figure S11), although patients with skin implication tended to survive more (0.64; 95% CI, 0.52–0.77) compared to those with gut (0.51; 95% CI, 0.35–0.66) and liver (0.57; 95% CI, 0.41–0.72) involvement (see Additional file, pp. 12–14). In the case of multiorgan implication, OS did not differ between subgroups (Figure S12, pp. 15). The analysis of data related to OS according to the grade and organs involved are summarized in Table S6 (see Additional file, pp. 41). Interestingly, the OS of patients with aGvHD infused with MSC was positively correlated with the MSC dose (P = 0.0214, Fig. 4b) and did not differ in relation to the first day of infusion of MSC from HSCT (P = 0.1034, Figure S13, pp. 16). Twenty-two percent (95% CI, 0.16–0.29, I2 = 44%) of the patients infused with MSC for the treatment of aGvHD developed cGvHD, without differences between children and adults (P = 0.62, Fig. 5).
For patients with aGvHD, 67% of 887 patients (95% CI, 0.61–0.74, I2 = 74%) responded to the MSC infusion (Figure S14A) and 39% (95% CI, 0.31–0.48, I2 = 81%) achieved CR (Figure S14B, pp. 17). The proportion of responders did not differ between children and adults nor with respect to the aGvHD grade (P = 0.17, Figure S15, pp. 18–19) or number of organs affected (OR, P = 0.14, Figure S16; CR, P = 0.89, Figure S17, pp. 20–21). However, the proportion of complete responders tended to be lower for patients with grade III–IV compared to grade II (RR, 0.76; 95%CI, 0.54–1.05; I2 = 0%, Figure S18, pp. 22). Regarding to organ-specific response, the proportion of patients that achieved OR (Figure S19A) and CR (Figure S19B) was higher in skin compared with liver and gut implication (see Additional file, pp. 23). In addition, the OR was significantly reduced in children (0.61; 95% CI, 0.51–0.71) compared with adults (0.76; 95% CI, 0.68–0.85) when the gut was affected (P = 0.03, Figure S20, pp. 24). In the case of multiorgan involvement, the proportion of patients that achieved OR and CR did not differ between subgroups. Subgroup analysis for OR and CR are summarized in Table S7 and S8, respectively (see Additional file, pp. 42–43). No significant correlation was found between the proportion of responder patients and day of MSC infusion after HSCT (Figure S21A) or MSC dose (Figure S21B, pp. 25).
Data related to the effect of the MSC infusion in the outcome of patients with cGvHD were collected from 10 studies (n = 75 patients). At the longest follow-up, 64% (95% CI, 0.47–0.80, I2 = 28%) of chronic patients were alive (Fig. 6a). Sixty-six percent (95% CI, 0.55–0.76, I2 = 0%) and 23% (95% CI, 0.12–0.34, I2 = 0%) achieved OR (Fig. 6b) and CR (Fig. 6c), respectively. A low number of studies reported data about both aGvHD and cGvHD. Data from five and seven studies showed no differences in OS (P = 0.74, Figure S22A) and response rates between patients with aGvHD and cGvHD infused with MSC, although patients that suffered cGvHD were less likely to achieve OR (0.77; 95% CI, 0.56–1.06, Figure S22B) and CR (0.66; 95% CI, 0.32–1.36, Figure S22C) than aGvHD patients (see Additional file, pp. 26–27).