Study selection

From the initial literature search, we retrieved relevant 3948 records from PubMed, Web of Science, Scopus, and the Cochrane Library. After the title and abstract screening of them we screened the full text of 295 articles. Only nine studies met our criteria31,32,33,34,35,36,37,38,39. Figure1 shows the PRISMA flow diagram of our search and selection process.

The nine studies were RCTs and enrolled a total of 422 multiple sclerosis patients. All studies were parallel in design except 4 studies were cross-over RCTs31,32,34,39. These cross-over trials were reviewed up to the point of cross-over. All studies infused stem cells intravenously except Petrou et al. that included an additional intrathecal SCT subgroup32. This study showed that intrathecal SCT was more effective than intravenous SCT, but we pooled the data of both routes as single study data. Of the included studies, only two studies used autologous hematopoietic SCT (AHSCT) in addition to immune ablative regimen prior to the transplantation37,38. Burt et al. compared SCT to DMTs (natalizumab, fingolimod, and dimethyl fumarate) in RRMS patients37, and Mancardi et al., compared SCT to mitoxantrone in relapsing and progressive MS patients38.

Supplementary Table S1 summarizes the characteristics of the included trials, Table 1 shows the demographic and baseline characteristics of these studies population, and Supplementary Table S3 shows efficacy endpoints reported at 6months.

We assessed seven domains in each study according to The Cochrane Collaborations tool for assessing risk of bias 1. The 9 studies were randomized but 4 studies32,35,38,39 didnt clarify the methods of random sequence generation. 6 RCTs confirmed concealment of patients allocation to the intervention31,32,33,34,36,37. Blinding of the outcome assessors was clearly stated in all studies except Nabavi et al.39 but blinding of participants and personnel wasnt fulfilled in three studies35,37,38. The reasons for incomplete outcome data are related to the treatment in Uccelli et al.31 and the reasons werent clearly described in Burt et al37. Two studies reported the outcomes in an incomplete way that limited their inclusion in the meta-analysis inducing a reporting bias33,38. The overall quality of the studies was good for 2 studies32,36, fair for 3 studies31,34,37, and poor for 4 studies33,35,38,39. Figure2 shows the risk of bias summary and graph.

Risk of bias assessment: (a) Risk of bias summary, (b) Risk of bias graph.

After analyzing the efficacy and safety outcomes for all studies collectively, we subdivided the results into studies that used immunosuppression before AHSCT37,38 and studies that transplanted mesenchymal stem cells (MSCs) without immunosuppression31,32,33,34,35,36,39 to minimize the procedural variations among the included trials.

The majority of the studies31,32,33,34,35,36,37,39 reported EDSS change for 211 patients in stem cell transplantation (SCT) arm and 176 controls. Because the time of reporting this outcome varied among the studies, we analyzed EDSS change at the last follow-up reported by each study. Our analysis showed nonsignificant difference between SCT group and the control group (MD=0.48, 95% CI [1.11, 0.14], p=0.13). There was great heterogeneity between studies (2=116.74, df=7, p<0.00001, I2=94%), so we pooled the data under the random-effects model (Table 2 and Supplementary Figure S1).

The subgroup analysis of the studies that used MSCs without immunosuppression also showed nonsignificant improvement (MD=0.3, 95% CI [0.87, 0.27], p=0.3). However, Burt et al37. that used immunosuppression before AHSCT revealed significant EDSS reduction (Supplementary Figure S1).

The results remained nonsignificant after the leave-one-out sensitivity analysis (Supplementary Figure L1).

The heterogeneity within the studies was not significant (2=1.61, df=1, p=0.2, I2=38%), and we adopted a random effect model. The reduction of EDSS in SCT group was significantly greater than the control group (MD=0.57, 95% CI [1.08, 0.06], p=0.03) (Table 2 and Fig.3a).

Forest plot of EDSS change from baseline at (a) 2months, (b) 6months, (c) 12months.

Adopting the random-effects model, the heterogeneity between the studies was significant (2=65.27, df=6, p<0.00001, I2=91%), and SCT showed nonsignificant improvement of EDSS compared to the control (MD=0.48, 95% CI [0.98, 0.03], p=0.07) (Table 2 and Fig.3b). MSCs without immunosuppression also resulted in nonsignificant EDSS reduction at 6months (MD=0.33, 95% CI [0.78, 0.11], p=0.14) (Fig.3b).

The effect estimate changed to (MD=0.62,95% CI [1.14, 0.09], p=0.02) favoring SCT over the control after excluding Nabavi et al.39 from the analysis (Supplementary Figure L2 and Table L1).

We adopted the random-effects model because heterogeneity was significant, and the difference between the SCT group and the control was not significant at 12months for both collective studies analysis and studies used MSCs without immunosuppression (p=0.06 and p=0.5, respectively). However, the study that used AHSCT plus immunosuppression37 showed significant improvement in patients disability (p<0.00001) (Table 2 and Fig.3c).

After performing a sensitivity analysis by excluding Fernandez et al.36, the results changed from nonsignificant to significant improvement in SCT arm (MD=1.69, 95% CI [1.94, 1.44], p<0.00001) (Supplementary Figure L3 and Table L1).

We compared the effect of SCT on patients disability depending on baseline EDSS. Six studies31,32,33,34,37,39 included 334 MS patients with baseline EDSS6.5, while two studies35,36 included 53 patients with baseline EDSS>6.5. Using a random effects model, both subgroups showed significant heterogeneity (p<0.00001 and p<0.00001). Both subgroups revealed nonsignificant effect of SCT on EDSS, (MD=0.41, 95% CI [1.11, 0.29], p=0.25) for baseline EDSS6.5 subgroup and (MD=0.68, 95%CI [2.68, 1.32], p=0.5) for baseline EDSS>6.5 subgroup (Table 2 and Supplementary Figure S2).

We pooled data of EDSS change from baseline to the last assessment time under a random-effects model, and the differences were nonsignificant for both low and high doses subgroups, (MD=0.31, 95% CI [1, 0.38], p=0.37) and (MD=0.57, 95% CI [1.94, 0.8], p=0.41), respectively. The studies of both subgroups showed significant heterogeneity (I2=95%, p<0.00001) for the low doses subgroup, and (I2=89%, p=0.0001) for the high doses subgroup (Table 2 and Supplementary Figure S3).

Adopting a random-effects model, stem cells from embryonic as well as adult origin showed nonsignificant effect on EDSS (p=0.17, and p=0.37, respectively), With significant heterogeneity among the studies (I2=88%, p=0.004), and (I2=94%, p<0.00001), respectively (Table 2 and Supplementary Figure S4).

We pooled data using a random-effects model. Five studies31,32,33,36,39, in which placebo was the control, showed substantialheterogeneity (I2=63%, p=0.03) and the difference between SCT and placebo was not significant (MD=0.09, 95% CI [0.46, 0.28], p=0.62). Three studies34,35,37, in which the control was active treatment, showed significant reduction of EDSS with SCT compared to the active drugs (MD=1.21, 95% CI [1.98, 0.43], p=0.002) and the heterogeneity was significant (I2=88%, p=0.0002) (Table 2 and Supplementary Figure S5).

Only two studies32,34 reported the number of relapses in the 6months following the intervention. Under a random-effects model, the heterogeneity was moderate (p=0.14, I2=53%), and the decrease in relapses number was nonsignificant (p=0.23) (Supplementary Figure S6).

Four studies31,32,34,37 assessed T25-FW in 154 and 136 patients in the SCT and control groups, respectively. We pooled data under a random-effect model, and heterogeneity was moderate (2=5.99, df=3, p=0.11, I2=50%). SCT resulted in a nonsignificant improvement in patients T25-FW scores compared to the control group (MD=0.69, 95% CI [1.93, 0.56], p=0.28), as shown in Fig.4.

Forest plot of T25-FW change from baseline.

In the studies that included mesenchymal SCT without immunosuppression, the improvement in patients T25-FW scores after SCT was not significant (MD=0.39, 95% CI [0.84, 0.06], p=0.09), but T25-FW significantly improved in the study that used AHSCT and immunosuppression37 (p=0.006). Figure4 demonstrates these analyses. The p value of the results didnt change after the one-study-removed sensitivity analysis (Supplementary Figure L4).

9-HPT was evaluated in four RCTs31,32,34,37. We used a random-effects model because heterogeneity was significant (p=0.0003, I2=84%). 9-HPT showed nonsignificant improvement in the collective analysis and the sub-analysis of MSCs without immunosuppression. However, Burt et al.37. revealed a significant improvement (p<0.00001) (Supplementary Figure S7). The results remained nonsignificant after sensitivity analysis (Supplementary Figure L5).

We pooled PASAT-3 scores assessed at the end of treatment in four trials under a random-effects model31,34,36,37. Heterogeneity was minimal (p=0.35, I2=9%), and the differences were nonsignificant in the collective analysis and the sub-analysis of autologous and mesenchymal SCT (p=0.35, p=0.96, and p=0.31, respectively) (Supplementary Figure S8). Effect estimate remained nonsignificant after one-study-removed sensitivity analysis (Supplementary Figure L6).

We analyzed the change in brain lesion volume from baseline to the end of the follow-up. Data were pooled under a random-effects model, heterogeneity was absent (p=0.38, I2=0%). Our analysis revealed a significant reduction in T2 lesions volume (MD=7.05, 95% CI [10.69, 3.4], p=0.0002). In the studies that used MSCs without immunosuppression, the reduction of brain lesions volume was nonsignificant (p=0.1) (Fig.5a).

Forest plot of radiological outcomes change from baseline (a) MRI T2-weighted lesions volume at the end of treatment, (b) MRI T2-weighted lesions number at 12months, (c) number of GELs at the end of treatment. *the study used immunosuppression before AHSCT.

The results became nonsignificant and changed to (MD=4.41, 95% CI [9.66, 0.85], p=0.1) after a sensitivity analysis performed by excluding Burt et al.37 (Supplementary Figure L7 and Table L1).

Adopting a random-effects model, the studies showed substantial heterogeneity (p=0.07, I2=70%). And the differences between SCT and the control after 12months were nonsignificant (p=0.99) (Fig.5b).

Five studies31,32,33,34,36 assessed this outcome. Four studies reported the change of GELs number from baseline at 6months except Fernandez et al.36 at 12months. We pooled data under a random-effects model and heterogeneity was not significant (2=7.81, df=4, p=0.1, I2=49%). Our analysis revealed nonsignificant differences in GELs number change (p=0.83) (Fig.5c). The results didnt change after sensitivity analysis (Supplementary Figure L8).

Seven studies31,32,33,34,36,37,38 reported adverse events that occurred during the follow-up period. Two studies35,39 didnt provide data about AEs. Nabavi et al. mentioned only pain at the site of bone marrow aspiration39. Our analysis revealed that the difference was nonsignificant between SCT and the control group regarding the incidence of most AEs. Administration-related AEs, including infusion site swelling, hematoma, and pain, were significantly more common in the SCT group compared to the control (N=25, RR=2.55, 95% CI [1.08, 6.03], p=0.034). On the other hand, the SCT group had a lower incidence of total infections (any infection during the follow-up period, including viral infections, respiratory, urinary infections, scabies, and other infestations) than the control group (N=60, RR=0.58, 95% CI [0.37, 0.9], p=0.02). Regarding the use of immunosuppression, AHSCT combined with immunosuppression was significantly associated with a higher incidence of blood and lymphatic system disorders (N=16, RR=2.33, 95% CI [1.23, 4.39], p=0.009). The analyses of the adverse events are shown in Table 3 and Supplementary Figures S9S14. No transplant-related mortality was noted in all trials during the follow-up period, except for two unrelated deaths compacted by Fernandez et al. in the placebo arm (one due to choking while feeding and the other due to respiratory infection)36.

We examined the publication bias among the studies that reported the effect of SCT on patients disability using the funnel plot test. Although there was funnel plot asymmetry, the test isnt reliable because the included studies were less than ten studies24 (Supplementary Figure S15).

Read the original post:
Efficacy and safety of stem cell transplantation for multiple sclerosis: a systematic review and meta-analysis of ... - Nature.com