Allogeneic stem-cell transplantation for multiple myeloma: a systematic review and meta-analysis from 2007 to 2017

Background Despite recent advances, multiple myeloma (MM) remains incurable. However, the appearance of allogeneic stem cell transplantation (allo-SCT) through graft-versus-myeloma effect provides a potential way to cure MM to some degree. This systematic review aimed to evaluate the outcome of patients receiving allo-SCT and identified a series of prognostic factors that may affect the outcome of allo-SCT. Patients/methods We systematically searched PubMed, Embase, and the Cochrane Library from 2007.01.01 to 2017.05.03 using the keywords ‘allogeneic’ and ‘myeloma’. Results A total of 61 clinical trials involving 8698 adult patients were included. The pooled estimates (95% CI) for overall survival (OS) at 1, 2, 3 and 5 years were 70 (95% CI 56–84%), 62 (95% CI 53–71%), 52 (95% CI 44–61%), and 46 (95% CI 40–52%), respectively; for progression-free survival were 51 (95% CI 38–64%), 40 (95% CI 32–48%), 34 (95% CI 27–41%), and 27 (95% CI 23–31%), respectively; and for treatment-related mortality (TRM) were 18 (95% CI 14–21%), 21 (95% CI 17–25%), 20 (95% CI 13–26%), and 27 (95% CI 21–33%), respectively. Additionally, the pooled 100-day TRM was 12 (95% CI 5–18%). The incidences of grades II–IV acute graft-versus-host disease (GVHD) and chronic GVHD were 34 (95% CI 30–37%) and 51 (95% CI 46–56%), respectively. The incidences of relapse rate (RR) and death rate were 50 (95% CI 45–55%) and 51 (95% CI 45–57%), respectively. Importantly, disease progression was the most major cause of death (48%), followed by TRM (44%). The results failed to show an apparent benefit of allo-SCT for standard risk patients, compared with tandem auto-SCT. In contrast, all 14 trials in our study showed that patients with high cytogenetic risk after allo-SCT had similar OS and PFS compared to those with standard risk, suggesting that allo-SCT may overcome the adverse prognosis of high cytogenetic risk. Conclusion Due to the lack of consistent survival benefit, allo-SCT should not be considered as a standard of care for newly diagnosed and relapsed standard-risk MM patients. However, for patients with high-risk MM who have a poor long-term prognosis, allo-SCT may be a strong consideration in their initial course of therapy or in first relapse after chemotherapy, when the risk of disease progression may outweigh the transplant-related risks. A large number of prospective randomized controlled trials were needed to prove the benefits of these therapeutic options. Electronic supplementary material The online version of this article (10.1186/s12935-018-0553-8) contains supplementary material, which is available to authorized users.


Background
Multiple myeloma (MM) is an incurable clonal plasma cell hematologic malignancy. It is usually a disease of the elderly and its' median age at diagnosis is 65 years [1]. Modern therapy for MM includes corticosteroids, immune-modulatory drugs (IMiDs) (thalidomide, lenalidomide, pomalidomide) [2], proteasome inhibitors (bortezomib, carfilzomib) [3], compounds targeting specific molecules and monoclonal antibodies. Besides, autologous stem cell transplantation (auto-SCT) in combination with high-dose chemotherapy could be considered as a frontline strategy for younger MM patients [4]. Although these therapies dramatically increased patients' response rate and survival rate [5], most patients could not maintain a sustained complete remission and relapse ultimately. Allogeneic SCT (allo-SCT) through graftversus-myeloma (GvM) effect was gradually emerging as a potential way to cure MM [6]. However, allo-SCT couldn't be widely used to treat MM patients [7] due to unusable donors, high risk of treatment related mortality (TRM) and the occurrence of GVHD during several decades. Major transplant innovations and new technologies emerged in 2000. The European Group for Blood and Marrow Transplantation (EBMT) registry reported that the usage rates of PBSC grew from 66 to 91%, while RIC/ NMA from 37 to 75% during the time of 1998-1999 to 2002-2003 [8]. Moreover, since 2000, novel drugs have been applied in different phases of allo-SCT, such as conditioning regimen and post-transplantation therapy, the introduction of which brought additional clinical benefits, improved responses rates and made the transplant safer. Additionally, with the increasing awareness of the risk stratification of MM and the improvement of detection techniques, it is now possible to identify highrisk MM patients more quickly and accurately [9]. High cytogenetic risk, as a poor prognostic factor, may encourage these patients receiving transplants earlier. Indeed, Nivison-Smith et al. reported the shorter length of time between diagnosis and transplant was a prognostic factor index for both improved OS and PFS [10]. Furthermore, the occurrence rate of TRM and GVHD has been reduced through advanced maintenance strategies, better supportive care, more suitable patient selection and strategies for GVHD prophylaxis. Therefore, allo-SCT maybe well tolerated and become an effective way to cure MM in the future. Till now, the indications which guided clinicians to apply allo-SCT to clinical practice are mainly derived from large registry and single-center retrospective studies. Thus we performed a systematic literature review and meta-analysis to evaluate the efficacy and safety of allo-SCT for MM during the last 10 years.

Search strategy
We searched MEDLINE, Embase, and the Cochrane Library database. We used 'allogeneic AND myeloma' for Medline and the Cochrane Library search, and the following terms for Embase search: 'allogeneic' .mp. AND 'myeloma' .mp. Adults, humans and English language articles were limited to our search. We conducted our search from 2007.01.01 to the 2017.05.03.
Exclusion criteria: (1) the use of cord blood as the stem cell source, (2) the inclusion of patients with various hematological malignancies without a separate description of the results of MM patients, (3) the use of a wide variety of transplant strategies rather than one clearly defined strategy that was similar for each patient, (4) lack of outcome data.

Study selection and data extraction
We only considered full-text articles. The titles and abstracts of the remaining articles were screened by the following exclusion criteria: reviews, meta-analyses, editorials, conference proceedings, no primary or secondary endpoints reported and commentaries. Study selection, quality assessment and data extraction were conducted by two reviewers independently using standardized forms. If there were disagreements, third investigator would adjudicate.

Study quality assessment
We followed 5 items to evaluate study quality: (1) conditioning regimens, (2) stem cell source, (3) donor, (4) GvHD prophylaxis regimen, and (5) disease status before allo-SCT. When articles provided one corresponding item, 1 was given to the study or otherwise 0. Only studies received 5 scores were deem as good quality, 4 scores were moderate quality and 3 scores were low quality.

Heterogeneity analyses and subgroup analyses
Study heterogeneity was assessed with the Cochran Q test, and the I 2 statistic was used to quantify it. If the p value of Cochran's Q test was < 0.1 and I 2 statistic was > 50%, it indicated the substantial heterogeneity was existent. We used a random-effect model to pool the data. If we found substantial heterogeneity, subgroup analyses were performed to explore the causes. Specifically, the subgroups transplantation period 1990s versus 2000s were considered. We identified the following prognostic factors that may affect the outcome of allo-SCT: cytogenetic risk (high-risk versus non-high-risk), remission status at the time of transplantation and post transplantation (CR versus non-CR), source of the transplanted stem cell (peripheral blood stem cell versus bone marrow).

Sensitivity analyses
Sensitivity analyses were performed to test the robustness of the results.

Publication bias
Publication bias was assessed using funnel plots, begg and Egger test. p values < 0.05 were considered statistically significant. Analyses were performed with Review Manager (version 5.1) and Stata SE/MP 11.0.

Result
Our initial search yielded 3144 articles. After screening, 393 duplicating studies were removed, and 2654 studies were excluded based on titles and abstracts. A further 36 studies were excluded for not fulfilling the inclusion criteria. Finally, a total of 61 citations with 8698 eligible patients were included in the meta-analysis. The sample size varied from 7 to 1667. Follow-up period ranged from 1 to 217.2 months. Patients' age ranged between 21 and 77 years. According to quality assessment scores, 29 studies scored 5, 25 scored 4, and 7 scored 3.
During the past decade, myeloablative conditioning (MA) has been largely abandoned due to the high treatment related mortality (TRM). However, reducedintensity conditioning (RIC) or non-myeloablative conditioning (NMA) regimens have high risk of relapse. To estimate whether the widespread adoption of RIC/NMA regimens could bring about better OS, we compared patients receiving allo-SCT with RIC/NMA regimen with those receiving MA regimens in OS. No evidence was found that RIC/NMA regimens improved OS compared to MA regimens (RR = 0.88, 95% CI 0.74-1.05) [10,13,25,29]. These results suggested that different conditioning regimens didn't affect OS.
High cytogenetic risk has been previously reported to be negative factor for OS and PFS. MM patients with high cytogenetic risk resisted to conventional chemotherapy, relapsed repeatedly after autologous stem cell transplantation (auto-SCT) and had a grim prognosis [53,54]. In order to solve this problem, many new treatment strategies have been developed over the past few decades, including novel agents (bortezomib and lenalidomide) and double auto-SCT [55,56]. Although overall response rates and OS have been increased, most patients still relapsed soon. All 14 trials involved in our study showed that high cytogenetic risk patients after allo-SCT had similar OS and PFS to those with standardrisk (RR = 0.83, 95% CI 0.67-1.03) [12, 14, 15, 21, 25, 31, 39, 42-44, 47, 50, 51, 57, 58]. However, most trials described the situation using descriptive language, only 5 provided concrete data of comparisons of OS between high cytogenetic risk patients and standard-risk patients [15,25,42,44,57]. A meta-analysis of OS provided by these 5 trials showed no statistical difference between high cytogenetic risk patients and standard-risk patients. No substantial heterogeneity was found among the included studies (p = 0.599, I 2 = 0). These findings indicate allo-SCT overcomes the adverse prognosis of high cytogenetic risk.
EBMT centers reported that PBSCs have replaced bone marrow (BM) and then become the primary source of grafts since 2000 [8]. In every trial of the present metaanalysis, most allo-SCT were performed with PBSC as source of graft, while a few allo-SCT with BM. Only 4 studies provided the comparisons of OS between PBSC and BM, and the results indicated that use of PBSC had faster engraftment kinetics and quicker immune reconstitution than BM, but these advantages didn't translate into higher OS (HR = 1.02, 95% CI 0.53-1.96) [31,50,59,60].
No evidence was found that genetic risk stratification (RR = 0. 89

Subgroup analyses and sensitivity analyses
Since major transplant innovation and new technologies happened in the year 2000, we took the year 2000 as a cut-off point of the subgroup analysis. We took the first year of transplantation period as a standard of 1990s and 2000s cut-off point, and found that substantial heterogeneity of 100-day TRM, 1-year TRM, 2-year TRM, 3-year TRM, and 3-year PFS mainly came from 1990s transplantation period (I 2 = 97.5%, p = 0 for 1990s, I 2 = 26.2%, p = 0.255 for 2000s; I 2 = 93.6%, p = 0 for 1990s, I 2 = 56.4%, p = 0.0031 for 2000s; I 2 = 70.1%, p = 0.005 for 1990s, I 2 = 13.2%, p = 0.327 for 2000s; I 2 = 96.6%, p = 0 for 1990s, I 2 = 0%, p = 0.552 for 2000s; I 2 = 97.4%, p = 0 for 1990s, I 2 = 62%, p = 0.007 for 2000s; respectively). This phenomenon most probably due to 1990s subgroup including a few patients performed allo-SCT after the year 2000. The articles involved didn't provide the individual data of those patients, so we failed to separate out these small amount of patient from 1990s subgroup. Sensitivity analyses were performed by excluding a study at a time in turn, and pooling the outcomes of the remaining studies. No material changes happened in all results.

Publication bias
We failed to identify obvious asymmetry with the exception of 5-year PFS, 5-year OS and exGVHD in all the funnel plots through visual inspection. In line with the funnel plots, Egger test find substantial publication bias' evidence in 5-year PFS, 5-year OS and exGVHD (p = 0.046, p = 0.036, p = 0.013, respectively).

Discussion
Since the first allo-SCT was performed by Donnall Thomas in 1957 [75], it has been chosen to be a salvage regimen for relapsed or refractory MM patients for several decades [76,77]. However, a large number of articles reported that allo-SCT failed to be used widely because of unusable donors, high TRM and GVHD. Compared with auto-SCT, allo-SCT is still desirable due to the higher rates of molecular responses, longer-term disease control and graft-versus-myeloma effect despite higher TRM. Additionally, the increasing use of RIC/NMA regimens and PBSC as source of graft, advances in supportive care and effective infection prevention programs since 2000 may facilitate allo-SCT to be a potential way to cure MM. Therefore, we performed a systemic review and meta-analysis of 61 clinical trials reported between 2007.01.01 and 2017.05.03 involving 8698 adult patients to examine the efficacy and safety of allo-SCT for MM.
High-risk (HR) patients seem to be relatively resistant to novel agents, and had only short-term or no response to high-dose chemotherapy with or without following auto-SCT [55,[78][79][80]. In addition, some recent studies also reported high-risk patients may acquire new clonal abnormalities, showing rapidly progressing relapses after induction therapy followed by upfront ASCT, even evolving into extramedullary relapses or secondary plasma cell leukemia [81,82]. The International Myeloma Working Group (IMWG) has demonstrated that newly diagnosed high-risk patients treated with conventional therapies had a median overall survival (OS) of only 2-and 4-year OS was only 33% [83]. Medical Research Council (MRC) Myeloma IX trial showed ultra-high risk MM defined by ISS II or III in the presence of > 1 adverse lesion, including adverse IGH translocations, + 1q21 and del(17p13), have a particularly poor outcome (a median PFS of only 9.9 months and a median OS of 19.4 months) after being treated with auto-SCT [84]. In contrast, all 14 trials in our study showed that patients with high cytogenetic risk after allo-SCT had similar OS and PFS compared to those with standard risk, suggesting that allo-SCT may overcome the adverse prognosis of high cytogenetic risk. Furthermore, a included prospective study showed that after performing auto-SCT followed by allo-SCT, high risk patients with del(17p)/t(4;14) had similar remission rate, PFS, OS and relapse rate to those without del(17p)/t (4;14). Even ultra-high risk patients obtained molecular complete remission [43]. Importantly, Barlogie et al. showed flat survival curves between 4 and 10 years post allo-SCT, which suggested a proportion of these high risk patients may experience prolonged disease control or perhaps cure [85].
Some negative factors were especially important consideration in the counseling, implementation, and post treatment management of allo-SCT. We found TRM at 100 days and 1 year were 12 (5-18)% and 18 (14-21)%, respectively. Although compared with 50% TRM when the allo-SCT was initially performed [8], the present TRM has been drastically reduced, it was still too high for standard-risk patients. It should be mentioned that the patients in our study were heavily pretreated, which probably explains the high TRM observed. Indeed, more than half of patients had received at least 2 prior lines of treatment and 25% had at least 2 prior auto-SCT. The costs of allogenic transplant for MM are greater than those for chemotherapy and autologous transplantation [86,87]. The median number of hospital days for allo-SCT are longer than that for auto-SCT [88]. The article we included did not mention quality of life (QOL), but many other articles showed comparisons between patients after transplantation and adults without cancer. These studies showed that patients who underwent transplantation have low or moderate impairment in physical, social, psychological, and emotional functioning, as well as overall QOL [89][90][91]. The decline in overall QOL for auto-SCT was transient, but it was a longer term for allo-SCT [92]. However, Bush et al. reported that 80% of survivors had returned to work or school to resume their roles at home and in the community at 2 years after transplantation [93]. Specifically, patient reported benefits include an enhanced appreciation for life, different priorities, love and appreciation for family and friends, and greater religious or spiritual beliefs [90,94,95]. These data suggest that patients are often able to reinterpret the adversity of allo-SCT into a meaningful life narrative despite reduced QOL. Though 5-year PFS of 27% after allo-SCT isn't an ideal outcome, it's still better than 5-year PFS of only 19% after tandem auto-SCT reported by other articles [49]. High-risk patients-either upfront or relapsed-may be candidates for the allo-SCT treatment, when the risk of disease progression may outweigh the transplant-related risks. A large number of prospective randomized controlled trials were needed to prove the benefits of these therapeutic options.
Even when limited to patients with high risk MM or as first or second line salvage, the risk of acute and chronic GVHD and the high rates of recurrence after allograft need to be addressed to make allografting a readily acceptable treatment options for MM patients (Figs. 1, 2, 3, 4, 5, 6).
We found aGVHD was an independent predictor for shorter OS and PFS, while cGVHD might prolong survival due to an accompanying GvM effect and can thus in a mild form be regarded as an advantageous feature. However, some studies have reported cGVHD was also related to chronic diarrhea, appetite loss and inferior QOL [90,96,97], and was the most important cause of TRM. Novel strategies for acute and chronic GVHD prevention need to be explored in the setting of allografting for MM. Kroger et al. reported High-dose ATG could decrease the risk of aGVHD without the improvement of relapses [64]. T cell depletion using CD34 selection also could reduce the occurrence of GVHD [98].
Currently, it is noticeable that allo-SCT still remains high relapse rate [50 (45-55)%] and disease progression dominates the cause of the death (48%). Post-transplant immune treatment may be the key to sustain remission, prolong PFS and reduce relapse rate。Post-transplant therapies consist of donor lymphocyte infusions (DLI), possibly combined with immune stimulatory drugs [99,100]. Monoclonal antibodies, checkpoint inhibitors, vaccines and additional cellular therapies such as CAR-T-cells and NK-cell therapy could be used in the future [101]. They specifically target the remaining myeloma cells and limit the risk of augmenting GVHD [101]. Use of novel agents post-allografting has also been explored. For example, due to its dual effect-preventing GVHD and preserving GVM-bortezomib may be ideal to use upfront in the conditioning regimen and/or for consolidation and maintenance in the allogeneic transplant setting [102,103].
We identified many prognostic factors that may affect the outcome of allo-SCT. Since RIC/NMA was firstly introduced in 2000, MA have been gradually replaced to reduce TRM. Although the widespread adoption of RIC/NMA regimens failed to bring about better OS and PFS in the present meta-analysis, but it evidently enhanced patients' over QOL [104,105]. PBSC serving as the source of hematopoietic stem cells increased donors' availability and made allo-SCT more accessible. We found PBSC was associated with faster engraftment kinetics and quicker immune reconstitution, but less benefits in OS and PFS when compared with BM. In addition, three large retrospective studies [17,30,32] indicated unrelated donors in MM patients had similar Fig. 1 Flow diagram of study selection method engraftment rate to HLA-matching donors and acceptable TRM. Unrelated donors owing to a better donors' availability may be the feasible option. In our analysis, four studies indicated that compared with direct allo-SCT, planned autologous transplantation before allo-SCT had lower TRM as well as higher OS and PFS. Since 2000, nearly all allo-SCTs have adopted RIC/NMA regimens, which had insufficient cytoreduction. To our opinion, planned auto-SCT leads to sufficient cytoreduction and offset the disadvantage of RIC/NMA, thereby having better anti-MM effect than direct allo-SCTs. We showed the quality of response was also an independent prognostic factor for better PFS and OS before and after allo-SCT. Future therapeutic strategies should pay more attention to patients in non-CR state after the induction therapy and post-transplantation therapy. Adoptive immunotherapy alone or in combination with novel drugs can be applied to help these patients reach CR state. We found patients' age may be an independent predictor for shorter PFS and OS. A large prospective study reported donor's age > 50 years mean worse OS (HR = 1.99, 95% CI 1. 22-3.25), which emphasized increasing donor's age impaired donor stem cells' repopulation and homing abilities [58].
Several limitations should be considered in this metaanalysis. (1) Because the total sample size applied to perform a specific analysis was too small, we included the results of univariate analysis. Therefore, the conclusion was inaccurate in some way. (2) Significant heterogeneity was founded in the present study. We performed subgroup analyses and sensitivity analyses to explore the causes, and found the transplantations period of 1990s may be one source of significant heterogeneity. Because the 1990 s' subgroup included a few patients who had allo-SCT after the year 2000, and the articles involved didn't provide the individual data of those patients, we were incapable to separate the data out. In addition, the heterogeneity may come from different conditioning regimens, GVHD prophylaxis and patient selection bias (age, comorbidity, stage of disease) among different studies. But 95% confidence interval of all results was narrow, which meant the conclusion was credible in a way (Tables 1, 2). Importantly, this study had its own remarkable profits. (1) To our knowledge, the present meta-analysis was the first and largest comprehensive review of the role of allo-SCT in the treatment of MM patients. We found the most suitable subgroup of patients for allo-SCT and the best therapeutic time window of allo-SCT in MM (2) In our analysis, 29/61 studies scored 5 that was deemed as good quality, resulting in a more preferable conclusion to some extent. (3) We investigated many factors which may emerge as predictors of survival outcomes in MM patients after allo-SCT. The present metaanalysis may provide indications to policy makers and holistic clinicians in applying allo-SCT to clinical practice (Additional files 1, 2, 3, 4, 5, 6,7,8,9,10,11,12,13,14,15,16,17).

Conclusion
Due to the lack of consistent survival benefit, allo-SCT should not be considered as a standard of care for newly diagnosed and relapsed standard-risk MM patients. However, for patients with high-risk MM who have a poor long-term prognosis, allo-SCT may be a strong consideration in their initial course of therapy or in first relapse after chemotherapy, when the risk of disease progression may outweigh the transplant-related risks. A  large number of prospective randomized controlled trials were needed to prove the benefits of these therapeutic options.

Additional files
Additional file 1: Table S1. Characteristics of studies included in the meta-analysis.
Additional file 2: Table S2. Patients' transplant outcomes of individual clinical trials.
Additional file 3: Table S3. Quality assessment of individual clinical trials.
Additional file 4: Table S4. Begg and Egger test of studies included in the meta-analysis.