Hematopoietic stem cell transplantation donor sources in the 21st century: choosing the ideal donor when a perfect match does not exist

The ability to perform hematopoietic stem cell transplantation (HSCT) hinges on the availability of a suitable donor. The best donor for HSCT is an HLA-matched sibling or unrelated donor. Unfortunately, on the basis of average family size, less than 30% of patients will have a matched sibling donor.¹  As ethnic diversity increases in Europe and North America, it is imperative to have a strategy to identify an alternative stem cell source when an adult matched unrelated donor (MUD) cannot be identified. At present there are 3 alternative donor options: a partially HLA-mismatched unrelated donor (MMURD), a haploidentical related donor, and an umbilical cord blood (UCB) stem cell product.

The heterogeneity of patients in observational studies in the literature does not lend itself to direct comparisons of alternative donor sources. Until such studies as the Blood and Marrow Transplant Clinical Trials Network (BMT-CTN) randomized comparison of UCB and haploidentical transplantation (BMT-CTN 1101, NCT01597778) are complete, physicians are left to cautiously interpret the existing data in making these important decisions. There are patient-related, disease-related, and transplant protocol–related factors that together uniquely affect the clinical outcomes of an individual patient. Here we summarize the existing body of evidence in an effort to guide clinicians in using the current literature to help make a decision on donor source for an individual patient when a matched related donor (MRD) or MUD are unavailable.

In this analysis, MMURD refers to an adult unrelated donor mismatched in at least one antigen or allele at HLA-A, -B, -C, or -DR. Mismatched related donor transplants are not included in this analysis. Older studies evaluated HLA compatibility on the basis of antigen matching, that is, by using anti-HLA antibodies or by low-resolution molecular typing. HLA typing precision improved to high-resolution molecular typing of the HLA locus, termed HLA allele matching, which became available over the last 2 decades. It is clear that allele-level typing gives more reliable results than antigen-level typing, so earlier observational studies using antigen level testing need to be interpreted cautiously. There is growing evidence that not all HLA mismatches are created equal. Permissive HLA mismatches seem to confer similar transplant-related outcomes when compared with matched donor sources, presumably reflecting the inability of the T cell to recognize an intrinsic HLA sequence difference,²  as well as the tendency of the mismatched HLA molecules to present similar minor histocompatibility proteins to the immune system. A nonpermissive HLA allele mismatch combination leads to poorer outcomes.^2-6  MUD transplants increase donor availability but take time to organize because of donor screening and graft retrieval.

In this review, haploidentical refers to a complete half mismatch (generally 3 of 6 or 4 of 8) from a related donor. These transplants have the advantage of speed because relatives are usually easy to contact for stem cell collection. The cost of collection is generally lower than in MMURD and UCB products. The major disadvantage of haploidentical donors is the HLA disparity. In T-cell–depleted haploidentical grafts, selection of a maternal over a paternal donor has been shown to result in better survival, because the maternal immune system is tolerized to fetal antigens during pregnancy.⁷  The benefit of maternal versus paternal donor is not as clear in T-cell–replete grafts.⁸  The use of a haploidentical sibling with noninherited maternal antigens has been associated with lower transplant-related mortality (TRM) and better graft-versus-host disease (GVHD) outcomes when compared with haploidentical sibling donors with noninherited paternal antigens.⁸ 

UCB stem cell products are cryopreserved and stored so they are readily available.⁹  The minimal number of T cells in a UCB product allows it to be used across HLA barriers. Typically UCB stem cell products are HLA mismatched at 1 to 6 antigens or alleles. The disadvantage is the small size of the product, which limits the stem cell dose in adults and often requires the use of a second UCB product. UCB units are also expensive because each unit must be bought from a bank that needs to recoup the costs of typing, cryopreservation, and storage.

When comparable conditioning regimens and cellular products are used, total nucleated cell (TNC) dose, engraftment time, and reliability with MMURD, MRD, MUD, and haploidentical donors are similar.^10-12  UCB products have the lowest effective cell dose based on the amount collected and losses incurred during cryopreservation and thawing. The median TNCs and CD34⁺ cells infused from a single UCB product is between 1.0 and 3.3 × 10⁷ cells per kilogram and 0.74 and 1.2 × 10⁵ cells per kilogram, respectively.^13-18  This is 10-fold fewer stem cells compared with adult donor stem cell products. In situations where body size is large, a strategy to increase the effective stem cell dose by the use of two UCB products is generally adopted. This increases the TNC and CD34⁺ cell dose infused and reduces the duration of cytopenias in adults but still results in slower count recovery than anticipated in MMURD and haploidentical transplant recipients.^19-21  One may reasonably anticipate at least a 7-day prolongation in time to neutrophil recovery in adults receiving a double UCB transplant when compared with peripheral blood reduced-intensity conditioning (RIC) MUD transplants (21.5 vs 13 days). Platelet recovery time is delayed from a median of 19 days in MUD to 41 days in UCB recipients.²²  Research to expand stem cells in UCB units has demonstrated an improved engraftment time with these strategies,^23,24  but this remains experimental.

Graft failure can be mediated by cellular or humoral immunity or it may reflect insufficient or damaged stem cells. Immunologically mediated rejection can be caused by sensitization of the recipient to nonshared HLA antigens. For instance, in all 3 alternative donor sources, the risk of graft failure is higher in transplant recipients who have donor-specific anti-HLA antibodies.^25-27 

There is an approximately 10% graft failure rate in MMURD transplants, significantly higher than that observed in MRD and MUD transplants.^28-31  Similar to MUD transplants, the risk of graft failure is higher with bone marrow (BM) than with peripheral blood stem cells (PBSCs) as a graft source in MMURD transplants (16% with BM vs 3% with PBSCs).^32,33  The direction or vector of the HLA mismatch may also be important. In patients with an HLA nonpermissive mismatch in the host-versus-graft vector, the risk of graft failure is increased when compared with permissive or MUD transplants.⁶ 

The Perugia group has reduced graft failure rates by increasing the CD34⁺ cell dose (so-called “mega-dose”) with a CD34⁺-selected PBSC graft, in which the median CD34⁺ cell dose was 13.8 × 10⁶ cells per kilogram (range, 5.1 to 29.7 × 10⁶ cells per kilogram).³⁴  This results in a primary engraftment failure rate of 9%. Rizzieri et al³⁵  observed a 6% graft failure rate in RIC haploidentical transplantation by infusing similarly large CD34⁺ cell doses (median, 13.5 × 10⁶ cells per kilogram), but in contrast to the Perugia regimen, they used in vivo T-cell depletion with alemtuzumab. Huang et al³⁶  used a combination of T-cell–replete BM and granulocyte colony-stimulating factor–primed PBSCs with an augmented myeloablative (MA) conditioning regimen that included anti-thymocyte globulin (ATG), which resulted in almost no primary engraftment failures. Drobyski et al¹¹  also observed a low graft failure rate of 4% by using augmented conditioning with an ex vivo T-cell–depleted BM graft. In patients receiving the Hopkins strategy of posttransplant cyclophosphamide with a T-cell–replete graft, the graft failure rate is 10% in MA³⁷  and 13% in RIC³⁸  transplants.

In UCB transplantation, there are few passively transferred T cells from the donor to protect against graft rejection, and this may be more problematic because of the low TNC and CD34⁺ cell doses. Graft failure is close to 10% in RIC UCB transplantation^21,39-41  and as high as 20% in MA UCB transplantation.¹⁶  If engraftment failure follows UCB transplant, there is no opportunity to return to the donor for more stem cells. Thus, either additional UCB needs to be used and the recipient must survive the additional period of cytopenias, or a haploidentical donor transplant may be attempted. Salvage RIC haploidentical transplantation has been used with some success after graft failure from UCB transplantation.⁴² 

UCB transplant is associated with the highest engraftment failure rate. Furthermore, options after UCB infusion for graft failure are limited because returning to the donor is not an option. Small studies have shown that haploidentical stem cells can be infused at the time of UCB transplant or at evidence of graft failure to promote engraftment.^43,44  In recent years, the graft failure rate in haploidentical transplantation has improved to levels comparable to those of MUD, MRD, and MMURD. If graft rejection does occur after haploidentical transplantation, it is also difficult to retransplant patients who have become sensitized to unshared alleles or antigens. Graft failure rates for alternative donor transplants are summarized in Tables 1 and 2.^45-57 

The HLA disparity that results in high engraftment failure rates with alternative donors also results in a higher rate of GVHD. Woolfrey et al⁵⁸  examined MMURD with PBSCs as a graft source and found an increased risk of acute grade 3 to 4 GVHD with single-allele MMURD when compared with matched transplants (relative risk [RR], 1.59; 95% CI, 1.20 to 2.09) but no difference in chronic GVHD (cGVHD). A similar study by Lee et al⁵⁹  in patients receiving a BM graft showed that single-allele MMURD transplants had more grade 3 to 4 acute GVHD (aGVHD) (RR, 1.34; 95% CI, 1.12 to 1.61) than MUD transplants. The majority of patients in these studies received MA conditioning, but a similar increase in aGVHD and not cGVHD has been observed with RIC MMURD when compared with MUD transplants.⁶⁰ 

In MA PBSC or marrow MMURD transplants, adding ATG to calcineurin inhibitors for GVHD prophylaxis results in a rate of grade 2 to 4 aGVHD of 30% to 40%.^11,28,61,62  In contrast, calcineurin inhibitor–based GVHD prophylaxis without ATG in MMURD transplants results in aGVHD rates of 50% to 80%.^29,63-65 

HLA class I allele mismatched as well as HLA-DRB1 allele mismatched transplants are associated with higher rates of aGVHD when compared with MUD transplants.^58,64,66  The difference in cGVHD between MMURD and MUD transplants is not as clear, but some reports do show a higher risk of cGVHD with HLA class I mismatched transplants.^64,67,68  In contrast, HLA-DQ and HLA-DP mismatches have not always been shown to worsen clinical outcomes when compared with matched donor recipients.^31,58,59  In patients who are otherwise matched at HLA-A, -B, -C, and -DRB1, there is no survival difference if there is an HLA-DQ mismatch, but HLA-DQ mismatch may worsen outcomes in patients who already have a 1- or 2-allele HLA mismatch.⁵⁹  HLA-DPB1 mismatches are associated with lower relapse rates but with an associated increase in GVHD and TRM.^69-71  HLA-C mismatched donor transplants have been associated with worse survival when compared with HLA-C matched donor transplants, likely because of more severe aGVHD.⁷² 

In studies using high-intensity conditioning by adding cytarabine and semustine to cyclophosphamide and busulfan or total-body irradiation, graft failure rates have been low, but aGVHD has been 40% to 60%, despite the use of ATG.^11,36,73  Although Perugia used mega-doses of CD34⁺-selected stem cells, the alternative approach of using ATG for GVHD prophylaxis and standard MA conditioning demonstrated low rates of aGVHD and cGVHD (8% and 3%, respectively) but with higher graft failure rates³⁴  as described in the “Engraftment failure” section. Both approaches continue to have drawbacks: slow posttransplant immune reconstitution in patients who receive T-cell–depleted transplants and aGVHD and cGVHD in those who receive T-cell–replete grafts.⁷⁴  In RIC haploidentical transplantation, the Johns Hopkins group pioneered the use of posttransplant cyclophosphamide in an effort to reduce GVHD rates. This approach has proven to be very effective, requiring no stem cell manipulation, a simple marrow collection, and well-tolerated conditioning with modest toxicity. The rate of acute grade 2 to 4, acute grade 3 to 4, and cGVHD was 34%, 6%, and 5%, respectively.³⁸ 

The less stringent HLA matching needed when selecting a UCB unit for transplantation is not associated with an increased risk of GVHD-related mortality. The risk of aGVHD appears to be slightly lower with the use of ATG with MA conditioning, but this difference is not as appreciable with RIC UCB transplants (Tables 1 and 2).

Grade 2 to 4 aGVHD was significantly higher in MMURD (85%) compared with UCB transplants (53%) at 100 days after HSCT for hematologic malignancies, but grade 3 to 4 aGVHD was not significantly different.⁶⁵  Similarly, cGVHD rates at 2 years after HSCT were significantly higher for MMURD transplantation when compared with UCB transplants (48% vs 26%, respectively). Smaller studies have not been able to show an appreciable difference in GVHD rates after UCB or MMURD transplantation.^75,76 

In parallel phase 2 trials of UCB or haploidentical transplantation using posttransplantation cyclophosphamide conducted by the BMT-CTN, the grade 2 to 4 GVHD rate was 40% for UCB transplants recipients and 32% for haploidentical transplant recipients. The grade 3 to 4 aGVHD rates were 21% and 0% for UCB and haploidentical transplants, respectively. cGVHD was also higher in UCB transplants: 25% compared with 13% at 1 year after transplant.⁴⁰ 

GVHD is most frequent in MMURD transplants and is comparable between UCB and haploidentical transplants when posttransplant cyclophosphamide is used for the latter. Studies in alternative donor transplants use a variety of GVHD prophylaxis strategies, which complicates a comparative analysis. With the high degree of HLA mismatch in haploidentical transplants, novel techniques to reduce GVHD are necessary such as high stem cell doses, in vivo or ex vivo T-cell depletion, and posttransplant cyclophosphamide. None of these strategies have been compared directly with each other, and the risk of graft failure needs to be weighed against the benefit of graft-versus-leukemia effect in each of these strategies. Tables 1 and 2 summarize GVHD rates in alternative donor transplants.

The reported TRM is highly variable and likely accounted for by the diversity in conditioning regimens, underlying disease, use of PBSCs or BM, and comorbidities at time of transplantation. Most studies do not account for these differences when reporting an overall TRM.

Multiple studies have demonstrated that the long-term nonrelapse mortality is significantly higher in MMURD when compared with MUD transplants. In a large Center for International Blood and Marrow Transplant Research (CIBMTR) study of more than 4000 patients receiving a transplant for chronic myeloblastic leukemia in chronic phase, the 5-year TRM was 31% in MRD, 38% in MUD, 50% in 1 HLA class I MMURD, and 48% in 1 HLA class II MMURD.⁶⁷  In another large retrospective study of 1800 patients, the RR of an allele MMURD was 1.4 (95% CI, 1.09 to 1.81) when compared with HLA allele-MUDs,⁵⁸  and this observation is consistent among other studies.^29,59,64 

In a study of patients receiving MA conditioning for a haploidentical transplant with ex vivo T-cell–depleted BM, the 2-year TRM in haploidentical transplants was 42%, which was not significantly different from MMURD (45%) but significantly higher than MUD transplants (23%).¹¹  In a more recent analysis by Wang et al¹²  in which patients with acute monoblastic leukemia (AML) or acute lymphoblastic leukemia received MA conditioning with ATG as part of GVHD prophylaxis, there was no appreciable difference in TRM between haploidentical (34%) and MRD transplant recipients (38%) at 2 years after transplant. Another study examined the impact of T-cell depletion with ATG on transplant outcomes and found that the use of ATG increased the TRM significantly from 16% to 42%.⁷⁷  In patients receiving the Perugia regimen, the TRM was 36.5%,³⁴  mostly because of infection, which reflected impaired T-cell immune reconstitution. In patients receiving posttransplant cyclophosphamide without ATG to prevent GVHD, the TRM has been reported as 16% at 100 days after transplant.³⁷ 

In heavily pretreated patients with Hodgkin disease, the TRM with RIC haploidentical transplant was 8% at 2 years, significantly better than MUD and MRD transplants in the study by Burroughs et al.⁷⁸  Low TRM has also been reported in RIC haploidentical transplants for patients receiving posttransplant cyclophosphamide (4% to 6%).^38,40  A small study in high-risk patients with AML who received RIC haploidentical transplantation did not demonstrate a difference in TRM when compared with MMURD transplants (10.1% vs 17.9%, respectively).¹⁰ 

Long-term TRM of UCB transplants has been comparable to MUD and MRD transplants in some studies,^13,41,79  but the upfront mortality associated with UCB transplants is higher.^80,81  Small studies that have compared TRM of UCB to MMURD could not demonstrate a significant difference, but these were small studies that were likely underpowered.^65,76 

The TRM in RIC UCB transplant recipients is reportedly higher than that observed in MUD and haploidentical transplants. In a study of 2 phase 2 parallel trials, the 2-year TRM with UCB was 24% compared with 7% in haploidentical transplant recipients.⁴⁰  It is difficult to compare stem cell sources because this was not a randomized trial, and posttransplant immune suppression differed by stem cell source. Nevertheless, the data do suggest that early TRM with UCB is higher than that with haploidentical transplants. This is likely due to the slower UCB transplant lymphohematopoietic reconstitution.⁸²  This higher rate of TRM in UCB is also observed when compared with MUD transplants.^{21,22,39,41,83} 

In all alternative donor transplants, the TRM is generally higher than that observed in matched donor transplants when similar conditioning regimens are used. In the MA setting, the TRM from haploidentical and UCB transplants is comparable with that for MMURD transplants. RIC reduces the risk of TRM in haploidentical and UCB transplants. Comparing RIC haploidentical and UCB transplants suggests that UCB has a higher TRM than haploidentical transplants, likely related to the slower engraftment and risk of infection-related fatality with UCB. Tables 3 and 4 summarize TRM rates in alternative donor transplants.

In MMURD, regardless of the indication for transplant and the conditioning regimen, patients generally have relapse rates comparable to those of MUD and MRD transplant recipients. A large retrospective CIBMTR analysis compared 521 patients who received a 1 or more allele MMURD to 3514 patients who received an MRD transplant. Relapse rates at 5 years were 14% in MRD, 12% in MUD, 11% in single class I mismatch and 9% in single class II mismatched donors, and these were not significantly different in multivariate analysis.⁶⁷ 

Relapse in haploidentical transplants is highly variable based on the GVHD and graft failure strategies used. In patients receiving an RIC haploidentical transplant with posttransplant cyclophosphamide to reduce GVHD, the relapse rate was 45% at 1 year after transplant.⁴⁰  The relapse rate for MA haploidentical transplant with posttransplant cyclophosphamide was 22% at a median follow-up of about 11 months, suggesting that the higher intensity conditioning can improve relapse rates in patients receiving posttransplant cyclophosphamide. In patients receiving the Perugia regimen, the cumulative incidence of relapse at 6 months was 25% for all patients but was significantly higher in patients transplanted in relapse (51%) when compared with those transplanted in remission (16%).³⁴ 

Patients receiving MA UCB transplants are associated with a lower risk of relapse (16%) when compared with matched and mismatched transplants (37% to 52%).⁶⁵  In patients receiving an RIC double UCB transplant, the 1-year relapse incidence was 31%. This appeared to be lower than the 45% observed in the parallel haploidentical transplant study⁴⁰ ; however, patients who received UCB had a higher TRM. Therefore, the relapse rates are not directly comparable, because more haploidentical patients were at risk for relapse.

Disease risk stratification and status at time of transplantation are ultimately very important in determining risk of relapse or progression after transplant, and it is difficult to compensate for these issues in noncomparative trials.

A large retrospective study of about 1300 patients receiving an HSCT for malignant hematologic diseases showed a 3-year survival of 40% in MMURD, a 20% disadvantage when compared with MUD transplants.⁶⁴  This relationship has been observed in multiple other studies of MMURD compared with matched donor sources (Tables 3 and 4). When comparing RIC and MA conditioning regimens, Woolfrey et al⁵⁸  were not able to demonstrate a difference in survival based on regimen intensity. Woolfrey et al went on to compare the results of their study with PBSC MMURD, and the results were reported by Lee et al⁵⁹  in which patients had a similar MMURD transplant but with a BM graft. No difference in overall survival (OS) was observed in single-antigen MMURD transplants with PBSCs when compared with single-antigen MMURD with BM (RR, 1.13; 95% CI, 0.93 to 1.40).

In RIC MMURD transplants, the OS difference when compared with matched donor sources is not as clear. Koreth et al⁶⁰  recently reported a large retrospective analysis of the CIBMTR in which a 3-year OS of 30.9% for RIC-allele MMURD was seen, significantly lower than that observed in MUD transplants (37.4%). A CIBMTR analysis of patients with non-Hodgkin lymphoma receiving RIC MMURD transplant showed that older age, shorter time from diagnosis to transplant, non–total-body irradiation conditioning regimen, ex vivo T-cell depletion, and HLA mismatch were associated with mortality.⁸⁵ 

The largest analysis to date in 756 adults receiving an MA haploidentical transplant for AML, chronic myeloblastic leukemia, or acute lymphoblastic leukemia had an excellent 3-year OS of 67%.⁷³  In patients with intermediate- or high-risk AML in first complete remission (CR1), 4-year OS was 77.5% and was significantly better than that for patients receiving chemotherapy alone.⁸⁶  Other studies have demonstrated similar survival outcomes with MA haploidentical transplants (Tables 3 and 4). When compared with patients who received MUD and MMURD transplants, patients who received a haploidentical transplant with an ex vivo T-cell–depleted BM graft had an OS at 2 years of 21%, significantly lower than both MUD (58%) and single-antigen MMURD (34%),¹¹  indicating that GVHD prevention strategy is an important variable.

In 83 patients with AML or myelodysplastic syndrome who received RIC haploidentical transplants with busulfan and fludarabine, including ATG as part of GVHD prophylaxis,⁸⁷  OS at a median follow up of 26 months was 60% in leukemia in CR1, 53% in CR2 and CR3, and 53% in myelodysplastic syndrome. Patients with refractory leukemia had a significantly lower OS at 9%. In high-risk patients with AML who received RIC haploidentical transplants and ATG for GVHD prophylaxis, 3-year OS was 65.7%,¹⁰  comparable to that in patients receiving a well-matched or partially matched MA HSCT in that study. In refractory or relapsed patients with Hodgkin disease, for most of whom a previous autologous transplant had failed, no difference in OS at 2 years was observed in MRD, MUD, or haploidentical transplants with RIC.⁷⁸ Table 4 summarizes the reported OS rates in RIC haploidentical transplants.

In two studies of patients receiving MA UCB transplants compared with MMURD transplants, the 3-year OS was 66% for UCB, significantly higher than that for MMURD transplants. In both studies, the TRM was lower and progression-free survival was higher with UCB compared with MMURD.^75,76  A CIBMTR observational analysis was not able to show a significant OS difference between UCB and MMURD transplants in patients receiving MA conditioning.¹³  In another study of UCB MA conditioning that included more than 500 patients, a 1-year OS of 37% was observed.⁸⁸  Other studies that reported OS in UCB HSCT with MA conditioning are summarized in Table 3. As in MMURD, the direction of HLA mismatch may be important in UCB transplantation. Unidirectional mismatch in the graft-versus-host direction alone may result in less TRM and better OS when compared with host-versus-graft unidirectional mismatches and bidirectional mismatches.⁸⁹ 

OS in patients receiving RIC UCB is between 34% and 63% in patients with hematologic malignancies.^39,41,90,91  In the BMT-CTN phase 2 parallel studies of UCB and haploidentical transplants, UCB appeared to have higher TRM (24% vs 7%) with comparable disease-free survival rates, resulting in an apparent OS benefit with haploidentical transplants (62% vs 54% with UCB transplants) at 1 year after transplantation. More TRM has also been observed with UCB transplants compared with MUD transplants, but similar disease-free survival and OS rates were observed.²² 

OS is a compound outcome of transplant-related complications and disease relapse. Patients receiving an MMURD have a higher rate of TRM than those receiving matched donor transplants, resulting in an overall lower OS. The range of OS rates after haploidentical transplants likely relates to the differences in transplant indication, severity of disease, and GVHD prophylaxis strategies. Compared with haploidentical and MMURD transplants, UCB transplantation results in higher TRM but similar progression-free survival and OS rates. Tables 3 and 4 summarize survival outcomes in alternative donor transplants.

Despite the large number of phase 2 and observational studies in the literature, the dearth of randomized trials makes the prioritization of an alternative donor difficult. The decision may in part reflect the research agenda of the transplantation center because no one source of stem cells is clearly superior to another. The anticipated outcomes of alternative donor transplantation are relatively predictable, and they are similar enough that randomized controlled trials will be critical to resolving the remaining issues. However, the large number of variables that influence outcomes makes it unrealistic to expect such trials to be conducted quickly enough or in large enough numbers to provide definitive recommendations. Nevertheless, there are themes that can be inferred by the current data to inform a decision.

If a donor is urgently required, UCB and haploidentical transplantation have the advantage over adult volunteer MMURD. In general, UCB products can be obtained promptly because they are cryopreserved and in inventory. Haploidentical family members are usually evaluated and scheduled for stem cell collection more promptly than unrelated donors. If upfront cost is a major concern, haploidentical donors have a clear advantage over both UCB products and MMURD. For patients who have had problems with Epstein-Barr virus, cytomegalovirus, or other infections, UCB may be less desirable because of both the delay in hematopoietic recovery and the lack of passively transferred cellular immunity. Strategies to speed hematopoietic recovery in UCB transplantation are very interesting, but they are unlikely to improve immunologic function, and they are likely to be quite expensive.

All MMURD transplants are not equal. Although in general, there is a higher risk of GVHD and outcomes similar to those of UCB and haploidentical transplantation, the ability to recognize permissive and nonpermissive mismatches and choosing donors with mismatches in the graft rejection direction rather than the GVHD direction may allow for graft vs leukemia effect with a risk of GVHD similar to that observed using matched donors.

Physicians are left with the current evidence to help them decide on what type of transplant would be best for their patient. Diagnosis, disease severity, disease status at time of transplant, conditioning regimens, graft source, and GVHD prophylaxis strategies are the factors that together influence clinical outcomes after transplantation. The outcomes have significantly improved with alternative donor transplantation, but much work remains to be done until they are proven to be noninferior to matched related and unrelated donor transplants.

Contribution: N.K. and J.H.A. performed the literature review and wrote and edited the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Joseph H. Antin, Dana-Farber Cancer Institute, 450 Brookline Ave, Boston, MA 02215; e-mail: jantin@partners.org.