Primary Results from BMT CTN 1702 - Clinical Transplant-related Long-term Outcomes of Alternative Donor Allogeneic Transplantation (CTRL-ALT-D)

Abstract

Purpose

The likelihood of finding an HLA-matched unrelated donor (MUD) for hematopoietic cell transplantation can be predicted using a donor search prognosis score. Patients without a MUD may use alternative donors (haploidentical related, mismatched unrelated, or umbilical cord blood).

Methods

This multicenter biologic assignment trial was conducted by the Blood and Marrow Transplant Clinical Trials Network (BMT CTN 1702). Eligibility criteria were broad to mirror clinical practice. The primary endpoint was 2-year survival from evaluability, compared between those Very Likely (>90%) and Very Unlikely (<10%) to find a MUD. All other patients, Less Likely to find a MUD, enrolled in an observational arm. Transplant outcomes were compared for all three groups.

Results

1751 evaluable subjects at 47 centers were Very Likely (54.7%), Less Likely (29.5%), and Very Unlikely (15.8%) to identify a MUD. Survival did not differ in univariate (HR 1.00, 95% CI 0.82–1.21, p=−0.98) or multivariate (HR 1.07, 95% CI 0.86–1.33, p=0.56) analyses between the Very Unlikely and Very Likely groups, measured through two years from the beginning of a search for a MUD or alternative donor. Of the transplanted patients (n=1179), MUD was used for 94% of the Very Likely, 38% of Less Likely and 9% of Very Unlikely patients. Multivariate analyses showed no differences in relapse, treatment-related mortality, disease-free survival, and acute and chronic GVHD for the three search prognosis groups after transplantation.

Conclusion

Using a donor search prognosis strategy to prioritize an alternative donor for patients Very Unlikely to find a MUD resulted in survival and transplant outcomes that were not statistically different compared to those Very Likely to find a MUD.

Introduction

Allogeneic hematopoietic cell transplantation (HCT) is standard of care for patients with high-risk or advanced hematologic malignancies and other blood diseases. The use of allogeneic HCT has expanded rapidly over the past decades owing to substantial advances in transplant procedures and supportive care.¹ However, providing an HLA-matched donor for every potential HCT patient in an optimal period of time remains a challenge. Approximately two-thirds of patients who need a transplant do not have a fully HLA-matched related donor (MRD) available²; most patients must rely on other donor choices. Historically, an 8/8 HLA-matched unrelated donor (MUD) has been accepted as the next best option,^3–6 and transplant outcomes after MUD grafts approximate those of MRD grafts.⁷

However, unrelated donor searches may be prolonged in order to find a “perfect match,” before use of an alternative (HLA mismatched) donor is considered. As a result, patients without a MRD may develop progressive disease⁸ or become medically unfit while searching for a MUD, which might have a negative impact on overall survival. Identification of a MUD may be especially challenging for genetically underrepresented racial/ethnic groups or mixed race individuals due to the composition of the unrelated donor registries,^2,9,10 need to do additional high-resolution HLA typing, and availability of potential donors to undergo testing and eventually donate.¹¹ With recent improvements in success using mismatched unrelated donors (MMUD),^12,13 haploidentical relatives^14,15 and cord blood,^16,17 more HCTs are being performed using these alternative sources.¹⁸ Choice of alternative donor type is likely determined by donor availability, clinical trial availability, patient medical status and center preference. However, we lack evidence about which is the best alternative donor source or when in the search process to include these donor types, and there is no consistent national practice or algorithm to guide the search strategy for all patients.

Work by Dehn and colleagues has defined a “donor search prognosis” based on HLA allele frequencies and race/ethnicity. This score predicts the likelihood of successfully identifying an 8/8 MUD at HLA-A, -B, -C, and -DR.^19,20 Patients who are Very Likely to find a MUD have a >90% likelihood of finding a MUD, while those who are Very Unlikely to find a MUD have a <10% chance. Patients who are Less Likely have an approximately 26% chance of finding a MUD. Worse search prognosis is associated with underrepresented racial/ethnic groups, as well as frequency of haplotypes, but not with other patient and disease biology characteristics that might influence the success of HCT. Because disease progression/relapse and cumulative toxicity of disease-directed therapy are major reasons that transplants are not performed, both the speed of getting to HCT²¹ and the outcome of transplantation are important determinants of which donor source is preferable.⁸ We hypothesized that using the donor search prognosis to guide the initial search strategy, where the Very Likely group pursues a MUD and the Very Unlikely group immediately pursues an alternative donor, would result in improved survival and transplant outcomes for the Very Unlikely group, adjusting for baseline clinical variables, despite having different donor sources. Data for patients in the Less Likely group were collected but no donor selection strategy was specified in the protocol, and they were not included in the primary comparison.

Methods

Study design

This was a multicenter biologic assignment study²² to test whether provision of a donor search prognosis score to guide the search strategy would improve outcomes by directing patients with a very low likelihood of finding an 8/8 HLA-A, -B, -C, and -DR-matched MUD to proceed to HCT expeditiously using the best alternative donor. (Supplementary Figure 1) This design was selected because randomization between continued search for a MUD and pivoting to an alternative donor was not practicable nor considered ethical. Comparison to an historical control could not be done because we lacked an appropriate comparator cohort with information about patients who did not proceed to transplant.

There were no requirements for age, disease status, conditioning regimen or graft-versus-host disease (GVHD) prophylaxis. Subjects could be enrolled at any time up until formal unrelated donor search was started and were declared “evaluable” once the center concluded that the patient did not have an HLA-identical or 1-antigen mismatched family donor and was still a transplant candidate.

Other eligibility requirements included diagnoses of acute myeloid leukemia (AML), acute lymphoblastic leukemia (ALL), myelodysplastic syndromes (MDS), non-Hodgkin lymphoma (NHL), Hodgkin lymphoma (HL), acquired aplastic anemia (AA) or sickle cell disease (SCD). Participants had to be considered suitable HCT candidates per institutional standards with intent to proceed to HCT within 6 months of enrollment and needed to commit to following the donor search strategy dictated by the search prognosis algorithm. Patients who had a prior allogeneic HCT or formal unrelated donor search were not eligible. The study was approved by the National Marrow Donor Program (NMDP) Institutional Review Board, and all participants provided written informed consent. The trial was registered in clinicaltrials.gov with NCT#: 03904134.

Donor search prognosis

Centers submitted recipient HLA type and self-reported race/ethnicity to the NMDP as soon as a patient was declared evaluable. The donor search prognosis results were returned to centers within 2 business days so it could be used to guide the search strategy.

Biostatistical considerations

Sample size

Sample size requirements for this study were based on the primary analysis comparing overall survival from evaluability between the Very Likely and Very Unlikely groups. The target enrollment of 1022 patients (730 Very Likely, 292 Very Unlikely, 679 events expected) provides > 85% power to detect a hazard ratio (HR) of 0.76, corresponding approximately to a 10% improvement in OS at two years. This calculation used a log-rank test with a two-sided significance level of 5%, and assumed baseline survival of 30% at 2 years, 5% exponential rate of loss to follow-up per year, 3 years of accrual, total study time of 4.5 years, and censored all patients at 2 years since few events are expected to occur after 2 years. Study design assumptions, particularly the ratio of patients in the Very Likely vs. Very Unlikely groups, were monitored throughout the study to ensure the power requirements were maintained. Patients in the Less Likely group, approximately 710 patients, followed a search strategy decided by the center and were observed for outcomes but were not part of the primary analysis.

Analysis plan

All analyses were performed using statistical software SAS V9.4 or R V3.6.3. Population characteristics were compared using Pearson chi-square or Kruskal-Wallis tests for categorical or continuous variables, respectively.

Survival through two years from the time of evaluability:

The primary analysis population included all patients registered who met criteria for evaluability, i.e., did not have a suitable matched or 1 allele or antigen mismatched family donor, and who were in the Very Likely or Very Unlikely groups. The primary endpoint of survival through two years after evaluability was measured from the time the patient was deemed evaluable and analyzed according to the donor search prognosis, regardless of the donor used, as an Intention-to-Treat analysis. The rationale for truncating the primary analysis at two years is to avoid bias because patients who did not undergo HCT were only followed for two years but transplanted patients continue to be followed through routine data collection by the Center for International Blood and Marrow Transplant Research (CIBMTR). The primary null hypothesis of the study was that there was no difference in overall survival between the Very Likely and Very Unlikely to find a MUD groups. Because of the potential bias resulting from this biologic assignment mechanism (using donor search prognosis rather than randomization assignment),²² the comparisons of overall survival between groups were done using a Cox proportional hazards model adjusted for the following pre-specified patient characteristics: age at enrollment, sex, Karnofsky performance status (KPS) at enrollment, race, ethnicity, disease, and disease stage at enrollment, if applicable, and interval from consent to evaluability. A p-value <0.05 was considered significant for the primary analysis. It was not possible to adjust this model for co-morbidities or disease risk index because this information was not collected until the time of transplant, so was not available for use in the primary analysis which included all evaluable patients, whether or not they underwent HCT. Adjusted overall survival was also described based on a Cox model stratified on search prognosis group and adjusting for the same variables.²³ Additional exploratory analyses examined interactions between each covariate and search prognosis group, tested for a center effect using the score test of Commenges and Anderson,²⁴ and checked the potential impact of Covid era on the outcomes by adjusting for whether the patient was declared evaluable before or after the Covid pandemic started.

Transplant outcomes:

Several secondary analyses were performed on the subset of patients in all three search prognosis groups who received a transplant, according to their search prognosis score. These outcomes were measured from the start of transplant and were censored at time of last contact.

Post-transplant survival and disease free survival (DFS) were summarized in each donor search prognosis group using the Kaplan-Meier estimate, and compared using a Cox proportional hazards model adjusted for the following pre-specified patient characteristics: recipient age at HCT, race/ethnicity, KPS at HCT, cytomegalovirus (CMV) serostatus, disease risk index (DRI) at HCT,²⁵ co-morbidity index (HCT-CI) at HCT,^26,27 and conditioning regimen intensity. Relapse, treatment-related mortality (TRM), grade II-IV and III-IV acute graft-versus-host disease (GVHD) and chronic GVHD were described using cumulative incidence, with relapse as a competing risk for TRM, and death as a competing risk for the others. Cause specific hazard rates for each outcome were compared between donor search prognosis groups using the Cox proportional hazards model, adjusted for the same pre-specified patient characteristics as above. Disease-free survival, relapse and TRM were not reported for non-malignant patients. We avoided adjustment for factors associated with the donor type such as donor age, graft type (bone marrow vs. peripheral blood), and GVHD prophylaxis regimen because this was an intention-to-treat analysis evaluating donor search prognosis groups and search strategy, not actual donor used. Adjusted posttransplant survival and DFS were summarized by search prognosis based on a stratified Cox model adjusting for the same variables, while adjusted cumulative incidence was described based on a stratified Fine-Gray model.²⁸ The figures are truncated at 24 months although all available data were used for the analysis of transplant outcomes.

Results

Participant characteristics

The study opened to enrollment in June 2019 and closed in May 2022. Fifty-five centers were activated, and 47 enrolled at least one subject (Supplemental Table 3). The study enrolled 2224 participants of whom 1751 were considered evaluable (Figure 1). Their characteristics are shown in Table 1 with additional details in Supplementary Table 1. There were 958 evaluable patients in the Very Likely donor search prognosis group, 517 in the Less Likely group, and 276 in the Very Unlikely group. Among evaluable patients, 827 have died, and 924 were still alive, whether or not transplanted, with a median of 27 months of follow-up (range=6–54 months, inter-quartile range=25–34 months).

Figure 1.

Consort diagram

*Not evaluable (>10 subjects): HLA-matched related donor (n=238), Missing reason (n=94), Died (n=62), Other reasons (n=27), Ineligible after enrollment (n=15), Patient elected non-transplant treatment (n=11)

Table 1.

Participant characteristics, for evaluable patients in the groups Very Likely and Very Unlikely to find a matched unrelated donor.

Characteristic	Very likely	Very unlikely	P Value
Number of patients	958	276
Number of centers	43	40
Sex, n (%)			0.85
Male	539 (56.3)	157 (56.9)
Female	419 (43.7)	119 (43.1)
Age at evaluability, years - n (%)			<.01
Median (min-max)	61.2 (0.7–81.3)	56.4 (2.6–76.4)	<.01
<=10	17 (1.8)	8 (2.9)
11–20	21 (2.2)	13 (4.7)
21–30	61 (6.4)	25 (9.1)
31–40	61 (6.4)	28 (10.1)
41–50	99 (10.3)	32 (11.6)
51–60	190 (19.8)	56 (20.3)
61–70	351 (36.6)	84 (30.4)
>70	158 (16.5)	30 (10.9)
Performance score at evaluability, n (%)			0.18
>=90	416 (43.4)	138 (50.0)
80	330 (34.4)	91 (33.0)
<80	207 (21.6)	46 (16.7)
Missing	5 (0.5)	1 (0.4)
Race/Ethnicity, n (%)			<.01
Hispanic White	50 (5.2)	43 (15.6)
Non-Hispanic White	848 (88.5)	145 (52.5)
Black/African American	16 (1.7)	40 (14.5)
American Indian/Alaska Native	4 (0.4)	2 (0.7)
Asian/Native Hawaiian/Pacific Islander	18 (1.9)	28 (10.1)
Other/Multiple race	1 (0.1)	4 (1.4)
Missing	21 (2.2)	14 (5.1)
Disease, n (%)			<.01
Acute myeloid leukemia	496 (51.8)	130 (47.1)
Acute lymphoblastic leukemia	145 (15.1)	62 (22.5)
Myelodysplastic syndrome	246 (25.7)	55 (19.9)
Hodgkin lymphoma	6 (0.6)	1 (0.4)
Non-Hodgkin lymphoma	32 (3.3)	20 (7.2)
Severe aplastic anemia	25 (2.6)	5 (1.8)
Sickle cell disease	3 (0.3)	2 (0.7)
Other acute leukemia	5 (0.5)	1 (0.4)
Disease status, acute leukemia, n (%)			0.49
CR1/CR2	211 (32.7)	73 (37.8)
CR3+/Primary induction failure/Relapse	101 (15.6)	30 (15.5)
No Treatment/Currently undergoing induction treatment/Unknown	332 (51.4)	90 (46.6)
Missing	2 (0.3)	0 (0.0)
Disease status, myelodysplastic syndrome, n (%)			0.46
CR/HI	21 (8.5)	7 (12.7)
NR/SD/Prog from HI/Rel from CR	89 (36.2)	16 (29.1)
Not assessed	136 (55.3)	32 (58.2)
Disease status, lymphoma, n (%)			0.24
Treatment responsive	9 (23.7)	8 (38.1)
Not Treatment responsive	29 (76.3)	13 (61.9)
Reason for transplant, Sickle cell disease, n (%)			0.33
Acute chest syndrome	1 (33.3)	0 (0.0)
Recurrent vaso-occlusive pain	2 (66.7)	1 (50.0)
Stroke	0 (0.0)	1 (50.0)
Time from consent to evaluability, months - n (%)			0.73
Median (min-max)	0.1 (0.0–12.3)	0.1 (0.0–8.8)	0.78
within 1 month	842 (87.9)	237 (85.9)
1–2 months	81 (8.5)	25 (9.1)
3–6 months	32 (3.3)	13 (4.7)
>6 months	3 (0.3)	1 (0.4)

Abbreviations: CR, complete response; HI, hematologic improvement; NR, no response; SD, stable disease

Survival from the time of evaluability in the Very Likely and Very Unlikely groups

The unadjusted survival at 1- and 2-years from evaluability was 69% (95% confidence interval [CI] 63–74%) and 56% (95% CI 50–62%) for the Very Unlikely group and 69% (95% CI 66–72%) and 55% (95% CI 52–58%) for the Very Likely group, respectively (HR 1.00, 95% CI 0.82–1.21, p=0.98, Supplementary Figure 2). There was no significant difference in the risk of death (HR = 1.07 for Very Unlikely vs. Very Likely, 95% CI 0.86–1.33, p=0.56, Table 2) in the covariate adjusted Cox proportional hazards model. Age, KPS, disease status, and time from consent to evaluability were all significantly associated with overall survival in this multivariate model. Although the aggregate variable for race/ethnicity was not significant, the point estimates for Hispanic, Black and Other patients were numerically higher but also not significant, Table 2. Figure 2 shows the adjusted survival curves by donor search prognosis. There were no significant interactions with the main effect, and the center effect was not significant (p=0.688). There was no significant effect of COVID-19 era on OS (p=0.657).

Table 2.

Multivariate results for overall survival, adjusted for a priori selected clinical characteristics, for the Very Likely and Very Unlikely groups

Variable	N/events	Hazard Ratio	95% CI	p-value
Donor Search Prognosis				0.555
Very likely	958/454	1.00
Very unlikely	276/128	1.07	(0.86–1.33)	0.555
Age (decades) (continuous)		1.17	(1.09–1.25)	<0.001
Karnofsky Performance Status				<0.001
90–100	554/221	1.00
80	421/210	1.31	(1.08–1.59)	0.006
< 80	253/148	1.83	(1.47–2.27)	<0.001
Unknown	6/3	2.39	(0.72–7.95)	0.670
Race/Ethnicity				0.670
White Non-Hispanic	993/474	1.00
Hispanic White	93/39	1.23	(0.87–1.74)	0.237
Black or African-American	56/28	1.33	(0.88–2.01)	0.172
Asian/Pacific Islander	46/18	0.98	(0.60–1.60)	0.942
Other	11/6	1.25	(0.55–2.82)	0.596
Unknown	35/17	1.07	(0.65–1.75)	0.790
Sex				0.300
Female	538/244	1.00
Male	696/338	1.09	(0.92–1.29)	0.300
Interval from consent to evaluability				0.038
0 to 1 month	1079/479	1.00
1 to 2 months	106/57	1.34	(1.01–1.77)	0.043
3+ months	49/28	1.40	(0.95–2.06)	0.087
Disease				0.098
Acute myeloid leukemia	626/298	1.00
Acute lymphoblastic leukemia/Other acute leukemia	213/73	0.75	(0.57–0.98)	0.035
Myelodysplastic syndrome	301/178	1.03	(0.58–1.85)	0.913
Non-Hodgkin lymphoma/Hodgkin lymphoma	59/28	0.93	(0.39–2.22)	0.873
Non-malignant disease	35/5	0.40	(0.16–1.00)	0.049
Acute leukemia disease status				0.008
CR1/CR2	284/121	1.00
CR3+/Primary induction failure/Relapse	131/66	1.61	(1.19–2.19)	0.002
No treatment/Currently undergoing induction treatment	424/184	1.10	(0.88–1.39)	0.402
Myelodysplastic syndrome disease status				0.382
Complete response/Hematologic improvement	28/13	1.00
No response/Stable disease/Progression from hematologic improvement/Relapse from complete response	105/62	1.39	(0.76–2.56)	0.285
Not assessed	168/103	1.51	(0.84–2.71)	0.170
Lymphoma disease status				0.458
Not treatment responsive	42/22	1.00
Treatment responsive	17/6	0.70	(0.27–1.80)	0.458

Abbreviations: CR, complete response

Figure 2.

Overall survival by donor search prognosis for the Very Likely and Very Unlikely groups from the time of evaluability.

The survival two years from evaluability was 65% (95% CI 60–70%) for the Less Likely group (Figure 3, panel A and Supplementary Table 2).

Figure 3.

Post-transplant outcomes by donor search prognosis, from the time of transplantation, adjusted for clinical covariates (a) KM curve of survival; (b) KM curve of disease-free survival (DFS), limited to malignant diseases; (c) cumulative incidence of treatment-related mortality (TRM), limited to malignant diseases; (d) cumulative incidence of relapse, limited to malignant diseases; (e) cumulative incidence of grade II-IV GVHD; (f) cumulative incidence of grade III-IV GVHD; (g) cumulative incidence of any chronic GVHD. KM=Kaplan-Meier, GVHD= graft-vs-host disease

Outcomes after transplantation by donor search prognosis

A total of 1179 participants received a transplant, 70% in the Very Likely, 66% in the Less Likely and 62% in the Very Unlikely. Median time from evaluability to transplant did not differ between the groups (3.3 months vs. 3.4 months vs. 3.3 months, respectively, p=0.36). Demographics and baseline characteristics for all transplanted participants by donor search prognosis are displayed in Supplementary Table 1. Ninety-four percent of patients in the Very Likely group received 8/8 MUD grafts; 4% received grafts from haploidentical family members and the rest from other donor types. Patients in the Very Unlikely group received grafts from haploidentical relatives (60.2%), MMUD (22.8%), and cord blood (7.7%); only 9.4% received an 8/8 MUD. The characteristics of the Less Likely group were more similar to the Very Unlikely group in terms of age, race/ethnicity, and disease than the Very Likely group, except that 37.6% received an 8/8 MUD graft. Use of post-transplant cyclophosphamide as GVHD prophylaxis was higher in the Very Unlikely (68.4%) and Less Likely (52.6%) groups compared to the Very Likely group (25.3%), p<0.01. One transplanted patient with no follow-up is included in the demographics tables but not in any of the outcome analyses.

Overall Survival

Univariate results are shown in Supplementary Table 2. The unadjusted Kaplan-Meier 2-year survival from transplant was 64% (95% CI 60–67%) in the Very Likely group, 65% (95% CI 60–70%) in the Less Likely group, and 62% (95% CI 54–69%) in the Very Unlikely group, overall p-value 0.70. In the multivariate Cox model, there was no significant difference in the risk of death after HCT by donor search prognosis (overall p=0.66, Table 3). Older age, lower KPS, higher DRI, and higher HCT-CI were all significantly associated with worse overall survival in this multivariate model, using factors determined in the a priori covariate-adjusted model. There were no interactions between these variables and the donor search prognosis. Figure 3, Panel a shows the adjusted survival curves by donor search prognosis. There was no significant center effect (p=0.878) and there was no interaction identified between donor search prognosis and the other variables.

Table 3.

Multivariate analyses of transplant outcomes, adjusted for a priori selected clinical characteristics, by donor search prognosis

	Very Likely	Less Likely	Very Unlikely
Relapse	1.0 overall P=0.66	1.15 (0.87–1.52) P=0.32	1.40 (1.01–1.96) P=0.047
Treatment-related mortality	1.0 overall P=0.89	1.09 (0.77–1.54) P=0.64	1.06 (0.69–1.64) P=0.78
Disease-free survival	1.0 overall P=0.24	1.11 (0.90–1.38) P=0.33	1.25 (0.96–1.62) P=0.10
Overall survival	1.0 overall P=0.66	1.03 (0.82–1.31) P=0.78	1.15 (0.86–1.52) P=0.36
Grade II-IV acute GVHD	1.0 overall P=0.93	0.97 (0.77–1.22) P=0.80	0.95 (0.71–1.27) P=0.71
Grade III-IV acute GVHD	1.0 overall P=0.38	1.33 (0.89–1.98) P=0.16	1.13 (0.67–1.91) P=0.65
Any chronic GVHD	1.0 overall P=0.42	0.85 (0.66–1.09) P=0.19	0.92 (0.67–1.26) P=0.59

Abbreviation: GVHD, graft-versus-host disease

Relapse, Treatment-Related Mortality and Disease-free survival (Malignant disease patients only)

There were no differences in the univariate results for relapse, TRM and DFS, as shown in Supplementary Table 2. The unadjusted 2-year cumulative incidence of TRM was 16% (95% CI 13–19%) in the Very Likely group, 18% (95% CI 14–22%) in the Less Likely group and 18% (95% CI 12–24%) in the Very Unlikely group, overall p=0.64. In the multivariate Cox models, there was no significant difference in the risk of relapse, TRM or DFS (Table 3, Figure 3, Panels b-d) by donor search prognosis, based on overall P values.

Acute and Chronic GVHD

No differences were seen in the unadjusted rates of GVHD, as shown in Supplementary Table 2. The cumulative incidences of grade III-IV acute GVHD by six months in the Very Likely, Less Likely, and Very Unlikely groups were 11%, 12% and 11%, respectively; any chronic GVHD by two years was 39%, 33% and 33%, respectively. In the multivariate cause specific Cox models, there was no significant difference in the risk of grade II-IV or grade III-IV acute GVHD or any chronic GVHD by donor search prognosis (Table 3, Figure 3, panels e-g)

Discussion

We conducted a large prospective multicenter biologic assignment study using the donor search strategy described to compare patients who are Very Likely or Very Unlikely to find an 8/8 MUD, where those Very Unlikely to have a MUD were supposed to proceed directly to an alternative donor search without a prolonged search for a MUD. Using this approach, we did not see a difference in survival from the time of evaluability. In addition, there were no differences between donor search prognosis groups in HCT outcomes including survival from the time of transplant, relapse, TRM, DFS, acute GVHD II-IV and III-IV, or any chronic GVHD, even though transplanted patients in the Very Likely group had 94% MUD grafts compared with only 9% MUD in the Very Unlikely group. Our results show that survival and transplant outcomes were not different using an available alternative donor when a MUD is not likely to be found. Multivariate analyses confirmed that patient and disease factors were more strongly associated with these endpoints than donor search prognosis, providing evidence that patient-intrinsic characteristics rather than donor or graft factors are the primary determinants of transplant success.

In a previous analysis, we did not find in multivariate analysis that patients who were in the Very Unlikely group, who were more likely to receive MMUD, haploidentical related or cord blood grafts, were more or less likely to be transplanted than those in the Very Likely group,²⁹ consistent with our finding that overall survival from the time of evaluability did not differ. This may be due to the protocol-specified donor search strategy that emphasized using the best donor available whenever a patient in the Very Unlikely group was ready for transplant rather than continuing to search for a MUD. These results could also be explained by increasing center comfort with performing transplants from HLA-mismatched donors, irrespective of the donor search prognosis score. We still found a center preference for MUD though, since over 90% of patients who were Very Likely to find such a donor ended up using one. Our results are especially encouraging for patients from genetically underrepresented racial and ethnic groups and mixed-race individuals as well as those with uncommon haplotypes who are disproportionally in the Very Unlikely group to find a MUD, despite the more than 40 million people in the worldwide registries.

Standard practice in the U.S. has been to prioritize HLA-identical sibling donors over MUDs over haploidentical relatives, MMUD or cord blood. In a survey of physicians and search coordinators conducted in 2018, 48% would allow 2 or more months of additional search time to try to identify a MUD in non-urgent cases.³⁰ Prior studies showed that patients unlikely to find a MUD, disproportionately from groups other than non-Hispanic Whites, were much less likely to proceed to transplantation^19,20 or were delayed in undergoing transplantation.²¹ Although multiple biases limit interpretation, one study showed that lack of transplantation once a search is started is associated with shorter survival.⁸ Thus, achieving a transplant rate, time to transplant and survival from evaluability for the Very Unlikely group that is not statistically different than the Very Likely group represents an improvement in access and outcomes.

Most centers would still choose an HLA-matched donor if one is available although outcomes with HLA-mismatched donors have been improving with innovations such as post-transplant cyclophosphamide,^31,32 in vivo T cell depletion,^33,34 graft manipulation,^35,36 and adding agents to the historical standard of a calcineurin inhibitor and methotrexate combination.^37–39 Some studies suggest HILA-mismatched donors can provide comparable outcomes to HLA-matched donors.^12,40 Other studies show more complications with HLA-mismatched donors and lower survival.^31,41,42 More studies comparing alternative donor types would be helpful since many patients will have a choice of alternative donors. While our results are encouraging for patients in need of a transplant who are Very Unlikely to find a MUD, overall survival was still between 60–70% for all groups at two years from evaluability, highlighting the need for more research to address remaining challenges.

Limitations of our study include changes in transplantation techniques during the course of the study with increasing use of MMUD and post-transplant cyclophosphamide for GVHD prophylaxis. Models were not adjusted for GVHD prophylaxis nor donor type. A detailed analysis of transplant outcomes according to actual donor used will be reported in a subsequent paper. In addition, the COVID-19 pandemic occurred during the study period and temporarily altered criteria for transplantation candidacy and cryopreservation of the stem cell product,⁴³ which may have affected the Very Likely and Very Unlikely groups differently, although we checked for and did not find a COVID-19 effect on our results. Also, despite attention throughout the study to the enrollment of under-represented racial and ethnic groups and even extending enrollment at sites with more diverse populations, the final evaluable population was still 73.3% non-Hispanic White. Race and ethnicity were not found to be statistically associated with survival in the multivariable analysis despite a numerically higher hazard ratio for Hispanic and Black patients.

In conclusion, survival and all other major transplant outcomes did not differ between patients Very Likely and those Very Unlikely to find an 8/8 MUD, supporting the donor search strategy used in this study. Our data support that patients Very Unlikely to have an 8/8 MUD should be directed to alternative donors without a prolonged search for a MUD.

Supplementary Material

PV Appendix Figure 1

Supplementary Figure 1. Study schema.

Figure footnote: Abbreviations: ITT, intention-to-treat; HLA, human leukocyte antigen; MUD, matched unrelated donor; URD, unrelated donor; HCT, hematopoietic cell transplantation; Haplo, haploidentical related donor; CBT, cord blood transplant; MMUD, mismatched unrelated donor

PV Appendix Figure 2

Supplementary Figure 2. Unadjusted and adjusted overall survival by donor search prognosis for the Very Likely and Very Unlikely groups from the time of evaluability.

Context.

Key Objective

Does a donor search strategy based on likelihood of a patient finding a matched unrelated donor (MUD) result in different survival and transplant outcomes for those very likely or very unlikely to find a MUD?

Knowledge Generated

Survival and transplant outcomes were not different for those very likely or very unlikely to find a MUD, even though 94% of transplanted patients in the very likely group received a transplant from a MUD compared to 9% in the very unlikely group. The rest of the very unlikely group received haploidentical related (60%), mismatched unrelated (23%) or cord blood (8%) transplants.

Relevance (written by Charles Craddock):

In this prospective study, early implementation of an innovative donor search strategy was shown to facilitate rapid access to an alternative donor transplant for patients whose probability of identifying a well matched unrelated donor can be predicted to be very low. Refining the donor search algorithm has the potential to increase access to transplant for patients whose likelihood of identifying a well matched donor is low and addresses long-standing issues of inequity for patients from minority ethnic groups.