Race and Survival in Unrelated Hematopoietic-cell Transplantation

Abstract

Background:

Survival after hematopoietic-cell transplantation depends on race/ethnicity and histocompatibility (HLA) between the patient and transplant donor. HLA sequence variation is a genetic construct of continental populations, but its role in accounting for racial disparities of transplant outcome is unknown.

Objective:

To determine disparities in transplant survivorship among patients of diverse race while accounting for patient and donor HLA variation.

Study Design:

A total of 26945 self-described Japanese, US Asian, White, Hispanic, and Black patients received an unrelated donor transplant for the treatment of a life-threatening blood disorder. The risk of mortality with and without adjustment for known HLA risk factors (number and location of donor mismatches; patient HLA-B leader genotype and HLA-DRβ peptide-binding motif) was studied using multivariable models. Survival after HLA-matched transplantation for patients with optimal leader and peptide-binding features was estimated for each race, as was the improvement in survival over calendar-year time by considering year of transplantation as a continuous linear variable.

Results:

The number, location and nature of donor HLA mismatches and the frequency of patient HLA-B and HLA-DRB1 sequence motifs differed by race. Japanese patients had superior survival compared to other races without consideration of HLA. Following HLA adjustment, three mortality risk strata were identified: Japanese and US Asian (low-risk); White and Hispanic (intermediate-risk), and Black patients (high-risk). Survival for patients with optimal donor and HLA characteristics was superior for Japanese, intermediate for US Asian, White and Hispanic, and lowest for Black patients. Five-year increments of transplant year were associated with greater decreases in mortality hazards for Black and Hispanic patients than for Japanese, US Asian and White patients.

Conclusion:

Transplant survivorship disparities are influenced by HLA as a genetic construct of race. A more complete understanding of the factors that influence transplant outcomes provides opportunities to narrow disparities for future patients.

Graphical Abstract

INTRODUCTION

Disparities in health outcomes across racially diverse populations are recognized in many disciplines in medicine and have been brought sharply into focus by the current COVID-19 pandemic.^1,2 Survivorship disparities are accentuated in hematopoietic-cell transplantation because the patient and transplant donor each contribute genetic and non-genetic factors that influence clinical outcome.³ In Western countries, patients of non-White background have higher risks of mortality and transplant-associated complications including graft-versus-host disease (GVHD) than White patients after related or unrelated donor transplantation.^4–9

Genetic differences across human populations have been extensively catalogued particularly for the loci that govern histocompatibility known as HLA genes (“Humaine-1 Locus A”).^10–12 HLA genes have evolved over the span of human evolution to meet their primary function in host defense against foreign antigens^13,14 and display striking population-specific patterns of variation.¹¹ Indeed, HLA polymorphism and race/ethnicity are inextricably linked. The extent and nature of HLA-A, -B, -C, -DRB1, -DQB1, -DPB1 allelic variation is descriptive of continental populations and in this way, HLA is a genetic construct of race/ethnicity. The very HLA diversity that is advantageous for host defense poses major challenges in transplantation. HLA governs graft acceptance by the patient and graft-versus-host recognition by the donor, and therefore precise matching of donors and limiting the number of mismatches is undertaken to lower the risks of graft rejection and GVHD.¹⁵ The likelihood of identifying HLA-matched donors correlates strongly with patient race/ethnicity and is disproportionately higher for some populations than others.¹⁶ International efforts to increase the size and racial diversity of donor registries have greatly improved the odds for non-White patients, but coverage of the full spectrum of HLA variation remains a challenge.¹⁷

The second consequence of ancestry-specific HLA variation pertains to the alleles and haplotypes of the patient’s germline. Two loci in particular encode polymorphisms that influence mortality: a methionine (M)/threonine (T) dimorphism of the HLA-B leader peptide,^15,18 and a phenylalanine (F) - glutamate (E) - tyrosine (Y) (“FEY”) amino acid motif located at residues 26–28-30 of the HLA-DRβ peptide-binding groove.¹⁹ Patients with TT leader genotype have been shown to have higher survival than patients with the MM leader genotype after HLA-B-matched and -mismatched transplantation.^15,18 In one study, MHC-wide single nucleotide polymorphism mapping identified the presence of FEY-positive HLA-DR molecules in patients as a risk factor for mortality.¹⁹ The frequencies of the leader and FEY motif vary across different races/ethnicities.^19,20

Despite ample evidence that donor HLA mismatching, HLA-DRβ FEY, and HLA-B leader each function as transplantation determinants and that their frequencies vary across diverse human populations, their full potential as genetic constructs of race has not been used to study racial disparities in transplantation survivorship. We leveraged the ancestry-informative nature of HLA to better understand differences in survival for self-described African, Asian, Hispanic, and White patients after unrelated donor hematopoietic-cell transplantation. Supplementation of self-defined race/ethnicity with HLA provided information on ancestry and outcomes disparities in a way that self-defined race alone could not.

MATERIALS AND METHODS

Study Design and Population

A total of 26945 patients received an unrelated donor transplant between 1988–2016 in centers in North America and Japan for the treatment of a life-threatening blood disorder for whom HLA typing, self-described race, and clinical data were contributed to the International Histocompatibility Working Group (IHWG) in Hematopoietic-cell Transplantation (Table S1). All transplantations had complete HLA typing and clinical data (Table S2). There were no exclusions based on patient or donor characteristics. Informed consent was obtained from patients and donors in accordance with the Declaration of Helsinki in each registry or institution, and consent of participation with the IHWG was obtained from each participating registry or institution before registration of the data. Protocols were approved by the institutional review boards of the National Institutes of Health Office for Human Research Protections, each participating IHWG center, the Fred Hutchinson Cancer Research Center, the Center for Blood and Marrow Transplant Research, and the Japanese Data Center for Hematopoietic Cell Transplantation.

Statistical Analysis

Self-described race for US transplants was available from registry data. Broad race/ethnicity groups were used to describe transplantations prior to 2014;^6,21 since 2014 the broad Hispanic group was defined as White plus Hispanic ethnicity. No further delineation based on residence, language, or other social constructs were made.²²

US transplants were compared to 8575 Japanese patients registered by the Japan Marrow Donor Program. HLA-A, -B, -C, -DRB1, -DQB1 and -DPB1 alleles were typed and allele ambiguity groups, HLA match grades (HLA-10/10; HLA-9/10; HLA-8/10 or less), allele/antigen mismatches, HLA-B leader genotype and HLA-DRB1 FEY motif were defined as previously described.^15,18,19,23

The goal of the current analysis was to estimate survival disparities among patients of different self-defined race after accounting for patient and donor HLA features that are known to influence mortality. The two major risks affecting survival, clinically severe grades III-IV acute GVHD and relapse, were also evaluated. Adjusted estimates of overall survival estimates were calculated as described.²⁴ Cox proportional-hazards models were used to assess the relationship between race and the risk of overall mortality and relapse. Death without relapse was censored in the models that fit the cause-specific hazard of relapse. Logistic regression was used to assess the relationship between race and the probability of grades III-IV acute GVHD. A re-assessment of non-HLA variables on transplant outcome was not the goal of the current analysis, and all regression models were adjusted for year of transplant, conditioning intensity, disease severity, patient cytomegalovirus serostatus, patient age, donor age, source of stem cells, patient sex, donor sex, use of total-body irradiation for conditioning, and use of T-cell depletion. When HLA-DP match status is included in models, the relative risks across the racial groups do not qualitatively change; furthermore, HLA-DPB1 matching was not available for approximately 20% of the pairs, and for these reasons we have not included HLA-DP in the models. Missing data for adjusted variables were included by creating a separate category for missing. Two-sided P values from regression models were obtained from the Wald test. The population with the largest number of patients (ie., White) was used as the reference group. Statistical interactions were assessed by including appropriate terms in the regression model. No adjustments were made for multiple comparisons, but single models with more than two categories for the main effect were subjected to a global test. If the global test was not statistically significant, pairwise comparisons were not made. All analyses were conducted using R version 3.6.3.

RESULTS

Mortality after transplantation was strongly influenced by both race and HLA factors. When HLA characteristics are not considered, Japanese patients had superior survival while Hispanic and Black patients each had inferior survival relative to White patients; the outcomes of US Asians and Whites were similar (Table 1, Figure 1A). A similar rank-order of risks of grades III-IV acute GVHD and relapse were observed across patient race and HLA match grades (Table S3).

Table 1.

Mortality after Unrelated Donor Hematopoietic-cell Transplantation

Patient race	Unadjusted for HLA		Adjusted for HLA		“Ideal” transplant*
	Number	Hazard ratio (95% CI; P value)	Number	Hazard ratio (95% CI; P value)	Number	Hazard ratio (95% CI; P value)
White	15080	1.0	13254	1.0	2413	1.0
Japanese	8575	0.69 (0.65– 0.73; <0.0001)	5499	0.69 (0.65– 0.74; <0.0001)	2579	0.64 (0.57– 0.73; <0.0001)
US Asian	563	0.92 (0.81– 1.03; 0.16)	469	0.86 (0.75– 0.99; 0.03)	149	0.93 (0.73– 1.19; 0.58)
Hispanic	1422	1.22 (1.14– 1.31; <0.0001)	1106	1.08 (1.00– 1.18; 0.07)	129	1.01 (0.80– 1.28; 0.92)
Black	1305	1.38 (1.29– 1.49; <0.0001)	978	1.19 (1.08– 1.30; 0.0002)	122	1.76 (1.39– 2.23; <0.0001)

^*Ideal transplant is defined by HLA-10/10 donor matching, patient TT leader genotype and patient FEY-negative HLA-DRB1 allele(s).

Figure 1.

Probability of survival for White, Japanese, US Asian, Hispanic and Black patients adjusted for factors listed in biostatistical methods. A. No adjustment for HLA. B. Adjustment for HLA. C. The ideal transplant.

Five key features of HLA are known risk factors for mortality: the total number of donor HLA mismatches; the location (locus) of the mismatch; mismatches between DNA-defined alleles versus serologically-distinct antigens; the HLA-B leader genotype (MM, MT, TT), and the presence of the HLA-DRB1 FEY motif in the peptide-binding groove.^15,18,19 Striking differences across the patient race groups were observed for each of these five features and illustrates their powerful ancestry-informative nature (Table 2; Table S4). The specific patient/donor mismatches in transplants mismatched for one allele or antigen (“HLA-9/10”) varied by patient race with virtually no sharing of mismatch combinations. The distribution of mismatched loci also differed across the patient race-groups. Whereas Japanese patients were more frequently mismatched at HLA-C, Hispanic and Black patients were more frequently mismatched at HLA-A; mismatches in US Asians were more similar to other US groups. In the current study, the frequency of FEY-positive patient HLA-DRB1 was higher among Black and Hispanic patients compared to White and Asian patients.

Table 2.

Key HLA Characteristics of the Study Population

Characteristic	Total number	White	Japanese	US Asian	Hispanic	Black
Match grade at HLA-A,B,C,DRB1,DQB1- n (%) *
HLA 10/10	11390	6767 (59;45)	3765 (33;44)	202 (2;36)	334 (3;23)	322 (3;25)
HLA 9/10	9916	6487 (65;43)	1734 (17;20)	267 (3;47)	772 (8;54)	656 (7;50)
HLA 8/10 or fewer	5639	1826 (32;12)	3076 (55;36)	94 (2;17)	316 (6;22)	327 (6;25)
Total n	26945	15080	8575	563	1422	1305
Mismatched gene (HLA-9/10) - n (%) *
HLA-A	2870	1757 (61;27)	351 (12;20)	104 (4;39)	310 (11;40)	348 (12;53)
HLA-B	1252	910 (73;14)	54 (4;3)	44 (4;16)	166 (13;22)	78 (6;12)
HLA-C	3218	1973 (61;30)	949 (29;55)	60 (2;22)	154 (5;20)	82 (3;13)
HLA-DRB1	693	468 (68;7)	118 (17;7)	7 (1;3)	59 (9;8)	41 (6;6)
HLA-DQB1	1883	1379 (73;21)	262 (14;15)	52 (3;19)	83 (4;11)	107 (6;16)
Total n	9916	6487	1734	267	772	656
Type of mismatch (HLA 9/10) - n (%) *
Allele	3040	1832 (60;28)	680 (22;39)	104 (3;39)	297 (10;38)	127 (4;19)
Antigen	6876	4655 (68;72)	1054 (15;61)	163 (2;61)	475 (7;62)	529 (8;81)
Total n	9916	6487	1734	267	772	656
Patient HLA-B leader genotype (HLA-9/10 - n (%) *
MM/MT	4300	3110 (72;48)	427 (10;25)	79 (2;30)	395 (9;51)	289 (7;44)
TT	5616	3377 (60;52)	1307 (23;75)	188 (3;70)	377 (7;49)	367 (7;56)
Total n	9916	6487	1734	267	772	656
Patient HLA-DRB1 FEY (HLA-9/10 - n (%) *
None	9618	6459 (67;100)	1665 (17;96)	264 (3;99)	675 (7;87)	555 (6;85)
Any	298	28 (9;<1)	69 (23;4)	3 (1;1)	97 (33;13)	101 (34;15)
Total n	9916	6487	1734	267	772	656

^*First percent number is row percent, second percent number is column percent.

For each patient race, the key HLA features correlated with outcome as previously observed.^15,19 Adjustment of the mortality models for the number and location of mismatches, leader genotype and HLA-DRB1 motif, led to a shift in the low-to-high rank order of mortality: Japanese and US Asian; White and Hispanic; Black patients (Table 1, Figure 1B). The percent difference between HLA adjusted and unadjusted hazards was largest for Black patients (14.3% decrease of mortality risk with adjustment for HLA) followed by Hispanic (11.2% decrease) and US Asians (6.2% decrease); the hazard ratios for Japanese versus White patients were similar with and without HLA adjustment. Taken together, these findings demonstrate that survival after unrelated donor transplantation is strongly influenced by both race and HLA features. Furthermore, the disparities in survival are evident between Asian (low mortality), White and Hispanic (intermediate risk), and Black (high risk) patients.

Allele and antigen mismatches differ for the number and nature of amino-acid differences within the peptide-binding region and immunogenicity.^15,18 Among HLA-9/10 transplants, Black patients had the lowest frequency of allele-mismatches (Table 2), but the low-to-high rank order of mortality for Japanese, US Asian, White, Hispanic, and Black patients was not altered by adjustment for allele versus antigen mismatches; the hazard ratios for Japanese, US Asians, Whites, Hispanics, Blacks without adjustment (0.70; 0.86; 1.0; 1.03; 1.14) were similar to hazard ratios with adjustment (0.71; 0.87; 1.0; 1.04; 1.14). Furthermore, the low-to-high rank order of mortality was preserved for transplants with one (HLA-9/10) and two or more (HLA-8/10) mismatches (Table S5).

As the specific functional HLA features depend on both the patient and donor, we tested the hypothesis that survival also depends on donor race. Within each US patient-race group, we tested the global hypothesis that mortality was the same across all donor races. These global tests yielded P = 0.60, P = 0.71, P = 0.36, and P = 0.14 for White, US Asian, Hispanic, and Black patients, respectively (all Japanese patients had a Japanese donor). There was insufficient evidence to conclude that donor race has a major impact on mortality. Hazard ratios for each US patient/donor race group are summarized in Table S6.

The analyses described above provide new information on survival of patients of different race and HLA features. To assess whether differences in survival are diminished or even abolished when the patient and donor have optimal HLA features, we evaluated survival in HLA-matched transplants for patients with low-risk TT HLA-B leader genotype and FEY-negative HLA-DRB1 alleles.^15,18,19 When all three characteristics are considered, the frequency of “ideal” characteristics varies by race: 30% (2579/8575) Japanese; 26% (149/563) US Asian; 16% (2413/15080) White; 9% (129/1422) Hispanic, and 9% (122/1305) Black patients. Relative to White patients, Japanese patients had superior survival, Blacks had inferior survival, and US Asian, Hispanic and White patients had similar intermediate survival (Table 1, Figure 1C). In summary, the largest disparity in survival among transplants with optimal HLA features is between Black patients and patients of other races.

Over the past two decades, the substantial growth of international donor registries together with new information on tolerable HLA mismatches has enabled more non-White patients to identify suitable donors.¹⁷ At the same time, major advances in clinical care have improved the overall success of transplantation. Whether patients of all races and ethnicities have benefited from these practice changes remains an important question in outcomes research. We hypothesized that improvements in survival with respect to calendar time would occur across all patient races and ethnicities and examined the slope of the hazard of mortality with respect to year across the various race groups (year of transplant modelled as a continuous linear variable). The risk of mortality associated with year of transplantation varied across patient-race groups but mortality decreased with time for each patient-race group (P < 0.0001 for Japanese, White, Hispanic and Black; P = 0.01 for US Asians). The accompanying decrease in the hazard of mortality for each 5-year increase in time was 0.84, 0.81, 0.87, 0.80, and 0.77 for Japanese, US Asian, White, Hispanic, and Black patients, respectively. Under the assumption of linearity, the rate of improvement in survival with respect to time was largest in Hispanics and Blacks, suggesting a narrowing of the racial disparities.

DISCUSSION

Survival after hematopoietic-cell transplantation is strongly influenced by patient and donor characteristics of which HLA and race/ethnicity play prominent roles.^4–9 The same HLA polymorphisms that define continental ancestries participate in the immune response and are critical to the success of hematopoietic-cell transplantation.^15,23,25 Yet, HLA has not been used as an ancestry-informative marker to more completely define health outcomes disparities in transplantation. The current study integrated patient and donor HLA features with self-identified race/ethnicity to examine survival of US and Japanese patients after unrelated donor hematopoietic-cell transplantation. We leveraged robust principles of HLA genetics, not as a re-validation of their role in transplantation but rather applied as ancestry-informative markers to understand disparities in survivorship across patients of different races. The combination of HLA features with self-identified race/ancestry provided a more complete delineation of genetic ancestry and enabled identification of populations at risk for survivorship disparities. The findings from the study inform clinical practice and future research in outcomes disparities.

The integration of patient and donor HLA genetics with race uncovered three strata of patients at low, intermediate, or high risk of mortality after transplantation: Japanese; US Asian, White, and Hispanic; Black patients. Although Japanese and US Asian patients had superior survival to other US populations, Japanese patients still had better outcomes compared to US Asians. There were inherent differences in the transplant procedure between Japanese and US patients including the exclusive use of bone marrow as the stem cell source in Japan during this period; however, among all recipients of marrow allografts, the relationship between survival and race remained qualitatively the same. The current study was not designed to tease out the potential underlying mechanisms for these differences, but patterns of donor HLA mismatching among US Asians were more similar to those of US White/Hispanic than to Japanese patients, potentially pointing to differences in donor selection strategies between the US and Japan. Extension of the current study to additional world populations with diverse HLA genetics will provide additional insight into racial differences in transplant survivorship.

The HLA alleles used to select transplant donors are also markers for long-range HLA haplotypes that could harbor additional variation of functional importance. The extensive genetic conservation of HLA haplotypes in Japanese patients and transplant donors is striking and in sharp contrast to the more diverse haplotypes observed in US patients and donors.^26,27 Future investigation into the pathways and mechanisms for FEY- and leader-associated risks will enrich understanding of the immunobiology of HLA in transplantation. Although the specific mechanisms leading to increased mortality with FEY-positive HLA-DRβ is not known at this time, it is intriguing to hypothesize that due to its putative impact on the bound peptide,¹⁹ survivorship disparities associated with FEY-positive HLA-DRβ may elicit differential T-cell recognition of the peptide-MHC complex in individuals of different race. Additional areas for future studies include the role of non-HLA genes within the polymorphic HLA haplotypes as a source of race-associated variation that could influence survival. Beyond the MHC, differences in other regions of the genome could be a source of variation contributing to survival differences across US continental populations.²⁸ In particular, population differences in cytokine genes ²⁹ may play a prominent role in the stress response of acute GVHD.³⁰ Of note, the current study sought to define transplant survivorship among patients who successfully identified an unrelated donor. Many patients of non-White background who lack suitable unrelated donors are successfully treated with haploidentical related donor or cord blood transplantation.¹⁶ The role of HLA ancestry-informative features in survival after haploidentical related donor and cord blood transplantation, and whether outcome disparities are narrowing over time for these stem cell sources, remain important questions for future investigation.

Consideration of the patient and donor’s HLA provides a genetic construct to self-described race/ethnicity, but important societal and environmental determinants of survival disparities remain. Although the patients in the current study had successfully identified suitable donors and had access to transplantation, we lacked data on socio-economic factors that could impact outcome. Socio-economic determinants of survivorship remain an important area of research for future studies. In the current analysis, the available data for patient/donor race permitted examination of race on the broadest level, but inter-racial differences within US Asian, Black and Hispanic populations could exist and remain an important source of variation within a given race. This has potential relevance to the geographic variation in genetic markers across the continental US and may shed light on differences within the major US populations²⁸ and the interplay of ancestry, geography and HLA variation in survivorship disparities. Furthermore, consideration of HLA together with race/ethnicity does not describe important social and life-style constructs including societal stress which influence health outcomes via effects on the immune system.³¹ Information on these and other societal constructs of race/ethnicity were not available for the current study populations and remain important questions for future investigation.

There are currently over 38 million registered unrelated donors worldwide.¹⁷ Efforts to increase the number of donors and breadth of HLA is an on-going collective goal of the international community to ensure that patients in need of a transplant have the best possible chances of identifying matched donors.³² Our data show that a better understanding of the gaps that remain when the “ideal” transplant setting is achieved can provide invaluable information to aid investigation of new factors that influence transplant survival. At the same time, donor matching, HLA-B TT leader genotype and FEY-negative HLA-DR peptide-binding motifs may reflect optimal characteristics for some populations more so than others, and introduces a new concept in HLA that what is ideal for one race may not necessarily fulfill the needs for another. The definition of an ideal transplant may require that genetic and non-genetic characteristics unique to a given race be taken into consideration. To the extent that donor race could be considered in prospective donor selection, our results did not show definitive differences across patient races but the numerical increases in mortality risk merit study in the future when a larger transplant experience is available. Finally, the current study population included patients who received a transplant from a matched or mismatched unrelated donor. The interplay of histocompatibility, race and survival of patients undergoing transplantation from other allogeneic stem cell sources and with different transplant conditioning and immunosuppressive regimens including unrelated donor and haploidentical related donor transplantation utilizing post-transplant cyclophosphamide and cord blood unit transplantation, is warranted.^33,34

The findings from the current study provide a starting point to narrow the gaps in racial disparity in transplantation. Survival is improving for all patients of all races and disparities among US patients are becoming less pronounced. At the same time, a paradigm for outcomes research in racial disparities is envisioned to include the HLA region for its unique contribution of ancestry-informative functional variation.

Highlights.

• Transplant survivorship disparities are influenced by HLA as a genetic construct of race.

• Survival is improving over time for patients of all races.

Data Availability

Request for data may be made to the corresponding authors. Data sharing for research purposes may have partial restrictions consistent with the informed consent of study participants from whom the data were obtained.