Abstract
Disparities in survival after allogeneic hematopoietic cell transplantation have been reported for some race and ethnic groups despite comparable HLA matching. Individuals’ ethnic and race groups, as reported through self-identification, can change over time due to multiple sociological factors. We studied the effect of two measures of genetic similarity in 1,378 recipients who underwent myeloablative first allogeneic hematopoietic cell transplantation between 1995 and 2011 and their unrelated 10-of-10 HLA-A, -B, -C, -DRB1 and DQB1 matched donors. The studied factors were: i) Donor and recipient genetic ancestral admixture, and ii) Pairwise donor/recipient genetic distance. Increased African genetic admixture for either transplant recipients or donors was associated with increased risk of overall mortality (HR=2.26, p=0.005 and HR=3.09, p=0.0002 respectively), Transplant Related Mortality (HR=3.3, p=0.0003 and HR=3.86, p=0.0001 respectively) and decreased Disease Free Survival (HR=1.9, p=0.02 and HR=2.46, p=0.002 respectively). The observed effect, albeit statistically significant, was relevant to small subset of the studied population and was notably correlated with self-reported African-American race. We were not able to control for other non-genetic factors such as access to healthcare or other socio-economic factors, however the results suggest the influence of a genetic driver. Our findings confirm what has been previously reported for African-American recipients and show similar results for donors. No significant association was found with donor/recipient genetic distance.
Keywords: Hematopoietic Stem Cell Transplantation, African-American race, Genetic admixture, Genetic ancestry, Race and ethnicity, Ancestry Informative Markers
Introduction
The search for an HLA matched unrelated donor typically yields individuals with the same self-identified race/ethnicity (SIRE) as the recipient, mainly due to the increased likelihood of finding an HLA match within the same SIRE group. However, it is not clear whether a donor of matched race/ancestral composition as the recipient specifically results in better survival. Additionally, some ethnic groups are defined at a very general level: for example, the majority of hematopoietic cell transplantations (HCTs) in the US are performed with recipients and donors with European ancestry mostly classified as White or Caucasian. Previous studies have addressed racial disparities in outcome for HLA-matched related and unrelated donor HCT, albeit with more focus on the recipient rather than donor ethnic background1–8. However, there are very few studies to date analyzing outcome with respect to genetic admixture or differences in ancestral groups between a donor and recipient as defined genetically.
Multiple issues and inaccuracies relate to race constructs and how these data are reported9,10. There is more genetic variation within than between race groups, and race does not reliably demarcate populations with discrete genetic characteristics9. In research studies, data on race may be collected by self-report, direct observation, proxy report (especially in the case of recipients) or extraction from records. Very frequently in the United States race can serve as a surrogate for socioeconomic, education and health insurance status9. Disparities by race exist in three areas related to HCT: donor availability, access to HCT and outcomes of HCT9. Self-identified African-American recipients are usually disadvantaged in all three areas. For the above reasons, this study examined the genetic ancestral characteristics as defined by Ancestry Informative Markers (AIMs) to avoid the confounding factors caused by the use of race. We hypothesized that recipient and/or donor genetic admixture was associated with HCT outcome.
Ancestry Informative Markers (AIMs) are a set of genetic markers that differ in allele frequencies across different populations, whether within or across world continents. Most variation is shared among populations, so for most AIMs the most common allele is the same in each population. These Single Nucleotide Polymorphism (SNP) markers may be used to categorize individuals into populations sharing similar allele frequency distributions and perhaps phenotypes (such as SIRE). While an individual’s genetic composition does not change during their lifetime, SIRE is a result of self-perception and can change over time11,12. This can introduce inconsistency in the process of matching stem cell recipients with unrelated donors and can potentially affect the outcome of HCT.
When considering HCT outcome influences beyond full HLA match status, other factors have been investigated, such as minor histocompatibility antigens (MiHAs), especially due to their role in the dynamics of graft versus host disease (GVHD) and graft versus tumor effect13,14. Mismatches in MiHAs between donors and recipients may induce GVHD15. These mismatches could be due to donor/recipient ethnic mismatch16. Driven by this concept, and acknowledging the limitations in self-reported race information, a prior study has examined whether race and ethnic match status between recipient and donor is associated with outcomes; however, no significant associations were reported17.
In this study, we estimated the overall genetic admixture of transplant donors and recipients by genotyping a set of selected AIMs18, and then tested whether genetic admixture was associated with HCT outcomes.
Study Population and Methods
Study Population
Data were obtained from the Center for International Blood and Marrow Transplant Research (CIBMTR). Recipients received a first HLA-A, B, C, DRB1 and DQB1 (10-of-10) allele matched unrelated donor myeloablative transplantation for hematological malignancy (AML, ALL, CML and MDS) (Table 1). All transplantations were performed between 1995 and 2011. All donors and recipients provided informed consent to participate in the CIBMTR research program and the study was approved by the Institutional Review Board of the National Marrow Donor Program.
Table 1.
Characteristics of myeloablative first transplants for patients with samples available for donor/recipient pairs and with disease of AML, ALL, CML or MDS and are 10/10 high-resolution matched using unrelated NMDP donors from 1995 to 2011
| Variable | N (%) |
|---|---|
| Number of patients | 1378 |
| Number of centers | 146 |
| Age, median (range), years | 39 (<1–70) |
| Age at transplant | |
| <20 years old | 212 (15) |
| 20–59 years old | 1108 (79) |
| 60 years old and older | 58 (4) |
| Recipient race | |
| Caucasian | 1277 (93) |
| African-American | 29 (2) |
| Asian/Pacific Islander | 28 (2) |
| Hispanic | 2 (<1) |
| Native American | 6 (<1) |
| Other/multiple/declined/unknown | 36 (2) |
| Male sex | 767 (56) |
| Karnofsky prior to transplant > 90 | 931 (68) |
| Disease at transplant | |
| AML | 461 (33) |
| ALL | 216 (16) |
| CML | 436 (32) |
| MDS | 265 (19) |
| Disease status at transplant | |
| Early | 989 (72) |
| Intermediate | 88 (6) |
| Advanced | 283 (21) |
| Other | 18 (1) |
| Graft type | |
| Bone marrow | 777 (56) |
| Peripheral blood | 601 (44) |
| ATG given | |
| No | 1033 (75) |
| Yes | 334 (24) |
| In vivo T-cell depletion | |
| No | 907 (66) |
| Yes | 350 (25) |
| Donor age | |
| 18–29 years old | 558 (41) |
| 30–49 years old | 739 (54) |
| 50 years old and older | 81 (6) |
| Donor/recipient CMV match | |
| Negative/negative | 486 (35) |
| Negative/positive | 420 (30) |
| Positive/negative | 180 (13) |
| Positive/positive | 274 (20) |
| Unknown | 18 (1) |
| Donor age, median (range), years | 33 (18–59) |
| GVHD prophylaxis | |
| Ex-vivo T-cell depletion | 113 (8) |
| CD34 Selection | 10 (1) |
| Cyclophosphamide | 12 (1) |
| TACROLIMUS + (MTX or MMF) ± other | 683 (50) |
| TACROLIMUS ± other | 51 (4) |
| TACROLIMUS alone | 10 (1) |
| CsA + (MMF or MTX) ± other | 462 (34) |
| CsA ± other (no MTX nor MMF) | 11 (1) |
| CsA alone | 6 (<1) |
| Other | 20 (1) |
| HLA-DPB1 typing | |
| Double mismatch | 271 (26) |
| Single mismatch | 589 (57) |
| Matched | 176 (17) |
| Missing/not typed | 342 |
| Donor/recipient sex match | |
| Male/male | 569 (41) |
| Male/female | 383 (28) |
| Female/male | 198 (14) |
| Female/female | 228 (17) |
| Median follow-up of survivors, month (range) | 91 (11–199) |
HLA and SNP Genotyping
HLA genotyping for all donors and recipients was completed at allele-level resolution for HLA-A, -B, -C, DRB1 and -DQB1 as previously reported19. SNP genotyping was performed at two different phases of the study over a period of approximately 18 months. In the initial pilot phase 300 donor/recipient pairs were genotyped for the chosen SNP AIMs panel18 using the Sequenom iPLEX platform20 complemented by Thermo Fisher’s Taqman assays. In the second phase, extended genotyping on the Illumina HumanOmniExpress BeadChip became feasible, providing over 719,000 SNPs. However, to maintain the integrity of the initial study protocol, we sustained work with the 500 AIMs panel18, 304 of which were genotyped on the Illumina array and the remaining 196 were imputed using the 1000 Genomes reference populations21 and the IMPUTE 222 software package. More details on SNP genotyping and data quality control are available in supplementary material.
Statistical methods
Genetic Admixture
Genetic admixtures were examined using the software Structure v2.3.423. Twenty Structure runs were performed without any prior population assignment using K=4 clusters, with 10,000 replicates and 100,000 burnin cycles using the 1000 Genomes dataset21 as reference parental populations. We chose K=4 clusters to model four principal admixtures in our cohort: African (AFR), European (EUR), South European/Amerindian (SE/A) and Asian (ASI). It was difficult to separate the South European (SE) and Amerindian admixtures, which are the major admixture components of most individuals identifying as Hispanic (HIS). As a result, SE/A was dealt with as a single genetic constituent forming the majority admixture in HIS individuals as reported in Results.
The optimum run chosen for the Structure analysis had the minimum log probability of the data. Table 2(a) and (b) show the admixture distributions in each of the SIRE groups in recipients and donors respectively. Shaded cells correspond to genetic admixtures that represent the majority component in the respective SIRE groups: for example, on average the EUR genetic admixture would represent a majority genetic admixture in an individual that self-identified as European Caucasian (CAU). To display the Structure output we used the software CLUMPP24 augmented as CLUMPAK25.
Table 2.
Median and interquartile range (IQR) for principal genetic admixture components in self-reported (A) recipient and (B) donor race groups. Shaded cells correspond to genetic admixtures that represent the majority component in the respective self-identified race group. African (AFR), European (EUR), South European/Amerindian (SE/A), Asian (ASI), African-American (AFA), European Caucasian (CAU), Hispanic (HIS), Asian or Pacific Islander (API)
| Table 2A: | ||||||||
|---|---|---|---|---|---|---|---|---|
| Self-identified Race | Genetic Admixture | |||||||
| AFR | EUR | SE/A | ASI | |||||
| Median | IQR | Median | IQR | Median | IQR | Median | IQR | |
| AFA N=29 | 0.81 | 0.7, 0.9 | 0.08 | 0.03, 0.1 | 0.06 | 0.02, 0.1 | 0.005 | 0.003, 0.008 |
| CAU N=1113 | 0.004 | 0.002, 0.007 | 0.97 | 0.9, 0.98 | 0.015 | 0.007, 0.05 | 0.004 | 0.002, 0.007 |
| HIS N= 63 | 0.005 | 0.003, 0.02 | 0.03 | 0.01, 0.14 | 0.764 | 0.5, 0.9 | 0.08 | 0.015, 0.2 |
| API N= 20 | 0.005 | 0.002, 0.01 | 0.005 | 0.001, 0.01 | 0.008 | 0.002, 0.2 | 0.97 | 0.7, 0.99 |
| Table 2B: | ||||||||
|---|---|---|---|---|---|---|---|---|
| Self-identified Race | Genetic Admixture | |||||||
| AFR | EUR | SE/A | ASI | |||||
| Median | IQR | Median | IQR | Median | IQR | Median | IQR | |
| AFA N= 21 | 0.82 | 0.7, 0.9 | 0.1 | 0.03, 0.14 | 0.05 | 0.03, 0.06 | 0.004 | 0.002, 0.005 |
| CAU N= 1077 | 0.004 | 0.002, 0.007 | 0.97 | 0.9, 0.98 | 0.014 | 0.007, 0.05 | 0.004 | 0.002, 0.008 |
| HIS N= 60 | 0.006 | 0.003, 0.02 | 0.04 | 0.02, 0.48 | 0.7 | 0.3, 0.9 | 0.05 | 0.007, 0.2 |
| API N=22 | 0.003 | 0.002, 0.005 | 0.003 | 0.002, 0.03 | 0.002 | 0.002, 0.03 | 0.98 | 0.8, 0.99 |
Donor/Recipient Pairwise Genetic Distance
The donor/recipient genetic proximity was estimated by using the AIMs panel chosen for the study analysis18. Initially, we applied principal components analyses (PCA) for dimensionality reduction (Figure 1C). The advantage of this approach is that the obtained principal components maximize the variance in the original high-dimensional dataset while minimizing the total squared reconstruction error26. To calculate the donor/recipient genetic distance we needed an adequate count of eigenvectors (principal components). The challenge was to choose a count that would preserve the signal in the data without accepting too much spurious noise. With this goal in mind, we calculated the correlation between the pairwise genetic distance calculated from the study genotypes and the pairwise Euclidean distances calculated from different counts of eigenvectors generated from the PCA (Supplemental figure 2A). This was followed by taking the first derivative (Supplemental figure 2B). The goal was to roughly capture the point of high convexity in the correlation plot which corresponds to a minimum first derivative (since the derivative never reaches an absolute zero value but fluctuates around it). The number of eigenvectors at this minimal derivative is meant to minimize the signal-to noise ratio in our data and was used to calculate the donor/recipient pairwise genetic distance.
Figure 1.
A. Population clusters in the study cohort as output by the software Structure. Study populations are listed first (AFA, API, CAU, HIS, NAM, OTH, MLT, DEC, UNK) followed by the reference 1000 Genomes populations (GBR through TSI). Each vertical line represents an individual’s combined admixture while each color represents a genetic admixture constituent (i.e. genetic cluster). Population labels refer to self-identification. Orange, blue, green and purple clusters represent EUR, AFR, ASI and SE/A genetic admixture respectively. Notice the similarity in genetic admixture of some self-identified CAU and HIS individuals. B. Same population clusters as in (A) but displaying only study populations. C. Genetic clusters of study subjects via PCA, color coded by reported SIRE. The abscissa and ordinate show the first and second principal components respectively. Groups that self-identify similarly roughly cluster together. Some groups cluster more closely than others. For example, the study subset that self-identified as AFA are depicted by the red X’s. The genetic admixture of individuals with relatively high AFR admixture formed the top right cluster. Physical distance between different points implies genetic distance. PCA – principal components analysis, SIRE – self-identified race/ethnicity. Study populations: AFA-African-American, API-Asian/Pacific Islander, CAU-European Caucasian, HIS-Hispanic, NAM-Native American, MLT-Multiple ethnicity, DEC-Declined, OTH-Other, UNK-Unknown. 1000 Genomes populations: GBR-British in England and Scotland, FIN-Finnish in Finland, CHS-Southern Han Chinese, PUR-Puerto Ricans from Puerto Rico, CLM-Colombians from Medellin, Colombia, IBS-Iberian Population in Spain, CEU-Utah Residents (CEPH) with Northern and Western European Ancestry, YRI-Yoruba in Ibadan and Nigeria, CHB-Han Chinese in Beijing, JPT-Japanese in Tokyo, LWK-Luhya in Webuye, Kenya, MXL-Mexican Ancestry from Los Angeles, ASW-Americans of African Ancestry in SW USA, TSI- Tuscany in Italy. EUR-European genetic admixture, AFR-African genetic admixture, ASI-Asian genetic admixture, SE/A-South European/Amerindian genetic admixture.
Outcomes analysis
The studied outcomes were overall survival (OS), disease free survival (DFS), relapse, transplant-related mortality (TRM), grade II–IV and III–IV acute GVHD and chronic GVHD. We also evaluated neutrophil engraftment defined as achieving an absolute neutrophil count ≥ 0.5 × 109/L for three consecutive days.
The probabilities of overall survival and disease-free survival were calculated using the Kaplan-Meier estimator27. The probabilities of transplant-related mortality, GVHD, relapse and neutrophil engraftment were calculated using the cumulative incidence estimator28. Death was the competing risk for neutrophil engraftment and GVHD. Relapse was a competing risk for transplant-related mortality, and transplant-related mortality was a competing risk for relapse.
Cox proportional hazard models29 were applied for acute and chronic GVHD, non-relapse mortality, relapse and overall mortality. The clinical variables were tested for affirmation of the proportional hazards assumption. A stepwise forward model building procedure was used to select the adjusted covariates for each outcome with a threshold of 0.05 for both entry and retention in the model. Interactions between the main variable and the adjusted covariates were also tested. All p-values are two-sided. To adjust for multiple testing, p-values ≤ 0.01 were considered to be statistically significant. Analyses were performed using SAS 9.3 (SAS Institute, Cary, NC).
Results
Recipient characteristics
Table 1 describes recipients, disease and transplant characteristics. The median recipient age was 39 years with 85% of recipients 20 years or older at the time of transplant, and 56% of the recipients were males. The majority of recipients (93%) reported Caucasian (CAU) SIRE with the remaining 7% distributed as 2% African American (AFA), 2% Asian/Pacific Islander (API), <1% Hispanics (HIS), <1% Native American and 2% Multiple/declined/other/unspecified SIRE
Diagnoses included acute myeloid leukemia (AML, 33%), acute lymphoblastic leukemia (ALL, 16%), chronic myeloid leukemia (CML, 32%) and myelodysplastic syndrome (MDS, 19%). Graft type was bone marrow (56%) or peripheral blood (44%). The majority of recipients underwent transplantation with an early disease status (72%).
Genetic admixture
Figures 1A and B show the patterns of genetic admixture observed in the study cohort. Each individual (donor or recipient) is represented by a vertical line and each cluster is uniquely colored. Another clustering form is the PCA plot shown in figure 1C. Both cluster forms depict genetic proximity while the color coding in figure 1C corresponds to the SIRE of donors and recipients. In general, self-identified API and AFA individuals tend to genetically group well. However, the genetic clusters of self-identified CAU and HIS tend to overlap as in figure 1C. This is also portrayed in figure 1A and B as similarity in admixture among some CAU and HIS individuals along the population border.
Table 2(A) and (B) show the median and interquartile range (IQR) of the principal genetic admixtures in relation to recipient and donor SIRE. In general, self-reported CAUs have a dominant European (EUR) genetic admixture, while the dominant admixtures in self-reported API and AFA are Asian (ASI) and African (AFR) respectively. For individuals reporting HIS ethnicity, SE/A was dealt with as a single genetic constituent forming the majority admixture (Table 2).
The linearity of all estimated genetic admixtures was assessed using the Supremum Test30. For recipients, the EUR and SE/A admixtures did not pass the linearity test with p=0.04 and p<0.0001, respectively, suggesting the absence of linear functional relationship with outcome. In contrast, the AFR and ASI groups passed with p=0.363 and p=0.467, respectively, so were further analyzed. For donors, all admixture groups passed the linearity test with p>0.33.
Overall Survival
Multivariate analyses demonstrated that higher recipient and donor AFR genetic admixture were independently adversely associated with OS post-transplant, irrespective of self-identified race and ethnic status (HR = 2.26, 95% CI 1.28–3.96, p-value=0.005 and HR = 3.09, 95% CI 1.7–5.64, p-value=0.0002 respectively). Multivariate analyses showed no significant associations with any of the other studied genetic admixtures (Tables 2 and 3).
Table 3.
Multivariate analysis* evaluating the effects of recipient African (AFR) and Asian (ASI) genetic admixture as a continuous variable. Bold values indicate statistically significant results.
| Genetic Admixture | Outcome | Hazard Ratio | 95% Confidence Interval | p-value |
|---|---|---|---|---|
| AFR | OS | 2.26 | 1.28–3.96 | 0.005 |
| ASI | OS | 0.65 | 0.32–1.33 | 0.240 |
| AFR | DFS | 1.90 | 1.11–3.25 | 0.020 |
| ASI | DFS | 0.98 | 0.53–1.80 | 0.937 |
| AFR | TRM | 3.30 | 1.72–6.32 | 0.0003 |
| ASI | TRM | 1.18 | 0.50–2.78 | 0.705 |
| AFR | Relapse | 0.97 | 0.35–2.64 | 0.945 |
| ASI | Relapse | 0.86 | 0.37–2.00 | 0.722 |
| AFR | aGVHD II–IV | 1.21 | 0.62–2.35 | 0.574 |
| ASI | aGVHD II–IV | 1.09 | 0.58–2.04 | 0.793 |
| AFR | aGVHD III–IV | 0.95 | 0.36–2.48 | 0.919 |
| ASI | aGVHD III–IV | 0.94 | 0.36–2.45 | 0.892 |
| AFR | cGVHD | 1.46 | 0.74–2.85 | 0.274 |
| ASI | cGVHD | 1.02 | 0.56–1.86 | 0.940 |
| AFR | Neutrophil engraftment | 1.26 | 0.74–2.15 | 0.338 |
| ASI | Neutrophil engraftment | 1.11 | 0.70–1.75 | 0.667 |
The overall survival model was adjusted for disease type, disease stage, patient age, donor age, patient-donor CMV serostatus, and stratified by graft type and Karnofsky score.
The disease-free survival model was adjusted for disease type, disease stage, patient age, donor age, patient-donor CMV serostatus, GVHD prophylaxis, and stratified by graft type and Karnofsky score. The transplant-related mortality model was adjusted for disease type, patient age, patient-donor CMV serostatus, GVHD prophylaxis, patient-donor sex match, and stratified by graft type and Karnofsky score. The relapse model was adjusted for disease type, disease stage, GVHD prophylaxis, HLA matches at HLA-DPB1. The aGVHD II–IV model was adjusted for ABO match, graft type, time from diagnosis to transplant, in-vivo TCD usage, year of transplantation, and stratified by disease type and GVHD prophylaxis. The aGVHD III–IV model was adjusted for ABO match, graft type, time from diagnosis to transplant, year of transplantation. The cGVHD model was adjusted for patient age, donor age, graft type, time from diagnosis to transplant, in-vivo TCD usage, year of transplantation, and stratified by disease type and GVHD prophylaxis. The neutrophil engraftment model was adjusted for patient age, donor age, Karnofsky score, TBI usage, TBI dose, and stratified by disease type, graft type and GVHD prophylaxis.
Our analysis revealed discrepancy between the results when AFR admixture was treated as a continuous variable compared to a categorical variable. This led us to test for an admixture cut-point. The optimal cut point was >14% for recipient AFR admixture and only included 2.8% of the population (N=34 recipients) (Figure 2A). Ninety percent of the self-identified AFA recipients were included in the >14% recipient AFR admixture group (data not shown). When testing for an admixture cut-point for donor AFR genetic admixture, the cut-point was >23% which only included 1.9% of the donor population (N=24 donors) but 89% of self-identified AFA donors (Figure 2B). The curves began splitting 6–12 months after transplant.
Figure 2.
Probability of overall Survival (OS) by African (AFR) genetic admixture: The 5-year probability of OS for (A) recipients with AFR genetic admixture ≤ 14% and >14% after 10/10 HLA matched transplants was 46% (95% CI 44 – 49) and 24% (95% CI 10 – 40), p=0.004 respectively. For recipient receiving transplants from (B) donors with AFR genetic admixture ≤ 23% and >23% the same probability was 47% (95% CI 44 – 49) and 13% (95% CI 3 – 28), p<0.001 respectively.
When considering recipient AFR admixture, univariate analyses showed that the 5-year probability of OS for recipients with AFR genetic admixture ≤ 14% and >14% was 46% (95% CI 44– 49) and 24% (95% CI 10 – 40), p=0.004 respectively (Supplemental table 1A). However, for recipients receiving transplants from donors with AFR genetic admixture ≤ 23% and >23% the same probability was 47% (95% CI 44 – 49) and 13% (95% CI 3 – 28), p <0.001 respectively (Supplemental table 1B).
Transplant-Related Mortality
Similar to the OS analysis, recipient or donor AFR admixture above the identified cutoffs was found to be strongly associated with higher TRM (HR = 3.3, 95% CI 1.72–6.32, p-value=0.0003 and HR = 3.86, 95% CI 1.96–7.58, p-value=0.0001, respectively) (Supplemental table 1A and 1B). Multivariate analyses showed no significant associations with any of the other studied genetic admixtures (Tables 3 and 4).
Table 4.
Multivariate analysis* evaluating the effects of donor African (AFR), Asian (ASI), European (EUR) and South European/Amerindian (SE/A) genetic admixture as a continuous variable. Bold values indicate statistically significant results.
| Genetic Admixture | Outcome | Hazard Ratio | 95% Confidence Interval | p-value |
|---|---|---|---|---|
| AFR | OS | 3.09 | 1.70–5.64 | 0.0002 |
| ASI | OS | 0.56 | 0.26–1.19 | 0.133 |
| EUR | OS | 1.13 | 0.89–1.43 | 0.312 |
| SE/A | OS | 0.80 | 0.61–1.05 | 0.115 |
| AFR | DFS | 2.46 | 1.38–4.37 | 0.002 |
| ASI | DFS | 1.10 | 0.61–1.98 | 0.751 |
| EUR | DFS | 1.10 | 0.88–1.38 | 0.409 |
| SE/A | DFS | 0.76 | 0.59–0.99 | 0.046 |
| AFR | TRM | 3.86 | 1.96–7.58 | 0.0001 |
| ASI | TRM | 1.03 | 0.43–2.46 | 0.950 |
| EUR | TRM | 0.97 | 0.72–1.30 | 0.845 |
| SE/A | TRM | 0.84 | 0.59–1.18 | 0.306 |
| AFR | Relapse | 1.30 | 0.44–3.85 | 0.641 |
| ASI | Relapse | 1.17 | 0.54–2.53 | 0.699 |
| EUR | Relapse | 1.34 | 0.95–1.90 | 0.096 |
| SE/A | Relapse | 0.63 | 0.42–0.96 | 0.029 |
| AFR | aGVHD II–IV | 1.00 | 0.45–2.25 | 0.993 |
| ASI | aGVHD II–IV | 1.14 | 0.61–2.14 | 0.686 |
| EUR | aGVHD II–IV | 0.85 | 0.66–1.08 | 0.184 |
| SE/A | aGVHD II–IV | 1.21 | 0.92–1.60 | 0.178 |
| AFR | aGVHD III–IV | 1.20 | 0.42–3.45 | 0.728 |
| ASI | aGVHD III–IV | 1.13 | 0.45–2.82 | 0.795 |
| EUR | aGVHD III–IV | 0.92 | 0.65–1.29 | 0.621 |
| SE/A | aGVHD III–IV | 1.07 | 0.72–1.58 | 0.739 |
| AFR | cGVHD | 1.50 | 0.65–3.48 | 0.343 |
| ASI | cGVHD | 0.74 | 0.38–1.44 | 0.379 |
| EUR | cGVHD | 0.94 | 0.73–1.21 | 0.362 |
| SE/A | cGVHD | 1.10 | 0.83–1.46 | 0.491 |
| AFR | Neutrophil engraftment | 1.25 | 0.72–2.19 | 0.431 |
| ASI | Neutrophil engraftment | 0.94 | 0.58–1.50 | 0.786 |
| EUR | Neutrophil engraftment | 1.07 | 0.89–1.29 | 0.458 |
| SE/A | Neutrophil engraftment | 0.90 | 0.73–1.11 | 0.318 |
The above models were adjusted by the same set of variables as listed in Table 3.
The 5-year risk of TRM for recipients with AFR admixture >14% was 51% (95% CI 34 – 68), higher than recipients with AFR admixture ≤ 14% (31%, 95% CI 28 – 33), p=0.02 (Figure 3A, and Supplemental table 1A). For recipients receiving transplants from donors with AFR genetic admixture ≤ 23% and >23%, risk of TRM at five years post transplantation was 30% (95% CI 28 – 33) and 67% (95% CI 47– 84), p<0.001 respectively (Figure 3B, and Supplemental table 1B). Survival curves based on donor AFR admixture started diverging immediately after transplant but curves based on recipient AFR admixture began diverging at 6–12 months post-transplant.
Figure 3.
Risk of Transplant related mortality (TRM) by African (AFR) genetic admixture: The 5-year risk of TRM for (A) recipients with AFR genetic admixture ≤ 14% and >14% after 10/10 HLA matched transplants was 31% (95% CI 28 – 33) and 51% (95% CI 34 – 68), p=0.023 respectively. For recipient receiving transplants from (B) donors with AFR genetic admixture ≤ 23% and >23%, risk of TRM was 30% (95% CI 28 – 33) and 67% (95% CI 47– 84), p<0.001 respectively.
Disease Free Survival
Univariate analysis revealed a significant association between donor AFR admixture cut point groups and the probability of DFS at one, three and five years (p<0.001 – Figure 4 and Supplemental table 4). The DFS curves began separating at 6–12 months after transplant. Donor AFR admixture above the cutpoint was adversely associated with DFS. After adjusting for the significant covariates, multivariate analysis found higher AFR donor admixture above the cutpoint was associated with worse DFS (HR 2.46, 95% CI 1.38–4.37, p=0.002) (Table 4). Recipient AFR admixture was not associated with DFS.
Figure 4.
Kaplan-Meier Disease-Free Survival (DFS) plots for recipients receiving transplants from donors with AFR genetic admixture ≤ 23% and >23%. Five year probability of DFS for donor AFR admixture ≤ 23% and >23% were 40% (95% CI 38–43) and 8% (95% CI 1–22) p<0.001 respectively.
Donor/recipient pairwise genetic distance
Following the process described in Methods, 88 eigenvectors were used to calculate the donor/recipient pairwise genetic distance. A multivariate analysis using genetic distance as a continuous variable revealed no significant associations with any of the studied clinical outcomes at p<0.01.
Discussion
In this study we explored the use of genetic admixture as a prognostic factor in the outcome of HLA-matched 10/10 HCT for malignant disease. We analyzed the effect of donor and recipient genetic admixture separately, as well as their interaction through a calculated pairwise genetic distance. Admixture was estimated using a published and well validated panel of AIMs18. To our knowledge, this is the first study to address the effect of genetically defined ancestral origin on HCT outcome, as opposed to the classical use of self-identified race/ethnicity.
While no significant associations were found for the donor/recipient genetic distance used here, AFR admixture in recipients and donors above the cut point was associated with lower OS and DFS and increased TRM. The observed effect was relevant to about 3% of the recipients and donors with AFR genetic admixture greater than the statistically identified cut-point; groups that contained >89% of the self-identified African-Americans. Our findings were not associated with significant increase in GVHD or relapse, suggesting that the main driver is potentially TRM, not related to GVHD, which occurred 6–12 months after transplant.
Several factors might have confounded the genetic distance analysis, including difference in population diversity among the studied groups, for example individuals of African ancestry are more genetically diverse than others, population gene pool size and donor/recipient interaction not covered by the calculated genetic distance.
The study cohort included 29 recipients and 21 donors that self-identified as AFA. The 29 AFA recipients received a graft from 18 AFA, seven CAU, three multiethnic and one HIS donor while the 21 self-identified AFA donors donated stem cells to 18 AFA, two CAU and one HIS recipient. On the other hand, the AFR admixture of 34 recipients and 23 donors surpassed the statistical thresholds driving our findings (Figures 2–4). These ‘high-risk’ groups self-identified as 26 AFA, five CAU and three HIS for recipients (supplementary table S2-a) and 17 AFA, one CAU, one multiethnic and 5 HIS for donors (supplementary table S2-b), demonstrating that not all AFA individuals in the study had an AFR genetic admixture high enough to be included in the population subset driving the observed effects. This illustrates the difference between classifying individuals by SIRE and using genetic criteria. For example, in figure 1C most individuals in the top right cluster reported AFA SIRE (red Xs) except for two reporting CAU (yellow circles). Similarly, two individuals reported AFA SIRE (red Xs) but their genetic admixture fell outside the general top right cluster and closer to the HIS mass. This suggests that their genetic admixture is more comparable to individuals with HIS ethnicity. Importantly, the statistical threshold identified by our study (>14% AFR admixture for recipients and >23% for donors) was derived from the study cohort and is by no means indicative of the AFR admixture in a typical self-identified AFA individual. The mean proportion of AFR ancestry in self-identified African-Americans in the US usually ranges from 73% to 93%31. Some individuals in our cohort tend to have slightly higher AFR ancestry (supplementary table S2). This could be due to recent immigration from Africa or being a first or second generation descendant of African immigrants.
The genetic admixture of HIS ethnicity is even more complex than AFA and differs by the geographic area of origin of the individual; it is therefore difficult to isolate a single majority ancestral admixture (as shown in Table 2) for individuals with HIS ethnicity. Typically, the genetic composition of people reporting HIS SIRE is a mix of EUR (mostly south European from the Spanish peninsula) and Amerindian with different proportions based on the geographic origin31. However, the proportions of genetic admixture are usually unknown to the individual (% EUR versus % Amerindian) and social factors usually influence how one self-identifies. This could partly explain the overlap in CAU/HIS clusters in figure 1C. Some HIS have an additional AFR component such as Puerto Ricans31. Furthermore, in our analysis it was difficult to separate the South European (SE) and Amerindian (A) admixtures. As a result, SE/A was dealt with as a single genetic constituent forming the majority admixture in most HIS individuals in the study population (Table 2).
Our study cohort demonstrated high correlation between increased AFR genetic admixture and self-identified AFA suggesting that our findings are consistent with studies showing that African-Americans experience less successful HCT outcomes4–9. Because of this correlation, it is unclear whether the deleterious effect of recipient AFR admixture is solely genetic in nature or also due to other non-genetic factors such as access to healthcare or other socio-economic factors. Supplementary table S2 shows the high-risk subgroup of donors and recipients driving the study findings. Only 19 out of the 34 recipients (56%) with AFR admixture > 14% received grafts from donors with AFR admixture > 23% (supplementary table S2-a) while all 24 donors with AFR admixture >23% donated to recipients with AFR admixture > 14% (Supplementary table S2-b). This suggests a recipient AFR genetic ancestry driver, although a larger sample is needed for further validation.
The fact that poor outcome was observed in transplants where the donor’s AFR admixture was higher than the defined threshold also suggests a genetic driver, especially for DFS where statistical significance was reached for the effect of donor but not recipient AFR admixture. We examined the center-reported primary causes of death in the high-risk groups, where 25 of the 34 high-risk recipients (73.5%) and 21 of 24 recipients receiving grafts from high-risk donors (87.5%) died. We found no significant differences between the distributions of causes of death in both groups as well as compared to the rest of the study cohort.
Our results support the possible use of SIRE as a proxy for genetic ancestry for some SIRE groups like AFA and API, since SIRE captures a majority of AFR and ASI genetic admixtures as shown in the relatively pure cluster colors of self-identified AFA and API in figures 1-A/B and tight red and orange clusters in figure 1C. This becomes more complicated in the case of CAU and HIS SIRE groups, especially in US cohorts where populations have undergone a fair amount of recent admixture and frequent discordance between self-identification and genetic admixture occurs, as shown by the overlap of the CAU and HIS clusters in figure 1C. One should be careful when considering the findings of this study in selecting 10/10 matched donors for HCT, especially if multiple 10/10 donors of different SIRE are available and the recipient is of AFA SIRE. While the findings are in favor of selecting a non-AFA donor, the sample size driving these findings is not sufficiently large to settle this issue. Further analysis is required to validate these findings.
We acknowledge some limitations with the presented analysis. Due to the nature of the study cohort (10/10 HLA allele-matched URD transplants) the study cohort included a small subset of individuals (7%) with non-CAU SIRE. A larger, more diverse sample could help validate our findings for the analyzed variables and potentially resolve some of the issues affecting analyses of others (e.g. linearity test). Additionally, the 10/10 HLA allele-matched selection criteria raised the odds of SIRE-matched donor-recipient pairs. Expanding the study to mismatched transplants could increase the diversity in the sample SIRE groups and SIRE match patterns.
Supplementary Material
Highlights.
This is the first study to address the effect of genetic ancestry on HCT outcome, as opposed to self-identified race/ethnicity.
Higher African genetic admixture in recipients and donors was associated with lower Overall Survival and Disease-Free Survival and increased Transplant-Related Mortality (TRM).
Our findings were not associated with significant increase in GVHD or relapse, suggesting that the main driver is potentially TRM, not related to GVHD, which occurred 6–12 months after transplant.
While the findings are statistically significant, higher number are needed to validate our study findings.
Acknowledgments
We would like to acknowledge Mark Albrecht for his valuable input on some analysis aspects. This work was supported by the Office of Naval Research grant N00014-11-1-0339. The CIBMTR is supported by Public Health Service Grant/Cooperative Agreement 5U24-CA076518 from the National Cancer Institute (NCI), the National Heart, Lung and Blood Institute (NHLBI) and the National Institute of Allergy and Infectious Diseases (NIAID); a Grant/Cooperative Agreement 5U10HL069294 from NHLBI and NCI; a contract HHSH250201200016C with Health Resources and Services Administration (HRSA/DHHS); two Grants N00014-15-1-0848 and N00014-16-1-2020 from the Office of Naval Research; and grants from * Actinium Pharmaceuticals, Inc.; Alexion; * Amgen, Inc.; Anonymous donation to the Medical College of Wisconsin; Astellas Pharma US; AstraZeneca; Atara Biotherapeutics, Inc.; Be the Match Foundation; * Bluebird Bio, Inc.; * Bristol Myers Squibb Oncology; * Celgene Corporation; Cellular Dynamics International, Inc.; Cerus Corporation; * Chimerix, Inc.; Fred Hutchinson Cancer Research Center; Gamida Cell Ltd.; Genentech, Inc.; Genzyme Corporation; Gilead Sciences, Inc.; Health Research, Inc. Roswell Park Cancer Institute; HistoGenetics, Inc.; Incyte Corporation; Janssen Scientific Affairs, LLC; * Jazz Pharmaceuticals, Inc.; Jeff Gordon Children’s Foundation; The Leukemia & Lymphoma Society; Medac, GmbH; MedImmune; The Medical College of Wisconsin; * Merck & Co, Inc.; * Mesoblast; MesoScale Diagnostics, Inc.; * Miltenyi Biotec, Inc.; National Marrow Donor Program; Neovii Biotech NA, Inc.; Novartis Pharmaceuticals Corporation; Onyx Pharmaceuticals; Optum Healthcare Solutions, Inc.; Otsuka America Pharmaceutical, Inc.; Otsuka Pharmaceutical Co, Ltd. – Japan; PCORI; Perkin Elmer, Inc.; Pfizer, Inc; * Sanofi US; * Seattle Genetics; * Spectrum Pharmaceuticals, Inc.; St. Baldrick’s Foundation; * Sunesis Pharmaceuticals, Inc.; Swedish Orphan Biovitrum, Inc.; Takeda Oncology; Telomere Diagnostics, Inc.; University of Minnesota; and * Wellpoint, Inc. The views expressed in this article do not reflect the official policy or position of the National Institute of Health, the Department of the Navy, the Department of Defense, Health Resources and Services Administration (HRSA) or any other agency of the U.S. Government.
Footnotes
Corporate Members
Disclosure of Conflicts of Interest
The authors have no relevant conflicts of interest to declare.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Davies SM, Kollman C, Anasetti C, et al. Engraftment and survival after unrelated-donor bone marrow transplantation: a report from the national marrow donor program. Blood. 2000;96(13):4096–4102. [PubMed] [Google Scholar]
- 2.Serna DS, Lee SJ, Zhang MJ, et al. Trends in survival rates after allogeneic hematopoietic stem-cell transplantation for acute and chronic leukemia by ethnicity in the United States and Canada. J Clin Oncol. 2003;21(20):3754–3760. doi: 10.1200/JCO.2003.03.133. [DOI] [PubMed] [Google Scholar]
- 3.Oh H, Loberiza FR, Zhang MJ, et al. Comparison of graft-versus-host-disease and survival after HLA-identical sibling bone marrow transplantation in ethnic populations. Blood. 2005;105(4):1408–1416. doi: 10.1182/blood-2004-06-2385. [DOI] [PubMed] [Google Scholar]
- 4.Baker KS, Davies SM, Majhail NS, et al. Race and socioeconomic status influence outcomes of unrelated donor hematopoietic cell transplantation. Biol Blood Marrow Transplant. 2009;15(12):1543–1554. doi: 10.1016/j.bbmt.2009.07.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Mielcarek M, Gooley T, Martin PJ, et al. Effects of race on survival after stem cell transplantation. Biol Blood Marrow Transplant. 2005;11(3):231–239. doi: 10.1016/j.bbmt.2004.12.327. [DOI] [PubMed] [Google Scholar]
- 6.Bhatia S, Sather HN, Heerema NA, Trigg ME, Gaynon PS, Robison LL. Racial and ethnic differences in survival of children with acute lymphoblastic leukemia. Blood. 2002;100(6):1957–1964. doi: 10.1182/blood-2002-02-0395. [DOI] [PubMed] [Google Scholar]
- 7.Ballen KK, Klein JP, Pedersen TL, et al. Relationship of race/ethnicity and survival after single umbilical cord blood transplantation for adults and children with leukemia and myelodysplastic syndromes. Biol Blood Marrow Transplant. 2012;18(6):903–912. doi: 10.1016/j.bbmt.2011.10.040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sekeres MA, Peterson B, Dodge RK, et al. Differences in prognostic factors and outcomes in African Americans and whites with acute myeloid leukemia. Blood. 2004;103(11):4036–4042. doi: 10.1182/blood-2003-09-3118. [DOI] [PubMed] [Google Scholar]
- 9.Majhail NS, Nayyar S, Santibañez ME, Murphy EA, Denzen EM. Racial disparities in hematopoietic cell transplantation in the United States. Bone Marrow Transplant. 2012;47(11):1385–1390. doi: 10.1038/bmt.2011.214. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Egede LE. Race, ethnicity, culture, and disparities in health care. J Gen Intern Med. 2006;21(6):667–669. doi: 10.1111/j.1525-1497.2006.0512.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Carter KNHM, Blakely T, et al. How much and for whom does self-identified ethnicity change over time in New Zealand? Results from a longitudinal study. Social Policy Journal of New Zealand. 2009;(36):32–45. [Google Scholar]
- 12.Ford ME, Kelly PA. Conceptualizing and categorizing race and ethnicity in health services research. Health Serv Res. 2005;40(5 Pt 2):1658–1675. doi: 10.1111/j.1475-6773.2005.00449.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Riddell SR, Murata M, Bryant S, Warren EH. Minor histocompatibility antigens--targets of graft versus leukemia responses. Int J Hematol. 2002;76(Suppl 2):155–161. doi: 10.1007/BF03165108. [DOI] [PubMed] [Google Scholar]
- 14.Larsen ME, Kornblit B, Larsen MV, et al. Degree of predicted minor histocompatibility antigen mismatch correlates with poorer clinical outcomes in nonmyeloablative allogeneic hematopoietic cell transplantation. Biol Blood Marrow Transplant. 2010;16(10):1370–1381. doi: 10.1016/j.bbmt.2010.03.022. [DOI] [PubMed] [Google Scholar]
- 15.Toubai T, Tawara I, Sun Y, et al. Induction of acute GVHD by sex-mismatched H-Y antigens in the absence of functional radiosensitive host hematopoietic-derived antigen-presenting cells. Blood. 2012;119(16):3844–3853. doi: 10.1182/blood-2011-10-384057. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Spierings E, Hendriks M, Absi L, et al. Phenotype frequencies of autosomal minor histocompatibility antigens display significant differences among populations. PLoS Genet. 2007;3(6):e103. doi: 10.1371/journal.pgen.0030103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ustun C, Bachanova V, Shanley R, et al. Importance of donor ethnicity/race matching in unrelated adult and cord blood allogeneic hematopoietic cell transplant. Leuk Lymphoma. 2014;55(2):358–364. doi: 10.3109/10428194.2013.800200. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Paschou P, Drineas P, Lewis J, et al. Tracing sub-structure in the European American population with PCA-informative markers. PLoS Genet. 2008;4(7):e1000114. doi: 10.1371/journal.pgen.1000114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Spellman S, Setterholm M, Maiers M, et al. Advances in the selection of HLA-compatible donors: refinements in HLA typing and matching over the first 20 years of the National Marrow Donor Program Registry. Biol Blood Marrow Transplant. 2008;14(9 Suppl):37–44. doi: 10.1016/j.bbmt.2008.05.001. [DOI] [PubMed] [Google Scholar]
- 20.Gabriel S, Ziaugra L, Tabbaa D. SNP genotyping using the Sequenom MassARRAY iPLEX platform. Curr Protoc Hum Genet. 2009;Chapter 2(Unit 2.12) doi: 10.1002/0471142905.hg0212s60. [DOI] [PubMed] [Google Scholar]
- 21.Auton A, Brooks LD, Durbin RM, et al. A global reference for human genetic variation. Nature. 2015;526(7571):68–74. doi: 10.1038/nature15393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Howie B, Fuchsberger C, Stephens M, Marchini J, Abecasis GR. Fast and accurate genotype imputation in genome-wide association studies through pre-phasing. Nat Genet. 2012;44(8):955–959. doi: 10.1038/ng.2354. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Hubisz MJ, Falush D, Stephens M, Pritchard JK. Inferring weak population structure with the assistance of sample group information. Mol Ecol Resour. 2009;9(5):1322–1332. doi: 10.1111/j.1755-0998.2009.02591.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Jakobsson M, Rosenberg NA. CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics. 2007;23(14):1801–1806. doi: 10.1093/bioinformatics/btm233. [DOI] [PubMed] [Google Scholar]
- 25.Kopelman NM, Mayzel J, Jakobsson M, Rosenberg NA, Mayrose I. Clumpak: a program for identifying clustering modes and packaging population structure inferences across K. Mol Ecol Resour. 2015;15(5):1179–1191. doi: 10.1111/1755-0998.12387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Hotelling H. Analysis of a complex of statistical variables into principal components. Journal of Educational Psychology. 1933;24(6):417–441. [Google Scholar]
- 27.Klein JP. Handbook of survival analysis. Boca Raton: CRC Press, Taylor & Franciss Group; 2014. [Google Scholar]
- 28.Gooley TA, Leisenring W, Crowley J, Storer BE. Estimation of failure probabilities in the presence of competing risks: new representations of old estimators. Stat Med. 1999;18(6):695–706. doi: 10.1002/(sici)1097-0258(19990330)18:6<695::aid-sim60>3.0.co;2-o. [DOI] [PubMed] [Google Scholar]
- 29.DRC. Regression model and life tables. J R Stat Soc. 1972;33:187–220. [Google Scholar]
- 30.Francq C, Horvath L, Zakoian J-M. Sup-Tests for Linearity in a General Nonlinear AR(1) Model. Econometric Theory. 2010;26(04):965–993. [Google Scholar]
- 31.Bryc K, Durand EY, Macpherson JM, Reich D, Mountain JL. The genetic ancestry of African Americans, Latinos, and European Americans across the United States. Am J Hum Genet. 2015;96(1):37–53. doi: 10.1016/j.ajhg.2014.11.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.