Race and ethnicity in decisions about unrelated hematopoietic stem cell donation

The use of unrelated hematopoietic stem cell (HSC) donors as a source of stem cells for patients ill with blood-related disorders continues to increase.¹  The National Marrow Donor Program (NMDP), the largest registry in the world for matching unrelated donors with patients in need of HSC transplantations, has facilitated more than 50 000 transplantations using unrelated donors in its 25 year history, with more than 5800 of those in the past year alone (NMDP program statistics). Heightened demand for unrelated donors has led to dramatic growth of the NMDP volunteer donor registry, which now numbers more than 10 million registrants, and 60 000 new volunteers are currently being registered monthly (NMDP program statistics). In this context, questions about what motivates people to join an unrelated HSC registry and what factors are associated with the decision of whether to continue toward donation after being identified as a potential match for a patient have become increasingly important.

Despite the large number of registered potential donors, the NMDP and unrelated HSC registries worldwide continue to face difficulties in identifying matched donors for some patients, in particular racial/ethnic minorities.²  A recent investigation found that although approximately 79% of white (WH) persons searching the NMDP registry find at least one 8-of-8 HLA allele-matched potential donor, only 50% of Asian/Pacific Islander (API), 44% of Hispanic (HIS), and 33% of African-American (AA) persons find a similarly HLA-matched potential donor (L. Gragert, M. Eapen, E. Williams, J. Freeman, S. Spellman, R. Baitty, R. Hartzman, D. Rizzo, M. Horowitz, D.L.C., M. Maiers, The current status of patient-donor HLA-matching for hematopoietic stem cell transplantation in the United States; manuscript in preparation). Racial and ethnic minority patients naturally have a smaller absolute pool of potential donors from which to choose given the association between HLA type and genetic heritage. Minority status is also often associated with less common HLA types.³  Finally, donor attrition rates from the NMDP registry are much higher for racial/ethnic minority groups than they are for nonminorities (approximately 60% attrition for minority groups vs 40% for WH persons; NMDP program statistics). Unfortunately, these factors combine to disproportionately affect racial/ethnic minority patients.^4,5  The reasons for higher ethnic minority attrition at key decision points leading to donation have not been examined systematically.

A critical decision point for potential donors leading to actual donation is confirmatory typing (CT) stage. Registry members identified as a potential match for a patient are contacted at the CT stage to undergo blood tests and to receive detailed information about their role as a potential donor, the probability of ultimately donating, and the medical risks of donation. At this decision point, WH persons are approximately 30% more likely to agree to move forward toward donation than are all other race/ethnic groups (NMDP program statistics). This difference in attrition at the CT stage has remained relatively constant throughout the past 2.5 decades of NMDP facilitation of HSC transplantations. Although it is clear that the WH-minority difference in attrition likely has little to do with simple membership in a minority group, factors that might be associated both with racial/ethnic group and risk of opting out of the registry have not been investigated fully.

We set out to examine factors potentially associated both with race/ethnicity and attrition from the registry above and beyond donor demographic characteristics. Based on previous investigations, we identified 3 classes of variables—culturally related, psychosocial, and donation-related—that have been associated either with race/ethnicity in the context of HSC or solid-organ donation or with the decision to opt out of the NMDP registry (see “Sudy measures” for more information on each variable). Culturally related factors included religious beliefs and potential religious objections to HSC donation,^6-9  family cohesion,^6,9-11  and mistrust either of the medical system at large or in the equitable allocation of HSC donations.^6,9,11-14  Psychosocial factors included psychological distress,¹⁵  self-esteem and mastery,^15,16  and health-related quality-of-life. Donation-related factors included other's reactions to potential donation (ie, encouragement to donate or discouragement from donating¹⁷ ), the extent to which the person views him/herself as a donor or believes that others view him/her as a donor,¹⁵  ambivalence about donation,^15,17  donation-related concerns,^6,15,18  and satisfaction with the decision of whether to continue toward donation.¹⁵  The overarching question to be addressed by the present study was whether associations between these 3 classes of factors and attrition would provide insights into the higher minority group attrition at this key decision point leading to donation.

In collaboration with the NMDP, we generated a random, race-stratified sample of people who were contacted at the CT stage and decided either to proceed toward donation or to opt out of the registry. Our goal was to address 3 specific questions: (1) Do potential donors in 5 race/ethnic groups differ across the 3 key classes of variables and do WH persons differ from each of the other groups? (2) Do potential donors who continue toward donation at the CT stage differ from those who opt out of the registry across the 3 classes of variables? (3) Within each race/ethnic group, what are the variables most strongly associated with the decision to opt out of the registry? We anticipated that the findings from this investigation would both inform the strategies for donor management in unrelated HSC registries and have potential implications in the context of increasingly prevalent unrelated solid-organ donation.

This investigation was reviewed and approved by the institutional review boards at the University of Pittsburgh and the NMDP. All participants signed informed consent before completing the study interview in accordance with the Declaration of Helsinki.

This cross-sectional investigation included NMDP potential donors who were contacted between April 2007 and October 2010 after preliminarily matching a patient in need of an HSC transplantation and decided either to continue toward donation or to opt out of the registry. Fifty-one NMDP-operated and network donor centers representing all 50 states and the Commonwealth of Puerto Rico participated. The sampling strategy involved stratifying first by CT-stage decision (continue or opt out) and then by race/ethnic group (WH, AA, API, HIS, or American Indian [AI]), and randomly sampling within these strata. Because this investigation was focused on examining factors associated with race/ethnicity, our sampling goal was to include a minimum of 15% of each of 3 major non-WH ethnic groups and 5% AIs. Adjustments to the sampling algorithm were made when necessary to ensure that we met target inclusion goals. Eligible participants were NMDP registry members 18-60 years of age who had not previously donated BM or stem cells. To maximize participation among minority groups, the interview and all other study materials (eg, consent forms and scripts used during participant recruitment) were translated into and, when needed, conducted in Spanish, Korean, Cantonese, or Mandarin. Potential participants were excluded if they did not speak English or one of these other languages or if they did not have access to a telephone.

Randomly sampled potential participants were contacted by telephone by staff at the NMDP coordinating center shortly after they made the decision of whether to continue toward donation to briefly explain the study and to obtain permission for the release of contact information to the University of Pittsburgh. Participants who continued toward donation were interviewed before donation. Potential participants received an average of 10 call attempts at the permission stage before being declared unreachable. University of Pittsburgh research staff contacted potential participants to further explain the study and to obtain verbal permission to send a consent form. Potential participants received an average of 14 call attempts at this stage before being declared unreachable. Consent forms were mailed and follow-up by telephone was conducted if the signed consent form was not returned within 3 weeks. Full consent packets were remailed if the form was not received within 6 weeks.

After receipt of signed consent, interviewers from the Survey Research Program at the University of Pittsburgh Center for Social and Urban Research contacted participants to conduct the 40-minute structured telephone interview in the respondent's preferred language. A Computer Assisted Telephone Interview (CATI) system was used to collect and directly enter the interview data. Data were stored on a secure server in a proprietary CATI system data file. Participants received a $40.00 honorarium after completing the interview.

Four broad categories of participant characteristics were assessed in the interview: (1) demographic, (2) culturally related, (3) psychosocial, and (4) donation-related characteristics. Specific factors within each broad category represented conceptual subcomponents of the broader category. Measures were previously validated scales with established psychometric properties either created for or used in other donation-related settings.

Demographic characteristics included race, sex, age, marital status, education level, employment status, and income.

Religion importance/influence were assessed with 2 standard items gauging the importance and influence of religious beliefs in the past 2 months. Items were averaged to create a scale ranging from 1-9, with a higher score indicating greater importance/influence of religious beliefs (Cronbach α = .91 in the present sample). Religious objections to HSC donation was assessed with an 8-item subscale of the Organ Donation Attitude Survey.¹⁹  Items assessed religious beliefs as they relate to HSC donation; for example: “I believe that stem cell donation is against my religious or personal beliefs.” Item scores were averaged to create a scale ranging from 1-5, with a higher score indicating more religious objections to donation (Cronbach α = .72). Mistrust of the medical system was assessed with the 10-item Health Care System Distrust Scale.²⁰  Items assessed the degree to which respondents believed the health care system is honest, confidential, and competent. Items were averaged to create a scale ranging from 1-5, with a higher score indicating less trust in the medical system (Cronbach α = 0.70). HSC allocation mistrust was assessed with 6 items adapted from the Organ Donation Attitude Survey.¹⁹  Items assessed issues of perceived fairness in the HSC allocation system; for example: “stem cells go to the person who needs them most regardless of their race.” Items were averaged to create a scale ranging from 1-5, with a higher score indicating less trust in the fairness of the allocation of HSCs (Cronbach α = .75). Family cohesion was assessed with the 16-item Bardis Familism Scale.²¹  Items assessed feelings of family loyalty and mutual support and 10 items dealt with nuclear and 6 items with extended family; for example, “a person should always consider the needs of their family as more important than their own needs.” Items were averaged to form a scale ranging from 1-5, with a higher score indicating stronger family cohesion (Cronbach α = 0.77).

Anxiety and depression were assessed with the anxiety and depression subscales of the Brief Symptom Inventory.^22,23  Each subscale comprised 6 items that were averaged to create composite scales ranging from 0-4, with higher scores indicating greater emotional distress (Cronbach α = .66 and 0.73, respectively). Self-esteem was assessed with the 8-item Rosenberg self-esteem scale.²⁴  Items assessed feelings of self-worth and were averaged to create a scale ranging from 1-4 with a higher score indicating greater feelings of self-worth (Cronbach α = .86). Mastery was assessed with the Mastery Scale²⁵  and included 7 items assessing whether people felt that they had influence over things that happened to them. Items were averaged to create a scale ranging from 1-4, with a higher score indicating greater perceived mastery (Cronbach α = .83). Perceived physical and mental health were assessed with the SF-8 physical and mental health composite scores.²⁶  Subscale scores ranged from 28-69, with higher scores indicating better physical and mental health.

Others' reaction to donation was assessed with 2 items used extensively in our previous research with donors.^27,28  One item asked whether respondents had been encouraged by anyone to donate (yes/no) and a second item asked if they had been discouraged by anyone from donating (yes/no). Ambivalence about donation was assessed with the 7-item Ambivalence Scale that has been used extensively in living solid-organ and tissue donors.^27,29-31  An example item is “how hard of a decision was it for you to decide whether to donate stem cells?” (1 = not at all hard to 4 = very hard). Five of the 7 items were reworded to past tense for potential donors who opted out of the registry. Items were averaged to form a scale ranging from 1-4, with a higher score indicating greater uncertainty/reluctance about donation (Cronbach α = .73). Self-definition as a donor was assessed with 2 items from a scale developed for blood donors³²  and modified for HSC donors.²⁸  Items ask if being an HSC volunteer means more than just donating HSCs and whether being a HSC donor is an important part of who they are. Item scores were averaged to create a scale ranging from 1-9, with a higher score indicating a higher level of self-definition as an HSC donor (Cronbach α = .66). Donor social role was assessed with 8 items adapted from Piliavin and Callero's³²  work with blood donors and used in our work with HSC donors.²⁸  Items assessed the extent to which potential and actual donors have incorporated being a potential BM donor into their social roles; for example, whether others were aware that they were a potential donor and whether others would be disappointed/surprised if they opted out of the registry. Items were averaged to form a scale ranging from 1-9, with a higher score indicating a stronger social-role as a HSC donor (Cronbach α = .78). Medical/family/other concerns were assessed with 4 medical concern items (eg, pain and anesthesia), 5 family/work concern items (eg, missing time from work and family would worry), and 6 other concerns (eg, payment for the procedure, transportation, and donation would go to someone with different beliefs). All items were scored as 0 = not endorsed or 1 = endorsed. Concerns were summed separately for the 3 types of concerns.^27,28,31  Risk of a serious donation-related complication was assessed with a single item asking the respondent's perception of the likelihood that a serious complication could occur during donation (1 = not at all likely to 4 = very likely^27,28 ). Satisfaction with the decision to (or not to) donate HSCs was measured by a 2-item scale used in previous work with BM donors and based on past work with other donor groups (eg, kidney donors³³  and blood donors³² ). Items asked participants how satisfied and how happy they were with their donation-related decision. Responses were averaged and a higher score indicated greater satisfaction with the donation decision (Cronbach α = .78).

Data were cleaned and exported from the CATI system to PASW Statistic 18, Release Version 18.0.3 (IBM) for analysis. To examine differences by race/ethnic group, we used general linear modeling to compare mean differences for continuous variables and the χ² test to examine percentage differences for categorical variables. Post hoc comparisons of WH participants versus each of the other race/ethnic groups were also performed. To account for the multiple post hoc comparisons for this set of analyses, we applied the Holm-Bonferroni correction to each of the 4 classes of variables.³⁴  When the correction was applied, post hoc WH-minority comparisons significant at the P = .001 level remained significant, but those with P > .001 were no longer statistically significant. In the tables, we have noted which comparisons were significant at 3 levels (P = .05, P = .01, and P = .001), but those with P > .001 should be interpreted with caution.

To examine overall differences by CT-stage decision, we used the t test for continuous variables and the χ² test for categorical variables. To examine the association of key study variables with CT-stage decision within race/ethnic groups, we conducted a series of 4 binary logistic regressions (1 each for the WH, AA, HIS, and API groups; the AI sample size was not adequate for this set of analyses). Before entering variables into the logistic regression models, we examined item intercorrelations for evidence of multicollinearity. High intercorrelations (> 0.50) among several variables led us to exclude the following variables: income, medical mistrust, mastery, anxiety, depression, donor social role, medical concerns, and satisfaction with donation. Demographic variables were entered first as a block, followed by culturally related, psychosocial, and donation-related variables.

Of potential participants continuing toward donation at the CT stage, 1063 met eligibility criteria and verbally agreed to participate (an additional 87 declined to participate); interviews were completed with 843 members (79%) of this group. Of the potential participants who opted out of the registry at the CT stage, 433 met our eligibility criteria and verbally agreed to participate (an additional 315 declined to participate); interviews were completed with 224 members (52%) of this group. The final sample included 28% WH, 19% AA, 25% HIS, 21% API, and 7% AI participants. Participation rates were virtually identical for WH, AA, HIS, and API participants: AI participants were 8%-12% more likely to be available for the interview than were members of the other 4 groups. Comparisons of available demographic characteristics (ie, sex and age) for all those who completed the interview versus all those sampled and eligible indicated that those interviewed were somewhat more likely to be female for 4 of the 5 groups (WH 52% vs 45%; AA 73% vs 60%; HIS 59% vs 56%; and API 55% vs 44%). The average age was similar for WH, AA, and API participants and the total sample (ie, < 1-year difference in average age); HIS and AI participants were an average of 2 years older than the total sample. Because of the tendency of our sample to include more female participants than the total sample, we conducted all analyses reported below separately for men and women and for the sample as a whole. The pattern of results for the 3 sets of analyses was virtually identical and thus we present results for the complete sample.

Results from analyses examining differences across the 5 race/ethnic groups are presented in Table 1. The test statistic for the overall difference across groups and its associated P value is presented in the 2 right columns of the table. Significant differences from the comparison of WH participants with each of the other race/ethnic groups are indicated by asterisks in the minority group columns. For example, findings for the importance of religious beliefs indicated that there was an overall difference in religious importance across the 5 groups (far right column) and that the WH group rated religious importance lower than did the AA and HIS groups (asterisks in the AA and HIS columns). Based on our correction for potential type I error because of multiple WH-minority pairwise comparisons, we discuss only pairwise findings significant at the P = .001 level in the text, although we do indicate higher P values of P = .05 and P = .01 in Table 1. There were overall differences across ethnic groups on all demographic variables except employment status. An examination of the specific WH-minority differences indicated that the WH group: (1) was less likely to include females than was the AA group, (2) was older than the API group, (3) was more likely to be married than the AA or API groups, (4) was more highly educated than the HIS group and less highly educated than the API group, and (5) had higher income than the AA, HIS, and AI groups.

Analyses of culturally related variables indicated that there were overall race/ethnic group differences on all variables. Comparisons of WH versus other groups indicated that the WH group reported that: (1) religion was less important in their lives than it was for the AA and HIS groups, (2) they had fewer religious objections to HSC donation than did each of the other groups, (3) they had less family cohesion than the HIS and API groups, and (4) they had less mistrust in the way that HSCs would be allocated than did the AA and API groups.

Analyses of psychosocial variables indicated that there were overall race/ethnic group differences only on self-esteem and mastery. Comparisons of WH versus other groups indicated that the WH group did report higher levels of self-esteem and mastery than the API group.

Analyses of donation-related variables indicated that there were overall race/ethnic group differences on all variables except satisfaction with the donation decision. Comparisons of WH with other groups indicated that the WH group: (1) were less likely to have been discouraged from donating than the API group, (2) were less likely to define themselves as donors than were the AA and HIS groups, (3) were less likely to have internalized a donor social role than were the HIS group, and (4) had fewer medical concerns than the API group and fewer other concerns than the HIS group.

Comparisons of those who continued toward donation with those who opted out of the registry at the CT stage are presented in Table 2. The far right column lists the test statistic and significance level. There were significant differences between the 2 groups across all variables except sex, the importance of religion, family cohesion, mistrust in the medical system, and having been encouraged or discouraged from donating. Among the demographic variables, those who opted out of the registry were more likely to be younger, unmarried, less educated, unemployed, and to have lower incomes than their counterparts. Among culturally related variables, those who opted out were more likely to have religious objections to donation and more mistrust about the equitability of HSC allocation than their counterparts. Among psychosocial variables, those who opted out were more anxious and depressed, had lower self-esteem and mastery, and had lower physical and mental health summary scores than their counterparts. Among donation-related variables, those who opted out of the registry had higher ambivalence, had less self-identification or social identification as a donor, had more medical and work/family and other concerns, believed that the chances of a donation-related complication were more likely, and were less satisfied with the donation-related decision than their counterparts.

The results of 4 logistic regression analyses (1 for each race/ethnic group) are presented in Table 3. The AI group was excluded from these analyses because of its relatively small sample size. Demographic variables were entered first as a block to examine the effects of the other 3 categories of variables on CT-stage decision above and beyond demographics. Table 3 includes 3 columns for each of the race/ethnic groups: the β value, the standard error of beta, and the odds ratio. An odds ratio greater than 1 indicates increased risk of opt out associated with that variable, whereas an odds ratio less than 1 indicates that the variable decreases risk (ie, has a protective effect). Among the demographic variables, the only variable significantly associated with opting out of the registry was education for AA participants; higher education was associated with a higher likelihood of attrition. Across all groups, higher levels of ambivalence were consistently associated with higher risk of attrition. In addition, for WH participants, greater self-identification as a donor was associated with lower risk of attrition; the apparent association of “other concerns” with lower likelihood of attrition from the registry resulted from a suppression effect because of the ambivalence variable and should be considered spurious.³⁵  For HIS participants, having been discouraged from donating was associated with increased attrition risk. For API participants, a poorer score on the SF-8 mental health composite was associated with increased attrition risk.

This investigation is the first to examine comprehensively the role of race/ethnicity in the decision of whether to proceed toward unrelated HSC donation after matching a patient. The investigation is important because, based on NMDP availability statistics, the higher attrition rates among members of minority groups disproportionately disadvantage minority patients searching for a matched donor. The reasons for these higher attrition rates remain unexplained.

Analyses examining differences in study variables across the race/ethnic groups suggest that the groups differ from each other, and in particular from WH persons, in multiple ways. These differences were most evident in the culturally and donation-related variables. Four of these factors seem especially important both in terms of their association with minority group membership and their association with increased risk of attrition. Compared with WH persons, minorities reported more religious objections to donation (AA, HIS, API, and AI groups), less trust that HSCs would be allocated equitably (AA and API groups), more concerns about donation (HIS and API groups), and a greater likelihood of having been discouraged from donating (API group). In contrast, 2 factors associated with minority group membership were associated with lower attrition risk: both the AA and HIS groups were more likely than WH persons to have incorporated being a potential donor into their self- and social identifications. These findings suggest that the messages used at donor drives and those that are part of the current NMDP media campaigns targeted at recruitment of minority donors and/or the approach to managing individual donors at the CT stage could be tailored to both overcome these potential barriers to donation and to capitalize on the factors associated with lower risk of attrition. Among API persons, for example, religious, HSC allocation, and donation-related concerns could be addressed directly at drives and through NMDP media campaigns. These concerns and questions about having been discouraged from donating could be raised and addressed in recruitment settings with the option of offering a “cooling-off” period if the potential registrant has residual concerns. For all groups, self- and social identification as a potential donor could be emphasized as a potential buffer against attrition.

Findings from multivariate analyses within race/ethnic groups were also intriguing. Despite the multiple variables associated both with race/ethnic group and with increased risk of attrition at the bivariate level, ambivalence emerged in the multivariate analyses as the factor most strongly associated with attrition regardless of group membership. Although a few other variables were independently associated with attrition (concerns [WH and AA groups], poorer mental health [API group], and having been discouraged from donation [HIS group]), ambivalence provided by far the strongest and most consistent associations. The ambivalence items that asked whether the potential donor would feel (or would have felt) disappointed/relieved if they found out they could not donate, whether they have (or had) doubts and worries or feel unsure about donating, whether they would want (or have wanted) to donate even if someone else could do it, and whether they wish (or had wished) that the patient was getting the stem cells from someone else seem to be tapping into a core construct that differentiates those who continue toward donation from those who opt out of the registry. The fact that ambivalence was so strongly associated with attrition but not with race/ethnic group membership implies that this is a critical and universal factor for all potential donors in their decision of whether to proceed toward donation.

The finding that ambivalence plays a critical role in donation-related decisions is consistent with our other research on potential and actual HSC donors and implies that, despite some key differences between race/ethnic groups, there is a comprehensive factor that may either encompass or overwhelm the effects of many of the race-specific variables associated with attrition. This suggests that although tailored messages could be developed to address factors identified as being important to a particular race/ethnic group in the bivariate analyses, assessing and intervening to reduce ambivalence among all potential donors would be an efficient complimentary strategy for reducing donor attrition. One key intervention point in the management of potential donors is at the time of donor recruitment. At this point, ambivalent potential donors could be identified and the concerns that may be producing ambivalence (eg, medical concerns about the donation process, religious objections, or mistrust of the medical system) could be addressed directly at that time. Alternatively, ambivalent potential donors could be offered a cooling-off period after which time they would need to recontact the NMDP to activate their registration. Such a cooling-off period is similar to what is recommended in the context of living solid-organ donation.³⁶  Potential donors who express ambivalence when they are contacted at the CT stage after having matched a patient could be offered ambivalence-reducing interventions. One such intervention that has shown promise among living solid-organ donors is the use of brief motivational interviewing techniques to reduce ambivalence.³⁷  As applied in this setting, the technique is not designed to encourage donation, but rather to resolve ambivalence and to ensure that potential donors feel comfortable with their decision regardless of whether they continue toward donation or not. Any interventions at recruitment or CT stage would need to take particular care to strike a balance between the resolution of ambivalence by addressing concerns and misconceptions and being coercive in encouraging donation.

One potential limitation to the present study is the relatively lower participation rate among potential donors who opted out of the registry. Lower participation rates in this group are not unexpected despite our use of state-of-the-art techniques (including multiple online methods for locating contact information, multiple attempts to contact, and contact attempts at various times of day including evenings and weekends) to enhance participation, because this group of people has terminated its relationship with the NMDP and may wish no further contact regarding potential HSC donation. A potential concern might be that study participants are not representative of the larger registry and, in fact, we did find that females were overrepresented in our sample. However, several things may mitigate this concern, including: (1) subanalyses by sex that provide virtually identical results for men, women, and the total sample; (2) the fact that the multivariate analyses controlled for the effects of participant demographic characteristics; (3) the robustness of the findings across ethnic groups, particularly those related to the association of ambivalence with attrition; and (4) the consistency of these findings with our past research involving HSC donors.^15,17,28  A second limitation is the cross-sectional design. Studies designed to follow potential donors prospectively from recruitment through the CT-stage decision would be valuable but are impractical at this time given the very small proportion of new registry members who are ultimately contacted as a potential match at the CT stage.

In conclusion, we believe that these findings are important and novel and that they have implications both for the management of HSC donor registries and for the broader field of living tissue and solid-organ donation. The fact that there appear to be both race/ethnic group–specific factors and a single factor (ambivalence) that appear to be strongly associated with HSC donation decision making for all groups suggests that common donor recruitment and management approaches and tailored messages and strategies are important.

This project was supported by a grant from the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD (R01 HL081405).

The contents do not reflect the views of the Department of Veterans Affairs or the United States Government.

National Institutes of Health

Contribution: G.E.S., J.G.B., L.M., and M.A.D. designed the research, organized the project and analyses, and wrote the manuscript; J.G.B., A.G.H., and S.O. collected and managed the data at the University of Pittsburgh, assisted with data analysis, and contributed to the conceptual organization of the manuscript; A.D. and D.S. contributed to the conceptual direction of the analyses; D.L.C., L.K.A., and R.J.K. facilitated and managed sampling and participant recruitment at the National Marrow Donor Program; and all authors interpreted the data and reviewed and approved the final version of the manuscript.

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

Correspondence: Galen E. Switzer, PhD, Division of General Internal Medicine, Iroquois Bldg, Ste 502, 3600 Forbes Ave, Pittsburgh, PA 15213; e-mail: switzerge@upmc.edu.