|Home | About | Journals | Submit | Contact Us | Français|
Cord blood transplantation (CB-T) is increasingly used as a treatment alternative for hematologic malignancies. However, how CB-T compares to related (RD-T) and unrelated donor transplantation (URD-T) is not established. We compared survival of 75 double unit CB-T, 108 RD-T, and 184 URD-T recipients transplanted over the same period for the treatment of hematologic malignancies. Patients had similar ages and disease-risk, and a similar percentage had acute leukemia. The incidence of day 180 transplant-related mortality (TRM) of 21% (95%CI:12-31) after CB-T was higher than that of RD-T recipients. However, this was compensated for by a low risk of TRM after day 180, and a relatively low incidence of relapse. Hence, the 2 year progression-free survival (PFS) of 55% (95%CI:45-68) after CB-T was similar to that after RD-T or URD-T (p = 0.573). In multivariate analysis, donor source had no influence on PFS, with the only significant factors being recipient age and disease-risk. In a sub-analysis of 201 patients with acute leukemia, CB-T, RD-T and URD-T recipients also had similar 2 year disease-free survival (p = 0.482). These data provide strong support for the further investigation of double unit CB grafts as an alternative hematopoietic stem cell source.
Cord blood transplantation (CB-T) may be curative for patients with high-risk or advanced hematologic malignancies1-4, and can extend hematopoietic stem cell (HSC) transplant access to racial and ethnic minorities5. In recent years, use of CB as an alternative HSC source has increased substantially6. Retrospective studies have been conducted analyzing outcomes after single unit CB-T and traditional HSC sources. Compared with 8/8 human leukocyte antigen (HLA)-matched unrelated donor (URD) bone marrow (BM) transplantation, Eapen et al reported a similar 5 year disease-free survival after 4-5/6 HLA-matched CB-T, and superior survival after 6/6 HLA-matched CB-T, in children2. Rocha et al found that single unit CB-T had similar outcomes to adult 6/6 HLA-matched URD BM transplantation recipients7. Laughlin et al reported a similar leukemia-free survival after single unit myeloablative CB-T in adults as compared to 5/6 HLA-matched URD BM transplantation8, whereas more recently Eapen et al have found comparable disease-free survival after single-unit CB-T and 7-8/8 allele-matched peripheral blood or BM transplantation in adults9. Despite these findings CB-T has not yet been widely adopted, likely due to the risk of delayed or failed engraftment after single unit CB-T. We have observed, however, a high incidence of sustained donor engraftment and promising progression-free survival (PFS) after double unit CB-T10. Therefore, we conducted a retrospective analysis of survival after related donor transplantation (RD-T), URD transplantation (URD-T), and double unit CB-T performed at Memorial Sloan-Kettering Cancer Center (MSKCC) over the same period for the treatment of hematologic malignancies. Our hypothesis was that the 2 years PFS is similar after transplantation of the three HSC sources.
This retrospective analysis was conducted on patients who underwent HSC transplantation at MSKCC between October 1, 2005 and June 30, 2009. Collection and analysis of patient demographics and transplant outcomes was approved by the Human Subject Institutional Review Board. Eligible patients included adult and pediatric recipients of first allograft for the treatment of hematologic malignancies. For adult donors HLA-match was assessed at 10 HLA-alleles with adequate donor-recipient HLA-match being 9-10/10 HLA-A, -B, -C, -DRB1, -DQ matched for RDs, 8-10/10 HLA-matched for T-cell depleted URDs, and 9-10/10 for an unmodified URDs. All CB-T recipients received double unit grafts. Units were selected according to 4-6/6 HLA-A, -B antigen, -DRB1 allele match to the recipient, the cryopreserved total nucleated cell (TNC) dose (at least 1.5 × 107/kg/unit), and the bank of origin10,11. Unit-unit HLA-match was not considered in CB unit selection. High resolution typing of CB units was done routinely but usually did not influence unit selection. URDs had priority as the HSC source if patients did not have a suitable HLA-matched related donor; CB was chosen if no suitably HLA-matched URDs were available within the required time period.
Eligible diagnoses for this analysis included all consecutively transplanted patients with acute leukemia in complete remission (CR1-3), myelodysplasia (MDS) with ≤ 5% blasts, chronic myelogenous leukemia (CML) in chronic or accelerated phase, chronic lymphocytic leukemia (CLL), non-Hodgkin and Hodgkin lymphoma, and multiple myeloma. Patients with refractory or relapsed acute leukemia, juvenile myelomonocytic leukemia, myeloproliferative disorders other than CML, and non-malignant diseases were excluded as were recipients of syngeneic transplants, second allografts, two prior autologous transplants, and < 9/10 HLA-matched RD, or < 8/10 HLA-matched URD.
All patients were cared for in high-efficiency particulate air filtered rooms and received similar supportive care. Pre-transplant conditioning varied according to patient’s age, diagnosis, remission status, extent of prior therapies, and co-morbidities, and consisted of high-dose, reduced intensity myeloablative, and non-myeloablative regimens (Table 1). Graft-versus-host disease (GVHD) prophylaxis for RD-T and URD-T recipients was either with T-cell depletion12 or calcineurin-inhibitor (CNI) based. By contrast, all CB-T recipients received CNI (predominantly cyclosporine-A) and mycophenolate mofetil (1 gram every 12 hours or 15 mg/kg if < 50kg every 12 hours intravenously), and post-transplant granulocyte-colony-stimulating factor, and none had anti-thymocyte globulin10. G-CSF was used in non-CB-T recipients according to protocol or physician preference.
Standard-risk disease for acute myelogenous leukemia (AML) or acute lymphoblastic leukemia (ALL) was defined as CR1 without high-risk cytogenetics or high-risk molecular abnormalities13,14, de novo MDS with an International Prognostic Scoring System score < 2, CML in first chronic phase, and chemotherapy sensitive lymphoma in less than second relapse for aggressive histologies or less than third relapse for indolent disease without prior autologous transplantation. All remaining patients were considered high-risk.
Patients were evaluable for engraftment as from day 14 post-transplant. Neutrophil and platelet recovery were defined as previously described10. Donor chimerism was determined serially on BM and blood after transplantation. Primary graft failure was the lack of donor-derived neutrophil recovery by day 45, or requirement for either a boost from the same donor or a second transplant for lack of count recovery. Secondary graft failure was defined as a fall in ANC to < 0.5 × 109/l for ≥ 14 consecutive days after donor-derived neutrophil recovery, or requirement for a stem cell boost from the same donor, a second transplant, or the use of anti-thymocyte globulin as therapy for severe cytopenias that developed after initial engraftment. Sustained donor engraftment was defined as sustained donor-derived neutrophil recovery, and included all patients without graft failure and patients with graft failure who spontaneously recovered or were successfully rescued.
Acute and late acute/ chronic GVHD were diagnosed clinically with histologic confirmation when possible. Staging of acute GVHD was based on International BM Transplant Registry criteria15. Late acute GVHD (including persistent acute GVHD after day 100 requiring ongoing therapy, or GVHD with acute features presenting for the first time after day 100), and chronic GVHD were defined according to the National Institutes of Health consensus criteria16. Late acute and chronic GVHD were analyzed together in patients who survived for at least 100 days with donor engraftment. Relapse was defined as recurrence or progression of disease over pre-transplant baseline, whereas transplant-related mortality (TRM) was defined as death from any cause in continued remission. Overall survival (OS) and PFS were defined according to standard criteria. The primary cause of death was defined according to the algorithm of Copelan et al17.
Data on patient characteristics and transplant-related outcomes were obtained from the prospectively maintained MSKCC BM Transplant database with additional chart review as required. Outcomes were analyzed as of December 31, 2009. Survivors had a median follow-up of 22 months (range 6-52) and this was similar between HSC sources. Patient and graft characteristics among the three HSC sources were compared using the Fisher’s exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. The incidence of neutrophil and platelet engraftment, acute GVHD, late acute/ chronic GVHD, TRM, and relapse were computed using the cumulative incidence function, and Gray’s test was used to assess the differences between HSC sources. OS and PFS were calculated using Kaplan-Meier methodology for each HSC source and were compared using the log-rank test. Endpoints at specific post-transplant time-points were compared using the Wald test. Multivariate Cox regression analysis was performed to ascertain if HSC source was a significant predictor of PFS controlling for age, disease risk, and method of GVHD prophylaxis. As 10/10 HLA-matched and mismatched URD-T recipients had similar outcomes (with the exception of a higher incidence of acute GVHD in the recipients of mismatched unmodified URD-T), all URD-T recipients were combined for the purposes of comparison to RD-T and CB-T. Analyses were completed using SAS 9.2 software (SAS Institute Inc., Cary, NC) and R version 2.9.2.
One hundred-and-eight RD-T, 184 URD-T, and 75 CB-T consecutive patients fulfilled eligibility criteria. Patient and graft characteristics are summarized in Table 1. All three groups had similar age, gender, weight, recipient cytomegalovirus (CMV) seropositivity, and disease risk. A similar percentage had acute leukemia, with a similar proportion of AML and ALL among the 3 HSC sources. A higher percentage of RD-T and URD-T recipients had MDS, and a higher percentage of CB-T recipients had lymphoma. Consequently, a higher percentage of CB-T recipients had had a prior autologous transplant. The majority of the patients in the 3 groups received myeloablative conditioning. Approximately two-thirds of RD-T and URD-T grafts were T-cell depleted, whereas all CB grafts consisted of double units and were unmodified.
Nearly all (98%) RD-T recipients received grafts that were 10/10 HLA-allele matched, and the majority (60%) of URD grafts were 10/10 HLA-allele matched. CB units were markedly HLA-mismatched at high resolution. While the donor-recipient HLA-match of CB units was 6/6 (n = 5), 5/6 (n = 82), or 4/6 (n = 63) at HLA-A and -B antigens and -DRB1 alleles, units were only a median of 6/10 HLA-allele matched (range 2-9/10) to the patient. CB-T recipients also received more than a log less cells as compared to non-CB-T recipients.
The incidences of engraftment after transplantation of the three HSC sources are compared in Table 2. Recipients of myeloablative CB-T had a slower neutrophil recovery than myeloablative RD-T or URD-T recipients (p < 0.001). However, there was no difference between the 3 transplant groups in the speed of neutrophil recovery after non-myeloablative conditioning. The 93% (95%CI:87-99) cumulative incidence of sustained neutrophil engraftment after CB-T was accounted for by one unit in nearly all patients, and was lower than the other HSC sources. Five CB-T recipients had graft failure. Four received myeloablative conditioning and had primary (n = 3) or early secondary (n = 1) graft failure. In these patients, early onset multi-organ failure on days 7 and 11 (n = 2), early CMV infection (n = 1), and human herpesvirus 6 viremia18 (n = 1) may have contributed to the failure of engraftment. One additional non-myeloablative CB-T recipient had primary graft failure with autologous recovery.
By contrast, only one graft failure was seen in a RD-T recipient. This patient had secondary graft failure 4 months after a myeloablative T-cell depleted 10/10 HLA-matched RD graft, but achieved sustained donor engraftment after a stem cell boost. None of the unmodified URD-T recipients had graft failure. However, among recipients of URD myeloablative T-cell depleted grafts, 10 (eight 8-9/10 HLA-matched, and two 10/10 HLA-matched) had secondary graft failure. Six of these patients were successfully treated, and one had spontaneous recovery of donor hematopoiesis. Thus, only 3 URD-T recipients had sustained failure of donor-derived neutrophil engraftment.
The cumulative incidence of day 180 platelet engraftment to > 50 × 109/l was 80% (95%CI:71-89) in CB-T recipients and occurred at a median of 51 days (range 35-182) for recipients of myeloablative conditioning and 38 days (range 21-59) for non-myeloablative recipients. This incidence of recovery was lower than both RD-T recipients [99% (95%CI:96-100)], and URD-T recipients [93% (95%CI:89-97)] (p <0.001). The subset of CB-T recipients alive at day 100, however, had a 97% (95%CI:92-100) incidence of sustained platelet engraftment by day 180. This was lower than the 100% (95%CI:98-100) rate observed in RD-T recipients (p = 0.035), but similar to the 98% (95%CI:95-100) rate seen in URD-T recipients (p = 0.215), alive at day 100
The cumulative incidence of grade II-IV acute GVHD at day 100 was 43% (95%CI:31-54) in CB-T recipients. This was not different to that of unmodified RD-T or unmodified URD-T recipients (p = 0.326) (Table 2).
The cumulative incidence of late acute/ chronic GVHD at 1 year in CB-T recipients was 28% (95%CI:18-38) (Table 2) and consisted predominantly of ongoing acute GVHD after day 100 or overlap syndromes. This was similar to the 31% (95%CI:15-47) 1-year incidence of unmodified RD-T recipients, but lower than the 44% (95%CI:31-58) 1-year incidence of unmodified URD-T recipients (p = 0.015). When compared to T-cell depleted grafts, the incidence of late acute/ chronic GVHD at 1 year in CB-T was higher than the 12% (95%CI:4-20) incidence of RD-T recipients, but the difference with the 19% (95%CI:12-27) incidence URD-T recipients did not reach significance (p = 0.072).
The overall cumulative incidence of relapse/ progression was not different between the three groups (p = 0.813) (Table 2). Only one of the 27 CB-T recipients with a myeloid malignancy relapsed. The risk for relapse or progression was higher among patients with high-risk disease (data not shown).
The 2 year TRM after CB-T, RD-T, and URD-T were not different: 25% (95%CI:15-35), 15% (95%CI:8-23), and 27% (95%CI:20-34), respectively (p = 0.183, Table 2 and Figure 1). Notably, the early TRM in the first 180 days after CB-T was 21% (95%CI:12-31). This compared to 8% (95%CI:3-14) after RD-T (p = 0.017), and 13% (95%CI:8-18) after URD-T (p = 0.123). However, this was compensated for by a relatively low TRM after day 180 in CB-T recipients.
With a median follow-up of 22 months (range 6-52), 65% (95%CI:55-77) of CB-T recipients were alive at 2 years. This OS was similar to that seen in RD-T recipients [70% (95%CI:61-80)], and URD-T recipients [62% (95%CI:54-70)] (Table 2, Figure 2). There was also no difference in the PFS (p = 0.573), with the 2 year PFS of 55% (95%CI:45-68) in CB-T, 66% (95%CI:57-76) in RD-T, and 55% (95%CI:48-64) in URD-T recipients as shown in Table 2 and Figure 3.
Multivariate Cox regression analysis showed that the HSC source was not associated with PFS [RD-T versus CB-T hazard ratio 0.68 (95%CI:0.39-1.20), p = 0.185; URD-T versus CB-T hazard ratio 0.78 (95%CI:0.47-1.29), p = 0.329]. The method of GVHD prophylaxis was also not significant [unmodified versus TCD hazard ratio 0.83 (95%CI:0.55-1.25), p = 0.362]. The only significant factors were age at transplantation [increasing age per 5 years hazard ratio 1.07 (95%CI:1.02-1.11), p = 0.005], and high-risk disease [high-risk versus standard-risk disease hazard ratio 2.48 (95%CI:1.33-4.60), p = 0.004].
In order to better understand the causes of early versus late mortality after CB-T, a detailed cause of death analysis was performed in these patients (Table 3). Overall, the percentage of CB-T recipients who died before day 180 post-transplantation was 25% (19/75) , as compared to 14% (15/108) of RD-T, and 16% (29/184) of URD-T recipients. Organ failure was the leading cause of early mortality in CB-T recipients (n = 7/19, 37%). Organs involved included lung (n = 4), liver (n = 2), and central nervous system (n = 1). This was followed by early deaths due to graft failure, relapse, and GVHD, with the least common primary cause of early mortality being infection. The leading cause of early mortality in RD-T recipients was relapse and in URD-T recipients was GVHD.
Notably, however, the increased early mortality after CB-T was compensated for by a decreased late mortality after day 180, which was observed in only 7/56 (13%) of CB-T recipients alive at day 180. This compared with 17/93 (18%) of RD-T, and 38/155 (25%) of URD-T recipients alive at day 180. The leading cause of late mortality in CB-T recipients was relapse (n = 5, 71%), with no late deaths from graft failure, organ failure, or infection. The leading cause of late mortality in both RD-T and URD-T recipients was also relapse, followed by GVHD.
A comparison of the causes of death in the first 2 years after transplantation of the 3 HSC sources is shown in Figure 4. Overall, relapse was the leading primary cause of death in each group. Further, GVHD was as common as graft failure as a primary cause of death after CB-T.
The subset of 202 patients (64 RD-T, 99 URD-T, and 39 CB-T) with acute leukemia was analyzed in order to compare outcomes of a more uniform patient group. A similar percentage of the 3 sub-groups were adults and they had similar genders, weights and rates of CMV seropositivity (Table 4). Nearly all patients received myeloablative conditioning. However, CB-T recipients had more advanced disease with only 19 (49%) in first CR as compared to 47 (73%) of RD-T and 59 (60%) URD-T recipients (p = 0.048). Outcomes are compared in Table 5. Rates of neutrophil and platelet engraftment, and acute GVHD were similar to those observed in the overall analysis. There was no difference in the incidence of late acute/ chronic GVHD among unmodified RD-T, unmodified URD-T, and CB-T recipients (p = 0.915).
As in the entire dataset, a relatively high mortality in the first 180 days post-transplant in CB-T recipients with acute leukemia was compensated by a reduced late mortality. Most notably, the incidence of relapse over the first 2 years post-transplant was low in CB-T recipients at only 5% (95%CI: 0-12) (Figure 5, p = 0.072 by log rank analysis). Statistical comparison of the 2 year relapse incidence using the Wald test showed this 5% incidence after CB-T was significantly lower than that of all RD-T recipients (p = 0.007), and URD-T recipients transplanted with T-cell depletion (p = 0.008), and similar to that seen after unmodified URD-T (p = 0.686). Over the entire follow-up period, the OS (p = 0.491), and disease-free survival (p = 0.482) in patients with acute leukemia were similar among the three transplant groups by log rank analysis.
Survival after CB-T has significantly improved in the last decade, especially in adults. Possible contributors to this success have been better patient selection, a larger global CB inventory, the use of fludarabine-based conditioning regimens, substitution of mycophenolate mofetil immunosuppression for corticosteroids19,20, the introduction of double units CB grafts10,19-21, and improved supportive care. We are investigating CB as an alternative HSC source and have found CB-T extends transplant access to racial and ethnic minorities5, and double unit CB-T affords promising survival10. Thus, a comparison of double unit CB-T to transplantation with traditional HSC sources is both timely and appropriate. While our study is retrospective and not randomized, it has the advantage that we can compare CB-T to both RD and URD transplantation that were either unmodified or T-cell depleted, and performed over the same time period in a single institution with similar supportive care measures.
We found a 93% incidence of sustained donor engraftment after CB-T. This is similar to prior series of double unit CB-T22,23. However, this engraftment incidence was lower than that of RD-T and URD-T recipients. Two of five CB-T recipients classified as having graft failure as their primary cause of death had multi-organ failure within the first 11 days post-transplant. This likely contributed to their poor count recovery given both patients were 100% donor in their day 21 bone marrow. Viral infections with CMV and human herpesvirus 6 may have contributed to two additional graft failures in CB-T recipients. A larger CB inventory to enable transplantation of units with higher TNC dose and better HLA-match will facilitate improved engraftment24. More effective viral prophylaxis could also be beneficial. We also found a high incidence of secondary graft failure in TCD URD-T recipients. However, 7/10 of these patients achieved sustained donor engraftment after stem cell boost or spontaneous recovery, contributing to a higher rate of sustained donor engraftment after URD-T. Platelet recovery after CB-T was also inferior to other HSC sources, although it was identical to URD-T recipients in patients alive at day 100. This reflects the vulnerability of platelet recovery in CB-T recipients who are critically ill early post-transplant, and demonstrates that platelet recovery in long-term survivors of CB-T is nearly always adequate.
CB-T recipients had a similar incidence of acute GVHD as seen after the transplantation of unmodified grafts from related or unrelated donors. In addition, their incidence of late acute/ chronic GVHD was similar to unmodified RD-T, and lower than unmodified URD-T recipients. However, GVHD was the primary cause of death in five CB-T recipients. By contrast, despite the prolonged neutropenia after CB-T, infection was the primary cause of death in only two patients. Therefore, GVHD is at least as much of a problem as graft failure, and a greater threat than infection as the primary cause of death after CB-T. Three of the 5 patients who died of GVHD had sub-therapeutic CNI levels early post-transplant (data not shown). Therapeutic CNI levels are likely critical in the prevention of GVHD after CB-T, as has been reported in RD and URD transplantation25. While we are now investigating increased dosing of mycophenolate mofetil to ensure therapeutic blood levels26, additional strategies are needed, especially for those who are unable to tolerate therapeutic levels of CNI due to renal impairment or other toxicities. A larger inventory of CB units will be helpful given the critical influence of HLA-match upon acute GVHD risk after single unit CB-T24, which will likely also apply to the engrafting unit in double unit CB-T recipients.
Overall, the incidence of relapse was similar among the groups. Interestingly, only one of 27 CB-T recipients with myeloid malignancies relapsed, and the 2 year incidence of relapse in CB-T recipients with acute leukemia was only 5%. Our observation of a reduced incidence of relapse in CB-T recipients with acute leukemia (p = 0.072) is consistent with the findings of Brunstein et al who found myeloablative double unit CB-T recipients with acute leukemia or CML had a reduced relapse incidence compared with related and unrelated donor transplant recipients27. In our study, while the difference between relapse risk in CB-T recipients with myeloid and lymphoid malignancies may have been contributed to by greater disease risk in the latter group, these observations support a potent graft-versus-leukemia effect after double unit CB-T as suggested by Verneris et al28. We hypothesize this could be related to the immune mediated graft-versus-graft interactions between the two CB units29 which could impact minimal residual disease. Moreover, the apparent robust protection against relapse in acute leukemia, combined with an increased risk of organ failure after high dose conditioning (Table 3), suggests investigation of reduced intensity double unit CB-T may be appropriate in leukemic patients in remission.
Notably, a higher incidence of TRM was seen in CB-T recipients in the first six months post-transplant as compared with RD-T recipients. This was most commonly due to early post-transplant organ failure. While deaths due to veno-occlusive disease and central nervous system toxicity were not attributable to prolonged count recovery (Table 3), a contribution from prolonged neutropenia must be considered in pulmonary deaths. Methods to speed neutrophil recovery and possibly reduced intensity conditioning will likely be required to ameliorate this toxicity. Notably, however, the increased early mortality risk in CB-T recipients was compensated for by a reduced late mortality risk. Therefore, CB-T recipients that survive the first six months post-transplant are unlikely to die of transplant-related causes. Further, CB-T recipients transplanted in remission have a relatively low risk of relapse. This relatively reduced risk of late mortality after CB-T contributed to a similar 2 year OS and PFS as observed in RD-T and URD-T recipients. Further, in multivariate analysis, the HSC source had no impact on PFS, with the only significant factors being high-risk disease and increasing age.
The similar PFS after CB-T compared to the transplantation of the traditional HSC sources is remarkable given the relatively low cell dose and degree of HLA-disparity of CB grafts. Our study, like that of Brunstein et al27 who reported comparable 5-year leukemia-free survival after double unit CB-T and related and unrelated donor transplantation, supports the use of double unit CB grafts for the transplantation of children and adults. Our results are of particular significance given, unlike URD-T, > 50% of our CB-T recipients have non-European ancestry5. Thus, CB-T can extend transplant access to racial and ethnic minorities and potentially achieve similar PFS. We acknowledge that the main limitation of this retrospective study is the heterogeneity of the patient population. CB-T recipients were, for example, a decade younger than RD-T or URD-T recipients. While it is encouraging that in the subset analysis of acute leukemia patients we found similar 2 year disease-free survival after transplantation of the three HSC sources, these results must be confirmed with larger studies of uniform patient populations. Further, our results may have been contributed to by a center effect24 given we have a specific interest in CB-T, and other centers without CB-T experience may not be able to replicate these findings. Nonetheless, the outcomes of our study support wider adoption of CB-T. We propose centers consider double unit CB-T as a potential treatment alternative in all patients with high-risk hematological malignancies who are allograft candidates and do not have any suitable related or unrelated donors, or require an urgent transplant (within 8-10 weeks of URD search initiation). Further, as center experience with CB-T increases, a randomized study between CB-T and URD-T may be appropriate in the future.
Andromachi Scaradavou, M.D., is the Medical Director of the National Cord Blood Program of the New York Blood Center as well as an Attending Physician at the Memorial Sloan-Kettering Cancer Center. The authors have no other relevant conflicts of interest to disclose.
This work was supported in part by the New York State Empire Clinical Research Investigator Program (D.P.), Gabrielle’s Angel Foundation for Cancer Research (J.N.B.), the Memorial Sloan-Kettering Cancer Center Society (J.N.B.), the Translational and Integrative Medicine Research Grant (J.N.B.), P01 CA23766 from the National Cancer Institute, National Institutes of Health.
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.