Factors affecting mortality following myeloablative cord blood transplantation in adults: a pooled analysis of three international registries

Page 1

Factors Affecting Mortality following Myeloablative Cord BloodTransplantation in Adults: a Pooled Analysis of ThreeInternational Registries

Yael C Cohen, M.D.1, Andromachi Scaradavou, M.D.2, Cladd E. Stevens, M.D.2, PabloRubinstein, M.D.2, Eliane Gluckman, M.D.3, Vanderson Rocha, M.D.3, Mary M. Horowitz,M.D.4, Mary Eapen, M.B., B.S.4, Arnon Nagler, M.D.5, Elizabeth J. Shpall, M.D.6, Mary J.Laughlin, M.D.7, Yaron Daniely, Ph.D.8, David Pacheco, M.A.3, Raya Barishev, B.Sc.9, LirazOlmer, M.A.9, and Laurence S. Freedman, Ph.D.91Institute of Hematology, Davidoff Cancer Center, Rabin Medical Center, Beilnson Hospital,Petach Tikva, Israel2National Cord Blood Program (NCBP), New York Blood Center (NYBC), New York, NY3Eurocord-Netcord registry office, Hospital Saint Louis, Paris, France4Center for International Blood and Marrow Transplant Research (CIBMTR), Medical College ofWisconsin, Milwaukee, WI5Sheba Medical Center, Hematology Bone Marrow Transplant Department, and CBB, ChaimSheba Medical Center, Tel Hashomer, Israel6M.D. Anderson Cancer Center, Bone Marrow Transplant Department, Houston, TX7Case Western Reserve University, Case Comprehensive Cancer Center, Cleveland, OH8Gamida Cell Ltd., Jerusalem, Israel9Gertner Institute for Epidemiology and Health Policy Research, Tel Hashomer, Israel

AbstractA retrospective analysis was conducted to examine factors affecting early mortality followingmyeloablative, single-unit cord blood transplantation (CBT) for hematological malignancies inadolescents and adults. Data were collected from the three main CBT registries pooling 514records of unrelated, single, unmanipulated, first myeloablative allogeneic CBTs conducted inNorth America or Europe from 1995 to 2005, with an HLA match ≥4/6 loci, in patients aged 12 to55. Overall 100-, 180- day and 1-year survival (Kaplan-Meier method) were 56%, 46% and 37%,respectively, with no significant heterogeneity across registries. Multivariate analysis showed celldose < 2.5×107/Kg (Odds Ratio [OR] 2.76, p<0.0001), older age (p=0.002), advanced disease(p=0.02), positive CMV sero-status (OR 1.37 p=0.11), female gender (OR 1.43, p=0.07) andlimited CBT center experience (<10 records contributed, OR 2.08, p=0.0003) to be associated withhigher 100-day mortality. A multivariate model predictive of 1-year mortality included similarprognostic factors except female gender. Transplant year did not appear as a significantindependent predictor. This is the first analysis to pool records from three major CBT registries inthe US and Europe. Despite some differences in practice patterns, survival was remarkably

Corresponding author: Yael Cohen, MD, [email protected] ; phone: +972-3-9377905 ; Fax: +972-3-9378046.

Conflict-of-interest disclosure: Yael Cohen and Yaron Daniely were employees of Gamida Cell at the time this work was conducted.Laurence Freedman and Liraz Olmer acted as statistical consultants, and Raya Barishev provided data management services toGamida Cell Ltd. on research projects in stem cell transplantation.

NIH Public AccessAuthor ManuscriptBone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

Published in final edited form as:Bone Marrow Transplant. 2011 January ; 46(1): 70–76. doi:10.1038/bmt.2010.83.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 2

homogeneous. The resulting model may contribute to better understanding factors affecting CBToutcomes.

KeywordsCord blood transplantation; registry; leukemia; mortality; leukemia

IntroductionAllogeneic hematopoietic stem cell transplantation (HSCT) is a curative procedure forseveral hematopoietic malignancies. However, wide application of this procedure is limitedby the availability of a suitably HLA-matched donor1. Recent studies demonstrate that theclinical outcomes after umbilical cord blood transplantation (CBT) are comparable tooutcomes after transplantation from an HLA-matched unrelated marrow donor, providedthat a sufficient cell dose per body weight is infused1–3. Consequently, cord blood isincreasingly used as an alternative hematopoietic stem cell source, to date it is estimated thatmore than 20,000 CBTs have been conducted, of which almost half were in adults4,5. Veryfew of these transplantations were done as part of randomized controlled trials andassessment of outcomes of CBT in adults to date is based predominantly on case series andregistry analyses1–5.

In this study we describe a pooled analysis of a historical cohort compiled from threeoutcome registries: the Center for International Blood and Marrow Transplant Research(CIBMTR)7,8, the New York Blood Center National Cord Blood Program (NYBC)8 andEurocord-Netcord4 registry. The objective of the pooled analysis was to identify factorsaffecting mortality in this population based on multivariate prognostic models for mortality.These models may provide insights into the relative importance of various prognostic factorsfor outcomes and consequently, may facilitate the comparison of clinical outcomes betweenseries with different patient case-mixes.

MethodsPooling Registry Data

Records for unrelated CBTs done with myeloablative conditioning and with a singleunmanipulated cord blood unit (CBU) in the years 1995 to 2005 were provided to the data-management center at Gertner Institute for Epidemiology and Health Policy Research by thethree outcome registries: CIBMTR, NYBC and Eurocord. Records were included in thepooled analysis only if they met the following eligibility criteria: Age 12–55 years;Transplantation for acute leukemia (unless documented refractory disease), chronicmyelogenous leukemia (CML) in chronic or accelerated phase, Hodgkin lymphoma (HL) ornon-Hodgkin lymphoma (NHL) in relapse or, myelodysplastic syndrome (MDS); No priorallogeneic transplantation; Myeloablative conditioning regimen; Single, unmanipulatedCBU; with HLA disparity ≤ 2/6 loci was defined according to the total number of antigenmismatches at HLA-A and HLA-B loci (defined by serologic or low-to-intermediate-resolution DNA typing) and the number of allele mismatches at HLA-DRB1; CBUcryopreserved cell dose ≥ 1×107 Total Nucleated Cells (TNC)/ kg body weight; Transplantyear 1995 to 2005; North American or European transplant center; Availability of 100-dayvital status. After three data sets were merged by the data management center into a single,final data set, logical checks were applied to confirm the integrity and consistency of thepooled data and any discrepancies were resolved. To achieve consistency across registries,several variables were re-coded. Transplant centers CBT experience was classified as“novice” (<10 patients registered in the database) or “experienced” (10 or more patients. In

Cohen et al. Page 2

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 3

Eurocord, often only infused cell dose or cell dose at collection were captured, andcrypreseved dose was estimated as 123% and 85% of the latter, respectively, based reportscomparing these values9–15. Duplicate records were identified based on date of birth, sex,disease, date of transplant and transplant center, and HLA.

Statistical MethodsThe outcomes for the main analysis were chosen to be the 100-day and 1-year mortalityrates. In addition, the 180-day mortality, engraftment failure (surviving to day 42 but notachieving ≥500 neutrophils/mm3 by that day) and acute GvHD rates were examineddescriptively. The risk factors for mortality that were investigated included patientdemographics, disease characteristics (e.g. type of hematological malignancy and diseasestatus), treatment and transplant details, registry, year of transplant (1995–1997, 1998–2001and 2002–2005) and transplant center CBT experience.

Survival curves were estimated using the Kaplan-Meier method16 and Greenwood’s formulafor standard errors17. Heterogeneity across registries was assessed using chi-squared testsfor categorical variables and the logrank test for survival18. Logistic regression was used tobuild the prognostic model for 100-day and for 1-year mortality20. The model was built inthree stages. First, we entered five “established” prognostic factors, associated with CBTmortality in previous publications (patient/CBU HLA disparity - 0,1 or 2 mismatches for upto 6 antigens/alleles at the 3 loci6,8,21; cryopreserved cell dose - less or greater than 2.5×107 /Kg8,21–23; disease risk - early leukemia, intermediate leukemia or advanceddisease1,21,23; age – 12–18, 19–39, 40–551,21,24,25; and CMV sero-status - positive ornegative1) into a backward stepwise procedure. The resultant model was defined as the“basic” prognostic model. Second, eight "exploratory” prognostic factors, with lesscompelling evidence for association with mortality (sex, time since diagnosis, donor-recipient ABO mismatch, prior autologous transplant, donor-recipient sex mismatch,myeloablative regimen, and GvHD prophylaxis regimen and transplant center CBTexperience) were added to the basic prognostic model using a forward stepwise procedure.Third, the resulting model was used to check on the potential confounding variables, registryand year of transplant, again using a forward stepwise procedure. At all stages, the critical p-value was 0.05.

Missing values were found for two important explanatory variables, CMV sero-status (16patients) and transplant center CBT experience (26 patients). Also, for the 1-year mortalitylogistic regression analysis, 30 patients (6%) had survival times censored between 100 daysand 1-year. These missing values were handled by multiple imputation26.

ResultsStudy Population

Seven hundred and forty two records were provided by the 3 registries (after exclusion of 19duplicate records), 514 of which were found to meet the study eligibility predeterminedcriteria and were included in the final analysis. Hematological malignancies included AcuteLymphoblastic Leukemia (211), Acute Myeloid Leukemia (144), CML (106), MDS (45),HL (2) and NHL (6). Reasons for exclusion (not mutually exclusive) were non-eligibledisease stage (129), unavailable 100-day vital status (49), non-myeloablative conditioning(17), year of transplant < 1995 (1), CBU criteria (manipulation or cell dose) (21) and non-eligible country (56). Heterogeneity in prognostic factors was found in HLA mismatch, priorautologous transplant, myeloablative regimen, GvHD prophylaxis, transplant centerexperience and year of transplant. These differences are likely to reflect the differinghistories of the registries and the differing clinical management policies across the transplant

Cohen et al. Page 3

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 4

centers, rather than artifacts arising from differences in data reporting or processing.Mortality rates over 1 year were similar across the three registries (Figure 1). Two hundredand twenty seven patients (44%) of the 514 died within 100 days of the transplant. Mortalityincreased to 54% (276 deaths) and 63% (319 deaths) after 180 days and 1 year, respectively.Overall 100-, 180-day and 1-year survival (Kaplan-Meier method) were 56%, 46% and37%, respectively, with no significant heterogeneity across registries. Reported 100-, 180-day and 1-year disease free survival (Kaplan-Meier method) were 54%, 42% and 33%,respectively. Eighty-one (17%) patients had engraftment failure; 177 (36%) had aGvHDgrade II-IV. Engraftment failure rates and aGvHD disease rates were also similar acrossregistries (Figure 1).

In a univariate analysis (Table 1), four of the five “established” prognostic factors weresignificantly related to mortality at 100 days and 1 year. Of the eight “exploratory”prognostic factors, transplant center CBT experience was significantly related to both 100-day and 1 year mortality, while sex appeared related to 100-day mortality only. Amongpotential confounding factors, year of transplant appeared to be related to mortality.Noteworthy, the proportion of patients with cell dose <2.5 ×107/Kg (associated with poorprognosis)8,21,22, declined sharply over these years, being 67%, 54% and 24% in 1995–7,1998–2001, and 2002–5, respectively (data not shown). The multivariate model for 100-daymortality included all of the above factors except for year of transplant, which was no longersignificant after cell dose was included in the model (Table 2). The most influential factorswere cell dose, age and center experience. The multivariate model for 1-year mortalityincluded CMV sero-status, cell dose, disease status, patient age and center experience.

DiscussionThis analysis pooled together records of unrelated single unmanipulated CBTs in adults andadolescents from the three major international CBT registries, and was aimed to identifyprognostic factors associated with CBT outcomes. This may provide insights into the effectsof patient case-mix on clinical outcome and support patient risk assessment. This is ofparticular importance in adult CBT research, considering the inherent challenges that limitthe conduct of large randomized controlled trials in this field. To the best of our knowledge,this is the largest cohort of adults with CBT in North America and Europe that has beenreported to date and the first to pool records from all three registries. The crude overall anddisease free survival outcomes are comparable to those reported previously in CB registryreports of single, myeloablative CBT in North America and Europe4,5. TRM is inferior toresults in Japanese studies3, it has been speculated to results from availability of higher celldose units, preferential racial genetic homogeneity and possibly some differences in GvHDprophylaxis.

Several published analyses have previously investigated factors affecting CBT clinicaloutcomes1,2,6,8,21–25,27,28. Many of these studies focused on a predominantly pediatricpopulation21,23–25,28 and analyzed treatment-related mortality or event-free survival ratherthan overall mortality. Some of the studies with an adult population were relativelysmall6,24, or focused on comparison to BM transplant.1

As has been widely recognized in previous work8,21–23, we found low cell dose to bestrongly associated with increased overall mortality. Using a cut-point of 2.5×107 cells/kg tomodel this effect captured most of the relationship. We also confirmed previous reports ofadvanced disease stage1,21,23 and CMV seropositivity1 association with worse survival.Most reports8,21,22,25,6,28 found greater HLA discrepancy is related with higher treatmentrelated mortality or reduced event free survival. Conversely, and similar to our findings,Arcese et al24 did not find an impact of HLA discrepancy on mortality. A possible

Cohen et al. Page 4

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 5

explanation for our not finding HLA discrepancy to be predictive of mortality is that thereexists a trade-off in the CBU selection process between cell dose and HLA disparity (i.e.smaller units more “acceptable” with lesser HLA disparity). Another explanation could bethat HLA matching has limited effect in patients who receive relatively low cell dose, asthese patients may die early due to infections, regardless to HLA impact. Indeed our pooledcohort, 50% of the transplants involved cell dose below 2.5×107/Kg TNCs; however, the testfor an interaction term between these two variables was negative. Among the “exploratory”factors investigated, only female sex was found to be weakly associated with survival at 100days. It was not predictive at 1 year. Female sex has been previously described24 as anegative prognostic factor by one group. Transplant center CBT experience (“centereffect”), was significantly related to survival outcomes at 100 days and at 1-year and wastherefore included in the final logistic models. Center effect has previously beendemonstrated in some reports21 but not in others25. The number of records contributed to thecohort by each transplant center was used as a surrogate measure for the transplant center’sCBT experience. We selected the median number of records as the threshold, definingexperienced centers as those contributing 10 or more records, and novice centers as thosecontributing less than 10 records. Sensitivity analysis showed that this cutoff yielded thestrongest association with outcomes (data not shown). This may reflect the effect of centerexperience on patient selection and/or transplant management. An alternative explanation isthat centers with initial poor outcomes abandoned CBT, and thus remained with lesserexperience.

Finally, we tested two potential confounders: registry and year of transplant. The threeregistries varied with regards to their data capture procedures; nevertheless, all incorporateda structured review process and provided sufficiently detailed data, enabling the applicationof uniform eligibility criteria at a patient record level. Some variations in practice patternsamong centers reporting to the three registries were demonstrated, yet despite thesevariations, remarkable homogeneity in mortality outcomes of across registries, up-to 1-yearpost transplantation was demonstrated: 42%, 45%, and 43% at 100 days and 59%, 65% and63% at 1-year, for CIBMTR, NYBC and Eurocord, respectively.

With regards to the year of transplant, the 100-day mortality rate was 57% in patients treatedin 1995–7, 42% in those treated in 1998–2001, and 37% in those treated in 2002–2005 (chi-squared on 2 degrees of freedom = 10.4, p=0.006). A similar pattern was observed in the 1-year survival outcomes. On further inspection we found that the proportion of patientstransplanted with cell dose <2.5 × 107/Kg also declined sharply over these years, being 66%,50% and 21% in 1995–7, 1998–2001, and 2002–5, respectively. This may well reflect theemerging recognition among the medical community of the negative impact of low cell dosein the cord blood graft on the clinical outcomes of CBT. Hence, when we added year oftransplant to the “basic” prognostic model, the relationship was no longer statisticallysignificant. The test for trend over the three time periods was also not significant. We alsofound some indication of improved patient selection over time, thus, the proportion ofpatients transplanted with relapsed leukemia was 26% in 1995–7, 18% in 1998–2001 and15% in 2002–5. These findings support the notion that the observed improvement in CBToutcomes over time in our cohort is attributable predominantly to better patient and CBUselection.

This analysis has several potential limitations: as in any observational study, additionalpotential unmeasured confounders may exist, e.g. cord blood bank, age of the CBU orstorage conditions. Data regarding myeloablative conditioning was incomplete. While thethree registries include much of the western world’s experience in single unit CBT withinthe defined age group and geographic regions, selection bias and reporting bias may not becompletely excluded. Thus CIBMTR have almost complete 3 month follow-up, NYBC has

Cohen et al. Page 5

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 6

90–95%, and Eurocord reports 80% 3-month follow-up. At the time the cohort wasassembled (2005) some of the follow-up data from the most recent years was stillunavailable to the registry. Even when 3-month follow-up was available, some records hadmissing data. The numbers of patients with Lymphoma, those who received prior autologoustransplants, fludarabine for conditioning were small and, consequently, we had limitedpower to detect an effect of those variables. As the registries varied with regards to their datastructure and definitions, particular efforts were directed towards aligning the data to enablean accurate and valid merge. The relatively high proportion of patients with CML in thisanalysis (20%) reflect management patterns at the time the registry, and are similar to thefindings of Laughlin et al5. Estimation of cryopreserved cell dose from the infused dose orthe dose at collection was a particular challenge and no doubt introduced inaccuracies.Indeed, when analyzing by registry, cell dose was significantly related to 100-day mortalityin CIBMTR (OR 2.6, 95% CI 1.04–6.5) and NYBC (OR 3.65, 95% CI 2.19–6.06), but failedto reach significance in Eurocord (OR 1.58; 95% CI 0.85–2.93). Data reporting wassupported by experienced CBT physicians, and registries were queried in cases ofinconsistency or lack of data, yet some inaccuracies may have occurred. Current efforts toharmonize forms across registries will facilitate such analyses in the future. Overallmortality was selected as the outcome variable rather than treatment related mortality, eventhough the latter accounts for a large proportion of the mortality. The mortality due todisease recurrence may in part be attributable to the CB GVL effect and to the transplantregimen. Therefore, we considered overall mortality as the CBT outcome with the greatestclinical significance. Finally, our analysis is limited to single myeloablative un-manipulatedcord blood transplantation. Novel strategies such as double cord blood transplantation30–32

and ex-vivo cord blood expansion33–34 continue to be investigated and may enable tobroaden the use of CBT in adults and adolescents. Analyses such as the one presented in thispaper could help estimating the contribution such interventions, based on a retrospectivecase-mix adjusted comparison to single CBT outcomes. This may be of particularimportance since the assignment of patients to a low dose single CBU transplant in arandomized controlled trial setting, would now days be unacceptable, for ethical reasons.

In conclusion, this is the first analysis to pool records from the three major CBT registries inthe US and Europe which arguably contain much of the world’s experience in adult singleunit CBT. Despite some differences in practice patterns, survival data showed remarkablehomogeneity. The findings suggest that the improvement over time in CBT outcomes maybe attributable to better patient and CBU selection. This analysis establishes a mechanismfor evaluating improvements in CBTs by future interventions.

AcknowledgmentsThe authors wish to thank Drs S. Merchav, T. Peled, Y. Margolin and Y. Stark for their valuable suggestionsthroughout the conduct of this study.

References1. Gluckman E, Rocha V. Cord blood transplantation: state of the art. Haematologica. 2009; 94:451–

454. [PubMed: 19336748]

2. Stanevsky A, Goldstein G, Nagler A. Umbilical cord blood transplantation: Pros, cons and beyond.Blood Reviews. 2009

3. Takahashi S, Iseki T, Ooi J, et al. Single-institute comparative analysis of unrelated bone marrowtransplantation and cord blood transplantation for adult patients with hematologic malignancies.Blood. 2004; 104:3813–3820. [PubMed: 15280199]

4. Rocha V, Labopin M, Sanz G, et al. Transplants of Umbilical-Cord Blood or Bone Marrow fromUnrelated Donors in Adults with Acute Leukemia. N Engl J Med. 2004; 351:2276–2285. [PubMed:15564544]

Cohen et al. Page 6

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 7

5. Laughlin MJ, Eapen M, Rubinstein P, et al. Outcomes after Transplantation of Cord Blood or BoneMarrow from Unrelated Donors in Adults with Leukemia. N Engl J Med. 2004; 351:2265–2275.[PubMed: 15564543]

6. Cornetta K, Laughlin M, Carter S, et al. Umbilical cord blood transplantation in adults: Results ofthe prospective cord blood transplantation (COBLT). Biol Blood Marrow Transplant. 2005;11:149–160. [PubMed: 15682076]

7. Gluckman E, Wagner J, Hows J, Kernan N, Bradley B, Broxmeyer HE. Cord blood banking forhematopoietic stem cell transplantation: an international cord blood transplant registry. BoneMarrow Transplant. 1993; 11:199–200. [PubMed: 8096776]

8. Eapen M, Rubinstein P, Zhang MJ, et al. Outcomes of transplantation of unrelated donor umbilicalcord blood and bone marrow in children with acute leukaemia: a comparison study. Lancet. 2007;369:1947–1954. [PubMed: 17560447]

9. Armitage S. Cord Blood Processing: Volume Reduction. Cell Preservation Technology. 2006; 4:9–16.

10. Querol, S.; Azqueta, C.; Torrico, C.; Amill, B.; Torrabadella, M.; Massuet, L.; Garcia, J. Freezingand Thawing: the Barcelona Cord Blood Bank Experience. Barcelona: EBMT; 2004. AnAutomated and Closed Procedure for Cord Blood Processing.

11. Solves P, Mirabet V, Planelles D, et al. Red blood cell depletion with a semiautomated system orhydroxyethyl starch sedimentation for routine cord blood banking: a comparative study.Transfusion. 2005; 45:867–873. [PubMed: 15934983]

12. Kogler G, Sarnowski A, Wernet P. Volume reduction of cord blood by Hetastarch for long-termstem cell banking. Bone Marrow Transplant. 1998; 22(Suppl 1):S14–S15. [PubMed: 9715874]

13. Gluckman E, Rocha V, Arcese W, et al. Factors associated with outcomes of unrelated cord bloodtransplant: guidelines for donor choice. Exp Hematol. 2004; 32:397–407. [PubMed: 15050751]

14. Sanz GF, Saavedra S, Planelles D, et al. Standardized, unrelated donor cord blood transplantationin adults with hematologic malignancies. Blood. 2001; 98:2332–2338. [PubMed: 11588027]

15. Long GD, Laughlin M, Madan B, et al. Unrelated umbilical cord blood transplantation in adultpatients. Biol Blood Marrow Transplant. 2003; 9:772–780. [PubMed: 14677117]

16. Kaplan EL MP. Nonparametric estimation from incomplete observations. Journal of the AmericanStatistical Association. 1958; 53:457–481.

17. Greenwood, M. A report on the natural duration of cancer. In: London, HSO., editor. Reports onPublic Health and Medical Subjects. Vol. Vol. 33. 1926. p. 1-26.

18. Peto R, Peto J. Asymptotically efficient rank invariance test procedures (with discussion). Journalof the Royal Statistical Society. 1972; Series A:185–206.

19. Pintilie, M. Competing Risks. A Practical Perspective. In: Stephen Senn, MS.; Bloomfield, Peter,editors. Series: Statistics in Practice. Chichester, UK: John Wiley and Sons; 2006. p. 55-70.In:Series Editors

20. Hosmer, DW., Jr; Lemeshow, S. Applied Logistic Regression. 2nd edition. Hoboken, NJ: JohnWiley & Sons, Inc; 2000. p. 1-43.

21. Rubinstein P, Carrier C, Scaradavou A, et al. Outcomes among 562 recipients of placental-bloodtransplants from unrelated donors. N Engl J Med. 1998; 339:1565–1577. [PubMed: 9828244]

22. Wagner JE, Barker JN, DeFor TE, et al. Transplantation of unrelated donor umbilical cord blood in102 patients with malignant and nonmalignant diseases: influence of CD34 cell dose and HLAdisparity on treatment-related mortality and survival. Blood. 2002; 100:1611–1618. [PubMed:12176879]

23. Locatelli F, Rocha V, Chastang C, et al. Factors associated with outcome after cord bloodtransplantation in children with acute leukemia. Eurocord-Cord Blood Transplant Group. Blood.1999; 93:3662–3671. [PubMed: 10339472]

24. Arcese W, Rocha V, Labopin M, et al. Unrelated cord blood transplants in adults with hematologicmalignancies. Haematologica. 2006; 91:223–230. [PubMed: 16461307]

25. Ballen KK, Haley NR, Kurtzberg J, et al. Outcomes of 122 diverse adult and pediatric cord bloodtransplant recipients from a large cord blood bank. Transfusion. 2006; 46:2063–2070. [PubMed:17176317]

Cohen et al. Page 7

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 8

26. Little, RJA.; Rubin, DB. Statistical Analysis with Missing Data. 2nd edition. Hoboken, NJ: JohnWiley and Sons; 2002. p. 209-218.

27. Rocha V, Gluckman E. Clinical use of umbilical cord blood hematopoietic stem cells. Biol BloodMarrow Transplant. 2006; 12:34–41. [PubMed: 16399582]

28. Wall DA, Carter SL, Kernan NA, et al. Busulfan/Melphalan/Antithymocyte Globulin Followed byUnrelated Donor Cord Blood Transplantation for Treatment of Infant Leukemia and Leukemia inYoung Children: The Cord Blood Transplantation Study (COBLT) Experience. Biol BloodMarrow Transplant. 2005; 11:637–646. [PubMed: 16041314]

29. Gluckman E, Rocha V. Donor selection for unrelated cord blood transplants. Curr Opin Immunol.2006; 18:565–570. [PubMed: 16893632]

30. Barker JN, Weisdorf DJ, DeFor TE, et al. Transplantation of 2 partially HLA-matched umbilicalcord blood units to enhance engraftment in adults with hematologic malignancy. Blood. 2005;105:1343–1347. [PubMed: 15466923]

31. Brunstein CG, Barker JN, Weisdorf DJ, et al. Umbilical cord blood transplantation afternonmyeloablative conditioning: impact on transplantation outcomes in 110 adults withhematologic disease. Blood. 2007; 110:3064–3070. [PubMed: 17569820]

32. Ballen KK, Spitzer TR, Yeap BY, et al. Double unrelated reduced-intensity umbilical cord bloodtransplantation in adults. Biol Blood Marrow Transplant. 2007; 13:82–89. [PubMed: 17222756]

33. de Lima M, McMannis J, Gee A, et al. Transplantation of ex vivo expanded cord blood cells usingthe copper chelator tetraethylenepentamine: a phase I/II clinical trial. Bone Marrow Transplant.2008; 41:771–778. [PubMed: 18209724]

34. Shpall EJ, Quinones R, Giller R, et al. Transplantation of ex vivo expanded cord blood. Biol BloodMarrow Transplant. 2002:368–376. [PubMed: 12171483]

Cohen et al. Page 8

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 9

Figure 1.Kaplan Meyer survival curves across the three registries: CIBMTR, NYBC and Eurocordshow similar survival over 1 year (chi-squared on 2 df = 1.67, p=0.4)

Cohen et al. Page 9

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Page 10

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Cohen et al. Page 10

Tabl

e 1

100-

day

mor

talit

y an

d 1

year

mor

talit

y ac

cord

ing

to p

rogn

ostic

fac

tors

: uni

vari

ate

anal

ysis

Fac

tor

Tot

al(n

=514

)D

ied

at 1

00 d

ays

Die

d at

1 y

ear

(n=4

84)

N (

%)

Chi

-sq

uare

d,df

, and

P-v

alue

*

N (

%)

Chi

-sq

uare

d,df

, and

P*

Est

ablis

hed

prog

nost

ic f

acto

rs

HL

A m

ism

atch

†0

mis

mat

ches

1 m

ism

atch

2 m

ism

atch

es

23 152

339

10 (

43)

66 (

43)

151

(45)

0.06

on

2 df

P=0.

9715

(68

)98

(69

)20

6 (6

5)

0.74

on

2df P=

0.69

Cel

l dos

e (1

07 /kg

)‡L

ess

than

2.5

At l

east

2.5

255

259

142

(56)

85 (

33)

27.2

on

1 df

P<0.

0001

174

(71)

145

(61)

5.18

on

1df P=

0.02

Dis

ease

Ris

k§E

arly

Leu

kem

iaIn

term

edia

te L

euke

mia

Adv

ance

d di

seas

e

116

291

107

52 (

45)

116

(40)

59 (

55)

7.40

on

2 df

P=0.

0264

(61

)17

5 (6

4)80

(76

)

6.34

on

2df

P=0.

04

Age

(ye

ars)

12–1

819

–39

40–5

5

215

220 79

82 (

38)

97 (

44)

48 (

61)

12.0

on

2 df

P=0.

003

123

(60)

137

(67)

59 (

80)

9.55

on

2df P=

0.00

8

CM

V s

ero-

stat

usN

egat

ive

Posi

tive

Unk

now

n

221

277 16

84 (

38)

135

(49)

8 (5

0)

5.74

on

1 df

P=0.

0211

2 (5

3)19

5 (7

6)12

(80

)

26.0

on

1df P<

0.00

01

Exp

lora

tory

pro

gnos

tic f

acto

rs

Sex

Mal

eFe

mal

e

296

218

120

(41)

107

(49)

3.71

on

1df P=

0.05

4

177

(64)

142

(68)

0.90

on

1df P=

0.3

Dia

gnos

is t

o tr

ansp

lant

tim

e (y

ears

)U

p to

11–

2O

ver

2U

nkno

wn

192

133

181 8

81 (

42)

56 (

42)

85 (

47)

5 (6

3)

1.09

on

2df P=

0.6

117

(66)

81 (

64)

116

(67)

5 (6

3)

0.33

on

2df P=

0.8

Pri

or a

utol

ogou

s tr

ansp

lant

Yes

No

Unk

now

n

52 443 19

25 (

48)

193

(44)

9 (4

7)

0.38

on

1df P=

0.5

35 (

67)

274

(66)

10 (

67)

0.05

on

1df P=

0.8

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

Page 11

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Cohen et al. Page 11

Fac

tor

Tot

al(n

=514

)D

ied

at 1

00 d

ays

Die

d at

1 y

ear

(n=4

84)

N (

%)

Chi

-sq

uare

d,df

, and

P-v

alue

*

N (

%)

Chi

-sq

uare

d,df

, and

P*

Don

or-r

ecip

ient

AB

O B

lood

Gro

upm

atch

Mat

chM

ism

atch

Unk

now

n

146

264

104

63 (

43)

121

(46)

43 (

41)

0.27

on

1df P=

0.6

91 (

67)

169

(67)

59 (

61)

0.00

09 o

n1

dfP=

0.98

Don

or-r

ecip

ient

Sex

mat

chM

atch

Mis

mat

chU

nkno

wn

203

222 89

92 (

45)

97 (

44)

38 (

43)

0.11

on

1df P=

0.7

138

(72)

129

(62)

52 (

62)

3.80

on

1df P=

0.05

Mye

loab

lati

on r

egim

en|

TB

I ba

sed

with

out F

LU

TB

I ba

sed

with

FL

UN

on-T

BI

with

FL

UN

on-T

BI

with

out F

LU

Unk

now

n

329 17 5

138 25

142

(43)

5 (2

9)2

(40)

63 (

46)

15 (

60)

1.67

on

3df P=

0.6

204

(65)

5 (3

1)4

(100

)87

(69

)19

(83

)

10.9

on

3df P=

0.01

GvH

D p

roph

ylax

is r

egim

enC

yclo

spor

ine

+ o

ther

, No

met

hotr

exat

eC

yclo

spor

ine

+ o

ther

+ m

etho

trex

ate

Tac

rolim

us +

oth

erO

ther

Unk

now

n

388 42 36 6 42

171

(44)

18 (

43)

10 (

28)

2 (3

3)26

(62

)

3.79

on

3df P=

0.3

242

(66)

25 (

61)

18 (

55)

2 (4

0)32

(89

)

3.10

on

3df P=

0.4

Tra

nspl

ant

cent

er C

BT

exp

erie

nce¶

novi

ce (

<10

rec

ords

)ex

peri

ence

d (1

0+ r

ecor

ds)

Unk

now

n

183

305 26

98 (

54)

121

(40)

8 (3

1)

8.91

on

1df P=

0.00

3

133

(79)

174

(60)

12 (

50)

18.4

on

1df P<

0.00

01

Pot

entia

l con

foun

ding

fac

tors

Reg

istr

yC

IBM

TR

NY

BC

Eur

ocor

d

86 262

166

36 (

42)

119

(45)

72 (

43)

0.35

on

2 df

P=0.

850

(59

)16

6 (6

5)10

3 (6

3)

1.67

on

2df

,P=

0.4

Yea

r of

tra

nspl

ant

1995

–199

719

98–2

001

2002

–200

5

112

279

123

64 (

57)

118

(42)

45 (

37)

10.9

1 on

2df P=

0.00

4

80 (

74)

169

(62)

70 (

68)

5.34

on

2df P=

0.07

* The

chi

-squ

ared

test

om

its a

ny u

nkno

wn

cate

gory

;

† HL

A d

ispa

rity

≤2/

6 lo

ci w

as d

efin

ed a

ccor

ding

to th

e to

tal n

umbe

r of

ant

igen

mis

mat

ches

at H

LA

-A a

nd H

LA

-B lo

ci (

defi

ned

by s

erol

ogic

or

low

-to-

inte

rmed

iate

-res

olut

ion

DN

A ty

ping

) an

d th

e nu

mbe

rof

alle

le m

ism

atch

es a

t HL

A-D

RB

1 (d

efin

ed b

y hi

gh-r

esol

utio

n D

NA

typi

ng);

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

Page 12

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Cohen et al. Page 12‡ C

ryop

rese

rved

cel

l dos

e w

as r

epor

ted

cons

iste

ntly

in N

YB

C a

nd C

IBM

TR

, in

Eur

ocor

d, C

ropr

esrv

ed c

ell d

ose

was

est

imat

ed in

som

e ca

ses

base

d on

123

% o

f in

fuse

d or

85%

of

colle

cted

cel

l doe

(se

ete

xt);

§ Dis

ease

ris

k w

as d

efin

ed a

s ea

rly

leuk

emia

(ac

ute

leuk

emia

in f

irst

com

plet

e re

mis

sion

or

CM

L in

fir

st c

hron

ic p

hase

), in

term

edia

te le

ukem

ia (

acut

e le

ukm

ia in

sec

ond

or s

ubse

quen

t com

plet

e re

mis

sion

,C

ML

in s

econ

d or

sub

sequ

ent c

hron

ic p

hase

or

MD

S); A

dvan

ced

dise

ase

(rel

apse

d le

ukem

ia o

r ly

mph

oma)

;

| Use

of

flud

arab

in w

as u

nava

ilabl

e in

CIB

MT

R r

ecor

ds;

# Tra

nspl

ant c

ente

rs w

ere

clas

sifi

ed a

ccor

ding

to th

eir

CB

T e

xper

ienc

e as

“no

vice

” (<

10 p

atie

nts

regi

ster

ed in

the

data

base

) or

“ex

peri

ence

d” (

10 o

r m

ore

patie

nts)

.

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.

Page 13

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Cohen et al. Page 13

Tabl

e 2

Mul

tivar

iate

adj

uste

d od

ds r

atio

s fo

r 10

0-da

y an

d 1

year

mor

talit

y fo

r th

e pr

ogno

stic

fac

tors

that

app

ear

in th

e m

ultiv

aria

te p

rogn

ostic

mod

el

Par

amet

er

100-

days

mod

el1-

year

mod

el

Odd

sra

tio

95%

CI

p-va

lue

Odd

sra

tio

95%

CI

p-va

lue

CM

VPo

sitiv

e V

s. N

egat

ive

1.37

0.93

–2.0

30.

112.

231.

50–3

.32

<0.

0001

Cel

l dos

e ×

107 /

Kg*

< 2

.5 V

s. >

2.5

2.76

1.88

–4.0

4<

0.00

011.

671.

13–2

.47

0.01

Dis

ease

ris

k†A

dvan

ced

dise

ase

Vs.

Ear

ly L

euke

mia

1.95

1.21

–3.1

50.

022.

331.

27–4

.27

0.02

Dis

ease

ris

kIn

term

edia

te L

euke

mia

Vs.

Ear

ly L

euke

mia

1.17

0.74

–1.8

71.

270.

79–2

.05

Age

19–3

9 V

s. 1

2–18

1.03

0.68

–1.5

70.

002

1.12

0.73

–1.7

10.

02A

ge40

–55

Vs.

12–

182.

641.

49–4

.68

2.43

1.31

–4.5

2

Sex

Fem

ale

Vs.

Mal

e1.

430.

98–2

.10

0.07

N/A

‡N

/AN

/A

Cen

ter

CB

T e

xper

ienc

e§N

ovic

e v

expe

rien

ced

2.08

1.39

–3.1

10.

0003

2.20

1.43

–3.3

80.

0003

* Cry

opre

serv

ed c

ell d

ose

was

rep

orte

d co

nsis

tent

ly in

NY

BC

and

CIB

MT

R, i

n E

uroc

ord,

Cro

pres

rved

cel

l dos

e w

as e

stim

ated

in s

ome

case

s ba

sed

on 1

23%

of

infu

sed

or 8

5% o

f co

llect

ed c

ell d

ose

(see

text

) ;

† dise

ase

risk

was

def

ined

as

earl

y le

ukem

ia (

acut

e le

ukem

ia in

fir

st c

ompl

ete

rem

issi

on o

r C

ML

in f

irst

chr

onic

pha

se),

inte

rmed

iate

leuk

emia

(ac

ute

leuk

emia

in s

econ

d or

sub

sequ

ent c

ompl

ete

rem

issi

on,

CM

L in

sec

ond

or s

ubse

quen

t chr

onic

pha

se o

r M

DS)

; Adv

ance

d di

seas

e (r

elap

sed

leuk

emia

or

lym

phom

a);

‡ N/A

Not

app

licab

le a

s se

x w

as n

ot s

igni

fica

ntly

ass

ocia

ted

with

1 y

ear

mor

talit

y in

the

mul

tivar

iabl

e m

odel

;

§ Tra

nspl

ant c

ente

rs w

ere

clas

sifi

ed a

ccor

ding

to th

eir

CB

T e

xper

ienc

e as

“no

vice

” (<

10 p

atie

nts

regi

ster

ed in

the

data

base

) or

“ex

peri

ence

d” (

10 o

r m

ore

patie

nts)

.

Bone Marrow Transplant. Author manuscript; available in PMC 2012 June 03.