Relationship between pediatric blood and marrow transplant ... - Nature

Bone Marrow Transplantation (2013) 48, 514–522 & 2013 Macmillan Publishers Limited All rights reserved 0268-3369/13 www.nature.com/bmt

ORIGINAL ARTICLE

Relationship between pediatric blood and marrow transplant center volume and day þ 100 mortality: Pediatric Blood and Marrow Transplant Consortium experience DS Taylor1, M Dharmar1, E Urquhart-Scott1, R Ryan2, MA Pulsipher3, A Gamis2, K Schultz4 and JP Marcin1 The number of patients receiving a BMT is currently being used as a factor in the accreditation process in determining whether a center can provide a high-quality BMT. Such criteria particularly impact pediatric BMT centers as most of them perform a relatively small number of BMTs. To determine whether patient volume is a valid marker of pediatric BMT center’s capabilities, the Pediatric Blood and Marrow Transplant Consortium (PBMTC) evaluated data from its registry to define the relationship between a pediatric transplant center’s patient volume and day þ 100 mortality. The analyses evaluated 2575 transplants from 60 centers reporting to the PBMTC between the years 2002 and 2004. The volume–outcome relationship was evaluated while adjusting for 46 independent data categories divided between nine variables that were known- or suspected-mortality risk factors. We found no association between transplant center volume and day þ 100 mortality in several analyses. A calculated intraclass correlation coefficient also indicated that differences in individual transplant center volume contributed to only 1% of the variance in day þ 100 mortality within the PBMTC. The results of this study suggest that factors other than transplant center volume contribute to variation in day þ 100 mortality among pediatric patients. Bone Marrow Transplantation (2013) 48, 514–522; doi:10.1038/bmt.2012.192; published online 12 November 2012 Keywords: pediatric; outcome; transplantation; mortality; volume; PBMTC

INTRODUCTION The ‘volume–outcome’ association refers to the relationship between the quantity of care that a hospital or physician provides and the quality of care that a patient receives. The volume– outcome relationship has been studied for a variety of disease processes and surgical conditions.1 Although studies have found that hospitals and physicians treating more patients, including cancer patients, may have better outcomes,2–4 other data support a clinically relevant relationship less consistently.5–9 This is particularly true among studies that have investigated volume– outcome relationships among hospitalized pediatric patients,10,11 including pediatric surgical cancer patients.12 In the context of pediatric solid organ transplantation, increased volume has been associated with better outcomes in the context of cardiac transplantation and found to be a surrogate marker for the use of anti-T cell induction therapy for renal transplantation, but was not associated with liver transplant outcomes done in the context of a Children’s Hospital.13–15 Volume remains a core criterion for pediatric BMT center accreditation by national and international organizations, including the Foundation for Accreditation of Cellular Therapy (FACT) and the Joint Accreditation Committee of the International Society for Cell Therapy.16 Payors for BMT also retain volume as a core criterion for ‘excellence’ as do state agencies (for example, California Children’s Services). Center minimum volume criteria is thus a significant issue not just for individual, smaller transplant centers, but also for clinical data accrual to national studies and, 1

potentially, for access to care by patients who would require significant travel if these procedures were available at only regional centers. Pediatric BMT procedures represent a unique clinical situation wherein collaboration between large numbers of centers is necessary to study relatively rare diseases.17 Limited patient numbers also compromise the evaluation of clinical situations, such as BMT procedures, that are impacted by multiple confounding variables. To date, data regarding pediatric BMTs are often derived from a single or limited institution experience, span a large time period and contain relatively homogeneous patient populations, or a combination, thereof.18–20 It is unclear whether the observations reported from these studies are representative of the broader pediatric BMT experience. Thus, a dense data set from a broadbased patient population is necessary to best investigate the volume–outcome relationship for pediatric BMT procedures. The Pediatric Blood and Marrow Transplant Consortium (PBMTC) is a voluntary collaboration among pediatric transplant centers in North America and, more currently, Australia. Between 2002 and 2004, the PBMTC made it a requirement that pediatric BMT centers that wished to maintain a membership submit patient information to the PBMTC registry. Although in place, the requirements for reporting resulted in a relatively high-quality database across a broad patient demographic that included patient information from centers that varied significantly in patient volume. The objective of this study was to evaluate the relationship between pediatric BMT center volume and day þ 100

Department of Pediatrics, University of California, Davis Children’s Hospital, Sacramento, CA, USA; 2Department of Pediatrics, Children’s Mercy Hospital, Kansas City, MO, USA; Department of Pediatrics, Division of Hematology/BMT, University of Utah School of Medicine/Primary Children’s Medical Center, Salt Lake City, UT, USA and 4Department of Pediatrics, British Columbia Children’s Hospital, Vancouver, British Columbia, Canada. Correspondence: Dr DS Taylor, Department of Pediatrics, University of California, Davis Children’s Hospital, 2516 Stockton Blvd, Sacramento, CA 95817, USA. E-mail: [email protected] Received 21 December 2011; revised and accepted 3 September 2012; published online 12 November 2012 3

Pediatric transplant volume–outcome relationship DS Taylor et al

515 mortality among a large cohort of transplant centers. Herein, we report a concentrated pediatric BMT experience and assess the contribution of individual center volume to the variability in day þ 100 mortality, using three different analyses and controlling for a large number of independent variables. MATERIALS AND METHODS Transplant centers Patient data were derived from 60 independent pediatric transplant centers participating in the PBMTC. Data reported from 2002 to 2004 to the Consortium are derived from all BMTs performed at participating centers, regardless of the patient’s enrollment in a clinical trial. Forty-eight of these centers participated in the National Marrow Donor Program, and thirty-two of the centers were approved by the Federation for the Accreditation of Cellular Therapy (FACT) during the study period. Transplant centers and their Principal Investigators are summarized in the Acknowledgements.

Patient population Included in the analysis were all BMT procedures reported to the PBMTC between 2002 and 2004, inclusive, representing the densest data set of transplants reported to the PBMTC registry. All patient cases were followed through a minimum of day þ 100 post transplant. Data submitted to the PBMTC were obtained directly from the same registration and day þ 100 data fields used by centers to submit to the Center for International Blood and Marrow Transplant Research. All unrelated donor transplant data were subject to audit by the National Marrow Donor Program. Although the PBMTC did not perform an independent audit, missing and inconsistent data fields were clarified through direct query to individual centers by PBMTC Registry personnel. A total of 2991 BMT procedures were conducted at 60 different pediatric BMT centers in North America and reported to the PBMTC during this period. Outcome data for survival at day þ 100 was available for 2718 (91%) patients. Excluded from the analysis were three patients from centers that reported o1 transplant per year during this time period (day þ 100 survival ¼ 100%), seven cases that lacked a patient date of birth and one case that lacked donor information. Also excluded from the analysis were transplants performed on patients older than 18 years of age. A total of 264 eligible patients who were p18 years of age (9%) lacked outcome data. The remaining 2575 patients were included in the analysis. One thousand four hundred and twenty (55.1%) of these procedures were conducted at FACT-accredited centers.

Data elements Data were collected prospectively and contributed to the PBMTC registry after obtaining informed consent from patient’s families and assent from patients, as appropriate, at each transplant center. Patient age was defined on the day of transplant. Pediatric BMT volume was defined as the average number of transplants per year. Engraftment was defined as the first of three consecutive days with an ANC 4500/mm3. The primary outcome variable was day þ 100 mortality and includes both transplant- and disease-related mortality. One-year mortality was not considered as an outcome variable as it is influenced by multiple disease- and managementspecific variables that were not included in the PBMTC data set. The investigators were blinded with respect to the association of specific data with the identity of the individual transplant centers. Forty-six independent data categories were selected a priori to investigate their relationship with day þ 100 mortality. These independent data categories were distributed between nine variables, as follows: three age groups (infant, o1 year; child, from 1 to 6 years; adolescent, from 7 to 18 years); five race–ethnicity groups (Caucasian, Asian, African American, Hispanic and others), two gender groups (male and female), seventeen diagnosis groups (ALL, AML, neuroblastoma, central nervous system tumor, immunodeficiency, myelodysplastic syndrome, aplastic anemia, Hodgkin’s disease, non-Hodgkin’s lymphoma, BM failure, metabolic disorders, hemoglobin/platelet disorders, chronic myelocytic leukemia, histiocytosis, other solid tumor, malignant disorders not otherwise specified and unknown diagnosis), two conditioning regimen groups (myeloablative and non-myeloablative), two graft manipulation groups (yes and no), eight donor type groups (autologous, syngeneic, HLA-matched sibling, HLA-matched other family member, HLA-mismatched sibling, HLA-matched unrelated donor, HLA-mismatched unrelated donor and allogeneic not otherwise specified), five graft source groups (BM, peripheral blood, cord blood, combination and unknown) and two repeat transplant groups (repeat autologous and repeat allogeneic). & 2013 Macmillan Publishers Limited

Statistical analysis Univariable, standard multivariable and hierarchical multivariable analyses were used to assess the volume–outcome relationship while adjusting for 46 independent data categories divided between nine variables that were known- or suspected-mortality risk factors, as detailed above. Continuous variables were analyzed using the Student’s t-test, and categorical data were analyzed using w2-analysis. We also tested for colinearity between independent variables and found no statistically significant interactions between them. Variables that included a category of ‘Unknown’ were included in the analysis and were further tested with simulations to confirm that their presence did not influence the significance of other categories within that variable (data not shown). It was decided, a priori, that all independent variables would be included in the final multivariable analyses to retain their influence, regardless of significance in the univariable analyses. To test the association between the primary outcome variable (day þ 100 mortality) and the primary independent variable of interest (transplant center volume), univariable and multivariable logistic regression analyses were conducted. We considered the association of volume as follows: volume as a continuous variable; volume as a categorical variable in seven equally distributed volume categories; volume as a dichotomous variable, using 20 total transplants per year, 10 total transplants per year, 10 allogeneic transplants per year and 10 ‘first-time’ allogeneic transplants per year as thresholds; and volume divided by quintile. We selected 20 transplants per year as a division point because this volume has been used as a minimum volume threshold for transplant ‘Center of Excellence’ designation by some third party payors and reimbursement by some state agencies (for example, California Children’s Services), and 10 first-time allogeneic transplants per year as a division point consistent with the FACT accreditation requirements for pediatric BMT centers. A hierarchical logistic regression using a random intercept model was also performed to adjust for patient ‘clustering.’ In this model, patients represented the primary level and hospitals (transplant centers) represented the secondary level. Further evaluation of individual transplant center contribution to day þ 100 mortality was accomplished by assessment of the intraclass correlation (ICC) using the analysis of variance method.21 This statistic provides an indication of the contribution of individual transplant centers to the total variation in outcome (day þ 100 mortality). The association of FACT accreditation with day þ 100 mortality was also investigated. These analyses were based on first-time allogeneic transplant volume. All patient information was managed consistent with the Health Insurance Portability and Accountability Act and the University of California Institution Review Board guidelines. Statistical analyses were conducted using Stata version 11 (StatCorp, LP, College Station, TX, USA) and R.

RESULTS Transplant and patient characteristics The 2575 cases evaluated include 893 (35%) autologous transplants, 1671 (65%) allogeneic transplants and 11 (0.4%) syngeneic transplants (Table 1). These data reflect a slightly higher frequency of unrelated donor (887 or 53%) as compared with related donor (713 or 43%) allogeneic transplants. ALL, AML and neuroblastoma were the most common diagnoses and, together, accounted for 51% of all transplants performed. The seven most common disease categories accounted for 74% of the transplants performed (Table 1). The mean age at transplant was 7.5 years. The median age was 7.5 years, with an interquartile range of 6 years and 13 years. Day þ 100 mortality Overall day þ 100 mortality rates for selected patient populations and center volume stratifications are shown in Table 2a and Table 2b, respectively. Thirty-six (60%) of the 60 centers included in the analysis performed o20 transplant procedures per year. These centers accounted for 1008 (39%) of the 2575 transplant procedures analyzed. The overall day þ 100 observed mortality was 14.3% with no clear association with center volume (Figure 1). Bone Marrow Transplantation (2013) 514 – 522


516 Table 1.

Selected patient populations and day þ 100 mortality

Table 2a.

BMT patient and transplant characteristics

BMT/patient characteristic

Number (%)

Patient population

N

Stem cell source BM Peripheral blood Umbilical cord blood Combination Not reported

888 1193 450 24 20

(34.5) (46.3) (17.5) (0.9) (0.8)

All BMT Autologous BMT Allogeneic BMT Matched sibling donor BMT Related donor BMT Unrelated donor BMT Umbilical cord BMT

2575 893 1671 590 713 887 450

Donor type Autologous Syngeneic Allogeneic no. (% total, % allogeneic) HLA identical sibling HLA-matched other relative HLA-mismatched sibling/relative HLA-matched unrelated donor HLA-mismatched unrelated donor Allogeneic, not otherwise specified

893 11 1671 590 39 84 449 438 71

(34.7) (0.4) (64.9, 100) (22.9, 35.3) (1.5, 2.3) (3.3, 5.0) (17.4, 26.9) (17.0, 26.2) (2.8, 4.2)

Conditioning Myeloablative Non-myeloablative

2442 (94.8) 133 (5.2)

Major diagnoses ALL AML Neuroblastoma Central nervous system tumor Immunodeficiency Myelodyspastic syndrome Other solid tumor

514 398 393 204 134 132 130

(20) (15.5) (15.3 (7.9) (5.2) (5.1) (5.0)

Mortality (%) 14.3 7.3 18.1 9.3 10.7 23.7 22.4

Transplant center volume and day þ 100 mortality

Table 2b.

Volume (BMT per year)

Centers (N)

40 to o10 10 to o20 20 to o30 30 to o40 40 to o50 50 to o60 60 to o70 Total

Transplants performed, N (%)

10 26 10 8 3 0 3 60

176 832 427 644 267

(6.8) (32.3) (16.6) (25) (10.4)

229 (8.9) 2575 (100)

Mortality, N (%) 22 118 70 84 32

(12.5) (14.2) (16.4) (13.0) (12.0)

43 (18.8) 369 (14.3)

45 40

Age (in years) Mean, s.d. Median 25th percentile, 75th percentile

1508 (58.6) 1045 (40.6) 22 (0.9) 7.5, 5.5 6 3, 13

Mortality (%)

35 Gender Male Female Not reported

30 25 20 15 Mean mortality = 14.3%

10 5 0 0

Age by groups (in years) Infant (o1) Child (1–6) Adolescent (7–18)

160 (6.2) 1143 (44.4) 1272 (49.4)

Ethnicity Caucasian Asian African American Hispanic Others Not reported

1649 74 253 329 141 129

Outcome Mean day to engraftment, s.d. Median day to engraftment 25th percentile, 75th percentile Day þ 100 mortality

18.1, 25.8 15 11, 20 369 (14.3)

(64.0) (2.9) (9.8) (12.8) (5.5) (5)

Impact of center size on day 100 mortality As shown in Table 3, statistical significance between transplant center volume and day þ 100 mortality was observed twice, but only in the univariable analysis. This was first observed when volume was grouped into seven volume categories, wherein the largest volume centers had a higher day þ 100 mortality rate (odds ratio (OR) ¼ 1.43, 95% confidence intervals (CI) 1.01–2.04, P ¼ 0.05). The second association was observed when volume was assessed by quartile, wherein the fourth quintile had a lower Bone Marrow Transplantation (2013) 514 – 522

10

20

30

40

50

60

70

Average BMTs per year

Figure 1. The average center volume (abscissa) and average day þ 100 mortality (ordinate) during the study period for an individual transplant center is represented as a single diamond. The mean day þ 100 mortality of all transplants performed is represented by a horizontal line. A full color version of this figure is available at the Bone Marrow Transplantation journal online.

day þ 100 mortality rate (OR ¼ 0.75, 95% CI 0.56–1.00, P ¼ 0.05). However, when the model adjusted for 46 data categories in both the standard multivariable analysis and in the two-level hierarchical multivariable analysis, there was no association found between transplant center volume and day þ 100 mortality, regardless of volume category (Table 3). Subgroup analyses were performed to evaluate whether volume-associated mortality differences were evident when only allogeneic transplants were considered. Results indicate no association between center volume and day þ 100 mortality, whether volume was considered as a continuous variable, categorized into four volume groups, assessed by quintile dichotomized at less than or X20 total transplants per year or dichotomized at less than or X10 allogeneic transplants per year (Table 4). Similarly, there was no association between transplant center volume and day þ 100 mortality when 971 unrelated donor transplant, mismatched unrelated donor and mismatched related donor transplants were considered, as shown in Table 5. & 2013 Macmillan Publishers Limited


517 Table 3.

BMT day þ 100 mortality: association of transplant center volume in univariable, multivariable and hierarchical multivariable analyses

Transplant center volume

Continuous 0 to 60 per year

Patients (N)

Observed mortality (%) day þ 100

Adjusted multivariable analysisb

Univariable analysis (unadjusted for covariatesa)

Two-level hierarchical adjusted multivariable analysisb

OR

95% CI

P

OR

95% CI

P

OR

95% CI

P

2575

14.3

1.04

0.96–1.13

0.33

1.00

0.92–1.10

0.97

0.98

0.87–1.10

0.73

Categorical o10 per year 10 to o20 per year 20 to o30 per year 30 to o40 per year 40 to o50 per year 50 to o60 per year 60 to o70 per year

176 832 427 644 267 0 229

12.5 14.2 16.4 13.0 12.0

0.84 0.98 1.21 0.86 0.80

0.53–1.33 0.78–1.24 0.91–1.61 0.67–1.13 0.54–1.17

0.47 0.88 0.18 0.28 0.25

Reference 1.05 1.18 0.87 0.79

0.61–1.81 0.67–2.09 0.49–1.51 0.41–1.50

0.85 0.67 0.49 0.41

Reference 1.05 1.17 0.87 0.76

0.60–1.84 0.64–2.14 0.48–1.58 0.38–1.53

0.87 0.60 0.65 0.44

18.8

1.43

1.01–2.04

0.05

1.29

0.69–1.54

0.42

1.23

0.61–2.51

0.56

Quintilesc 1st 2nd 3rd 4th 5th

515 515 515 515 515

12.6 15.7 16.3 11.7 15.3

0.83 1.15 1.21 0.75 1.11

0.63–1.11 0.88–1.5 0.93–1.58 0.56–1.00 0.84–1.45

0.22 0.31 0.15 0.05 0.47

Reference 1.27 1.18 0.81 1.09

0.87–1.85 0.81–1.71 0.54–1.21 0.74–1.61

0.21 0.38 0.31 0.67

Reference 1.28 1.18 0.80 1.05

0.85–1.93 0.78–1.78 0.51–1.26 0.67–1.65

0.24 0.43 0.34 0.83

Dichotomous 10d 0 to o10 per year X10 per year

176 2399

12.5 14.5

1 1.18

0.75–1.88

0.47

Reference 1.02

0.61–1.71

0.94

Reference 1.01

0.58–1.75

0.97

Dichotomous 20e 0 to o20 per year X20 per year

1008 1567

13.9 14.6

1 1.06

0.85–1.33

0.61

Reference 0.95

0.74–1.22

0.70

Reference 0.95

0.71–1.27

0.72

a

Abbreviations: CI ¼ confidence interval; OR ¼ odds ratio. Univariable analysis: odds ratios are relative to all other categories when greater than two categories and lowest volume category when only two categories. bAdjusted multivariable analysis and two-level hierarchical adjusted multivariable: odds ratios are relative to lowest volume category, while simultaneously including all other independent variables in the multivariable analysis. cQuintiles determined according to total number of transplants performed per year. dDichotomous 10-transplant centers were divided between those that performed o10 transplants per year and those that performed X10 transplants per year. eDichotomous 20-transplant centers were divided between those that performed o20 transplants per year of any type and those that performed X20 transplants per year of any type. Results of only the allogeneic transplants are analyzed.

Equivalent results were obtained if volume was dichotomized at five allogeneic transplants per year or based on total transplant volume (allogeneic and autologous), even though the lowest volume centers in these analyses performed o5 allogeneic BMT per year (data not shown). In addition, an ICC was calculated and found to be 0.0108 (95% CI 0.00–0.02), indicating that differences between transplant center volumes contributed 1.1% of the overall observed differences in day þ 100 mortality after adjusting for other known- and suspected-mortality risk factors. Impact of FACT accreditation This analysis included only the results from 1441 ‘first-time’ allogeneic transplants. As shown in Table 6, transplant center volume was associated with decreased day þ 100 mortality only in the univariable analysis if volume was defined by quintiles, but not after adjusting for other independent variables, as shown in the hierarchical multivariable analysis. Six hundred and thirty (43.7%) of these transplants were completed at centers that were not FACT accredited and had an overall day þ 100 mortality of 17%. There was no association between FACT accreditation and day þ 100 mortality in either the univariable or two-level hierarchical multivariable analysis (OR ¼ 1.12, 95% CI 0.86–1.48 and OR ¼ 1.02, 95% CI 0.75–1.40, respectively). DISCUSSION The three primary findings from our analyses are as follows: (1) small volume transplant centers contribute a significant & 2013 Macmillan Publishers Limited

percentage of pediatric transplants within the PBMTC consortium, (2) there is no association between transplant center volume and day þ 100 mortality and (3) among the 60 PBMTC participating centers contributing to this data set, the variability in day þ 100 mortality is 98.9% dependent on factors not specific to individual center size or their accreditation. Although Loberiza et al.22 suggested that center-specific factors, such as 420 allogeneic transplants per physician per year, correlates with an improved day þ 100 and 1-year mortality in adult allogeneic transplants for leukemia, the current analysis suggests that these factors are either similar between centers within the PBMTC or are not important in the pediatric transplant population. The European Blood and Marrow Transplant (EBMT) pediatric BMT volume–outcome relationship has also been reported. Klingebiel et al.18 analyzed 102 haploidentical BMT procedures done at 36 centers over 10 years for pediatric patients with very high-risk AML. Although center volume o23.1 transplants per year did not impact non-relapse mortality, a limited multivariable analysis suggested a correlation with decreased 5-year leukemiafree survival secondary to increased relapse incidence. It remains unclear whether this difference was a direct result of disease variables, the transplant center or differences in post-transplant patient management. Miano et al.19 analyzed 31 713 pediatric transplants done over 32 years. In a multivariable analysis containing four dichotomous variables, a small, increased risk of day þ 100 mortality was noted for centers with volume o10 versus those performing at least 10 allogeneic transplants per year. Although the limited study population of the former report precluded a more extensive analysis, inclusion of more variables in Bone Marrow Transplantation (2013) 514 – 522


518 Table 4.

Transplant center volume and allogeneic transplant day þ 100 mortality: univariable and hierarchical multivariable analyses

Transplant center volumea

Allogeneic transplants, N (%)

Observed mortality day þ 100, N (%)

Two-level hierarchical adjusted multivariable analysisc

Univariable analysis (unadjusted for covariatesb) OR

95% CI

P

OR

95% CI

P

302 (18.1)

1.05

0.89–1.24

0.58

0.97

0.78–1.20

0.78


1671 (100)

Categoricald o10 per year 10 to o20 per year 20 to o30/year 30 to o40 per year

341 624 515 191

(20.4) (37.3) (30.8) (11.4)

58 119 83 42

(17.0) (19.1) (16.1) (22.0)

0.91 1.11 0.82 1.32

0.67–1.25 0.86–1.44 0.62–1.08 0.92–1.91

0.57 0.41 0.17 0.14

Reference 1.11 0.89 1.05

0.74–1.64 0.57–1.39 0.57–1.94

0.62 0.60 0.88

Quintilesd 1st 2nd 3rd 4th 5th

334 334 334 334 335

(20.0) (20.0) (20.0) (20.0) (20.0)

54 64 69 53 62

(16.2) (19.2) (20.7) (15.9) (18.5)

0.85 1.09 1.23 0.82 1.04

0.61–1.17 0.81–1.49 0.91–1.67 0.60–1.14 0.76–1.41

0.31 0.56 0.17 0.24 0.82

Reference 1.19 1.32 0.90 0.94

0.75–1.87 0.84–2.08 0.55–1.49 0.56–1.58

0.46 0.23 0.69 0.83

Dichotomous 10d o10 per year X10 per year

341 (20.4) 1330 (79.6)

58 (17.0) 302 (18.4)

1 1.10

0.80–1.50

0.57

Reference 1.02

0.71–1.48

0.91

Dichotomous 20e o20 per year X20 per year

607 (36.3) 1064 (63.7)

106 (17.5) 196 (18.4)

1 1.07

0.82–1.39

0.63

Reference 1.00

0.72–1.37

0.98

a

b

Abbreviations: CI ¼ confidence interval; OR ¼ odds ratio. Number of allogeneic transplants per year. Univariable analysis: odds ratios are relative to all other categories when greater than two categories and lowest volume category when only two categories. cAdjusted multivariable analysis: data stratified by transplant center using random intercept model. Odds ratios are relative to lowest volume category, while simultaneously including all other independent variables in the multivariable analysis. The multivariable analysis was performed as described in Materials and methods in the absence of the data category of autologous transplant and repeat autologous transplant. In addition, the diagnosis category of central nervous system tumor was dropped due to colinearity. d Categorical, quintiles and dichotomous 10-transplant centers were divided according to number of allogeneic transplants performed per year by increments of 10 transplants per year (categorical), by quintile (quintiles) or between those that performed o10 allogeneic transplants per year and those that performed X10 allogeneic transplants per year (dichotomous 10). eDichotomous 20-transplant centers were divided between those that performed o20 transplants per year of any type and those that performed X20 transplants per year of any type. Results of only the allogeneic transplants are analyzed.

the latter multivariable analysis may have better excluded the possibility of a type 1 error. Corroboration of the EBMT center volume findings in these two studies using either a hierarchical analysis at the level of the center or calculation of an ICC, as provided in the present analysis, would help validate their observations. Nevertheless, it is interesting to speculate that volume may be a surrogate marker for differences that exist between EBMT centers that do not exist between PBMTC centers. The day þ 100 mortality rates reported in the PBMTC database are consistent with mortality from BMTs reported by others. The day þ 100 overall mortality rate (including both disease- and transplant-related mortality) for the 887 unrelated donor transplants was 23.7% in this data set, and is similar to the Seattle experience of 26.1% day þ 100 overall mortality rate for 88 children treated with unrelated marrow transplantation for ALL between 1987 and 1999, and the 21% transplant-related mortality reported for patients grafted after 1998 by Dini et al.23,24 The 18.1% overall mortality (including both disease- and transplantrelated mortality) for all allogeneic BMT in this PBMTC cohort is also consistent with a 13% day þ 100 transplant-related mortality reported by the EBMT during the same time period.19 The PBMTC overall day þ 100 mortality rate of 22.4% for 450 umbilical cord transplants is also consistent with the 23–39% day þ 100 transplant-related mortality rate among a total of 192 similar transplants performed between 1994 and 2002, as reviewed by Rocha et al.25 on behalf of the Eurocord and the EBMT Group. Thus, the transplant day þ 100 survival rates within PBMTC participating centers are comparable to other institutions and consortia, and indicates that greater inclusion of smaller transplant Bone Marrow Transplantation (2013) 514 – 522

centers, as modeled by the PBMTC, does not worsen reported transplant results. The current analyses demonstrate the utility of structured, aggregate data regarding pediatric BMT procedures and their outcomes, and highlight the implicit challenge in studying this diverse cohort with uncommon diseases that are geographically widespread. The implementation of mechanisms to insure greater inclusion of smaller centers can augment patient accrual to clinical trials, and insure that these procedures are available to meet patient needs and increases direct and long-term care of transplant patients by their primary transplant facility. An important strength of this study is inclusion of many patient level data variables relevant to patient BMT outcomes. We included 46 categories divided between 9 variables, not just those thought to be significant based on a univariable analysis, to adjust for such factors that are both known and suspected to contribute to risk of day þ 100 mortality. Sub-analyses of higher risk patient populations within this data set also failed to demonstrate a volume–outcome relationship. The lack of a center effect was corroborated by both a hierarchical multivariable analysis and calculation of an ICC. Although the inclusion of a large number of independent variables may decrease the risk of a type 1 error, the 95% CI can be larger than would be present in a more limited evaluation and possibly obfuscate the presence of small, but significant differences.26 There are several potential limitations of this study. Despite internal validity checks of data that are submitted to the PBMTC, these data were not subjected to an independent audit. Regardless, we believe that the observations noted herein are & 2013 Macmillan Publishers Limited


519 Table 5.

Transplant center volume versus URD and mRD or mURD transplant day þ 100 mortality: univariable and hierarchical multivariable analyses


URD/mURD/mRD transplants, N (%)




971 (100)

226 (23.3)

95% CI

P

Two-level hierarchical adjusted multivariable analysisc OR

95% CI

P

0.98

0.81–1.19

0.83

0.98

0.75–1.28

0.87

Categoricald o10 per year 10 to o20 per year 20 to o30 per year 30 to o40 per year

170 337 323 141

(17.5) (34.7) (33.3) (14.5)

36 88 66 36

(21.2) (26.1) (20.4) (25.5)

0.86 1.27 0.78 1.15

0.58–1.29 0.93–1.73 0.57–1.08 0.77–1.74

0.48 0.13 0.14 0.49

Reference 1.31 1.01 1.21

0.79–2.18 0.58–1.77 0.57–2.61

0.29 0.97 0.62

Quintilese 1st 2nd 3rd 4th 5th

194 194 194 194 195

(20.0) (20.0) (20.0) (20.0) (20.0)

42 53 47 39 45

(21.7) (27.3) (24.2) (20.1) (23.3)

0.89 1.31 1.07 0.79 0.99

0.61–1.30 0.92–1.88 0.74–1.54 0.54–1.17 0.68–1.43

0.55 0.14 0.73 0.24 0.94

Reference 1.30 1.08 0.97 0.95

0.75–2.25 0.62–1.90 0.53–1.79 0.49–1.83

0.35 0.79 0.93 0.87

Dichotomous 10d o10 per year X10 per year

170 (17.5) 801 (82.5)

36 (21.2) 190 (23.3)

1 1.16

0.77–1.73

0.48

Reference 1.19

0.74–1.91

0.47

Dichotomous 20f o20 per year X20 per year

303 (31.2) 668 (68.8)

73 (24.1) 153 (22.9)

1 0.94

0.68–1.29

0.69

Reference 0.96

0.64–1.42

0.83

Abbreviatons: CI ¼ confidence interval; mRD ¼ mismatched related donor; mURD ¼ mismatched unrelated donor; OR ¼ odds ratio; URD ¼ unrelated donor. a Number of allogeneic transplants per year. bUnivariable analysis: odds ratios are relative to all other categories when greater than two categories and lowest volume category when only two categories. cAdjusted multivariable analysis: data stratified by transplant center using random intercept model. Odds ratios are relative to lowest volume category, while simultaneously including all other independent variables in the multivariable analysis. The multivariable analysis was performed as described in Materials and methods in the absence of the data category of autologous transplant and repeat autologous transplant. In addition, the diagnosis category of central nervous system tumor was dropped due to colinearity. dCategorical and dichotomous 10-transplant centers were divided according to number of allogeneic transplants performed per year by increments of 10 transplants per year (categorical) or between those that performed o10 and those that performed X10 allogeneic transplants per year (dichotomous 10). Results of only the URD, mURD and mRD transplants are analyzed. eQuintiles determined according to number of URD, mURD and mRD transplants performed per year. fDichotomous 20-transplant centers were divided between those that performed o20 transplants per year of any type and those that performed X20 transplants per year of any type. Results of only the URD, mURD and mRD transplants are analyzed.

not impacted. First, 434% of the transplants reported (unrelated donor transplants) were subject to audit by the National Marrow Donor Program. Second, the data submitted to the PBMTC were derived directly from the same registration and day þ 100 data fields that centers submitted to the Center for International Blood and Marrow Transplant Research; data that were also audited. Third, the primary outcome variable, day þ 100 mortality, is a highly objective measure unlikely compromised by subjective interpretation, and mortality rates within the PBMTC data set are consistent with those reported by others, as outlined above. Fourth, the observed lack of variability in risk-adjusted outcomes attributable to transplant centers, as observed by both a low ICC and no statistical evidence of a volume–outcome relationship in several analyses, supports the integrity of the data despite absence of a formal audit. A second potential limitation is the necessary exclusion of 264 cases, for lack of outcome data and their possible impact on the results reported herein. A detailed comparison of the included and excluded cases revealed a small, statistically significant difference in only the stem cell source and age variables (data not shown). Within these variables, the infant age category had the greatest impact on the results secondary to its importance as an independent risk factor, and skewed distribution between the volume categories with excluded data disproportionately from the 30–40 BMT per year category. Although this may have resulted in a small improvement in the observed day þ 100 outcome of the 30–40 BMT per year volume category, simulations indicated that & 2013 Macmillan Publishers Limited

there were no conditions under which this variable could change the significance of this volume category or alter the conclusions of this study. All other categories for which there was a significant difference between included and excluded populations were less important as an independent risk factor, more evenly distributed between the volume categories, or both. The outcome variable used in the current study (day þ 100 mortality) is a common endpoint that is generally less influenced by the natural history of the underlying diseases, differences in post-transplant supportive care delivered by non-transplant services and more reflective of acute transplant morbidities. The fact that there was little change in our results despite three different analyses evaluating the impact of transplant center volume is most likely reflective of higher variability between patient-specific factors than between center-specific factors within the PBMTC. Although we did not find an overall average volume–outcome relationship among 60 pediatric BMT centers in the PBMTC database, there could still be individual centers that have higher or lower than expected day þ 100 mortality rates. Our analyses do not suggest that all centers are equal; rather, our analyses conclude that on average, transplant center volume does not significantly contribute to differences in PBMTC center performance as measured by day þ 100 mortality. The current findings suggest that the majority of differences in day þ 100 mortality observed between centers are attributable to identifiable patient- and transplantspecific variables rather than variability between centers. These results may reflect similarity in environment and practice patterns Bone Marrow Transplantation (2013) 514 – 522


520 Table 6.

First allogeneic transplant day þ 100 mortality: univariable and hierarchical multivariable analyses


First allogeneic transplant N (%)



Continuous 40 to 40 per year Quintilesd 1st 2nd 3rd 4th 5th Dichotomous 10e o10 per year X10 per year FACT accreditation No Yes

1441 (100)

P

OR

95% CI

P

0.98

0.72–1.10

0.28

0.83

0.65–1.06

0.14

(16.3) (20.1) (20.8) (19.1) (13.5)

0.86 1.19 1.26 1.10 0.66

0.61–1.22 0.86–1.65 0.91–1.74 0.79–1.53 0.46–0.95

0.41 0.29 0.58 0.38 0.03

Reference 1.33 1.36 1.21 0.71

0.84–2.12 0.86–2.14 0.75–1.93 0.43–1.17

0.22 0.19 0.43 0.18

430 (29.8) 1011 (70.2)

72 (16.7) 187 (18.5)

1 1.12

0.84–1.52

0.43

Reference 1.10

0.79–1.54

0.57

630 (43.7) 811 (56.3)

107 (17.0) 152 (18.7)

1 1.12

0.86–1.48

0.39

Reference 1.02

0.75–1.40

0.86

288 288 288 288 289

(20.0) (20.0) (20.0) (20.0) (20.1)

259 (18.0)

95% CI

Two-level hierarchical adjusted multivariable analysisc

47 58 60 55 39

Abbreviations: CI ¼ confidence interval; FACT ¼ Foundation for Accreditation of Cellular Therapy; OR ¼ odds ratio. aNumber of first allogeneic transplants per year. bUnivariable analysis: odds ratios are relative to all other categories when greater than two categories and lowest volume category when only two categories. cAdjusted multivariable analysis: data stratified by transplant center using random intercept model. Odds ratios are relative to lowest volume category, while simultaneously including all other independent variables in the multivariable analysis. The multivariable analysis was performed as described in the Materials and methods in the absence of the data categories of autologous transplant, repeat allogeneic transplant, repeat autologous transplant, malignant disorder not otherwise specified and central nervous system tumor for the lack of data elements. dQuintiles determined according to number of first allogeneic transplants performed per year. eDichotomous 10-transplant centers were divided between those that performed o10 first allogeneic donor transplants per year and those that performed X10 first allogeneic transplants per year.

between PBMTC transplant centers that result from all programs being within tertiary care pediatric institutions, and that the transplant procedure is not dependent on a mechanical skill set that requires repetition to maintain proficiency. The current findings are most consistent with recent literature evaluating the volume–outcome relationship for acute hospital admissions and medical procedures that are amenable to evidence-based guidelines and clinical pathways.5,7,8,10–12 These data support the de-emphasis of center volume as a criterion for accreditation or determination of excellence and support risk-adjusted analysis of individual center performance. Further analyses are required to better determine those specific factors that positively contribute to variability in day þ 100 mortality and a similarly detailed analysis employing a more comprehensive data set to assess impact of center volume on 1-year mortality. CONFLICT OF INTEREST The authors declare no conflict of interest.

ACKNOWLEDGEMENTS This work was funded in part by a grant from the Children’s Miracle Network. Voluntary submission of transplant data to the PBMTC by Principal Investigators at participating institutions is gratefully acknowledged and summarized, below.

Participating BMT center Alberta Children’s Hospital, Calgary, Alberta1 All Children’s Hospital, Saint Petersburg, FL1 Ann and Robert H Lurie Children’s Hospital of Chicago, Chicago, IL1 British Columbia Children’s Hospital, Vancouver, BC1 Bone Marrow Transplantation (2013) 514 – 522

Pediatric Principal Investigator* Lewis, Victor Nieder, Michael Duerst, Reggie Davis, Jeffrey

Participating BMT center

Pediatric Principal Investigator*

CancerCare Manitoba, Winnipeg, Manitoba2 Cardinal Glennon Children’s Medical Center, St. Louis, MO1 Children’s Healthcare of Atlanta, Atlanta, GA1 Children’s Hospital, New Orleans, LA1 Children’s Hospital Medical Center of Akron, Akron, OH1 Children’s Hospital of Pittsburgh of UPMC, Pittsburgh, PA1 Children’s Hospital and Research Center at Oakland1 Children’s Hospitals and Clinics of Minnesota, Minneapolis, MN2 Children’s Mercy Hospitals and Clinics. Kansas City, MO1 Children’s National Medical Center, Washington, DC1 Cincinnati Children’s Hospital Medical Center1 Columbia University Medical Center, New York, NY1 Cook Children’s Medical Center, Fort Worth, TX1 CS Mott Children’s Hospital, Ann Arbor, MI2 Floating Hospital for Children at Tufts Medical Center, Boston, MA2 Helen DeVos Children’s Hospital at Spectrum Health, Grand Rapids, MI1 Hackensack University Medical Center, Hackensack, NJ2 Hospital Sainte-Justine, Montreal, Quebec1 Hospital for Sick Children, Toronto, Ontario1 Kosair Children’s Hospital, Louisville, KY2 Loma Linda University Medical Center, Loma Linda, CA1 Lucille Packard Children’s Hospital Stanford University, Palo Alto, CA2

Schroeder, Marlis Ferguson, William Haight, Ann Yu, Lolie Kuerbitz, Steven Goyal, Rakesh Walters, Mark Richards, Michael Dalal, Jignesh Kamani, Naynesh Davies, Stella Garvin, James Eames, Gretchen Levine, John Grodman, Howard Pietryga, Daniel Brochstein, Joel Duval, Michel Doyle, John Cheerva, Alexa Morris, Joan Agarwal, Rajni

& 2013 Macmillan Publishers Limited


521 Participating BMT center

Pediatric Principal Investigator*

Mayo Clinic, Rochester, MN2 MD Anderson Cancer Center, Houston, TX2 Methodist Children’s Hospital of South Texas, San Antonio, TX2 Miami Children’s Hospital, Miami, FL1 Nationwide Children’s Hospital, Columbus, OH1 Nemours Children’s Clinic, Jacksonville, FL1 New York Medical College, Valhalla, NY2 New York University Langone Medical Center, New York, NY2 Oregon Health and Science University, Portland, OR2 Penn State Hershey Children’s Hospital, Hershey, PA2 Primary Children’s Medical Center, Salt Lake City, UT2 Princess Margaret Hospital for Children, Perth, Western Australia1 Rady Children’s Hospital, San Diego, CA1 Rainbow Babies and Children’s Hospital, Cleveland, OH2 Riley Hospital for Children, Indianapolis, IN2 Roswell Park Cancer Institute, Buffalo, NY2 St. Louis Children’s Hospital, St. Louis, MO2 Starship Children’s Hospital, Auckland, New Zealand1 The Montreal Children’s Hospital of the MUHC, Montreal, Quebec2 The Steven and Alexandra Cohen Children’s Medical Center of NY, New Hyde Park, NY1 University of Alabama at Birmingham2 University of California—Davis, Sacramento, CA2 University of California, San Francisco Medical Center, San Francisco, CA2 University of Iowa Hospitals and Clinics, Iowa City, IA2 University of Miami Miller School of Medicine— Sylvester Cancer Center, Miami, FL1 University of Minnesota Medical Center, Minneapolis, MN2 University of Mississippi Medical Center, Jackson, MS2 University of Nebraska Medical Center, Omaha, NB2 University of Oklahoma Health Sciences Center, Oklahoma City, OK2 University of Rochester, Rochester, NY2 University of Texas Southwestern Medical Center, Dallas, TX2 University of Wisconsin Hospitals & Clinics, Madison, WI2 Vanderbilt University, Nashville, TN2 Wayne State University, Detroit, MI2

Khan, Shakila Worth, Laura Grimley, Michael Fort, John Termuhlen, Amanda Joyce, Michael Ozkaynak, Fevzi Gardner, Sharon Nemecek, Eneida Lucas, Ken Pulsipher, Michael Cole, Cathy Kadota, Richard Weirsma, Susan Haut, Paul Bambach, Barbara Shenoy, Shalini Teague, Lochie Mitchell, David Sahdev, Indira Sande, Jane Taylor, Douglas Cowan, Morton Goldman, Fred Kleiner, Gary Baker, Scott Megason, Gail Gordon, Bruce Selby, George Mullen, Craig Aquino, Victor DeSantes, Ken Frangoul, Haydar Savasan, Sureyya

*-The Pediatric Principal Investigator of the BMT Program at the time of data submission to the PBMTC is acknowledged. 1-Indicates Pediatric BMT Center 2-Indicates Joint Adult and Pediatric BMT Center

Author contributions: DST is responsible for project concept, experimental design, database structure for analysis, as well as a co-participant in data analysis, critical interpretation of results and principal author of primary manuscript. MD introduced the ICC to the methods, performed the associated statistical analysis and provided critical participation to the writing of the final manuscript. EUS is responsible for extensive data organization, database management and literature research. RR provided data elements for the database and was responsible for internal data quality control. MAP provided critical interpretation of results and review of the manuscript. AG provided critical structure to the PBMTC to permit the organized collection of patient data, initiated the PBMTC data registry, provided critical interpretation of results and review of the final manuscript. KS provided PBMTC leadership that insured data submission to data registry, oversight of center


participation in data registry, critical input to study design, result interpretation and review of the manuscript. JPM provided primary statistical expertise and mentoring necessary to perform data analysis, critical assistance in the design of the analysis and interpretation of the results. This author also contributed significantly to writing both the primary and final manuscripts.

REFERENCES 1 Shahian DM, Normand SL. The volume-outcome relationship: from Luft to Leapfrog. Ann Thorac Surg 2003; 75: 1048–1058. 2 Gooiker GA, van Gijn W, Post PN, van de Velde CJ, Tollenaar RA, Wouters MW. A systematic review and meta-analysis of the volume-outcome relationship in the surgical treatment of breast cancer. Are breast cancer patients better of with a high volume provider? Eur J Surg Oncol 2010; 36(Suppl 1): S27–S35. 3 Borowski DW, Bradburn DM, Mills SJ, Bharathan B, Wilson RG, Ratcliffe AA et al. Volume-outcome analysis of colorectal cancer-related outcomes. Br J Surg 2010; 97: 1416–1430. 4 Lipscomb J. Transcending the volume-outcome relationship in cancer care. J Natl Cancer Inst 2006; 98: 151–154. 5 Yasunaga H, Matsuyama Y, Ohe K. Volume-outcome relationship in rectal cancer surgery: a new perspective. Surg Today 2009; 39: 663–668. 6 Christiansen CL, Morris CN. Improving the statistical approach to health care provider profiling. Ann Intern Med 1997; 127(8 Pt 2): 764–768. 7 Kalant N, Shrier I. Volume and outcome of coronary artery bypass graft surgery: are more and less the same? Can J Cardiol 2004; 20: 81–86. 8 Marcin JP, Li Z, Kravitz RL, Dai JJ, Rocke DM, Romano PS. The CABG surgery volume-outcome relationship: temporal trends and selection effects in California, 1998-2004. Health Serv Res 2008; 43(1 Pt 1): 174–192. 9 Peterson ED, Coombs LP, DeLong ER, Haan CK, Ferguson TB. Procedural volume as a marker of quality for CABG surgery. JAMA 2004; 291: 195–201. 10 Bazzani LG, Marcin JP. Case volume and mortality in pediatric cardiac surgery patients in California, 1998-2003. Circulation 2007; 115: 2652–2659. 11 Marcin JP, Song J, Leigh JP. The impact of pediatric intensive care unit volume on mortality: a hierarchical instrumental variable analysis. Pediatr Crit Care Med 2005; 6: 136–141. 12 Gutierrez JC, Koniaris LG, Cheung MC, Byrne MM, Fischer AC, Sola JE. Cancer care in the pediatric surgical patient: a paradigm to abolish volume-outcome disparities in surgery. Surgery 2009; 145: 76–85. 13 Davies RR, Russo MJ, Hong KN, Mital S, Mosca RS, Quaegebeur JM et al. Increased short- and long-term mortality at low-volume pediatric heart transplant centers: should minimum standards be set? Retrospective data analysis. Ann Surg 2011; 253: 393–401. 14 Schurman SJ, Stablein DM, Perlman SA, Warady BA. Center volume effects in pediatric renal transplantation. A report of the North American Pediatric Renal Transplant Cooperative Study. Pediatr Nephrol 1999; 13: 373–378. 15 Tracy ET, Bennett KM, Danko ME, Diesen DL, Westmoreland TJ, Kuo PC et al. Low volume is associated with worse patient outcomes for pediatric liver transplant centers. J Pediatr Surg 2010; 45: 108–113. 16 Chabannon C, Pamphilon D, Vermylen C, Gratwohl A, Niederwieser D, McGrath E et al. Ten years after the first inspection of a candidate European centre, an EBMT registry analysis suggests that clinical outcome is improved when hematopoietic SCT is performed in a JACIE accredited program. Bone Marrow Transplant 2011; 47: 15–7. 17 Pulsipher MA, Horwitz EM, Haight AE, Kadota R, Chen AR, Frangoul H et al. Advancement of pediatric blood and marrow transplantation research in North America: priorities of the Pediatric Blood and Marrow Transplant Consortium. Biol Blood Marrow Transplant 2010; 16: 1212–1221. 18 Klingebiel T, Cornish J, Labopin M, Locatelli F, Darbyshire P, Handgretinger R et al. Results and factors influencing outcome after fully haploidentical hematopoietic stem cell transplantation in children with very high-risk acute lymphoblastic leukemia: impact of center size: an analysis on behalf of the Acute Leukemia and Pediatric Disease Working Parties of the European Blood and Marrow Transplant group. Blood 2010; 115: 3437–3446. 19 Miano M, Labopin M, Hartmann O, Angelucci E, Cornish J, Gluckman E et al. Haematopoietic stem cell transplantation trends in children over the last three decades: a survey by the paediatric diseases working party of the European Group for Blood and Marrow Transplantation. Bone Marrow Transplant 2007; 39: 89–99. 20 Yusuf U, Frangoul HA, Gooley TA, Woolfrey AE, Carpenter PA, Andrews RG et al. Allogeneic bone marrow transplantation in children with myelodysplastic syndrome or juvenile myelomonocytic leukemia: the Seattle experience. Bone Marrow Transplant 2004; 33: 805–814. 21 Ridout MS, Demetrio CG, Firth D. Estimating intraclass correlation for binary data. Biometrics 1999; 55: 137–148.

Bone Marrow Transplantation (2013) 514 – 522


522 22 Loberiza Jr FR, Zhang MJ, Lee SJ, Klein JP, LeMaistre CF, Serna DS et al. Association of transplant center and physician factors on mortality after hematopoietic stem cell transplantation in the United States. Blood 2005; 105: 2979–2987. 23 Woolfrey AE, Anasetti C, Storer B, Doney K, Milner LA, Sievers EL et al. Factors associated with outcome after unrelated marrow transplantation for treatment of acute lymphoblastic leukemia in children. Blood 2002; 99: 2002–2008.

Bone Marrow Transplantation (2013) 514 – 522

24 Dini G, Cancedda R, Giorgiani G, Porta F, Messina C, Uderzo C et al. Unrelated donor marrow transplantation in childhood: a report from the Associazione Italiana Ematologia e Oncologia Pediatrica (AIEOP) and the Gruppo Italiano per il Trapianto Midollo Osseo (GITMO). Haematologica 2002; 87((8 Suppl): 51–57. 25 Rocha V, Sanz G, Gluckman E. Umbilical cord blood transplantation. Curr Opin Hematol 2004; 11: 375–385. 26 Smith AH, Bates MN. Confidence limit analyses should replace power calculations in the interpretation of epidemiologic studies. Epidemiology 1992; 3: 449–452.
