Using Cost Effectiveness Analysis to Determine Inventory Size for a National Umbilical Cord Blood Bank

Using Cost Effectiveness Analysis to Determine Inventory Size for a National Umbilical Cord Blood Bank Martin Maiers, Bioinformatics National Marrow Donor Program, USA

NMDP Minneapolis, MN, USA NATIONAL MARROW DONOR PROGRAM 2 Creating Connections. Saving Lives.

Messages Cord Blood is an alternative to adult unrelated donors The more units are stored, the greater the likelihood that transplant candidates will match a unit Storing cord blood units is costly Benefits of increasing cord inventory must be weighed against the costs 3

Approach Estimate the likelihood that transplant candidates will match to a living unrelated marrow donor or a cord blood unit as a function of cord blood inventory Calculate the life years gained for each transplant type by match level using historical data Develop a model of the cord blood inventory level to estimate total costs as a function of the number of stored units 4

Motivation >22 U.S. public banks Charge a fee per transplanted unit ~ 15,000USD Complicated to search independently U.S. Congress commissioned a report from the Institute of Medicine to make recommendations on the governance and structure of a national cord blood bank This cost-effectiveness analysis was commissioned by the IOM to address the question: How many units should a national bank place in inventory? 5

Cord Blood Transplantation + Does not require healthy donors to undergo an invasive and discomforting procedure + HLA match requirements considered less stringent + Patients can undergo transplantation immediately instead of having to wait for the living unrelated donor search process to run its course - The volume of stem cells obtained from a single cord unit is limited - Patient weight must be considered - Used more often for pediatric patients 6

Cost/Benefit The larger the inventory, the greater the likelihood that transplant candidates will match to a stored unit Processing new cord blood units is costly The access benefits must be weighed against the storage costs 7

Assumptions Current inventory: effectively 50,000 Inventory sizes in the range 50,000 100,000 150,000 200,000 300,000 TNC Dose 2.5 x 10 7 8

CBU TNC Histogram (N=36,135) 0 500 1000 1500 2000 2500 3000 5 15 25 35 45 55 65 75 85 95 105 115 125 135 145 155 165 175 185 195 205 215 225 235 245 255 265 275 285 295 305 315 325 335 345 355 365 375 385 395 TNC (x10^7) count

Model - HLA Matching 1. Use the expectation maximization (EM) algorithm to estimate the population frequency of HLA haplotypes types by racial/ethnic group 2. Calculate matching likelihood separately for donor & cord HLA-A & -B antigen level, DRB1 allele level 3. Sum over HLA types and racial groups to derive nationallyrepresentative estimates Donor match probabilities incorporate donor availability based on historical patterns Donor match probabilities based on 2,267,366 fully typed US donors Cord match incorporates a requirement of TNC dose of 2.5 10 7 /kg Patient bodyweight is modeled using empirical weight distribution Kollman C, et al. Transplantation. 2004;78:89-95. 10

Results - Matching: Adult 6/6 Cord 6/6 Donor or 6/6 Cord 5/6+ Cord 6/6 Donor or 5/6+ Cord 4/6+ Cord 300,000 200,000 150,000 100,000 50,000 6/6 Donor or 4/6+ Cord 0 0.2 0.4 0.6 0.8 1 Proportion of patients aged 20+ finding a match 11

Results - Matching: Pediatric 6/6 Cord 6/6 Donor or 6/6 Cord 5/6+ Cord 6/6 Donor or 5/6+ Cord 4/6+ Cord 300,000 200,000 150,000 100,000 50,000 6/6 Donor or 4/6+ Cord 0 0.2 0.4 0.6 0.8 1 Proportion of patients aged <20 finding a match 12

Model - Life Years Number of CBU transplants Estimate: 8,200 unrelated transplant candidates Medical urgency (assume 25% are urgent) Urgent: consider cord blood transplants exclusively Non-urgent: adult donor preferred Matching more stringent for donors: donor transplants are considered only if the donor matches to the recipient at 7/8 or 8/8 alleles P(8/8 HRM 6/6 LRM) = 56% P(7/8 HRM 6/6 LRM) = 26% P(7/8 HRM 5/6 LRM) = 24% 13 Flomenberg N, et al. Blood. 2004;104:1923-1930.

Adult Patients 5,467 Adult Patients 4,100 Non-Urgent 1,367 Urgent 4,048 5/6+ donor match 52 no donor match cord match 6/6 94 5/6 447 4/6 608 no cord match 218 HR donor match 1512 8/8 1026 7/8 no HR donor 1512 cord match 6/6 0 5/6 0 4/6 10 no cord match 42 cord match 6/6 52 5/6 332 4/6 780 no cord match 346 14

Pediatric Patients 2,733 Pediatric Patients 2,050 Non-Urgent 683 Urgent 1,350 6/6 donor match 700 no donor match cord match 6/6 155 5/6 365 4/6 149 no cord match 14 HR donor match 756 8/8 no HR donor 594 cord match 6/6 7 5/6 331 4/6 327 no cord match 42 cord match 6/6 206 5/6 336 4/6 52 no cord match 4 15

Projected Number: Annual Cord Transplants 300,000 200,000 150,000 100,000 50,000 Pediatric 4/6 Cord Blood 5/6 Cord Blood 6/6 Cord Blood 0 100 200 300 400 300,000 200,000 150,000 100,000 50,000 Adult 4/6 Cord Blood 5/6 Cord Blood 6/6 Cord Blood 16 0 100 200 300 400

Model - 5 year Survival 1 0.8 0.6 0.4 0.2 0 8/8 donor 7/8 donor 6/6 cord 5/6 cord 4/6 cord no transpl ant Adult 0.36 0.25 0.36 0.25 0.17 0.15 Pediatric 0.49 0.41 0.49 0.41 0.34 0.15 Pediatric Adult 17

Cost Model Cost Model Parameters Parameter Description Value Unit Inventory parameters N 0 Initial inventory 50,000 cord units λ Discard rate 0.50 proportion y Length of storage before discard 20 years Cost parameters c PS Cost of initial processing, stored unit $1,500 dollars c PD Cost of initial processing, discarded unit $500 dollars c S Cost of storage, annual $50 dollars c T Cost of transplantation $220,000 dollars E Endowment $9,000,000 dollars A Annual administrative cost $1,000,000 dollars r Discount rate 0.03 proportion 18

Cost Model C 0 = [(1 λ)c PS + λc PD ] N N 0 1 λ Cost of increasing cord blood inventory to a target size C = [(1 λ)c PS + λc PD ]U + c s N + A Annual costs for a cord blood bank TC( N ) = C + rc 0 + c T T ICER(N,N') = TC(N') TC(N) LY(N') LY(N) Total cost of transplants Incremental Cost- Effectiveness Ratio: cost per life-year-gained by increasing inventory from N to N 19

Cost per life year gained Cost Per Life Year Gained as a Function of Cord Inventory Cord Inventory 50,000 100,000 150,000 200,000 300,000 Total costs (millions) Total $146 $162 $175 $187 $210 Incremental NA $16 $13 $12 $23 Life years gained Total 77,003 77,437 77,672 77,830 78,042 Incremental NA 434 234 158 212 Incremental cost per life year NA $37,667 $55,873 $75,953 $106,948 20

21

Results The cost per life year gained associated with increasing inventory from 50,000 to 100,000 units is $38,000 - $75,000 from 100,000 to 150,000 units is $59,000 - $142,000 22

Conclusion Assuming the monetary value of a year of life exceeds $100,000: Expanding the cord blood inventory above current levels to 100,000-150,000 units is cost-effective 23

Future Analyses Selection of a graft source Cord as first choice Model international search Enhance matching model high-resolution haplotypes, refine race/ethnic categories Explore race/ethnic equity Effects of targeted recruitment Enhance survival model Include cell dose/matching interaction Multiple cord transplants 24

Acknowledgements Working Group David H. Howard, Ph.D. Emory University David Meltzer, M.D., Ph.D. University of Chicago Mary M. Horowitz, M.D. Center for International Blood and Marrow Transplant Research Brent Logan, Ph.D. Center for International Blood and Marrow Transplant Research Martin Maiers, B.S. National Marrow Donor Program Loren Gragert, B.S., B.A. National Marrow Donor Program Michelle Setterholm, B.S. National Marrow Donor Program Craig Kollman, Ph.D. Jaeb Center for Health Research Funding National Academy of Sciences, USA ONR N000 14-99-2-0006 25