Brent R Logan1,2, Martin J Maiers3, Rodney A Sparapani1, Purushottam W Laud1, Stephen R Spellman3, Robert E McCulloch4, Bronwen E Shaw2. 1. Division of Biostatistics, Institute for Health and Equity, Medical College of Wisconsin (MCW), Milwaukee, WI. 2. Center for International Blood and Marrow Transplant Research (CIBMTR), Medical College of Wisconsin, Milwaukee, WI. 3. National Marrow Donor Program and Center for International Blood and Marrow Transplant Research, Minneapolis, MN. 4. School of Mathematical and Statistical Sciences, Arizona State University, Tempe, AZ.
Abstract
PURPOSE: Donor selection practices for matched unrelated donor (MUD) hematopoietic cell transplantation (HCT) vary, and the impact of optimizing donor selection in a patient-specific way using modern machine learning (ML) models has not been studied. METHODS: We trained a Bayesian ML model in 10,318 patients who underwent MUD HCT from 1999 to 2014 to provide patient- and donor-specific predictions of clinically severe (grade 3 or 4) acute graft-versus-host disease or death by day 180. The model was validated in 3,501 patients from 2015 to 2016 with archived records of potential donors at search. Donor selection optimizing predicted outcomes was implemented over either an unlimited donor pool or the donors in the search archives. Posterior mean differences in outcomes from optimal donor selection versus actual practice were summarized per patient and across the population with 95% intervals. RESULTS: Event rates were 33% (training) and 37% (validation). Among donor features, only age affected outcomes, with the effect consistent regardless of patient features. The median (interquartile range) difference in age between the youngest donor at search and the selected donor was 6 (1-10) years, whereas the number of donors per patient younger than the selected donor was 6 (1-36). Fourteen percent of the validation data set had an approximate 5% absolute reduction in event rates from selecting the youngest donor at search versus the actual donor used, leading to an absolute population reduction of 1% (95% interval, 0 to 3). CONCLUSION: We confirmed the singular importance of selecting the youngest available MUD, irrespective of patient features, identified potential for improved HCT outcomes by selecting a younger MUD, and demonstrated use of novel ML models transferable to optimize other complex treatment decisions in a patient-specific way.
PURPOSE: Donor selection practices for matched unrelated donor (MUD) hematopoietic cell transplantation (HCT) vary, and the impact of optimizing donor selection in a patient-specific way using modern machine learning (ML) models has not been studied. METHODS: We trained a Bayesian ML model in 10,318 patients who underwent MUD HCT from 1999 to 2014 to provide patient- and donor-specific predictions of clinically severe (grade 3 or 4) acute graft-versus-host disease or death by day 180. The model was validated in 3,501 patients from 2015 to 2016 with archived records of potential donors at search. Donor selection optimizing predicted outcomes was implemented over either an unlimited donor pool or the donors in the search archives. Posterior mean differences in outcomes from optimal donor selection versus actual practice were summarized per patient and across the population with 95% intervals. RESULTS: Event rates were 33% (training) and 37% (validation). Among donor features, only age affected outcomes, with the effect consistent regardless of patient features. The median (interquartile range) difference in age between the youngest donor at search and the selected donor was 6 (1-10) years, whereas the number of donors per patient younger than the selected donor was 6 (1-36). Fourteen percent of the validation data set had an approximate 5% absolute reduction in event rates from selecting the youngest donor at search versus the actual donor used, leading to an absolute population reduction of 1% (95% interval, 0 to 3). CONCLUSION: We confirmed the singular importance of selecting the youngest available MUD, irrespective of patient features, identified potential for improved HCT outcomes by selecting a younger MUD, and demonstrated use of novel ML models transferable to optimize other complex treatment decisions in a patient-specific way.
Authors: Hideki Nakasone; Bita Sahaf; Lu Tian; Tao Wang; Michael D Haagenson; Kelsi Schoenrock; Spenser Perloff; Christine E Ryan; Fang Wu; Stephen R Spellman; Stephanie J Lee; Jerome Ritz; David B Miklos Journal: Haematologica Date: 2015-10-22 Impact factor: 9.941
Authors: David B Miklos; Haesook T Kim; Katherine H Miller; Luxuan Guo; Emmanuel Zorn; Stephanie J Lee; Ephraim P Hochberg; Catherine J Wu; Edwin P Alyea; Corey Cutler; Vincent Ho; Robert J Soiffer; Joseph H Antin; Jerome Ritz Journal: Blood Date: 2004-12-21 Impact factor: 22.113
Authors: C Kollman; C W Howe; C Anasetti; J H Antin; S M Davies; A H Filipovich; J Hegland; N Kamani; N A Kernan; R King; V Ratanatharathorn; D Weisdorf; D L Confer Journal: Blood Date: 2001-10-01 Impact factor: 22.113
Authors: Bronwen E Shaw; Brent R Logan; Stephen R Spellman; Steven G E Marsh; James Robinson; Joseph Pidala; Carolyn Hurley; Juliet Barker; Martin Maiers; Jason Dehn; Hailin Wang; Mike Haagenson; David Porter; Effie W Petersdorf; Ann Woolfrey; Mary M Horowitz; Michael Verneris; Katharine C Hsu; Katharina Fleischhauer; Stephanie J Lee Journal: Biol Blood Marrow Transplant Date: 2018-02-14 Impact factor: 5.742
Authors: D Przepiorka; D Weisdorf; P Martin; H G Klingemann; P Beatty; J Hows; E D Thomas Journal: Bone Marrow Transplant Date: 1995-06 Impact factor: 5.483
Authors: Craig Kollman; Stephen R Spellman; Mei-Jie Zhang; Anna Hassebroek; Claudio Anasetti; Joseph H Antin; Richard E Champlin; Dennis L Confer; John F DiPersio; Marcelo Fernandez-Viña; Robert J Hartzman; Mary M Horowitz; Carolyn K Hurley; Chatchada Karanes; Martin Maiers; Carlheinz R Mueller; Miguel-Angel Perales; Michelle Setterholm; Ann E Woolfrey; Neng Yu; Mary Eapen Journal: Blood Date: 2015-11-02 Impact factor: 22.113