Nonparametric failure time: Time-to-event machine learning with heteroskedastic Bayesian additive regression trees and low information omnibus Dirichlet process mixtures.

Publisher: Biometric Society Country of Publication: England NLM ID: 0370625 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1541-0420 (Electronic) Linking ISSN: 0006341X NLM ISO Abbreviation: Biometrics Subsets: MEDLINE

Publication: Alexandria Va : Biometric Society
Original Publication: Washington.

Software* ; Machine Learning*; Humans ; Bayes Theorem ; Proportional Hazards Models ; Uncertainty ; Models, Statistical ; Computer Simulation

Many popular survival models rely on restrictive parametric, or semiparametric, assumptions that could provide erroneous predictions when the effects of covariates are complex. Modern advances in computational hardware have led to an increasing interest in flexible Bayesian nonparametric methods for time-to-event data such as Bayesian additive regression trees (BART). We propose a novel approach that we call nonparametric failure time (NFT) BART in order to increase the flexibility beyond accelerated failure time (AFT) and proportional hazard models. NFT BART has three key features: (1) a BART prior for the mean function of the event time logarithm; (2) a heteroskedastic BART prior to deduce a covariate-dependent variance function; and (3) a flexible nonparametric error distribution using Dirichlet process mixtures (DPM). Our proposed approach widens the scope of hazard shapes including nonproportional hazards, can be scaled up to large sample sizes, naturally provides estimates of uncertainty via the posterior and can be seamlessly employed for variable selection. We provide convenient, user-friendly, computer software that is freely available as a reference implementation. Simulations demonstrate that NFT BART maintains excellent performance for survival prediction especially when AFT assumptions are violated by heteroskedasticity. We illustrate the proposed approach on a study examining predictors for mortality risk in patients undergoing hematopoietic stem cell transplant (HSCT) for blood-borne cancer, where heteroskedasticity and nonproportional hazards are likely present.
(© 2023 The International Biometric Society.)

Biostatistics. 2020 Jan 1;21(1):69-85. (PMID: 30059992)
Bioinformatics. 2011 Feb 1;27(3):359-67. (PMID: 21148161)
JAMA. 1982 May 14;247(18):2543-6. (PMID: 7069920)
J Chronic Dis. 1987;40(5):373-83. (PMID: 3558716)
Biol Blood Marrow Transplant. 2015 Aug;21(8):1479-87. (PMID: 25862591)
Stem Cells Cloning. 2010 Aug 26;3:105-17. (PMID: 24198516)
Stat Methods Med Res. 2020 Jan;29(1):57-77. (PMID: 30612519)
Nat Med. 2005 Oct;11(10):1026-8. (PMID: 16211027)
Comput Stat Data Anal. 2010 Sep 1;54(9):2172-2186. (PMID: 24363478)
J Clin Oncol. 2017 Apr 10;35(11):1154-1161. (PMID: 28380315)
Biostatistics. 2020 Jan 1;21(1):50-68. (PMID: 30052809)
JCO Clin Cancer Inform. 2021 May;5:494-507. (PMID: 33950708)
Stat Med. 2016 Jul 20;35(16):2741-53. (PMID: 26854022)

U24 CA076518 United States CA NCI NIH HHS; United States HL NHLBI NIH HHS; United States CA NCI NIH HHS; United States HL NHLBI NIH HHS

Keywords: BART; LIO prior hierarchy; Thompson sampling variable selection; accelerated failure time; constrained DPM; hematopoietic stem cell transplant; nonproportional hazards; survival analysis

Date Created: 20230318 Date Completed: 20231221 Latest Revision: 20241202

20241204

PMC10505620

10.1111/biom.13857

36932826