A two-stage approach for dynamic prediction of time-to-event distributions

Xuelin Huang, Fangrong Yan, Jing Ning, Ziding Feng, Sangbum Choi, Jorge Cortes

Research output: Contribution to journalArticle

Abstract

Dynamic prediction uses longitudinal biomarkers for real-time prediction of an individual patient's prognosis. This is critical for patients with an incurable disease such as cancer. Biomarker trajectories are usually not linear, nor even monotone, and vary greatly across individuals. Therefore, it is difficult to fit them with parametric models. With this consideration, we propose an approach for dynamic prediction that does not need to model the biomarker trajectories. Instead, as a trade-off, we assume that the biomarker effects on the risk of disease recurrence are smooth functions over time. This approach turns out to be computationally easier. Simulation studies show that the proposed approach achieves stable estimation of biomarker effects over time, has good predictive performance, and is robust against model misspecification. It is a good compromise between two major approaches, namely, (i) joint modeling of longitudinal and survival data and (ii) landmark analysis. The proposed method is applied to patients with chronic myeloid leukemia. At any time following their treatment with tyrosine kinase inhibitors, longitudinally measured BCR-ABL gene expression levels are used to predict the risk of disease progression.

Original languageEnglish (US)
Pages (from-to)2167-2182
Number of pages16
JournalStatistics in Medicine
Volume35
Issue number13
DOIs
StatePublished - Jun 15 2016
Externally publishedYes

Fingerprint

Biomarkers
Prediction
Trajectory
Joint Modeling
Model Misspecification
Survival Data
Prognosis
Leukemia
Longitudinal Data
Leukemia, Myelogenous, Chronic, BCR-ABL Positive
Landmarks
Parametric Model
Progression
Smooth function
Recurrence
Protein-Tyrosine Kinases
Inhibitor
Gene Expression
Disease Progression
Monotone

Keywords

  • Biomarker
  • Dynamic prediction
  • Landmark analysis
  • Longitudinal data
  • Survival analysis
  • Time-dependent covariate

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Cite this

A two-stage approach for dynamic prediction of time-to-event distributions. / Huang, Xuelin; Yan, Fangrong; Ning, Jing; Feng, Ziding; Choi, Sangbum; Cortes, Jorge.

In: Statistics in Medicine, Vol. 35, No. 13, 15.06.2016, p. 2167-2182.

Research output: Contribution to journalArticle

Huang, Xuelin ; Yan, Fangrong ; Ning, Jing ; Feng, Ziding ; Choi, Sangbum ; Cortes, Jorge. / A two-stage approach for dynamic prediction of time-to-event distributions. In: Statistics in Medicine. 2016 ; Vol. 35, No. 13. pp. 2167-2182.
@article{161e372ceac147d0ae6e68b3ca9de77f,
title = "A two-stage approach for dynamic prediction of time-to-event distributions",
abstract = "Dynamic prediction uses longitudinal biomarkers for real-time prediction of an individual patient's prognosis. This is critical for patients with an incurable disease such as cancer. Biomarker trajectories are usually not linear, nor even monotone, and vary greatly across individuals. Therefore, it is difficult to fit them with parametric models. With this consideration, we propose an approach for dynamic prediction that does not need to model the biomarker trajectories. Instead, as a trade-off, we assume that the biomarker effects on the risk of disease recurrence are smooth functions over time. This approach turns out to be computationally easier. Simulation studies show that the proposed approach achieves stable estimation of biomarker effects over time, has good predictive performance, and is robust against model misspecification. It is a good compromise between two major approaches, namely, (i) joint modeling of longitudinal and survival data and (ii) landmark analysis. The proposed method is applied to patients with chronic myeloid leukemia. At any time following their treatment with tyrosine kinase inhibitors, longitudinally measured BCR-ABL gene expression levels are used to predict the risk of disease progression.",
keywords = "Biomarker, Dynamic prediction, Landmark analysis, Longitudinal data, Survival analysis, Time-dependent covariate",
author = "Xuelin Huang and Fangrong Yan and Jing Ning and Ziding Feng and Sangbum Choi and Jorge Cortes",
year = "2016",
month = "6",
day = "15",
doi = "10.1002/sim.6860",
language = "English (US)",
volume = "35",
pages = "2167--2182",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "13",

}

TY - JOUR

T1 - A two-stage approach for dynamic prediction of time-to-event distributions

AU - Huang, Xuelin

AU - Yan, Fangrong

AU - Ning, Jing

AU - Feng, Ziding

AU - Choi, Sangbum

AU - Cortes, Jorge

PY - 2016/6/15

Y1 - 2016/6/15

N2 - Dynamic prediction uses longitudinal biomarkers for real-time prediction of an individual patient's prognosis. This is critical for patients with an incurable disease such as cancer. Biomarker trajectories are usually not linear, nor even monotone, and vary greatly across individuals. Therefore, it is difficult to fit them with parametric models. With this consideration, we propose an approach for dynamic prediction that does not need to model the biomarker trajectories. Instead, as a trade-off, we assume that the biomarker effects on the risk of disease recurrence are smooth functions over time. This approach turns out to be computationally easier. Simulation studies show that the proposed approach achieves stable estimation of biomarker effects over time, has good predictive performance, and is robust against model misspecification. It is a good compromise between two major approaches, namely, (i) joint modeling of longitudinal and survival data and (ii) landmark analysis. The proposed method is applied to patients with chronic myeloid leukemia. At any time following their treatment with tyrosine kinase inhibitors, longitudinally measured BCR-ABL gene expression levels are used to predict the risk of disease progression.

AB - Dynamic prediction uses longitudinal biomarkers for real-time prediction of an individual patient's prognosis. This is critical for patients with an incurable disease such as cancer. Biomarker trajectories are usually not linear, nor even monotone, and vary greatly across individuals. Therefore, it is difficult to fit them with parametric models. With this consideration, we propose an approach for dynamic prediction that does not need to model the biomarker trajectories. Instead, as a trade-off, we assume that the biomarker effects on the risk of disease recurrence are smooth functions over time. This approach turns out to be computationally easier. Simulation studies show that the proposed approach achieves stable estimation of biomarker effects over time, has good predictive performance, and is robust against model misspecification. It is a good compromise between two major approaches, namely, (i) joint modeling of longitudinal and survival data and (ii) landmark analysis. The proposed method is applied to patients with chronic myeloid leukemia. At any time following their treatment with tyrosine kinase inhibitors, longitudinally measured BCR-ABL gene expression levels are used to predict the risk of disease progression.

KW - Biomarker

KW - Dynamic prediction

KW - Landmark analysis

KW - Longitudinal data

KW - Survival analysis

KW - Time-dependent covariate

UR - http://www.scopus.com/inward/record.url?scp=84954134178&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84954134178&partnerID=8YFLogxK

U2 - 10.1002/sim.6860

DO - 10.1002/sim.6860

M3 - Article

C2 - 26748812

AN - SCOPUS:84954134178

VL - 35

SP - 2167

EP - 2182

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 13

ER -