Dynamic Prediction in Clinical Survival Analysis

by van Houwelingen; Hans

• ISBN13: 9781439835333
• ISBN10: 1439835330
• Edition: 1st
• Format: Hardcover
• Copyright: 2011-11-09
• Publisher: CRC Press

Summary

In the last twenty years, dynamic prediction models have been extensively used to monitor patient prognosis in survival analysis. Written by one of the pioneers in the area, this book synthesizes these developments in a unified framework. It covers a range of models, including prognostic and dynamic prediction of survival using genomic data and time-dependent information. The text includes numerous examples using real data that is taken from the authors'¬" collaborative research. R programs are provided for implementing the methods.

Table of Contents

Preface	p. xi
About the Authors	p. xv
Prognostic models for survival data using (clinical) information available at baseline, based on the Cox model	p. 1
The special nature of survival data	p. 3
Introduction	p. 3
Basic statistical concepts	p. 5
Predictive use of the survival function	p. 9
Additional remarks	p. 13
Cox regression model	p. 15
The hazard function	p. 15
The proportional hazards model	p. 18
Fitting the Cox model	p. 21
Example: Breast Cancer II	p. 24
Extensions of the data structure	p. 26
Alternative models	p. 30
Additional remarks	p. 33
Measuring the predictive value of a Cox model	p. 35
Introduction	p. 35
Visualizing the relation between predictor and survival	p. 35
Measuring the discriminative ability	p. 38
Measuring the prediction error	p. 42
Dealing with overfitting	p. 49
Cross-validated partial likelihood	p. 51
Additional remarks	p. 54
Calibration and revision of Cox models	p. 57
Validation by calibration	p. 57
Internal calibration	p. 58
External calibration	p. 59
Model revision	p. 66
Additional remarks	p. 68
Prognostic models for survival data using (clinical) information available at baseline, when the proportional hazards assumption of the Cox model is violated	p. 71
Mechanisms explaining violation of the Cox model	p. 73
The Cox model is just a model	p. 73
Heterogeneity	p. 74
Measurement error in covariates	p. 79
Cause specific hazards and competing risks	p. 81
Additional remarks	p. 84
Non-proportional hazards models	p. 85
Cox model with time-varying coefficients	p. 85
Models inspired by the frailty concept	p. 91
Enforcing parsimony through reduced rank models	p. 94
Additional remarks	p. 98
Dealing with non-proportional hazards	p. 101
Robustness of the Cox model	p. 101
Obtaining dynamic predictions by landmarking	p. 105
Additional remarks	p. 116
Dynamic prognostic models for survival data usingtime-dependent information	p. 119
Dynamic predictions using biomarkers	p. 121
Prediction in a dynamic setting	p. 121
Landmark prediction model	p. 124
Application	p. 126
Additional remarks	p. 132
Dynamic prediction in multi-state models	p. 135
Multi-state models in clinical applications	p. 135
Dynamic prediction in multi-state models	p. 139
Application	p. 142
Additional remarks	p. 151
Dynamic prediction in chronic disease	p. 153
General description	p. 153
Exploration of the EORTC breast cancer data set	p. 154
Dynamic prediction models for breast cancer	p. 161
Dynamic assessment of "cure"	p. 164
Additional remarks	p. 168
Dynamic prognostic models for survival data using genomic data	p. 169
Penalized Cox models	p. 171
Introduction	p. 171
Ridge and las,so	p. 172
Application to Data Set 3	p. 174
Adding clinical predictors	p. 179
Additional remarks	p. 181
Dynamic prediction based on genomic data	p. 185
Testing the proportional hazards assumption	p. 185
Landmark predictions	p. 186
Additional remarks	p. 191
Appendices	p. 193
Data sets	p. 195
Data Set 1: Advanced ovarian cancer	p. 195
Data Set 2: Chronic Myeloid Leukemia (CML)	p. 196
Data Set 3: Breast Cancer I (NKI)	p. 199
Data Set 4: Gastric Cancer	p. 200
Data Set 5: Breast Cancer II (EORTC)	p. 203
Data Set 6: Acute Lymphatic Leukemia (ALL)	p. 205
Software and website	p. 211
R packages used	p. 212
The dynpred package	p. 213
Additional remarks	p. 215
References	p. 217
Index	p. 233
Table of Contents provided by Ingram. All Rights Reserved.