Joint Models for Longitudinal and Time to Event Data With

Applications in R, 1st Edition

Visit the link below to download the full version of this book:

https://medidownload.com/product/joint-models-for-longitudinal-and-time-to-event
-data-with-applications-in-r-1st-edition/

Click Download Now

 DNA Microarrays and Related Genomics Monte Carlo Simulation for the
Techniques: Design, Analysis, and Pharmaceutical Industry: Concepts,
Interpretation of Experiments Algorithms, and Case Studies
David B. Allison, Grier P. Page, Mark Chang
T. Mark Beasley, and Jode W. Edwards Multiple Testing Problems in
Dose Finding by the Continual Pharmaceutical Statistics
Reassessment Method Alex Dmitrienko, Ajit C. Tamhane,
Ying Kuen Cheung and Frank Bretz
Elementary Bayesian Biostatistics Randomized Clinical Trials of
Lemuel A. Moyé Nonpharmacological Treatments
Frailty Models in Survival Analysis Isabelle Boutron, Philippe Ravaud, and
Andreas Wienke David Moher

Generalized Linear Models: A Bayesian Sample Size Calculations in Clinical

Perspective Research, Second Edition
Dipak K. Dey, Sujit K. Ghosh, Shein-Chung Chow, Jun Shao
and Bani K. Mallick and Hansheng Wang

Handbook of Regression and Modeling: Statistical Design and Analysis of

Applications for the Clinical and Stability Studies
Pharmaceutical Industries Shein-Chung Chow
Daryl S. Paulson Statistical Evaluation of Diagnostic
Interval-Censored Time-to-Event Data: Performance: Topics in ROC Analysis
Methods and Applications Kelly H. Zou, Aiyi Liu, Andriy Bandos,
Ding-Geng (Din) Chen, Jianguo Sun, Lucila Ohno-Machado, and Howard Rockette
and Karl E. Peace Statistical Methods for Clinical Trials
Joint Models for Longitudinal and Time- Mark X. Norleans
to-Event Data: With Applications in R Statistics in Drug Research:
Dimitris Rizopoulos Methodologies and Recent
Measures of Interobserver Agreement Developments
and Reliability, Second Edition Shein-Chung Chow and Jun Shao
Mohamed M. Shoukri Statistics in the Pharmaceutical Industry,
Medical Biostatistics, Third Edition Third Edition
A. Indrayan Ralph Buncher and Jia-Yeong Tsay

Meta-Analysis in Medicine and Health Translational Medicine: Strategies and

Policy Statistical Methods
Dalene Stangl and Donald A. Berry Dennis Cosmatos and Shein-Chung Chow

K13371_FM.indd 4 5/9/12 1:36 PM

 Joint Models for
Longitudinal and
Time-to-Event Data
With Applications in R

Dimitris Rizopoulos

K13371_FM.indd 5 5/9/12 1:36 PM

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2012 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Version Date: 20120504

International Standard Book Number-13: 978-1-4398-7287-1 (eBook - PDF)

This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and
publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a pho-
tocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
To Roula and my parents

To the memory of my grandmother

This page intentionally left blank
Preface

Joint models for longitudinal and time-to-event data have become a valuable
tool in the analysis of follow-up data. These models are applicable mainly in
two settings: First, when the focus is on the survival outcome and we wish to
account for the effect of an endogenous time-dependent covariate measured
with error, and second, when the focus is on the longitudinal outcome and
we wish to correct for nonrandom dropout. Due to the capability of joint
models to provide valid inferences in settings where simpler statistical tools
fail to do so, and their wide range of applications, the last 25 years have
seen many advances in the joint modeling field. Even though interest and
developments in joint models have been widespread, information about them
has been equally scattered in articles, presenting recent advances in the field,
and in book chapters in a few texts dedicated either to longitudinal or survival
data analysis. However, no single monograph or text dedicated to this type of
models seems to be available.
The purpose in writing this book, therefore, is to provide an overview of
the theory and application of joint models for longitudinal and survival data.
In the literature two main frameworks have been proposed, namely the ran-
dom effects joint model that uses latent variables to capture the associations
between the two outcomes (Tsiatis and Davidian, 2004), and the marginal
structural joint models based on G estimators (Robins et al., 1999, 2000).
In this book we focus on the former. Both subfields of joint modeling, i.e.,
handling of endogenous time-varying covariates and nonrandom dropout, are
equally covered and presented in real datasets. Motivated by the background
and collaborations of the author, these examples come from the medical field
and in particular from biomarker research. Nevertheless, joint models of this

vii
viii

type can be employed whenever there is interest in the association between a

longitudinal and an event time outcome, and therefore have numerous appli-
cations in other fields as well. In addition, several extensions are presented,
including among others, different types of association structure between the
longitudinal and survival outcomes, inclusion of stratification factors, incorpo-
ration of lagged effects, the handling of exogenous time-dependent covariates,
and competing risks settings. For the longitudinal part, primary focus is given
on continuous longitudinal outcomes and the use of linear mixed-effects mod-
els.

All the analyses included in the book are implemented in the R software
environment for statistical computing and graphics, using the freely avail-
able package JM written by the author. This package fits a wide range of
joint models, including the extensions mentioned above. However, it does not
cover all of the types of joint models that have been proposed in the liter-
ature, and therefore, some extensions, such as joint models for categorical
longitudinal markers and multiple longitudinal markers are introduced in the
text, but without any software illustration. The results presented in the text
were produced using version 1.0-0 of package JM and R 2.15.0. The current
version of the package can be obtained from the Comprehensive R Archive
Network (http://cran.r-project.org). Because of platform dependencies,
the analysis results may be expected to vary slightly with different computers
or operating systems, and future versions of R and the package. All the code
used in the text is available from the supporting Web site

http://jmr.r-forge.r-project.org

The author would appreciate being informed of typos, errors and improve-
ments to the contents of this book, and may be contacted via electronic mail
at

[email protected]

Prerequisites

The contents in this book assume familiarity with basic concepts of statistical
data analysis at the level of an introductory course in statistics, including
standard maximum likelihood theory, and a strong background in regression
analysis. Some background knowledge of mixed effects models and survival
analysis would also be beneficial but is not required. Similarly, with respect to
R, the book does not assume any prior knowledge, but basic familiarity with
the language would be helpful.
 ix

Typographical Conventions
R language objects, commands and output referenced throughout this book
are printed in a monospace typewriter font. Code that could be interactively
entered at the R command line is formatted as:
> x <- 5

where > denotes the R command-line prompt and everything else is what the
user should enter. In addition, R expressions that do not fit in a single line
will be appropriately indented like this:
> jointModel(lmeFit, coxFit, timeVar = "time",
method = "weibull-PH-aGH", parameterization = "both")

When referring to a function within the main text, it will be formatted in

a typewriter font and will have parentheses after the function name, e.g.,
jointModel(). Similarly, function arguments and value specified for these
arguments will also be formatted in a typewriter font, but with no parentheses
at the end, e.g., method = "weibull-PH-aGH".
To save space, some of the R output has been edited. Omission of complete
lines is indicated by
. . .

but some blank lines have been removed without indication.

Acknowledgements
Many pieces of the work presented in this monograph are based on joint re-
search with Geert Verbeke and Geert Molenberghs, with whom it has been my
great pleasure to collaborate. I would also like to thank Emmanuel Lesaffre
for useful discussions on joint modeling topics, and for creating a stimulating
work environment. Special thanks also goes to all users of JM whose input
has greatly benefitted the development of the package.
With regard to this book in particular, I would like to thank the anony-
mous reviewers for valuable feedback on earlier versions of the manuscript, and
John Kimmel at Chapman and Hall/CRC for his help and support during the
whole production process. This monograph was written using a combination
of Friedrich Leisch’s Sweave package, the LATEX document preparation sys-
tem, and, of course, R.

Dimitris Rizopoulos
Rotterdam, April 2012
x
Contents

Preface vii

1 Introduction 1
1.1 Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Motivating Studies . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Primary Biliary Cirrhosis Data . . . . . . . . . . . . . 2
1.2.2 AIDS Data . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3 Liver Cirrhosis Data . . . . . . . . . . . . . . . . . . . 4
1.2.4 Aortic Valve Data . . . . . . . . . . . . . . . . . . . . 6
1.2.5 Other Applications . . . . . . . . . . . . . . . . . . . . 7
1.3 Inferential Objectives in Longitudinal Studies . . . . . . . . . 8
1.3.1 Effect of Covariates on a Single Outcome . . . . . . . 9
1.3.2 Association between Outcomes . . . . . . . . . . . . . 9
1.3.3 Complex Hypothesis Testing . . . . . . . . . . . . . . 10
1.3.4 Prediction . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.5 Statistical Analysis with Implicit Outcomes . . . . . . 11
1.4 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Longitudinal Data Analysis 15

2.1 Features of Longitudinal Data . . . . . . . . . . . . . . . . . 15
2.2 Linear Mixed-Effects Models . . . . . . . . . . . . . . . . . . 16
2.2.1 Estimation . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.2 Implementation in R . . . . . . . . . . . . . . . . . . . 21
2.3 Missing Data in Longitudinal Studies . . . . . . . . . . . . . 27
2.3.1 Missing Data Mechanisms . . . . . . . . . . . . . . . . 29
2.3.2 Missing Not at Random Model Families . . . . . . . . 32
2.4 Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . 34

3 Analysis of Event Time Data 35

3.1 Features of Event Time Data . . . . . . . . . . . . . . . . . . 35
3.2 Basic Functions in Survival Analysis . . . . . . . . . . . . . . 37
3.2.1 Likelihood Construction for Censored Data . . . . . . 40
3.3 Relative Risk Regression Models . . . . . . . . . . . . . . . . 41
3.3.1 Implementation in R . . . . . . . . . . . . . . . . . . . 42
3.4 Time-Dependent Covariates . . . . . . . . . . . . . . . . . . 44
3.5 Extended Cox Model . . . . . . . . . . . . . . . . . . . . . . 46

xi
xii

3.6 Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . 49

4 Joint Models for Longitudinal and Time-to-Event Data 51

4.1 The Basic Joint Model . . . . . . . . . . . . . . . . . . . . . 51
4.1.1 The Survival Submodel . . . . . . . . . . . . . . . . . 52
4.1.2 The Longitudinal Submodel . . . . . . . . . . . . . . . 54
4.2 Joint Modeling in R: A Comparison with the Extended Cox
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3 Estimation of Joint Models . . . . . . . . . . . . . . . . . . . 61
4.3.1 Two-Stage Approaches . . . . . . . . . . . . . . . . . . 61
4.3.2 Joint Likelihood Formulation . . . . . . . . . . . . . . 62
4.3.3 Standard Errors with an Unspecified Baseline Risk
Function . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3.4 Optimization Control in JM . . . . . . . . . . . . . . . 67
4.3.5 Numerical Integration . . . . . . . . . . . . . . . . . . 68
4.3.6 Numerical Integration Control in JM . . . . . . . . . . 72
4.3.7 Convergence Problems . . . . . . . . . . . . . . . . . . 74
4.4 Asymptotic Inference for Joint Models . . . . . . . . . . . . 78
4.4.1 Hypothesis Testing . . . . . . . . . . . . . . . . . . . . 78
4.4.2 Confidence Intervals . . . . . . . . . . . . . . . . . . . 84
4.4.3 Design Considerations . . . . . . . . . . . . . . . . . . 86
4.5 Estimation of the Random Effects . . . . . . . . . . . . . . . 87
4.6 Connection with the Missing Data Framework . . . . . . . . 89
4.7 Sensitivity Analysis under Joint Models . . . . . . . . . . . . 92

5 Extensions of the Standard Joint Model 97

5.1 Parameterizations . . . . . . . . . . . . . . . . . . . . . . . . 97
5.1.1 Interaction Effects . . . . . . . . . . . . . . . . . . . . 98
5.1.2 Lagged Effects . . . . . . . . . . . . . . . . . . . . . . 100
5.1.3 Time-Dependent Slopes Parameterization . . . . . . . 103
5.1.4 Cumulative Effects Parameterization . . . . . . . . . . 106
5.1.5 Random-Effects Parameterization . . . . . . . . . . . 111
5.2 Handling Exogenous Time-Dependent Covariates . . . . . . . 115
5.3 Stratified Relative Risk Models . . . . . . . . . . . . . . . . . 119
5.4 Latent Class Joint Models . . . . . . . . . . . . . . . . . . . 122
5.5 Multiple Failure Times . . . . . . . . . . . . . . . . . . . . . 128
5.5.1 Competing Risks . . . . . . . . . . . . . . . . . . . . . 128
5.5.2 Recurrent Events . . . . . . . . . . . . . . . . . . . . . 134
5.6 Accelerated Failure Time Models . . . . . . . . . . . . . . . . 137
5.7 Joint Models for Categorical Longitudinal Outcomes . . . . . 139
5.7.1 The Generalized Linear Mixed Model (GLMM) . . . . 139
5.7.2 Combining Discrete Repeated Measures with Survival 141
5.8 Joint Models for Multiple Longitudinal Outcomes . . . . . . 142
 xiii

6 Joint Model Diagnostics 145

6.1 Residuals for Joint Models . . . . . . . . . . . . . . . . . . . 145
6.1.1 Residuals for the Longitudinal Part . . . . . . . . . . 145
6.1.2 Residuals for the Survival Part . . . . . . . . . . . . . 148
6.2 Dropout and Residuals . . . . . . . . . . . . . . . . . . . . . 155
6.3 Multiple Imputation Residuals . . . . . . . . . . . . . . . . . 158
6.3.1 Fixed Visit Times . . . . . . . . . . . . . . . . . . . . 158
6.3.2 Random Visit Times . . . . . . . . . . . . . . . . . . . 162
6.4 Random-Effects Distribution . . . . . . . . . . . . . . . . . . 169

7 Prediction and Accuracy in Joint Models 171

7.1 Dynamic Predictions of Survival Probabilities . . . . . . . . 171
7.1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . 171
7.1.2 Estimation . . . . . . . . . . . . . . . . . . . . . . . . 172
7.1.3 Implementation in R . . . . . . . . . . . . . . . . . . . 174
7.2 Dynamic Predictions for the Longitudinal Outcome . . . . . 182
7.3 Effect of Parameterization on Predictions . . . . . . . . . . . 186
7.4 Prospective Accuracy for Joint Models . . . . . . . . . . . . 194
7.4.1 Discrimination Measures for Binary Outcomes . . . . 195
7.4.2 Discrimination Measures for Survival Outcomes . . . . 197
7.4.3 Prediction Rules for Longitudinal Markers . . . . . . . 198
7.4.4 Discrimination Indices . . . . . . . . . . . . . . . . . . 200
7.4.5 Estimation under the Joint Modeling Framework . . . 202
7.4.6 Implementation in R . . . . . . . . . . . . . . . . . . . 205

A A Brief Introduction to R 219

A.1 Obtaining and Installing R and R Packages . . . . . . . . . . 219
A.2 Simple Manipulations . . . . . . . . . . . . . . . . . . . . . . 220
A.2.1 Basic R Objects . . . . . . . . . . . . . . . . . . . . . 220
A.2.2 Indexing . . . . . . . . . . . . . . . . . . . . . . . . . . 222
A.3 Import and Manipulate Data Frames . . . . . . . . . . . . . 224
A.4 The Formula Interface . . . . . . . . . . . . . . . . . . . . . . 226

B The EM Algorithm for Joint Models 229

B.1 A Short Description of the EM Algorithm . . . . . . . . . . . 229
B.2 The E-step for Joint Models . . . . . . . . . . . . . . . . . . 230
B.3 The M-step for Joint Models . . . . . . . . . . . . . . . . . . 231

C Structure of the JM Package 235

C.1 Methods for Standard Generic Functions . . . . . . . . . . . 235
C.2 Additional Functions . . . . . . . . . . . . . . . . . . . . . . 237

References 239

Index 257
xiv
Chapter 1
Introduction

1.1 Goals
In follow-up studies different types of outcomes are typically collected for each
sample unit, which may include several longitudinally measured responses, and
the time until an event of particular interest occurs. The research questions
of interest in such studies often require the separate analysis of the recorded
outcomes, but in many occasions interest may also lie in studying their as-
sociation structure. A frequently encountered example of the latter case can
be found in biomarker research, where many clinical studies are designed to
identify biomarkers with strong prognostic capabilities for event time out-
comes. Standard examples include among others, human immunodeficiency
virus (HIV) research in which interest lies in the association between CD4
cell counts or viral load and the time to Acquired immune deficiency syn-
drom (AIDS), liver cirrhosis studies which investigate the association between
serum bilirubin and the time to death, and prostate cancer studies in which
interest lies in the association between prostate-specific antigen (PSA) levels
and the time to the development of prostate cancer. An important inherent
characteristic of these medical conditions is their dynamic nature. That is,
the rate of progression is not only different from patient to patient but also
dynamically changes in time for the same patient. Thus, the true potential of
a biomarker in describing disease progression and its association with survival
can only be revealed when repeated evaluations of the marker are considered
in the analysis.