MJC - Flexible Parametric Joint Modelling 2011

Introduction
Gauss-Hermite quadrature
A new joint model
stjm
Applications
Future work
References
Flexible parametric joint modelling of

longitudinal and survival data
FPM Workshop 10th November 2011
Michael J. Crowther1 ,
Keith R. Abrams1 and Paul C. Lambert1,2
1 Centre
for Biostatistics and Genetic Epidemiology

Department of Health Sciences
University of Leicester, UK.
2 Department
of Medical Epidemiology and Biostatistics

Karolinska Institutet
Stockholm, Sweden.
michael.crowther@le.ac.uk
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
1 / 27
Applications
Future work
References
Outline
Outline
Introduction to joint modelling

A new joint model
Example application - Primary Biliary Cirrhosis (PBC)
Future work
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
2 / 27
Applications
Future work
References
Background
Background
Longitudinal response data affected by informative

dropout
Inclusion of time-varying covariates in survival analyses
Michael J. Crowther
FPM Workshop
3 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
Background
Background

dropout
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
3 / 27
Applications
Future work
References
Background
Background

dropout
Approaches:
Latent class approach (Proust-Lima and Taylor, 2009)
Shared parameter models - dependence through shared
random effects (Wulfsohn and Tsiatis, 1997)
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
3 / 27
Applications
Future work
References
Background
Background

dropout
Approaches:
Latent class approach (Proust-Lima and Taylor, 2009)
Shared parameter models - dependence through shared
random effects (Wulfsohn and Tsiatis, 1997)
Michael J. Crowther
FPM Workshop
3 / 27
Introduction
A new joint model
Applications
stjm
Future work
References
Example dataset
Example dataset
Primary Biliary Cirrhosis (PBC) dataset - 312 patients

with 1945 repeated measurements of serum bilirubin
(Murtagh et al., 1994).
158 randomised to receive D-penicillamine and 154 to
placebo
Interested in the treatment effect after adjusting for the
longitudinal biomarker
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
4 / 27
Applications
stjm
Future work
References
Data structure
Data structure
Longitudinal Data
Survival Data
yij = {yi(tij); j = 1,,mi}
Ti = min(Si, Ci), di = Si Ci
eg. Linear mixed

effects model
eg. Proportional
hazards model
Subject-specific
random effects
bi ~ MVN(0,)
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
5 / 27
stjm
Applications
Future work
References
Data structure
Data structure
. list id logb drug _t0 _t _d if id==3 | id==20, noobs sepby(id)
id
logb
drug
_t0
_t
_d
3
3
3
3
.3364722
.0953102
.4054651
.5877866
D-penicil
D-penicil
D-penicil
D-penicil
0
.48187494
.99660498
2.0342789
.48187494
.99660498
2.0342789
2.7707808
0
0
0
1
20
20
20
20
1.629241
2.727853
2.406945
3.465736
placebo
placebo
placebo
placebo
0
.49556455
.91446722
3.6797721
.49556455
.91446722
3.6797721
3.7126274
0
0
0
1
Michael J. Crowther
FPM Workshop
6 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
Data structure
Data structure
KaplanMeier survival estimates
1.0
Survival probability
0.8
0.6
0.4
0.2
0.0
0
10
15
Followup time
95% Confidence Interval
Michael J. Crowther
Introduction
Log(serum bilirubin) < 1

Log(serum bilirubin) > 1
FPM Workshop
A new joint model
stjm
7 / 27
Applications
Future work
References
Data structure
stjmgraph logb, panel(id)
15
Longitudinal response
0
2
2
0
2
Event
Censored
10
5
Time before censoring
15
10
5
Time before event
Figure: Longitudinal trajectories. Adjusted timescale.

Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
8 / 27
Applications
Future work
References
Model specification
Longitudinal submodel
Linear mixed effects model (Laird and Ware, 1982):

yi (tij ) = Wi (tij ) + eij ,
eij N(0, e2 )
Wi (tij ) = xi (tij ) + zi (tij )bi + ui

.
xtmixed logb time drug || id:
time, cov(unstr)
Increase flexibility through the use of fixed/random

fractional polynomials of time (Royston and Altman,
1994).
Michael J. Crowther
FPM Workshop
9 / 27
Introduction
A new joint model
Applications
stjm
Future work
References
Model specification
Survival submodel
Many choices proposed including standard parametric

and, of course, Cox proportional hazards models
hi (Ti ) = h0 (Ti )exp(Wi (Ti ) + vi )
where vi is a set of baseline covariates and, for example;
Wi (Ti ) = 0i + 1i Ti + ui
is termed the association parameter; in this case we

assume the association is based on the current value of
the biomarker
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
10 / 27
Applications
stjm
Future work
References
Model specification
We can now write down the full joint likelihood:

!
#
"Z
mi
n
Y
Y
f (yi (tij )|bi , ) f (bi |)f (Ti , di |bi , ) dbi
i=1
where
j=1

yi (tij ) Wi (tij )
f (yi (tij )|bi , ) = (2e2 )1/2 exp
,
2
2e
(
)
bi V 1 bi
1/2
exp
,
f (bi |) = (2|V |)
2
and
Z
f (Ti , di |bi , ) = [h0 (Ti )exp(Wi (t) + vi )]di exp
Michael J. Crowther
Introduction
Ti
h0 (u)exp(Wi (u) + vi )du
FPM Workshop
A new joint model
11 / 27
Applications
stjm
Future work
References
Model specification
We can now write down the full joint likelihood:

"Z
!
#
mi
n
Y
Y
f (yi (tij )|bi , ) f (bi |)f (Ti , di |bi , ) dbi
i=1
where
j=1

yi (tij ) Wi (tij )
f (yi (tij )|bi , ) = (2e2 )1/2 exp
,
2
2e
(
)
bi V 1 bi
1/2
exp
,
f (bi |) = (2|V |)
2
and
Z
f (Ti , di |bi , ) = [h0 (Ti )exp(Wi (t) + vi )]di exp
Michael J. Crowther
FPM Workshop
Ti
0
h0 (u)exp(Wi (u) + vi )du
11 / 27
Introduction
A new joint model
Applications
stjm
Future work
References
Numerical method to approximate analytically intractable

integrals (Pinheiro and Bates, 1995)
Z
x 2
wq f (xq )
q=1
Can be extended to multivariate integrals i.e. multiple

random effects
Michael J. Crowther
Introduction
f (x)dx
m
X
FPM Workshop
A new joint model
12 / 27
stjm
Applications
Future work
References
u ~ N(0,1)
Michael J. Crowther
Introduction
0
u
FPM Workshop
A new joint model
13 / 27
stjm
Applications
Future work
References
u ~ N(0,1)
Michael J. Crowther
0
u
FPM Workshop
13 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
u ~ N(0, )
Michael J. Crowther
Introduction
0
u
FPM Workshop
A new joint model
13 / 27
stjm
Applications
Future work
References
u ~ N(0, )
Michael J. Crowther
Introduction
0
u
FPM Workshop
A new joint model
13 / 27
stjm
Applications
Future work
References
u2
u1
Michael J. Crowther
0
u
FPM Workshop
13 / 27
Introduction
A new joint model
Michael J. Crowther
Introduction
A new joint model
Michael J. Crowther
Introduction
Michael J. Crowther
Future work
13 / 27
stjm
0
u
A new joint model
Applications
Future work
FPM Workshop
References
13 / 27
stjm
0
u
References
FPM Workshop
0
u
Applications
FPM Workshop
stjm
Applications
Future work
References
13 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
A new joint model
Survival submodel
Flexible parametric survival model (Royston and Parmar,

2002; Lambert and Royston, 2009)
Modelled on the log cumulative hazard scale using
restricted cubic splines (Durrleman and Simon, 1989)
log{H0 (t)} = s{log(t)|, k0 }
Can evaluate the likelihood directly
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
14 / 27
Applications
Future work
References
Motivation
Marginal survival 1 degree of freedom
.8
.6
.4
.2
KaplanMeier curve
Marginal survival
0
0
2
3
Followup time
Figure: Predicted marginal survival function from joint model with 1 degree of
freedom, overlaid on the Kaplan-Meier survival curve.
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
15 / 27
Applications
Future work
References
Motivation
Marginal survival 5 degrees of freedom
.8
.6
.4
.2
KaplanMeier curve
Marginal survival
0
0
2
3
Followup time
Figure: Predicted marginal survival function from joint model with 5 degrees of
freedom, overlaid on the Kaplan-Meier survival curve.
Michael J. Crowther
FPM Workshop
16 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
Motivation
Survival model linear predictor:

log{H(Ti |bi , vi )} = i = s{log(Ti )|, k0 } + Wi (Ti ) + vi
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
17 / 27
Applications
Future work
References
Motivation
Survival model linear predictor:

log{H(Ti |bi , vi )} = i = s{log(Ti )|, k0 } + Wi (Ti ) + vi
Transform to the hazard and survival scales:

1 ds{log(Ti )|, k0 }
dW (Ti )
h(Ti |bi , vi ) =
+
exp(i )
Ti
d log(Ti )
dTi
S(Ti |bi , vi ) = exp{exp(i )}
(Crowther et al., 2011)
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
17 / 27
Applications
Future work
References
Syntax
Implementation in Stata
stjm longdepvar [varlist ], panel(varname ) df(#)
[gh(#) ...]
Michael J. Crowther
FPM Workshop
18 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
Syntax
[gh(#) ...]
Longitudinal submodel:
ffracpoly(numlist ) - Fixed FPs of time

rfracpoly(numlist ) - Random FPs of time
[varlist ] - Baseline covariates
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
18 / 27
Applications
Future work
References
Syntax
[gh(#) ...]

Survival submodel:
df(#)/knots(numlist ) - Baseline cum. hazard

survcov(varlist ) - Baseline covariates
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
18 / 27
Applications
Future work
References
Syntax
[gh(#) ...]
Survival submodel:

df(#)/knots(numlist ) - Baseline cum. hazard
survcov(varlist ) - Baseline covariates
Association:
nocurrent - Current value is the default

derivassoc(1) - 1st derivative
sepintassoc/sepassoc(numlist) - Random
coefficient, e.g. random intercept
Michael J. Crowther
FPM Workshop
18 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
Applications
Future work
References
Predictions
Predictions
predict newvarname, option
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
19 / 27
Predictions
Predictions
Longitudinal:
xb/fitted - Fitted values

residuals - Subject level residuals
rstandard - Standardised residuals
reffects/reses - Empirical Bayes predictions of
random effects
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
19 / 27
Applications
Future work
References
Predictions
Predictions
Longitudinal:

random effects
Survival:
hazard - Hazard function

survival - Survival function
cumhazard - Cumulative hazard function
martingale - Martingale residuals
stjmcondsurv - Conditional survival
Michael J. Crowther
FPM Workshop
19 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
Predictions
Predictions
Longitudinal:

random effects
Survival:
hazard - Hazard function

survival - Survival function
cumhazard - Cumulative hazard function
martingale - Martingale residuals
stjmcondsurv - Conditional survival
Predictions can be evaluated at measurement/survival times,

or user specified times.
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
19 / 27
stjm
Applications
Future work
References
Application 1
Application to PBC dataset

. stjm logb trt, panel(id) gh(5) survm(fpm) df(1) rfracp(1) survcov(trt) nolog
note: delayed entry models are being fitted
-> gen double timevar_1 = X^(1)
(where X = _t0)
Joint model estimates
Panel variable: id
Log-likelihood = -1907.0077
Number of obs.
=
Number of panels =
Coef.
Longitud.:
timevar_1
trt
_cons
Std. Err.
P>|z|
1945
312
[95% Conf. Interval]
.2014373
-.1321578
.5764177
.0134045
.1094385
.0793
15.03
-1.21
7.27
0.000
0.227
0.000
.1751651
-.3466534
.4209925
.2277096
.0823378
.7318429
-.0112922
.7024496
-3.576883
.1884277
.0947584
.2760118
-0.06
7.41
-12.96
0.952
0.000
0.000
-.3806038
.5167266
-4.117856
.3580193
.8881726
-3.03591
1.100995
.0943797
11.67
0.000
.9160146
1.285976
Survival:
XB
trt
_rcs1
_cons
association
current
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
20 / 27
stjm
Applications
Future work
References
Application 1
Random effects table

Random effects Parameters
Estimate
Std. Err.
id: Unstructured
sd(timevar_1)
sd(intercept)
corr(timevar_1,intercept)
.1990849
.9935064
.4601163
.0148383
.0423497
.0674666
.1720267
.9138753
.3182637
.2303991
1.080076
.5818161
sd(Residual)
.3460383
.0066226
.3332986
.3592649
Longitudinal submodel: Linear mixed effects model

Survival submodel: Flexible parametric model with 1 degree of freedom
Integration method: Adaptive Gauss-Hermite quadrature using 5 nodes
Michael J. Crowther
FPM Workshop
21 / 27
Introduction
A new joint model
Applications
stjm
Future work
References
Predictions
Individual level predictions

stjmcondsurv, panel(id) id(16) fu(20)
Predicted conditional survival

Patient 16
3.5
1.0
3.0
1.0
3.0
0.8
1.0
0.4
0.5
0.0
2.0
0.6
1.5
1.0
0.4
0.5
0.0
0.2
0.5
0.6
1.5
0.8
2.5
2.0
log(serum bilirubin)
2.5
log(serum bilirubin)
Patient 70
3.5
0.2
0.5
1.0
0.0
0
10
15
Followup time
1.0
20
0.0
0
10
15
Followup time
20
Predicted conditional survival
95% pointwise confidence interval
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
22 / 27
Applications
stjm
Future work
References
Application 2
Application to PBC dataset

. stjm logb trt, panel(id) gh(5) survm(fpm) df(1) rfracp(1) ///
> survcov(trt) derivassoc(1) nolog
note: delayed entry models are being fitted
-> gen double timevar_1 = X^(1)
(where X = _t0)
Joint model estimates
Number of obs.
=
Panel variable: id
Number of panels =
Log-likelihood = -1908.3346
Coef.
Longitud.:
timevar_1
trt
_cons
Std. Err.
P>|z|
1945
312
.199051
-.1330277
.5462605
.0092538
.0837321
.063101
21.51
-1.59
8.66
0.000
0.112
0.000
.1809138
-.2971395
.4225849
.2171881
.0310841
.6699361
-.0124688
.8948762
-3.974368
.1925987
.1432524
.3677949
-0.06
6.25
-10.81
0.948
0.000
0.000
-.3899553
.6141066
-4.695232
.3650178
1.175646
-3.253503
.9457228
2.237211
.1200589
1.101684
7.88
2.03
0.000
0.042
.7104118
.077949
1.181034
4.396473
Survival:
XB
trt
_rcs1
_cons
association
current
deriv1
(output omitted )
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
23 / 27
Applications
stjm
Future work
References
Application 2
1.0
0.8
0.6
0.4
0.2
0.0
0
10
15
Followup time
Fitted survival function
KaplanMeier curve
Figure: Predicted marginal survival function for those on treatment.

Michael J. Crowther
FPM Workshop
24 / 27
Introduction
A new joint model
stjm
Applications
Future work
References
Application 2
Standardised residuals
4
1
1
2
Longitudinal fitted values
Figure: Standardised residuals against longitudinal fitted values.

Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
25 / 27
Applications
Future work
References
Future work
References
Future work
Future work
Extension to competing risks

Extension to include a cure proportion
Longitudinal categorical responses
EM algorithm
Clustering
Michael J. Crowther
Introduction
FPM Workshop
A new joint model
stjm
26 / 27
Applications
Future work
References I
M. J. Crowther, K. R. Abrams, and P. C Lambert. Flexible parametric joint modelling of longitudinal and survival
data. Submitted, 2011.
S. Durrleman and R. Simon. Flexible regression models with cubic splines. Statistics in Medicine, 8(5):551561,
1989.
N. M. Laird and J. H. Ware. Random-effects models for longitudinal data. Biometrics, 38(4):963974, 1982.
P. C Lambert and P. Royston. Further development of flexible parametric models for survival analysis. The Stata
Journal, 9:265290, 2009.
P. Murtagh, E. Dickson, M. Van Dam, G. Malincho, and P. Grambsch. Primary biliary cirrhosis: Prediction of
short-term survival based on repeated patient visits. Hepatology, 20:126134, 1994.
Jos
e C. Pinheiro and Douglas M. Bates. Approximations to the log-likelihood function in the nonlinear mixed-effects
model. Journal of Computational and Graphical Statistics, 4(1):pp. 1235, 1995. ISSN 10618600.
C. Proust-Lima and J. M. G. Taylor. Development and validation of a dynamic prognostic tool for prostate cancer
recurrence using repeated measures of posttreatment psa: a joint modeling approach. Biostatistics, 10(3):
535549, 2009.
P. Royston and D. G. Altman. Regression using fractional polynomials of continuous covariates: Parsimonious
parametric modelling. Journal of the Royal Statistical Society. Series C (Applied Statistics), 43(3):429467,
1994.
P. Royston and M. K. B. Parmar. Flexible parametric proportional-hazards and proportional-odds models for
censored survival data, with application to prognostic modelling and estimation of treatment effects. Stat Med,
21(15):21752197, 2002.
M. S. Wulfsohn and A. A. Tsiatis. A joint model for survival and longitudinal data measured with error.
Biometrics, 53(1):330339, 1997.
Michael J. Crowther
FPM Workshop
27 / 27

MJC - Flexible Parametric Joint Modelling 2011

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

MJC - Flexible Parametric Joint Modelling 2011

Încărcat de

Drepturi de autor:

Formate disponibile

Introduction

A new joint model

Flexible parametric joint modelling of

for Biostatistics and Genetic Epidemiology

of Medical Epidemiology and Biostatistics

Introduction to joint modelling

Longitudinal response data affected by informative

A new joint model

Longitudinal response data affected by informative

Longitudinal response data affected by informative

Longitudinal response data affected by informative

A new joint model

Primary Biliary Cirrhosis (PBC) dataset - 312 patients

A new joint model

yij = {yi(tij); j = 1,,mi}

eg. Linear mixed

A new joint model

A new joint model

Log(serum bilirubin) < 1

A new joint model

stjmgraph logb, panel(id)

Figure: Longitudinal trajectories. Adjusted timescale.

A new joint model

Linear mixed effects model (Laird and Ware, 1982):

Wi (tij ) = xi (tij ) + zi (tij )bi + ui

xtmixed logb time drug || id:

Increase flexibility through the use of fixed/random

A new joint model

Many choices proposed including standard parametric

is termed the association parameter; in this case we

A new joint model

We can now write down the full joint likelihood:

h0 (u)exp(Wi (u) + vi )du

A new joint model

We can now write down the full joint likelihood:

h0 (u)exp(Wi (u) + vi )du

A new joint model

Numerical method to approximate analytically intractable

Can be extended to multivariate integrals i.e. multiple

A new joint model

A new joint model

A new joint model

A new joint model

A new joint model

A new joint model

A new joint model

A new joint model

A new joint model

A new joint model

Flexible parametric survival model (Royston and Parmar,

Can evaluate the likelihood directly

A new joint model

Marginal survival 1 degree of freedom

A new joint model

Marginal survival 5 degrees of freedom

A new joint model

Survival model linear predictor:

A new joint model

Survival model linear predictor:

Transform to the hazard and survival scales:

A new joint model

ffracpoly(numlist ) - Fixed FPs of time

ffracpoly(numlist ) - Fixed FPs of time

df(#)/knots(numlist ) - Baseline cum. hazard