||1R21CA239168-01A1 Interpret this number
||Fred Hutchinson Cancer Research Center
||Joint Modeling of Longitudinal Physical Activity and Diet Data and Survival
The proposed project is in response to PAR-18-857, Diet and Physical Activity Assessment Methodology.
We are primarily interested in developing and applying innovative statistical methods for joint modeling of
incidence or mortality outcomes and longitudinal physical activity and dietary intake data. Physical activity
and diet are important behavioral factors involved in the etiology of many chronic diseases, such as cardio-
vascular disease and cancer. We are motivated by longitudinal physical activity measurement issues involved
in the Women's Health Initiative (WHI) and the Health, Eating, Activity, and Lifestyle (HEAL) study. The WHI
observational study (OS) is a prospective study of health outcomes among 93,676 postmenopausal women
enrolled between 1993 and 1998 in 40 U.S. clinical centers. The HEAL study enrolled approximately 1,200
women with early stage breast cancer, diagnosed between 1997-1998, and who have been followed up for
more than 15 years. Many nutrient or physical activity measures may have a zero value (or a low de-
tectable value) among a group of individuals. Our preliminary simulation results demonstrate that a
naive method without taking into account measurement error, or a standard bias correction for mea-
surement error without taking into account a subset of individuals with zero values may lead to bias
in the effect estimation in regression analysis. Speciﬁc aims of this proposal include: (i) To develop and
apply methods to adjust for measurement error in joint modeling of binary incidence or mortality outcomes
and longitudinal measures of physical activity and dietary intake data, which may be zero among some indi-
viduals. (ii) To develop and apply methods to adjust for measurement error in joint modeling of time-to-event
survival outcomes and longitudinal measures of physical activity and dietary intake data, which may be zero
among some individuals. (iii) To develop and apply methods for time-varying association between time-to-
event outcomes and longitudinal measures of physical activity and dietary intake data. The proposed models
and methods will be applied to the physical activity and dietary data from both the WHI-OS and the HEAL
study. Our new methods will be written in user friendly subroutines via R, which can be used by statisticians,
epidemiologists, and other public health professionals. The methods developed in the proposal will have
general applications to other studies where longitudinal physical activity and dietary data are available.
Multinomial logistic regression with missing outcome data: An application to cancer subtypes.
, Hsu L.
Statistics in medicine, 2020-10-30; 39(24), p. 3299-3312.
Semiparametric regression calibration for general hazard models in survival analysis with covariate measurement error; surprising performance under linear hazard.
, Song X.
Biometrics, 2020-06-18; , .