• Text size

Life Course Epidemiology (LCE)

Print 

Life course epidemiology tries to understand and establish the dynamic character of the association and changes in risk factors and exposures with disease outcome over a life time of a person. Concepts like accumulation of risk, critical and sensitive periods, trajectory and path analysis are important issues.
The focus is on observational longitudinal studies.  A  high level of methodological knowledge and research is required in the field of longitudinal data analysis for studies that cover a large life span of patients or healthy individuals.


Programme Leaders   Mission and Description of the Programme  

Mission
To develop, present, and discuss methodology for life course epidemiology. Life course epidemiology can be defined as the study of long-term effects on chronic disease risk of exposures during gestation, childhood, adolescence, young adulthood and later adult life. In epidemiological practice this means longitudinal follow-up research. Since ageing starts at fertilization, these studies should not be limited to specific age groups (e.g. the elderly) alone, but instead should have include or combine cohort studies with different age groups.

The type of methodology should be suitable or applicable to cohort studies like LifeLines to be able to understand the association of changes in risk factors and disease outcome. LifeLines is a prospective population-based study among ultimately 165.000 inhabitants from the three northern provinces of the Netherlands, using a three-generation family design. The overall aim of LifeLines is to unravel the interaction between genetic and environmental factors in the development of multifactorial diseases over the life course.

Description of the Programme
Traditional epidemiology typically focuses on the association of risk factors and exposures with disease outcome, using cross-sectional, case-control, or cohort studies, without specifically addressing the dynamic character of changes in risk factors and exposures over a life time of a person.

LCE wants to assess the relationship between changes in risk factors over time and disease outcomes. In epidemiological practice this can be best studied with longitudinal follow-up research. Since ageing starts at fertilization, these studies should not be limited to adults or even the elderly, where most disease obviously occurs.

A cornerstone of LCE’s research concerns LifeLines, which is a prospective population-based study among ultimately 165.000 inhabitants from the three northern provinces of the Netherlands, using a three-generation family design. The overall aim of LifeLines is to unravel the interaction between genetic and environmental factors in the development of multifactorial diseases over the life course. Besides LifeLines, the LCE program contains the vast long lasting experience of population cohort studies at the UMCG, ranging from the oldest population study in the Netherlands (Vlagtwedde-Vlaardingen study) to the largest (LifeLines). Other important cohort studies included in LCE are TRAILS, Prevend, and GECKO.

This program focuses on one technique in epidemiology to unravel healthy ageing, i.e. longitudinal observational cohort studies. The common interest and expertise concerns the methodology to analyze data from these, often large-scale, datasets. Medical statistics as well as genetic epidemiology, both important methodological disciplines, are included in LCE.  Research themes include 

  1. Methodological approaches for longitudinal data analysis of cohort studies
  2. Etiology and prognosis of chronic age-related diseases

The research is conducted in multidisciplinary collaboration of methodological experts within LCE and disease-specific clinical experts. Members of the program are primarily based in the Department of Epidemiology, ICPE and Public Health. The researchers have a strong common interest in both research themes and aims.

Statistical methodology focuses on longitudinal data analysis of cohort studies, in particular on generalized, non-linear, and linear mixed effects models. This type of models entail a large set of statistical models that could determine the association of risk factors with longitudinal disease outcomes and describe the dynamics of risk factors and health outcome simultaneously. It would include variance component models, latent variable models in its widest form, non-linear growth curves, and many other types of models. The family of models could also deal with different types of health outcomes (continuous, ordinal, binary). Furthermore, the program would also focus on multivariate type of analysis to be able to analyze the longitudinal data in risk factors and health outcomes. This class of methods would include structural equation modeling, causal search inference, and other pathway analysis.

Relevance to Healthy Ageing  

LCE is the core of Healthy Ageing research in human populations. Following groups of unselected individuals or specific patients over the life span, and investigate their risk factors for incidence of disease and disease complications (or absence of disease) will identify relevant risk factors and predictors for Healthy Ageing. An important aspect of the LCE program is the inclusion of non-clinical, that is societal, research groups. Healthy Ageing is more than the absence of disease. Social capital, informal care, distance to health care facilities and other societal factors contribute to health and well-being at older – but also younger - ages.

An example of LCE’s research projects   

The need of statistical methodology for Life Course Epidemiology
Researcher: Nazanin Nooraee

LCE data can be either continuous or discrete, but a more common type of outcome data is typically of the ordinal form. The different types of outcome variables require different statistical methods. Although these methods have been well developed for continuous and discrete outcomes, they have not been fully developed for ordinal outcomes.
In many applications, multiple ordinal outcomes on each subject are combined into a sum score, also referred to as a scale. These scales are typically analyzed as continuous outcomes, but they frequently do not satisfy the necessary underlying statistical conditions. This approximate analysis may lead to bias in the estimation of the effect of risk factors on disease outcome. The consequence may be that either an irrelevant risk factor is accepted or a relevant risk factor is ignored. An alternative approach to this statistical analysis is finding a set of continuous variables that lies underneath the categorical outcomes. These type of models are referred to as latent variable models and they introduce a one-to-one relationship between continuous variables and categorical outcomes.
A real challenge is modeling the correlation structure of the latent variable. Research in this area is important to LCE, since it may better describe the association of time varying risk factors and disease status measured with ordinal outcomes. One important aspect in the longitudinal data analysis is handling the missing data, since they occur in almost all studies. Reasons for missing data are mistakes in the input of data, non-response to several items from a questionnaire, drop-out of subjects, laboratory errors, and others. Although modern statistical methods have been developed to handle missing data, these methods are not always implemented due to its complexity and lack of standard software. Less sophisticated methods, which have been demonstrated to give inappropriate estimates of the effects, are still popular among researchers. One of the research topics is to investigate what type of missing data approach can be considered best in analyzing longitudinal data when specific conditions on the missing data patterns are satisfied. This area is less developed for longitudinal multivariate ordinal data. Again, this research may lead to better understanding of how the LCE factors influence disease outcomes or quality of life.

Principal Investigators / nr of PhD students