ABSTRACT
This chapter is devoted to models for the analysis of longitudinal data defined as time-dependent repeated data. The dependent variable Y of interest is measured repeatedly at different times on the same subjects. Such longitudinal data are frequent in epidemiology. For example, in observational cohort studies and clinical trials of HIV infected patients, the concentration of CD4 + T lymphocytes (CD4) (CD4 counts) and the plasma viral load are measured repeatedly during the follow-up. The analysis of the time course of these two markers that are predictors of the clinical progression of the disease enables the natural history of the disease to be described, and the effectiveness of treatments to be evaluated. In many other diseases, repeated measures of biomarkers are used to track the progression of a patient’s condition and anticipate the occurrence of clinical events. For instance, in prostate cancer, an increase of the prostate specific antigen (PSA) is a predictor of prostate cancer relapse after an initial treatment. The study of normal and pathological cognitive aging in the elderly is also based on repeated measures of cognitive tests. In the three above examples, the variables of interest are quantitative variables either continuous (CD4 counts, viral load, PSA) or discrete (cognitive tests). There are also numerous examples of longitudinal binary data, such as daily measures of truancy among children or of asthma attacks in asthmatic patients.
