Improved longitudinal data analysis for cross-over design settings, with a piecewise linear mixed-effects model

Abstract

Abstract Repeated measures data are commonly encountered in a wide variety of disciplines including business, agriculture and medicine. The fact that observations from the same unit, in general, will not be independent poses particular challenges to the statistical procedures used for the analysis of such data. In the statistical literature, analysis of cross-over designs is mainly centred around a single response variable measured at the end of each period after treatment. Less commonly, cross-over design studies are used in more complex settings, for example, repeated measurements collected within each center across a number of centers or within individual’s treatment period(s). A single measurement response analysis approach may lead to loss of information that otherwise would be captured during patients follow-up, thus affecting precision in estimation. To circumvent this limitation, we propose the application of a piecewise linear mixed-effects model. We analyze data from a cross-over design, where both systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured repeatedly for each patient within each period. These are continuous variables assumed to arise from the family of Gaussian multivariate distributions. The objective of the study was to investigate changes in the two response variables over time and to detect the role of two treatment dosages of Iodine in household salt associated with a more rapid decrease in the two outcomes.