Special considerations in analyzing the longitudinal study Using a longitudinal study design, we can repeatedly and simultaneously measure both the time-varying prognostic factors and the outcome(s) of interest, allowing direct relationships to be established. Examples of prognostic factors that often vary over time may include a biomarker of disease activity, or alterations in treatment. Prognostic factors that fluctuate or emerge later during the course of the disease can only be understood using the longitudinal design. In contrast, by measuring the outcomes repeatedly over time– for each individual– the longitudinal design captures important prognostic information about within-individual differences and allows these effects to be distinguished from between-individual differences.

The effects of within-individual differences cannot be differentiated from that of between-individual differences in cross-sectional studies. In studies where outcomes are assessed once, the differences in outcomes are usually attributed to differences between individuals. Questions that can be addressed by a longitudinal studyĪ longitudinal study can answer questions about the sources of variability in observed outcomes.

We will focus on application of these methods to a real-life clinic-based rheumatic disease cohort. This paper aims to: 1) Provide a review of common methods used to analyze longitudinal trajectory data and 2) Demonstrate how to interpret results from longitudinal trajectory analysis. These complex methods are harder to understand and use. Although some of these methods have been available for many years, they are not commonly used in the literature. Longitudinal studies require special longitudinal statistical analysis. By measuring patients’ outcomes repeatedly (by definition, on ≥ 3 occasions), a longitudinal study provides information about the shape of outcome trajectory, e.g., whether the disease goes into remission, waxes and wanes or remains persistently active.

To understand the disease course, information about how outcomes change over time is necessary. Two patients may have similar outcomes at a point in time but how they arrived at their outcomes may have been very different. When the outcome of interest is not a terminal event (e.g., death), but a dynamic one (e.g., disease activity, functional status), a cross-sectional view is usually not the best way to look at the data. Outcome studies in rheumatic diseases have generally focused on single-occasion outcome assessments, e.g., 5-year outcome.

Skin trajectory analysis