Statistical Approaches to Longitudinal Data Analysis in Neurodegenerative Diseases: Huntington’s Disease as a Model

Abstract

Understanding the overall progression of neurodegenerative diseases is critical to the timing of therapeutic interventions and design of effective clinical trials. Disease progression can be assessed with longitudinal study designs in which outcomes are measured repeatedly over time and are assessed with respect to risk factors, either measured repeatedly or at baseline. Longitudinal data allows researchers to assess temporal disease aspects, but the analysis is complicated by complex correlation structures, irregularly spaced visits, missing data, and mixtures of time-varying and static covariate effects. We review modern statistical methods designed for these challenges. Among all methods, the mixed effect model most flexibly accommodates the challenges and is preferred by the FDA for observational and clinical studies. Examples from Huntington’s disease studies are used for clarification, but the methods apply to neurodegenerative diseases in general, particularly as the identification of prodromal forms of neurodegenerative disease through sensitive biomarkers is increasing.

Statistical considerations for drawing conclusions about recovery

“…38 Further, mixed-effect models would allow researchers to model change based on more data points, irregularly timed data points, individuals with missing data, and even model change non-linearly if needed. 39 Mixed-effect regression models are not a panacea of course, they have their own limitations and are computationally more complex, but we do think they would be a more appropriate tool teasing apart individual differences in stroke recovery. Recent work has made strides in this direction 16 and we think this work is an excellent example of how we can meaningfully investigate variation in recovery following stroke without burdening ourselves with the mathematical artifacts of difference scores.…”

Mood disorders and prospective suicidality in young adults: a population‐based cohort study

“…GEE have several advantages when analyzing data from longitudinal studies, including robustness against the wrong choice of correlation matrix, ability to use all available data for analysis, and being especially suitable for the longitudinal analysis of categorical data . While being able to handle potentially incorrectly specified models, GEE models still assume that data are missing completely at random, which can be an issue with excessive missingness . In the model used here, any suicidality was the outcome, and the follow‐up wave (baseline or follow‐up) was used as the time measure.…”

Modeling Longitudinal Outcomes: A Contrast of Two Methods

“…The problems that arise in evaluating change vary with data structure and trade-offs one may face in selecting from available methods. Garcia and Marder (2017) discuss three problems common to longitudinal data: (1) correlations within the data, (2) irregularly timed measurements, and (3) missing data. Correlations within data can exist when individuals are measured repeatedly over time or when individuals are clustered within the data.…”

“…In weighing the effects of these different constraints, Garcia & Marder (2017) discuss other methods beyond mixed-effect regressions and repeated measures ANOVA (i.e., change scores, multivariate ANOVA, and generalized estimating equations), but the key distinction between ANOVA and other methods is that ANOVA addresses questions of differences, but does not explicitly model time. In contrast, mixed-effects regression explicitly models trajectories that are fit to the available data.…”

“…In general, we argue that mixed-effects regression is a preferable approach for modeling change over time for its advantages in: (1) correlations within the data, (2) irregularly timed measurements, and (3) missing data. However, employing mixedeffects regression is not suitable for all situations and does come with trade-offs (Garcia & Marder, 2017;Molenberghs & Verbeke, 2001). To use mixed-effects regression, researchers need additional knowledge of how to specify appropriate models and need to appreciate the implications of choices made in the model-building process.…”