Statistical approaches for optimizing outcome measures and covariates for the development of disease progression modeling in Huntington's disease

Xiaomeng Jiang

Disease progression models are important tools in the drug development process, and can decrease time from first-in-human trials to approval for therapeutic use. These models can be used for investigating disease mechanisms, evaluating therapeutic efficacy, guiding hypothesis development, and clinical trial design. Comprehensive disease progression models require sensitive, specific and reproducible outcome measures, and often are informed by covariate factors that modulate disease symptomology. After reviewing the state of the art with respect to disease progression models of neurodegenerative disorders, an effort was undertaken to develop a novel statistical framework for optimizing outcome measures and identifying significant covariates of Huntington's disease as a model neurodegenerative disorder. Using data mined from a publically-available HD database (>3000 patients), the strength of relationships between 19 clinical outcome measures and an objective measure of HD severity, disease burden score (DBS), was evaluated. All outcome measures in the motor, functional, and cognitive domains were significantly (p<0.05) correlated with DBS and best described by sigmoidal models. No statistical relationships between behavioral outcome measures and DBS were identified. Based upon these results, the best two outcome measures for the motor, functional and cognitive domains were selected for inclusion in the covariate analysis portion of the project. Several patient-specific factors were selected for evaluation as potential covariates, based on published evidence for association with HD symptomology; other factors were evaluated as negative controls. Using multiple least-squares regression with stepwise forward-addition, gender, BMI, and use of alcohol, caffeine, tetrabenazine, antipsychotics, or antidepressants generally were identified as significant covariates of motor, functional and cognitive outcome measures. Smoking, handedness, marital status, amphetamines, proton pump inhibitors and histamine-2 receptor antagonists were generally not associated with HD severity. Depending on the outcome measure, models that included all relevant covariates could explain as much as 58% of the variability in the relationship between outcome measure and DBS. These results indicate that standard statistical approaches are appropriate for identifying optimal outcome measures and important covariates for inclusion in formalized disease progression models. This novel framework could inform and streamline modeling efforts, and is generally applicable regardless of the disease or therapeutic class under investigation.

Statistical approaches for optimizing outcome measures and covariates for the development of disease progression modeling in Huntington's disease

Files and links (1)

Abstract

Metrics

Details