Table 3.

Advantages and disadvantages of different survival type methods for modeling age of motor-onset when genetic mutation status is known.

Method	Advantages	Disadvantages
Parametric methods (Gutierrez and MacDonald, 2002; Langbehn et al., 2004; Zhang et al., 2011)	• Simple implementation. • Predicts likeliness of onset occurring by a given age as an explicit function of subject-specific features. • Predicts probability of onset in s-year intervals (e.g., 5-year intervals). • Estimates penetrance rates at different CAG repeat-lengths. • Provides clinically meaningful divisions of subjects to make crosssectional comparisons.	• Current parametric methods only incorporate CAG repeat-length and age-at-study entry, although many other subject-specific features have been identified to influence age-at-onset. • Searching for explicit relationships between multiple subject-specific features and age-at-onset is challenging. • Prone to model misspecification.
Semiparametric Cox model	• Available in standard software. • Relates linear effect of subject-specific features to a ratio of hazard functions through a log-transformation. • Baseline hazard function left completely unspecified to allow model flexibility.	• Proportional hazards assumption not always satisfied in HD studies (Langbehn et al., 2004).
Nonparametric Kaplan-Meier	• No assumptions about underlying age-at-onset distribution. • Estimates age-at-onset distribution stratified by different subject-specific features (e.g., CAG repeat-length).	• No explicit relationship between the effect of CAG repeat-length and the age-at-onset distribution.
Nonparametric Kernel and Smoothing Splines (Ma and Wang, 2014b)	• Effect of subject-specific features estimated using smooth functions that do not necessarily adhere to any particular explicit formula.	• Computationally demanding. • No explicit final model available and hence no clinically meaningful divisions of subjects.