Table 3.
Studies Using Clustering Methods to Stratify Attention-Deficit/Hyperactivity Disorder
| Study | Subjects (N) | Measures | Algorithm | No. of Clusters (Method) | Cluster Descriptions | External Validation |
|---|---|---|---|---|---|---|
| Fair et al., 2012 (15) | ADHD (285) and TDC (213) | Neuropsychologic scores | CD (33) | 6 for ADHD (determined implicitly by the algorithm) | Response time variability (+) | None |
| Working memory (–), memory span (–), inhibition (–), and output speed (–) | ||||||
| Working memory (–), memory span (–), inhibition (–), and output speed (–), minor differences in remaining measures | ||||||
| Temporal processing (–) | ||||||
| Arousal (–) | ||||||
| Arousal (–), minor differences in remaining measures | ||||||
| Karalunas et al., 2014 (14) | ADHD (247) and TDC (190) | Personality measures (e.g., temperament) | CD | 3 (determined implicitly by the algorithm) | Mild | Physiological (e.g., cardiac) measures, resting state fMRI and 1-year clinical outcomes |
| Surgent (positive apporach motivation) | ||||||
| Irritable (negative emotionality, anger, and poor soothability) | ||||||
| Gates et al., 2014 (16) | ADHD (32) and TDC (58) | fMRI (functional connectivity) | CD | 5 (determined implicitly by the algorithm) | Subgroups characterized in terms of functional connectivity profiles | None |
| Costa Dias et al., 2015 (17) | ADHD (42) and TDC (63) | fMRI (reward related functional connectivity) | CD | 3 (determined implicitly by the algorithm) | Subgroups characterized in terms of functional connectivity profiles | Clinical variables and reward sensitivity |
| Van Hulst et al., 2015 (67) | ADHD (96) and TDC (121) | Neuropsychological scores | LCCA | 5 (BIC) | Quick and accurate | Parent ratings of behavioral problems |
| Poor cognitive control | ||||||
| Slow and variable timing | ||||||
| Remaining 2 groups were too small to characterize | ||||||
| Mostert et al., 2015 (106) | ADHD (133) and TDC (132) | Neuropsychological scores | CD | 3 (determined implicitly by the algorithm) | Attention (–), inhibition (–) | Clinical symptoms and case history |
| Reward sensitivity (+) | ||||||
| Working memory (–) and verbal fluency (–) |
External validation is defined as a data measure used to validate the derived classes that is of a different type to the data use to derive the classes. Wherever possible, we follow the authors’ own nomenclature for describing clusters, and a (+) or (–) indicates relative improvement or deficit in the specified variable.
ADHD, attention-deficit/hyperactivity disorder; BIC, Bayesian information criterion; CD, community detection; fMRI, functional magnetic resonance imaging; LCCA, latent class cluster analysis; TDC, typically developing control.