Background There are hundreds of publications arguing for the next best cognitive test for HD, yet little uniformity exists among professionals on which criteria determines “best”.

Aims Data from over 1300 participants enrolled in the PREDICT-HD study completed annual cognitive exams from 1 to 12 years. Cross-sectional comparisons with healthy controls, longitudinal changes over varying time epochs, as well as efforts to use cognitive test scores as predictors of various outcomes will be shown to challenge the field to develop uniform criteria that might expedite efficient review of the literature as well as the future design of new tests and their validation for HD clinical trials.

Methods Findings suggest that the “best cognitive test” varies with the design of the clinical trial as well as the sample recruited for study. The PREDICT-HD participants were stratified into High, Medium, and Low groups based on cumulative genetic toxicity at study entrance. Linear mixed effects regression examined the intercept and longitudinal trend for 29 cognitive variables. All models controlled for age, education, and gender. Models were evaluated using Akaike's information criterion and likelihood ratio test. Annualised rates of change were obtained.

Results Findings showed that utility of cognitive tasks varied in several ways. Some tests were most robust as methods for participant selection (eg, most different from controls or most “predictive” of entering the rapid progressive phase seen immediately prior to motor diagnosis). Other cognitive tests were considered “best” in the amount of change while controlling for practice effects found in controls.

Conclusions Findings are discussed in terms of clinical trial design.