Background TRACK-HD is a multinational prospective observational study of Huntington's disease (HD) that examines clinical and biological findings of disease progression in individuals with premanifest HD (preHD) and early-stage HD. We aimed to describe phenotypic changes in these participants over 36 months and identify baseline predictors of progression. Methods Individuals without HD but carrying the mutant huntingtin gene (classed as preHD-A if ≥10·8 years and preHD-B if <10·8 years from predicted onset), participants with early HD (classed as HD1 if they had a total functional capacity score of 11–13 and HD2 if they had a score of 7–10), and healthy control individuals were assessed at four study sites in the Netherlands, the UK, France, and Canada. We measured 36-month change for 3T MRI, clinical, cognitive, quantitative motor, and neuropsychiatric assessments and examined their prognostic value. We also assessed the relation between disease progression and the combined effect of CAG repeat length and age. All participants were analysed according to their baseline subgroups. Longitudinal results were analysed using a combination of repeated-measure weighted least squares models and, when examining risk of new diagnosis, survival analysis. Findings At baseline, 366 participants were enrolled between Jan 17, and Aug 26, 2008, and of these 298 completed 36-month follow-up: 97 controls, 58 participants with preHD-A, 46 with preHD-B, 66 with HD1, and 31 with HD2. In the preHD-B group, several quantitative motor and cognitive tasks showed significantly increased rates of decline at 36 months, compared with controls, whereas few had at 24 months. Of the cognitive measures, the symbol digit modality test was especially sensitive (adjusted mean loss 4·11 points [95% CI 1·49–6·73] greater than controls; p=0·003). Among psychiatric indicators, apathy ratings specifically showed significant increases (0·34 points [95% CI 0·02–0·66] greater than controls; p=0·038). There was little evidence of reliable change in non-imaging measures in the preHD-A group, with the exception of the speeded tapping inter-tap interval (0·01 s [95% CI 0·01–0·02] longer than controls; p=0·0001). Several baseline imaging, quantitative motor, and cognitive measures had prognostic value, independent of age and CAG repeat length, for predicting subsequent clinical diagnosis in preHD. Of these, grey-matter volume and inter-tap interval were particularly sensitive (p=0·013 and 0·002, respectively). Longitudinal change in these two measures was also greater in participants with preHD who received a diagnosis of HD during the study compared with those who did not, after controlling for CAG repeat length and age-related risk (p=0·006 and 0·0003, respectively). In early HD, imaging, quantitative motor, and cognitive measures were predictive of decline in total functional capacity and tracked longitudinal change; also, neuropsychiatric changes consistent with frontostriatal pathological abnormalities were associated with this loss of functional capacity (problem behaviours assessment composite behaviour score p<0·0001). Age and CAG repeat length explained variance in longitudinal change of multimodal measures, with the effect more prominent in preHD. Interpretation We have shown changes in several outcome measures in individuals with preHD over 36 months. These findings further our understanding of HD progression and have implications for clinical trial design. Funding CHDI Foundation.