FROM LANCET NEUROLOGY
An ensemble of prediction models developed using genetic data and both baseline molecular and clinical variables from patients with Parkinson’s disease, as well as healthy controls, confirmed established predictors of the disease and identified new ones in a longitudinal cohort study.
Furthermore, the models were shown via simulated, randomized, placebo-controlled trials to have utility for clinical trial design and evaluation, as well as for clinical disease monitoring and treatment, Jeanne C. Latourelle, DSc , of GNS Healthcare, Cambridge, Mass., and her colleagues reported in Lancet Neurology.
The findings are important given the substantial heterogeneity in the presentation of Parkinson’s symptoms, which makes it difficult for care providers to give accurate prognoses and for researchers to develop drugs to modify the disease course, the investigators said, noting that “increasing evidence supports a complex interplay between genetic, biological, and molecular abnormalities of the disease, accounting for this heterogeneity between patients.”
“Understanding the causal and physiological factors that contribute to this variability in the evolution of symptoms of Parkinson’s disease is, therefore, a high priority area of Parkinson’s disease research,” they wrote (Lancet Neurol. 2017 Sep 25. doi: 10.1016/S1474-4422[17]30328-9 ).
The investigators developed the models by applying a Bayesian, machine-learning, multivariate predictive inference platform – known as Reverse Engineering and Forward Simulation – to data from the Parkinson’s Progression Markers Initiative (PPMI) study; they used these models to predict the annual rate of change in combined scores from the Movement Disorder Society–Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) parts II and III. The investigators tested the overall explanatory power of the models using the coefficient of determination (R2); their models also replicated novel findings from an independent clinical cohort from the Longitudinal and Biomarker Study in Parkinson’s disease (LABS-PD).
In 117 healthy age- and sex-matched controls and 312 patients with early-stage Parkinson’s disease from the PPMI study, the model ensemble showed strong performance (fivefold cross-validated Pearson R2, 41%), the investigators said.
In the 317-patient LABS-PD validation cohort, the performance of the models was reduced, but still statistically significant (R2, 9%).
Individual predictive features – including significant replication of higher baseline motor score on MDS-UPDRS, male sex, increased age, and novel Parkinson’s disease–specific epistatic interaction, which all were indicative of faster motor progression in the disease – were confirmed in the LABS-PD cohort, they said. The most useful predictive marker of motor progression was genetic variation (R2, 2.9%), the investigators said, and the cerebrospinal fluid biomarkers at baseline also significantly affected prediction of motor progression, although more modestly (R2, 0.3%).
In 5,000 trial simulations, the incorporation of predicted rates of motor progression into the final models of treatment effect reduced the variability in study outcome, which allowed for the detection of significant differences in sample sizes up to 20% smaller than in trials of drug-naive patients.
The investigators noted that the Bayesian model inference using Reverse Engineering and Forward Simulation was particularly good at predicting Parkinson’s motor progression in early disease stages, which has “immediate relevance toward enabling more effective trial recruitment and clinical disease management.”
This study was supported by grants from the Michael J. Fox Foundation for Parkinson’s Research and the National Institute of Neurological Disorders and Stroke. PPMI is funded by the Michael J. Fox Foundation for Parkinson’s Research, and it also has many industry funding partners. Dr. Latourelle and several of her coauthors are or were employees of GNS Healthcare.
