Efficacy of a Prediction Model-based Algorithm to PREVENT Drug-induced Impulse Control Disorders in Parkinson's Disease

Impulse control disorders and related behaviors (ICDRBs) are characterized by pathological gambling, compulsive shopping or eating, and hypersexuality, but other related behaviors have been described, e.g. hobbyism, and punding. ICDRBs are frequent in Parkinson's Disease (PD), affecting up to 50% of the patients after 5 years with major medical, social, and legal impact, with life changing consequences for patients and caregivers. The main risk factor is dopaminergic therapy, particularly the cumulative dose of dopamine agonists (DA). On the other hand, the dopaminergic therapy is necessary to control motor symptoms, and DA have demonstrated efficacy in delaying motor complications occurring in PD. Ideally, dopaminergic therapy would have to be adjusted to the individual risk of developing ICRDBs to maximize the benefit/risk ratio of each drug. However, despite several clinical risk factors associated with the risk of ICDRBs (in addition to the dopaminergic therapy), it is still not possible to predict their risk at the individual level, and not every patient treated with dopaminergic medications will develop ICDRBs. A machine learning algorithm to predict ICDRBs, based on clinical data, validated by cross-validation on independent replication cohorts has been developed. The PREVENT-ICD study proposes to test the efficacy of a new application, ICD-Shield, based on an algorithm to predict and prevent ICDs,in a multicenter randomized controlled trial to prevent ICDRBs in PD patients by proposing to the clinician treatment adjustment according to the risk predicted by the algorithm, as compared to the standard of care (SoC)