Rationale and Objective: Current approaches for diagnosing cognitive dysfunction/impairment in Parkinson's disease (PD) utilize a limited approach of classifying the patients into categories of mild cognitive impairment (abbreviated as PD-MCI) and dementia (abbreviated as PD-D), both of which are then limited to labels of either "Yes" or "No". In sharp contrast, cognitive impairment in PD can vary for different patients in terms of severity on a continuous scale. Furthermore, cognitive impairment in PD is heterogenous in nature, given that there is selective impairment of specific cognitive functions and only a subset of patients develop it early in the disease. In addition, the PD-MCI and PD-D approaches of classifying cognitive dysfunction make a rather superficial use of neuropsychological/cognitive testing data for diagnostic conclusions, and thus fail to capture the richness of these data. The categorical PD-MCI and PD-D classifications also have limited ability to accurately reflect on progression, as patients can remain labelled only in either of these two categories several years despite having worsening in their cognitive function. Thus, several key aspects of cognitive dysfunction in PD are not fully addressed by the currently available classification schemes.

The proposed study aims to address the above issues by developing and testing a new approach to studying cognitive impairment in PD and is titled "Cognitive Functions Impairment as a Novel Paradigm for Delineating Cognitive Dysfunction in Parkinson's Disease (PD-CFI)." Specifically, we plan to apply a novel statistical modelling technique, called partial ordered set (POSET), for developing the PD-CFI classification scheme (scores) using extensive data developed specifically for studying cognitive impairment in PD. We also plan to validate the PD-CFI classification scheme using an artificial intelligence (machine learning) method in a multi-center clinical research study. Of note, the POSET models have been used successfully in a similar fashion in Alzheimer's disease, and we have done preliminary work to show the feasibility of this approach in PD. Furthermore, the proposed research project will be implemented within the framework of our Ontology-driven Machine Learning Informatics System for PD (ORMIS-PD), which is a technological platform funded by the Department of Defense for capturing PD-specific patient information.

This research is a collaboration among a movement disorders neurologist with clinical informatics expertise (study principal investigator Deepak K. Gupta, M.D., M.S., at the University of Vermont), a computer scientist with artificial intelligence expertise (site principal investigator Satya S. Sahoo, Ph.D., at Case Western Reserve University), and a movement disorders neurologist with clinical trials experience (site principal investigator Amie Hiller, M.D., M.C.R., at the Portland VA Medical Center), along with two neuropsychologists (Brenna Cholerton, Ph.D., at Stanford University, Abigail Ryan, Ph.D., at the University of Vermont Medical Center) and a biostatistician researcher (Curtis Tatsuoka, Ph.D., at University of Pittsburgh) serving as consultants.

Applicability and Impact: This proposal addresses the section of "Cognitive dysfunction relevant to Parkinson's disease" under the Focus Area of "Biological mechanisms or biomarkers of non-motor symptoms that could lead to the development of treatments for Parkinson’s disease" of this funding opportunity. A great majority of patients with PD are likely to benefit from this research. Specifically, cognitive impairment can occur in up to 30% of newly diagnosed patients and up to 80% of patients after 20 years of living with PD. Potential applications of this research in clinical care and clinical trials are, but not limited to: (1) new method of diagnosing cognitive impairment in PD in clinical care, clinical trials, and other clinical research; (2) detailed characterization of impairment of different cognitive functions; (3) identification of PD patients early on who are more likely than others to develop cognitive impairment after new diagnosis of PD; and (4) predict the rate of progression of cognitive impairment on a continuous scale. We anticipate that the deliverables of this project, specifically PD-CFI classification scheme for cognitive dysfunction in PD, will be ready to serve as a patient-related outcome by the end of this project.

Contribution to Advancing Parkinson's Disease Research and Care: In the short term, this research project will have two major impacts: (1) Address critical needs to develop novel approaches to study and classify the cognitive impairment in PD at levels of specific cognitive functions and on a continuous scale, and (2) Enable computer-aided diagnosis of cognitive impairment and predict its progression for an individual patient at point-of-care. In the long term, this research project will produce data to (1) support application of PD-CFI approaches to further distinguish subtypes of PD and differentiate early on from atypical parkinsonian disorders (such as Lewy body dementia), and (2) enable development of treatments by serving as a biomarker (proxy) of progression of cognitive impairment in clinical trials.