Use of robust norming to create a sensitive cognitive summary score inde novoParkinson’s disease

Abstract

ABSTRACTBackground and ObjectivesCognitive impairment is common at all stages in Parkinson’s disease (PD). However, the field is hampered by consensus over which neuropsychological tests to use and how to utilize the results generated by a cognitive battery. An option that combines the richness of a neuropsychological battery with the simplicity of a single test score is a cognitive summary score (CSS). The objective was to determine if a CSS created using robust norming is sensitive in detecting early cognitive deficits inde novo, untreated PD.MethodsUsing baseline cognitive data from PD participants and healthy controls (HCs) in the Parkinson’s Progression Markers Initiative, these steps were taken: (1) creating a robust HC subgroup that did not demonstrate cognitive decline over time; (2) using the robust HC subgroup to create regression-based internally-derived standardized scores (z-scores) for six cognitive scores across five tests; and (3) creating a CSS by averaging all standardized test z-scores.ResultsPD participants scored worse than HCs on all cognitive tests, with a larger effect size (PD versus HCs) when the comparison group was the robust HC subgroup compared with all HCs. Applying internally-derived norms rather than published norms, the largest cognitive domain effect sizes (PD vs. robust HCs) were for processing speed/working memory (Cohen’s d= -0.55) and verbal episodic memory (Cohen’s d= -0.48 and -0.52). In addition, using robust norming shifted PD performance from the middle of the average range (CSS z-score= -0.01) closer to low average (CSS z-score= -0.40), with the CSS having a larger effect size (PD vs. robust HC subgroup; Cohen’s d= -0.60) compared with all individual cognitive tests.DiscussionPD patients perform worse cognitively than HC at disease diagnosis on multiple cognitive domains, particularly information processing speed and verbal memory. Using robust norming increases effect sizes and lowers the scores of PD patients to “expected” levels. The CSS performed better than all individual cognitive tests. A CSS developed using a robust norming process may be sensitive to cognitive changes in the earliest stages of PD and have utility as an outcome measure in clinical research, including clinical trials.