24 Functional Distinctions in Metabolic Network Patterns in Parkinson's Disease

Abstract

Objective:

Cognitive decline is a common non-motor feature of Parkinson's disease (PD). However, the underlying mechanisms of cognitive impairment in PD require further elucidation. FDG PET imaging data analyses have revealed distinct brain metabolic patterns associated with the cognitive features of PD. The PD cognition-related pattern (PDCP) and default mode network (DMN) are two overlapping, but topographically distinct, networks that may serve as biomarkers of cognitive decline in PD. Decreased activity of the resting-state DMN and increased expression of the PDCP are associated with cognitive impairment in PD. Studies have consistently demonstrated the association between neuropsychological memory test performance and PDCP expression. Thus, we examined whether memory performance could offer additional value in predicting PDCP expression in PD patients. We hypothesized that DMN and memory performance would predict greater variance in PDCP expression than the DMN alone.

Participants and Methods:

Participants included 48 PD patients ages 46-80 (mean (SD) Age: 61.9 (8.1), Education: 15.0 (2.8), IQ: 112.5 (14.9), DRS total: 136.7 (5.8)). All participants completed the FDG PET and neuropsychological evaluation 8-12 hours after their last dose of Levodopa. Neuropsychological memory testing included the California Verbal Learning Test (CVLT) z score of sum of learning trials. PDCP and DMN values were z scores generated from normal controls in previous studies. Data were analyzed using linear regression analyses.

Results:

A hierarchical regression was performed to predict PDCP as a function of DMN and CVLT learning performance. Variables were entered in two separate blocks. The first block included DMN as a predictor, and the overall regression was significant (R2 = 0.55, F(1, 39)= 47.0, p < 0.001). As hypothesized, DMN significantly predicted PCDP expression (β= -0.74, p < 0.001). The second block of the regression included CVLT learning memory performance. Both DMN and CVLT performance explained a significant amount of variance in PDCP (R2 change = 0.05, F(2, 39)= 27.6, p < 0.001). CVLT performance significantly predicted PDCP (β= -0.22, p =0.048). The final model accounted for 60.0% of the variance inPDCP.

Conclusions:

Disruptions in functional connectivity within brain networks have become increasingly recognized as mechanisms responsible for cognitive impairment in patients. As demonstrated in previous studies, our results indicated that DMN loss is a strong predictor of PDCP expression, likely due to the networks' overlapping spatial regions. However, we found that the addition of memory performance to the model could explain a small amount of variance (5%) over and above DMN expression. Overall, the current findings demonstrate a functional (i.e., learning) distinction between population-specific (PDCP) and more general brain networks (DMN).