Healthy controls and PD subjects did not significantly differ in terms of age or strength of right handedness; nevertheless, controls displayed slightly higher educational levels [F(1, 80) = 5.03, p < 0.05; table ]. PD patients were nondemented based on their MMSE scores and did not display clinically significant symptoms of anxiety or depression. However, as expected, PD patients scored lower on the MMSE than controls [F(1, 80) = 11.69, p < 0.001] and endorsed more symptoms of anxiety [F(1, 80) = 42.34, p < 0.001] and depression [F(1, 80) = 30.97, p < 0.001; see table for means and SD]. PD patients displayed higher UPDRS total scores (t = −11.12, p < 0.001) and UPDRS motor scores (t = −10.66, p < 0.001) in the off state compared to the on state, verifying adequate medication washout. However, they did not differ in sleep or self-reported measures of anxiety or depression between medication states (see table for means and SD). Since education differences were identified between the PD and healthy control groups, education was investigated as a covariate in all models below.
The multivariate model for attention/executive subtests was significant for between-groups differences [F(10, 71) = 2.97, p < 0.005]; however, education did not significantly adjust the variance in dependent measures and thus was not utilized as a covariate. All attention/executive subtests were statistically significant between the groups, demonstrating lower z-scores for PD subjects compared to controls, with the exception of the WAIS-IV Digit Span Forward (DSF) and the Stroop Color and Color-Word Tests (see table for univariate values). A comparison of the effect sizes revealed the greatest significance for the D-KEFS Tower Total Completion Time, followed by the WAIS-IV LNS and Stroop Word scores. However, an evaluation of the z-scores revealed a very subtle decline in the attention/executive measures for early-stage PD patients of less than 1 SD below the mean of healthy controls. A medication effect was only evident for the Digit Span Total [DST; F(1, 80) = 5.01, p = 0.012, η2 = 0.10] and the DSF [F(1, 80) = 6.56, p = 0.028, η2 = 0.08]. PD patients displayed significantly greater impairment on the DST (DSToff = −0.588; DSTon = −0.356) and the DSF (DSFoff = −0.322; DSFon = −0.101) in the off medication state, although they did not differ from controls on the DSF and barely reached significance for the DST.
The multivariate model for language subtests was significant for between-groups differences [F(6, 75) = 2.94, p = 0.012] but not for within-subjects medication effects. All language subtests were significant between-groups, with the exception of D-KEFS Letter Fluency (table ). PD patients displayed lower scores on all language subtests; however, the RBANS Picture Naming subtest displayed the greatest level of impairment and the largest effect size, with PD patients falling −1.45 SD below the means of normal controls (table ). Conversely, across fluency measures, PD patients were only −0.27 to −0.55 SD below normal controls. When education was utilized as a covariate, as it significantly adjusted for language functions at the multivariate level [F(6, 74) = 3.26, p = 0.007], the D-KEFS Fluency measures were no longer significant. The RBANS Picture Naming scores remained highly significant (p < 0.001), while the RBANS Semantic Fluency scores only approached significance (p = 0.057).
The multivariate model revealed a between-groups significance across memory subtests [F(6, 75) = 4.52, p < 0.001], but within-subjects effects for medication as well as education as a covariate were not significant. PD subjects scored lower than normal controls in all learning and memory measures on the RBANS, although effect sizes were the largest for Story Delayed Recall, List Recognition, and List Recall (table ). Impairment indices based on z-scores revealed that early-stage PD patients scored from −0.76 to −1.41 SD below the mean of normal controls.
The multivariate model revealed between-groups significance across visuomotor/visuospatial subtests [F(8, 73) = 2.89, p = 0.007], but there was no within-subjects significance for medication. PD patients scored significantly below normal controls on all subtests, with the exception of RBANS Line Orientation. Effect sizes were greatest for D-KEFS Trails 5 (Motor), D-KEFS Trails 2 (Number), and RBANS Coding. However, impairment indices were greatest for D-KEFS Trails 3 and 4 (Letter and Number-Letter alternation), displaying the most significant deviation relative to controls (−1.79 and −1.8) across the entire neuropsychological battery. Although education significantly adjusted for the variance in the model when utilized as a covariate [F(8, 72) = 2.63, p = 0.014], the significance of the subtests remained unchanged.
Based on effect sizes and impairment indices, the following neuropsychological measures from each cognitive domain were selected as stepwise predictors for a multiple regression model to identify the best predictors of cognitive impairment in early-stage PD patients relative to normal controls: LNS, D-KEFS Tower Completion Time, Stroop Word, RBANS Naming, RBANS Memory (List Learning, List Recall, Story Recall, and Figure Recall), D-KEFS Trails 2, 4, and 5, and RBANS Coding. After controlling for age and education [accounting for only 6.9% of the variance; F(2, 79) = 2.91, p = 0.061], Trails 5 [Motor; F(3, 78) = 9.31, p < 0.0001] and RBANS List Recall [F(3, 78) = 10.06, p < 0.0001] were the best predictors of early-stage PD cognitive impairment, accounting for an additional 19.5 and 7.9% of the variance, respectively. In a second regression model that eliminated visuomotor tasks (i.e., Trails), RBANS List Recall [18.6% of the variance; F(3, 78) = 8.89, p < 0.0001], LNS [an additional 8.5%; F(4, 77) = 9.86, p < 0.0001], and Figure Recall [an additional 3.9%; F(5, 76) = 9.24, p < 0.0001] were the best predictors and explained a combined 37.8% of the variance in terms of cognitive impairment in early-stage PD (table ).
Predictors of group (PD vs. controls) and neuropsychological performance for PD patients in the off medication state based on stepwise multiple regression analyses
To expand on the above regression analyses, the performance of each significant predictor was utilized in separate regression models to determine disease-related predictors of cognitive performance in early-stage PD. The predictors entered into the equation for each model (after forcing age and education into the first step of the regression) included daily levodopa dosage, disease duration, and either on or off scores for the UPDRS Motor, Parkinson's Disease Questionnaire 39 (PDQ-39), Epworth Sleepiness Scale (EP), Parkinson's Disease Sleep Scale (PDSS), Beck Anxiety Inventory (BAI), and Beck Depression Inventory II (BDI-II). Cognitive performance was modeled both for on and off medication states, with the corresponding appropriate variables in either the on or off medication state utilized as predictors. The first set of regression analyses for the off medication state is presented in table . Age and education significantly accounted for 27.1% of the variance [F(2, 37) = 8.06, p < 0.001] for RBANS List Recall in the off state, while the PDQ-39 in the off medication state explained an additional 12% of the variance [F(3, 36) = 8.76, p < 0.001]. Following age and education [accounting for 24.4% of the variance; F(2, 37) = 5.79, p = 0.007], RBANS Figure Recall in the off state was best predicted by disease duration, explaining an additional 11.5% of the variance [F(3, 36) = 6.52, p = 0.001]. WAIS-IV LNS performance in the off state was predicted by age and education [25.3% of the variance; F(2, 37) = 6.08, p = 0.005], followed by the PDQ-39 in the off state, explaining an additional 8.5% of the variance [F(3, 36) = 5.95, p = 0.002]. Finally, 30.2% of the variance in the D-KEFS Trails 5 performance in the off state was explained by age and education [F(2, 37) = 7.79, p = 0.002], with the BAI explaining an additional 9.8% of the variance [F(3, 36) = 7.76, p = 0.001].
The same dependent cognitive measures were evaluated in the on medication state, but the disease predictors were not significant with the exception of D-KEFS Trails 5, which measures motor speed. After accounting for the variance from age and education [32% of the variance; F(2, 37) = 8.69, p = 0.001], the UPDRS motor score in the on state accounted for an additional 9.6% of the variance in the D-KEFS Trails 5 performance in the on state [F(3, 36) = 8.56, p = 0.001].