The Longitudinal Associations between Cognition, Mood and Striatal Dopaminergic Binding in Parkinson’s Disease

Abstract

Introduction:

Cognitive decline and mood symptoms are common in Parkinson’s disease (PD). Reductions in striatal dopaminergic binding have been associated with worse cognition and mood. We investigated whether this association persists throughout the disease progression in newly diagnosed PD.

Methods:

Four-year data from Parkinson Progression Markers Initiative (PPMI) were used. Groups consisted of left and right limb dominantly affected PD patients, and controls. Longitudinal relationships between cognition, mood and striatal binding ratios were assessed by repeated measures correlations.

Results:

Reduced binding was associated with general cognitive decline in controls; reduced processing speed and increased mood symptoms in PD. Anxiety was associated with striatum only in left limb dominantly affected PD. Dominantly affected limb side did not impact striatum and cognition association.

Conclusion:

There are longitudinal associations between striatum, processing speed and anxiety. Dopamine transporter availability imaging may have some prognostic value for cognition and mood in PD.

Introduction

Cognitive impairment and mood disturbances are common non-motor symptoms of Parkinson’s disease (PD), which can precede the onset of motor symptoms (Goldman & Postuma, 2014). Although non-motor symptoms in PD significantly impact quality of life, they are frequently under-recognized and undertreated (Gallagher, Lees, & Schrag, 2010). As PD progresses, severity of mood disturbances increases (Rana, Ansari, M Qureshi, & Rahman, 2018) and 80% of individuals develop dementia over the course of 20 years (Aarsland, Andersen, Larsen, & Lolk, 2003). Currently little is known about the role of dopamine in the clinical course of non-motor symptoms. Understanding how progression of dopaminergic loss contributes to changes in cognition and mood can be vital for developing prevention and management strategies for these symptoms.

Dopamine transporter availability (DaT) imaging is helpful in the differential diagnosis of Parkinson’s disease (Bajaj, Hauser, & Grachev, 2013; R. de la Fuente-Fernández & Lövblad, 2014). Early dopaminergic loss can be detected by this imaging method, and bilateral striatal dopaminergic loss has been reported in unilateral PD (Tissingh et al., 1998). The accuracy of an abnormal DaTscan is identical to that of clinical diagnosis, with a sensitivity of 98% and a specificity of 67% in early PD, and with specificity increasing up to 94% in established PD diagnoses (Raúl de la Fuente-Fernández, 2012). In clinical practice, abnormal DaTscan can be determined visually or by quantitative assessments. Commonly, binding of 2 standard deviations below healthy controls is considered abnormal (Brigo, Matinella, Erro, & Tinazzi, 2014), although quantitative values for the diagnosis or the disease stages have yet to be determined.

Studies using DaTscan support the role of dopaminergic denervation in cognitive deficits in PD. Reduced baseline caudate dopamine binding can differentiate cognitively intact PD and PD-mild cognitive impairment (Ekman et al., 2012), and predict future cognitive decline in PD (Weil, Lashley, Bras, Schrag, & Schott, 2017). Reductions in striatal binding have also been implicated in more severe depression/anxiety in de novo PD (Erro et al., 2012). However, longitudinal impact of striatal DaT remains unclear, with only one study showing that striatal DaT does not predict anxiety over two years in de novo PD (Rutten et al., 2017). Longitudinal imaging of striatal DaT offers an opportunity to characterize the contribution of striatal dopaminergic denervation to cognition and mood as the disease progresses.

Asymmetry of motor symptoms is common in PD, and is observed early in the disease (Djaldetti, Ziv, & Melamed, 2006). However, results on how asymmetry relates to cognition and mood are not consistent (see (Verreyt, Nys, Santens, & Vingerhoets, 2011) for a review). Asymmetries in substantia nigra dopaminergic degeneration and striatal dysfunction underlie motor asymmetry (Leenders et al., 1990). Therefore, predominantly affected limb side may be important while evaluating the role of right and left striatum in non-motor symptoms. In fact, whereas some studies report correlations with the left striatum for cognition or mood (Weintraub et al., 2005), others show associations with the right striatum (Ekman et al., 2012). Conflicting reports may be due to inclusion of both left and right limb dominantly affected PD patients in the same study group while looking at unilateral striatal function. For research specifically targeting the striatum, grouping across patients with opposite motor asymmetries (i.e., left versus right limb dominantly affected) may confound our ability to parse relationships between non-motor symptoms and striatal function.

In this study, we investigated the relationship between reduction of left and right striatal DaT availability, and changes in cognition and mood with disease progression over time. We employed data from the Parkinson’s Progression Markers Initiative (PPMI), which collects longitudinal data from de novo untreated PD patients. As cross-sectional approaches may limit understanding of the course of relationships between variables over time, repeated measures correlation analyses were employed to evaluate associations between DaT and cognition or mood, with all factors measured at multiple time points. Using this method, we investigated whether longitudinal striatal DaT has prognostic value for cognition and mood. We also investigated if these associations in PD differ from healthy aging.

Methods

Data were downloaded in May 2018 from the PPMI website (http://www.ppmi-info.org/). The PPMI is an international, multi-site study, approved by the Institutional Review Boards of all participating sites. Written informed consents were obtained from all participants. Details on the PPMI aims and design have been published elsewhere (Marek et al., 2011) and can be found on the PPMI website (http://www.ppmi-info.org/study-design).

We had three participant groups; PD patients with left limbs dominantly affected (PD-L) (n=154), PD patients with right limbs dominantly affected (PD-R) (n=213) and healthy controls (HC) (n=113). All participants were right-handed. Clinical assessments included Hoehn and Yahr staging (HYS) and Movement Disorders Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) Part III.

Cognitive and mood assessments

Cognitive battery consisted of the Montreal Cognitive Assessment (MoCA) for global cognition, Hopkins Verbal Learning Test-Revised (HVLT-R) for verbal learning and memory, Benton Judgment of Line Orientation (JLO) for visuospatial function, Semantic Fluency (Animal Naming) and Letter Number Sequencing (LNS) for executive function/working memory, and Symbol Digit Modalities Test (SDMT) for processing speed/attention. Mood assessments included State-Trait Anxiety Inventory for Adults (STAI) and Geriatric Depression Scale (GDS-15). To match the DaTscan time points; cognitive/mood and clinical assessment scores from baseline, years 1, 2 and 3 were used.

¹²³I-FP-CIT SPECT

Striatal binding ratios (SBRs) were downloaded fully processed from the PPMI website. To summarize information provided by PPMI, dopamine transporter imaging (DaTscan) single photon emission computerized tomography (SPECT) was performed annually. Participants were imaged 4 ± 0.5 h following the injection with a dose range of 111 to 185 MBq or 3.0 to 5.0 mCi of Ioflupane I123 injection. Raw data were acquired into a 128 x 128 matrix stepping each 3° for a total of 120 (or 4° for a total of 90) projections in a window centered on 159 ± 10% KeV. Data from individual centers were transferred to the Institute for Neurodegenerative Disorders for processing and calculation of SBRs. Iterative (HOSEM) reconstruction was done without filtering and subsequent processing was done with HOSEM reconstructed files. Attenuation correction ellipses were drawn on images and a Chang 0 attenuation correction was applied using a site specific mu. Following the attenuation correction, a standard Gaussian 3D 6.0 mm filter was applied. Files were normalized to standard Montreal Neurologic Institute space. The transaxial slice with the highest striatal uptake was determined and eight hottest striatal slices around were averaged to obtain a single slice image. Regions of interest were placed on the left and right caudate, left and right putamen as target regions; and occipital cortex as the reference region. Count densities were extracted for each region. SBRs were calculated as (target region/reference region)-1.

Statistical analysis

Statistics were performed using IBM SPSS version 23 (Armonk, NY) and SAS version 9.4 (SAS Institute Inc., Cary, NC). Variables are reported as mean (standard deviation) or percentage. Demographics and disease features were compared using the Chi-squared, Kruskal Wallis and Mann Whitney U tests. Changes in cognition and mood were analyzed with repeated measures analysis of variance using Bonferroni correction. Mixed effects model (Hamlett, Ryan, & Wolfinger, 2004) was used to estimate the longitudinal correlation coefficients between the two variables (SBRs and cognition/mood scores) with repeated observations for each participant group individually. The key component of the SAS code was the procedure, PROC MIXED, to estimate the variance-covariance matrix by assuming a completely general covariance matrix in the statistical model. If the correlation for the SBR and cognitive or mood score was found to be significant among the three groups, correlation coefficients were compared between each group using Fisher r-to-Z transformation (Lenhard & Lenhard, 2014).

As there were significantly less women and more men in the PD-R group, significantly different correlation coefficients between HC, PD-L and PD-R were further evaluated for men and women following the same longitudinal correlation coefficient estimation and correlation coefficient comparison approach. p<0.05 was considered statistically significant.

Results

Demographics and disease features

Demographics and disease features are shown in Table 1.

Table 1.

Demographics and disease features

	PD-L (n=154)	PD-R (n=213)	HC (n=113)	Group comparisons
Age	60.3 (9.86)	62.3 (9.59)	60.8 (12.2)	x2(2)=4.59, p=.10
Sex, female %	43 %	31 %	43 %	x2(2)=7.52, p=0.023*
Years of education	15.4 (3.07)	15.6 (3.10)	16.3 (3.04)	x2(2)=5.65, p=.059
Disease duration, months	6.20 (6.40)	6.60 (6.46)	n/a	U=14296, p=.30
Hoehn and Yahr stage	1.63 (0.51)	1.50 (0.50)	n/a	U=14340, p=.017*
MDS-UPDRS Part III	21.6 (8.38)	19.9 (8.77)	n/a	U=14116, p=.023*
Right caudate SBR	1.98 (0.69)	2.23 (0.70)	2.60 (0.71)	x2(2)=50.3, p<.001*
Left caudate SBR	2.08 (0.68)	2.17 (0.72)	2.63 (0.69)	x2(2)=44.0, p<.001*
Right putamen SBR	0.95 (0.58)	1.10 (0.66)	1.65 (0.74)	x2(2)=62.2, p<.001*
Left putamen SBR	0.98 (0.61)	1.03 (0.67)	1.61 (0.74)	x2(2)=52.3, p<.001*

PD-L: left side predominant Parkinson’s disease patients, PD-R: right side predominant Parkinson’s disease patients, HC: healthy controls, MDS-UPDRS: Movement Disorders Society-Unified Parkinson’s Disease Rating Scale, SBR: striatal binding ratio, n/a: not applicable. Variables are reported as mean (standard deviation) or percentage.

^*Statistically significant.

Age and years of education were similar across groups. Sex was different between groups; with more males in PD-R compared to the PD-L and HC groups. Disease duration was similar between PD groups. PD-L had more severe motor impairment than PD-R. Regarding SBR, HC had higher SBRs than PD-L (right caudate U=4029.0, left caudate U=4491.0, right putamen U=3759.0, left putamen U=4247.0; p<.001 for all) and PD-R (right caudate U=7578.0, left caudate U=6738.0, right putamen U=6411.0, left putamen U=6221.5; p<.001 for all). Right caudate and right putamen SBRs were significantly lower in PD-L compared to PD-R (right caudate U=12007.5, p=.001, right putamen U=12582.0, p=.005). Left caudate and left putamen SBRs were not different between PD groups.

Repeated measures ANOVA

Results are summarized in Table 2 and Figure 1.

Table 2.

Main and interaction effects of participant group and time for cognition and mood

	Participant group- Main effect	Significant participant group differences	Time- Main effect	Participant group and time- Interaction effect
Montreal Cognitive Assessment	F(2,283)=13.3, p<.001*	PD-L<HC PD-R<HC	F(3,849)=7.82, p<.001*	F(6,849)=1.18, p=.32
Hopkins Verbal Learning Test-Revised- Learning	F(2,283)=12.0, p<.001*	PD-L<HC PD-R<HC	F(3,849)=4.43, p=.004*	F(6,849)=0.61, p=.72
Hopkins Verbal Learning Test-Revised- Delayed recall	F(2,283)=8.74, p<.001*	PD-L< HC PD-R<HC	F(3,849)=0.49, p=.69	F(6,849)=0.59, p=.74
Benton Judgment of Line Orientation	F(2,283)=0.30, p=.74		F(3,849)=5.45, p=.001*	F(6,849)=0.88, p=.51
Semantic fluency	F(2,283)=0.86, p=.43		F(3,849)=1.11, p=.34	F(6,849)=1.16, p=.33
Letter Number Sequencing	F(2,283)=6.30, p=.61		F(3,849)=1.20, p=.31	F(6,849)=0.59, p=.74
Symbol Digit Modalities Test	F(2,283)=6.30, p=.002*	PD-L<HC PD-R<HC	F(3,849)=5.05, p=.002*	F(6,849)=1.99, p=.065
State-Trait Anxiety Inventory for Adults-State	F(2,283)=14.3, p<.001*	PD-L>HC PD-R>HC	F(3,849)=0.59, p=.62	F(6,849)=0.69, p=.66
State-Trait Anxiety Inventory for Adults-Trait	F(2,283)=9.33, p<.001*	PD-L>HC PD-R>HC	F(3,849)=1.24, p=.29	F(6,849)=0.57, p=.76
Geriatric Depression Scale	F(2,283)=3.07, p=.048*		F(3,849)=1.35, p=.26	F(6,849)=0.85, p=.53

PD-L: left side predominant Parkinson’s disease patients, PD-R: right side predominant Parkinson’s disease patients, HC: healthy controls. The number of participants in the statistical model were; PD-L: 81, PD-R:118, HC:87.

^*Statistically significant.

Figure 1.

Cognitive and mood score changes in each group

Cognition

PD groups had lower scores on MoCA, HVLT-R learning and delayed recall, and SDMT as compared to HC. There were no differences between PD-L and PD-R. JLO, semantic fluency and LNS were not significantly different between groups. MoCA, HVLT-R learning, JLO, and SDMT significantly decreased over time in all groups. Other test scores were not affected by time. There was no interaction effect of group and time.

Mood

PD groups had higher STAI-State and Trait scores as compared to HC. Although GDS scores were higher in the PD groups, differences did not reach statistical significance (p=0.055 for PD-L vs HC; p=0.059 for PD-R vs HC). There were no differences between PD-L and PD-R. There was no significant effect of time on mood scores, nor significant interaction effect of group and time.

Repeated measures correlations

Results are summarized in Table 3.

Table 3.

Correlation coefficients of significant repeated measures correlations between SBRs, and cognition/mood

	Right caudate			Left caudate			Right putamen			Left putamen
	PD-L	PD-R	HC	PD-L	PD-R	HC	PD-L	PD-R	HC	PD-L	PD-R	HC
MoCA	r=.055, p=.32	r=.061, p=.23	r=.21, p=.005*	r=.10, p=.071	r=−.004, p=.93	r=.25b, p<.001*	r=.060, p=.20	r=.090, p=.045*	r=.21, p=.004*	r=.074, p=.12	r=.012, p=.80	r=.18, p=.012*
HVLT-R learning	r=.071, p=.22	r=.036, p=.48	r=.043, p=.59	r=.088, p=.13	r=.010, p=.84	r=.078, p=.32	r<.001, p=.99	r=−.019, p=.66	r=.065, p=.38	r<.001, p=.99	r=−.048, p=.25	r=.028, p=.71
HVLT-R delayed recall	r=.010, p=.86	r=.037, p=.46	r=.12, p=.15	r=.031, p=.58	r=.009, p=.87	r=.16, p=.040*	r=−.060, p=.22	r=.004, p=.93	r=.15, p=.057	r=−.050, p=.31	r=−.038, p=.36	r=.088, p=.27
JLO	r=.023, p=.69	r=.057, p=.24	r=.078, p=.34	r=.056, p=.34	r=−.027, p=.56	r=.087, p=.29	r=.002, p=.97	r=.070, p=.11	r=.022, p=.79	r=.047, p=.34	r=−.013, p=.77	r=.018, p=.83
Semantic fluency	r=.051, p=.35	r=.10, p=.034*	r=.11, p=.18	r=.044, p=.43	r=.082, p=.083	r=.19, p=.011*	r=.043, p=.38	r=.11, p=.006*	r=.16, p=.022*	r=.045, p=.36	r=.053, p=.19	r=.17, p=.017*
LNS	r=.13, p=.020*	r=.13, p=.007*	r=.12, p=.13	r=.096, p=.091	r=.12, p=.010*	r=.20, p=.024*	r=.13, p=.005*	r=.11, p=.010*	r=.12, p=.086	r=.11, p=.020*	r=.10, p=.019*	r=.18, p=.012*
SDMT	r=.20, p<.001*	r=.22, p<.001*	r=.32, p<.001*	r=.22, p<.001*	r=.24, p<.001*	r=.40, p<.001*	r=.090, p=.064	r=.21, p<.001*	r=.35a, p<.001*	r=.087, p=.075	r=.20c, p<.001*	r=.39a,b, p<.001*
STAI-State	r=−.11, p=.039*	r=.038, p=.43	r=.054, p=.51	r=−.19b, p<.001*	r=.053, p=.27	r=.004, p=.96	r=−.033, p=.48	r=.083, p=.056	r=−.008, p=.92	r=−.13b, p=.006*	r=.079, p=.062	r=−.072, p=.34
STAI-Trait	r=−.13c, p=.029*	r=−.025, p=.63	r=.15, p=.071	r=−.20b,c, p<.001*	r=−.004, p=.93	r=.050, p=.53	r=−.066, p=.17	r=.006, p=.90	r=.040, p=.59	r=−.14, p=.004*	r=.019, p=.67	r=−.048, p=.51
GDS	r=−.024, p=.67	r=−.11, p=.018*	r=−.011, p=.86	r=−.093, p=.10	r=−.086, p=.058	r=.025, p=.70	r=.031, p=.55	r=−.078, p=.071	r=.043, p=.49	r=−.046, p=.38	r=−.065, p=.12	r=.038, p=.54

PD-L: left side predominant Parkinson’s disease patients, PD-R: right side predominant Parkinson’s disease patients, HC: healthy controls, MoCA: Montreal Cognitive Assessment, HVLT-R: Hopkins Verbal Learning Test- Revised, JLO: Benton Judgment of Line Orientation, LNS: Letter Number Sequencing, SDMT: Symbol Digit Modalities Test, STAI: State-Trait Anxiety Inventory for Adults, GDS: Geriatric Depression Scale.

^*Statistically significant.

^aSignificantly different than PD-L.

^bSignificantly different than PD-R.

^cSignificantly different than HC.

In HC, MoCA positively correlated with all SBRs. HVLT-R delayed recall positively correlated with left caudate. LNS positively correlated with left caudate and left putamen. Semantic fluency positively correlated with left caudate, right putamen, and left putamen. SDMT positively correlated with all SBRs. HVLT-R learning, JLO and mood did not correlate with any SBR.

In PD-L, LNS positively correlated with right caudate, right putamen and left putamen. SDMT positively correlated with right and left caudate. STAI-State negatively correlated with right and left caudate, and left putamen. STAI-Trait negatively correlated with right and left caudate, and left putamen. MoCA, HVLT-R learning and delayed recall, JLO, semantic fluency, and GDS did not correlate with any SBR.

In PD-R, MoCA positively correlated with right putamen. LNS positively correlated with all SBRs. Semantic fluency positively correlated with right caudate and right putamen. SDMT positively correlated with all SBRs. GDS negatively correlated with right caudate. HVLT-R learning and delayed recall, JLO, and STAI did not correlate with any SBR.

Pairwise comparisons of the above correlation coefficients by group revealed several significant differences between groups, as summarized in Table 3 (p<.05 for all). In sum, HC and PD-L differed for SDMT/bilateral putamen, and STAI-Trait/bilateral caudate correlations, and showed trend-level differences for SDMT/left caudate (p=.055) and STAI-State/left caudate (p=.051). HC and PD-R differed for MoCA/left caudate and SDMT/left putamen correlations. PD-L and PD-R differed for STAI-State/left caudate, STAI-Trait/left caudate, and STAI-State/left putamen correlations.

Sex impact on group differences for correlations

MoCA/left caudate correlation was significant for HC men (r=.28, p=.002) but not for HC women, or PD-R women or men. Correlation coefficients were not significantly different between women and men in any group.

SDMT/left putamen correlation was significant for HC men and women (r=.42, p<.001; r=.34, p<.001), PD-L men (r=.15, p=.020), PD-R men and women (r=.19, p<.001; r=.25, p<.001); but not PD-L women. Correlation coefficients were not significantly different between women and men in any group.

STAI-State/left caudate correlation was significant for PD-L women (r=−.26, p=.002) and PD-R women (r=.21, p=.022), but not PD men. Coefficients were significantly different between PD-R women and men (p=.004), but not different between PD-L women and men. Coefficients were significantly different between PD-L and PR-R women (p=.004).

STAI-State/left putamen correlation was significant for PD-L women (r=−.22, p=.002) and PD-R women (r=.25, p=.002), but not PD men. Correlation coefficients were significantly different between PD-R women and men (p=.012), but not different between PD-L women and men. Coefficients were significantly different between PD-L women and PD-R women (p=.004).

STAI-Trait/left caudate correlation was significant for only PD-L women (r=−.31, p<.001), but not for PD-R women or PD men. Correlation coefficients were significantly different between PD-L women and men (p=.041); PD-L and PD-R women (p=.008).

Discussion

In this study we performed repeated measures correlation analyses to evaluate the longitudinal associations between cognition, mood and striatal DaT in PD and HC. In healthy aging, striatal dysfunction was associated with slower processing and general cognitive decline, while striatal dysfunction in PD was related to slower processing speed and mood disturbances. These findings relate striatal dysfunction to both overlapping and different cognitive and mood changes in HC and PD. Similar cognitive performance and correlation results between left and right limb dominantly affected PD suggest dominantly affected limb side does not strongly impact cognition, at least early in the disease. On the other hand, anxiety was associated with striatal dysfunction only in PD-L, which may indicate important early differences in mood symptom profiles.

Longitudinal correlation results for healthy adults were in line with and extend previous cross-sectional studies (Cropley, Fujita, Innis, & Nathan, 2006). In our HC, striatal dopaminergic denervation related to worse global cognition, memory, executive function and processing speed; but not mood. Global cognition and processing speed were positively associated striatum bilaterally, but memory and working memory showed positive associations with left hemisphere only. The latter results are not surprising considering PPMI employs exclusively verbal tests, and both verbal working memory (Reuter-Lorenz et al., 2000) and verbal memory (Nyberg et al., 2016) have known left hemisphere lateralization. However, prior studies included visuospatial tests, and reported that spatial working memory is associated with right lateralization (Reuter-Lorenz et al., 2000), while visual memory tests positively correlated with the left caudate (Nyberg et al., 2016). Future studies using both verbal and visual tests may help determine whether the left caudate is particularly important for verbal memory during aging, or whether it relates to memory function more generally.

Looking at the correlation differences between HC and PD patients in our study, positive correlations between striatal DaT availability and cognition were not different between the groups except for processing speed. Significant positive correlations with processing speed were restricted to caudate in PD-L. In contrast, processing speed related to both caudate and putamen in the PD-R and HC groups. For the left putamen, this positive correlation between SBR and processing speed was stronger for HC compared to PD-R. Even though cognition has been typically attributed to caudate while putamen is associated with motor functions (Chung et al., 2018), both regions have been shown to contribute to cognitive and motor slowing in PD (Sawamoto et al., 2007). Our findings suggest both caudate and putamen DaT can be markers for processing speed, with perhaps a more dominant role of caudate for processing speed in PD.

Evaluating significant correlations for each PD group separately revealed a more dominant association of right striatum with cognition. For PD-R, global cognitive scores and executive function showed specific relationships with only right striatum. Although these relationships were not significant in PD-L, comparison of the correlation coefficients of PD-R and PD-L groups revealed that the correlation coefficients were not significantly different. Positive right caudate and executive functioning correlation in de novo PD have been previously shown in a study with 10 PD patients with an unknown ratio of PD-L and PD-R (Brück et al., 2001). Right striatum may have a more dominant role in cognition in PD, but this hypothesis requires further evaluation.

Cognitive deficits present early in PD, and their prevalence increases with age, PD duration and severity (Litvan et al., 2011). De novo PD patients in our study had worse scores than HC on global cognition, verbal learning and memory, and processing speed. Increased anxiety and depression scores in our PD sample are consistent with depression and anxiety often preceding the onset of PD (Goldman & Postuma, 2014). Over time, global cognition, memory, visuospatial functioning and processing speed deterioration were found in both HC and PD, suggesting changes were not disease-related but due to aging. Decline in these domains have been shown in healthy aging (Harada, Natelson Love, & Triebel, 2013), and there are reports showing that early stage PD is not related to deterioration of cognition and mood (Spalletta et al., 2014). In fact, the lack of differential decline in PD patients in our study may be due to short disease duration in our longitudinal sample.

Despite the lack of differences in the progression of cognitive and mood scores between groups, we were able to show different associations between the SBRs and changes in cognition and mood. These may point to possible compensatory mechanisms recruited by PD patients, and may support the value of DaTscan for cognitive and psychological prognosis. Previously, caudate DaT availability was shown to predict future cognitive impairment, although it did not have any predictive value for MoCA scores (Caspell-Garcia et al., 2017). The lack of association between MoCA scores and SBRs in the PD groups supports a nonlinear relationship between global cognition and striatal DaT availability; other disruptions in PD may interact with dopaminergic loss and affect cognition differently. A previous study reported associations between memory and caudate binding in 12 PD patients with a mean disease duration of 7 years (Jokinen et al., 2009). Perhaps the disease duration affects the relationship between cognition and striatal DaT, including whether it is different between PD-L and PD-R; associations may change or become more apparent over time.

Little is known about the pathophysiology of anxiety in PD, and DaTscan studies have provided inconsistent results (Erro et al., 2012; Rutten et al., 2017; Weintraub et al., 2005). We observed that anxiety had significant negative correlations with SBRs only in the PD-L group. The difference of SBR associations in PD-L compared to PD-R and HC, suggest involvement of striatal dopaminergic denervation in more severe anxiety in PD-L. Reductions in SBRs of the same regions have previously been shown in de novo PD (Erro et al., 2012). Although this previous study excluded the effects of motor lateralization, we assessed the PD groups separately and observed different associations. Nevertheless, the two PD groups were not matched for sex and disease severity; which are both risk factors for anxiety in PD. Female sex and more severe disease is associated with increased risk for anxiety in PD (Broen et al., 2018; Leentjens et al., 2011). With disease progression, multiple neurotransmitter systems are disrupted in PD, which may lead to recruitment of the dopaminergic system at a different level for anxiety (Barone, 2010). Future studies may benefit from investigating the pathophysiology of anxiety by including the effect of disease severity, rather than grouping all PD patients with various motor impairment stages into one big heterogeneous group.

As sex was not matched between the PD groups, we further investigated the impact of sex on anxiety, which revealed results in line with previous studies (Broen et al., 2018; Leentjens et al., 2011). Only women had significant SBR correlations with anxiety. With respect to state anxiety, correlation with left striatum was significantly different for PD-R women and men. Women had significant negative correlations between state anxiety and left striatum, while this relationship was not significant for men. Within women with PD, the correlations between state anxiety and left striatum differed in PD-L and PD-R. While the PD-L group had negative correlations between the state anxiety score and left striatum SBR, this was positive in the PD-R group. This may suggest that the left striatum association with state anxiety depends on the more affected hemisphere. As the left striatum was less affected in the PD-L group, state anxiety seemed to increase with more dopaminergic loss in the left. State anxiety may be a primary contributor to PD-R and PD-L differences reported in previous studies. Trait anxiety scores were related to left caudate only in PD-L women, with these women showing significantly negative correlation versus non-significant findings in other groups. Sex differences should be investigated further to clarify if neural correlates of anxiety differ between women and men with PD.

In our sample, dopaminergic denervation of striatum was not associated with depression. Depression in PD has been previously linked with dopaminergic as well as noradrenergic denervation in the limbic system, rather than striatum (Remy, Doder, Lees, Turjanski, & Brooks, 2005). Thus, one may argue that dopaminergic system plays a role in depression in PD via denervation of different regions as the disease progresses.

Our study has several strengths and limitations. One important strength is the use of a relatively new statistical method to investigate longitudinal associations between two measures. By using scores on separate cognitive tests, and anxiety and depression scales instead of a global cognitive or a mood score; we were able to determine specific associations between each one of these domains and striatal dopaminergic binding. Additionally, by classifying patients based on the more dominantly affected limb side, we believe we were able to provide more reliable findings regarding the correlations of left and right striatum. PPMI allows researchers to access longitudinal multi-site data collected from both patients and healthy adults, which include multiple aspects of PD. Although this database provides rich data, the cognitive battery is somewhat limited. A more detailed assessment of cognition and mood can provide more reliable results regarding each of the domains specifically. Additionally, due to the study’s longitudinal nature, there is also attrition over time, which can reduce the power of the statistical analyses. A common problem with HC groups should also be considered while interpreting the results. There may be a selection bias for the HC as these are participants choosing to be included in studies with a specific focus on disease, and long evaluations regarding the multiple features of the disease. Thus, the HC’s motive of participation in the study, and whether these controls remain to be healthy without carrying any risk factors for the disease has not been thoroughly investigated in our study. Finally, while interpreting our results, it is important to remember that although motor symptoms are asymmetric, the disease affects both hemispheres. Therefore, the predominantly affected limb side does not indicate the ipsilateral hemisphere is not affected. On the other hand, left striatum SBRs were similar between the PD groups, and only the right striatum SBRs differed between the groups. In addition to the SBR findings, the PD-L group had more severe disease. These findings suggest that the PD groups were not matched for dopaminergic binding status and the PD-L group had more dopaminergic loss.

In conclusion, our findings revealed that the association between SBRs and changes in cognition and mood persists with early disease progression in PD, without a strong association with motor symptom lateralization. In contrast, motor symptom lateralization may have a specific effect on anxiety, though this will require further investigation. Weak yet significant SBR relationships with cognitive domains and mood in our study require further investigation. Future studies may benefit from stratifying patients based on disease duration to evaluate different compensation strategies in PD throughout the disease progression. Combining imaging techniques to evaluate the contribution of multiple neurotransmitter systems and their interactive effects on cognition/mood outcomes in PD will also help clarify the pathophysiology more reliably, as progressive dopaminergic loss is not the sole cause of non-motor symptoms in PD.