Differentiation in theta and gamma activation in weight-shifting learning between people with parkinson’s disease of different anxiety severities

Abstract

Anxiety and postural control deficits may be related in people with Parkinson’s disease (PwPD). However, the association between anxiety levels and weight-shifting control remains ambiguous. This study investigated whether 1) weight-shifting control differed between PwPD with and without anxiety, and 2) the learning effect of weight-shifting differed between the two populations. Additionally, we evaluated cortical activities to investigate neural mechanisms underlying weight-shifting control. Twenty-eight PwPD (14 anxiety, 14 nonanxiety) participated in a 5-day weight-shifting study by coupling the bearing weight of their more-affected leg to a sinusoidal target at 0.25 Hz. We tested the weight-shifting control on day 1 (pretest), day 3 (posttest), and day 5 (retention test) with a learning session on day 3. The error and jerk of weight-shifting trajectory and the theta and gamma powers of electroencephalography in prefrontal, frontal, sensorimotor and parietal-occipital areas were measured. At the pretest, the anxiety group showed larger error and smaller jerk of weight-shifting with greater prefrontal theta, frontal gamma, and sensorimotor gamma powers than the nonanxiety group. Anxiety intensity was correlated positively with weight-shifting error and theta power but negatively with weight-shifting jerk. Reduced weight-shifting error with increased theta power after weight-shifting learning was observed in the nonanxiety group. However, the anxiety group showed decreased gamma power after weight-shifting learning without behavior change. Our findings suggest differential weight-shifting control and associated cortical activation between PwPD with and without anxiety. In addition, anxiety would deteriorate weight-shifting control and hinder weight-shifting learning benefits in PwPD, leading to less weight-shifting accuracy and correction.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11357-024-01236-7.

Introduction

Deficient postural control is a critical motor symptom in Parkinson’s disease (PD), particularly in people with PD (PwPD) who also have anxiety [1]. The ability to perform symmetrical, rhythmic weight-shifting is critical for safe and effective ambulation [2]. Ineffective postural control with asymmetry is often observed in PwPD. Due to abnormal amplitude and timing responses, PwPD typically shift their weight asymmetrically with small amplitudes, even when on medication, compared to age-matched individuals [3].

Anxiety is a common psychiatric comorbidity in PwPD, even in the early stage of the disease [4]. The frontostriatal pathways organize motor, cognitive, and emotional functions [5]. Functional magnetic resonance imaging study results have suggested that impaired dopaminergic input to the core limbic system is associated with abnormal emotional processing in PwPD [6]. The neural circuitry that connects the limbic system to the basal ganglia stimulates the central nervous system to generate motor outputs [6]. The strength of this stimulation is based on the emotional significance of a stimulus [7]. Thus, this neural connection may explain the potential relationship between postural control and anxiety in PD. For example, when PwPD are standing on a foam surface, patients with high anxiety exhibit a greater movement velocity at the center of pressure than patients with low anxiety [8]. Additionally, during quiet standing, PwPD with high anxiety. exhibit greater variability in center of gravity displacement than do PwPD with low anxiety. [1].

In previous studies demonstrating that anxiety impairs postural control, postural control was measured with static standing postural tasks. However, it remains largely unclear whether anxiety affects dynamic postural control (e.g., weight-shifting or walking). For PwPD, ability of body weight transfer was positively correlated to executive function, such as attentional control and working memory, whereas the quiet standing performance was not correlated to executive function [9]. Moreover, attentional capacity and working memory ability are deteriorated in individuals with anxiety compared to those without anxiety [10]. Previous studies revealed that anxiety affects both working memory and attention, as well as learning ability, in PwPD with high anxiety. [11, 12]. For example, PwPD with high anxiety show worse performance on the digit span test of the Wechsler Memory Scale III than those with low anxiety [11]. Moreover, PwPD with high anxiety demonstrate lower postural learning benefits compared to PwPD with low anxiety. After 10 trials of practice for keeping standing balance on a stabilometer, PwPD with a higher level of nervousness showed shorter time in keeping postural balance than PwPD with a lower level of nervousness [12].

Continuous dynamic postural control relies on attentional resources and working memory [13]. In PwPD, dysfunctions in working memory may result from impaired frontostriatal circuitry [5]. Consequently, differences in cortical activation are expected between PwPD with anxiety and PwPD without anxiety. when they are asked to control their dynamic posture accurately. This study was aimed to investigate the effect of anxiety and learning on rhythmic weight-shifting movements in PwPD and to measure their cortical activation. We focused on cortical activation in the theta and gamma bands because cortical oscillations in the two bands are pertinent to working memory [14, 15], which is intrinsically linked to attentional control and postural control in the frontal, sensorimotor and parietal-occipital areas in PwPD [16, 17]. Based on previous studies, we reasoned that theta and gamma activities would be enhanced in challenging working memory tasks because of the greater demand for attentional resources and information processing. We designed this study to determine the effect of anxiety on weight-shifting learning, and to reveal related neural mechanisms. We hypothesized that PwPD with anxiety would show greater weight-shifting errors and greater cortical activation in the theta and gamma bands than PwPD without anxiety. In addition, we hypothesized that PwPD with anxiety would have less learning benefits for weight-shifting control than PwPD without anxiety because anxiety would limit attentional control and working memory.

Methods

Participants

We recruited 14 PwPD with anxiety (mean age: 65.08 ± 6.74 years) and 14 PwPD without anxiety. (mean age: 66.15 ± 6.93 years). The inclusion criteria were (i) a diagnosis of idiopathic PD, according to the United Kingdom PD Society Brain Bank clinical diagnostic criteria [18], (ii) modified Hoehn and Yahr stages 2 − 3, (iii), a Mini Mental State Examination (MMSE) score ≥ 26 points [19], and (iv) PD onset at age > 40 years [20]. PwPD with onset of PD before the age of 40 are often categorized into young onset PD. PwPD with young onset often have different genetic and clinical characteristic from PwPD in general. Anxiety symptoms were defined based on the Hospital Anxiety and Depression Scale: Anxiety subscale (HADS-A). Participants with HADS-A scores ≥ 8 were allocated to the anxiety group, and those with HADS-A scores < 8 were allocated to the nonanxiety group [21]. Participants were excluded when they had a history of brain surgery, other diseases and conditions that could influence balance ability (e.g. neurological diseases other than PD, musculoskeletal disease of lower extremities, or cardiopulmonary diseases), or a score > 2 on item 3.13 (posture) of the Movement Disorder Society-sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS). A score > 2 on item 3.13 indicated moderate or severe stooped posture or scoliosis, which could affect control of weight-shifting. We also excluded participants who were currently taking antidepressant medication because depression symptom could influence the ability of postural control [22]. All participants had normal or corrected-to-normal vision. Table 1 presents participant demographics and clinical characteristics. All procedures in the experiment were approved by the National Taiwan University Hospital Research Ethics Committee (Clinical Trial Registration No.: NCT03895125), and all participants provided written informed consent.

Table 1.

Demographic and clinical characteristics of people with Parkinson’s disease, grouped by those with and those without anxiety (nonanxiety)

Characteristic	Anxiety group	Nonanxiety group	p value
Age (years)	65.08 ± 6.74	66.15 ± 6.93	0.68
Age range (years)	56.5 − 78.58	54.67 − 80.33	-
Sex (male/female)	5/9	9/5	0.26
Disease duration (years)	7.56 ± 4.13	6.20 ± 3.91	0.47
Modified H-Y stage	2.53 ± 0.41	2.32 ± 0.54	0.25
MDS-UPDRS part III	16.50 ± 9.05	18.14 ± 9.70	0.65
MMSE	28.71 ± 1.33	28.86 ± 1.03	0.75
HADS-A	11.64 ± 3.20	2.5 ± 1.51	< 0.001
HADS-D	3.36 ± 1.55	2.57 ± 1.28	0.16

Data are presented as the mean ± standard deviation, unless otherwise indicated. Modified H&Y stage = Modified Hoehn and Yahr stage, MMSE = Mini-Mental Status Examination; MDS-UPDRS = Movement Disorder Society Unified Parkinson’s Disease Rating Scale. HADS-A = Hospital Anxiety and Depression Scale: Anxiety subscale. HADS-D = Hospital Anxiety and Depression Scale: Depression subscale

The sample size calculation was calculated by G*power software. Based on data from a previous study [23], 26 participants (13 participants per group) would be sufficient to detect effects of anxiety on postural control (Cohen’s d = 1.34, power = 0.90, α = 0.05). Considering 10% dropout rate, a sample size of 28 or 29 participants was required.

Experimental setup and data recording

All examinations were performed with participants in the on-medication state. There were 3 sessions in this study. Session 1 (Day 1) was the pretest. Session 2 (Day 3) included weight-shifting learning and the posttest. The rest time between the weight-shifting learning session and the posttest was 10 min. Session 3 (Day 5) was presented as the retention test (Fig. 1).

Fig. 1.

Study procedure

For the weight-shifting task, participants were instructed to stand with their feet shoulder-width apart on a level surface to perform rhythmic weight-shifting movements. A force plate (9260AA6, Kistler, Switzerland) was placed under the lower limb of the more-affected side (Fig. 2A). The scores for MDS-UPDRS items 3.3 (rigidity), 3.7 (toe tapping), 3.8 (leg agility), and 3.17 (rest tremor amplitude) were summed to obtain total scores in the right and left lower extremities. The side with the higher score was defined as the more-affected side. When performing the weight-shifting task, a 22-inch monitor was positioned at eye level with a distance of 1 m in front of the participants. The monitor displayed 2 different colored circles (white and red) with 2-cm diameters. The white circle represented the weight-shifting target, which moved horizontally across the field; this represented the timing for the shift in percentage of total body weight from foot to foot. The white circle was set at a pace of 0.25 Hz (i.e., 2 s to shift 10% to 90% of body weight to the more-affected foot and 2 s to shift the body weight back again). The pace was chosen because 2 s per transition approximated slow walking in PwPD [3]. Participants were asked to match the movement of the target circle (white circle) by controlling their shift in weight, which was tracked with the red circle, to and from the more-affected lower limb (Fig. 2B). The instruction to the participants was “keeping the red circle as close as possible to the white circle during the whole testing trial by shifting your body weight”. For the pretest, posttest and retention test, each trial lasted 32 s, and 6 trials were conducted for each participant. For weight-shifting learning, the participants performed 6 blocks of weight-shifting tasks, with 6 trials per block. The interblock resting period was 2 min.

Fig. 2.

Weight-shift control test. (A) Experimental setup. The force plate measures the weight on the more-affected side. The white circle is the target for weight-shifting. The red circle tracks the weight borne on the participant’s more affected side. (B) Circles move horizontally along the percent body-weight scale

During the weight-shifting task, cortical activities were recorded with a SymAmps RT 64–channel amplifier (NeuroScan Inc., USA). Electroencephalography (EEG) electrodes were placed based on the 5–5 electrode system of the International Federation. To monitor vertical and horizontal eye movements and blinks, electrodes were placed above the left eyebrow, below the left eye, and horizontally on the outer canthi of both eyes. The impedances of all electrodes were below 5 kΩ, and they were referenced to linked electrodes placed on the mastoids of both sides. Weight-bearing data and EEG data were synchronized and digitized at a sampling rate of 1 kHz.

In addition to the weight-shifting task, we conducted functional reach tests in the forward direction and lateral directions toward the less- and more-affected sides. For forward reach test, participants were asked to stand with their feet shoulder-width apart, raised the arm of less-affected side in front of them to shoulder height, and reached forward as far as possible without taking a step or losing balance [24]. For lateral (leftward and rightward) reach tests, participants abducted their left or right arm to 90 degrees and reached leftward or rightward as far as possible, without taking a step or losing balance [25]. The functional reach test is a clinical measurement that has been used to assess the capacity of limits of stability (i.e., margin of stability) [26]. This test is also related to the measurement of movement of the center of mass and center of pressure for dynamic postural stability [27]. Reaching distances were measured with a digital laser distance meter (DM 200, Zircon Corporation, Malaysia).

Data analysis

For the weight-shifting task, the weight data were first pre-processed using a zero-phase low-pass filter (cutoff frequency: 6 Hz). The amount of error for weight-shifting was represented as the root mean square value of the mismatch between the target body weight distribution and the recorded weight distribution. In addition, we evaluated jerk (J) motions, defined as the time derivative of acceleration. Then, we calculated the squared mean of the jerk (JSM) of the weight-shifting trajectory. The JSM has been used as an index of movement smoothness during the movement time (MT) [28]. A low JSM value represents smoother movement with less movement correction [29].

$$JSM=\underset{0}{\stackrel{\mathit{MT}}{\int }}{J}^2/2dt$$

The EEG data were pre-processed with Curry 8 software (NeuroScan Inc., USA). At first, the EEG data were filtered between 1 and 50 Hz. Vertical and horizontal eye movements and blinks were detected using threshold method from the channels of VEOG and VEOH, respectively. Because the signal amplitude of eye blink is usually around 100–300 µV [30], we set 200 µV as the threshold value for vertical eye movements and blinks, and 100 µV as the threshold value for horizontal eye movements and blinks in the present study. EEG epochs were selected from 200 ms before the peak of detected eye movement or blink to 500 ms after the peak of detected eye movement or blink in all channels. The eye movement and blinks were subsequently corrected through the application of Independent Component Analysis (ICA) from the selected epochs [31], and remained eye movement/blink free EEG data were unchanged. We chose ICA for correcting ocular artifacts because it would not over- or under-correct the data [32]. In addition, an experienced researcher visually inspected each trial of conditioned EEG data to make sure the EEG data were artifact-free and were not over-corrected (e.g. The EEG signal was over-flat or had an inversed EEG wave around the time duration of the detected ocular artifact) for the whole trial. Then, principal component analysis (PCA) was performed to transform the data into a set of principal components (PCs) for the prefrontal (Fp1, Fpz, Fp2, AF3, AF4), frontal (F1, F3, F5, F7, F11, Fz, F2, F4, F6, F8, F12, FC1, FC3, FC5, FCz, FC2, FC4, FC6), sensorimotor (C1, C3, C5, Cz, C2, C4, C6, CP1, CP3, CPz, CP2, CP4, CP6), and parietal-occipital (P1, P3, P5, P7, Pz, P2, P4, P6, P8, PO3, PO7, POz, PO4, PO8, O1, Oz, O2) areas, since these brain areas are critical for attentional control and dynamic postural control in PwPD [16, 17, 33]. In each cortical area, the relative significance of the corresponding PCs was ranked according to their eigenvalues (λ₁, λ₂,…, λ_n), which explained the variance of the whole set of EEG data. The eigenvectors (U₁, U₂,…, U_n) acted as weighting coefficients; these were used to calculate the relative contribution of each PC to the EEG for all electrodes. The loading factors (W₁, W₂,…, W_n) of the PCs in each electrode were calculated as follows:

$$Wi={(Ui}^2/{\sum }_{i=1}^n{\mathit{Ui}}^2)\times 100\%$$

In this formula, n indicates the number of channels in the specific areas. In the present study, only the major PC (PC1) was analyzed further because it accounted for more than 70% of the total variance in the EEG data for each cortical area (Supplementary Table 1) [34]. PC1 was analyzed to determine its relative power in theta (4–8 Hz) and gamma (30–50 Hz) bands at the frontal, prefrontal, sensorimotor, and parietal-occipital areas. The relative powers of the theta and gamma bands were determined by dividing the absolute power in each spectral band by the total power of all spectral bands. All parameters of weight-shifting performance (error and JSM) and EEG data (theta power and gamma power) were averaged from 6 testing trials.

Statistical analysis

Based on the Shapiro–Wilk test, all parameters in the present study fitted normal distribution. The effects of anxiety (independent variables: anxiety and nonanxiety) and learning (independent variables: pretest, posttest, retention test) on the dependent variables of weight-shifting error, weight-shifting JSM, and distances in the functional reach test were examined with two-way mixed analysis of variance (ANOVA). When necessary, a post hoc test with least significant difference (LSD) comparisons was performed. On the other hand, to emphasize the anxiety effect on weight-shifting control, we further used Pearson correlation to examine the correlation of HADS-A and weight-shifting performance, functional reach tests, and relative power of EEG based on the pretest data of all participants. The Bonferroni correction would be used for multiple correlation estimations. The level of significance was set at p < 0.05. Signal processing and statistical analyses were completed using MATLAB v. R2012a (MathWorks, MA, USA) and SPSS v. 21 (SPSS Inc., USA). All data are presented as the mean ± standard error.

Results

Behavior performance

There was no missing data and no data was discarded due to technical issues for behavior data in the present study. Therefore, the behavior data of the weight-shifting task in all 6 testing trials were included in the analysis. The mean values of weight-shifting error and weight-shifting JSM of the two groups are depicted in Fig. 3A and B. The ANOVA results revealed that the learning main effect, group main effect and the interaction effect between learning and group affected the size of the tracking error (learning: F_(2,52) = 5.989, p = 0.005, η_p² = 0.187; group: F_(1,26) = 9.490, p = 0.005, η_p² = 0.267; interaction: F_(2,52) = 3.445, p = 0.039, η_p² = 0.117). Post hoc analysis showed that weight-shifting errors were smaller in the nonanxiety group than in the anxiety group at the pretest (p_LSD = 0.037, d = 0.588), posttest (p_LSD = 0.007, d = 0.776), and retention test (p_LSD = 0.001, d = 0.955). In addition, the nonanxiety group exhibited smaller weight-shifting errors at the posttest (p_LSD = 0.004, d = 0.842) and at the retention test (p_LSD < 0.001, d = 1.064) than at the pretest, whereas weight-shifting learning did not affect weight-shifting errors in the anxiety group. Weight-shifting JSM was also affected by the main effects of learning (F_(2,52) = 4.550, p = 0.015, η_p² = 0.153) and group (F_(1,26) = 7.110, p = 0.013, η_p² = 0.226) and their interaction effect (F_(2,52) = 5.099, p = 0.010, η_p² = 0.148). Post hoc analysis showed that the nonanxiety group had a larger weight-shifting JSM at the posttest than at the pretest (p_LSD = 0.002, d = 0.898) and the retention test (p_LSD = 0.003, d = 0.872). However, although it seems to have an increased trend of weight-shifting JSM across the three testing time points (retention test vs. pretest, p = 0.064), weight-shifting learning did not significantly affect the weight-shifting JSM in the anxiety group. The weight-shifting JSM was larger in the nonanxiety group than in the anxiety group at the pretest (p_LSD = 0.034, d = 0.598) and the posttest (p_LSD < 0.001, d = 1.252) but not at the retention test (p_LSD = 0.514).

For functional reach tests, there was no significant interaction, learning or group effect on the forward reach distance (group: F_(1,26) = 2.560, p = 0.112; learning: F_(2,52) = 2.197, p = 0.121; interaction: F_(2,52) = 0.515, p = 0.601) or the lateral reach distance toward the less-affected side (group: F_(1,26) = 1.213, p = 0.281; learning: F_(2,52) = 1.479, p = 0.237; interaction: F_(2,52) = 2.943, p = 0.060) (Fig. 4A, B). The lateral reach toward the more-affected side was affected by the main effects of learning (F_(2,52) = 12.493, p = 0.020, η_p² = 0.299) and group (F_(1,26) = 13.426, p = 0.043, η_p² = 0.418) and varied with the interaction effect (F_(2,52) = 12.634, p = 0.003, η_p² = 0.207) (Fig. 4C). After weight-shifting learning (posttest), the nonanxiety group improved their reaching distance to the more-affected side compared to the pretest (p_LSD < 0.001, d = 1.176), and the improvement lasted until the retention test (retention test > pretest, p_LSD = 0.005, d = 0.824). In contrast, although it seemed to have an increasing trend of reach distance across the three testing time points (retention test vs. pretest, p_LSD = 0.071), the reaching distance toward the more-affected side was not significantly affected by weight-shifting learning in the anxiety group. In addition, the nonanxiety group had a longer lateral reach distance toward the more-affected side than the anxiety group at the pretest (p_LSD = 0.009, d = 0.766), posttest (p_LSD < 0.001, d = 1.179) and retention test (p_LSD = 0.002, d = 0.969).

Fig. 4.

Distances of functional reach in the (A) forward direction and lateral direction toward the (B) less-affected side and the (C) more-affected side. Data are presented as the mean ± standard error. **p_LSD < 0.01; ***p_LSD < 0.001

Brain activity

There was no missing data and no data was discarded due to technical issues for EEG data in the present study. Therefore, the EEG data of the weight-shifting task in all 6 testing trials were included in the analysis. Table 2A summarizes the values of the relative power of the theta band in the prefrontal, frontal, sensorimotor, and parietal-occipital areas. Only the values of prefrontal theta power were significantly affected by the main effects of learning (F_(2,52) = 0.920, p = 0.005, η_p² = 0.222), group (F_(1,26) = 1.940, p < 0.001, η_p² = 0.489), and their interaction (F_(2,52) = 3.160, p = 0.031, η_p² = 0.108). The nonanxiety group had a lower prefrontal theta power than the anxiety group at the pretest (p_LSD = 0.020, d = 0.660). After weight-shifting learning, the prefrontal theta power increased in the nonanxiety group (p_LSD = 0.025, d = 0.592), and the learning effect lasted to the retention test (p_LSD = 0.021, d = 0.698). However, the values of prefrontal theta power among the three testing time points were not significantly different in the anxiety group (Fig. 5A).

Table 2.

Comparison of the relative powers in (A) theta and (B) gamma bands in the prefrontal, frontal, sensorimotor, and parietal-occipital brain areas measured in the anxiety group and nonanxiety group at the pretest, posttest and retention test

Data are presented as the mean ± standard error

Fig. 5.

Comparison of the relative powers measured in people with PD and anxiety and in people with PD without anxiety (nonanxiety group) in (A) prefrontal theta power, (B) frontal gamma power, and (C) sensorimotor gamma power. Data are presented as the mean ± standard error. *p_LSD < 0.05; **p_LSD < 0.01; ***p_LSD < 0.001

Table 2B summarizes the values of the relative power of the gamma band in the prefrontal, frontal, sensorimotor, and parietal-occipital areas. The frontal gamma power was significantly affected by the main effects of learning (F_(2,52) = 2.096, p = 0.021, η_p² = 0.065) and group (F_(1,26) = 5.337, p = 0.030, η_p² = 0.029) and varied with the interaction effect (F_(2,52) = 3.262, p = 0.030, η_p² = 0.028). Post hoc analysis showed that the anxiety group had a greater frontal gamma power than the nonanxiety group at the pretest (p_LSD = 0.008, d = 0.767) (Fig. 5B). In addition, learning modulation was observed only in the anxiety group. The frontal gamma power decreased after weight-shifting learning in the anxiety group (posttest < pretest, p_LSD = 0.014, d = 0.807), but the frontal gamma power increased during the retention test (retention test > posttest, p_LSD = 0.043, d = 0.673). Similar to that in the frontal area, the gamma power in the sensorimotor area was also significantly affected by the main effects of learning (F_(2,52) = 4.143, p = 0.022, η_p² = 0.145), group (F_(1,26) = 3.834, p < 0.001, η_p² = 0.155), and their interaction (F_(2,52) = 4.953, p = 0.011, η_p² = 0.174). Post hoc analysis showed that the anxiety group had a greater sensorimotor gamma power than the nonanxiety group at the pretest (p_LSD = 0.016, d = 0.700) and posttest (p_LSD = 0.036, d = 0.576) (Fig. 5C). Learning modulation was observed only in the anxiety group with less sensorimotor gamma power at the posttest (p_LSD < 0.001, d = 1.086) and retention test (p_LSD = 0.013, d = 1.060) relative to the pretest.

Correlation between HADS-A and behavior performance or brain activity

Table 3A and B present the correlation results between HADS-A and weight-shifting performance, distances of the functional reach tests, and relative power in the theta and gamma bands. For the correlation between HADS-A and weight-shifting performance, because there were two parameters of weight-shifting performance (error and JSM), the significance of Pearson correlation was p = 0.025 using the Bonferroni correction. Interestingly, the HADS-A score was positively correlated with the weight-shifting error (r = 0.437, p = 0.020) but negatively correlated with the weight-shifting JSM (r =  − 0.550, p = 0.002). For the correlation between anxiety level and functional reach distances, because there were three reaching directions (forward, more-affected side, and less-affected side), the significance of Pearson correlation was p = 0.017 using Bonferroni correction. Only the distance to the more-affected side showed a negative correlation with the HADS-A score (r =  − 0.590, p = 0.001). However, there was no significant correlation between the HADS-A score and the distances of forward reach or lateral reach to the less-affected side. For the correlation between HADS-A and EEG relative powers, because there were two bands of EEG power (theta and gamma), the significance of Pearson correlation was p = 0.025 using the Bonferroni correction. The association between anxiety level and EEG relative power was only shown in the theta band in the prefrontal area, with a positive correlation between the HADS-A score and the theta band in the prefrontal area (r = 0.474, p = 0.014).

Table 3.

Correlations between Hospital Anxiety and Depression Scale: Anxiety subscale (HADS-A) scores and (A) behavioral measures and (B) neural measures in people with Parkinson’s disease

Discussion

First, the present study demonstrated that weight-shifting control was affected by anxiety in PwPD. Before weight-shifting learning, larger errors and a smaller JSM of weight-shifting movements were observed in PwPD with anxiety than in PwPD without anxiety. In addition, PwPD with anxiety had greater theta power in the prefrontal area and greater gamma power in the frontal and sensorimotor areas. Second, the learning effect was different between PD with anxiety and PD without anxiety. Weight-shifting learning led to changes in weight-shifting errors and JSM values in the nonanxiety group but not in the anxiety group. With weight-shifting learning, the theta power was modulated in the nonanxiety group, whereas the gamma power was modulated in the anxiety group. This study is the first to reveal the effects of anxiety on performance and learning weight-shifting control and an underlying neural mechanism in PwPD.

The effect of anxiety on weight-shifting control before learning

Although it has been shown that PwPD with anxiety (or high anxiety) have worse postural control than PwPD without anxiety (or low anxiety) in a static standing posture [1, 8], the present study specifically focused on dynamic movement (i.e., weight-shifting movement), which is more crucial for ambulation in PD [35, 36]. We found that when performing a rhythmic weight-shifting movement, PwPD with anxiety had less tracking accuracy and jerk than those without anxiety. Previous studies in healthy adults established that elevated anxiety levels lead to less postural stability and impaired behavioral flexibility [37, 38]. According to attentional control theory, anxiety would hinder working memory and degrade motor control processing [39]. The theory proposes that individuals with anxiety require more effort to avoid shortcomings in performance accuracy than individuals without anxiety, because of reduced attentional resource of working memory and decreased efficiency of information processing. However, PwPD also have to devote more attentional resources to postural control due to malfunctioning basal ganglia [40]. Thus, the attentional resource should be more limited in PD with anxiety than PD without anxiety. This reduced ability of dynamic postural control due to anxiety could be evidenced by the negative correlation between the weight-shifting JSM value and the HADS-A score and the positive correlation between the weight-shifting errors and HADS-A score (Table 3A). A smaller JSM value indicates that the movement is performed with less correction [29]. Our findings implied that among PwPD, high anxiety impaired the ability to correct movement trajectory, which resulted in greater weight-shifting errors.

For the reaching distance tests, only the lateral reaching distance toward the more-affected side was affected by anxiety, with greater anxiety linked to shorter reaching distances (Table 3A). A shorter reaching distance in the anxiety group implies that PwPD with anxiety put less weight-bearing on the more-affected side than PwPD without anxiety. This may be because anxiety would deteriorate body awareness of movement further [41]. Although a previous study showed that the ambulation performance (e.g. walking speed, cadence, and stride length) in early stage PD were positively correlated to limits of stability distance in both forward and lateral directions [36], the amplitude of medial–lateral postural sway could be a better predictor of future falling risk than the amplitude of anterior–posterior sway in older individuals with anxiety or fear of falling [42]. Our findings of decreased lateral reaching distance and unvaried forward reaching distance were similar to those of Takahashi et al. [27]. They found that the lateral reach distance was shorter in older adults with depression than in those without depression; however, the reaching distance in the forward direction was not affected by depression. In the present study, depression levels were similar in the two PD groups; thus, the association between postural control and neuropsychiatric symptoms was extended beyond depression and involved anxiety issues, particularly toward the more-affected side in PD.

The learning effect on weight-shifting control between anxiety and nonanxiety groups

In addition to between-group differences in the weight-shifting performance and reaching distance at baseline, the present study further indicated that the changes in weight-shifting performance and reaching distance after weight-shifting learning were different between PwPD with anxiety and PwPD without anxiety. The nonanxiety group increased the JSM value at posttest, indicating that they performed more postural correction immediately after weight-shifting learning [29]. Integrating the results of weight-shifting error and JSM (Fig. 3A, B), our findings implied that the decreased tracking error in the nonanxiety group might be due to intensive correction for weight-shifting movement after learning. Also, the distance of lateral reaching to the more-affected side increased after weight-shifting learning (Fig. 4C). PD is one of disease populations with motor asymmetry for dynamic postural control [17]. Previous studies reported that PwPD have axial kinaesthetic deficits and they tend to rely less on their affected limb when performing postural balance tasks [43, 44]. During the learning trials, the participants had to track the target by adjusting their weight bearing on more-affected leg, which could make participants devote more consciousness to control the weight-shifting movement toward more-affected side, particularly in PwPD without anxiety. This finding is similar to the result of Sheikh et al.’s study in stroke population [45]. With weight-shift training, the stroke patients increased weight bearing on the affected side and improved gait symmetry. Notably, the JSM value decreased in the retention test relative to the posttest in the nonanxiety group, implying that PwPD without anxiety could track weight-shifting targets in a more efficient way with less conscious control (e.g., more automatic control) after acquiring the weight-shifting skill [46]. However, for PwPD with anxiety, neither the weight-shifting performance nor the reaching distance improved after learning.

Fig. 3.

Functions compared between the anxiety and nonanxiety groups of people with PD at the pretest, posttest and retention test. (A) Weight-shifting error, (B) weight-shifting JSM. Data are presented as the mean ± standard error. *p_LSD < 0.05; **p_LSD < 0.01; ***p_LSD < 0.001

To fine tune postural control (such as weight-shifting movement), PwPD rely on more conscious control (or more attentional resources) for planning postural strategies than unaffected adults [47]. Enhanced anxiety may impair attentional control [39], leading to inefficient processing of motor learning. Usually, people with anxiety have a lower ability perception (lower self-efficacy of their motor ability) [48], which strongly impacts their balance performance and learning capacity [49]. Greater ability-related concerns and nervousness regarding a motor task would deteriorate balance performance and learning. Thus, our observation of ineffective learning for weight-shifting movement in the anxiety group may be because of their poor attentional control and ability perception.

Brain activity modulation between the anxiety and nonanxiety groups

The present study showed that the anxiety group had a greater theta power in the prefrontal area than the nonanxiety group at the pretest. Moreover, the anxiety score was positively correlated with the theta power in the prefrontal area. Recent studies have suggested that theta oscillation reflects the demands of working memory and sustaining attention [50, 51]. Greater theta power in the prefrontal area has been observed during high-demand cognitive tasks or high-challenging postural tasks in healthy adults [52]. Indeed, studies have shown that theta powers in the prefrontal cortex and hippocampus are proportional to the level of information processing [51, 53]. High anxiety may increase the conscious load during movement control due to insufficient attention resources and working memory [39]. For example, high-anxiety individuals show greater frontal theta power than low-anxiety individuals while they perform a decision-making task [54]. Hence, the higher theta power in the anxiety group at pretest may imply deficient working memory or overload conscious control for weight-shifting movement and lead to larger weight-shifting errors and less weight-shifting corrections (i.e., smaller weight-shifting JSM). On the other hand, although excessive theta power might indicate an overload of attentional engagement, theta activity could be beneficial for the consolidation of motor learning [55]. A positive association between theta power and motor learning was found in a complex motor learning task of a sequential visual isometric pinch task in young healthy adults [56]. After one day of weight-shifting learning, the theta power in the prefrontal area increased in the nonanxiety group (Fig. 5A), indicating that the nonanxious PwPD may still have learning capability by memory consolidation and motor plan updates [56]. However, theta modulation was not observed in the anxious group. It was likely that anxiety had already taxed the majority of the limited central capacity [57], which left no room for theta power modulation with a postural learning effect.

In addition to theta power, we also observed greater gamma power in the frontal and sensorimotor areas in the anxiety group than in the nonanxiety group. In EEG studies, frontal gamma power values increased when people were under time pressure or anxiety-inducing conditions [58, 59]. Gamma power in the frontal regions is associated with the demand of targeting executive control and is linked to heightened attentive states [60]. On the other hand, greater gamma power in the sensorimotor area and basal ganglia is associated with reduced motor and cognitive flexibility [61, 62]. Additionally, gamma oscillations in the sensorimotor cortex could be strongly linked to negative emotional processing and related to motor functioning in the context of attentional reorienting [63, 64]. The greater gamma power in PwPD with anxiety may indicate that these patients experienced greater mental pressure when controlling weight-shifting trajectories than PwPD without anxiety. Although there was no significant behavior improvement in the anxiety group after weight-shifting learning, the decreased gamma power in the frontal and sensorimotor areas might imply a potential enhancement in motor-cognitive flexibility and attentional control. A further study is needed to explore whether a longer learning period will lead to behavior improvement in PwPD with anxiety.

Methodological concerns and study limitations

First, we measured the HADS-A score to differentiate the anxiety group and nonanxiety group. However, the level of anxiety was not measured when the participants performed the weight-shifting task. Second, the participants shifted their body weight at a pace of 0.25 Hz. Faster weight-shifting may also be necessary in daily life, for example, when dodging a pedestrian or a car when crossing a road. Third, we focus on short-term learning effect in the present study. Future research should quantify the level of anxiety (e.g., by measuring skin conductance responses) while performing the weight-shifting task, and the weight-shifting should be tested at different velocities.

Conclusion

The present study demonstrated that high anxiety levels were detrimental to weight-shifting control in PwPD. Additionally, there were fewer learning benefits for weight-shifting control in PwPD with anxiety than in PwPD without anxiety. Our findings highlighted neural mechanisms involved in weight-shifting behavior and linked them to postural control in association with anxiety level in PwPD. Psychiatric comorbidity should be considered in clinical practice, particularly in postural control training for PwPD.

Supplementary Information

Below is the link to the electronic supplementary material.

Authors’ contribution

Conceptualization: R.M.W. and C.Y.H. Methodology: Y.T.H. and C.Y.H. Formal analysis: Y.T.H. and C.Y.H. Writing-original draft preparation: Y.T.H. and C.Y.H. Writing-review and editing: Y.T.H., R.M.W. and C.Y.H. Funding acquisition: C.Y.H.

Funding

This work was supported by a grant from the National Science and Technology Council, Taiwan (grant no. NSTC 112–2314-B-002–136-MY3).

Data availability

The data that support the findings of this study are available on request from the corresponding author.

Declarations

Ethics approval

All procedures performed in this study were in accordance with the ethical standards of the Helsinki declaration and approved by the National Taiwan University Hospital Research Ethics Committee.

Consent for publication

Not applicable.

Consent to participate

Written informed consent was obtained from all individual participants included in the study.

Conflicts of interest

The authors declare no competing interests.