In this retrospective study, 119 patients (165 unique hemispheres) were recruited from an IRB approved movement disorders database (UF INFORM). Subjects selected were diagnosed with PD and implanted with the Medtronic Percept PC neurostimulators attached to unilateral or bilateral DBS leads in either the GPi (96 subjects, 135 hemispheres) or the STN (23 patients, 30 hemispheres). A total of 88 males and 29 females were included in this study, with an average age at surgery of 67.8 ± 9.0 (mean ± std) years and an average disease duration of 13.8 ± 6.5 (mean ± std) years. Table 1 summarizes the subjects characteristics for the cohort.
In each subject, neural activity was recorded chronically in participants’ home environment while under both therapeutic chronic medication and stimulation conditions. Therefore, for each subject and each hemisphere, we selected the first chronic neural recording, following DBS implantation, lasting at least 5-consecutive-day worth of data and out stimulation settings through the recordings. The average recording duration was 17.8 ± 19.7 (mean ± std) consecutive days commencing at 13 ± 35 days after pulse generator device implantation, which happens 4 weeks after DBS lead implantation at our center. The bipolar contact pair used for sensing was determined by the contact(s) used for continuous stimulation therapy. The frequency of oscillatory activity sensed chronically was selected for each subject by the clinician during DBS programming sessions. The clinicians typically selected the most prominent peak on the power spectral density generated after a short survey recording. The neural frequency tracked ranged from 5.86 Hz to 67.38 Hz across this cohort. For analysis, frequencies were grouped in bands defined as theta/alpha ( < = 12 Hz), low-beta (> 12 and < = 20 Hz), high-beta (> 20 and < = 30 Hz), and gamma ( > = 31 Hz). In GPi subjects, the theta/alpha band and low-beta band were most commonly reordered (46.7% and 37.0% respectively), while in STN the low-beta band was most commonly recorded (63.3%). In some participants, the gamma band was tracked either due to the presence of a physiology peak or documentation of dyskinetic symptoms.
In addition to sensing configuration, stimulation settings were also extracted for the analysis since most patients were receiving continuous high-frequency stimulation therapy during the chronic recordings. Therapy settings varied across subjects, with stimulation pulse width ranging from 60 to 90 µsec, the stimulation frequency ranging from 110 Hz to 180 Hz, and stimulation amplitude ranging from 0 mA to 4.5 mA.
Beta power was usually reduced at night but could also be increased, especially in GPi
Figure 1A shows chronic GPi high beta (24.41 Hz ± 2.50 Hz) activity recorded in an individual subject over 5 consecutive days. Power fluctuations were observed with a consistent reduction during night-time. The reduction manifested around midnight. Vertical lines indicate the time of the events marked, and ranged from 7am to midnight (EDT), corresponding to a period of higher beta power during the daytime period. The polar plots illustrate the circadian rhythms, showing that beta power was higher during the day (when the events were marked). Interestingly, an increased power at night was also observed in other patients. Figure 1B shows an example of beta (19.53Hz ± 2.50Hz) in a different patient with GPi recordings, consistently increased at night and decreased during the day when the subject marked events. The normalized power collected in each GPi (N = 135) plotted over 24h and sorted by highest power in daytime, is shown on the circadian heatmap (Fig. 1C). Visual inspection shows the presence of a strong circadian rhythm in most recordings. We found a decreased power at night in 57.8% of the recordings (78 hemispheres) while an increased power was observed in 28.9% (39 hemispheres). In 18 hemispheres (13.3%), there was no statistically significant difference between day and night (see Methods).
As expected, STN power was more consistently reduced at night compared to daytime when compared to the GPi (76.7%, 23 hemispheres), with only 5 hemispheres (16.7%) showing increased power at night-time. Two hemispheres (6.7%) exhibited no statistically significant difference between day and night-time. Figure 2 shows individual examples of STN beta power which was reduced at night (A), as well as STN beta power increased at night (B). The STN power fluctuation of each recording (n = 30) is shown on the circadian heatmap (Fig. 2C).
Basal ganglia circadian rhythm depends on the frequency band recorded.
Given the variability in circadian rhythm (increase vs decrease power at night), especially in GPi, we investigate the effects of sensing frequency. For each recording, the change in power between day and night was calculated and plotted against the center of the sensing frequency (Fig. 3A/D).
In GPi, non-parametric tests within each canonical spectral band (theta/alpha, low-beta, high-beta, and gamma) indicated that changes in neural signals between day and night were significantly different from 0 (no-change) in both high-beta (median = 1.29, p = 0.044, FDR corrected) and gamma (median = 1.43, p = 0.013, FDR corrected). The positive median values indicate that GPi power in high-beta and gamma bands were more likely to decrease at night. Alpha and Low-Beta power were either decreased or increased during night-time, as indicated by a median non-significantly different from 0 (alpha, p = 0.576 and Low-Beta p = 0.291, FDR corrected). A one-way ANOVA to compare across spectral bands revealed a significant effect of sensing frequency in GPi (F = 6.42, p < 0.001). Post-hoc analysis with Tukey’s HSD showed that the circadian rhythm in the theta/alpha band showed more increased in night-time than high-beta (q=-0.81, p = 0.023) and gamma (q=-1.31, p = 0.002), and low-beta showed more increased in night-time than gamma (q=-1.07, p = 0.019) (Fig. 3B).
In STN, non-parametric tests showed that low-beta (median = 1.29, p = 0.033, FDR corrected) and high-beta (median = 1.29, p = 0.022, FDR corrected) were significantly different from 0. The positive median revealed that STN power in low- or high-beta bands was more likely to decrease at night, while theta/alpha power was either increased or decreased. A one-way ANOVA to compare across spectral bands indicated a significant effect of sensing frequency (F = 4.11, p = 0.014). Post-hoc analysis with Tukey’s HSD showed that the circadian rhythm in high-beta decreased more in night-time than theta/alpha (q=-1.71, p = 0.013) (Fig. 3E).
These results suggest that different frequency bands of brain activity exhibited circadian rhythms of different magnitudes and directions. In both targets, high-beta power was more consistently reduced at night, while theta/alpha power was either increased or reduced at night. However, low-beta band, especially in the GPi, was more likely to increase at night (32.6%) than in STN (11.1%).
Extended-release medications modulate circadian rhythm.
Medications and stimulation are known to modulate basal ganglia activity in PD, especially the beta band. Therefore, a general linear model (GLM) was built to test the effects of stimulation parameters and medications on the direction of beta power (13-30Hz) circadian rhythms in GPi and STN. In particular, we investigated whether the change in beta power (dependent variable) during daytime versus night-time was significantly modulated by the following independent variables: the levodopa equivalent daily dose (LEDD), the use of medications (dopaminergic extended-release, benzodiazepines, and melatoninergic medications), the total energy delivered (TEED) (a measure of the amount of stimulation based on amplitude, pulse width, and frequency), the contact used for stimulation (dorsal/ventral), and the hemisphere recorded (left/right).
Table 2 summarizes the results of the GLM. The model indicated a significant effect of extended-release dopaminergic medications on beta power circadian rhythm (z = 2.46, p = 0.016), while all other variables (stimulation parameters and other medications) did not significantly affect the change in beta power during daytime versus night-time. In GPi, the group of subjects using extended-release dopaminergic medication showed a statistically significant greater decrease in power at night compared those who were not taking extended-release medication (W = 112, p < 0.001) (Fig. 3C). Interestingly, the GLM built for STN recordings did not reveal any independent variable that contributed significantly to the direction of circadian rhythm in STN. The use of extended-release dopaminergic medication did not influence the direction of circadian rhythm in STN (z = 0.61, p = 0.540).
Although the TEED did not significantly affect beta power circadian rhythms at the group level, we observed that stimulation parameters could affect circadian rhythms within individual subjects. In particular, larger stimulation amplitude might reduce the range of neural signal fluctuations as shown in Figure S1A. However, the influence of stimulation varies considerably across patients.