The effect of Levodopa and Stimulation on post-surgery Freezing of Gait in STN-DBS Parkinson's Disease patients: a clinical and kinematic analysis

doi:10.21203/rs.3.rs-4058858/v1

Download PDF

Article

The effect of Levodopa and Stimulation on post-surgery Freezing of Gait in STN-DBS Parkinson's Disease patients: a clinical and kinematic analysis

https://doi.org/10.21203/rs.3.rs-4058858/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Despite the long-term efficacy of high-frequency (HFS) subthalamic nucleus deep brain stimulation (STN-DBS) on appendicular symptoms, its benefit on freezing of gait (FOG) is less clear. Mechanisms and optimal therapeutic approaches to this type of FOG remain unclear.

Objective

Assess acute post-surgery FOG response to levodopa and stimulation

Methods

17 PD STN-DBS patients with a FOG score (item 3.11) ≥ 2 in the MedON/StimON condition were evaluated under 5 experimental conditions, including a low frequency (60Hz) condition maintaining the same total energy delivered. In each condition, gait and FOG episodes (#FOG) were assessed using clinical (including a 3x14 meters Stand-Walk-Sit task) and kinematic metrics using a set of Inertial Measurement Units (IMUs).

Results

At a cohort level, compared to MedOFF/StimOFF, #FOG was significantly reduced in the MedONStimON 130Hz condition. A high variability in individual responses were seen regarding individual responses to LD or stimulation. While ~ 29% of patients worsened their FOG with LD and were rescued by DBS, ~ 18% presented the reverse pattern. No significant differences were observed in #FOG when low and high frequency were compared, however MDS-UPDRS axial subscores were significantly lower in 60Hz condition. Gait variability emerged as the strongest kinematic dimension associated with FOG. A convolutional neural network model trained to identify FOG episodes on sensor data from an independent cohort of PD presented a good correlation with clinical FOG metrics (r > 0.54).

Discussion

FOG presenting in the Best-Functional state after surgery is mostly a therapy-resistant FOG partially improved by stimulation and medication. The clinical and kinematic heterogeneity in FOG responses to LD and stimulation (including frequency) should be clinically considered. IMU based tools can provide powerful methods to identify FOG episodes, study gait phenotypes and clarify the circuit mechanisms of FOG, whose treatment remains an unmet clinical need.

Health sciences/Neurology/Neurological disorders/Parkinson's disease

Health sciences/Signs and symptoms/Neurological manifestations

Freezing of gait (FOG) is characterized by a sudden and transient reduction in the forward progression of the feet despite the intention to walk^1–5. This prevalent and debilitating motor symptom in Parkinson's disease (PD) significantly impairs the quality of life, elevates fall risk, and contributes to increased disability. ⁶

While subthalamic nucleus deep brain stimulation (STN-DBS) has demonstrated noteworthy benefits in improving appendicular symptoms, its impact on axial symptoms, including FOG, appears to diminish over the long-term follow-up. ^7–10 Regarding FOG, despite an initial benefit observed in the first years post-surgery, with stimulation effectively mimicking the preoperative effects of dopaminergic medication, this positive effect progressively diminishes. ^7–14 Consequently, FOG emerges in the best functional state (Medication ON and Stimulation ON), despite the pre-surgery good response to dopaminergic medication.^14–16

The etiology of FOG emerging in the best functional state after surgery remains unclear. It has been hypothesized that this phenomenon reflects disease progression with affection of nondopaminergic pathways, leading to a loss of levodopa (LD) and stimulation responses.^8,17 Alternatively, suboptimal targeting^18–20, stimulation-induced effects^15,21, structural brain lesions due to electrode placement, or inadequate stimulation settings or LD dose, have also been proposed as mechanisms for post-surgery best functional state FOG.⁸

Other stimulation paradigms, including Low-frequency stimulation (LFS), have been suggested as a potential strategy to ameliorate FOG^15,21–25, but the use of patient-reported questionnaires limit the proper assessment of FOG. ^26–28 Given the paroxysmal nature and the different phenotypes of FOG, subjective metrics may not be sufficient to capture all phenomenological complexity and particularly differential modulation by therapy.^29–31 To address these challenges, integrating unbiased assessment tools (kinematic analysis and wearable sensors) for gait analysis becomes crucial.²⁹

In the current study we used clinical and inertial-sensor based tools to study FOG and FOG response to stimulation and medication, in advanced PD patients submitted to STN-DBS. We focused our study on patients that present FOG in the best functional state (Medication ON/Stimulation ON), a major clinical problem in DBS clinics. We expanded our main objective by performing multiple assessments to assess individual responses of FOG to medication, high-frequency (130 Hz) and energy matched low-frequency (60 Hz) stimulation and developing a machine-learning powered classifier of FOG that allowed us to identify kinematic variables related to FOG across conditions, potentially informative in future studies.

A total of 17 PD patients were included in the study: 71% male, disease onset at 51.9 ± 5.9 years and age at bilateral STN-DBS 62.2 ± 5.9 years. Patients were evaluated 7.7 ± 4.6 years after surgery, at a time where mean age and disease duration were 69.9 ± 6.2 and 17.9 ± 5.1 years, respectively. (Table 1). FOG emerged on average 24.1 ± 2.11 months after surgery. Before surgery, none of the patients presented severe gait or FOG in their Best ON state (scores 0.6 ± 0.6 of item 3.10 and 0.4 ± 0.4 of item 3.11, during pre-DBS LCT). Using the same dose of levodopa, the MDS-UPDRS III improvement upon the LCT was significantly lower when compared with the baseline assessment (20.1 ± 13.4% vs 56.1 ± 14.0%, p = 0.001). A similar result was found for the axial sub-score of MDS-UPDRS part III (19 ± 37% vs 78 ± 21% vs, p < 0.001). Total MDS-UPDRS part III score, in both OFF and ON state, were significantly worse compared with pre-surgery evaluation. (Fig. 1, Table 1). Stimulation parameters at the time of the study are shown in Supplementary Table S1 and, in patients whose reconstruction was possible, no electrode was found to be misplaced (Supplementary Figure S1).

Table 1

Clinical and demographic characteristics
Sex: male	12 (70%)
Age disease onset, yrs	51.9 ± 5.9
Age at surgery, yrs	62.2 ± 5.9
Disease duration at surgery, yrs	11.59 ± 3.1
Time-to-FoG (months)	24.1± 21.1
Age at evaluation (yrs)	24.1± 21.1
Disease duration at study inclusion (yrs)	17.9 ± 5.1
DBS duration at study inclusion (yrs)	7.7 ± 4.6
MDS UPDRS OFF pre-op*	51.3 ± 14.3
MDS UPDRS ON pre-op*	23.5 ± 11.3
Item 3.10 OFF pre-op*	1.4 ± 1.5
Item 3.11 OFF pre-op**	2.0 ± 2.0
Item 3.10 ON pre-op*	0.6 ± 0.6
Item 3.11 ON pre-op*	0.4 ± 0.4
LEDD pre-op (mg)	1298.8 ± 453.1
LEDD at study inclusion (mg)	838 ± 453.1
LD dose LCT (mg) (pre-op)	433.1 ± 137.5
LD dose LCT (mg) (at study inclusion)	433.1 ± 137.5
Data are expressed as percentage or mean and standard deviation as appropriate. Abbrev: Yrs, years; FoG, freezing of gait; DBS, Deep brain stimulation; MDS UPDRS, Movement Disorder Society Unified Parkinson’s Disease Rating Scale; LEDD, Levodopa Equivalent Daily dose; LCT, levodopa challenge test; * n = 16 n = 9 *

1 – FOG assessment in Med OFF/Stim OFF vs Med ON/Stim ON

The number of FOG episodes was significantly reduced in the MedON/StimON condition when compared with the MedOFF/StimOFF (6.3 ± 7.9 vs 17.4 ± 15.7, p < 0.001, Fig. 1A) paired with the reduction in the SWS time (76.5 ± 54.0 vs 239.5 ± 198.0 p = 0.004, Fig. 1B). A similar finding was observed for the total MDS-UPDRS score (62.0 ± 9.9 vs 38.6 ± 8.5, p = 0.003), axial sub-score (9.8 ± 3.8: 7.4 ± 4.4, p = 0.0247), FOG (3.4 ± 0.8 vs 2.0 ± 1.4, p = 0.007) and postural instability (1.9 ± 1.5 vs 0.6 ± 1.1, p = 0.011) scores. (Fig. 1, Supplementary Table S2, Supplementary Table S3)

2 - Assessing individual effect of stimulation and LD on FOG

2.1 Improvement at the #FOG episodes with both stimulation and medication at a cohort level

We found that either stimulation or LD (Med OFF/Stim ON or Med ON/ Stim OFF) significantly reduced the total MDS-UPDRS part III score (Med OFF/Stim OFF: 62.0 ± 9.9 ; Med OFF/Stim ON 45.91 ± 8.6; Med ON/Stim OFF: 49.353 ± 10.451, p < 0.0001, Fig. 2A). This was paired with a significant reduction in the #FOG episodes with any strategy (Med OFF/Stim OFF: 17.395 ± 15.710; Med OFF/Stim ON: 11.235 ± 13.022; Med ON/Stim OFF: 9.279 ± 7.506, p = 0.0164, Fig. 2B) in line with the observations of SWS time (Fig. 2C). In fact, a strong correlation between the number of FOG episodes and SWS time was observed across all condition (Fig. 2D).

A high heterogeneity of responses on #FOG episodes was observed with either stimulation (maximum worsening of 287% and maximum improvement of 100%) or medication (minimum worsening of -600% and improvement of 94.12%). However, an average improvement of 75.5 ± 16.7% was obtainable with at least one therapeutic intervention (MED-only or STIM-only) (Fig. 2E). The magnitude of this improvement was not related to the severity of FOG (#FOG episodes) in the Med OFF/Stim OFF condition, excluding a floor or ceiling effect (r=-0.11, p = 0.67, Fig. 2F).

Medication or stimulation also significantly improved non-axial MDS-UPDRS subscores (Fig. 2H), whilst reduction in axial subscores only achieved statistical significance on the MedOFF/StimON condition (Fig. 2G). Abnormal involuntary movements were significantly increase with Levodopa, but not with Stimulation (Fig. 2I). Detailed description of clinical metrics can be found in Supplementary Table S2 and S4.

Although there was always at least one condition where FOG improved (Figs. 2B and 2E), an increase in the number of FOG episodes under stimulation was observed in 3 (17.7%) patients whose freezing improved with medication whilst 5 patients (29.4%) worsened their freezing with Levodopa but were rescued by stimulation (Fig. 2J-I). These observations motivated a better description of these groups.

2.2 Exploring the differential responses of FOG to LD and Stimulation at an individual level

In the five patients presenting a LD-resistant FOG a significant effect of LD on MDS-UPDRS non-axial score was seen (Fig. 3A) excluding ineffective intake as a cause. The #FOG episodes was significantly lower in the MedOFF/StimON that in the MedON/StimOFF condition (1.8 ± 2.4 vs 15.8 ± 6.8, p = 0.003) supporting that there was the possibility to recover FOG in these patients (Fig. 3A, left). Interestingly, when these patients were evaluated in the MedON/StimON condition, the addition of medication to the effective stimulation led to a trend in the increase of #FOG episodes (Supplementary Figure S2A, Left) reinforcing a possibly deleterious effect of LD in FOG in a subgroup of patients. A similar pattern is observed on MDS-UPDRS III axial score, with stimulation significantly reversing the worsening induced by LD (MedOFF/StimOFF: 7.8 ± 4.8; MedON/StimOFF: 9.6 ± 3.9; MedOFF/StimON : 6.2 ± 4.4, p = 0.0278) (Fig. 3A, Supplementary Figure S2C, left). The improvement on non-axial MDS-UPDRS III seen with stimulation or medication (Fig. 3A, mid-left) was improved even more in the ON/ON condition (Supplementary Figure S2B, Left). This supports a pathophysiological dissociation between FOG and other motor symptoms.

On the STIM-resistant subgroup, stimulation induced an increase on the #FOG episodes, when compared to medication (MedOFF/SimOFF: 13.61 ± 6.73; MedON/StimOFF: 4.55 ± 2.86; MedOFF/StimON : 26.00 ± 11.53, p = 0.0278, Fig. 3B, left), supporting again that FOG rescue was possible. In this group, although the comparisons are limited by the low patient number, both LD and stimulation, resulted on a non-significant reduction on non-axial subscore (MedOFF/StimOFF: 55.0 ± 6.1; ON/OFF: 44.00 ± 10.82; OFF/ON: 42.67 ± 8.50, p = 0.194 (Fig. 3B), but also a trend for an addictive effect of both therapies in this specific outcome is observed (Supplementary Figure S2B, middle). Adding the effects of LD to stimulation completely rescued the lack of FOG response (Supplementary Figure S2B, middle).

In the largest group that presented a FOG response with both strategies, this was broadly seen across all motor domains (Fig. 3C).

In both these sub-groups, the different responses patterns were not related to the severity of motor symptoms in the baseline (OFF/OFF) state (Fig. 3D). Detailed description of these groups can be found in Supplementary Tables S5, S6 and S7.

3 - The role of HFS and LFS on FOG

As previously described, combined stimulation and LD resulted in a significant decrease in the number of FOG episodes. Globally, the number of FOG episodes (HFS: 6.32 ± 7.91; LFS: 5.08 ± 8.26, p = 0.410) and SWS time (HFS: 76.52 ± 54.03; LFS: 71.84 ± 54.23, p = 0.836) were not significantly different between the HFS and LFS conditions (Fig. 4A and 4B). On the contrary, the effects on MDS-UPDRS non-axial and axial sub-scores, were distinct. While HFS was associated to significantly better non-axial scores (HFS: 31.18 ± 7.27; LFS: 35.12 ± 8.75, p = 0.004, Fig. 4C), axial scores were significantly better with LFS (HFS: 7.41 ± 4.36; LFS: 5.59 ± 2.57, p = 0.004, Fig. 4D) without any significant change in the AIMS score (Fig. 4E).

Variability in patient response to different stimulation frequencies was observed (Fig. 4F): in 58.8% of the patients, LFS was superior to HFS in reducing the number of FOG episodes (Fig. 4G), with a decrease in axial MDS-UPDRS sub-score (Fig. 4I) and a trend for an increase in non-axial MDS-UPDRS score (Fig. 4J). By contrast, 25.5% of patients had a better FOG response to HFS condition compared to LFS (Fig. 4F-J). Detailed description of these groups can be found in Supplementary Table S2 and S3.

4- Using IMU-based approaches to study gait and FOG

4.1 Identification of FOG and kinematic features related with FOG

Using a Convolutional neural network (CNN) in an independent group of 21 PD patients, a model for automatic quantification of the percentage of FOG during straight gait (% FOG) was developed. This detection was not based on specific gait variables, but on the analysis of the inertial sensor signal (Fig. 5A-D, Supplementary Methods) and had a good specificity and sensitivity for freezing detection compared to clinical assessment (Fig. 5-E).

The % FOG significantly decreased from Med OFF/Stim OFF to Med ON/Stim ON condition (61.51% ± 9.99 vs 24.43% ± 24.43, p = 0.0128) (Fig. 5F left). The number of FOG episodes and the SWS time significantly correlated with the %FOG detected by IMUs across conditions (Fig. 5A right, details on correlations can be found in Supplementary Table S8). There was a significant correlation between the %FOG and individual kinematic gait metrics across conditions (Fig. 5B, details on correlations can be found in Supplementary Table S9) with metrics related to gait variability having the strongest correlations with FOG. As power was very limited to perform comprehensible analysis, dimension reduction was performed using a Principal Component Analysis (PCA) with the set of 30 pre-specified kinematic variables. The first Principal Component (PC1) was able to explain 75% of the gait variability (Fig. 5H), it was enriched with dimensions of gait variability and asymmetry (Fig. 5I) and was correlated with multiple FOG metrics across conditions (Fig. 5J-K, Supplementary Table S10). Another argument suggesting that these dimensions capture relevant FOG-related kinematics emerge from comparison across frequency conditions. In the LFS-better group (n = 10), values of PC1 were lower with 60 Hz (low freezing) compared with 130 Hz (high freezing), whereas in the HFS-better group (n = 4), the reverse trend was observed (Supplementary Figure S3 C-D).

4.2 Using 3D kinematics do describe gait alterations

Supplementary Tables S3, S4, S11 on this manuscript provide a detailed description of the selected kinematic variables across the 5 study conditions. In the current study we found that LFS, when compared to HFS, was related to a global reduction of gait and axial symptoms as assessed by the MDS-UPDRS (Fig. 4D) with no difference in FOG (Fig. 4A). An exploratory analysis found that gait of patients on LFS tended to be faster with a significantly higher step cadence (p = 0.032), but no significant differences on step length (p = 0.640) or width (p = 0.330) (Supplementary Figure S4 left). LFS was also characterized by a lower stance time (p = 0.0219), and double support time variability (p = 0.022) translating into gait cycles with lower variability (Supplementary Figure S4 Right). This could not be explained by a difference in the number of FOG episodes (Fig. 4A).

The emergence and management of severe FOG in the best functional state after surgery remains a challenge in the long-term follow-up of STN-DBS patients. In this study, we aimed to understand the impact of stimulation and LD on post-surgery FOG, in patients previously responsive to LD. We have found that FOG presenting in the Best-Functional state after surgery is mostly a therapy-resistant FOG, and that stimulation and medication are still able to partially improve the severe gait alterations observed in the treatment-OFF condition. Importantly, a strong heterogeneity in FOG responses to specific treatments is observed. Despite the lack of benefit in the number of FOG episodes with LFS, axial scores were significantly improved, suggesting a role on LFS in the management of gait.

Individual subjects have distinct FOG responses to Stimulation or Levodopa

In our study, we identified that best functional state FOG is mostly a treatment-resistant FOG and not treatment-induced FOG. Consequently, severe FOG and associated gait impairments observed in the off-treatment condition are, to a certain extent,rescued by therapy in all patients.

We found that either LD or stimulation improved OFF/OFF FOG, and that a synergistic effect was observed when they were combined. This suggests that patients presenting FOG in MedON/StimON state 7 years after DBS still benefit from the effects of stimulation and LD, even if with a lower magnitude of response than before surgery: Additionally, these findings do not suggest, at least at a group level, a deleterious effect of stimulation. and do not suggest, at least at a group level, a deleterious effect of stimulation. This observation aligns with prior research findings, indicating that while FOG may persists in the MedON/StimON condition, both its severity and frequency decrease in comparison to the OFF treatment-state.^{10,15,32–34}

We hypothesized that this attenuated response to treatment is related to disease progression, with a decline in LD sensitivity reflecting the extension of the neurodegenerative process to non-dopaminergic pathways.^8,35 The lower magnitude of motor response in the post-op LCT compared to pre-op LCT, particularly regarding axial symptoms, as well the increased severity of motor scores in the treatment OFF states, supports this argument. This is in line with previous studies in STN-DBS patients, where a progressive decline in axial symptoms, including freezing was observed in the medium to long-term follow-ups. ^7,9,10 However, our study does not inform us whether patients would have equivalent benefit compared to pre-DBS were much higher doses of LD used for the pos-DBS LCT.

Different effects of STN-DBS and LD at motor outcomes^36–39 have been described and, more recently, STN-DBS and Levodopa were shown to modulate brain motor circuits differently.⁴⁰ Here, we expand these observations by proposing that symptom-specific improvements must be accounted when addressing disease mechanisms. Our observations may have practical clinical implications: currently, patients with LD-resistant FOG and gait impairment are currently excluded from surgical programs^41,42 even if some isolated finding have suggested that axial non-LD responsive signs may improve with STN-DBS ⁴³. Our quasi-experimental design may complement these findings. We describe a subgroup of ~ 30% of patients in our cohort whose FOG was not improved with medication but was rescued with stimulation. In fact, in these patients the number of FOG episodes nearly doubled while on medication, even if bradykinesia and rigidity improved. Although we can’t exclude that a low dose of LD was used in the LCT, the pattern of response suggests a lack of effect of LD and an effect of STN-DBS specifically in axial signs. This may suggest that exclusion of DBS surgery based solely on the response to LD may limit their accessibility to an effective treatment.

Interestingly, the dissociation between therapy-specific responses is observed mostly for axial signs, suggesting that FOG and axial symptoms may behave differently than the rest of PD motor symptoms.

The exact pathophysiologic mechanisms involved in FOG are not well understood, with multiple theoretical models⁵ and clinical triggers identified⁴. By systematically exploring the response of the same patient in the same self-paced scenario across multiple therapeutic manipulations, we are limited to extract conclusions regarding freezing in other contexts but we benefit from constraining contextual variability. This allow us to argue that it is very likely that therapeutic responses of FOG can be obtained by manipulating different nodes of circuit dysfunction (with stimulation or medication). ^44,45

Stimulation Frequency can lead to important differences in FOG response to stimulation

LFS has been appointed as a possible strategy to manage refractory-FOG in STN-DBS patients.^{15,22,23,46,47} These studies have different designs and outcome measures, but from most of them the clinical variability in FOG response is clear, even if they focus on younger and less severe patients than our sample. In our study, we found no difference between HFS and LFS, even if LFS was associated with a significant reduction in axial MDS-UPDRS III scores. The benefits of LFS in axial scores, including improvements in symptoms such as dysarthria and dysphagia, have also been previously demonstrated. .^22,48 Although variability, unblinding and small sample limited comparisons, nearly 60% of patients had less FOG episodes with LFS at equivalent delivered energy levels than HFS. In this subgroup of patients, when LFS and HFS were compared, a significant reduction in MDS-UPDRS III axial scores with a trend for an increase in non-axial was also seen. The effect of LFS on axial symptoms was also captured by the kinematic analysis. Here, variables reflecting mostly the overall gait bradykinesia (as speed or step length) were not specifically modulated by LFS (Suppl Fig. S4), whilst variables known to be associated with axial signs^49–52, such as gait cycle variability metrics, were reduced by 60 Hz stimulation, in accordance with a previous observation with 80 Hz stimulation.⁵³ Previous studies have shown that the optimal contacts for 60 Hz stimulation were situated more ventrally than the ones used for 130 Hz stimulation.^47,54 Here, one can envisage that whilst HFS of the descending nigropontine projections and outflow tracts of the pedunculopontine nucleus (PPN) can negatively impact gait, LFS stimulation of the same structure will have a positive impact, as suggested by the beneficial effects on gait observed with PPN-DBS at low frequency stimulation (25 Hz).^54–56

We used the same contacts for both HFS and LFS, suggesting that even using the same contact location, different FOG responses can be obtained. It would be crucial to characterize the distinct circuit engagement related to symptom improvement in order to refine stimulation targeting and refinement.

Nonetheless it appears to be clear that axial signs are specifically prone to modulation by LFS, suggesting that axial and appendicular symptoms may emerge from distinct brain circuits involved in motor control.

Our results reinforce the clinical evidence that different patients will benefit from different therapeutic strategies and that an individualized and systematic approach exploring patients under both medication and stimulation conditions may help in individualizing the role of stimulation and levodopa resistance in the etiology of FOG (Fig. 6). This heterogeneity might be a source of bias in clinical trials assessing stimulation parameters strategies and should be considered when calculating sample size.

Automatic FOG detection correlates well with clinical FOG metrics

FOG detection and quantification have traditionally been carried out using clinical scales, often relying on retrospective self-assessment.^29,57,58 Even when evaluated by a trained clinician, a significant inter-rater variability concerning both the number and duration of FOG episodes exists.⁵⁹ To find more robust and objective metrics, wearable devices, such as IMUs, have been increasingly used for FOG quantification and detection.²⁹ Additionally, machine-learning algorithms have been increasingly used to provide an automated detection of FOG.^60,61 We created a model for FOG detection using the continuous IMU data from previously assessed PD patients⁶², which was fed into a convolutional neural network (CNN), an approach that has previously proved its capacity in capturing complex movement and accurately detecting FOG.⁶³ The results where then applied to the current sample of DBS patients. Our model for automatic FOG detection has shown to capture well the clinical phenomenon, with a good correlation with clinical detection of FOG episodes across all experimental conditions, which comes in line with previous works showing good correlation between automatic-FOG detection methods and clinical metrics.⁶¹ In addition, the automatic detection of the percentage of FOG in gait significantly decreased from Med OFF / Stim OFF condition to Med ON / Stim ON condition, suggesting it is sensitive to treatment modulation. In future, replication of our results in different populations and bigger sample is needed, as well as the assessment of its ability to detect subclinical FOG.

A discussion about the true episodic nature of FOG and the artificial split between episodic and continuous gait disorders, is ongoing.⁵ Objective gait analysis, especially when performed in ambulatory settings throughout long time-periods, can provide useful insights regarding the overall gait structure, clarifying how the presence of continuous alterations on the background gait may predispose to the emergence of a clinical episode of FOG. PD patients with FOG present higher gait variability and asymmetry than PD patients without FOG, even outside FOG episodes.^64–67 An increase in gait variability has been linked to an elevated risk of negative gait outcomes, including risk of falls,^49,68,69 suggesting that increased gait variability discloses a significant disruption of normal gait patterns. Here one can picture that the presence of this severe background and continuous gait deficits will progressively accumulate until a certain threshold is achieved and a FOG episode occurs – threshold model. ⁵

In our work, the gait metrics who were most strongly associated with both clinical or automatic FOG metrics were gait variability, asymmetry, and non-linear metrics, with this holding true across all tested conditions This was also evident with PC1, that being mostly composed by variability and asymmetry metrics, showed a significant decrease with medication and stimulation closely followed the behavior of clinical observed freezing events. This may argue that the existence of severe gait alterations will be associated to a decrease on the freezing threshold, and “normalization” of this background gait alterations could potentially translate into a lower likelihood of FOG.

Our kinematic gait analysis was made including freezing episodes, that may potentially contribute to the estimation of kinematic variables. Importantly, in the LFS/HFS scenario, the number of FOG episodes is similar, and kinematic patterns distinct, suggesting that even in this scenario kinematics can captures non-FOG changes. To properly ascertain the background gait in PD freezers a comparison with an age and disease duration-matched non-freezer group should be performed. Our results favors the idea that cumulative and severe gait alterations will culminate on a clinical FOG event, that will only represent the “tip of the iceberg” of profound and subclinical gait disturbances. Recently, stride time variability was used as a biomarker to modulate adaptative DBS, with positive effects in FOG.⁷⁰ Enriching this “kinematic biomarker” with comprehensive set of gait variables known to capture FOG could provide an attractive treatment strategy.

Several limitations should be considered when interpreting the findings of this study. Firstly, the small sample size restricts the generalizability of the results and may diminish statistical power to detect significant effects. Additionally, the absence of randomization and blinding in the evaluation process introduces potential biases into the study design. The LD dose administered during the LCT was determined based on pre-surgery assessments and may underestimate the actual dosage required. Accurately estimating the post-surgery LD dose post STN-DBS surgery is challenging due to post-surgery LEDD reduction. Consequently, many STN-DBS studies employ fixed LD doses of 200–250 mg for the LCT. In this study, an average LD dose of 433 mg was utilized, exceeding the dosages typically employed in prior research. Our kinematic gait analysis incorporated freezing episodes, potentially impacting the observed results. Nevertheless, consistent relationships between analyzed metrics and FOG were observed in all the analysis performed, aligning with prior research findings. ^51,71,72

In summary, at a group level, post-surgery FOG still responds to LD and DBS, even if the LD-induced improvement observed pre-surgery cannot be fully reproduced. A high inter-individual variability exists on response to treatment, claiming for an individualized approach to FOG. Importantly, freezers who not response to medication may still respond to stimulation: a pivotal observation regarding patients’ selection for surgery. Automatic detection of FOG is reliable and correlates well with clinical assessment. The identification of an association between gait variability-related metrics and FOG may provide valuable insights into the biomechanics of gait in FOG patients and may be used in the future to monitor disease progression and optimize therapeutic strategies.

Objective and study design

A cross-sectional and unblinded study assessing the effect of levodopa and stimulation in FOG emerging after STN-DBS, in five different experimental conditions.

Primary Outcome: the difference in the number of FOG episodes between Med OFF/Stim OFF and Med ON/Stim ON 130 Hz conditions.

Study Population

We reviewed the electronical medical records off all PD patients operated at our center since the beginning of the DBS program (2006), to identify those with Med On/Stim ON FOG. The presence of FOG was defined by a score ≥ 2 on item 3.11 (freezing) of the MDS-UPDRS part III on the Med ON/Stim ON state. FOG was considered only after 6 months into the post-surgical period, as we considered 6 months the time required for treatment stabilization. We called into clinic for assessment those still alive and available for evaluation. Patients unable to walk without ambulatory aids in the Med ON/Stim ON state, and those with severe osteo-articular or other non-neurological issues affecting gait, were excluded. All patients had been selected for surgery according to the CAPSIT-PD protocol, and none had levodopa-resistant axial signs before surgery during best ON state. DBS surgery followed standard stereotactic techniques, and postoperative neuroimaging confirmed lead position. DBS programming parameters were optimized regularly by a movement disorders specialist.

Clinical Evaluation

Clinical evaluations were conducted under five different conditions following a pre-determined order: 1) Medication OFF/Stimulation ON (Med OFF/Stim ON),, 2) Medication OFF/Stimulation OFF (Med OFF/Stim OFF) 3) Medication ON/Stimulation OFF (Med ON/ Stim OFF), 4) Medication ON/Stimulation ON at 130 Hz High-frequency stimulation (Med ON/Stim ON HFS), and 5) Medication ON/Stimulation ON at 60 Hz Low-frequency stimulation (Med ON/Stim ON LFS).

The full assessment lasted approximately 4 hours in a single morning session. Patients were evaluated on the "practical OFF drug" condition after 12 hours of medication withdrawal. A levodopa challenge test (LCT) was performed using the same dose of LD as used on the pre-surgery LCT. The Stim ON condition used the stimulation settings that had shown the best clinical results over the previous 6 months for each patient, while we adjusted the total energy delivered when testing LFS according to the total electrical energy delivered (TEED) formula: TEED (1 second) = voltage² X frequency x pulse width/impedance).^15,73. No patient was on LFS at the time of evaluation. A 30-minute interval was maintained between frequency changes and after stimulation arrest.

Motor evaluations using the MDS-UPDRS part III⁷⁴, gait/axial sub-score (MDS-UPDRS part III items 3.9–3.12)⁷⁴, non-gait/axial sub-score (MDS-UPDRS III – gait/axial sub-score), Hoehn and Yahr scale⁷⁵ and the Stand-Walk and Sit test (SWS-test)¹⁵ were performed on each of the five conditions. The SWS test is a standardized, timed test where subjects walk a 14-meter distance between sitting and standing. The overall duration of the test (time to walk 14 meters, SWS time) and the number of FOG episodes (#FOG episodes) were recorded. FOG was defined as a transient incapacity to move forward, despite the intention to walk, including both akinetic and "trembling in place" forms.^1,76 Patients performed the SWS test three times, and the data was averaged.

We defined responsiveness to LD and stimulation as any decrease on the number of FOG episodes in the SWS (i.e, a clinical improvement). Accordingly, patients who worsened or had no change with LD or stimulation were classified as LD-resistant or stimulation-resistant, respectively.

Imaging procedures - Electrode reconstruction and localization

To ensure electrode placement across subjects, we reconstructed the DBS electrodes of all patients whose neuroimaging data was possible to retrospectively retrieve (n = 11). This was performed through the advanced processing pipeline in Lead-DBS⁷⁷. For each patient, a post-operative CT scan was linearly coregistered to a pre-operative structural MRI (T1w), and then transformed into the ICBM 2009b NLIN asymmetric MNI space⁷⁸ both using Advanced Normalization Tools⁷⁹. A brain shift correction was employed and electrode trajectories were reconstructed automatically using the PaCER algorithm⁸⁰, following manual readjustments. Group visualization was performed using Lead Group with the DISTAL Atlas segmentation⁸¹.

Development of an inertial sensor-based classifier for FOG

We used data collected from previous studies ⁶² to develop a model for automatic detection of FOG. In summary, 20 PD patients (freezers and non-freezers) wearing seven wearable sensors (inertial measurement units, IMUs) fixed to different body parts performed a self-paced gait task. 180° degrees turns were removed from the analysis leaving only straight-line walking. Clinical annotation of FOG episodes (presence and duration) was made by a PD expert clinician (RB) based on video recordings.

Each IMU consisted of a tri-axial accelerometer, a gyroscope and a magnetometer (Xsens Technologies, Enschede, The Netherlands), that was fixed to the patient’s body using a Velcro elastic band. The inertial sensors were positioned in pelvis, right and left thighs, legs and feet. We used the raw signal collected by the IMUs and applied a supervised learning methodology on labelled data. This strategy trained a model to distinguish between different labels: FOG and non-FOG.

Continuous IMU data was processed into a one-second dataset, in which all 9 dimensions were split into non-overlapping sequential 1-second slices, each corresponding to 100 time-points, given the collection rate of 100Hz. Additionally, each resulting slice was then associated with the clinician label of belonging to a FOG episode or not. Considering the continuous nature of sensor data, an algorithm for identifying FOG events was envisaged as a time series (TS) classification problem. Deep learning (DL) has been increasingly used in the TS area, especially for multivariate time series problems.⁸² Convolutional Neural Network (CNN) based architecture was used to build a FOG detection model, consisting of three sequential convolution modules ⁸³The model was built in 13 patients and 7 were used to test the model on unseen data. FOG % was the main output of the model, representing the percentage of FOG presenting during straight gait.

A detailed description of the development and validation of this model is provided in Supplementary Methods.

Wearable sensor-based gait analysis:

During the SWS test, patients wore the same seven sensors as previously described in the same locations (pelvis, right and left thighs, legs and feet), during each of the five stimulation/medication conditions. The data collected from the IMUs were processed using the KINETIKOS (Coimbra, Portugal) cloud-based platform to reconstruct each subject's body motion using a 3D kinematic biomechanical model. Each trial out of the 3 SWS trials were individually computed and results were averaged. A final dataset consisted of 30 variables organized into 4 domains (spatio-temporal, asymmetry, variability, and non-linear metrics) selected based on their relevance in the literature (Supplementary Table S12). Spatiotemporal features were measured on both sides of the body, and the “worst-side” score was selected based on clinical assessment. For spatiotemporal variables, coefficients of variation (e.g. standard deviation of variable X score / mean of variable X score) and asymmetries [(variable X score on the right side – variable X score on the left side) / (variable X score on the right side + variable X score on the left side)] were calculated.

Statistical Analysis

The primary outcome of the study was the difference in the number of FOG episodes between Med OFF/Stim OFF and Med ON/Stim ON 130 Hz conditions. Secondary outcomes were: a) the change in the number of FOG episodes between the Med OFF/Stim OFF and the Med ON/Stim OFF and the Med OFF/Stim ON condition; b) the differences in the number of FOG episodes between the 130Hz and the 60 HZ conditions c) the association between the automatic detection of FOG and FOG assessed by clinical gait metrics. Exploratory analysis was performed to dissect kinematic dimensions related to FOG.

The statistical analysis was conducted in alignment with the prespecified hypothesis defined in our primary and secondary outcomes. This was done to mitigate the risk of Type II errors. Summary statistics were presented as mean (± SD) and median (IQR). To compare groups, paired non-parametric statistics were used (Wilcoxon signed rank test for comparisons of 2 groups and Friedman test for more than 2 groups with corrections to multiple comparisons performed using Dunn’s test). Spearman correlation was used to assess the association between kinematic and clinical variables. A Principal Component Analysis (PCA) used a dimensionality reduction method on the individual gait metrics to facilitate interpretation of data. Significance level was set at 0.05. Data processing was conducted using Matlab R2021a, Python 3.8 and R 4.2.2. Statistical analysis used Graphpad Prism 10.1.1.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Author Contribution Statement

RB designed the research project, including study design and statistical analysis, collected and analyzed the data and was a major contributor in writing the manuscript. PB designed the statistical analysis, analyzed the data and reviewed the manuscript. PPL reviewed the manuscript. CCR collected data and reviewed the manuscript. AV reviewed the manuscript. LCG reviewed the manuscript. BM analyzed kinematic data and developed the model for automatic FOG detection. SA analyzed kinematic data and developed the model for automatic FOG detection. VP analyzed kinematic data and developed the model for automatic FOG detection. MMR reviewed the manuscript. RM analyzed kinematic data and developed the model for automatic FOG detection. DM analyzed image data and performed reconstruction of DBS electrodes using Lead-DBS software. MM designed the research project, including study design and statistical analysis, analyzed the data and was a major contributor in writing the manuscript. MC designed the research project, including study design and statistical analysis, analyzed the data and was a major contributor in writing the manuscript. All authors read and approved the final manuscript.

Acknowledgements:

We thank all the participants and their families for their cooperation.

RB is supported by Fundação para a Ciência e Tecnologia (FCT) through a Ph.D. Scholarship (SFRH/BD/143797/2019) and Prémio João Lobo Antunes by Santa Casa da Misericórdia de Lisboa. The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript.

MM is supported by grant EurDyscover – EJPRD19-135, funded by FCT/MCTES through the European Joint Programme for Rare Disease (EJPRD/0001/2019), by the Michael J. Fox Foundation (MJFF-023180) and by the PsyPal project (grant agreement number 101137378) funded by the European Union's Horizon Research programme. The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript.

Ethical Compliance Statement:

This study was approved by the local ethics committee of the Comissão de Ética do CHULN e do CAML (Nº 345/21). Written, informed consent for this study was obtained and documented from all participants in accordance with the Declaration of Helsinki and its later amendments. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this work is consistent with those guidelines.

Competing Interests:

Author RB has received in the past 3 years, payment, honoraria or other support from AbbVie. Author RM. is a shareholder of a Kinetikos’s affiliated company, which owns the cloud-based platform used in this study. Author MM has received in the past 3 years, payment, honoraria or other support from Bial, Pharmacademy, Evidenze and AbbVie. Author MC has received in the past 3 years honoraria and sponsorship from Abbvie, Bial, Boston, Medtronic, Zambon and Ipsen. All other authors declare no financial competing interests.

Lewis S, Factor S, Giladi N, Nieuwboer A, Nutt J, Hallett M. Stepping up to meet the challenge of freezing of gait in Parkinson’s disease. Transl Neurodegener. 2022;11(1):23. doi:10.1186/s40035-022-00298-x
Okuma Y, Yanagisawa N. The clinical spectrum of freezing of gait in Parkinson’s disease. Mov Disord. 2008;23 Suppl 2:S426-430. doi:10.1002/mds.21934
Nonnekes J, Snijders AH, Nutt JG, Deuschl G, Giladi N, Bloem BR. Freezing of gait: a practical approach to management. Lancet Neurol. 2015;14(7):768-778. doi:10.1016/S1474-4422(15)00041-1
Schaafsma JD, Balash Y, Gurevich T, Bartels AL, Hausdorff JM, Giladi N. Characterization of freezing of gait subtypes and the response of each to levodopa in Parkinson’s disease. Eur J Neurol. 2003;10(4):391-398. doi:10.1046/j.1468-1331.2003.00611.x
Nieuwboer A, Giladi N. Characterizing freezing of gait in Parkinson’s disease: models of an episodic phenomenon. Mov Disord. 2013;28(11):1509-1519. doi:10.1002/mds.25683
Perez-Lloret S, Negre-Pages L, Damier P, et al. Prevalence, determinants, and effect on quality of life of freezing of gait in Parkinson disease. JAMA Neurol. 2014;71(7):884-890. doi:10.1001/jamaneurol.2014.753
Fasano A, Romito LM, Daniele A, et al. Motor and cognitive outcome in patients with Parkinson’s disease 8 years after subthalamic implants. Brain. 2010;133(9):2664-2676. doi:10.1093/brain/awq221
Fasano A, Aquino CC, Krauss JK, Honey CR, Bloem BR. Axial disability and deep brain stimulation in patients with Parkinson disease. Nat Rev Neurol. 2015;11(2):98-110. doi:10.1038/nrneurol.2014.252
Schlenstedt C, Shalash A, Muthuraman M, Falk D, Witt K, Deuschl G. Effect of high-frequency subthalamic neurostimulation on gait and freezing of gait in Parkinson’s disease: a systematic review and meta-analysis. Eur J Neurol. 2017;24(1):18-26. doi:10.1111/ene.13167
Castrioto A, Lozano AM, Poon YY, Lang AE, Fallis M, Moro E. Ten-year outcome of subthalamic stimulation in Parkinson disease: a blinded evaluation. Arch Neurol. 2011;68(12):1550-1556. doi:10.1001/archneurol.2011.182
Rodriguez-Oroz MC, Obeso JA, Lang AE, et al. Bilateral deep brain stimulation in Parkinson’s disease: a multicentre study with 4 years follow-up. Brain. 2005;128(Pt 10):2240-2249. doi:10.1093/brain/awh571
Bove F, Mulas D, Cavallieri F, et al. Long-term Outcomes (15 Years) After Subthalamic Nucleus Deep Brain Stimulation in Patients With Parkinson Disease. Neurology. Published online June 2, 2021:10.1212/WNL.0000000000012246. doi:10.1212/WNL.0000000000012246
Zampogna A, Cavallieri F, Bove F, et al. Axial impairment and falls in Parkinson’s disease: 15 years of subthalamic deep brain stimulation. NPJ Parkinsons Dis. 2022;8(1):121. doi:10.1038/s41531-022-00383-y
Barbosa R, Guedes LC, Cattoni MB, et al. Long-term follow-up of subthalamic nucleus deep brain stimulation in patients with Parkinson’s disease: An analysis of survival and disability milestones. Parkinsonism Relat Disord. 2024;118:105921. doi:10.1016/j.parkreldis.2023.105921
Moreau C, Defebvre L, Destée A, et al. STN-DBS frequency effects on freezing of gait in advanced Parkinson disease. Neurology. 2008;71(2):80-84. doi:10.1212/01.wnl.0000303972.16279.46
Ferraye MU, Debû B, Pollak P. Deep brain stimulation effect on freezing of gait. Mov Disord. 2008;23 Suppl 2:S489-494. doi:10.1002/mds.21975
Markham CH, Diamond SG. Long-term follow-up of early dopa treatment in Parkinson’s disease. Ann Neurol. 1986;19(4):365-372. doi:10.1002/ana.410190410
Tommasi G, Lopiano L, Zibetti M, et al. Freezing and hypokinesia of gait induced by stimulation of the subthalamic region. J Neurol Sci. 2007;258(1-2):99-103. doi:10.1016/j.jns.2007.03.002
El Ouadih Y, Marques A, Pereira B, et al. Deep brain stimulation of the subthalamic nucleus in severe Parkinson’s disease: relationships between dual-contact topographic setting and 1-year worsening of speech and gait. Acta Neurochir (Wien). Published online October 27, 2023. doi:10.1007/s00701-023-05843-9
Temiz G, Santin M des N, Olivier C, et al. Freezing of gait depends on cortico-subthalamic network recruitment following STN-DBS in PD patients. Parkinsonism Relat Disord. 2022;104:49-57. doi:10.1016/j.parkreldis.2022.10.002
Xie T, Kang UJ, Warnke P. Effect of stimulation frequency on immediate freezing of gait in newly activated STN DBS in Parkinson’s disease. J Neurol Neurosurg Psychiatry. 2012;83(10):1015-1017. doi:10.1136/jnnp-2011-302091
Xie T, Vigil J, MacCracken E, et al. Low-frequency stimulation of STN-DBS reduces aspiration and freezing of gait in patients with PD. Neurology. 2015;84(4):415-420. doi:10.1212/WNL.0000000000001184
Xie T, Bloom L, Padmanaban M, et al. Long-term effect of low frequency stimulation of STN on dysphagia, freezing of gait and other motor symptoms in PD. J Neurol Neurosurg Psychiatry. 2018;89(9):989-994. doi:10.1136/jnnp-2018-318060
Sidiropoulos C. Low-frequency stimulation of STN-DBS reduces aspiration and freezing of gait in patients with PD. Neurology. 2015;85(6):557. doi:10.1212/WNL.0000000000001823
Mügge F, Kleinholdermann U, Heun A, Ollenschläger M, Hannink J, Pedrosa DJ. Subthalamic 85 Hz deep brain stimulation improves walking pace and stride length in Parkinson’s disease patients. Neurol Res Pract. 2023;5(1):33. doi:10.1186/s42466-023-00263-7
Karl JA, Ouyang B, Goetz S, Metman LV. A Novel DBS Paradigm for Axial Features in Parkinson’s Disease: A Randomized Crossover Study. Mov Disord. 2020;35(8):1369-1378. doi:10.1002/mds.28048
Karl JA, Ouyang B, Verhagen Metman L. A Novel Dual-Frequency Deep Brain Stimulation Paradigm for Parkinson’s Disease. Neurol Ther. 2019;8(2):483-489. doi:10.1007/s40120-019-0140-5
Seger A, Gulberti A, Vettorazzi E, et al. Short Pulse and Conventional Deep Brain Stimulation Equally Improve the Parkinsonian Gait Disorder. J Parkinsons Dis. 2021;11(3):1455-1464. doi:10.3233/JPD-202492
Delval A, Tard C, Rambour M, Defebvre L, Moreau C. Characterization and quantification of freezing of gait in Parkinson’s disease: Can detection algorithms replace clinical expert opinion? Neurophysiol Clin. 2015;45(4-5):305-313. doi:10.1016/j.neucli.2015.09.009
Lewis SJG, Factor SA, Giladi N, et al. Addressing the Challenges of Clinical Research for Freezing of Gait in Parkinson’s Disease. Mov Disord. 2022;37(2):264-267. doi:10.1002/mds.28837
Mancini M, Bloem BR, Horak FB, Lewis SJG, Nieuwboer A, Nonnekes J. Clinical and methodological challenges for assessing freezing of gait: Future perspectives. Mov Disord. 2019;34(6):783-790. doi:10.1002/mds.27709
Cebi I, Scholten M, Gharabaghi A, Weiss D. Clinical and Kinematic Correlates of Favorable Gait Outcomes From Subthalamic Stimulation. Front Neurol. 2020;11:212. doi:10.3389/fneur.2020.00212
Kim R, Kim HJ, Shin C, et al. Long-term effect of subthalamic nucleus deep brain stimulation on freezing of gait in Parkinson’s disease. J Neurosurg. 2019;131(6):1797-1804. doi:10.3171/2018.8.JNS18350
Gavriliuc O, Paschen S, Andrusca A, Helmers AK, Schlenstedt C, Deuschl G. Clinical patterns of gait freezing in Parkinson’s disease and their response to interventions: An observer-blinded study. Parkinsonism Relat Disord. 2020;80:175-180. doi:10.1016/j.parkreldis.2020.09.043
Piboolnurak P, Lang AE, Lozano AM, et al. Levodopa response in long-term bilateral subthalamic stimulation for Parkinson’s disease. Mov Disord. 2007;22(7):990-997. doi:10.1002/mds.21482
Su ZH, Patel S, Gavine B, et al. Deep Brain Stimulation and Levodopa Affect Gait Variability in Parkinson Disease Differently. Neuromodulation. 2023;26(2):382-393. doi:10.1016/j.neurom.2022.04.035
Timmermann L, Braun M, Groiss S, et al. Differential effects of levodopa and subthalamic nucleus deep brain stimulation on bradykinesia in Parkinson’s disease. Mov Disord. 2008;23(2):218-227. doi:10.1002/mds.21808
Rocchi L, Chiari L, Horak F. Effects of deep brain stimulation and levodopa on postural sway in Parkinson’s disease. J Neurol Neurosurg Psychiatry. 2002;73(3):267-274. doi:10.1136/jnnp.73.3.267
Ferraye MU, Debû B, Fraix V, et al. Effects of subthalamic nucleus stimulation and levodopa on freezing of gait in Parkinson disease. Neurology. 2008;70(16 Pt 2):1431-1437. doi:10.1212/01.wnl.0000310416.90757.85
Mueller K, Urgošík D, Ballarini T, et al. Differential effects of deep brain stimulation and levodopa on brain activity in Parkinson’s disease. Brain Commun. 2020;2(1):fcaa005. doi:10.1093/braincomms/fcaa005
Artusi CA, Lopiano L, Morgante F. Deep Brain Stimulation Selection Criteria for Parkinson’s Disease: Time to Go beyond CAPSIT-PD. J Clin Med. 2020;9(12):3931. doi:10.3390/jcm9123931
Defer GL, Widner H, Marié RM, Rémy P, Levivier M. Core assessment program for surgical interventional therapies in Parkinson’s disease (CAPSIT-PD). Mov Disord. 1999;14(4):572-584. doi:10.1002/1531-8257(199907)14:4<572::aid-mds1005>3.0.co;2-c
Yamada K, Goto S, Hamasaki T, Kuratsu JI. Effect of bilateral subthalamic nucleus stimulation on levodopa-unresponsive axial symptoms in Parkinson’s disease. Acta Neurochir (Wien). 2008;150(1):15-22; discussion 22. doi:10.1007/s00701-007-1451-3
Ehgoetz Martens KA, Hall JM, Georgiades MJ, et al. The functional network signature of heterogeneity in freezing of gait. Brain. 2018;141(4):1145-1160. doi:10.1093/brain/awy019
Li N, Suo X, Zhang J, et al. Disrupted functional brain network topology in Parkinson’s disease patients with freezing of gait. Neurosci Lett. 2021;759:135970. doi:10.1016/j.neulet.2021.135970
Annic A, Moreau C, Salleron J, et al. Predictive factors for improvement of gait by low-frequency stimulation in Parkinson’s disease. J Parkinsons Dis. 2014;4(3):413-420. doi:10.3233/JPD-130337
Khoo HM, Kishima H, Hosomi K, et al. Low-frequency subthalamic nucleus stimulation in Parkinson’s disease: A randomized clinical trial. Movement Disorders. 2014;29(2):270-274. doi:10.1002/mds.25810
di Biase L, Fasano A. Low-frequency deep brain stimulation for Parkinson’s disease: Great expectation or false hope? Movement Disorders. 2016;31(7):962-967. doi:10.1002/mds.26658
Schaafsma JD, Giladi N, Balash Y, Bartels AL, Gurevich T, Hausdorff JM. Gait dynamics in Parkinson’s disease: relationship to Parkinsonian features, falls and response to levodopa. J Neurol Sci. 2003;212(1-2):47-53. doi:10.1016/s0022-510x(03)00104-7
Hausdorff JM. Gait dynamics in Parkinson’s disease: common and distinct behavior among stride length, gait variability, and fractal-like scaling. Chaos. 2009;19(2):026113. doi:10.1063/1.3147408
Hausdorff JM, Schaafsma JD, Balash Y, Bartels AL, Gurevich T, Giladi N. Impaired regulation of stride variability in Parkinson’s disease subjects with freezing of gait. Exp Brain Res. 2003;149(2):187-194. doi:10.1007/s00221-002-1354-8
Hausdorff JM. Gait dynamics, fractals and falls: finding meaning in the stride-to-stride fluctuations of human walking. Hum Mov Sci. 2007;26(4):555-589. doi:10.1016/j.humov.2007.05.003
Mendonça MD, Barbosa R, Seromenho-Santos A, Reizinho C, Bugalho P, CHLO Movement Disorders Surgery Group. Early use of 80 Hz subthalamic stimulation in Parkinson’s disease as an alternative for High-frequency stimulation induced gait changes and postural instability. Brain Stimul. 2018;11(3):620-622. doi:10.1016/j.brs.2017.12.005
Ramdhani RA, Patel A, Swope D, Kopell BH. Early Use of 60 Hz Frequency Subthalamic Stimulation in Parkinson’s Disease: A Case Series and Review. Neuromodulation: Technology at the Neural Interface. 2015;18(8):664-669. doi:10.1111/ner.12288
Ferraye MU, Debû B, Fraix V, et al. Effects of pedunculopontine nucleus area stimulation on gait disorders in Parkinson’s disease. Brain. 2010;133(1):205-214. doi:10.1093/brain/awp229
Moro E, Hamani C, Poon YY, et al. Unilateral pedunculopontine stimulation improves falls in Parkinson’s disease. Brain. 2010;133(1):215-224. doi:10.1093/brain/awp261
Snijders AH, Nijkrake MJ, Bakker M, Munneke M, Wind C, Bloem BR. Clinimetrics of freezing of gait. Mov Disord. 2008;23 Suppl 2:S468-474. doi:10.1002/mds.22144
Snijders AH, Haaxma CA, Hagen YJ, Munneke M, Bloem BR. Freezer or non-freezer: clinical assessment of freezing of gait. Parkinsonism Relat Disord. 2012;18(2):149-154. doi:10.1016/j.parkreldis.2011.09.006
Cockx H, Klaver E, Tjepkema-Cloostermans M, van Wezel R, Nonnekes J. The Gray Area of Freezing of Gait Annotation: A Guideline and Open-Source Practical Tool. Mov Disord Clin Pract. 2022;9(8):1099-1104. doi:10.1002/mdc3.13556
Giannakopoulou KM, Roussaki I, Demestichas K. Internet of Things Technologies and Machine Learning Methods for Parkinson’s Disease Diagnosis, Monitoring and Management: A Systematic Review. Sensors (Basel). 2022;22(5):1799. doi:10.3390/s22051799
Reches T, Dagan M, Herman T, et al. Using Wearable Sensors and Machine Learning to Automatically Detect Freezing of Gait during a FOG-Provoking Test. Sensors (Basel). 2020;20(16):4474. doi:10.3390/s20164474
Bastos P, Meira B, Mendonça M, Barbosa R. Distinct gait dimensions are modulated by physical activity in Parkinson’s disease patients. J Neural Transm (Vienna). 2022;129(7):879-887. doi:10.1007/s00702-022-02501-9
Kwon H, Clifford GD, Genias I, et al. An Explainable Spatial-Temporal Graphical Convolutional Network to Score Freezing of Gait in Parkinsonian Patients. Sensors. 2023;23(4):1766. doi:10.3390/s23041766
Plotnik M, Hausdorff JM. The role of gait rhythmicity and bilateral coordination of stepping in the pathophysiology of freezing of gait in Parkinson’s disease. Mov Disord. 2008;23 Suppl 2:S444-450. doi:10.1002/mds.21984
Plotnik M, Giladi N, Hausdorff JM. Bilateral coordination of walking and freezing of gait in Parkinson’s disease. Eur J Neurosci. 2008;27(8):1999-2006. doi:10.1111/j.1460-9568.2008.06167.x
Plotnik M, Giladi N, Balash Y, Peretz C, Hausdorff JM. Is freezing of gait in Parkinson’s disease related to asymmetric motor function? Ann Neurol. 2005;57(5):656-663. doi:10.1002/ana.20452
Nanhoe-Mahabier W, Snijders AH, Delval A, et al. Walking patterns in Parkinson’s disease with and without freezing of gait. Neuroscience. 2011;182:217-224. doi:10.1016/j.neuroscience.2011.02.061
Callisaya ML, Blizzard L, Schmidt MD, et al. Gait, gait variability and the risk of multiple incident falls in older people: a population-based study. Age Ageing. 2011;40(4):481-487. doi:10.1093/ageing/afr055
Ma L, Mi TM, Jia Q, Han C, Chhetri JK, Chan P. Gait variability is sensitive to detect Parkinson’s disease patients at high fall risk. Int J Neurosci. 2022;132(9):888-893. doi:10.1080/00207454.2020.1849189
Melbourne JA, Kehnemouyi YM, O’Day JJ, et al. Kinematic adaptive deep brain stimulation for gait impairment and freezing of gait in Parkinson’s disease. Brain Stimulation: Basic, Translational, and Clinical Research in Neuromodulation. 2023;16(4):1099-1101. doi:10.1016/j.brs.2023.07.003
Barbe MT, Amarell M, Snijders AH, et al. Gait and upper limb variability in Parkinson’s disease patients with and without freezing of gait. J Neurol. 2014;261(2):330-342. doi:10.1007/s00415-013-7199-1
Frazzitta G, Pezzoli G, Bertotti G, Maestri R. Asymmetry and freezing of gait in parkinsonian patients. J Neurol. 2013;260(1):71-76. doi:10.1007/s00415-012-6585-4
Mazzone P, Lozano A, Stanzione P, et al. Implantation of human pedunculopontine nucleus: a safe and clinically relevant target in Parkinson’s disease. Neuroreport. 2005;16(17):1877-1881. doi:10.1097/01.wnr.0000187629.38010.12
Goetz CG, Tilley BC, Shaftman SR, et al. Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS): scale presentation and clinimetric testing results. Mov Disord. 2008;23(15):2129-2170. doi:10.1002/mds.22340
Hoehn MM, Yahr MD. Parkinsonism: onset, progression and mortality. Neurology. 1967;17(5):427-442. doi:10.1212/wnl.17.5.427
Nutt JG, Bloem BR, Giladi N, Hallett M, Horak FB, Nieuwboer A. Freezing of gait: moving forward on a mysterious clinical phenomenon. Lancet Neurol. 2011;10(8):734-744. doi:10.1016/S1474-4422(11)70143-0
Neudorfer C, Butenko K, Oxenford S, et al. Lead-DBS v3.0: Mapping deep brain stimulation effects to local anatomy and global networks. Neuroimage. 2023;268:119862. doi:10.1016/j.neuroimage.2023.119862
Fonov V, Evans AC, Botteron K, et al. Unbiased average age-appropriate atlases for pediatric studies. Neuroimage. 2011;54(1):313-327. doi:10.1016/j.neuroimage.2010.07.033
Avants BB, Epstein CL, Grossman M, Gee JC. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal. 2008;12(1):26-41. doi:10.1016/j.media.2007.06.004
Husch A, V Petersen M, Gemmar P, Goncalves J, Hertel F. PaCER - A fully automated method for electrode trajectory and contact reconstruction in deep brain stimulation. Neuroimage Clin. 2018;17:80-89. doi:10.1016/j.nicl.2017.10.004
Ewert S, Plettig P, Li N, et al. Toward defining deep brain stimulation targets in MNI space: A subcortical atlas based on multimodal MRI, histology and structural connectivity. Neuroimage. 2018;170:271-282. doi:10.1016/j.neuroimage.2017.05.015
Ismail Fawaz H, Forestier G, Weber J, Idoumghar L, Muller PA. Deep learning for time series classification: a review. Data Min Knowl Disc. 2019;33(4):917-963. doi:10.1007/s10618-019-00619-1
Tang W, Long G, Liu L, Zhou T, Jiang J, Blumenstein M. Rethinking 1D-CNN for Time Series Classification: A Stronger Baseline.; 2020.

There is a conflict of interest Author RB has received in the past 3 years, payment, honoraria or other support from AbbVie. Author RM. is a shareholder of a Kinetikos’s affiliated company, which owns the cloud-based platform used in this study. Author MM has received in the past 3 years, payment, honoraria or other support from Bial, Pharmacademy, Evidenze and AbbVie. Author MC has received in the past 3 years honoraria and sponsorship from Abbvie, Bial, Boston, Medtronic, Zambon and Ipsen. All other authors declare no financial competing interests

SupplementaryTable1.docx
SupplementaryTable2.docx
SupplementaryTable3.docx
SupplementaryTable4.docx
SupplementaryTable5.docx
SupplementaryTable6.docx
SupplementaryTable7.docx
SupplementaryTable8.docx
SupplementaryTableS9.docx
SupplementaryTable10.docx
SupplementaryTable11.docx
SupplementaryTableS12.docx
figuresS1.pdf
Supplementary Figure S1: Reconstruction of DBS electrodes (11/17 patients). DBS electrode reconstruction confirmed a similar placement within the subthalamic nucleus (STN) for all eleven patients assessed. Mean MNI coordinate of the right electrode’s most ventral/distal contact: x: 10.98 ± 1.40 mm; y: -14.44 ± 1.11 mm; z: -9.26 ± 1.03 mm; Mean MNI coordinate of the left electrode’s most ventral/distal contact: x: -11.04 ± 1.63 mm; y: -14.51 ± 1.22 mm; z: -8.85 ± 1.29 mm
figuresS2.pdf
Supplementary Figure S2: #FOG episodes (A) MDS-UPDRS III non axial score (B) MDS-UPDRS III axial score (c) and AIMS scores (D) per condition and per subgroup of patients, where each colour represent a different subgroup of patient: Patients worsening with medication are depicted in blue (left), patients worsening under stimulation are depicted in orange (middle) and patients improving under both stimulation and medication are depicted in yellow (right): #FOG episodes, number of FOG episodes; MED only, Medication ON/Stimulation OFF; STIM only, Medication OFF/Stimulation ON; Med+Stim, MedicationON/StimulationON 130HZ; * p<0.05, ** p<0.01; ***p>0.001
figuresS3.pdf
Supplementary Figure S3 –PCI scores (and normalized PC1 scores) per condition and par subgroup of patients are depicted with different subgroups being represented by different colors: Patients worsening with medication are depicted in blue, patients worsening with stimulation are depicted in orange, patient responding to both stimulation and medication are represented in yellow, patients benefiting from LFS are represented in purple, the ones better under HFS are represented in green and rose depicts the subgroup of patients with equal improvements under HFS and LFS (A, B,C, D ) PC, principal component; #FOG, number of FOG episodes; SWS time, SWS gait time; %FOG, FOG % detected using a automatic model; OFF, medicationOFF/stimulationOFF, medication only, medicationON/stimulation OFF; stimulation only, medicationOFF/stimulationON; MED+STIM 130, medication ON/Stimulation ON 130 Hz; MED+STIM 60, medication ON/Stimulation ON 60 Hz; *, p<0.05
figuresS4.pdf
Supplementary Figure S4 - Mean values of spatio-temporal variables and the corresponding variability metric are presented at the Med+Sim 130 Hz condition and at the Med+Stim 60 Hz condition. Med+Stim 130 Hz, Medication ON/Stimulation ON 130 Hz; Med+Stim 60 Hz, Medication ON/Stimulation ON 60 Hz; *, p<0.05

Download PDF

Version 1

posted

You are reading this latest preprint version

The effect of Levodopa and Stimulation on post-surgery Freezing of Gait in STN-DBS Parkinson's Disease patients: a clinical and kinematic analysis

Status:

Version 1

Abstract

Background

Objective

Methods

Results

Discussion

Figures

Introduction

Results

1 – FOG assessment in Med OFF/Stim OFF vs Med ON/Stim ON

2 - Assessing individual effect of stimulation and LD on FOG

3 - The role of HFS and LFS on FOG

4- Using IMU-based approaches to study gait and FOG

Discussion

Individual subjects have distinct FOG responses to Stimulation or Levodopa

Stimulation Frequency can lead to important differences in FOG response to stimulation

Automatic FOG detection correlates well with clinical FOG metrics

Methods

Objective and study design

Study Population

Clinical Evaluation

Imaging procedures - Electrode reconstruction and localization

Development of an inertial sensor-based classifier for FOG

Wearable sensor-based gait analysis:

Statistical Analysis

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1