A Randomized Controlled Trial of Exercise for Parkinsonian Individuals with Freezing of Gait

Abstract

Background

Exercises with motor complexity induce neuroplasticity in individuals with Parkinson’s disease (PD), but its effects on freezing of gait (FOG) are unknown.

Objective

To verify if adapted resistance training with instability (ARTI-exercises with motor complexity) will be more effective than traditional motor rehabilitation (TMR-exercises without motor complexity) in improving FOG severity, outcomes linked to FOG, and brain function.

Methods

Freezers were randomized either to the ARTI group (n=17) or to the active control group (TMR n=15). Both training groups performed exercises three times a week for 12 weeks. The primary outcome was the new-FOG questionnaire (NFOGQ). Secondary outcomes were FOG-ratio (turning task), cognitive inhibition (Stroop-III test), motor signs (Unified Parkinson’s Disease Rating Scale part-III [UPDRS-III]), quality of life (PD Questionnaire [PDQ-39]), anticipatory postural adjustment (APA-leg lifting task) and brain activation during a functional magnetic resonance imaging (fMRI) protocol of simulated APA task. Outcomes were evaluated before and after interventions.

Results

Only ARTI improved all of the outcomes (P<0.05). ARTI was more effective than TMR in improving FOG-ratio, motor signs, quality of life, APA amplitude, and brain activation (P<0.05). Our results are clinically relevant because improvement in NFOGQ (−4.4 points) and UPDRS-III (−7.4 points) scores exceeded the minimally detectable change (TMR group data) and the moderate clinically important difference suggested for PD, respectively. The changes in mesencephalic locomotor region activation and in APA amplitude explained the changes in NFOGQ scores and in FOG-ratio following ARTI, respectively.

Conclusions

ARTI is able to cause significant clinical improvement and brain plasticity in freezers.

INTRODUCTION

Freezing of gait (FOG) affects more than one-third of individuals with Parkinson’s disease (PD),^{1, 2} and it is a common causes of falls, dependency, and poor quality of life.^{1, 3} FOG is a complex phenomenon and its pathophysiology remains unknown. FOG typically occurs during step initiation and turns,⁴ likely due to abnormal postural preparation for step initiation.^{4, 5}

Individuals with FOG (freezers) show delayed step initiation associated with repetitive anticipatory postural adjustments (APAs) as if they cannot inhibit their postural preparation and release their stepping motor program.⁶ In fact, FOG has been associated with small and prolonged APAs during step initiation⁷ and deficits in cognitive inhibition.⁵ Thus, we hypothesize that improving APAs and cognitive inhibition could contribute to decreased FOG severity.

Deficits in APAs and cognitive inhibition in freezers are linked to impairments in mesencephalic locomotor region (MLR). For example, the pedunculopontine nucleus, one of the major nuclei of the MLR, has neurons related to APAs preceding step initiation as well as to the step itself.⁸ Freezers have greater grey matter atrophy of the MLR than non-freezers⁹ and the more right hemisphere reduction of the pedunculopontine nucleus tract volume in freezers, the poorer the performance on Stroop task (i.e., impaired cognitive inhibition).¹⁰ Freezers have a decreased Blood Oxygenation Level Dependent (BOLD) activation within the MLR during a functional magnetic resonance imaging (fMRI) protocol that simulated a walking task,¹¹ which was correlated to FOG severity.¹¹ Taken together, these results suggest that increasing MLR activation during APAs could decrease FOG severity. To verify MLR activation during APAs, we used a leg lifting task in an event-related fMRI protocol that was recently validated by our group¹² as having similar characteristics as step initiation.

The effects of exercise interventions on brain function in freezers are equivocal, as one study has demonstrated increased BOLD activation in frontal-parietal areas after 4 weeks of exercise,¹³ while another study failed to demonstrate changes in brain activation after a 12-week exercise intervention.¹⁴ No study has investigated the effects of exercise intervention on MLR activation and gait initiation (APAs) in freezers. Thus, as abnormal APAs and dysfunction in cognitive inhibition are associated with FOG^{5, 7} and linked to impairments in MLR, we hypothesized that exercise interventions aiming to improve these aspects could increase MLR activation and to decrease FOG severity.

Recently, we found that Resistance Training with Instability (RTI), an intervention with high motor complexity (i.e., exercises that simultaneously require high cognitive, proprioceptive, and motor control demands)^{15, 16}, improved not only APA and cognitive function, but also motor symptoms and quality of life^{16, 17} in individuals with PD without FOG. RTI may be particularly useful for freezers when adapted (ARTI) to involve free weight exercises that require APAs (e.g., lunge exercise) and cognitive inhibition (e.g., dual task exercise) performed on unstable devices. Previous studies have shown more brain activation after high than low motor complexity exercises,¹⁵ so we hypothesized greater MLR activation in an event-related fMRI¹² after ARTI than traditional motor rehabilitation (TMR), a control intervention without motor complexity exercises.

Therefore, this randomized controlled trial (RCT) has two aims. First, we compared the effects of ARTI and active control group (TMR) on FOG severity assessed subjectively (New Freezing of Gait Questionnaire [NFOGQ] scores - primary outcome) and objectively (FOG-ratio during a turning task), APA magnitude (leg lifting task), cognitive inhibition (Stroop-III test), and brain activation during APAs in an event-related fMRI. Second, we tested if improved APAs, cognitive inhibition, and brain activation could explain decrease in FOG severity after ARTI. Clinical secondary outcomes (motor symptoms and quality of life) and brain activation of the cerebellar locomotor area (CLR) and supplementary motor area (SMA) that are linked to FOG and gait^18–20 were also assessed.

MATERIALS AND METHODS

Study Design and Participants

This was a prospective, single-center, parallel-group, single-blinded RCT conducted between June 2018 and April 2019. Freezers were recruited from the Movement Disorders Clinic in the School of Medicine at University of São Paulo. The diagnosis of idiopathic PD was confirmed by a movement disorders specialist in accordance with UK Parkinson’s Disease Society Brain Bank diagnostic criteria.²¹ A movement disorders specialist assessed probable FOG using videos of objective tests (e.g., step-over obstacles, turning clockwise and counter clockwise, and walking through a doorway) and the NFOGQ.²² Individuals with PD were classified as freezers if FOG was observed in the videos and if they answered affirmatively item 1 (Do you experience FOG?) and scored >1 on item 3 (How frequently do you experience freezing episodes during turning?) and 5 (How frequently do you experience episodes of freezing when initiating the first step?) of the NFOGQ during ON medication status. Eligibility criteria were: 1) Hoehn and Yahr stage range 3–4; 2) stable dopaminergic therapy 2 months before and during the period of study; 3) 49–85 years of age; 4) able to walk safely for 20 m without walking aids; 5) absence of neurological disorders (other than PD); 6) absence of significant arthritis, musculoskeletal or vestibular disorders, severe tremor, claustrophobia, and metal in the body; 7) high quality of brain volumes acquired during the fMRI (head motion <1mm);²³ 8) Mini-Mental State Examination score >23;²⁴ 9) no physical exercise practice in the three months preceding study commencement. Individuals gave their written informed consent to participate. The study was approved by University’s Ethical Committee (School of Physical Education and Sport - ref. 2011/12), registered at the National Clinical Trial (UTN-U1111–1215-9956) and at the Brazilian Clinical Trials Registry (RBR-83VB6B).

Study procedures

Freezers were assessed (all outcomes including fMRI) and trained in the clinically defined “on” state (fully medicated) within 1.5 to 2h of taking their morning dose of dopaminergic medication. The primary outcome measure was the NFOGQ score. Secondary outcome measures included FOG-ratio, cognitive inhibition, motor symptoms, quality of life, APA amplitude and duration, and brain function. On the first day, a physical therapist blind to the experimental design assessed FOG severity using the NFOGQ²² and a 2-minute turning task (alternating 360 degree turns to the right and to the left),²⁵ and motor symptoms using the Unified Parkinson’s Disease Rating Scale part III (UPDRS-III).²⁶ On the second day, a physical therapist (trained by a neuropsychologist) blind to the experimental design assessed cognitive inhibition using the Stroop Color-Word Test-Victoria version²⁷ (both time and error were used to estimate the cognitive interference scores on Stroop-III test²⁸), and quality of life using the Parkinson’s Disease Questionnaire (PDQ-39).²⁹ On the third day, individuals performed the leg lifting task in an event-related fMRI to assess APA, as previously published.¹² Amplitude and duration of APAs and brain activation (BOLD signal) were recorded during the leg lifting task. See Supplementary Material for a detailed description of the assessments.

Randomization

After baseline assessments, freezers were classified (a statistician blind to experimental design) into quartiles regarding their NFOGQ scores. Freezers from each quartile were randomly assigned to the active control group (TMR) or the experimental group (ARTI). Post-test assessments were performed 24 hours after the end of the training period (12 weeks).

Image acquisition and processing for maps of brain activation

Images were obtained using a 3.0 T MR system (Achieva, Philips Medical Imaging, The Netherlands), 32-channel head coil (80 mT/m gradient maximum amplitude). BOLD-sensitive images were acquired using T2*- weighted gradient echo planar imaging (EPI), TR = 2.000 ms, TE = 30 ms, 40 slices, 3.0 mm slice thickness, 0.3 mm interslice gap, 3.0 mm isotropic voxels, 240 volumes (acquisition time: 10 min). Anatomical T1-weighted 3D images were used for reference and image registration (T1-FFE; TR=7ms, TE=3.2ms, 180 slices, FA=8, 1mm isotropic voxels). fMRI data was processed using FSL software version 6.0 (www.fmrib.ox.ac.uk/fsl/)³⁰. The volumes were processed by movement correction and calculation of mean displacement (MCFLIRT), spatial smoothing (FWHM = 5 mm) and spatial normalization to standard space (affine, 12 DoF).^{31, 32} As previously recommended, the level of head motion was less than 1 mm to avoid erroneous inference on neuronal function²³. MCFLIRT is a motion correction tool based on FLIRT (FMRIB’s Linear Image Registration Tool), which is a fully automated robust and accurate tool for linear (affine) intra- and inter-modal brain image registration used in FEAT (FMRI Expert Analysis Tool) from FSL. Advantages of MCFLIRT are accurate to 0.1mm in tests, end-slice corruption is minimized, plots of motion parameters and RMS movement can be produced (FEAT reports the motion parameters and can use them as regressors), and it works by serial registrations of each EPI in a time-series to the halfway timepoint image. Further, the calculated motion parameters by MCFLIRT tool during the pre-procedure were added in the statistic model as confounders/covariates, using the standard option, which considered only 6 parameters of motion correction (3 for translation and 3 for rotation) during the analysis, in addition to the task-related regressors.^{31, 33} The event of interest was the time from the onset of the first stimulus (circle for mental preparation – see Supplementary Material for a detailed description) until 1 s after leg lifting onset. In the first level analysis, a linear model was implemented to estimate BOLD signal in the event of interest compared to resting periods. In the second level, the contrast post>pre was acquired for each subject. Contrast maps comparing ARTI and TMR were obtained using a mixed-effects model to include within-subject variances of parameter estimates. Significance was set at 1% for single voxel level and at 5% (corrected for multiple comparison using the Gaussian Random Field theory) at mass-cluster level for group analyses. The beta of BOLD signal change of regions of interest (MLR, CLR, and SMA) and of the group activation map was extracted using the featquery processing routine from FSL. Peak coordinates of these regions are shown in Table 1.

Table 1.

Peak coordinates of the region in the group activation map and of the regions of interest of the right hemisphere.

Neural Region	Size (mm)	x	y	z
Inferior and middle temporal gyrus (MITG)	6mm	60	−26	−18
*Mesencephalic locomotor region (MLR)	6mm	6	−30	−19
*Cerebellar locomotor region (CLR)	6mm	7	−52	−16
#Supplementary motor area (SMA)	8mm	3	−13	61

MNI coordinates for the peaks of intensity in each region of right hemisphere.

^*Peak coordinates were identified based on a previous study (Fling et al., 2014)²⁰.

^#Peak coordinate was identified based on a previous study (Shine et al., 2013)¹⁹.

Interventions

Both TMR and ARTI groups trained 3 days per week for 12 weeks (36 training sessions) in different facilities. Each training session lasted between 80–90 minutes. TMR, an active control group, consisted of exercises with focus on stretching, gait, balance, posture, and lower- and upper-limbs free weight exercises. ARTI is an adaptation of our RTI program for PD that has been previously published.¹⁶ See Supplementary Material for detail of the ARTI and TMR programs. ARTI consisted of seven lower-limb and upper-limb free weight exercises (half-squat, plantar flexion, chest press, knee-lifting stand, lunge, reverse fly, and dual-task squat) combined with unstable devices (i.e., foam pad, dyna discs, balance discs, BOSU®, and Swiss Ball). Throughout the 12-wk period, there was a progressive increase in motor complexity (see Supplementary Table 1 and Figure 1 for exercise progression and pictures, respectively). ARTI sessions were individualized and monitored by trainers knowledgeable in working with individuals with PD. TMR sessions were group-based (up to 8 individuals), monitored by a physical therapist knowledgeable in working with individuals with PD.

Figure 1.

The trial profile with schematic representation of participant recruitment and allocation. TMR= traditional motor rehabilitation; ARTI= adapted resistance training with instability for freezers.

Statistical analyses

Considering an effect size (ES) of 0.50 based on a previous study that reported a significant interaction effect for the NFOGQ scores,³⁴ an error of 0.05, a desired power of 0.80, and a correlation among repeated measures of 0.5, the power analysis determined that 14 individuals per group were needed (G*Power v. 3.0.10, Universität Kiel, Germany). We projected an attrition rate of 30%, which resulted in our target sample size of 20 patients per group to start the study.

We calculated the minimal detectable change (MDC) on NFOGQ scores using NFOGQ data from the control group (pre- and post-test), as follows: MDC₉₅ = 1.96 × √2 × standard error of measurement.³⁵

Normality and the presence of extreme observations were assessed through Shapiro-Wilk test and box-plots, respectively. Non-normal data (FOG-ratio and APA amplitude and duration) were log transformed.

To compare the characteristics between groups at baseline, we used independent t-tests.

To test for the effects of training protocols on outcomes (NFOGQ scores, FOG-ratio, APA amplitude and duration values, Stroop-III test values, UPDRS-III scores, and PDQ-39 scores), linear mixed models were used having group (TMR and ARTI) and time (pre and post) as fixed factors, and individuals as a random factor. Whenever a significant F-value was obtained, a Tukey’s adjustment was used for multi-comparison purposes.

Within-group (pre- to post-changes) and between-groups (post-changes) ES were calculated using Cohen’s d³⁶ for each outcome. ESs were classified as small (ES≤0.49), medium (ES 0.50–0.79), and large (ES≥0.80).

To compare the beta of BOLD signal change of the area identified in the map group contrasts and of the regions of interest (MLR, CLR, and SMA) for the contrast post > pre between the groups, we used independent t-tests.

A linear multiple regression (forward stepwise method) was used to verify if changes in independent variables (APA, cognitive inhibition, and beta of BOLD signal change of the regions of interest [MLR, CLR, and SMA] and of the area identified in the map group contrasts) could explain changes in FOG severity (NFOGQ scores and FOG-ratio – dependent variables) following ARTI. Only dependent variables with a low collinearity with previous selected independent variables entered in the regression model.

Results are presented as mean ± standard deviation (SD). Statistical procedures were implemented using SAS 9.2^® (Institute Inc., Cary, NC, USA) and the level of significance was set at P≤0.05.

RESULTS

Participants

One hundred and two individuals volunteered for the study and signed the written consent (Figure 1). Fifty-four did not fulfill inclusion criteria, three died, and five had transportation problems that precluded their participation in the study. Forty individuals performed baseline testing and were randomized into each group (TMR and ARTI). Thirty-two participants (TMR n=15; ARTI n=17) composed the final sample as eight participants dropped out prior to post-assessment (TMR n=5; ARTI n=3). Of the ARTI group, one participant experienced lower limb pain (no medical intervention required) and did not want to continue in the study; one participant dropped due to family problems; and one participant had surgery due to a fall getting out of his car. Of the TMR group, we were unable to contact two participants; two participants had family problems; and one participant had surgery due to a fall in her house.

At baseline, there were no between-group differences in any demographic, anthropometrical, and clinical characteristic (Table 2), and outcomes (Table 3).

Table 2.

Characteristics of the freezers at baseline, by group (mean ± SD).

	TMR (n=15)	ARTI (n=17)	P value
Demographic
Men/women (number)	9/6	12/5
Age (years)	66.8 ± 8.9	64.6 ± 10.5	0.542
Educational level (years)	10.5 ± 5.8	12.9 ± 5.8	0.253
Anthropometrical
Body mass (kg)	67.5 ± 9.1	73.4 ± 13.5	0.162
Height (cm)	1.6 ± 0.1	1.6 ± 0.1	0.577
Body mass index (kg/m2)	25.3 ± 2.7	26.8 ± 4.1	0.235
Clinical
Mini-Mental State Examination (score)	25.6 ± 2.0	25.5 ± 1.7	0.986
Range	24 – 30	24 – 29
Years since diagnosis (years)	10.0 ± 5.6	7.7 ± 4.0	0.204
Range	3 – 25	2 – 16
Hoehn and Yahr staging scale (a.u)	3.2 ± 0.4	3.1 ± 0.3	0.296
3	11	15
4	4	2
Symptom-dominant side (R/L/B)	1/3/11	3/2/12
Postural instability and gait disturbance (score)	8.9 ± 2.6	7.7 ± 1.9	0.153
Range	4 – 13	4 – 12
L-Dopa equivalent daily dose (mg/day)	854.8 ± 251.9	772.5 ± 275.8	0.342
Range	500 – 1300	300 – 1200
Antiparkinsonian medication
L-Dopa (mg/day)	503.3 ± 185.6	437.5 ± 211.7	0.364
Range	200 – 800	200 – 1000
Amantadine (mg/day)	159.1 ± 66.4	163.6 ± 80.9	0.886
Range	100 – 300	100 – 300
Entacapone (mg/day)	175.0 ± 41.8	157.1 ± 53.5	0.513
Range	100 – 200	100 – 200
MAOB inhibitors (mg/day)	3.9 ± 4.2	2.4 ± 3.2	0.408
Range	0.25 – 10	0.125 – 10
Dopamine agonists (mg/day)	2.3 ± 2.7	2.1 ± 1.8	0.846
Range	0.5 – 9	0.75 – 5

TMR= Traditional Motor Rehabilitation; ARTI= adapted resistance training with instability; L-Dopa = levodopa; a.u = Arbitrary units; R = right; L = left; B = both. MAOB = monoamine oxidase type B inhibitors.

Table 3.

Outcomes at pre- and post-training assessments for each group of freezers.

Groups	Outcomes	Change from Pre- to Posttraining		Difference at Posttraining: ARTI vs. TMR
		MD (95% CI)	P value	MD (95% CI)	P value
	Primary outcome
	NFOGQ (score)
TMR pre	22.3 ± 6.0
TMR post	22.0 ± 5.9	−0.3 (−2.0 to 2.6)	0.989
ARTI pre	21.6 ± 5.7
ARTI post	17.2 ± 3.1	−4.4 (−6.8 to −1.9)	<.0001	−4.8 (−0.2 to 9.9)	0.069
	Secondary outcome
	FOG-ratio (a.u.)
TMR pre	12.8 ± 8.8
TMR post	20.9 ± 8.5	8.1 (2.4 to 13.8)	0.002
ARTI pre	9.8 ± 8.2
ARTI post	4.8 ± 3.8	−5.0 (−10.3 to 0.4)	0.078	−16.1 (−23.3 to −8.7)	<.0001
	Stroop-III test (a.u.)
TMR pre	72.1 ± 44.4
TMR post	72.5 ± 41.1	0.4 (−6.4 to 5.6)	0.997
ARTI pre	79.1 ± 40.3
ARTI post	65.1 ± 31.4	−14.0 (−19.6 to −8.2)	<.0001	−7.4 (−30.5 to 45.3)	0.950
	UPDRS-III (score)
TMR pre	51.4 ± 10.6
TMR post	54.8 ± 11.6	3.4 (−8.7 to 2.0)	0.347
ARTI pre	46.4 ± 11.4
ARTI post	39.0 ± 10.2	−7.4 (−12.5 to −2.4)	0.001	−15.8 (−26.4 to −5.1)	0.001
	PDQ-39 (%)
TMR pre	43.2 ± 11.0
TMR post	47.0 ± 10.8	4.2 (−2.0 to 9.5)	0.314
ARTI pre	36.0 ± 15.6
ARTI post	27.6 ± 9.6	−8.4 (−13.8 to −2.9)	0.001	−19.4 (−31.0 to −7.7)	<.0001
	*APA amplitude (a.u.)
TMR pre	38.9 ± 17.5
TMR post	34.2 ± 15.2	−4.7 (−7.7 to 17.0)	0.731
ARTI pre	34.1 ± 11.7
ARTI post	53.8 ± 23.6	19.7 (8.5 to 30.9)	<.0001	19.6 (1.4 to 37.6)	0.030
	*APA duration (ms)
TMR pre	250.2 ± 76.2
TMR post	237.7 ± 57.1	−12.5 (−63.4 to 88.4)	0.969
ARTI pre	249.8 ± 85.3
ARTI post	221.1 ± 62.4	−28.7 (−39.7 to 97.1)	0.664	−16.6 (−56.4 to 89.8)	0.923

TMR = Traditional Motor Rehabilitation; ARTI = Adapted Resistance Training with Instability; MD = Mean Difference; CI = Confidence Interval; NFOGQ = New Freezing of Gait Questionnaire; FOG-ratio = Freezing of gait ratio; UPDRS-III = Unified Parkinson’s Disease Rating Scale part III motor subscale score; PDQ-39 = Parkinson’s Disease Questionnaire; a.u = Arbitrary units; APA = Anticipatory Postural Adjustment.

^*APA assessed inside scanner in 13 and 16 freezers, who performed TMR and ARTI, respectively.

Outcomes

FOG severity (primary outcome)

The NFOGQ scores showed a significant group × time interaction (F_{[1, 30]}=8.49, P=0.0067). Only the ARTI group decreased the NFOGQ scores (P<.0001, ES=0.80) presenting a trend toward lower scores than the TMR group (P=0.069, ES=2.14) at post-training (Table 3). The FOG-ratio also showed a significant group × time interaction (F_{[1, 30]}=20.67, P<.0001). The TMR group increased the FOG-ratio at post-training (P=0.002, ES=0.92), whereas the ARTI group presented a trend toward lower FOG-ratio at post-training (P=0.078, ES=0.60). Additionally, the ARTI group was effective in decreasing FOG-ratio when compared to TMR group at post-training (P<.0001, ES=1.88) (Table 3).

Cognitive inhibition

The Stroop-III test values showed a significant group × time interaction (F_{[1, 30]}=22.28, P<.0001). Only the ARTI group improved the Stroop-III test values (P<.0001, ES=0.35) at post-training. There was no between-group difference (P>0.05) (Table 3).

Motor signs

The UPDRS-III scores showed a significant group × time interaction (F_{[1, 30]}=15.91, P=0.0004). Only the ARTI group decreased the UPDRS-III scores (P<.0001, ES=0.66), presenting lower scores than the TMR group (P=0.001, ES=1.35) at post-training (Table 3).

Quality of life

The PDQ-39 scores showed a significant group × time interaction (F_{[1, 30]}=17.13, P=0.0003). Only the ARTI group decreased the PDQ-39 scores (P=0.0001, ES=0.54), presenting lower scores than the TMR group (P<.0001, ES=1.78) at post-training (Table 3).

APA amplitude and duration during the leg lifting task

APA amplitude showed a significant group × time interaction (F_{[1, 27]}=16.03, P=0.0004), but APA duration did not (F_{[1, 27]}=0.19, P=0.667). Only the ARTI group (n=16) increased APA amplitude (P<.0001, ES=1.69), reaching larger APA amplitudes (P=0.030, ES=1.29) than the TMR group (n=13) at post-training (Table 3). APA duration did not show significant differences between groups (Table 3).

Group activation map and Beta values during the leg lifting task

Results of the group activation map for the contrast (post>pre) showedhigher BOLD activation for the ARTI group (n=16) than the TMR group (n=13) in the region comprising the middle/inferior temporal gyrus (MITG) of the right hemisphere (0.76±0.37 and −0.02±0.50, respectively) (Figure 2A), as the APA task was lateralized to the left side. Peak coordinates (MITG) are shown in Table 1.

Figure 2.

(A) Group activation map of the adapted resistance training with instability (ARTI) > traditional motor rehabilitation (TMR) for the contrast post > pre showing higher BOLD signal in the right hemisphere for the middle and inferior temporal gyrus (MITG). (B) Beta of BOLD signal change of the MITG, mesencephalic locomotor region (MLR), cerebellar locomotor region (CLR), and supplementary motor area (SMA) in the right hemisphere for the contrast post > pre of each group. Results of 13 and 16 freezers who performed TMR and ARTI, respectively. a.u = Arbitrary unit.

Analysis of the beta of BOLD signal change for the contrast post>pre showed higher values (P<0.05) for the ARTI group than for the TMR group in the MLR (0.70±0.30 and −0.01±0.51, respectively) and CLR (0.61±0.36 and −0.11±0.68, respectively) of the right hemisphere. However, the SMA of the right hemisphere did not show significant changes (Figure 2B). Peak coordinates of these regions are shown in Table 1.

Additional analyses (see Supplementary Table 2) included post-hoc Spearman correlations between the change (post-pre) in UPDRS-III tremor items (20 and 21) and head motion with the change in beta values of the CLR³⁷ and MLR, which did not show any significant correlation (P>0.20). These analyses were performed because tremor amplitude in individuals with PD has previously been associated with increased beta values in the cerebellum,³⁷ and possibly with MLR.

Linear multiple regression approach

This analysis was performed only for the ARTI group due to the significant primary and secondary outcomes. The beta of BOLD signal changes of the MLR was the best predictor of changes in NFOGQ scores (R²=0.56, P=0.0007). Changes in Stroop-III test values entered the model, but it did not reach significance (R²=0.09, P=0.083). The regression analysis identified the changes in APA amplitude (R²=0.51, P=0.001) as the only predictor of changes in FOG-ratio.

Adherence and Adverse events

Adherence to the training protocols was high, as the TMR group performed 35.4±1.3 sessions (98%) and the ARTI group performed 36.0±0.0 sessions (100%). Only one adverse effect was reported during ARTI sessions (lower limb pain while performing lunge exercise). No adverse effect was reported during the TMR classes.

DISCUSSION

We report the results from the first RCT of rehabilitation for freezers using FOG severity (objectively and subjectively assessed), outcomes linked to FOG, as well as brain function. We found significant improvement in FOG severity, APA amplitude, cognitive inhibition, motor signs, quality of life, and brain activation of areas related to FOG and gait during the leg lifting task for the ARTI group, but not for the active control group (TMR). Our results are clinically relevant for freezers because improvement in NFOGQ (−4.4 points) and UPDRS-III scores (−7.4 points) after ARTI were greater than the MDC of 2.7 points from TMR group data (Table 3) and the moderate clinically important difference of 6.7 points suggested for PD,³⁸ respectively. These clinical improvements are supported by improved quality of life only for the ARTI group.

ARTI may be effective in improving FOG severity for four reasons. First, the ARTI program incorporates exercises that require APAs (lunge exercise) and inhibition of inappropriate actions while performing dual-task exercises (simultaneously squatting, alternating arm movements, and resisting loads on unstable surfaces while maintaining balance), which are likely to trigger freezing. These characteristics are a widely recognized ‘behavioural and cognitive signature’ linked to FOG.³⁹ In fact, changes in APA amplitude explained 51% of the improvement in FOG-ratio following ARTI. Although changes in Stroop-III test did not reach significance to predict improvements in NFOGQ, we observed increased cognitive inhibition after ARTI, in contrast to 7 weeks of cognitive training for freezers.⁴⁰

Second, ARTI exercises were performed on unstable devices that require high sensorimotor integration (i.e., proprioception, vestibular, and visual information), while TMR was performed on a stable rigid surface. Thus, ARTI should cause more activation of MITG, which includes visual temporal areas.^{41, 42} Visual temporal areas are important for freezers due to their increased reliance on visual and sensory information to walk,^{43, 44} however, they have an inadequate integration of sensory input, proprioceptive deficits, and visual information during walking.^{45, 46} Thus, increased activation in MITG of the right hemisphere after ARTI (Figure 2A) suggests improved visual sensory information integration in freezers.^41–43

Third, higher activations in the MLR and CLR, which are linked to FOG and gait was observed after ARTI, confirming our hypothesis that improved MLR activation during the APA task was the best predictor of change in NFOGQ scores after training. As abnormal APAs anddysfunction in cognitive inhibition are linked to impairments in MLR,^{8, 10} our results suggest that exercises requiring APA, cognitive inhibition, and motor complexity are effective in increasing MLR activation, thus, decreasing FOG severity and motor symptoms. Our RCT was the first to demonstrate increased MLR activation in freezers after exercise, which explained 56% of the improvement in NFOGQ scores. Our results demonstrate that the increased MLR activation during the APA task after ARTI is critical for coupling posture and gait to decrease freezing.

A fourth reason for ARTI being more effective than TMR in improving FOG severity is the fact that pedunculopontine nucleus has a reduced functional connectivity with the cerebellum and MITG,⁴¹ which are important areas involved in FOG and gait, and presented increased activation after ARTI (Figure 2B). CLR is a structure that regulates automatic posture/gait control.^{47, 48} Lesions causing FOG are located within the CLR¹⁸ and freezers have greater cerebellar grey matter atrophy than non-freezers.⁴⁹ Increase in CLR activation during the APA task after ARTI may indicate restored automaticity in freezers.⁵⁰ Although studies demonstrated that gait⁵¹ and the internally guided lower limb movement⁵² in PD required increased activation in CLR and cerebellum, respectively, maybe due to lack automatic control from the basal ganglia, evidence shows that both pathological and compensatory processes in the cerebellum could play a role in PD gait.^{53, 54} Split-belt treadmill training was used in the attempt to modulate cerebellar input during gait in freezers, but freezers were unable to adapt stride time to compensate for differences in leg speed,⁵⁵ suggesting that freezers may fail to adequately recruit cerebellar networks, which contribute to gait impairment. Our findings demonstrate that ARTI is able to increase activation of this important brain area linked to automaticity and FOG. Regarding SMA, our findings did not show any effect of exercise on SMA activation during simulated APA task (Figure 2B). Freezers have an increased activity of SMA during fMRI protocols of simulated walking tasks with motor arrests,^{11, 56} while another study showed decreased SMA activation in fMRI protocol of simulated walking tasks with motor arrests.¹⁹ Thus, the SMA involvement in impairments of leg movement initiation and walking in freezers require more investigations. Taken together, the four reasons aforementioned can explain the decreased FOG severity following ARTI.

Regarding methodological procedures, freezers were assessed and trained in the “on” state. As levodopa improves FOG,⁵⁷ APAs,⁵⁸ and neural circuits linked to gait automaticity in freezers,⁵⁹ one of the inclusion criteria was a stable drug regimen 2 months before starting the study and also during the study period, which was kept by participants. Moreover, the assessments were performed in the morning after the first levodopa dosage, which was similar between individuals (data not shown). In addition, the APA task was lateralized to the left (support) leg only although, the majority of the participants had both sides affected (Table 2). However, freezers tend to show predominant involvement of right-sided brain circuitry,^{10, 41, 43, 60} which reinforces the importance of the APA task lateralized to the left leg and the improvement of right-sided brain circuitry after ARTI. Finally, to avoid lessebo effect, our groups trained in different facilities, as previously stated. Participants were blinded to the expected outcomes, unaware of the study hypothesis and were explicitly instructed not to discuss their training program with the researchers to keep researchers blinded to group assignment.

The present study has limitations. First, even though a non-intervention control group would help to clarify the actual magnitude of the standard of care of many individuals with PD, we believe that such a group would be unethical (from a research point of view), as individuals with PD usually get worse following the aforementioned standard care.⁶¹ Thus, as ARTI produced greater changes in FOG severity than TMR (active control group), a corollary from our study is that the changes would be even greater when compared to a standard care group (no intervention). Second, even though our study showed robust changes in FOG severity after ARTI, but not after TMR, the total sample size of the present trial was small (32 participants); thus, a larger RCT trial is needed to validate the reported benefits of the ARTI on FOG severity. Third, it was not possible to observe the carryover effects of ARTI on outcomes at follow-up interval after the 12 weeks. Finally, ARTI was an individualized training, while TMR was a group-based training. Findings from a recent meta-analysis suggest that supervised training at facilities produces more long-term gains in gait,⁶² than home. It is also possible that an individualized training approach (one on one) might enhance these gains more than group-based training, but this requires further investigation.

CONCLUSIONS

ARTI is effective in improving FOG severity, outcomes linked to FOG (APA and cognitive inhibition), brain activation, motor signs, and quality of life in freezers. Our results are clinically relevant for freezers because improvement in NFOGQ and UPDRS-III scores exceeded the MDC from TMR group data and the moderate clinically important difference suggested for PD,³⁸ respectively. The changes in MLR activation and APA amplitude explained changes in NFOGQ scores and FOG-ratio following ARTI, respectively. Thus, ARTI is an innovative intervention resulting in clinical improvement and brain plasticity in people with PD who have FOG.