Freezing of Gait in People with Parkinson’s Disease: Nature,Occurrence,and Risk Factors

Participants and setting

The present study was a secondary analysis of baseline data from the Rivastigmine to Stablise Gait in Parkinson’s disease (ReSPonD) randomized controlled trial conducted in South West England, which investigated the effect of the drug rivastigmine on gait stability [20, 21]. Participants were included if they: a) had been on stable PD medication for 2 weeks prior to enrolment, b) were able to walk 18 m without a walking aid and c) had at least 1 fall in the previous year. A history of freezing was not an exclusion criterion and a small number (not formally captured) were using (or had used) apomorphine and had had deep brain stimulation. People were excluded if they a) had an absolute contradiction to, or had previously taken acetylcholinesterase inhibitors, b) had any other neurological, visual or orthopaedic problems that significantly interfered with gait or c) had dementia. All participants provided informed consent prior to study participation. Ethical approval was obtained from the South West Research Ethics Committee and a Clinical Trial Authorisation (CTA) was granted from the Medicines and the UK Healthcare Regulatory Agency (MHRA).

Gait assessment

The walking trials were conducted on a 22 m level outdoor covered flat concrete walkway (with no regular markings that could provide external cues) with participants assessed in the “on” medication state as indicated subjectively by participants during daytime hours. Participants walked in their normal outdoor footwear and completed three walking trials in three conditions (blocks), the order of which was block randomised. The three conditions comprised: a) walking with no additional cognitive task, b) walking while naming words beginning with a single letter (easy Dual-Task), and c) walking while naming words alternating between two letters of the alphabet (hard Dual-Task). The number of trials was selected to capture the maximum potential data whilst balancing the fatigue and performance abilities of the participants. Participants were given standardised instructions before every walk that indicated they should walk at their usual pace. For the dual task walks the standardised instructions were to commence both tasks on ‘go’ and no prioritisation instructions were given as to walking or the cognitive task. Participants were not instructed to tell the researcher if they were pausing to think about the words. We included the two secondary task conditions, because it has been shown harder secondary cognitive tasks have greater influences on gait than easier secondary cognitive tasks [22] The verbal fluency tasks were based on the controlled oral word association test (COWAT) [23]. Following these nine walks, participants then completed two walks in a specific paradigm designed to elicit FOG. This comprised rising from a chair, walking 9 m, passing between two chairs placed 50 cm apart, continuing for a further 9 m before turning 360° to the right, then 540° to the left and then walking back along the walkway, passing between the chairs, and finally sitting down on the chair (total distance for each walk: 36 m). Figure 1 shows the set-ups for the dual task and FOG elicitation conditions.

OPEN IN VIEWER

Fig.1

The set-ups for the dual task and FOG elicitation conditions.

Gait disruption annotation and classification

Each walk was videotaped (Panasonic HDC-SD800; frame rate of 50 Hz) by a researcher walking behind the participant. FOG was classified on these videos by a researcher trained in the recognition of FOG and the sub-classification used in this study by a Parkinson’s clinician. Custom software was built in MATLAB version 7.12.0 R2011a (Mathworks, Inc., MA, USA) to facilitate the synchronisation of visual recording of the walk with the triaxial accelerometry captured data. Using this software, a researcher could annotate events that occurred during the walking paradigms. Alterations in gait pattern were annotated with respect to type of freeze, location and duration by a researcher. Events were coded according to consistent, mutually exclusive criteria (see Supplementary Table 1) and where they occurred on the trajectory (see Supplementary Table 2). The researcher used both the video and the wearable device data to precisely annotate the initiation and termination of each event such as turning, walking and FOG according to the standardized instructions. The DynaPort Hybrid (McRoberts, Netherlands) system was attached to the waist with an elastic strap between the posterior superior iliac spines, sampling at 100 Hz. Based on the recorded data, the accelerometer settings were +/–2 g and gyroscope +–229 deg/sec.

FOG was defined as an episode where participants were unable to initiate or continue locomotion for no apparent reason. In some trials participants stopped walking during the tasks because they had forgotten the motor or cognitive task required and sought further instructions; these events were termed ‘cognitive stops’. Other gait disruptions such as tripping or near falls were classified as stumbles. Stumbles were timed from the onset of the gait disturbance until the resumption of normal “steady state” walking which often took several recovery steps and/or a pause. For the no-additional task, easy Dual-Task and hard Dual-Task conditions, gait was evaluated over the full length of the 22 m walkway with gait disruptions occurring in the initial and final 2 m labelled as gait initiation and destination hesitation disruptions respectively.

Freeze severity was classified as a) markedly smaller steps/shuffling (mild); b) trembling in place, associated with no forward motion (moderate); and c) total akinesia (severe) as defined by Schaafsma et al. [3]. For events with mixed manifestation (e.g., complete akinesia followed by a few small steps) we documented only a single event using a hierarchy for which complete akinesia was recorded in precedence to trembling in place and trembling in place was recorded in precedence to small steps. Freezes during the FOG elicitation condition were classified according to the activity being undertaken: a) start hesitation/initiation; b) turning; c) walking in the narrow gap between the chairs; d) destination hesitation: approaching the half way point before the turn and the chair at the end of the walk; and e) open space. The definitions of all gait disruptions are provided in Supplementary Table 1.

Additional assessments

Participants were assessed with respect to demographic and disease-related factors including body mass index (BMI), falls in the previous year and self-reported FOG episodes in the month prior to enrolment. PD characteristics were defined by the Unified Parkinson’s Disease Rating Scale part II and III (MDS-UPDRS), Postural Instability/Gait Difficulty-Score (PIGD; derived from UPDRS part III) and Levodopa Equivalent Dose (LED) [24]. Cognition was assessed with the Montreal Cognitive Assessment (MoCA) [25], executive function with the Frontal Assessment Battery (FAB) [26], quality of life (QoL) with the short version of Health Related Quality of Life Questionnaire (EuroQoL’s EQ-5D-5 L) [27], depressive symptoms with the Geriatric Depression Scale (GDS) [28], and falls efficacy with the Iconographical Falls Efficacy Scale (Icon-FES) [29]. The Physiological Profile Assessment (PPA) [30] was used to assess five parameters of physiologic performance as a measure of physiologic fall risk: visual contrast sensitivity (assessed using the Melbourne Edge Test); proprioception (measured using a lower limb-matching task, with errors in degrees recorded using a protractor inscribed on a vertical clear acrylic sheet placed between the legs), quadriceps strength (measured isometrically in the dominant leg with participants seated with the hip and knee flexed 90°), simple reaction time (measured using a light as the stimulus and a finger-press as the response), and postural sway (path length, measured using a sway meter recording displacements of the body at the level of the pelvis with participants standing on a 15 cm thick foam rubber mat with eyes open). Additionally, participants performed the coordinated stability test (COSTAB), an assessment of controlled leaning balance [31]. The MoCA was selected because it provides good assessment of visuospatial and executive function [25]. The PPA, coordinated stability test and the FAB were selected as they have previously been shown to be risk factors for falls in people with Parkinson’s disease [32]. All assessments were performed by one researcher (EH).

Statistical analysis

The sample size was chosen to detect a treatment effect difference of 0.6 standardised (Z score) units for step time variability (the primary outcome measure in the randomised controlled trial) with 80% power and at a two-sided 5% significance [20]. The occurrences of gait interruptions for the walking conditions are presented in tabular and graphical forms. Medical, physical and cognitive measures were contrasted between those who experienced one or more freezes across the walking trials (freezers), and those who experienced no freezes (non-freezers); with Pearson’s Chi-Square tests, Student’s t-tests and Mann-Whitney-U tests as appropriate. Binary logistic regression analysis was performed to identify significant and independent predictors of freezing status, i.e., freezers vs. non-freezers as defined above. The model derived was considered “explanatory” in nature, given the inclusion only of variables that could elucidate why freezes occur. The significance level was set at p < 0.05. Statistical analyses were performed using IBM SPSS Statistics 24 (SPSS Inc., Chicago, IL).

Gait interruptions

Thirteen people were unable to complete all 3 attempts at the simple (no-additional task) walk (12 people completed 2 walks, 1 completed 1 walk). Fourteen people were unable to complete all three attempts at the easy Dual-Task walk (13 people completed 2 attempts, 1 completed 1 walk). Fifteen people were unable to complete all three attempts at the hard Dual-Task walk (14 people completed 2 attempts, 1 completed 1 attempt). Eighteen people were unable to complete both attempts at the FOG elicitation condition (10 people completed 1 attempt, 8 people completed no attempts).

Fifty-four participants (42%) experienced a total of 391 freezes throughout the simple and DT walks and the FOG elicitation condition. Seventy-six participants experienced no freezes (58.5%), 25 (19.2%) experienced between one and three freezes, 17 (13.1%) experienced between 4–10 freezes, eight (6.2%) experienced between 11–20 freezes, 3 (2.3%) experienced between 21 to 30 freezes, and one (0.8%) experienced >30 freezes (Fig. 2).

OPEN IN VIEWER

Fig.2

Number of freezes by participants across the four trial conditions.

Of those who experienced FOG, 44 (81.5%) had multiple episodes. No sudden motor ‘offs’ were witnessed. Twenty-four FOG episodes occurred during the simple walk, 67 during the easy Dual-Task walk, 86 during the hard Dual-Task walk and 214 during the FOG elicitation condition (Fig. 3).

OPEN IN VIEWER

Fig.3

Number of freezes, cognitive stops and stumbles per person across the four trial conditions.

The median (inter-quartile range) duration of freezes were 5.0 s (2.8 – 12.7 s), 3.8 s (2.3–35.2 s), 7.1 s (3.6 – 26.3 s) and 4.9 s (2.3 – 13.4 s) in the simple, easy Dual-Task, hard Dual-Task and FOG elicitation condition walks respectively. Seventy freezes (17.9%) were classified by type as small steps, 188 (48.1%) were trembling legs and 133 (34.0%) were total akinesia. Most freezes in the FOG elicitation condition occurred when turning (66.7%), approaching the destination (11.7%) on initiation of walking (9.4%) (Fig. 4). Only 3.3% of freezes from four participants occurred when walking between the two chairs.

OPEN IN VIEWER

Fig.4

Location of freezing episodes in the freezing of gait elicitation condition.

Other interruptions to gait included 97 cognitive stops experienced by 49 participants and 73 stumbles experienced by 23 participants. No cognitive stops occurred in the simple walk, 20 in the easy DT walk, 50 in the hard DT walk and 27 in the FOG condition walk. Two stumbles occurred during the simple walk, 17 during the easy DT walk, 29 during the hard DT walk and 25 in the FOG elicitation condition. Cognitive stops and stumbles had median (inter-quartile range) durations of 4.5 s (3.6–7.7 s) and 2.4 s (2.0–3.6 s), respectively. No falls occurred during the walking trials.

Freezer versus non-freezer comparisons

Table 1 shows the medical, physical, and psychological measures for the freezers, non-freezers and total sample. Participants who self-reported freezing in the previous month were significantly more likely to freeze during the walking trials. Sensitivity was high (0.85) but specificity was low (0.42) indicating that many of those who reported freezing did not exhibit FOG during the test walks. Compared with non-freezers, the freezers were more likely to have PIGD subtype, longer disease duration, higher UPDRS part II and part III scores and have a higher LED. Freezers also experienced more falls in the previous year and had poorer executive function (lower frontal assessment battery score), increased sway, slower reaction times, poorer coordinated stability, more depressive symptoms, lower QoL and greater fear of falling. Freezers and non-freezers did not differ with respect to age, sex, BMI, vision, quadriceps strength and overall cognitive function (MoCA scores). No significant difference was detected between those who did and did not exhibit cognitive stops for any measures.

Stepwise logistic regression identified PD disease duration, reduced coordinated stability and poor proprioception as independent and significant determinants of freezing status; adjusted odds ratios for these variables are presented in Table 2.

Characteristics of gait disruptions

Across the standard walks, the incidence of freezing increased with increasing complexity of cognitive dual-tasks. Of those who experienced freezing during walking, the majority had multiple episodes and notably a quarter froze on 10 or more occasions. The number of freezes during the FOG condition was more than twice the number that occurred during hard DT walks, despite the total distance walked being only marginally longer: 72 m vs. 66 m respectively. This suggests that the execution of more complex motor activity (gait initiation, gait termination and turning) induces freezing episodes more frequently than the execution of a secondary cognitive task and supports the hypothesis that both automatic and controlled processes are more severely impaired in freezers [33, 34]. The relatively small number of freezes that occurred in narrow quarters between chairs contrasts with previous findings [3]. This might have resulted from differing definitions of freeze circumstance whereby previous studies have reported freezes that occur before, rather than between constrictions [3]. It may also be that two chairs positioned closely together to form a gap have less visual impact than larger frames such as doorways. Trembling legs was documented as the most common type of freezing, a finding consistent with previous studies that have used accelerometers to categorise FOG episodes [16, 35]. However, we acknowledge the relative occurrence of trembling leg and akinetic FOG episodes is unclear due to difficulties in reliably detecting akinetic gait with accelerometers and the potential misclassifications of akinetic FOG episodes and cognitive stops.

Factors associated with FOG

Freezers were significantly more likely to experience cognitive stops and stumbles in the walking trials, which emphasizes their instability and associated elevated risk of falls [5, 36]. Many measures related to PD disease severity (duration of disease, UPDRS II and III scores, LEDs) were identified as risk factors for freezing in bivariate analyses. Self-reported freezing episodes and PD disease severity (as measured with UPDRS that includes freezing as a component), had only moderate predictive value for freezing events in the walking trials. The low specificity of the FOG questionnaire item for freezing episodes most likely relates to participants being assessed in the “on” state when their balance and mobility was optimal, whereas the FOG questionnaire item reflects a one-month time period that would have included end of levodopa dose motor fluctuations. Thus, our test paradigm most likely identifies those more severely afflicted with this phenomenon.

Several physical measures were associated with objectively measured freezing. In the physical profile assessment, which was developed to determine fall risk in older people [30], freezers performed significantly worse in the reaction time and sway tests and as a consequence, had higher composite fall risk scores. Freezers also had poorer controlled leaning balance, and this measure, along with poor proprioception and PD disease duration, were identified as significant and independent risk factors for freezing in the multivariate logistic regression analysis. The controlled leaning balance test used here requires continuous balance adjustments, sustained attention, visual processing and is dependent on processing speed as well as lower limb strength and compact sway [33]. It thus represents a composite measure that involves both automaticity and higher level functioning [36]. The findings complement studies addressing gait in PD freezers that show cognitive load impacts planning of the final steps needed to avoid an obstacle, shifting from more automatic to more consciously controlled balance [37], and that freezing of gait is associated with asymmetric, less rhythmic and uncoordinated gait patterns [38].

The significant association between freezing and impaired proprioceptive processing is consistent with previous studies that proprioceptive loss is associated with impaired motor performance in people with PD [39] and that people with PD who freeze demonstrate an inability to adapt to changes in normal sensory availability [40]. Enhancing proprioceptive input may offer a feasible strategy to address FOG. Pereirra et al. showed that vibration stimulation applied to the triceps surae tendon of the lower affected leg could reduce the severity of FOG episodes [41]. This effect may be mediated via enhanced proprioceptive information processing insofar as steps were more frequently reinitiated following a FOG episode in the contralateral limb.

Freezers did not differ to non-freezers with respect to overall cognition assessed with the MoCA but had significantly lower FAB scores indicating executive dysfunction. Fasano et al. reported in their neuroimaging review that freezers had significant grey matter atrophy in multiple brain areas, especially in the frontal gyrus/supplementary motor area [42] and the link between freezing and executive function has been previously demonstrated [33, 37]. Similar to previous work in this area, freezing was associated with more depressive symptoms, greater fear of falling and lower QoL [4, 39]. The multiple associations uncovered lend further support to the notion that FOG is a multidimensional phenomenon that involves complex motor and cognitive impairments and interactions directly and indirectly related to the neuronal disease [4, 40].

Clinical implications

Our findings have clinical implications in that the sensorimotor, balance and cognitive impairments found to be associated with freezing. These impairments may be ameliorated by current or emergent behavioural and / or rehabilitative strategies and their potential to reduce the occurrence of freezing warrants investigation. These could comprise exercise to improve controlled leaning balance, step initiation and inhibition, proprioceptive cueing as well as counselling programs that emphasise greater concentration in the performance of motor tasks [43].

Strengths and limitations

The main strength of this study is the volume of captured gait interruptions that included over 550 freezes, cognitive stops and stumbles from a series of walks that varied in cognitive and motor complexity. We used definitions and terminology similar to that described in recent study that also used video recording to document FOG episodes [44]. One difference relates to the commencement of FOG where Gilat defined FOG starting when the foot no longer takes an effective step, whereas we took the last heel strike of the last full step as the FOG commencement point as it is easier to precisely identify. A further strength is the use of the gold standard method of video recordings with ancillary reference to synchronised accelerometry data to classify freezing episodes [19, 45]. Thus, given the assessments included cognitive and motor factors known to induce FOG episodes, the incidence and nature of the freezes and other gait disturbances observed should generalise well to those experienced in everyday life. Further research contrasting the determinants of objectively measured and self-reported FOG as well as the determinants FOG circumstances and FOG severity would build on our findings [19].

Our study also has limitations. First, it lacks an inter-rater reliability assessment of the gait disruption categorization, and whilst the rater was trained and strict definitions were adhered to, misclassification of some freezes could, however, have occurred. For example, we may have misclassified a gait interruption as a freeze if the patient did not verbally articulate a difficulty. Second, the sample comprised people with PD who had suffered at least one fall in the previous year, so the findings may not generalise to people with PD who do not fall. Third, participants were tested in a practically defined motoric “on” state, but the average time after medication intake was not recorded. Finally, we acknowledge that the cross-sectional nature of this study means that it is not possible to establish causal relationships between predictor and outcome variables.

Conclusions

This video annotation study captured over 550 gait interruptions and documented the frequency and duration of freezes, cognitive stops and stumbles across walks of variable difficulty. A specific FOG elicitation condition elicited freezes far more frequently than a standard walk whilst performing a concurrent cognitive task. The multiple motor and cognitive factors identified as being associated with freezing, including the inclusion of poor proprioception and impaired controlled leaning balance in the logistic regression model provide new insights into this debilitating PD symptom and may contribute to potential new targets for rehabilitation.