Comparing Backward Walking Performance in Parkinson’s Disease with and without Freezing of Gait—A Systematic Review

Introduction: Parkinson’s disease (PD) is a neurodegenerative disease characterized by motor symptoms and gait impairments. Among them, freezing of gait (FOG) is one of the most disabling manifestations. Backward walking (BW) is an activity of daily life that individuals with PD might find difficult and could cause falls. Recent studies have reported that gait impairments in PD were more pronounced in BW, particularly in people presenting FOG. However, to the best of our knowledge, no systematic review has synthetized the literature which compared BW performance in PD patients with and without FOG. Objective: The aim of this study was to evaluate the differences in BW performance between PD patients with FOG and PD patients without FOG. Methods: Two databases, PubMed and Web of Science, were systematically searched to identify studies comparing BW performance in PD patients with and without FOG. The National Institutes of Health (NIH) tool was used to assess the quality of the studies included. Results: Seven studies with 431 PD patients (179 PD with FOG and 252 PD without FOG) met the inclusion criteria and were included in this review. Among them, 5 studies reported walking speed, 3 studies step length, stride length and lower limb range of motion, 2 studies functional ambulation profile, toe clearance height, swing, and stance percent and 1 study reported the decomposition index and stepping coordination. Compared to PD patients without FOG, PD patients with FOG showed slower walking speed and reduced step length in 3 studies, shorter stride length, lower functional ambulation profile and decreased ankle range of motion in 2 studies, and smaller swing percent, higher stance percent, worse stepping coordination, greater decomposition between movements, and lower toe clearance height in one study. Conclusion: Despite the small number of included studies, the findings of this review suggested that PD patients with FOG have worse gait performance during the BW task than PD without FOG.


Introduction
Parkinson's disease (PD) is a progressive neurodegenerative disorder [1] characterized by motor symptoms such as slow movement, muscle stiffness, and tremor affecting the performance of motor tasks, including walking [2]. Gait impairments are rather common in PD, particularly with disease progression [3] and are associated with an altered quality of life and an increased risk of falls [4]. These include continuous alterations such as reduced step length, low gait speed, increased cadence and variability, and episodic disturbances such as freezing of gait (FOG) [3]. FOG is a debilitating symptom of PD [5], defined as a "brief, episodic absence or marked reduction of forward progression of the feet despite the intention to walk" [6]. It impairs mobility, alters quality of life, leads to falls and is category focused on terms related to backward walking such as "backwards walking" OR "backward walking" OR "backward gait" OR "backward locomotion" OR "backwards locomotion" OR "retrowalking" OR "retro-walking". The third category specified FOG and comprised the following terms: "freezing of gait" OR "fog" OR "frozen gait". For the final search, these 3 keyword's categories were combined as follows: (1) AND (2) AND (3). Search fields were restricted to the abstract, title, and keywords.

Study Selection
Two independent reviewers (TM and MC) screened the titles and abstracts of all studies to identify potentially relevant articles. After the removal of duplicates, selected full-text articles were then reviewed and independently screened for eligibility according to the inclusion criteria mentioned above. If disagreements occurred between the two independent reviewers, consensus will then be achieved through discussions or contact with a third reviewer (NV) to arbitrate the disagreement.

Data Extraction
After the end of the selection process, two independent reviewers (TM and MC) extracted data from all included studies. Information was extracted on (1) the study's characteristics, (2) the sample characteristics, (3) the measure of FOG, (4) the measure of BW, and (5) the main results of the comparison of BW performance in PD patients with and without FOG. Reviewers were not blinded to the authors or journals when extracting data. Any discrepancies between these independent reviewers were resolved at a consensus meeting. If disagreement persisted, a third reviewer (NV) was consulted for a final decision.

Methodological Quality
Two independent reviewers (TM and MC) assessed the quality of the included studies. The National Institutes of Health (NIH) Quality Assessment Tool for Observational Cohort and Cross-sectional Studies (NIH, 2014) was applied to assess the methodological quality of the included studies. It consists of answering yes, no, or other to each of the 14 items. Then, the overall quality of a study was evaluated by assigning a good, fair, or poor rating to each study. Any disagreements between the two independent reviewers were resolved by discussion, with the involvement of a third reviewer (NV) when necessary, who was contacted to arbitrate the disagreement to reach the final rating.

Study Selection
The database search yielded 18 potentially relevant studies. One additional study was identified through hand searching. After removing duplicates (n = 8), 11 studies remained. After the screening of titles and abstracts, 2 studies were excluded, and 9 studies were reviewed for eligibility. After full-text screening, 2 studies were excluded. The remaining 7 studies fulfilled the eligibility criteria and were included in this systematic review [7,9,[17][18][19][20][21]. The study selection process is illustrated in Figure 1.

Methodological Quality
The overall quality assigned to the 7 studies was "fair" according to the NIH tool (Table 1). Table 2 shows general information about the studies included in this systematic review, including authors, year of publication, country, title, journal, and funding sources.

Sample Characteristics
The 7 studies included covered a total of 431 patients with PD, among whom 248 (57%) were male. All studies included both men and women participants.

Studies Population
The 7 studies included covered a total of 179 PD patients (42%) with FOG (PD + FOG) and 252 PD patients (58%) without FOG (PD − FOG). Details about included studies, sample size, demographic, and anthropometric and clinical characteristics are shown in Table 3.

Assessments of Freezing of Gait
Two methods were used to assess the presence and severity of FOG in patients with PD: Freezing of Gait Questionnaire (FOG-Q) [29] and the new Freezing of Gait Questionnaire (NFOG-Q) [30].  Method 1. FOG-Q [29] was used in 3 studies (43%) [9,17,18], including a total of 187 participants, namely 105 PD without FOG (56%) and 82 PD with FOG (44%). The FOG-Q is a self-administrative scale which consists of 6 items, each item scored from 0 to 4 with a maximum score of 24 points. A higher score represents more severe freezing. Four items question the duration and frequency of FOG, and two items question general gait impairment. Participants were considered in the PD + FOG group, if they had a score > 1 on item 3 ("Do you feel that your feet get glued to the floor while walking, making a turn or when trying to initiate walking (freezing)?"), suggesting a frequency of FOG at least once a week [29].
Method 2. NFOG-Q [30] was used in 4 studies (57%) [7,[19][20][21], including a total of 244 participants, namely 147 PD without FOG (60%) and 97 PD with FOG (40%). The NFOG-Q was based on the FOG-Q version. An initial part was added to enable the detection of FOG, and the exclusion of subjects without symptoms, from the actual score of FOG severity and impact. A short video was also added to help in clarifying diverse types and durations of FOG. Participants were considered in the PD + FOG group if they answered "yes" on item 1 ("Did you experience freezing episodes over the past month?") [30]. Table 4 reports description of the BW tasks.

Experimental Procedure
Instructions. In one study (14%) [9], participants were instructed to walk at their normal pace, 3 times in each direction: forward then backward. In one study (14%) [17], participants performed single-task conditions at their normal or comfortable pace 3 times in each direction: forward then backward. Secondly, they performed one trial of a mental arithmetic task with forward then backward. In 2 studies (29%) [18,21] participants were asked to walk backward at a comfortable pace. In 3 studies (43%) [7,19,20], the instructions for the tasks were not reported.
There were no included studies in which participants were asked to perform BW only. In two studies (29%) [9,20] participants were asked to perform two walking tasks: backward and forward walking. In Hackney and Earhart's (2009) study, participants walked forward then backward. The order of presentation of the conditions was not reported by Son et al. (2018).
In one study (14%) [21], participants were asked to perform three walking tasks: BW, forward walking, and forward walking with a dual task (listing as many words as they could think of that start with a certain letter). By condition, trials were grouped into blocks, five trials per block, and the blocks were randomized.  In one study (14%) [17] participants were asked to perform four walking tasks: backward and forward walking in single-and dual-task conditions (performing a mental arithmetic task aloud that consisted of counting backward from 100 by threes, from 50 by fours, and from 75 by sixes). Participants walked forward then backward.
In one study (14%) [18], participants were asked to perform 6 walking tasks: BW, forward walking, turning right and left in a small radius circle and turning right and left in a large radius circle. These 6 walking tasks were performed in random order.
In one study (14%) [7], participants were asked to perform 4 walking tasks: BW, forward walking, and 360-degree turning in both directions. The order of presentation of the conditions was not reported.
In one study (14%) [19], participants were asked to perform 4 walking tasks: BW, forward walking, MDS-UPDRS-III item 12 (for postural responses), and the Mini Balance Evaluation Systems Test items 4 and 5 (for postural responses). The order of presentation of the conditions was not reported. Table 5 describes the data acquisition method and measured parameters during BW.
Two studies (29%) [7,20] reported step time (s) and asymmetry index of step time and length. Two studies (29%) [9,17] reported swing and stance percent, base of support (m), and functional ambulation profile which quantify gait variability. One study (14%) [7] reported stride time (s). One study (14%) [17] reported gait asymmetry. The other study (14%) of Hackney and Earhart reported double support percentage [9]. One study (14%) [18] reported stepping coordination measured as phase coordination index, a temporal gait variable that quantifies the accuracy and consistency of left to right stepping phases by assessing bilateral coordination of gait. It was calculated as the summation of 2 components: consistency of phase generation and temporal accuracy in producing anti-phased stepping through all steps [18]. One last study (14%) [21] reported cycle timing, % gait cycle of the hip, and knee and ankle joints measuring the timing of maximum and minimum angles.
Kinematic outcomes computed during BW in individuals with PD with and without FOG are presented in Table 6.
One study (14%) [21] measured the decomposition index which was calculated to determine percentage of the gait cycle in which a joint remained fixed while another one is moving.

Spatiotemporal BW Parameters
The distribution of the kinematics and spatiotemporal parameters in included studies was reported in Table 6.
Spatiotemporal outcomes during BW in individuals with PD with and without FOG were summarized in Table 5.
In one study (14%) [20], walking speed in PD + FOG was significantly slower by 23% in both the more affected side (MAS) and the less affected side (LAS) compared with PD − FOG (MAS and LAS: 0.46 ± 0.13 and 0.46 ± 0.13, respectively, in PD + FOG; 0.60 ± 0.14 and 0.60 ± 0.14, respectively, in PD − FOG; p < 0.05) [20]. The median walking speed was also lower by 26% in PD + FOG compared with PD − FOG (p < 0.05) in another study [19]. In addition, PD + FOG walked slower than PD − FOG (p < 0.00625) in both single-and dual-task conditions [17]. Figure 2 illustrates the average walking speed measured in PD patients with and without FOG.
One study (14%) [21] measured the decomposition index which was calculated to determine percentage of the gait cycle in which a joint remained fixed while another one is moving.

Spatiotemporal BW Parameters
The distribution of the kinematics and spatiotemporal parameters in included studies was reported in Table 6.
Spatiotemporal outcomes during BW in individuals with PD with and without FOG were summarized in Table 5.

Kinematic BW Parameters
Range of motion. PD + FOG group had significantly decreased ROM in the ankle joint in the most affected side [7,20] [20].

Discussion
To the best of our knowledge, this is the first systematic review summarizing published studies that compared BW performance in PD patients with and without FOG. The main findings of our review suggest that PD + FOG exhibited marked differences in BW compared to PD − FOG. PD + FOG had a slower walk with shorter strides and steps and a lower functional ambulation profile compared with PD − FOG. Furthermore, PD + FOG had worse stepping coordination, reduced swing percent, and greater stance percent. They also had a decreased ROM in the ankle and hip joints, greater movement decomposition and lower toe clearance height than PD − FOG. These findings could be explained by a reduction in proprioception, an impairment in attention and cognitive functions, or an alteration in visuospatial and cerebellar network in PD + FOG patients.
BW requires greater proprioception than forward walking due to the lack of visual control. In PD patients, an alteration of proprioception has been described [31] and, consequently, individuals with PD are more dependent on visual information [32] which is more pronounced in those with FOG [33]. This could explain why PD patients have greater impairments while walking backward [17] and the marked differences between PD with and without FOG [20].
Another factor that could explain our findings is the increased cognitive demand for BW. BW is a complex task, demanding more investment in attention and cognitive resources. As reported previously, individuals with PD might experience cognitive impairment, mainly related to attention and executive functions [34] and these have been closely related to walking and mobility performance [35]. To this end, studies examining the cognitive differences between PD + FOG and PD − FOG have suggested that PD + FOG had worse executive, attentional, and visuospatial performance [36][37][38][39][40][41][42][43]. Furthermore, FOG severity was negatively correlated with performance on cognitive tests assessing executive functions, suggesting that FOG progression is associated with frontal lobe dysfunction [36,43]. This is also supported by neuroimaging studies suggesting a common pattern of grey matter atrophy between executive functions and FOG [44]. Additionally, PD + FOG performed worse than PD − FOG in dual task [40], that is the performance of two tasks simultaneously. Dual task is a common paradigm to test motor-cognitive interaction and a worse performance during dual task has been linked to reduced cognitive resources and lower attention and executive [45].
In PD, the coordination of bilateral stepping during gait quantified by the phase coordination index is reduced, especially in PD + FOG [18,46]. This result, however, is particularly marked during BW [18]. Gait coordination, rhythmicity, and asymmetry have been related to cognitive functions and proprioception; therefore, a more markedly increased gait asymmetry during BW could also be associated to the increase in attentional demands [20]. This is further supported by the finding that PD + FOG demonstrated more gait asymmetry than PD − FOG in BW under dual task conditions (Hackney and Earhart 2010).
This review has reported that during the gait cycle, PD + FOG decompose movement between joints more than PD − FOG [21]. This difference in decomposition indices between PD + FOG and PD − FOG could be related to the cerebellar involvement in FOG [47]. PD + FOG, indeed, have been reported to have an abnormal functional connectivity network of pedunculopontine nucleus affecting the corticopontine-cerebellar pathways and visual temporal circuits [47]. These alterations, in the context of a demanding walking task and increased gait variability (i.e., BW), might lead PD + FOG to decompose their movement patterns to improve their stability [21]. Plotnik et al. (2008) suggested that in patients with PD, when alterations in gait surpass a certain limit, this might trigger FOG. This limit could be regulated by the attention used by PD patients in gait tasks, environmental stressors, and postural stability.
During BW, compared to forward walking, individuals tend to decrease their speed, most likely due to the inability to see the gait direction [10]. Moreover, during BW in healthy participants, range of motion of the hip, knee, and ankle have been reported to be reduced compared to forward walking [10]. The same study considered the ankle joint as the main joint responsible for propulsion and shock absorption during BW [10]. The findings from our review are in line with these results. PD patients, indeed, markedly reduced their hip, knee, and ankle ROM during BW, most likely due to the reduced proprioception, and this is significantly higher in PD + FOG [7,20], supporting the idea that this subgroup of patients have a more marked alteration of proprioceptive inputs.
Finally, the worse performance and higher variability of PD + FOG compared to PD − FOG, mainly evident in stance and swing percent, could be explained, at least partly, by their longer disease duration (10.5 ± 1.00 years vs. 6.4 ± 0.57 years) and greater balance impairment (Berg Balance Scale score: 46.8 ± 0.85 vs. 50.0 ± 0.61) [9].
Taken together, the present findings support the use of BW in PD patients and particularly in the subgroup of patients with FOG. These patients, indeed, have already been reported to be more prone to falls and subsequent injuries [22]. Moreover, PD + FOG have been reported to have more severe disease progression and non-motor symptoms [48] and a higher disability with a lower quality of life [49].
The findings that, in this group of patients, several walking parameters linked to all walking domains (pace, rhythm, variability, asymmetry, and postural control [50] are altered during BW and that these alterations are more pronounced than in forward walking, support the implementation of BW during clinical and instrumented evaluation, as well as considering including this paradigm in mobility research and clinical trials evaluating walking and mobility in patients reporting FOG. In addition, the alterations observed during BW could be less responsive to dopaminergic therapy, thus making this assessment more stable and less influenced by pharmacological state [13].

Limitations and Future Directions
The present study has some limitations. It was difficult to reproduce FOG in the laboratory [7,20]. Thus, the actual number of FOG episodes was unknown because FOG was rarely observed during the intervention [7]. Future studies that can detect FOG episodes in an actual daily environment are needed to understand the effect of FOG under different gait conditions [7]. The results are not generalizable due to the limited number of included studies and the small sample size therein [7,20,21]. Assessments were done during either the OFF-medication state [7,[18][19][20] or the ON-medication state [9,17,21]. Thus, assessments in both states could be useful as FOG episodes mainly occur in the OFF-state and some in the ON-state [51]. There was a high number and heterogeneity of spatiotemporal and kinematic parameters in each article, thus, a strict comparison of results is challenging.

Conclusions
This systematic review summarized the literature about BW performance of PD patients with and without FOG. PD with FOG had worse BW performance than PD without FOG. PD with FOG walked slower with shorter stride and step than PD without FOG. Also, during the gait cycle, PD + FOG decompose joint movement more than PD − FOG. As well, PD + FOG demonstrated greater variability in gait, swing, and stance percent and showed worse stepping coordination. PD with FOG had greater reduction in ROM, mainly at the ankle joint than PD − FOG. This could be due to reduced proprioception and limited attention and cognitive resources from PD patients during a challenging mobility task, or due to altered cerebellar network and visuospatial processing. These results suggest that BW assessment could be relevant to better characterize PD patients and that this paradigm could add valuable information to clinical and instrumented evaluation as well as research and clinical trial. Funding: This research was partly funded by the French National Research Agency, within the framework of the "Investissements d'avenir" program (ANR-10-AIRT-05 and ANR-15-IDEX-02). The sponsors had no involvement in study design, the collection, analysis, and interpretation of data, or in writing the manuscript. This work also forms part of a broader translational and interdisciplinary GaitAlps research program (N.V.).

Institutional Review Board Statement:
Review articles do not require Institutional Review Board (IRB) approval if the data reviewed are public (including private and government databases) and if the articles reviewed have received IRB approval previously.