Freezing of gait in Parkinson’s disease: pathophysiology, risk factors and treatments

Two studies using the Parkinson’s Progression Markers Initiative (PPMI) data included 390 (393) patients with newly diagnosed PD at baseline [16, 17]. During a median follow-up of 4.0 years, male sex was found to be an independent risk factor of FOG (Hazard ratio (HR) = 1.512, P = 0.046) [16]. This finding is consistent with another study that included 41 patients with a mean follow-up disease duration of 9.51 ± 3.18 years [25]. Of these 41 PD patients, 15 patients developed FOG at the end of follow-up, in which 14 patients were males. Besides these follow-up studies, a previous cross-section study that analyzed 6620 PD patients also found that males were more likely to have FOG than females (Odds ratio (OR) = 1.19, P = 0.011) [27]. A number of studies have reported the existence of gender difference on specific motor or non-motor symptoms in PD. Estrogen may play a protective role in PD by influencing dopamine synthesis and release or modulating dopamine receptor expression and function [28]. However, this finding was not observed in some other studies [18‐24, 26].

Zhang et al’s study followed 248 early PD patients without FOG at baseline for 3 years and found that patients with a lower education level (≤9 years of education) were more prone to have FOG (OR = 0.012, P < 0.001) [22]. This study provided an explanation that patient with a higher education had a better understanding of PD and better compliance, leading to better treatments. However, this finding was not found in other three follow-up studies [19, 23, 24]. Although cognitive impairment is involved in the pathogenesis of FOG, whether level of education can be used as an indicator of cognitive reserve is still under debate [29].

Eight studies recorded the age of onset and showed that there was no relationship between the age of onset and FOG development [16, 17, 20‐23, 25, 26]. Seven studies recorded age at baseline [18, 19, 21‐25], in which only one study showed that older age was a risk factor of FOG, but the statistical significance was weak (P = 0.054) [18]. Two previous cross-sectional surveys including 6620 and 683 PD patients respectively also found that age was not the predictor of FOG in PD [27, 30].

It is well recognized that the duration of the disease is a crucial risk factor of FOG. FOG tends to occur in the later stages of PD. Early onset FOG should be suspected of other parkinsonism, such as progressive supranuclear palsy [31]. In a retrospective study of 800 patients with early PD, only 7.1% of patients experienced FOG [20]. However, with prolonged disease progression and disease duration, FOG can affect 53% of patients in the advanced stages of the disease [32]. A previous prospective study [20] and several cross-sectional studies [27, 33, 34] also found that patients with longer disease duration were more likely to develop FOG episodes. However, other studies failed to identify disease duration as a risk factor of FOG [16‐19, 21]. The reason might be that the baseline disease duration instead of the whole disease duration till the end of follow-up were analyzed between non-freezers and transitional freezers, and most patients included were early PD patients who had relatively short disease duration at baseline.

The Freezing of Gait-Questionnaire (FOGQ) total [18](OR = 2.34, P = 0.01), gait speed (on state) assessed by wearable sensors (OR = 0.01, P = 0.032) [19], lower limbs as the onset site (OR = 2.632, P = 0.013) [20, 23] were found to be risk factors of FOG, respectively. In addition, several studies found that Movement Disorder Society-Sponsored Revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) motor score especially postural instability and gait difficulty (PIGD) score could predict FOG development [16, 17, 21]. These findings indicated that the gait impairment had already existed even patients didn’t present freezing episodes at the early stage of disease. This can be explained by the pathophysiology of gait control, which has been reviewed previously [8, 10, 35]. Lesions in any part of gait control system may cause abnormal gait performance. When gait impairment involved the key regions particularly important for gait, such as PPN, or when compensation was not achieved with disease progression, patients would suffer from FOG.

Postural instability and gait difficulty (PIGD) was identified as a risk factor of FOG in PD patients [16, 17, 21]. PD patients with PIGD not only showed more severe striatal dopaminergic terminal loss, but also presented with extrastriatal non-dopaminergic denervation compared to patients without FOG [36]. Additionally, PIGD symptoms were apparently associated with grey matter atrophy in motor-related regions and decreased functional connectivity [37]. Cognitive impairment was also significantly correlated with PIGD phenotype [38]. These mechanisms might all contribute to FOG development. However, another study did not show the difference in PIGD phenotype between non-freezers and transitional freezers at baseline [19]. If motor phenotype was classified as tremor dominant and akinetic rigid phenotype, results were controversial [22, 25]. In Zhang et al’s study, patients with akinetic-rigid phenotype were more likely to suffer from FOG 3 years later (OR = 4.881, P = 0.024) [22]. However, Djaldetti R et al’s study found no difference of motor phenotype between non-freezers and transitional freezers at baseline [25]. Ehgoetz et al. classified motor phenotype as tremor-dominant (TD) or non-TD phenotype, and didn’t reveal that the motor phenotype was a risk factor of FOG [18].

Motor fluctuation was found to be a risk factor of FOG in Forsaa’s study. This study included 232 PD patients without FOG at baseline and found that non-freezers who presented with motor fluctuations at study entry were at more than 3-fold increased risk to develop incident FOG during the 12-year follow-up period as compared with patients without motor fluctuations at baseline [21]. Another study included 225 PD patients and did not reveal didn’t reveal motor fluctuation as a risk factor of FOG, but found that frequency of motor fluctuation at baseline was much higher in transitional freezers than non-freezers (20% vs. 10%, P = 0.056) [23]. Motor fluctuations are supposed to be more closely related to the severity of dopamine depletion in the basal ganglia [39], which can partly explain the phenomenon that motor fluctuation may precede the onset of FOG.

Postural control and balance [40, 41] were affected more significantly in PD patients with FOG compared with non-FOG group. FOG often occurs when turning also suggests that the postural control impairment probably contributes to freezing. However, whether balance impairment is an accompanying symptom of FOG or a risk factor of FOG remains unclear. In Giladi N et al’s study, balance was assessed by two UPDRS subitems, and they found balance impairment predicted FOG development (HR = 1.94, P = 0.0005) [20]. However, Herman T et al. using Berg balance scale and Activities-specific balance confidence scale to assess balance and fear of falling failed to confirm balance impairment as a risk factor in logistic regression model, although transitional freezers had significantly worse Berg balance scale score than non-freezers at baseline [19]. For festination and falls, only one study reported that presence of festination or falls at baseline was risk factors of FOG [23]. The potential mechanism underlying the relationship between gait festination and freezing in PD has been reviewed and discussed in previous reviews [42, 43]. Briefly, festination is hypothesized to be caused by the progressive delays in the timing of phasic motor cues from the internal globus pallidus to the SMA and PM. Thus the step becomes shorter and shorter. When the defect reaches the limit presenting with extremely slow or absence of the phasic cues, the motor cortical regions are not provided with phasic cues necessary to enable them to generate force for the next step in sequence, leading to the subsequent motor block-FOG.

Cognitive impairment as an independent risk factor of FOG has been found by several studies [16, 17, 22]. The cognitive function domains with predictive value for FOG include baseline processing speed, learning ability [24] and visuospatial/executive abilities [23]. The involvement of cognitive impairment in FOG pathophysiology has been explained by previously proposed cognitive model [11, 14]. This model emphasized conflict-resolution deficit in executive dysfunction. Under normal condition, people will prevent premature action and delay the response selection until resolving the conflict. In contrast, when patients fail to process response conflict, impose faster response decision but with greater incongruence, FOG is triggered. Additionally, the beneficial effect of cognitive training for FOG also reflected that the pathophysiological mechanism of FOG linked to cognitive dysfunction [44]. Future studies would better enroll the PD patients without FOG at baseline, and test whether cognitive training can prevent or delay FOG development. It is worth noting that although much evidence supported cognitive impairment was a risk factor of FOG, there were still several studies didn’t confirm this finding [18‐21].

One study including 57 PD patients without FOG at baseline showed that depression was a strong risk factor for FOG development (OR = 10.93, P = 0.003). Eighty percent of the subjects who had marked depressive symptoms at baseline (Geriatric Depression Scale≥5) developed FOG at mean 5 years’ follow-up. In contrast, only 27% of those with few depressive symptoms at baseline became freezers (p < 0.001) [19]. A longitudinal study recruited 221 PD patients with over 1 year follow-up found that the anxiety score and FOGQ total score were the strongest predictors, and these two factors could significantly predict FOG in the next 15 months with 82% accuracy [18]. Ehgoetz et al. used the virtual reality-based functional magnetic resonance imaging approach also confirmed the association of limbic network with FOG. They found that during freezing, the coupling between limbic network (cortical and subcortical) and ventral striatum, and the coupling between limbic and cognitive control network were increased compared to normal foot tapping. In contrast, anti-coupling between the cortical and subcortical limbic network and within the subcortical network was found during freezing [45]. However, other studies did not reveal depression or anxiety as a risk factor of FOG, but found the depression or anxiety score was significantly higher at baseline in transitional freezers than non-freezers [16, 23].

Kim et al’s study using PPMI database found that REM Sleep Behavior Disorder (RBD) screening questionnaire (RBDSQ) score was higher in transitional freezers compared with non-freezers, and univariable analysis also confirmed that RBD may increase the risk for FOG development. However, multivariable analysis in the same study failed to establish its risk factor effect [16]. Although gait disorders have been found in polysomnography-confirmed RBD patients [46, 47] and it was hypothesized that FOG and RBD might share a common physiopathology [48], the relationship between RBD and FOG was not confirmed [16, 18, 24]. In other two studies, RBDSQ at baseline didn’t show significant difference between transitional freezers and non-freezers [18, 24], which was supported by a previous finding that there was no freezing difference between the patients with probable RBD and without RBD [49]. Other studies explored the relationship of insomnia or daytime sleepiness with FOG. One study demonstrated that insomnia assessed by Hamilton Depression Rating Scale (OR = 2.418, P = 0.036) was an independent risk factor for FOG development [22]. Three studies evaluated daytime sleepiness, but the results were inconsistent [16, 23, 24]. In Bank et al’ s study, daytime sleepiness assessed by Epworth Sleepiness Scale (ESS) was identified as a risk factor of future FOG [24]. However, Kim et al’ s study failed to establish its risk factor effect, although they found that ESS score was higher in transitional freezers compared with non-freezers, and ESS may increase the risk for FOG development in univariable analysis [16]. Another study didn’t find the difference of daytime sleepiness between non-freezers and transitional freezers at baseline [23].

Speech disturbance [50] was more frequent in patients with FOG. The DATATOP (Deprenyl and tocopherol antioxidative therapy of parkinsonism) study also found that patients with speech problems at baseline were more likely to develop FOG at the end of follow-up [20]. Another study evaluated hallucination using UPDRS part I subitem and found that hallucination was a risk factor of FOG [23]. These findings need to be confirmed by future studies and which type of speech disorder and hallucination related to FOG needs to be evaluated.

Several studies have found that presynaptic striatal dopaminergic depletion examined by dopamine transporter (DAT) uptake predicted the later development of FOG in de novo PD [16, 17, 25]. Caudate DAT uptake was more significantly related to FOG development compared with putamen [16]. The combined model integrating the caudate DAT uptake, PIGD score, and cerebrospinal fluid (CSF) Aβ42 together can predict FOG within 4 years after diagnosis of PD (area under the curve 0.755, 95% CI 0.700–0.810). These findings demonstrated a direct association between the initial presynaptic striatal dopamine loss and the subsequent development of FOG in patients with early PD.

Recently one study retrospectively reviewed medical records of 268 patients with de novo PD (follow-up > 3 years), and evaluated the longitudinal effects of WMH on the development of FOG using cox regression model [26]. Results showed that after adjusting for age, sex, striatal DAT availability and levodopa equivalent dose, the PD group with moderate to severe WMH showed a higher risk of developing FOG (HR = 3.29, P < 0.001) than the PD patients with minimal WMH. This study was consistent with previous studies which showed that increased WMH burden was associated with severe motor deficits, especially axial motor impairments [51, 52]. Although the mechanisms linking WMH and motor disability are not fully understood, diffuse white matter damage including major cortico-cortical, motor-related cortico-fugal and several striato-frontal tracts are involved in the gait control network, which may contribute to FOG development.

One recent study [16] using the PPMI data included 393 newly diagnosed PD patients without FOG at baseline aimed to explore whether CSF biomarker changed prior to FOG occurrence. Beta-amyloid 1-42 (Aβ42), α-synuclein, total tau, phosphorylated tau, and the calculated ratio of Aβ42 to total tau in CSF at baseline was evaluated. Cox proportional-hazards regression analysis was performed to identify factors predictive of FOG. They found that in multivariable Cox analysis, only Aβ42 of those CSF biomarkers was associated with the development of FOG (HR = 0.997, P = 0.009). This study proposed the possible mechanisms underlying Aβ and FOG. Firstly, Aβ plaque influences the neural circuitry associated with FOG. Secondly, Aβ may promote other proteins such as α-synuclein to misfold and accelerate disease progression; Thirdly, Aβ pathology is related to cognitive dysfunction, and cognitive dysfunction is a contributor for FOG. Those explanations were made based on brain Aβ plaque pathology. Although Aβ in CSF inversely correlated with Aβ plaque formation in brain reported by Fagan et al. [53], there is lack of direct evidence of brain Aβ pathology and FOG in this study.

Robust evidence supports that dopamine replacement therapy with levodopa is the first choice for FOG treatment in PD [55]. Levodopa can significantly decrease the frequency and number of FOG episodes [3, 59]. FOG is more common and more prolonged during the off-state than during the on-state [3, 30, 96], and levodopa can reduce both off time and FOG severity [3]. To be note, the “threshold” to improve FOG is higher than the threshold to improve other motor signs in some patients, thus increased levodopa dosage without deterioration of other motor symptoms may potentially be effective [97]. However, there still exist numerous patients whose FOG is resistant to levodopa [30]. Continuous intra-jejunal infusion of LCIG is an effective advanced therapy for the treatment of motor fluctuations in PD [98, 99]. LCIG is administered via percutaneous endoscopic gastrostomy with a jejunal extension tube. The suspension form of carbidopa and levodopa can be delivered continuously into the intestine and plasma levodopa concentration can maintain stable. One retrospective study including 65 advanced PD showed that with LCIG treatment, FOG presented in 22% of patients at 1-year follow-up compared to 46% at the baseline [60]. Another retrospective study including 91 advanced PD with 18 ± 8.4 months’ follow-up also showed that gait disorders (freezing, festination, postural instability) improved in 61.4% of patients [61]. These results were supported by other clinical trials [62, 63]. Besides, long-term effectiveness of LCIG on FOG has also been demonstrated by two prospective studies. Sensi M et al. evaluated the long-term outcome in 28 PD patients, in which 17 patients reached the 24-month follow-up. Results showed that FOGQ at 6 months (P = 0.001) and 2 years (P = 0.03) were significantly lower than the pre-treatment condition [64]. Vijiaratnam N et al. also found that FOGQ improved 24% (11.9 ± 5.5 vs. 9.1 ± 5.3, P = 0.007) at 6 months in 25 PD patients [65]. Possible mechanism is that increasing “On” time by LCIG may lead to benefits on FOG when freezing episodes happen during the “Off” time. Importantly, two pilot studies found that LCIG can also improve levodopa-resistance FOG [56, 66]. One study including 5 PD with levodopa-resistance FOG who were treated with 24 h LCIG therapy for 6 months, and results showed median 360° turn time, fall frequency and FOGQ improved [66]. The other was a single section study including 7 PD with FOG. Evaluations were performed in “On” state either during oral levodopa (60–90 min after intake of usual morning levodopa dose) or LCIG (60–90 min after starting LCIG infusion). UPDRS item 14 freezing significantly improved with LCIG in all patients compared to oral levodopa (P = 0.026). FOGQ also significantly improved with LCIG (oral levodopa:19.1 ± 1.4, LCIG:10.4 ± 1.6, P = 0.017) [56]. There are several potential explanations for the beneficial effect of LCIG on levodopa-resistant FOG. First, LCIG is administered via the jejunum, which provides greater bioavailability of levodopa compared with oral administration. This may enable better absorption and more levodopa in the brain than administration of levodopa via the oral route. Second, continuous delivery of LCIG provides stable plasma levodopa concentrations and avoids the pulsatile dopamine stimulation in striatum by oral levodopa. Additionally, LCIG can also be delivered at night. Adding nocturnal LCIG infusion improves sleep and better sleep quality improves daytime motor functions including FOG [56, 66].

An open-label uncontrolled study evaluated transdermal patch rotigotine as monotherapy in untreated PD patients for 6 months and found that transdermal patch rotigotine improved all aspects of gait compared to baseline, including straight walking, gait initiation and turning [100]. Studies of DA as an adjunct to levodopa also found beneficial effects on gait speed in add-on therapy of apomorphine (sublingual) and pramipexole compared to levodopa alone [101, 102]. An open label study found 3 months pramipexole treatment as an add-on to levodopa or single administration, FOGQ was significantly improved from 5.4 ± 5.3 to 4.2 ± 4.1in 36 PD patients [67]. Another open label study compared effects of transdermal patch of rotigotine, pramipexole LA and ropinirole CR on FOG, and showed that transdermal patch of rotigotine treatment for 2 months significantly improve FOG [68]. In this study, 47 patients (24 patients with FOG, 51%) were treated with rotigotine (maintenance doses of 9-27 mg/day), 33 patients (16 patients with FOG, 48%) received pramipexole LA (1.5-4.5 mg/day) and 31 patients (14 patients with FOG, 45%) received ropinirole CR (8-16 mg/day). The FOGQ score was recorded during off time in patients with wearing off. FOGQ score were significantly decreased from 30.1 ± 1.8 at baseline to 21.0 ± 1.2 after 2 months rotigotine treatment, whereas pramipexole LA and ropinirole treatment did not alter FOGQ score compared to baseline. Such amelioration of FOG did not depend on reduction of off time after treatment, as off time improvements were similar among these treatments. One possibility is that rotigotine has higher binding affinity to D1/D5 receptors compared to pramipexole and ropinirole. The distinct pharmacological profile and 24-h steady hemodynamics of rotigotine might play a role in the therapeutic mechanism of FOG in PD patients with wearing off. Although subcutaneous continuous apomorphine infusion providing continuous dopaminergic stimulation similar to LCIG, the results of a clinical trial were disappointing [69]. Both the rotigotine transdermal patch and continuous apomorphine infusion are more convenient and reversible than LCIG; therefore, future research is needed to clarify whether they have beneficial effects on refractory FOG.

Clinical trials have shown that selegiline was associated with reduced risk of future FOG [20, 103]. Lijima et al. used wearable sensors to record the daily walking profiles of 14 PD patients before and after selegiline add-on therapy, and used FOGQ to evaluate their FOG. Results revealed that FOGQ and UPDRSIII significantly improved after 3 months selegiline add-on or selegiline dose-increasing therapy (FOGQ, 9.2 ± 4.9 vs 7.9 ± 5.1, P = 0.01; UPDRSIII 18.1 ± 9.0 vs. 13.1 ± 10.0, P < 0.005). Hypokinesia index was also achieved by a small device that attached to the waist of the patients that measured three dimensionally the accelerations accompanied by limb and trunk movements and accelerations during gait. The results showed that improvements in hypokinesia and gait disorders did not occur in parallel. They suggested the beneficial effect was due to selegiline increased the synthesis of phenylamine, which might facilitate the release of noradrenaline. Hypokinesia of limb and trunk movements were not improved in this study indicating the beneficial effect exerted by selegiline add-on therapy was not by dopaminergic pathway [70]. Rasagiline, as an effective monotherapy in early PD and add-on therapy in PD patients with motor fluctuations [71, 104‐106], has also been studied on FOG. In the LARGO study [71], a prospective, double-blind, randomized, placebo-controlled study enrolled 687 PD patients, in which 278 patients had FOG. Oral rasagiline (1 mg once daily) provided a significant improvement in UPDRS-PIGD (P = 0.034) and UPDRS-freezing scores (P = 0.045). In addition, in an open-label, multicenter study, 42 patients with FOG were treated with 1 mg rasagiline daily as an add-on therapy, and FOGQ was assessed as primary outcome [72]. Results showed that patients treated with rasagiline had a statistically significant decrease in FOGQ score after 1, 2 and 3 months of therapy. However, the proportion of levodopa-unresponsive FOG in these studies was unclear. Only one study investigated the effect of MAO-B inhibitor on levodopa unresponsive FOG. This study found that 6 of 14 PD patients with intractable FOG showed reduction in FOG number and duration after a 90-day course of 1 mg daily rasagiline add-on therapy [73].

MPH influences both the dopaminergic and norepinephrine systems and thus is supposed to play a positive role in FOG treatment. One multicenter, parallel, double-blind, placebo-controlled, randomized study enrolled 69 PD patients who had severe gait disorders and FOG despite receiving an optimized, stable dose of levodopa and subthalamic nucleus (STN) stimulation. The results showed that MPH (1 mg/kg per day) treatment for 90 days was effective on levodopa-resistant FOG. In this study, FOG was assessed by FOGQ and the number of freezing episodes while taking walking trajectory [74]. Another study including17 advanced PD patients who presented gait disorders, despite their use of optimal dopaminergic doses and STN stimulation also showed that a daily dose of 1 mg/kg of MPH for 3 months significantly improved gait in the absence of levodopa as assessed by the walking speed, the number of steps and the number of freezing episodes [75]. Besides, one study tested the immediate effect of low dose MPH on five advanced PD and found the beneficial effect of MPH on FOG. The patients were tested during “off” state before and 2 h after the intake of 10 mg MPH while walking an “8 trajectory”. The total walking time, total freezing time, number of freezing episodes and the non-freezing walking time were assessed and all these parameters were improved [76]. But the result of another 6-month placebo-controlled, double-blind, crossover study was disappointing [77]. Twenty-seven subjects with PD and moderate gait impairment were randomly assigned to MPH (maximum, up to 80 mg/day) or placebo for 12 weeks and crossed over after a 3-week washout. Results showed MPH did not improve FOG as assessed by FOGQ and the freezing diary. Further well-designed studies were needed to explore the effectiveness of MPH on FOG.

Caffeine, which is an adenosine receptor antagonist, has been shown to be effective for the treatment of FOG [107]. Istradefylline is a selective adenosine A2A receptor antagonist, which has also been found to be useful in reducing “off” time in PD [108], thus be tested for FOG therapy. Fourteen PD patients were treated with 20 mg of Istradefylline in the morning. FOGQ scores were significantly decreased 1 month after Istradefylline treatment (9.79 ± 7.16) when compared to those before the intervention (12.14 ± 5.82, P = 0.030) [78]. Recently, Mutsumi Iijima et al. conducted a multicenter, open-label, prospective interventional study which evaluated changes in total gait-related scores of the MDS-UPDRS II/III, FOGQ and NFOGQ in 31 PD patients treated with istradefylline. Patients with advanced PD who were treated with levodopa preparations and had wearing-off symptoms and gait disorders complicated with FOG were enrolled. Istradefylline was orally administrated for 12 weeks at 20 mg/day for the first 4 weeks, followed by 20 mg/day or an increased dose of 40 mg/day for 8 weeks. Istradefylline significantly improved MDS-UPDRS Part III (ON-state) gait-related items total score after 12 weeks treatment compared with baseline (− 1.1 ± 2.0, P = 0.007), FOGQ (− 2.0 ± 3.1, P = 0.002), and NFOGQ (− 2.2 ± 5.1, P = 0.026) [79]. Istradefylline showed promising in treating FOG, but further studies involving larger patient numbers and randomized, controlled design are needed.

Depression is found to be a risk factor for FOG development and limbic system has been proven to be involved in FOG pathophysiology, which were discussed previously. However, only one study has studied the effect of antidepressants on FOG therapy. This prospective, multicenter, open-label, randomized study assigned PD patients with mild to severe depression into selective serotonin reuptake inhibitors (SSRIs) (paroxetine or escitalopram, 27 enrolled, 25 analyzed) or serotonin and norepinephrine reuptake inhibitors (SNRIs) (duloxetine, 28 enrolled, 27 analyzed) group. The mean changes (from baseline) of FOGQ scores at 6 and 10 weeks were statistically significant in both the SSRI (6 weeks, − 2.2, P = 0.018; 10 weeks, − 2.9, P = 0.012) and SNRI (6 weeks, − 2.1, P = 0.047; 10 weeks, − 3.4, P = 0.010) groups [80], indicating that FOG was significantly improved. No significant differences were observed between the SSRI and SNRI groups. Depression symptom was also improved. Further studies are needed to evaluate the effects of antidepressants on FOG.

One open label, controlled study randomized 16 PD patients with FOG into L-DOPS co-administered with entacapone group (n = 6), entacapone alone group (n = 5), and L-DOPS alone group (n = 5) in addition to their original medication. Results showed that the combination of L-DOPS and entacapone as an add-on therapy significantly improved FOG based on the visual analogue scale at 4 weeks compared with baseline while L-DOPS alone did not. The improvement was observed only in levodopa-resistant FOG patients, suggesting that dysfunction of noradrenergic neurons plays a significant role in levodopa-resistant FOG [81]. A prospective, open label, noncontrolled study including 13 PD with FOG, found that L-DOPS initially 100 mg/day with a weekly increase of 100 mg up to 600-900 mg/day maintenance, FOG improved in half of patients [82].

A retrospective study reviewed PD patients who received amantadine specifically for FOG and found that 10 of 11 patients reported subjective improvement in FOG after initiating amantadine. Median amantadine dosage was 100 mg twice daily, and treatment duration was 20 months (range 6-66 months). Four patients reported reduction in benefit after 4 months, indicating that this effect may be transient [83]. Although the mechanism of amantadine in PD is unclear, dopaminergic and non-dopaminergic mechanisms are both involved. Additionally, the effect of intravenous amantadine on FOG has been explored, and the results were inconsistent [84‐86]. In Kim et al’s open label study, 6 PD patients with intractable freezing was administrated with intravenous amantadine (200 mg in 500 cm³of saline solution given over a 3-h period), twice a day for 2 days along with the pre-existing anti-parkinsonism medication. Results showed a marked improvement on FOG (FOGQ score ≤ − 4) in 5 of 6 PD patients [85]. However, two double-blind, randomized, placebo-controlled studies including 42 PD and 10 PD patients with FOG respectively found that intravenous amantadine had no beneficial effect on FOG [84, 86].

Atomoxetine is a selective norepinephrine reuptake inhibitor that enhances noradrenergic transmission, thus it is supposed to be effective for FOG. However, a randomized double-blind, placebo-controlled study enrolled 5 patients did not find its efficiency on FOG [87]. Another eight-week open label study enrolled 10 PD patients with dopamine-unresponsive FOG also failed to demonstrate its efficiency [88].

An underlying loss of cholinergic function contributes to freezing [109], thus the effect of acetylcholinesterase inhibitors on FOG has been tested. One open controlled study included 41 patients with PD with dementia and randomized patients to a galantamine treatment group in which patients received galantamine in addition to ongoing treatment (21 patients) and a control group (20 patients) who continued ongoing treatment. Results showed 24-week galantamine (8 mg twice daily) add-on therapy significantly improved freezing assessed by UPDRS freezing subitem (3.0 ± 0.5 to 2.3 ± 0.4, P = 0.03) [89]. But recently the double blind placebo-controlled study did not find FOG improvement after acetylcholinesterase inhibitor rivastigmine treatment [90].

Two double blind placebo-controlled studies using subjective and objective measures found no significant FOG improvement with BTX-A injections into calf muscles [91, 92], albeit the beneficial effect has been observed previously in two open label studies [93, 94]. Apart from BTX-A, the effect of BTX-B on FOG was also disappointing [95], thus this approach is therefore discouraged.

A meta-analysis summarized the short and long-term effects of bilateral STN-DBS with HFS (usually 130 Hz) on gait and FOG in PD [118]. Seven studies with available data of FOG (UPDRS item 2.14) were included [119‐125]. The results showed that gait and FOG could be improved for more than 4 years in the Med-Off/Stim-On condition. However, no beneficial effect was found for the Med-On/Stim-On compared with Med-ON/Stim-Off. This is probably due to the selection criteria for DBS candidates was levodopa responsiveness of global motor performance as well as gaits. These results confirmed the lack of a synergistic effect of medication and DBS [118]. Other studies also reported that STN-HFS improved FOG in most patients [126‐128]. However, STN-DBS worsened or induced the new appearance of FOG and/or gait akinesia during the immediate postoperative period in some patients [129, 130]. Such paradoxical deterioration of gait and akinesia is rare, and this side effect is probably related to misplaced contacts [131]. Therefore, when levodopa-responsive FOG patients is complicated by dose related side effects of levodopa treatment, high frequency STN-DBS can be considered. Meanwhile, antidepressant treatment can be tried if depression exists, as depression may negatively affect the outcome of HFS on FOG [132].

LFS (most commonly 60 Hz) STN-DBS has been tried to treat PD patients with FOG and have shown its short-term or even long-term beneficial effects on improving FOG and other axial symptoms compared with HFS [133‐135], although some studies argued that there was no significant difference between HFS and LFS for controlling FOG [136, 137]. Currently, there are lack of meta-analysis to examine efficiency of low frequency STN-DBS on FOG, thus caution should be used in interpreting the results of these clinical studies and drawing conclusions. However, LFS should be recommended in STN-DBS patients with refractory axial symptoms (particularly FOG) at HFS, as long as tremor doesn’t worsen significantly with LFS [138‐140].

One randomized double-blinded study evaluated the long-term effect of PPN–DBS with a follow-up period for 24–48 months and reported that the benefits on FOG and falls lasted at least 4 years in four of the six patients [141]. On behalf of the Movement Disorders Society PPN–DBS Working Group, Thevathasan et al. reviewed the literature from 1990 to 2017 comprising fewer than 100 cases and summarized the outcomes of PPN-DBS application in patients with PD [142]. Although firm conclusions cannot be drawn on therapeutic efficacy, the literature suggested that PPN-DBS could improve medication refractory gait freezing and falls [143‐148]. These cases were from different surgical centers; therefore, there was great variability in clinical methodology. Caudal or rostral PPN, unilateral or bilateral implantation, high or low frequency stimulation are factors that may affect the outcome. Thus, consensus about the optimal methodology is needed.

A crossover, double-blind, randomized controlled trial enrolled 12 PD patients with gait and balance impairment resistant to optimized dopaminergic and STN-DBS treatment found that combined stimulation of the STN and substantia nigra pars reticulata (SNr) using interleaved stimulation of both sites at a high frequency of 125 Hz improved FOG at 3-month follow-up [149]. In another randomized, crossover pilot study, Valldeoriola and colleagues modified this approach by co-stimulating STN and SNr at different frequencies in 6 patients, and investigated the effects on FOG and other PD motor symptoms [150]. The results suggested that combined HFS-STN and LFS-SNr stimulation had the best effect on various measures of FOG, compared to stimulation of a single target alone. However, no comparative data between low- and high-frequency SNr stimulation (neither SNr alone nor combined with STN) is available, whether a low frequency at SNr would be superior over high frequency remains unclear.