Freezing of Gait Levodopa Response Pattern in Parkinson's Disease Provides Clues to Pathophysiology

ABSTRACT

Freezing of gait (FOG) is a common and enigmatic feature of Parkinson's disease (PD) because of its episodic and unpredictable nature. It is now clear that FOG is not a monolithic phenomenon but instead exhibits substantial heterogeneity across patients, suggesting the existence of subtypes. Among the heterogeneous features are levodopa response patterns and nonmotor features, cognitive impairment and anxiety/depression. It remains an open question as to whether these phenotypes are the result of different pathophysiology. In this paper, we develop the hypothesis that levodopa response patterns may identify FOG subpopulations tied to nonmotor symptoms and alterations in different neurotransmitter systems. Here, we review the levodopa response patterns of FOG seen in PD based on a rigorous levodopa challenge paradigm using a 40% higher dose of levodopa and blood levodopa levels to demonstrate that the majority of patients are either levodopa responsive (OFF‐FOG) or unresponsive (ONOFF‐FOG). The literature demonstrates that executive and affective changes are not universal in FOG and actually relate closely to levodopa response patterns, OFF‐FOG is associated with affective disorders, and ONOFF‐FOG is related to cognitive decline. In turn, OFF‐FOG and affective disorders appear to be associated with brain noradrenergic degeneration, whereas ONOFF‐FOG and cognitive decline are associated with cholinergic loss. These observations suggest different therapeutic targets by subtype. We suggest that levodopa challenge testing may help stratify FOG patients in trials and mechanistic studies.

Short abstract

In this paper, we develop the hypothesis that levodopa response patterns may identify FOG subpopulations tied to nonmotor symptoms and neurotransmitter systems. Executive and visuospatial dysfunction and cholinergic loss are associated with levodopa‐unresponsive FOG. Anxiety/depression and noradrenergic degeneration appear more associated with levodopa‐responsive FOG.

Abbreviations

[C]MeNER

[C]MeNER (2S,3S)(2‐[α‐(2‐methoxyphenoxy)benzyl]morpholine)

BAI

Beck Anxiety Inventory

BDI‐II

Beck Depression Inventory‐II

FOG

freezing of gait

LC

locus coeruleus

MoCA

Montreal Cognitive Assessment

MRI

Magnetic resonance imaging

NE

norepinephrine

NET

norepinephrine transporter

NFOG‐Q

new freezing of gait questionnaire

OFF

FOG‐levodopa‐responsive freezing of gait

ONOFF

FOG‐levodopa‐unresponsive freezing of gait

PD

Parkinson's disease

PET

positron emission tomography

PPN

pedunculopontine nucleus

SCID

Structured Clinical Interview DSM

UPDRS‐III

Unified Parkinson's Disease Rating Scale

VAChT PET

[18F]FEOBV‐[18F]‐fluoroethoxybenzovesamicol

1. Introduction

Parkinson's disease (PD) is the second most common neurodegenerative disease, which is largely defined pathologically by the loss of dopaminergic neurons in the substantia nigra pars compacta and is levodopa‐responsive. Freezing of gait (FOG) is a common and enigmatic feature of PD because of its episodic and unpredictable nature (Nutt et al. 2011). It is one of the most significant unmet needs, causing falls with injury and leading to loss of independence with social isolation. Pathophysiology remains elusive, and treatment is a challenge.

It is now clear that FOG is not a monolithic phenomenon but instead exhibits substantial heterogeneity across patients, spanning responsiveness to dopaminergic therapy that varies despite the general response of other cardinal features of disease as well as other gait parameters (Virmani, Pillai, et al. 2023), and nonmotor features in addition to the acknowledged variability in motor manifestations (Gilat et al. 2026; Schaafsma et al. 2003; Snijders et al. 2008). The most pertinent nonmotor features include cognitive decline and anxiety/depression (Nutt et al. 2011; Giladi and Hausdorff 2006). There is also variation in levodopa response (McKay et al. 2019; Tosserams et al. 2025). Furthermore, imaging studies have shown findings that can relate to nearly the entire brain (Bharti et al. 2019), and numerous gait physiology measure changes between episodes have been associated with FOG (Snijders et al. 2016; Virmani, Landes, et al. 2023).

It remains an open question as to whether these phenotypes are the result of different pathophysiology. Using a specifically developed questionnaire and functional MRI imaging, Ehgoetz Martens and colleagues (Ehgoetz Martens, Shine, et al. 2018; Ehgoetz Martens, Hall, et al. 2018) identified three independent subtypes that emerge from distinct upstream dysfunctions and ultimately lead to FOG: asymmetric‐motor, anxiety, and attentional set‐shifting impairments. In this paper, we develop the hypothesis that these patterns (at least two of them) relate to the levodopa response pattern of FOG and are likely to have different pathophysiology.

2. Patterns of Levodopa Response

There is a growing body of evidence suggesting that levodopa response patterns of FOG may reflect distinct phenotypes and offer insight into underlying pathophysiology. At least three patterns have been described (McKay et al. 2019; Tosserams et al. 2025; Espay et al. 2012). The first pattern is levodopa‐responsive FOG (OFF‐FOG), in which freezing emerges in the OFF‐medication state and resolves with dopaminergic therapy in the ON state. In these patients, FOG closely tracks overall motor responsiveness, consistent with a predominantly dopaminergic mechanism.

The second pattern is FOG that remains unresponsive to dopaminergic therapy even when other Parkinsonian motor signs show robust medication‐related improvement (ONOFF‐FOG). In this pattern, FOG remains roughly at the same level in both the ON state and the OFF state. McKay et al. (2019) assessed 60 PD patients with and without FOG both in the practically defined OFF state (≥ 12 h off medications) (Langston et al. 1992) and after supratherapeutic doses of dopaminergic medications approximately 40% higher than the standard morning dose (395 ± 243 mg levodopa equivalent; range,133–1348 mg) and showed that FOG persisted in 19 of 30 FOG patients who had achieved an otherwise clinically meaningful improvement of 20–80 (mean 45%) in MDS‐UPDRS‐III total score (F _1,90 = 53.17, p < 0.0001). Serum levodopa levels increased from 0.3 ± 0.4 ng/mg in the OFF state to 27.9 ± 16.8 ng/mg in the ON state and were highest in patients with ONOFF‐FOG. Nearly 80% of these patients developed dyskinesia as well. These findings indicate that persistent FOG was not attributable to inadequate levodopa dosing or delayed pharmacokinetic effects such as impaired gastrointestinal absorption. Taken together, these results demonstrate preserved dopaminergic responsiveness for PD motor signs despite clear unresponsiveness of FOG. Although FOG persistent in the ON‐state has sometimes been attributed to suboptimal levodopa dosing and labeled “pseudo‐ON FOG,” rigorous levodopa challenge studies strongly argue against this explanation. In this work, we use the term ONOFF‐FOG to describe those with levodopa‐resistant or unresponsive FOG.

A third pattern is FOG that is absent during the OFF state and is present during the ON state (ON‐FOG) (Espay et al. 2012). This includes FOG caused by levodopa, FOG that is worsened by levodopa, or diphasic FOG that appears during the transition period between OFF and ON states (Perez Parra et al. 2020). Dyskinesias have also been documented to exhibit treatment phase–dependent patterns similarly (Zeng et al. 2023).

Given that levodopa challenge testing remains uncommon in clinical practice for practical reasons, many prior studies have treated FOG as a single category, typically without documenting levodopa responsiveness or incorporating it into inclusion criteria. Further, estimating the relative prevalence of these response patterns is difficult. Although it is clear based on our clinical observations that ON‐FOG is significantly rarer than the other two patterns, it is hard to determine how often each pattern truly occurs without systematic OFF–ON testing in unselected clinic series or similar clinical surveillance. In one paper, based on patient report of FOG response pattern, the frequencies were 62% for OFF‐FOG, 36% for ONOFF‐FOG, and 2% for ON‐FOG (Amboni et al. 2015), and in another based on results of levodopa challenge, the frequencies were 40% OFF‐FOG, 57% ONOFF‐FOG, and 3% ON‐FOG (McKay et al. 2019).

3. No Strict Requirement for Temporal Progression From OFF‐FOG to ONOFF‐FOG

Because many parkinsonian signs are initially mild and levodopa‐responsive before becoming more severe and less responsive, it has often been assumed that, in all patients, FOG evolves along a continuum from levodopa‐responsive OFF‐FOG to more severe, levodopa‐unresponsive ONOFF‐FOG (Nonnekes et al. 2015). In contrast, others have argued that ONOFF‐FOG represents an independent phenotype that can emerge without preceding OFF‐FOG (Snijders et al. 2016; Amboni et al. 2015). Available data support the latter interpretation. FOG itself can occur in the absence of PD due to focal lesions (Fasano et al. 2017) or other conditions (Factor et al. 2006), providing strong evidence that initial basal ganglia deficits are not a strict requirement. Examining the clinical and demographic features of ONOFF‐FOG versus OFF‐FOG patients in PD provides additional clues. McKay et al. (2019) demonstrated that the mean duration of FOG in their ONOFF‐FOG group was 3.3 years as opposed to 4.4 years for the OFF‐FOG group. If ONOFF‐FOG was a later result of a cascade of changes, then the duration would be expected to be longer. This finding has been supported by other studies (Factor et al. 2014; Ferraye et al. 2013). Further, review of clinical records of 10 patients reported by McKay et al. indicated that ONOFF‐state freezing was present from the time of FOG onset (McKay et al. 2019). Further supporting the possibility of separable phenotypes, imaging data reviewed below suggest that ONOFF‐FOG is more closely associated with extra‐striatal cholinergic denervation, whereas OFF‐FOG is associated with noradrenergic terminal loss; however, longitudinal biomarker studies are needed to determine whether these patterns represent independent trajectories or divergent stages of disease evolution. Comparable pathophysiological data for ON‐FOG remain sparse, representing an important knowledge gap.

4. Nonmotor Features and Levodopa Response Pattern

Nonmotor features are more prevalent in those who experience FOG (Ehgoetz Martens, Lukasik, et al. 2018) than those who do not, with the frequency increasing as the severity of FOG advances (Giladi and Hausdorff 2006; Amboni et al. 2010). Most important in relation to subtyping FOG are cognitive and affective dysfunction. It has been suggested that these may be predictive of the onset of FOG or possibly share neurobiology (Ehgoetz Martens, Lukasik, et al. 2018). We will discuss how these features are associated with the levodopa response pattern and particular neurotransmitter circuitry.

4.1. Cognition

It is generally believed that patients with PD and FOG have worse cognitive function than those without FOG and that they may share neurobiology (Ehgoetz Martens, Lukasik, et al. 2018). This impairment spans multiple cognitive domains but appears most pronounced in executive and attentional function, as well as visuospatial processing (Factor et al. 2014; Monaghan, Gordon, et al. 2023; Yogev‐Seligmann et al. 2008; Vercruysse et al. 2012; Naismith et al. 2010; Nantel et al. 2012). Two prominent models of FOG pathophysiology have cognitive dysfunction in a central role.

Fahn (1995) described FOG as “By analogy to a telephone call, the freezing phenomenon might be compared to a busy signal in which the message cannot be delivered.” This insightful description points to what is now referred to as the interference model, which indicates that dopaminergic loss creates a vulnerable context in which cognitive overload of inputs to the basal ganglia disrupts information processing, leading to crosstalk between normally segregated motor and cognitive circuits and increased inhibitory output from the basal ganglia, culminating in FOG (Lewis and Barker 2009; Nieuwboer and Giladi 2013). Clinically, this mechanism is supported by the observation that adding cognitive demands (dual tasking such as serial subtraction) during walking increases FOG (Nutt et al. 2011; Vandenbossche et al. 2013; Monaghan, Ragothaman, et al. 2023).

A second model is the cognitive control model, which attributes FOG to impaired cognitive control of locomotion (Nieuwboer and Giladi 2013; Morris et al. 2020). In this model, motor plans are generated in (intact) frontal cortical regions, but dopaminergic loss within fronto‐striatal circuits impairs their appropriate selection and suppression (conflict resolution) (Shine et al. 2013; Mink 1996). As a result, individuals with FOG exhibit greater automatic activation of incorrect responses and reduced suppression of competing responses during incongruent tasks, leading to increased freezing, presumably because incompatible motor programs are engaged simultaneously and disrupt normal gait execution (Nieuwboer and Giladi 2013; Vandenbossche, Deroost, Soetens, et al. 2012; Vandenbossche, Deroost, Soetens, Coomans, et al. 2012).

A recent meta‐analysis on the relationship between cognitive change and FOG found 145 studies with a total sample size of 9010 subjects, 4240 with FOG and 4770 without FOG (Monaghan, Gordon, et al. 2023). Numerous studies demonstrated that, on average, global cognition, executive function/attention, and visuospatial function were significantly worse in those with FOG than without FOG, but with large statistical heterogeneity across study effects. Effects ranged from small to medium, with the greatest effect seen with executive function testing.

Notably, roughly half of the reviewed studies did not identify significantly worse executive or attentional function in patients with FOG—suggesting that cognitive impairment is not uniformly present across all individuals with FOG. This inconsistency implies that additional heterogeneous features of FOG may moderate the cognitive–FOG relationship. We propose that the levodopa response pattern is one such feature. Closer examination of several studies indicates that the association between cognitive dysfunction and FOG may be strongest, or potentially confined to, patients with ONOFF‐FOG.

To more thoroughly interrogate this hypothesis, we re‐reviewed all 145 studies in the 2023 systematic review (Monaghan, Gordon, et al. 2023) and conducted a thorough (albeit nonsystematic and not preregistered) secondary review of manuscripts published after 2022 that reported cognitive data in people with OFF‐FOG and ONOFF‐FOG. Studies were included if the levodopa response pattern was either explicitly reported or could be reliably inferred from the study description (e.g., patients described as having persistent ON‐state freezing despite otherwise meaningful levodopa response). We identified seven relevant manuscripts; results are presented in Supporting Information S1. We also note that Factor et al. (2014) and Moretti et al. (2011) each contribute multiple effect sizes from single samples. Therefore, findings from these studies should be interpreted with appropriate caution.

Although data remain somewhat heterogeneous, results broadly support the hypothesis that individuals with OFF‐FOG exhibit better cognitive function than those with ONOFF‐FOG. For example, Moretti et al. (2011) showed that patients with “ON” FOG exhibited worse executive dysfunction than those “without ON‐freezing (but with OFF freezing)” in executive, attentional, and visuospatial function. Ricciardi et al. (2014) and Amboni et al. (2015) contributed supportive data as well; results were broadly consistent with this pattern, though effect sizes in Ricciardi et al. were variable and generally nonsignificant. Although these studies varied in their assessments, they nevertheless had consistent results demonstrating an association between ONOFF‐FOG and executive dysfunction. Notably, one interventional study reported that cognitive training improved FOG only in the ON medication state, further strengthening the link between ONOFF‐FOG and executive dysfunction (Walton et al. 2018). Refining the point further, one negative study from the systematic review (Nantel et al. 2012) considered patients with OFF‐FOG and found no difference in executive function between those with OFF‐FOG and those without FOG. Finally, data from four studies from our group and others have explicitly compared cognitive assessment results across ONOFF‐FOG, OFF‐FOG, and NO‐FOG in PD (Factor et al. 2014; Turner et al. 2021; Landes et al. 2022; Goldstein et al. 2018), including data reported in abstract form (Supporting Information S2) (Goldstein et al. 2018) and therefore not included in Supporting Information S1, and also demonstrate that ONOFF‐FOG patients had significantly poorer executive and visuospatial function than OFF‐FOG and No‐FOG groups (Table 1). In three of the studies, the ONOFF‐FOG was compared to and found to perform more poorly than the OFF‐FOG and NO‐FOG groups (Factor et al. 2014; Turner et al. 2021; Goldstein et al. 2018), whereas one study just compared ONOFF‐FOG with OFF‐FOG groups (Landes et al. 2022). An important limitation is that available studies vary in whether they adjust for disease duration, motor severity, global cognitive status, or education. Therefore, although the pattern is consistent with greater executive and visuospatial dysfunction in ONOFF‐FOG, future adequately powered studies should determine whether these associations persist after controlling for disease duration, MoCA or comparable global cognitive measures, and dopaminergic motor responsiveness. Nevertheless, these results suggested a pattern whereby the levodopa response pattern may be a relevant component of the FOG phenotype, particularly with respect to cognitive contributions.

TABLE 1.

Studies examining groups with ONOFF‐FOG, OFF‐FOG, and No‐FOG and reporting associations between ONOFF‐FOG and cognitive dysfunction.

References	ONOFF‐FOG (patient number)	OFF‐FOG (patient number)	NO‐FOG (patient number)	Executive	Visuospatial	Memory
Factor et al. (2014)	16	20	99	Trails B‐A*	JOLO**
Goldstein et al. (2018)	22	16	17	Letter fluency,** digit span backward**	JOLO	Composite (total recall, delayed recall, recognition discrimination)*
Turner et al. (2021)	14	23	17	Set‐shifting errors*
Landes et al. (2022)	26	36	43	MoCA,** frontal assessment battery,* SCOPA‐Cog*

Note: Goldstein et al. was published in abstract form. See Supporting Information S2.

Abbreviations: JOLO = Judgement of Line Orientation; MoCA = Montreal Cognitive Assessment; OFF‐FOG = levodopa‐responsive FOG; ONOFF‐FOG = levodopa‐unresponsive FOG; SCOPA‐Cog = scales for outcomes in Parkinson's disease—cognitive.

^* p < 0.05.

^** p < 0.01.

Taken together, these results strongly suggest that the levodopa response pattern may be a critical factor in examining associations between executive functioning and FOG and, more specifically, that executive function deficits may be strongest among, or even limited to, patients with ONOFF‐FOG, but this remains to be established in adequately powered studies.

Anatomical studies provide additional support for an association between executive dysfunction and ONOFF‐FOG, pointing to degeneration of cholinergic systems that integrate attention, cognition, and gait. Acetylcholine circuits are important in the maintenance of executive function and attention, with sources from the basal nucleus, striatum, prefrontal cortex, and pedunculopontine nucleus (PPN) in PD (Pasquini et al. 2025; Wilson et al. 2021). An association has also been shown between PPN/lateral dorsal tegmental complex, thalamic, and basal forebrain corticopetal cholinergic denervation, with falls and gait slowing in PD (Muller and Bohnen 2013; Bohnen and Albin 2011). Furthermore, Bohnen and colleagues have shown that levels of acetylcholine are reduced in those with FOG, compared with those without, in the striatum, temporal, and mesiofrontal limbic regions (Bohnen et al. 2019). It is therefore feasible that cholinergic dysfunction underpins both cognitive deficits and FOG (Chou et al. 2025). In a recent study, using the vesicular acetylcholine transporter positron emission tomography radioligand for [18F]‐fluoroethoxybenzovesamicol ([18F]FEOBV) (VAChT PET), 36 PD subjects with FOG had their FOG assessed through video assessments ON and OFF levodopa (Chou et al. 2025). With whole‐brain voxel‐based analyses, ONOFF‐FOG was associated with bilateral extra‐striatal cholinergic terminal reductions, particularly thalamo‐limbic structures, involved in multisensory and cognitive integration (attentional control) of gait and postural control, as well as spatial navigation. These findings suggest a close relationship between cognitive decline, reduced VAChT binding in multiple pathways, and FOG levodopa unresponsiveness. Further, Hvingelby et al. (2025) also performed a cross‐sectional PET imaging study using 18F‐FEOBV and found lower levels of cholinergic activity in the thalamus, hippocampus, striatum, anterior cingulate, amygdala, and brainstem were associated with ONOFF‐FOG, suggesting that 18F‐FEOBV binding is a predictor of moderate strength of ONOFF‐FOG (R ² = 0.46975, p = 0.045) and interactions between cholinergic and executive dysfunction in generating ONOFF‐FOG. Although the N was small and there were questions regarding defining and assessing ONOFF‐FOG, these data appear to further support the findings of Chou et al. (2025).

4.2. Affective Disorders

Anxiety and depression are common nonmotor features of PD, impacting 30%–70% of patients (Ehgoetz Martens et al. 2014; Factor et al. 2017). It has been shown that a relationship exists between anxiety and severe gait impairments (Ehgoetz Martens et al. 2014; Ehgoetz Martens, Hall, et al. 2016; Hall et al. 2015), and anxiety or depression has been shown to frequently co‐occur with FOG (Giladi and Hausdorff 2006; Ehgoetz Martens, Hall, et al. 2016; Hall et al. 2015; Burn et al. 2012; Herman et al. 2023). Data supporting this relationship are from the use of questionnaires, interviews, biological measures of anxiety, and situations where anxiety is induced. Affective disorders could impact FOG through the interference model in an analogous fashion to cognitive deficits whereby, on a background of dopamine depletion, abnormal processing of emotional input (fear) could overload the capacity of the basal ganglia to manage competing inputs through interconnections between the limbic (nucleus accumbens) and motor circuits (Lewis and Barker 2009; Lewis and Shine 2016), which, in turn, interfere with motor outputs resulting in increased inhibition and FOG (Ehgoetz Martens et al. 2014).

Some of the most compelling evidence for an association between affective disorders and FOG comes from studies that directly manipulate anxiety and measure FOG as an outcome. When directly manipulated experimentally, the presence of anxiety has been shown to increase the number and severity of FOG episodes rather than the anxiety simply being a response to FOG (Ehgoetz Martens et al. 2014). In such studies, anxiety‐inducing virtual environments appear to reliably elicit a greater number of FOG episodes (Ehgoetz Martens et al. 2014). It is important to point out, however, that anxiety‐inducing virtual environments are unlikely to isolate affective load alone. Such paradigms often also increase attentional demand, conflict monitoring, visuospatial processing, or dual‐task burden. Thus, VR‐induced FOG may reflect an interaction between affective arousal and cognitive overload. This has been supported by fMRI studies demonstrating cortico‐striatal decoupling, particularly involving frontal areas supporting the impact of executive dysfunction, the interference and cognitive control models (Bardakan et al. 2022). An additional report indicated that both anxiety and depression were significant predictors of FOG (Ehgoetz Martens, Hall, et al. 2016; Ehgoetz Martens, Szeto, et al. 2016). As support of this finding, those PD patients with mild FOG demonstrated more anxiety symptoms than patients without FOG and less than those with severe FOG, suggesting that anxiety may be an important independent contributor to the underlying pathophysiology of FOG (Ehgoetz Martens, Hall, et al. 2016).

Despite experimental evidence linking anxiety to FOG, consistent associations at the group level have not been demonstrated. As with cognitive decline, affective disorders have been frequently but inconsistently associated with FOG. Although not studied as rigorously as cognitive dysfunction, a systematic review from 2019 (Witt et al. 2019) found 26 studies examining the relationship between anxiety and FOG in PD. Only 16 showed a significant relationship between an anxiety outcome measure and either the presence or severity of FOG. More recently, one study (Landes et al. 2022) measured FOG using instrumented continuous gait assessment and classified participants as NO‐FOG (n = 43), OFF‐FOG (n = 36), or ONOFF‐FOG (n = 26). Anxiety and depression were assessed using the Hamilton scales, and no differences were observed between groups.

Why is this? Perhaps the main reason that many authors have suggested, and which we concur, is that more comprehensive or PD‐specific anxiety assessments are needed to clarify the relationship between anxiety and FOG. Commonly used instruments such as the Hamilton or Beck scales, which were developed for general or geriatric populations, may lack specificity in PD, as they include items related to motor function such as feeling “unsteady” that may inflate anxiety scores in this population.

A second reason is that we believe this association may only really hold (or be the strongest) among those with levodopa‐responsive, OFF‐FOG. We note that in the review mentioned earlier, although two studies did motor testing in the ON and OFF states (Ehgoetz Martens et al. 2014), none of the studies reported on the responsiveness of FOG itself.

In another study, which was reported by abstract (Supporting Information S3), McKay and Factor (2018) examined whether anxiety or depression was associated with levodopa response patterns of FOG in two populations, one where FOG response to levodopa was by self‐report and one where subjects were examined through levodopa challenge (McKay et al. 2019)—(no FOG vs. levodopa‐responsive FOG [OFF‐FOG] and vs. levodopa‐unresponsive FOG [ONOFF‐FOG]) using a more comprehensive approach. Here, N = 125 PD patients were assessed for depression and anxiety using a Structured Clinical Interview DSM (SCID) for diagnosis of depression or anxiety disorder, as well as with the Beck Depression Inventory‐II (BDI‐II) and the Beck Anxiety Inventory (BAI), common screening instruments. BDI‐II and BAI scores were dichotomized about cut points from the literature as indicative of depression and anxiety, respectively. Analyses controlled for age, sex, education, MoCA score, UPDRS‐III score, disease duration, and medications. Current depression on SCID was associated with significantly increased odds of OFF‐FOG (OR [95% CI]: 4.84 [1.24–19.00]; p = 0.02). A similar, marginally significant association was identified for current anxiety and OFF‐FOG (3.90 [0.92–16.50]; p = 0.07). In contrast, associations between depression or anxiety and unresponsive FOG were not seen (OR: 0.91, depression; 1.05, anxiety). Although similar patterns were identified for dichotomized BDI‐II and BAI scores, effect sizes were attenuated, and p‐values were larger (Table 2).

TABLE 2.

Associations between depression and anxiety and FOG subtype (Goldstein et al. 2018).

	OFF‐FOG vs. NO‐FOG		ONOFF‐FOG vs. NOFOG
Variable	OR [95% CI]	p	OR [95% CI]	p
Current depression (SCID)	4.84 [1.24–19.00]*	0.02	0.91 [0.10–8.50]	0.93
Current anxiety (SCID)	3.90 [0.92–16.50]	0.07	1.05 [0.10–10.96]	0.97
BDI‐II score a	1.01 [0.95–1.08]	0.69	1.03 [0.96–1.10]	0.39
BAI score b	1.05 [0.98–1.11]	0.14	1.03 [0.97–1.10]	0.32
BDI‐II > 13	1.31 [0.47–3.65]	0.60	1.11 [0.39–3.17]	0.85
BAI > 14	2.26 [0.83–5.58]	0.11	1.53 [0.56–4.17]	0.41

Abbreviations: BAI = Beck Anxiety Inventory; BDI‐II = Beck Depression Inventory version II; OFF‐FOG = levodopa‐responsive FOG; ONOFF‐FOG = levodopa‐unresponsive FOG; SCID = structured interview for DSM‐5.

^a N = 125.

^b N = 114.

^* p < 0.05.

These results led us to the hypothesis that affective disorders may be differentially associated with OFF‐FOG, with the implication that OFF‐FOG may reflect distinct underlying pathophysiology with potentially less interaction with limbic cortico‐basal ganglia pathways.

What might be the neuroanatomical basis for an association between OFF‐FOG and affective disorders? The locus coeruleus (LC) is the primary source of central norepinephrine (NE) (~70%) and has widespread projections involved in arousal, mood, stress, and autonomic regulation (Ordway et al. 1997; Tejani‐Butt 1992; Pifl et al. 2012; Factor et al. 2025). Through these circuits, it plays a major role in processing anxiety‐provoking situations and, in turn, the development of anxiety and depression (Goddard et al. 2010). In PD, the LC demonstrates α‐synuclein pathology early and at variable levels (Braak et al. 2003) and ends with significant loss of neurons in the late stages. This loss has numerous impacts on the dopamine system as well, including exacerbating motor dysfunction in PD models, contributing to neuronal degeneration, and altering levodopa response (Factor et al. 2025). There is mounting evidence that supports the role of NE on FOG (Factor et al. 2025; McKay et al. 2023; Ono et al. 2016). Considering the relationship between NE and DA systems, the impact of NE on FOG may relate to the levodopa response pattern. Noradrenergic markers of anxiety, including elevated heart rate, skin conductance, and pupillary change, are associated with FOG in PD (Taylor et al. 2022). This circuit is also proposed to play a key role in the development of FOG through distinct compensatory networks involving the limbic system, which plays a key role in emotional processing (Factor et al. 2025; McKay et al. 2023). Imaging studies have shown a clear relationship between FOG and limbic connections to the striatum (Ehgoetz Martens, Hall, et al. 2018; Gilat et al. 2018; D'Cruz et al. 2021).

This relationship, similar to the occurrence of anxiety and depression, seems to be specifically related to a pattern of levodopa response as demonstrated through a recent study that examined NE transporter (NET) binding via brain PET using the radioligand [¹¹C]MeNER (2S,3S)(2‐[α‐(2‐methoxyphenoxy)benzyl]morpholine) ([¹¹C]MeNER) to evaluate changes in NET density associated with FOG (McKay et al. 2023). Fifty‐two patients were included and classified via rigorous levodopa challenge as PD without FOG (N = 16), OFF‐FOG (N = 10), ONOFF‐FOG (N = 21), and primary progressive (non‐PD) FOG (N = 5). OFF‐FOG, specifically, was linked to reduced expression in multiple brain regions. The levodopa‐responsive FOG group exhibited reduced whole‐brain NET binding compared with the no FOG group (−16.8%, p = 0.021). Region‐specific examinations demonstrated decreased NET expression in the left and right frontal and temporal lobes, thalamus, amygdalae, and LC, with the strongest effect seen in the right thalamus (p < 0.038). Linear regression analysis revealed a correlation between decreased NET binding in the right thalamus and more severe FOG based on the NFOG Questionnaire score (p < 0.022). The noradrenergic regions of the thalamus are present in the midline, central intralaminar, and medial mediodorsal nuclei, which project to the limbic circuit, which has been previously linked to FOG and relates to depression and anxiety. The reduced binding in OFF‐FOG, in particular, is consistent with the close relationship between NE and dopamine loss, as previously noted, which is also consistent with the finding that anxiety and panic are frequent features of OFF episodes. These findings suggest a close relationship between anxiety/depression, reduced NET binding in limbic pathways, and responsiveness of FOG to levodopa, more specifically, levodopa‐responsive FOG.

A recent MRI study demonstrated results that support the relationship between the LC‐NE circuits and OFF‐FOG (Sun et al. 2025). To examine the extent of neurodegeneration of LC in PD subjects with and without FOG, the authors used free water corrected DTI T1‐weighted MRI imaging to study several indices of microstructural changes in the LC and resting state fMRI to examine LC functional connectivity through its extensive neural projections. Subjects included 52 healthy controls, 79 PD patients without FOG, and 110 with FOG (48 OFF‐FOG and 62 ONOFF‐FOG), and they were examined in the OFF‐medication state 12 h after their last levodopa dose. Approximately half were reimaged in the ON‐medication state. Although they divided the FOG group into OFF‐FOG and ONOFF‐FOG, details are scant on how they did this. There is no mention of a levodopa challenge. No difference in LC volume was seen between groups. The FOG group, however, had significantly higher free water values and lower fractional anisotropy than the no FOG and healthy control groups, indicating greater neurodegeneration, perhaps relating to prolonged neuroinflammation. Severity of FOG as measured by the NFOG‐Q was positively correlated with free water values in the FOG group. The PD group with FOG had significantly reduced LC connectivity to various regions within the visual cortex in the OFF‐medication state. There was no difference in functional connectivity in the OFF‐FOG and ONOFF‐FOG groups. However, the LC functional connectivity to the visual cortex was enhanced in the ON‐medication state in the OFF‐FOG group only. These findings support a close relationship between LC‐NE circuits and levodopa‐responsive FOG. There was no discussion in the paper on affective disorders within the subject groups.

5. Possible Mechanistic Considerations

We offer some potential mechanistic explanations for the association between acetylcholine and ONOFF‐FOG and noradrenergic deterioration and OFF‐FOG, but they remain speculative due to the lack of established causal relationships.

Cholinergic denervation likely contributes to levodopa‐unresponsive FOG through parallel effects on cognition and gait control, with cognitive deficits emerging as a consequence of cholinergic loss but not necessarily lying on the causal pathway to FOG. This would suggest more of a correlational relationship. The possible mechanisms relating cholinergic denervation to levodopa unresponsiveness in PD patients with FOG are discussed in detail by Chou et al. (2025). Cholinergic systems contribute to attentional control, cue detection, error processing, and multisensory balance performance (Chou et al. 2025; Roytman et al. 2025; Bohnen et al. 2022), which are crucial for the maintenance of stable gait and posture. Chou et al. (2025) found in their VAChT PET study that ONOFF‐FOG patients had more widespread extra‐striatal degeneration of brain cholinergic systems, specifically involving thalamo‐limbic structures, which play a major role in multisensory integration. Although it is suggested that FOG may result from the combination of striatal dopamine loss and subsequent extra‐striatal cholinergic denervation, it has been suggested that cholinergic denervation in these neural networks may, in and of themselves, interfere with sensorimotor processing and cause balance and FOG problems that are refractory to levodopa therapy (Roytman et al. 2025; Bohnen et al. 2022). ONOFF‐FOG patients had more cholinergic deficits in several key areas involved in sensorimotor processing, including the higher order limbic regions of hippocampi, fimbria, posterior cingulate region, entorhinal cortex, and retrosplenial cortex (Chou et al. 2025), all consistent with the loss of basal forebrain cholinergic projections (Mesulam and Geula 1988). Further, lower level points of integration between vestibular, auditory, and visual processing are in the thalamus, which is also shown to have lower cholinergic activity (Chou et al. 2025). Cholinergic denervation in ONOFF‐FOG cases was also seen in the dorsomedial frontal cortex, which is associated with not only sensory but also higher order motor control functions, which are generally not levodopa responsive (Chou et al. 2025; Giladi et al. 2007). The cognitive control hypothesis suggests that cognitive dysfunction may contribute to FOG; however, an alternative explanation is that cognitive dysfunction and (ONOFF‐) FOG could arise in parallel as consequences of cholinergic denervation. This distinction may explain why FOG is commonly observed in atypical parkinsonism but not in dementia per se (Parmera et al. 2025).

The contribution of LC pathology to PD‐associated neuropsychiatric symptoms has been discussed here and elsewhere (Weinshenker 2018). Broadly, although NE does not encode motor commands, it rescales cortico‐striatal transmission by modulating neuronal gain, signal‐to‐noise, and input responsiveness. In the setting of dopaminergic impairment, this adrenergic modulation may serve a compensatory role, effectively amplifying motor commands to stabilize output and prevent FOG by increasing LC‐NE tone. When the NE system is also compromised, however, this compensation becomes contingent on dopaminergic state, such that it may be sufficient in the ON state but fails in the OFF state, leading to OFF‐FOG. The LC extensively projects to, either directly or indirectly, and regulates the function of motor centers including the cerebellum, primary and secondary motor cortices, red nucleus, and motor thalamus (Waterhouse et al. 2022). Ample evidence also implicates the LC‐NE system in motor dysfunction; we and others have shown that reducing LC‐NE transmission can impair motor function and motor learning, whereas increasing central adrenergic signaling promotes these functions (Thomas and Palmiter 1997; Rommelfanger et al. 2007; Yin et al. 2021). Because most of these data were derived from rodent neuroanatomical studies, they should be extrapolated with caution and need to be confirmed in nonhuman primate and human brains. In humans, it is clear that there is substantial damage to the NE system in PD, which complicates things further (McMillan et al. 2011).

Particularly germane to FOG in PD, it has been proposed that engagement of the LC‐NE system may be critical under demanding conditions requiring swift yet accurate movements. For example, the LC is powerfully activated by stress, and a common trigger of FOG is when the visual environment changes, especially when navigating obstacles, turning, and at transitions in floor texture, color, or pattern (Nutt et al. 2011). These demands, which require precise sensorimotor integration, may constitute stressful challenges for people with PD. Individuals with more compromised LC function may be especially prone to FOG triggered by transient environmental or cognitive stressors, whereas ONOFF‐FOG may reflect a less context‐dependent, more stochastic pattern of occurrence.

Although levodopa is generally considered through the lens of dopamine‐replacement therapy, it is worth noting that because dopamine is the precursor to NE in noradrenergic cells, levodopa may have the capacity to rescue NE deficiency in PD as well, although published clinical data supporting this notion are lacking. We speculate that a PD‐impaired LC could contribute to OFF‐FOG, whereas elevation of NE synthesis capacity by levodopa may ameliorate symptoms.

One other potential link between levodopa responsiveness and LC‐NE circuitry was described by Sun et al. (2025) who found that LC functional connectivity to the visual cortex was impaired in the OFF‐medication state in FOG patients compared with the no FOG and healthy control groups. There was no difference in the LC connectivity between OFF‐FOG and ONOFF‐FOG. However, dopaminergic medication significantly enhanced LC functional connectivity to the visual cortex in the OFF‐FOG group but not in the ONOFF‐FOG group. Such an increased connectivity could improve sensory (visual) perceptual performance and, in turn, improve FOG. The ONOFF‐FOG did not experience such an increased connectivity, and as indicated by their subtyping, the FOG did not improve, hence suggesting that the LC‐NE connectivity changes were specific to the OFF‐FOG group. This would suggest a different pathophysiology for the two levodopa‐responsive subtypes.

6. Therapeutic Implications

These findings have implications related to therapy beyond just the responsiveness to levodopa and its place in treating FOG. Interestingly, there have been several studies of noradrenergic agents in PD for the treatment of FOG, with surprisingly mixed results. We think the key missing ingredient was suboptimal patient selection, particularly the absence of stratification by levodopa response pattern. As a result, differences in patient mix may have driven the divergent outcomes rather than true differences in therapeutic effect. For example, the noradrenergic precursor droxidopa (aka L‐DOPS) was examined in the 1980's in a double‐blind, placebo‐controlled trial in PD and demonstrated that approximately 25% of the patients with FOG had moderate to marked improvement (Narabayashi et al. 1987). Although these results were not replicated (Fahn 1995), they led to the approval of droxidopa specifically for FOG in Japan. A limitation of these earlier noradrenergic studies is the lack of characterization of medication state during testing and the FOG levodopa‐response subtype. As a result, it remains unclear whether participants predominantly exhibited OFF‐FOG or ONOFF‐FOG at outcome assessment, which may have obscured subtype‐specific treatment effects. In addition, two pilot studies have examined the noradrenergic agent atomoxetine for FOG in PD in a small number of subjects (Revuelta et al. 2015; Jankovic 2009). In both (negative) studies, the reports specify that the included patients were evaluated in the ON state for objective measures (but also assessed by the FOG‐Q subjectively) and probably had ONOFF‐FOG and therefore would not have been expected to benefit from noradrenergic therapy anyway. A new trial of atomoxetine in PD patients with FOG is currently underway. There has also been interest in the use of cholinesterase inhibitors for the treatment of gait abnormalities and falls based on data indicating an association as addressed above, but there have not been any trials thus far specifically designed to assess FOG (Mancini et al. 2015; Henderson et al. 2016). Noninvasive brain stimulation may represent another subtype‐relevant therapeutic approach. Studies targeting prefrontal regions, particularly the dorsolateral prefrontal cortex, have reported improvements in FOG or gait outcomes in PD (Potvin‐Desrochers and Paquette 2021). Given the association between ONOFF‐FOG and executive/attentional dysfunction, prefrontal stimulation approaches may be especially relevant for the stratification of patients in whom cognitive‐control deficits contribute to freezing. The same is probably true for subthalamic nucleus deep brain stimulation, where it has been shown that levodopa‐responsive freezing responds positively (Moreau et al. 2008). The data reviewed above suggest such trials may be feasible. The use of a levodopa challenge paradigm as part of the screening process for future trials will allow for the appraisal of response in subtypes of FOG or, even further, the enrichment of study populations for specific therapies, for example, ONOFF‐FOG for cholinergic agents and OFF‐FOG for noradrenergic drugs and an appropriate pattern for nonmotor profiles.

7. Summary and Conclusion

Ehgoetz Martens et al. suggested that there may be subtypes of FOG in PD, two of which are related to the nonmotor correlates of affective disorders and cognitive disturbance (Ehgoetz Martens, Shine, et al. 2018). Data support that both nonmotor features, although well known to occur in FOG, occur perhaps in selected populations. In PD, it is well known that cognitive impairment relates to cholinergic neuron degeneration in the basal nucleus, caudate nucleus, cerebral cortex, and PPN and that affective disorders are related to NE change secondary to LC degeneration and impacting limbic function. Here, we hypothesize that levodopa response patterns may identify FOG subpopulations tied to these neurotransmitter systems. Executive and visuospatial dysfunction and cholinergic loss are associated with ONOFF‐FOG (Chou et al. 2025). Anxiety and depression, and likely noradrenergic degeneration, appear more associated with OFF‐FOG (McKay et al. 2023). These observations suggest different therapeutic targets by subtype (Figure 1). The scheme is simplified and hence preliminary since we know that these nonmotor features may overlap (Ehgoetz Martens, Hall, et al. 2016) and some ONOFF‐FOG cases do not have executive dysfunction (Ricciardi et al. 2014), perhaps indicating that additional tiers of subtyping may exist and require further vetting. As such, the levodopa response pattern is unlikely to capture all clinically relevant heterogeneity. Some patients with ONOFF‐FOG may have prominent affective symptoms, some patients with OFF‐FOG may have cognitive impairment, and some in both subtypes may have neither. At present, validated thresholds for further stratification into affective/nonaffective or cognitive/noncognitive subgroups are lacking. Future studies combining levodopa challenge testing with standardized cognitive and PD‐specific affective assessments could determine whether such nested subtypes improve the prediction of pathophysiology or treatment response. There are other clinical features of FOG that are heterogeneous in patients, such as whether the FOG is akinetic or kinetic, whether postural disturbance is present or absent, and including these would be important in further stratification. Also, the role of particular test types such as dual task testing in the stratification of cases is important. Does it particularly impact ONOFF‐FOG or both types in the same manner? Would it provide additional information for subtyping? This needs to be examined further. These hypotheses are testable in clinical studies, which should provide important information on the pathogenesis of FOG. Still, for now, based on current findings, we believe that levodopa challenge testing may help to better stratify FOG patients for clinical trials and mechanistic studies.

FIGURE 1.

Distinct freezing of gait phenotypes in PD. This figure contrasts L‐dopa–responsive (OFF‐FOG) and L‐dopa–unresponsive (ONOFF‐FOG) freezing of gait in PD across levodopa response pattern, nonmotor features, neurochemical binding deficits, and therapeutic implications. OFF‐FOG is characterized by freezing that improves with dopaminergic medication, prominent affective symptoms, and reduced norepinephrine transporter (NET) binding reflecting noradrenergic terminal loss especially in the thalamus (TH; dark green) and also involving frontal cortices (FC; light green), temporal lobe (TL; light green), amygdala (AMG; light green), and locus coeruleus (LC; light green). ONOFF‐FOG is characterized by freezing that persists despite dopaminergic medication, cognitive deficits in attentional, executive, and visuospatial domains, and reduced vesicular acetylcholine transporter (VAChT) binding localized to thalamo‐limbic and posterior medial regions, including the hippocampus (HPC; orange), fimbria (FIM; orange), posterior cingulate cortex (PCC; orange), and thalamus (TH; orange). These dissociable profiles motivate noradrenergic versus cholinergic therapeutic strategies.

Levodopa testing is central to identifying levodopa‐responsive versus levodopa‐unresponsive FOG. However, we appreciate that levodopa challenge testing is resource and time consuming for the patient and the clinician. Hence, its completion can represent a challenge. In research protocols, it is often utilized when testing agents that are being developed to treat motor fluctuations, as shown, for instance, in studies of apomorphine (Olanow et al. 2020). Hence, the levodopa challenge in this setting is shown to be feasible. We believe that, where the accuracy of identifying ON and OFF states is paramount, this paradigm should be followed. However, in a busy clinical practice, although it is utilized in some instances (although not with supratherapeutic doses of levodopa), for example, during DBS programming, it is generally more challenging. A more feasible alternative is to complete separate visits, one in the practically defined OFF state and another in the usual care ON condition in order to see both scenarios. Incorporating cognitive loading such as dual task walking, affective provocation, within these medication‐state paradigms may further clarify whether cognitive and affective mechanisms differentially map onto ONOFF‐FOG and OFF‐FOG.

Author Contributions

David Weinshenker: writing – original draft, writing – review and editing, conceptualization. J. Lucas Mckay: conceptualization, writing – original draft, writing – review and editing, investigation, data curation, visualization. Stewart A. Factor: conceptualization, funding acquisition, writing – original draft, writing – review and editing, investigation, supervision, data curation, methodology, project administration, visualization. Jeanne Powell: writing – original draft, writing – review and editing, visualization. Daniel S. Peterson: investigation, formal analysis, writing – original draft, writing – review and editing, data curation, visualization. Andrew S. Monaghan: investigation, formal analysis, writing – review and editing, visualization, data curation.

Funding

This work was supported by the Parkinson's Foundation, The Curtis Family Fund, The Sartain Lanier Family Foundation, CS Foundation, Miracle for Mom Foundation, The McCamish Parkinson's Disease Innovation Program at Georgia Institute of Technology, and National Institutes of Health (1R01AG086533).

Ethics Statement

As this is an Opinion paper (Editorial), no new data were created or analyzed. Ethical approval was not required.

Conflicts of Interest

Stewart A. Factor has received honoraria from Biogen, Takeda, IQVIA; grants from the Sun Pharmaceuticals Advanced Research Company, Aspen, Neurocrine, Rho Inc., CHDI Foundation, Michael J. Fox Foundation, NIH 1 P50 NS123103‐01, NIH 1R01NS125294‐01, Parkinson Foundation; and royalties from Demos, Blackwell Futura, Springer for textbooks, and UpToDate. Daniel S. Peterson has received grant funding under 1R01AG086533. The other authors have no conflicts of interest.

Supporting information

Data S1: Cognitive performance by FOG levodopa response subtype: Individual study effect sizes.

Figure S2: Copy of poster from October 2018: Cognitive correlates of levodopa unresponsive versus responsive freezing of gait subtypes. Reference 45.

Figure S3: Copy of poster from October 2018: Variation in anxiety and depression with freezing of gait subtype in Parkinson's disease.

Factor, S. A. , Weinshenker D., Powell J., Monaghan A. S., Peterson D. S., and Mckay J. L.. 2026. “Freezing of Gait Levodopa Response Pattern in Parkinson's Disease Provides Clues to Pathophysiology.” European Journal of Neuroscience 64, no. 1: e70610. 10.1111/ejn.70610.

Data Availability Statement

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.