GRAPHICAL ABSTRACT

INTRODUCTION

Parkinson’s disease (PD) is a neurodegenerative disease in older adults associated with debilitating motor and nonmotor symptoms [1]. Patients with PD (PwP) experience balance and gait deficits in advanced stages, including freezing of gait (FoG) and falls, which adversely affect their ability to safely carry out daily activities [1]. Cognitive deficits are also common throughout the progression of PD and can impact patients’ attention, memory, and executive functioning [2]. Cognitive deficits may interfere with a patient’s ability to maintain gait and balance, especially in situations that differ from normal forward walking, such as turning in a small room or hallway or taking a step backward to move out of someone’s way [3]. These movements are not as automatic as normal forward walking, partly because they rely more on the proprioceptive and vestibular systems rather than the visual system [4].

In PwP, gait impairments may be influenced not only by the disease itself but also by age-related changes, such as increased sedentarism, sensory deficits, and decreased muscle strength, which impact balance and coordination [5]. Many PwP may also have difficulty with more complex movements, such as turning, side-stepping, or walking backward [3]. Retro-walking, another name for backward walking, requires more attention, proprioception, and vestibular input than normal forward walking does to maintain balance and coordination due to unreliable visual input [6]. Although most people exhibit increased difficulties with retro-walking versus forward walking, previous research by Hackney and Earhart [3] showed significant deficits in retro-walking PwP at a magnitude that surpassed the deficits seen in forward walking. This finding suggests that PwP may have even more pronounced deficits during retro-walking versus forward walking than individuals without PD. Additionally, walking requires increased executive demands and prefrontal input as we age [7], which becomes even more pronounced in those with PD [8]. Retro-walking has been used in physical therapy and rehabilitation settings to increase gait speed [9] and improve balance [10] in PwP, individuals affected by stroke, and other patients. In a study of patients with mild-to-moderate PD, retro-walking training for eight weeks improved normal gait speed, cadence, and stride length more than did forward walking training alone for the same time period, suggesting that retro-walking training may be more beneficial for improving gait parameters than forward walking alone [9]. Studying retro-walking in conjunction with forward walking provides a better picture of daily balance and walking challenges for PwP because it considers movements that are not easily assessed by forward walking alone but may be crucial to a patient’s daily routine.

Both motor and nonmotor impairments are refractory to dopaminergic treatment at some point in disease progression, implying a nondopaminergic pathology or additional system involvement in these deficits [2]. The cholinergic system is a prominent system, with deficits identified in symptoms of PD such as dementia, falling, and FoG [11]. Prior research utilizing the vesicular acetylcholine transporter (VAChT) [¹⁸F]fluoroethoxybenzovesamicol ([¹⁸F]FEOBV) positron emission tomography (PET) tracer revealed cholinergic terminal deficits connected to isolated episodic postural instability and gait disturbances (PIGD) motor features, such as falls and FoG, in regions of the thalamus, hippocampus, striatum, amygdala, and specific cortical regions [12]. These findings point to a possible framework of a system-level disorder in PwP, with cardinal motor features elicited from deficits in and interconnections between various components of the basal ganglia–cortex–thalamus-cerebellum system, rather than from simple dopaminergic dysfunction of the basal ganglia [13].

Additionally, denervation of the cholinergic system is strongly associated with impairments in attention, memory, and language in PwP [2]. This denervation may be partly responsible for cognitive impairment in PwP, which may also play a role in PIGD symptoms. Reduced executive performance in PwP has been linked to gait impairment, implying the need for cognitive and motor integration in gait and balance tasks [14]. Taken together, denervation of the cholinergic system in PwP may help explain the framework of impaired cognitive integration, particularly motor-attentional integration. Given its demand for nonautomated gait and postural stability, retro-walking may be an avenue to examine that integration as it relates to cholinergic vulnerability in PwP.

The objective of this study was to investigate the cholinergic substrate involved in forward and retro-walking in PwP. We examined regional cerebral VAChT [¹⁸F]FEOBV PET binding and its associations with forward walking and retro-walking times using whole-brain voxel-based correlation analyses. We hypothesized that retro-walking relies more heavily on the integrity of the cholinergic system, as it likely reflects a more complex network involving balance and cognitive control than forward walking does.

MATERIALS & METHODS

Protocol approval and subject consent

This study (ClinicalTrials.gov Identifier: NCT05459753) was reviewed and approved by the Institutional Review Boards of the University of Michigan and Ann Arbor Veterans Affairs Healthcare System (University of Michigan study HUM00197590). The study was performed in accordance with the Declaration of Helsinki. All participants provided written informed consent prior to their participation.

Subjects

We retrospectively collected data from PwP who underwent [¹⁸F]FEOBV brain PET imaging and both retro-walking and forward-walking gait assessments. All studies from which we retrospectively collected data shared similar inclusion and exclusion criteria. The inclusion criteria were as follows: participants had received a diagnosis of PD according to the UK Brain Bank diagnostic criteria [15], were over 50 years old (30 males, 14 females; mean age=68.6±4.9 years), and had mild-to-moderate disease severity as defined by Hoehn and Yahr (H&Y) stages 1–3. Among the 44 PwP retrospectively included, 20 had confirmation with both [¹¹C]DTBZ and [¹¹C]PE2I, 11 had confirmation with [¹¹C]DTBZ only, 10 had confirmation with [¹¹C]PE2I only, and 3 did not undergo a presynaptic dopaminergic brain scan. Dementia was an exclusion criterion, defined according to the criteria of Emre et al. [16] Other exclusion criteria included a history of severe symptomatic leg or back musculoskeletal pain, the use of cholinergic or anticholinergic medications (such as donepezil, galantamine, rivastigmine, benztropine, or trihexyphenidyl), a history of stroke or other neurological conditions with residual sensorimotor deficits, and claustrophobia or contraindications for MRI.

We retrospectively retrieved the maximum number of control participants available with the required assessments to z score the main outcome variables: forward- and retro-walking performance and neuropsychological test scores. For gait normalization, we included 42 controls with walking assessments, matched to patients by age, height, weight, and sex (age: 57–78 years; sex: 20M/22F; height: 148.6–182.9 cm; weight: 48.7–127.0 kg). For neuropsychological normalization, we used 44 controls with complete cognitive data, matched by age, education, and sex (age: 50–82 years; sex: 30M/14F; education: 12–20 years).

Clinical assessments and outcomes of interest

Subjects completed motor testing in the dopaminergic medication “off” state after withholding their morning dose of dopaminergic medication, if applicable. Gait testing included 8.5 m normal forward walking and 8.5 m retro-walking. Forward and retro-walking were measured in seconds; therefore, higher values indicate slower (worse) performance. We z scored these variables relative to a healthy control cohort (n=42), adjusting for height, weight, age, and sex. Z scoring was performed using a residualization-based approach previously described [17] to correct for height, weight, age, and sex. The resulting z scores were used in all analyses.

The Movement Disorder Society-revised Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) [18] motor examination was performed by a certified staff member when the patient was in the dopaminergic medication “off” state. Using the MDS-UPDRS items, PIGD scores were derived according to Stebbins et al.’s guidelines [19]. To further study PIGD features, focusing on balance and dynamic gait, we also considered the Mini-Balance Evaluation Systems Test (MiniBESTest) total and subscale scores, an assessment that captures dynamic balance and gait functions using 14 items grouped into different balance domain subscores [20]. These subscores included anticipatory postural control, reactive postural control, dynamic gait, and sensory orientation. The participants also completed cognitive measures while on their dopaminergic medications, including the Montreal Cognitive Assessment (MoCA; permission was granted to use the MoCA in our research) [21] and the Parkinson’s Disease Cognitive Rating Scale (PD-CRS) [22]. Additionally, participants completed questionnaires, including the Short Activities-Specific Balance Confidence (SABC) scale [23], PD Questionnaire (PDQ)-39 [24], Instrumental Activities of Daily Living (IADL) [25], and the PD Cognitive Function Rating Scale (PD-CFRS) [26].

To investigate the contribution of cognition, we retrieved neuropsychological test data for whom it was available. Each test was z scored with the same procedure as described for gait outcome measures using a control cohort matched for age, sex, and education (n=44). Z scores for tests where a higher value indicated poorer performance (Stroop and Trail Making Test times) were multiplied by -1 to ensure that a higher z score value indicated better performance across all tests. We then derived five cognitive domain scores by averaging the z scores obtained across the individual tests corresponding to each domain: attention (Delis–Kaplan Executive Function System [DKEFS] Stroop 2 [word reading] time, Trail Making Test 2 [number sequencing] time, and Wechsler’s Adult Intelligence Scale digit symbol, and digit back task scores), executive function (DKEFS Stroop 4 [inhibition/switching] time, Trail Making Test 4 [number/letter switching] time, and Wechsler’s matrix reasoning score), language (verbal fluency animal naming test score), memory (California Verbal Learning Test short-term, long-term memory scores, and Wechsler’s Memory Scale total score), and visual processing (Judgment of Line Orientation test total score).

Imaging acquisition and preprocessing

Brain images were acquired in the dopaminergic “on” state, if applicable. PET imaging using the [¹⁸F]FEOBV PET tracer was performed in 3D imaging mode with a Biograph 6 TruPoint PET/CT scanner (Siemens Molecular Imaging, Inc.), and brain MRI was performed using a 3 Tesla Philips Achieva system (Philips) as previously reported [27]. The [¹⁸F]FEOBV PET tracer was prepared as previously described [28]. Beginning three hours after an intravenous bolus dose injection of 8 mCi [¹⁸F]FEOBV PET tracer, delayed dynamic imaging was performed over 30 minutes (in six 5-minute frames) [29]. Using MRI-PET registration statistical parametric mapping (SPM) software (SPM12; Welcome Trust Centre for Neuroimaging, University College, London, England [https://www.fil.ion.ucl.ac.uk/spm/software/spm12/]), images were corrected for any scatter and/or motion to limit image abnormalities [27]. [¹⁸F]FEOBV PET parametric images were obtained using a supratentorial white matter reference approach as previously reported [30,31]. Initially, the images from MRI were segmented to define the cortical, subcortical gray matter and white matter volumes of interest using the FreeSurfer software suite version 7.4.1, which is documented and freely available for download online (http://surfer.nmr.mgh.harvard.edu/). To minimize motion artifacts, PET frames were aligned within each subject using rigid-body transformation. Parametric [¹⁸F]FEOBV binding images were generated following a previously validated method [27,32]. To ensure accurate image alignment, a two-step registration process was employed. The images from MRI were first resampled to match the PET resolution. The average of the six delayed PET frames was subsequently rigidly coregistered to the corresponding subject’s images using SPM12, with visual confirmation for accuracy. The reference region (supratentorial white matter) was defined by excluding voxels below the ventricles and near cortical areas, followed by erosion. Parametric images were then generated by dividing voxel values in the respective activity images by the mean activity in this reference region.

Whole-brain VAChT PET voxelwise correlation analyses and forward and retro-walking

We ran two voxelwise regression models using forward- and retro-walking time z scores as independent variables and [¹⁸F] FEOBV brain PET data as the dependent variables. The levodopa equivalent dose (LED) and disease duration were included as covariates of no interest to control for their effects. The significance threshold was set at p<0.05 (uncorrected), with familywise error (FWE) correction applied at the cluster level (p<0.001). Any significant cluster was used as a binary mask to extract [¹⁸F] FEOBV uptake from MNI space parametric images with Muller-Gartner partial-volume correction.

Statistical analysis

To minimize violations of regression assumptions, we inspected the distribution of outcome variables—[¹⁸F]FEOBV uptake and retro-/forward-walking time z scores—and applied appropriate transformations for skewed variables before analysis.

We conducted a series of linear regression models to investigate the associations between cognitive and PIGD-related measures—including PIGD scores (Stebbins et al. [19]) and the Mini-BESTest total and subscale scores—and two main outcomes: 1) [¹⁸F]FEOBV uptake in the cluster identified through voxelwise analysis and 2) retro- and forward-walking time z scores. All the models included LED and disease duration as covariates. For each regression model, we examined residual distributions and tested for heteroscedasticity. When assumptions were violated, we applied robust linear regression.

Predictors that were significantly associated with both [¹⁸F] FEOBV uptake and retro-/forward-walking performance were then entered into the mediation models. These models tested whether cognitive and/or PIGD-related measures (M) mediated the relationship between cholinergic terminal density (X) and walking performance (Y). Mediation analyses were performed using the open-source R programming language “mediation” package, with 1,000 bootstrap iterations applied as a strategy to ensure stable parameter estimation and robustness.

RESULTS

Clinical and demographic information

Table 1 presents the clinical and demographic characteristics of the participants included in the present study. Thirty-seven PwP were taking dopaminergic medications for their PD, with a mean LED of 713.4±473.2 mg. The mean disease duration was 7.9 years (±6.8 years), and the mean modified H&Y stage was 2.34 (±0.48; mild to moderate disease severity). The mean motor examination score on part three of the MDS-UPDRS was 46.4±13.2. PwP walked nearly two times slower during retro-walking than during forward walking: 16.33±6.87 seconds vs. 8.45±2.27 seconds.

Cognitive and PIGD-related predictors of forward-and retro-walking times

The forward- and retro-walking time z scores showed positive skewness (forward: 2.41; retro: 1.03) and were log-transformed to achieve normality.

Forward walking predictors

Slower forward-walking (i.e., greater walking time) was significantly associated with lower scores in the attention (β=-0.50, p<0.001) and visual (β=-0.48, p<0.001) cognitive domains. Additionally, higher PIGD scores (β=0.40, p=0.002; robust regression) and lower MiniBESTest total scores (β=-0.31, p=0.036) were also significant predictors of increased walking time. Among the MiniBESTest subdomains, only the dynamic gait subscore reached significance (β=-0.42, p=0.003) (Table 2).

Retro-walking predictors

Slower retro-walking (i.e., greater walking time) was significantly associated with lower scores in the attention domain (β=-0.49, p=0.002) and lower MiniBESTest total scores (β=-0.53, p<0.001). Among the MiniBESTest subdomains, both reactive postural control (β=-0.49, p=0.001) and dynamic gait (β=-0.46, p=0.003) subscores were significantly associated with slower retro-walking. PIGD scores (β=0.31, p=0.062) showed a trend-level association (Table 3).

[¹⁸F]FEOBV PET voxelwise results

Forward-walking time did not show a significant voxelwise association with [¹⁸F]FEOBV brain uptake and was therefore not included in the subsequent mediation analyses. In contrast, slower retro-walking (i.e., greater walking time) was significantly associated with lower [¹⁸F]FEOBV binding in a subcortical–frontal–temporal cluster encompassing the orbitofrontal cortex (right more than left), anterior cingulate, bilateral insula, bilateral prefrontal cortex, left temporal pole, left dorsolateral prefrontal cortex, left caudate nucleus, putamen, and thalamus (Figure 1).

The statistical threshold for voxelwise analysis was set at p<0.05 (uncorrected), with a cluster-level FWE correction.

Cognitive and PIGD-related predictors of [¹⁸F]FEOBV cluster uptake

To explore potential determinants of cholinergic terminal density in the subcortical–frontal–temporal cluster identified through voxelwise analysis (and associated with retro-walking performance; see paragraph above), we ran a set of regression models. Greater [¹⁸F]FEOBV uptake was significantly associated with better attention domain z scores (β=0.44, p=0.006) and higher MiniBESTest total scores (β=0.39, p=0.012). Among the MiniBESTest subdomains, better reactive postural control (β=0.40, p=0.011), anticipatory postural control (β=0.32, p=0.043), and sensory orientation (β=0.39, p=0.010) subscores were significantly associated with greater tracer uptake in these regions (Table 4).

Results of the mediation analyses

Only attention z scores, MiniBESTest total scores, and reactive postural control subscores were included in the mediation models (Figure 2), as they were the only variables significantly associated with both [¹⁸F]FEOBV binding and retro-walking performance.

The association between [¹⁸F]FEOBV binding in the subcortical–frontal–temporal cluster and retro-walking performance was mediated by attentional functioning. The indirect effect via attention was significant (β=-0.16, 95% confidence interval; CI [-0.40, -0.02], p=0.034), making the direct effect marginally significant (β=-0.28, 95% CI [-0.60, 0.00], p=0.052). Attention accounted for 36% of the total effect (95% CI [0.04, 1.01], p=0.034).

A similar mediation pattern was found for the MiniBESTest total score. Both the indirect (β=-0.17, 95% CI [-0.36, -0.02], p=0.022) and direct (β=-0.24, 95% CI [-0.54, -0.01], p=0.034) effects were significant, with 42% of the total effect mediated (95% CI [0.07, 0.93], p=0.020). Notably, the reactive postural control subscore alone fully mediated the association (β=-0.15, 95% CI [-0.33, -0.02], p=0.014), accounting for 38% of the effect (95% CI [0.06, 1.03], p=0.016), making the direct effect marginally significant (β=-0.25, 95% CI [-0.57, 0.01], p=0.060).

DISCUSSION

Our study revealed that patients with mild to moderately severe PD required nearly twice as much time to complete the same distance retro-walking compared to forward walking, which is in line with previous research reporting that all participants experienced deficits in retro-walking compared with forward walking, with PwP experiencing deficits at a greater magnitude [3]. This finding indicates that retro-walking requires significantly more processing of neural resources than does forward walking. Retro-walking is unique in that visual sensory input has become unreliable, and effective postural control becomes more dependent on multisensory (proprioceptive, somatosensory, and vestibular) and cognitive integration with motor programming.

Linear regression analyses revealed that greater deficits in the attentional cognitive domain in PwP significantly predict slower forward walking, which suggests that normal walking involves similar mechanisms to the attentional-motor integration required to perform two tasks at once. Slower forward walking was also significantly associated with poorer performance in the visual cognitive domain, which includes both visuospatial and visuoconstructive abilities. This finding supports the role of visual processing as a fundamental integrative input for gait [33]. In contrast, visual cognition was not associated with retro-walking performance in keeping with the inability to rely on visual-related inputs and cognitive processing when walking backward. Forward walking performance was also associated with higher PIGD scores, as measured using the Stebbins et al.’s method [19], and with lower MiniBESTest total scores—particularly the dynamic gait subscore. This association may reflect the fact that both the PIGD score and the MiniBESTest dynamic gait assessment include functions that are relevant for forward walking.

Retro-walking time was significantly associated with greater attentional impairment (but not visual deficits) and lower MiniBESTest scores—especially the dynamic gait and reactive postural control subscores. In contrast, PIGD scores showed only a trend-level association, likely due to the mild PIGD severity in our cohort (approximately 27% classified as PIGD predominant). Balance impairments such as reactive postural control typically emerge later and may not be well captured by the MDS-UPDRS in early stages, which may explain why the MiniBESTest measures were more strongly linked to retro-walking performance. Reactive postural control is the ability to recover to an upright posture following an external perturbation, such as being pushed [33], and is indicative of PIGD symptoms in PwP. This finding suggests that the ability to recover balance is associated with retro-walking abilities independent of forward walking ability, which sheds light on the additional balancing capabilities required to successfully remain upright during retro-walking. Retro-walking requires increased balance coordination to maintain speeds similar to those of forward walking. The majority of falls in the elderly population occur from backward perturbations, which may be related to deficits in the ability to take a backward step, such as when turning or sitting down in a chair [34], pointing to the link between postural control and the ability to retro-walk. In the aging population, retro-walking measures are also more sensitive at differentiating fallers from non-fallers than are forward walking measures [35]. We recently published a paper investigating the association between retro-walking and falls in a larger cohort of PwP [34]. This cohort included the current sample plus an additional 34 patients, allowing us to increase the number of subjects since [¹⁸F]FEOBV imaging was not part of that study. In this work, we demonstrated that poorer retro-walking performance was indeed associated with falls. These results further support previous reports linking falls, postural instability, and retro-walking ability.

In the present study, we found that retro-walking time was associated with cholinergic nerve density in a topographically defined set of regions forming a subcortical–frontal–temporal cluster. This cluster included the orbitofrontal cortex, bilateral middle frontal cortex, anterior cingulate cortex, right gyrus rectus, left basal forebrain cholinergic regions, bilateral insular cortex, right frontal-temporal operculum, left temporal pole, left amygdala, left caudate nucleus, putamen and globus pallidus, and left dorsolateral prefrontal cortex (Figure 1). Forward gait time did not show any significant associations with cholinergic innervation, suggesting that in early or mild disease stages, it may rely less on cholinergic integrity, possibly reflecting contributions from the dopaminergic or other monoaminergic systems. While dopaminergic involvement could not be directly assessed in our cohort due to heterogeneity in dopaminergic nerve terminal PET acquisitions, future studies should explore the interplay between the cholinergic and dopaminergic systems, which likely contributes complementarily to gait regulation.

The set of brain regions in which cholinergic synaptic density was associated with retro-walking time showed substantial overlap with cholinergic changes previously reported to be involved in reactive postural control [35]. Specifically, this includes the anterior cingulate cortex, prefrontal cortices, caudate nucleus, and insular cortex—areas that have also been implicated in functional imaging studies examining postural control [36]. This convergence suggests that these regions may play a key role in retro-walking performance, likely due to the task’s reliance on balance-related abilities, particularly reactive postural control. Given that gait relies on complex multisensory and higher-order integrative processes [33], particularly in conditions where visual cues are limited or unavailable (e.g., walking backward), the complexity and demands of retro-walking may more strongly engage neural circuits involved in rapid postural adjustments. In particular, this may include regions associated with vestibular processing, which are critical for maintaining balance and spatial orientation in the absence of reliable visual input [37]. Part of the cholinergic regional contribution to retro-walking time also demonstrated specificity for attention-related cognitive performance [38]. In particular, the cingulate cortex, insula/operculum, temporal pole, and medial thalamic regions were implicated. These areas are known to support higher-order cognitive functions, including attentional control, stimulus detection, and environmental awareness. This implies that the ability to flexibly switch attention and monitor sensory inputs becomes especially critical in situations where visual cues are unreliable or absent. The role of this brain region in balance and attentional brain mechanisms was further supported by our mediation analyses, which revealed that both attentional performance and reactive postural control fully mediated the associations between cholinergic nerve density in the subcortical–frontal–temporal cluster and retro-walking. These results suggest that high-order integrative processes—particularly those involving attentional shifting and postural control—constitute key pathways through which the cholinergic system influences retro-walking ability.

An apparent missing brain region in our findings was the cerebellum. Despite its key role in balance and coordination [35,36], it was not significantly associated with retro-walking performance. This may be because retro-walking, in this context, relies more on attentional and reactive postural control resources that may involve the frontotemporal cortices, striatum (caudate nucleus), and thalamus [35]. Alternatively, the observed pattern may reflect a convergence of attentional, balance, and vestibular functions, with the identified regions serving integrative roles across these domains rather than representing isolated systems.

Finally, on a translational note, our results may provide neurobiological rationale for the evidence supporting retro-walking as a therapeutic intervention in PwP [9]. Retro-walking has been shown to improve lower limb strength, balance, and gait parameters more effectively than forward walking alone [39,40]. By identifying cholinergic and attentional-postural control networks associated with retro-walking performance, our findings help explain these clinical benefits and suggest that future research is warranted to explore retro-walking as a targeted rehabilitation strategy engaging specific neural systems.

This study has several limitations, including a small sample size, a population with primarily mild-to-moderate disease, and heterogeneous measures of presynaptic dopaminergic function. Future studies should examine whether retro-walking training in PwP improves gait and balance parameters, such as reactive postural control, attentional motor integration tasks, and balance confidence. By doing so, clinicians may be able to incorporate retro-walking training into patient care, which could improve patients’ strength and reduce the risk of fall-related injury. These findings could benefit both patients and clinicians in enhancing care and reducing fall-related burdens on patients and caregivers. Future research should also examine how dopaminergic system changes are related to retro-walking ability to better understand the interconnections between dopamine, acetylcholine, retro-walking, and PIGD motor features in PwP.

Notes

Conflicts of Interest

The authors have no financial conflicts of interest.

Funding Statement

This work was funded by National Institutes of Health (R01 AG073100, R01 NS070856, P50 NS091856, P50 NS123067), Department of Veterans Affairs grant (I01 RX001631), the Michael J. Fox Foundation, and the Parkinson’s Foundation.

Acknowledgments

We thank study participants, cyclotron operators and study staff for their help with this study.

Author Contributions

Conceptualization: Nicolaas Ida Bohnen, Miriam van Emde Boas, Chatkaew Pongmala. Data curation: Alexis Griggs, Giulia Carli, Taylor Brown, Stiven Roytman. Formal analysis: Alexis Griggs, Giulia Carli, Taylor Brown, Prabesh Kanel, Stiven Roytman. Funding acquisition: Nicolaas Ida Bohnen. Investigation: Alexis Griggs, Giulia Carli, Taylor Brown, Prabesh Kanel, Stiven Roytman, Chatkaew Pongmala, Nicolaas Ida Bohnen. Methodology: Nicolaas Ida Bohnen, Chatkaew Pongmala, Miriam van Emde Boas. Project administration: Nicolaas Ida Bohnen. Resources: Nicolaas Ida Bohnen. Software: Giulia Carli, Stiven Roytman. Supervision: Nicolaas Ida Bohnen. Visualization: Giulia Carli, Stiven Roytman. Writing—original draft: all authors. Writing—review & editing: all authors.

Figure 1.

Voxelwise correlation between the retro-walking time z scores and [¹⁸F]fluoroethoxybenzovesamicol positron emission tomography brain images. The threshold was set at p uncorrected <0.05, cluster-level familywise error correction p<0.001. The significant cluster is presented in neurological convention and overimposed on the MNI T1 template.

Figure 2.

Mediation analysis results. The mediation model parameters were estimated using 1,000 bootstrap iterations. The levodopa equivalent dose and disease duration were included as covariates in all the models. MiniBESTest, Mini-Balance Evaluation Systems Test; [¹⁸F]FEOBV, [¹⁸F]fluoroethoxybenzovesamicol; ACME, average causal mediation effect; ACE, average direct effect.

REFERENCES

• 1. Muslimovic D, Post B, Speelman JD, Schmand B, de Haan RJ; CARPA Study Group. Determinants of disability and quality of life in mild to moderate Parkinson disease. Neurology 2008;70:2241–2247.ArticlePubMed
• 2. Bohnen NI, Yarnall AJ, Weil RS, Moro E, Moehle MS, Borghammer P, et al. Cholinergic system changes in Parkinson’s disease: emerging therapeutic approaches. Lancet Neurol 2022;21:381–392.ArticlePubMedPMC
• 3. Hackney ME, Earhart GM. Backward walking in Parkinson’s disease. Mov Disord 2009;24:218–223.ArticlePubMedPMC
• 4. Vaugoyeau M, Hakam H, Azulay JP. Proprioceptive impairment and postural orientation control in Parkinson’s disease. Hum Mov Sci 2011;30:405–414.ArticlePubMed
• 5. Allen NE, Canning CG, Sherrington C, Fung VS. Bradykinesia, muscle weakness and reduced muscle power in Parkinson’s disease. Mov Disord 2009;24:1344–1351.ArticlePubMedPDF
• 6. Thomas MA, Fast A. One step forward and two steps back: the dangers of walking backwards in therapy. Am J Phys Med Rehabil 2000;79:459–461.PubMed
• 7. Nóbrega-Sousa P, Gobbi LTB, Orcioli-Silva D, Conceição NRD, Beretta VS, Vitório R. Prefrontal cortex activity during walking: effects of aging and associations with gait and executive function. Neurorehabil Neural Repair 2020;34:915–924.ArticlePubMedPDF
• 8. Muslimovic D, Post B, Speelman JD, Schmand B. Cognitive profile of patients with newly diagnosed Parkinson disease. Neurology 2005;65:1239–1245.ArticlePubMed
• 9. Grobbelaar R, Venter R, Welman KE. Backward compared to forward over ground gait retraining have additional benefits for gait in individuals with mild to moderate Parkinson’s disease: a randomized controlled trial. Gait Posture 2017;58:294–299.ArticlePubMed
• 10. Kim K, Lee S, Lee K. Effects of progressive body weight support treadmill forward and backward walking training on stroke patients’ affected side lower extremity’s walking ability. J Phys Ther Sci 2014;26:1923–1927.ArticlePubMedPMC
• 11. Bohnen NI, Kanel P, Müller MLTM. Molecular imaging of the cholinergic system in Parkinson’s disease. Int Rev Neurobiol 2018;141:211–250.ArticlePubMedPMC
• 12. Bohnen NI, Kanel P, Zhou Z, Koeppe RA, Frey KA, Dauer WT, et al. Cholinergic system changes of falls and freezing of gait in Parkinson’s disease. Ann Neurol 2019;85:538–549.ArticlePubMedPMCPDF
• 13. Caligiore D, Helmich RC, Hallett M, Moustafa AA, Timmermann L, Toni I, et al. Parkinson’s disease as a system-level disorder. NPJ Parkinsons Dis 2016;2:16025.ArticlePubMedPMCPDF
• 14. Correno MB, Hansen C, Chardon M, Milane T, Bianchini E, Vuillerme N. Association between backward walking and cognition in Parkinson disease: a systematic review. Int J Environ Res Public Health 2022;19:12810.ArticlePubMedPMC
• 15. Hughes AJ, Daniel SE, Lees AJ. Improved accuracy of clinical diagnosis of Lewy body Parkinson’s disease. Neurology 2001;57:1497–1499.ArticlePubMed
• 16. Emre M, Aarsland D, Brown R, Burn DJ, Duyckaerts C, Mizuno Y, et al. Clinical diagnostic criteria for dementia associated with Parkinson’s disease. Mov Disord 2007;22:1689–1707.ArticlePubMed
• 17. Bohnen NI, Roytman S, van der Zee S, Carli G, Michalakis F, Luker A, et al. A multicenter longitudinal study of cholinergic subgroups in Parkinson disease. Nat Commun 2025;16:5655.ArticlePubMedPMCPDF
• 18. Goetz CG, Tilley BC, Shaftman SR, Stebbins GT, Fahn S, Martinez-Martin P, et al. Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS): scale presentation and clinimetric testing results. Mov Disord 2008;23:2129–2170.PubMed
• 19. Stebbins GT, Goetz CG, Burn DJ, Jankovic J, Khoo TK, Tilley BC. How to identify tremor dominant and postural instability/gait difficulty groups with the movement disorder society unified Parkinson’s disease rating scale: comparison with the unified Parkinson’s disease rating scale. Mov Disord 2013;28:668–670.ArticlePubMed
• 20. Franchignoni F, Horak F, Godi M, Nardone A, Giordano A. Using psychometric techniques to improve the Balance Evaluation Systems Test: the mini-BESTest. J Rehabil Med 2010;42:323–331.ArticlePubMedPMC
• 21. Smith T, Gildeh N, Holmes C. The Montreal Cognitive Assessment: validity and utility in a memory clinic setting. Can J Psychiatry 2007;52:329–332.ArticlePubMedPDF
• 22. Pagonabarraga J, Kulisevsky J, Llebaria G, García-Sánchez C, Pascual-Sedano B, Gironell A. Parkinson’s disease-cognitive rating scale: a new cognitive scale specific for Parkinson’s disease. Mov Disord 2008;23:998–1005.ArticlePubMedPDF
• 23. Peretz C, Herman T, Hausdorff JM, Giladi N. Assessing fear of falling: can a short version of the Activities-specific Balance Confidence scale be useful? Mov Disord 2006;21:2101–2105.ArticlePubMed
• 24. Bushnell DM, Martin ML. Quality of life and Parkinson’s disease: translation and validation of the US Parkinson’s Disease Questionnaire (PDQ-39). Qual Life Res 1999;8:345–350.ArticlePubMedPDF
• 25. Lawton MP, Brody EM. Assessment of older people: self-maintaining and instrumental activities of daily living. Gerontologist 1969;9:179–186.ArticlePubMed
• 26. Kulisevsky J, Fernández de Bobadilla R, Pagonabarraga J, Martínez-Horta S, Campolongo A, García-Sánchez C, et al. Measuring functional impact of cognitive impairment: validation of the Parkinson’s disease cognitive functional rating scale. Parkinsonism Relat Disord 2013;19:812–817.ArticlePubMed
• 27. Bohnen NI, Kanel P, Koeppe RA, Sanchez-Catasus CA, Frey KA, Scott P, et al. Regional cerebral cholinergic nerve terminal integrity and cardinal motor features in Parkinson’s disease. Brain Commun 2021;3:fcab109.ArticlePubMedPMCPDF
• 28. Shao X, Hoareau R, Hockley BG, Tluczek LJM, Henderson BD, Padgett HC, et al. Highlighting the versatility of the tracerlab synthesis modules. Part 1: fully automated production of [18F]labelled radiopharmaceuticals using a Tracerlab FXFN. J Labelled Comp Radiopharm 2011;54:292–307.ArticlePubMedPMC
• 29. Petrou M, Frey KA, Kilbourn MR, Scott PJ, Raffel DM, Bohnen NI, et al. In vivo imaging of human cholinergic nerve terminals with (-)-5-(18)F-fluoroethoxybenzovesamicol: biodistribution, dosimetry, and tracer kinetic analyses. J Nucl Med 2014;55:396–404.ArticlePubMed
• 30. Aghourian M, Legault-Denis C, Soucy JP, Rosa-Neto P, Gauthier S, Kostikov A, et al. Quantification of brain cholinergic denervation in Alzheimer’s disease using PET imaging with [18F]-FEOBV. Mol Psychiatry 2017;22:1531–1538.ArticlePubMedPDF
• 31. Nejad-Davarani S, Koeppe RA, Albin RL, Frey KA, Müller MLTM, Bohnen NI. Quantification of brain cholinergic denervation in dementia with Lewy bodies using PET imaging with [18F]-FEOBV. Mol Psychiatry 2019;24:322–327.ArticlePubMedPMCPDF
• 32. Kanel P, van der Zee S, Sanchez-Catasus CA, Koeppe RA, Scott PJH, van Laar T, et al. Cerebral topography of vesicular cholinergic transporter changes in neurologically intact adults: a [18F]FEOBV PET study. Aging Brain 2022;2:100039.ArticlePubMedPMC
• 33. Roytman S, Paalanen N, Carli G, Marusic U, Kanel P, van Laar T, et al. Multisensory mechanisms of gait and balance in Parkinson’s disease: an integrative review. Neural Regen Res 2025;20:82–92.ArticlePubMedPMC
• 34. Pongmala C, Stonsaovapak C, van Emde Boas M, Bohnen NI. Differential associations between retro walking versus forward walking, axial motor features and body composition in Parkinson’s disease. Parkinsonism Relat Disord 2025;137:107924.ArticlePubMedPMC
• 35. Bohnen NI, Marusic U, Roytman S, Paalanen N, Michalakis F, Brown T, et al. Dynamic balance and gait impairments in Parkinson’s disease: novel cholinergic patterns. Brain Commun 2024;6:fcae286.ArticlePubMedPMCPDF
• 36. Dijkstra BW, Bekkers EMJ, Gilat M, de Rond V, Hardwick RM, Nieuwboer A. Functional neuroimaging of human postural control: a systematic review with meta-analysis. Neurosci Biobehav Rev 2020;115:351–362.ArticlePubMed
• 37. Roytman S, Paalanen N, Griggs A, David S, Pongmala C, Koeppe RA, et al. Cholinergic system correlates of postural control changes in Parkinson’s disease freezers. Brain 2023;146:3243–3257.ArticlePubMedPMCPDF
• 38. van der Zee S, Müller MLTM, Kanel P, van Laar T, Bohnen NI. Cholinergic denervation patterns across cognitive domains in Parkinson’s disease. Mov Disord 2021;36:642–650.ArticlePubMedPMCPDF
• 39. Flynn TW, Soutas-Little RW. Mechanical power and muscle action during forward and backward running. J Orthop Sports Phys Ther 1993;17:108–112.ArticlePubMed
• 40. Hoogkamer W, Meyns P, Duysens J. Steps forward in understanding backward gait: from basic circuits to rehabilitation. Exerc Sport Sci Rev 2014;42:23–29.PubMed

Figure & Data

References

Citations

Citations to this article as recorded by  

Comments on this article