Table 2.

Brain imaging and other modalities and their association with the future development of FOG.

Authors, year	Study design	Sample size (n total, % who developed FOG)	FOG ascertainment	Duration of Follow-up (years)	Exposure Predictors	Main findings	Remarks and limitations	Total quality score
Chung, Lee et al.34	Retrospective cohort	n 268, FOG in 52 (19.4%)	Self-report	5.9 ± 1.6, (WMH-) 5.0 ± 1.4, (WMH + )	WMH visual rating scale scores	After adjusting for age, sex, DAT, and LED, WMH + group showed a higher risk of developing FOG (HR, 3.29; 95%CI, 1.79–6.05; p < 0.001) than the PD-WMH- group. (PD-WMH- n = 198; FOG in 27 (14%); in PD-WMH + n = 70; FOG in 25 (36%).	Ischemia classification system is not very accurate. Lack of DTI data. FOG was not assessed objectively. Individual variability in the intensity of vascular risk factors.	7
Chung, Yoo et al.52	Retrospective cohort	n 139, FOG in 38 (27.3%)	Self-report & clinical observation	At least 3 years	Age at disease onset, (DAT) availability, motor and non-motor symptoms.	Older-onset PD group had higher risk for FOG than the younger-onset group.	Patients’ subdivision into groups is somewhat arbitrary. FP-CIT PET may not be ideal surrogate marker for nigrostriatal dopaminergic degeneration. FOG was not assessed using objective gait measures. Atypical parkinsonism cases might be misdiagnosed as idiopathic PD. Due to the lack of follow up data clinical progression was assessed based on the longitudinal changes in LED, which might not accurately reflect the motor status.	8
Chung et al.88	Prospective cohort	n 205, FOG in 51 (24.9%)	Self-report & Clinical observation	6.31 ± 1.59 years	Motor reserve	Greater motor reserve estimates, as determined by initial motor deficits and striatal dopamine depletion, were associated with a lower risk for FOG.	Motor-symptom laterality and interhemispheric symmetry of dopamine loss were not taken into account. FOG retrospective determination was based on medical records, so mild cases can be missed. Disease duration could be measured with error. There were no follow-up assessments for motor deficits or striatal DAT depletion. The longitudinal changes in LED may not be the ideal measures for disease progression. DAT availability may not purely reflect the nigrostriatal dopaminergic denervation.	8
Chung, Yoo, Shin et al.89	Retrospective cohort	n 248, FOG in 64 (25.8%)	Self-report & clinical observation	6.00 ± 1.85 years in PDEPVS−Group 6.35 ± 2.13 years in PD-EPVS + group	High number (>10) of enlarged perivascular spaces in the basal ganglia (EPVS + ).	FOG was higher in the PD EPVS + group; FOG in 35% vs. 20.5% in the EPVS- group.	Visual scoring of BGPVS only provided qualitative estimates of the PVS extent. Enlarged BG-PVS were only assessed at the baseline evaluation. Minimal FOG might not be detected early. Direct measures of motor deficits were not collected in a longitudinal manner.	8
D’Cruz et al.90	Prospective cohort	2 studies: MRI- n 45, FOG in 9 (20%) f-MRI- n 41, FOG in 7 (17%)	Self-report	2 years	Structural MRI. Resting state fMRI Connectivity between sub-thalamic nuclei/cortical areas.	Converters had higher LEDD and higher MDS-UPDRS part I and total scores. Freezers and converters showed significant local shape inflations compared to non-converters in the right and left thalamus. Gender, LEDD, and cluster from left thalamus were significant predictors of conversion to FOG.	Low conversion rate- only 20%. Differences in disease stage. Results may relate to a particular FOG phenotype.	7
Djaldetti et al.35	Retrospective cohort	n 41, FOG in 15 (34.8%)	UPDRS part II	2 years	DAT binding in striatum, age at symptom onset, age at DaTSCAN, disease type, UPDRS part III score during “on” at baseline and end of follow-up, duration of levodopa treatment, dosage of levodopa at end of follow-up.	DAT uptake values were significantly lower in the putamen and striatum in patients with FOG.	Relatively small number of cases. Retrospective design. Cognitive tests and assessment for depression, anxiety, and other behavioral features were not administered. Most of the patients were already under treatment at the time of imaging.	8
Kim et al.36	Retrospective cohort	n 390, FOG in 143 (36.7%)	MDS-UPDRS	4 years	Baseline dopamine transporter (DAT) uptake in caudate nucleus and putamen.	DAT uptakes in the caudate nucleus and putamen predicted the development of FOG. Severe reduction of DAT uptake in caudate nucleus and putamen had a higher incidence of FOG than mild and moderate reduction. Male sex, higher PIGD score, and lower MoCA score were also significant predictors of FOG.	Initial DAT scan may not reflect a precise prediction of disease progression.	8
Kim et al.37	Prospective cohort	n 339, FOG in 142, (42%)	MDS-UPDRS	6 years	APOE ε4, motor, mood and cognitive symptoms	FOG in 45 out of 88 (51%) with positive APOE ε4 allele vs. 97 out of 251 (39%) in the negative group. CSF Aβ42, male sex, age at onset, MDS-UPDRS motor score, PIGD score, MoCA score, GDS score, and DAT uptakes in the caudate nucleus and putamen were significantly related to FOG.	Results may be different if amyloid levels are measured by PET imaging rather by CSF Aβ42. FOG may be underestimated after starting dopaminergic medications.	7
Kim, Lee et al.39	Prospective cohort	n 393, FOG in 136 (35%)	MDS-UPDRS	4 years	CSF parameters, Demographic and clinical data, motor and non-motor symptoms, DAT imaging.	CSF Aβ42 was associated with FOG. Male sex, MDS-UDPRS motor, PIGD score, MoCA score, and caudate DAT uptake were also significantly predictive of FOG. *The combined model integrating the PIGD score, caudate DAT uptake, and CSF Aβ42 achieved a better discriminative ability than any factor alone. Cumulative incidence of FOG: 16.5%, 21.0%, 28.2%, and 36.7% at the 1, 2, 3, and 4-year follow up, respectively.	The data does not measure the extent of amyloid pathology. Relatively short follow-up period. FOG might be misestimated through MDS-UPDRS. FOG may be underestimated after dopaminergic drugs initiation.	5
Kim and Jeon, 91	Retrospective cohort	n 361, FOG in 189 incidence rates of 22, 37, 48, and 63% at the 2-, 4-, 6- and 8-year follow-ups, respectively	MDS-UPDRS items 2.13 and 3.11	8 years	Nfl, demographics, motor and non-motor symptoms, DAT scan.	NfL levels are a novel risk factor for FOG.	FOG might be misestimated through self-report. FOG could not be differentiated according to medication states.	5
Li et al.92	Prospective cohort	n 40, FOG in 20, (50%) matched to no-FOG	NFOG-Q item 1	5 years	Topological organization of whole brain functional networks (fMRI).	The PD with FOG group exhibited decreased nodal centrality in the left middle frontal gyrus (MFG).	Relatively small sample size. FOG was self-reported rather than objectively measured. A small part of patients received optimized anti-PD medication at baseline, so FOG during the follow-up may have been interfered by pharmacological effects.	8
Li et al.93	Retrospective cohort	n 158, FOG in 66 (42%)	UPDRS items 2.13 OR item 3.11	5 years	Demographics, clinical and laboratory characteristics, MRI T1 imaging.	T1-weighted imaging predicts FOG. PD patients who developed FOG had decreased olfactory function, depression, a more severe disease degree, dysfunction of daily living and movement, postural instability, and gait difficulty at baseline. Men are more prone to FOG.	Only T1 MRI analysis was performed. A limited amount of clinical data. Different machine-learning methods might change the results in diverse cohorts. Participants were mainly from European and American populations.	5
Dadar et al.38	Prospective cohort	n 423, FOG in 177 (42%)	MDS-UPDRS items 2.13 and 3.11	5 years	T1w MRI: DBM for SN Atrophy. Segmentation of WMH. Putamen and caudate DAT levels. Cerebrospinal fluid (CSF) amyloid β.	WMH load mediated the impact of amyloid β on future FOG.	No consistent follow-up information available for all participants. Lack of gait characteristics. Limited spectrum of disease.	8
Sarasso et al.94	Prospective cohort	n 22, FOG in 11 (50%)	Observation by neurologist & FOG Questionnaire	2 years	Demographics, motor and non-motor symptoms, MRI biomarkers.	FOG-converters showed greater PIGD, more severe limb/axial rigidity and more frequent dyskinesia. Prediction model including dyskinesia, PIGD scores and parietal clustering coefficient were the best predictor of FOG conversion.	Study samples were small and matched for UPDRS-III. There was no longitudinal MRI data in healthy subjects that could control for the aging effects. 1.5 T MRI scanner was used, which has lower spatial resolution and BOLD signal to noise ratio resolution. PD patients were evaluated in ON status. A longer follow-up might improve the possibility to monitor FOG conversion. Prediction model might suffer from overfitting of data due to small sample size.	8
Wieler et al.95	Prospective cohort	n 19, FOG in 7 (37%)	UPDRS II	3 years	Nigral iron content (MRI), motor and non-motor symptoms.	FOG was associated with motor impairment at baseline (UPDRS III scores), time to complete a 14 m walk, Berg Balance scale and timed up and go, more rapid deterioration in motor function, higher levodopa equivalent dose at baseline and increased iron content in lateral SN pars compacta.	FOG might be underestimated since it is mostly self-reported. There was no differentiation between “ON” freezing and “OFF” freezing. Small sample.	6

SN substantia nigra, BG basal ganglia, DBM deformation-based morphometry, WMH white matter hyperintensities, DTI diffusion tensor imaging, EPVS enlarged perivascular spaces, NfL neurofilament light chain. For additional abbreviations please also see Table S1 in the supplement.