Patients were consecutively recruited throughout the year from outpatient clinics specializing in movement disorders at the Second Xiangya Hospital of Central South University. Before analysis, data completeness was assessed, and participants with missing MRI, serum 25-hydroxyvitamin D, or questionnaire data were excluded from the corresponding analyses. Because the proportion of missing data was minimal (<5%), no data imputation was performed. In the end, 34 patients with idiopathic PD were included in this cross-sectional study. All included patients met the movement disorder society (MDS) clinical diagnostic criteria for PD (Postuma et al., 2015). Patients were followed for a minimum of 12 months, with the diagnosis reassessed and confirmed at the end of the follow-up period. Exclusion criteria at enrollment included the diagnosis of secondary Parkinson-plus syndromes, parkinsonism, gene-related PD, use of vitamin D supplements, history of any chronic disease, or the use of medications that could potentially impact the absorption of vitamin D. Additionally, 21 healthy controls (HCs), matched to PD patients for age, sex, and years of education, were recruited from volunteers and the spouses of PD patients. All controls were cognitively normal, with no symptoms of PD or history of neurological diseases. Controls were excluded if they had a first- or second-degree relative with PD, and the remaining exclusion criteria were the same as those for the PD group. The relatively small sample size was mainly due to the strict inclusion criteria and the availability of eligible PD patients who completed all imaging and serum assessments. Considering the exploratory nature of this study and our earlier research, which reported a medium sized sample linking vitamin D to PD (PD: N = 330, HC: N = 209) (Lv et al., 2021), the current sample size is considered sufficient to potentially detect meaningful group differences with reasonable statistical power. The study procedures are described in Figure 1. The study was approved by the ethics committee of the Second Xiangya Hospital of Central South University. Informed written consent was obtained from all subjects prior to their inclusion in the study.

The demographic characteristics of HCs and PD patients, including age, age of onset, sex, years of education, height, and weight, were collected. Clinical characteristics for all PD patients were assessed, encompassing disease duration, drug history, the Unified Parkinson’s Disease Rating Scale (UPDRS II, III, IV, and total), the modified Hoehn and Yahr Staging Scale (H&Y stage), the Parkinson’s Disease Questionnaire (PDQ39), the Mini-Mental State Examination (MMSE). Based on MMSE scores, patients were categorized into PD with normal cognition (PD-NC) and PD-CI. Following criteria established by previous studies (Zhang et al., 1990; Katzman et al., 1988; Marder, 2010; Williams-Gray et al., 2006), MMSE scores were classified as follows: ≤17 for illiterate subjects, ≤20 for grade-school literate, and ≤23 for those with junior high school education or higher.

The acquisition of blood samples and rs-fMRI data was performed on the same day as the neuropsychological testing. Serum 25-hydroxyvitamin D [25(OH)D] levels were measured using liquid chromatography-mass spectrometry (LC-MS). To minimize potential confounders affecting serum vitamin D levels, we adopted the same methodological controls as our previous study (Lv et al., 2021). Specifically, patients who used vitamin D supplements, had chronic illnesses, or were taking medications that could influence vitamin D absorption were excluded. Participants were recruited continuously throughout the year, ensuring a balanced distribution across different months, thereby minimizing the confounding effect of seasonal variation in vitamin D levels. As demonstrated in our previous work (Lv et al., 2021), this recruitment strategy effectively eliminated the seasonal bias in serum vitamin D measurements. All evaluations for PD patients, including MRI studies and neuropsychological tests, were conducted in the OFF-medication state. This OFF-medication state was achieved by the withdrawal of dopaminergic medications 12 to 18 h prior to testing.

The statistical analyses were conducted using the Statistical Package for Social Sciences (SPSS), version 26 (IBM SPSS Statistics 26). For continuous data that met the criteria for normal distribution and homogeneity of variance, differences between two groups were assessed using the independent samples t-test, while differences among three or more groups were evaluated using analysis of variance (ANOVA). In cases where the data did not meet these criteria, the Mann–Whitney U test and Kruskal–Wallis test were employed for two-group and multi-group comparisons, respectively. Categorical variables were analyzed using the χ² test or Fisher’s exact test, as appropriate. Correlation analyses were performed using multiple regression analysis, controlling for age and gender, to identify the association between MMSE scores and vitamin D levels in patients with PD. Statistical significance was defined as p < 0.05.

The MRI data was acquired using a Siemens MAGNETOM Skyra 3T MRI scanner. Functional images were obtained through Echo Planar Imaging (EPI) sequences with the following parameters: repetition time (TR) = 2,500 ms, echo time (TE) = 25 ms, flip angle (FA) = 90°, matrix size = 64 × 64, field of view (FOV) = 240 mm × 240 mm, slice thickness = 3.5 mm, and 39 slices with no gap between slices during the resting condition. Additionally, axial T1 structural images were acquired using T1-weighted imaging (T1WI) three-dimensional magnetization-prepared rapid acquisition gradient echo (3D MP-RAGE) sequences. The specific parameters for these sequences were: TR = 1,900 ms, TE = 2.01 ms, FA = 9°, slice thickness = 1 mm, 176 slices, FOV = 256 mm × 256 mm, and a matrix size of 256 × 256. During the MRI scans, subjects were instructed to keep their eyes closed while avoiding falling asleep. To minimize head motion and reduce noise, foam pads were used to stabilize the head.

VBM analysis was conducted using FMRIB’s FSL software suite.¹ The methodological steps are detailed as follows: (1) Image quality assessment: The T1-weighted images for each subject were visually inspected to ensure there were no significant quality issues; (2) Skull dissection and brain extraction: Each subject’s T1 image underwent skull stripping to remove non-brain tissues, focusing on isolating the brain matter; (3) GM Template construction: (1) Segmentation: Post skull stripping, each subject’s image was segmented into GM, white matter (WM), and cerebrospinal fluid (CSF); (2) Initial template formation: The GM images were initially aligned to the Montreal Neurological Institute (MNI152) standard space using linear registration to create a preliminary GM template; (3) Final template refinement: These GM images were then non-linearly registered to the preliminary GM template, resulting in the final group-specific GM template. (4) Registration and smoothing: All individual GM images were registered to the group-specific GM template obtained in the previous step. The registered images were then smoothed at varying kernel sizes to produce t-value maps with different levels of smoothness. (5) Statistical analysis and multiple comparison correction: Group comparisons were conducted using the “randomize” tool in FSL, employing a non-parametric permutation approach. Each comparison involved 5,000 permutations to ensure robust statistical analysis and correction for multiple comparisons.

After pre-processing, to account for potential confounding factors, age and sex were included as nuisance variables in the comparisons between PD subgroups and HCs identified through cluster analysis. Significant clusters were identified using the threshold-free cluster enhancement (TFCE) method (Smith and Nichols, 2009) with statistical significance defined as p < 0.05. The results of the group comparisons were visualized using FSLview. To pinpoint the vulnerable brain regions with GM that are highly sensitive to vitamin D levels, we conducted a Kendall’s tau-b correlation analysis. This analysis controlled for age and gender, aiming to explore the relationship between GMV and vitamin D levels.

DPABI (Yan et al., 2016) (http://rfmri.org/DPABI; version, V5.1_201201), Data Processing Assistant for Resting-State fMRI (DPARSF 5.1), SPM12² and REST_V1.8_130615 (Harrison et al., 2011) software based on Matlab R2019B are used to analysis rs-fMRI data (Yan et al., 2016; Chao-Gan and Yu-Feng, 2010). The standard processing steps in DPARSF included: (1) convert data from DICOM format to NIFTI format; (2) remove the first 10 time points; (3) time and space correction; (4) head motion correction; (5) normalize by using DARTEL toolbox; (6) smooth: spatial smoothing with an isotropic 6-mm full width at half maximum (FWHM) Gaussian kernel; (7) after normalization and smoothing, the waveform of each voxel was finally used for removal of the linear trends of time courses and for temporal band-pass filtering (0.01 to 0.08 Hz).

For the network construction of vulnerable areas associated with vitamin D, we selected vulnerable GM regions as seeds to perform voxel-wise FC analyses for PD sub-groups and HCs. The voxel-wise FC analyses were conducted by computing the temporal correlations between the mean time series of each seed region and the time series of each voxel across the whole brain. This process yielded correlation coefficients (r values), which were subsequently transformed into Z values using Fisher’s r-to-z transformation to ensure that the correlation coefficients followed a normal distribution. These Z values were then used to calculate the T values for the FC strength between the regions of interest (ROIs) and other brain regions throughout the whole brain (Chao-Gan and Yu-Feng, 2010). For the FC maps, we performed a two-tailed one-sample t-test on the Z maps. In the FC analysis, the AlphaSim method was employed to correct for multiple comparisons. The significance threshold was set at a p-value of <0.005 before correction and <0.05 after correction. Additionally, analyses were performed to examine the correlations between FC and disease severity, as well as cognitive impairment. We further performed binary logistic regression analysis, adjusting for age and H&Y stage, to investigate the impact of SFG FC on cognitive impairment in PD. Prior to model estimation, the variance inflation factor (VIF) was used to assess multicollinearity, and predictors with a VIF greater than 10 were excluded. For each model, odds ratios (OR) with 95% confidence intervals were calculated. To adjust for bias due to small sample size and extreme data, we further applied Firth logistic regression using the profile penalized log-likelihood method in RStudio to explore the effect of SFG FC on cognitive impairment in PD patients. All analyses were performed in both SPSS (Version 27.0) and RStudio (Copyright^© 2025 by Posit Software, PBC). The threshold for statistical significance was established at p < 0.05.

The demographic and clinical characteristics, as well as vitamin D levels, of the 34 PD patients and 21 HCs are summarized in Table 1. No significant differences were found between the HC group and the PD subgroup in terms of age, BMI, and sex. Additionally, there were no statistically significant differences between the PD-CI and PD-NC group regarding disease duration, UPDRS scores (II, III, IV, and total), Hoehn and Yahr stage, Levodopa Equivalent Daily Dose (LEDD), and Parkinson’s Disease Questionnaire (PDQ-39) scores (p > 0.05). The onset age of patients in the PD-CI group was older than that of patients in the PD + NC group. Furthermore, the PD-CI subgroup had significantly fewer years of education compared to the PD-NC subgroup [6 (5, 9) vs. 10 (9, 12); Bonferroni corrected p = 0.002]. The PD-CI subgroup also exhibited lower MMSE scores. No significant differences were observed in the levels of 25(OH)D₃ between the two subgroups. Similarly, there were no significant differences in the levels of 25(OH)D, although the levels of 25(OH)D were lower in the PD-CI group compared to the PD-NC group, which may be related to the small sample size.

Compared to the HCs, individuals with PD exhibited notable atrophy in both the cerebral cortex and cerebral white matter. Moreover, the PD cohort demonstrated widespread GMV loss across the entire brain. Specifically, notable abnormalities were detected in several regions, including the middle frontal gyrus (MFG), superior frontal gyrus (SFG), hippocampus, left thalamus, middle temporal gyrus, cingulate gyrus, frontal pole, angular gyrus, postcentral gyrus, supramarginal gyrus, precentral gyrus, caudate, temporal pole, and cerebellum (right VI) (TFCE corrected p < 0.05). Supplementary Figure 1 provides a visual representation of these results.

In this study, we identified brain regions where GMV differed between PD patients and HCs, defining these areas as ROIs. Using ROI-based Kendall’s tau-b correlation analysis, while controlling for age and gender, we sought to identify specific GMV regions in PD patients associated with vitamin D levels. Our analysis revealed a significant association between 25(OH)D levels and a particular voxel (see Figure 2, TFCE corrected p < 0.05, cluster > 112) located at coordinates (MIN, x = 28, y = 8, z = 46) (r = −0.406, corrected p = 0.021). This voxel predominantly encompassed the bilateral MFG and SFG. These findings establish the MFG and SFG as vulnerable regions in PD patients linked to vitamin D levels.

In the VBM analysis, GMV differences between PD patients and HCs were more pronounced in the SFG compared to the MFG (see Figure 2). Furthermore, we analyzed whole-brain functional connectivity (FC) patterns across the HC, PD-NC, and PD-CI groups using MFG as the ROI and found no significant association between cognitive function and the MFG functional network (data not provided). To further assess the impact of abnormal vitamin D levels on relevant resting-state networks, we selected clusters encompassing the bilateral SFG, specifically the medial SFG (SFGmed) and medial orbital SFG (SFGmorb) regions, as four ROIs. These regions were used to analyze FC patterns across the entire brain among the three groups. Spherical seed regions with a radius of 2 mm were created based on the AAL3 template (Rolls et al., 2020) within MATLAB R2019B software.

In this study, correlation analyses were conducted to explore the interrelations among disease severity, cognitive function, and functional connectivity correlation coefficient (fc-CC) values in patients with PD. Our findings demonstrate that abnormal fc-CC values within the SFG functional network are associated with several factors, including disease severity (Hoehn and Yahr stage), UPDRS II scores, cognitive dysfunction, and vitamin D levels (Pearson correlation analysis, p < 0.05) (see Figure 7). Specifically, abnormalities in the fc-CC values of the left SFGmed functional network were significantly correlated with H&Y stage (r = −0.418, p = 0.014), and MMSE scores (r = 0.372, p = 0.03). In parallel, anomalies in the fc-CC values of the right SFGmed functional network were linked to H&Y stage (r = −0.347, p = 0.044). Furthermore, deviations in the fc-CC values of the left SFGmorb functional network were correlated with UPDRS II scores (r = −0.359, p = 0.037), and H&Y stage (r = −0.404, p = 0.018). Similarly, irregularities in the fc-CC values of the right SFGmorb functional network were found to be related to vitamin D levels. These findings underscore the complex interplay between neural connectivity and various clinical and cognitive features in PD, highlighting the potential of fc-CC values as biomarkers for disease progression and symptomatology.

Supplementary Table 1 presents the results of a logistic regression analysis assessing the impact of FC in the SFG network and vitamin D levels on cognitive impairment in PD. SFGmed.L Cluster-2 shows a consistent negative association with cognitive impairment across all three models, with a statistically significant regression coefficient in Models 1 (B = −27.125, p = 0.026), 2 (B = −24.253, p = 0.040), and 3 (B = −22.817, p = 0.055), the latter approaching statistical significance. Additionally, in Model 3, 25(OH)D₃ is significantly associated with cognitive impairment (B = −0.173, p = 0.047). A trend toward significance is observed for SFGmed.R Cluster-2 (B = 0.185, p = 0.082), indicating a possible but non-significant positive relationship with cognitive impairment.

Further Firth logistic regression analyses provided additional insights into the relationships between SFG functional network connectivity and cognitive impairment in PD (Table 3). In Model 1, which did not adjust for confounders, connectivity in the left SFGmed network was significantly associated with cognitive impairment, with a negative coefficient suggesting a protective effect (B = −21.234, Chisq = 5.456, p = 0.020). Model 2, which adjusted for age, confirmed this relationship (B = −18.173, Chisq = 4.396, p = 0.036). Interestingly, model 3, which further adjusted for H&Y stage, suggested that vitamin D levels might exert a protective effect against cognitive impairment (B = −0.123, Chisq = 3.607, p = 0.058). In this model, the left SFGmed network showed a trend toward significance (B = −16.364, Chisq = 3.607, p = 0.058), though it did not reach the conventional threshold for statistical significance. The connectivity of the SFGmorb network, both left and right, showed weaker associations with cognitive impairment.

Importantly, these regression findings are consistent with the group comparison results shown in Table 2. As demonstrated in Table 2, the PD-CI group exhibited lower fc-CC values in the left SFGmed.L Cluster-2 compared with the PD-NC group, indicating reduced functional connectivity in patients with cognitive impairment. Correspondingly, the regression analysis revealed a negative association between SFGmed.L Cluster-2 connectivity and cognitive impairment, suggesting that higher connectivity is linked to a lower risk of cognitive impairment. Together, these findings consistently indicate that the potential of SFG functional network connectivity—particularly within the left SFGmed—is associated with cognitive decline in PD.

Our study revealed significant brain atrophy in individuals with PD, particularly in the cerebral cortex and white matter, along with widespread GMV loss in regions such as the frontal, temporal, and cingulate gyri, hippocampus, thalamus, and cerebellum. These observations are consistent with previous literature (Jia et al., 2015; Zheng et al., 2022; Xu et al., 2020). The findings highlight the extensive neurodegenerative impact of PD, correlating with clinical symptoms and cognitive decline seen in patients (Zhou et al., 2019; Pan et al., 2024; Li et al., 2022). The widespread atrophy and specific regional degeneration emphasize the multifaceted nature of PD, affecting both motor and cognitive domains. Future research should focus on longitudinal studies to track these changes over time and investigate potential interventions to mitigate neurodegeneration. Understanding the precise mechanisms of atrophy in these regions could pave the way for targeted therapies aimed at preserving brain volume and function, ultimately improving the quality of life for individuals with PD.

This study found a significant correlation between the reduction in GMV of the SFG and vitamin D levels, suggesting that vitamin D levels may be associated with cognitive function in PD patients through their effects on the structure and function of the SFG. The SFG, located on the superior portion of the frontal lobe, is involved in various cognitive functions, including working memory, attention, and executive function management. GMV reduction in the SFG of PD patients may be related to impairments in these cognitive functions. Vitamin D may contribute to this process through its neuroprotective effects, modulating brain regions vital for executive functions.

Vitamin D is a fat-soluble vitamin known for its role in bone health. Recent research has highlighted its importance in brain function and neuroprotection (Rimmelzwaan et al., 2016; Barichella et al., 2022; Wang et al., 2015). It exerts its action via vitamin D receptors (VDR) present in neurons and glial cells, their expression is maximum in the hippocampus, thalamus, hypothalamus, substantia nigra, orbitofrontal-cortex, cingulate, and amygdala which are the areas essential for cognition and motor control (Moretti et al., 2018; Annweiler et al., 2013; Veenstra et al., 1998). And study demonstrated that VDR polymorphisms were associated with cognitive decline in PD (Gatto et al., 2016). Other studies have shown that low levels of vitamin D are associated with cognitive decline in PD patients (Peterson et al., 2013; Pignolo et al., 2022). The potential benefits of vitamin D for PD patients have also garnered significant attention. Suzuki et al. (2013) conducted a randomized controlled trial and found that PD patients who received daily vitamin D supplementation experienced a stabilization of the severity of PD. However, even less is known about the effects of vitamin D supplementation on cognition in PD patients (Fullard and Duda, 2020). In PD patients, low levels of vitamin D might affect cognitive function through multiple mechanisms. Firstly, vitamin D has anti-inflammatory and antioxidant properties, which can reduce neuroinflammation and oxidative stress, both of which play significant roles in the pathogenesis of PD (Fullard and Duda, 2020; Calvello et al., 2017). Neuroinflammation and oxidative stress, commonly seen in the brains of PD patients, can lead to neuronal damage and death, consequently impairing cognitive functions (Bloem et al., 2021). Secondly, vitamin D can regulate the expression of neurotrophic factors and promote the growth and survival of neurons (Pertile et al., 2018). Thirdly, vitamin D is involved in calcium homeostasis regulation, and calcium signaling is crucial for neurotransmission and synaptic plasticity (Mayne and Burne, 2019; Latimer et al., 2014). While several possible mechanisms exist for the effect of vitamin D on cognitive in PD, the evidence is preliminary, and more study is needed to make any conclusions.

Additionally, we also found that the abnormal connections between the SFG and multiple functional networks might contribute to the cognitive deficits observed in PD patients. These findings are consistent with research in both the general population and AD patients, where vitamin D deficiency has been linked to accelerated brain aging and impaired cognitive function (Terock et al., 2022; Sultan, 2022; Shih et al., 2020). Our study further suggests that abnormal fc-CC values within the SFG functional network are associated with PD severity, cognitive dysfunction, and vitamin D levels. Specifically, we found that the abnormal functional network connections between the SFG and several networks—including the visual network, DMN, executive control network, sensorimotor network, and memory network—are characteristic of the PD-CI group. These findings are consistent with our logistic regression analysis, which also highlighted a significant association between functional connectivity in the left SFGmed.L Cluster-2 and cognitive impairment in PD patients.

These results align with previous studies in AD, which have shown that vitamin D deficiency is associated with disruptions in brain connectivity, particularly in the frontal, occipital, and temporal lobes (Fu et al., 2025). Additionally, research has demonstrated a significant correlation between changes in vitamin D levels and volumetric alterations in cortical regions such as the hippocampus, amygdala, and nucleus accumbens in AD (Li et al., 2025). These findings were further corroborated in AD mouse models, where vitamin D supplementation significantly delayed hippocampal and nucleus accumbens atrophy, while also reducing Aβ accumulation (Li et al., 2025). This suggests that vitamin D plays a crucial role in protecting against neurodegeneration and maintaining brain structure and function, as indicated by our findings, suggests a potential protective mechanism that could extend to PD patients. Given the similarities between PD and AD in terms of brain degeneration, future research should investigate whether vitamin D supplementation can help mitigate cognitive decline and brain atrophy in PD, similarly to its effects observed in AD patients.

Building on this, our study highlights the association between vitamin D levels, specific brain networks, and cognitive dysfunction in PD patients. Disruptions in these networks could provide further insights into the mechanisms underlying cognitive decline in PD. These networks are essential for integrating sensory information, maintaining cognitive flexibility, and performing complex tasks. Specifically, the visual network processes visual information, Relatedly, decreased visual performance was associated with cognitive decline and widespread white matter macrostructural changes in PD patients (Zarkali et al., 2021). The DMN remains active during rest and is associated with self-referential and introspective thoughts, Meanwhile, the altered connectivity pattern in the DMN may play a crucial role in the neurophysiological mechanism of cognitive decline in patients with PD (Chen et al., 2022; Zarifkar et al., 2021). The executive control network is responsible for high-order cognitive functions like planning and decision-making, the sensorimotor network involves motor control and sensory processing, and the memory network is responsible for encoding and retrieving memories. Disruptions in these networks can lead to the cognitive deficits observed in PD patients (Dong et al., 2020; Leh et al., 2010; Wang et al., 2021). Abnormal connections between the SFG and these networks might lead to disruptions in information integration and transmission, thereby affecting cognitive function in PD patients. However, due to the cross-sectional nature of this study, it is important to note that the observed relationships between vitamin D levels, SFG atrophy, and cognitive decline are associative and not causal. Longitudinal studies are essential to clarify the temporal dynamics, explore causality, and identify potential confounding factors, such as disease severity, that could influence both vitamin D levels and brain atrophy in PD. Future longitudinal studies are essential to validate these associations. Additionally, research should focus on investigating whether vitamin D supplementation can influence cognitive outcomes and brain atrophy over time in PD patients.

This study has some strengths. Firstly, the association between SFG atrophy and variations in vitamin D levels is particularly intriguing, suggesting a possible nutritional or metabolic influence on brain health in PD patients. This link opens avenues for potential dietary or supplementation interventions that could mitigate or slow the progression of atrophy and cognitive decline. Another strength lies in the study’s examination of functional network integrity. By not only measuring structural changes but also assessing the functional connectivity of the SFG, the research offers a more comprehensive understanding of the brain alterations occurring in PD-CI. This dual approach underscores the complexity of the disease and the interplay between structural and functional brain changes, providing a more holistic view of PD pathology. The correlation of SFG functional deterioration with disease severity and cognitive deficits further strengthens the study. This finding underscores the relevance of the SFG in the clinical progression of PD and suggests that SFG integrity could potentially serve as a biomarker for disease progression and cognitive decline. This potential biomarker utility is valuable for both clinical assessment and the evaluation of therapeutic interventions. Lastly, the study’s consideration of vitamin D levels adds a significant dimension to the research. By exploring the association between vitamin D and brain changes, the study highlights a modifiable factor that could influence disease outcomes. This focus on a potentially modifiable risk factor is particularly valuable for developing preventive or therapeutic strategies aimed at improving the quality of life for PD patients.

Additionally, we acknowledged the study has several limitations. Firstly, as a cross-sectional study, it cannot establish causality. Whether low vitamin D levels directly cause structural changes in the SFG and cognitive decline, or whether other mediating factors are involved, remains to be answered by longitudinal studies and mechanistic research. Secondly, the sample size in this study is relatively small, mainly due to the strict inclusion criteria and the availability of eligible PD patients who completed all imaging and serum assessments. While this sample size is considered sufficient to detect meaningful group differences with acceptable statistical power, based on the exploratory nature of the study and the medium effect sizes observed in our previous research, future studies should incorporate larger sample sizes for further validation. Lastly, the lack of replication and external validation in this study limits the generalizability of our results, highlighting the need for future studies with replication samples and validation groups.