Motor dysfunction in Parkinson’s patients: depression differences in a latent growth model

Objective This study aims to utilize latent growth model (LGM) to explore the developmental trajectory of motor dysfunction in Parkinson’s disease (PD) patients and investigate the relationship between depression and motor dysfunction. Methods Four-year follow-up data from 389 PD patients were collected through the Parkinson’s Progression Marker Initiative (PPMI). Firstly, a univariate LGM was employed to examine the developmental trajectory of motor dysfunction in PD patients. Subsequently, depression levels were introduced as covariates into the model, and depression was further treated as a parallel growth latent variable to study the longitudinal relationship between motor dysfunction and depression. Results In the trajectory analysis of motor dysfunction, the fit indices for the quadratic growth LGM model were χ2 = 7.419, df = 6, CFI = 0.998, TLI = 0.997, SRMR = 0.019, and RMSEA = 0.025, indicating that the growth trend of motor dysfunction follows a quadratic curve rather than a simple linear pattern. Introducing depression symptoms as time-varying covariates to explore their effect on motor dysfunction revealed significant positive correlations (β = 0.383, p = 0.026; β = 0.675, p < 0.001; β = 0.385, p = 0.019; β = 0.415, p = 0.014; β = 0.614, p = 0.003), suggesting that as depression levels increase, motor dysfunction scores also increase. Treating depression as a parallel developmental process in the LGM, the regression coefficients for depression intercept on motor dysfunction intercept, depression slope on motor dysfunction slope, and depression quadratic factor on motor dysfunction quadratic factor were 0.448 (p = 0.046), 1.316 (p = 0.003), and 1.496 (p = 0.038), respectively. These significant regression coefficients indicate a complex relationship between depression and motor dysfunction, involving not only initial level associations but also growth trends over time and possible quadratic effects. Conclusion This study indicates a quadratic growth trajectory for motor dysfunction in PD, suggesting a continuous increase in severity with a gradual deceleration in growth rate. The relationship between depression and motor dysfunction is complex, involving initial associations, evolving trends over time, and potential quadratic effects. Exacerbation of depressive symptoms may coincide with motor function deterioration.

Methods: Four-year follow-up data from 389 PD patients were collected through the Parkinson's Progression Marker Initiative (PPMI).Firstly, a univariate LGM was employed to examine the developmental trajectory of motor dysfunction in PD patients.Subsequently, depression levels were introduced as covariates into the model, and depression was further treated as a parallel growth latent variable to study the longitudinal relationship between motor dysfunction and depression.
Results: In the trajectory analysis of motor dysfunction, the fit indices for the quadratic growth LGM model were χ2 = 7.419, df = 6, CFI = 0.998, TLI = 0.997, SRMR = 0.019, and RMSEA = 0.025, indicating that the growth trend of motor dysfunction follows a quadratic curve rather than a simple linear pattern.Introducing depression symptoms as time-varying covariates to explore their effect on motor dysfunction revealed significant positive correlations (β = 0.383, p = 0.026; β = 0.675, p < 0.001; β = 0.385, p = 0.019; β = 0.415, p = 0.014; β = 0.614, p = 0.003), suggesting that as depression levels increase, motor dysfunction scores also increase.Treating depression as a parallel developmental process in the LGM, the regression coefficients for depression intercept on motor dysfunction intercept, depression slope on motor dysfunction slope, and depression quadratic factor on motor dysfunction quadratic factor were 0.448 (p = 0.046), 1.316 (p = 0.003), and 1.496 (p = 0.038), respectively.These significant regression coefficients indicate a complex relationship between depression and motor dysfunction, involving not only initial level associations but also growth trends over time and possible quadratic effects.

Conclusion:
This study indicates a quadratic growth trajectory for motor dysfunction in PD, suggesting a continuous increase in severity with a gradual deceleration in growth rate.The relationship between depression and motor

Introduction
Parkinson's disease (PD) is a chronic progressive neurological disorder, with its prevalence increasing with age.The World Health Organization predicts that by 2030, the global number of PD patients will reach approximately 8.67 million, with around 4.94 million in China (Ma et al., 2014).By 2040, the diagnosis numbers are expected to double (Titova and Chaudhuri, 2018).The main pathological changes in PD involve degeneration and reduction of dopaminergic neurons in the substantia nigra pars compacta (Munhoz et al., 2015).Dopamine, a neurotransmitter, is crucial for coordinating and controlling motor function.Therefore, the decline in dopamine levels directly affects motor function.The primary clinical symptoms of PD are motor disturbances, including bradykinesia, resting tremor, rigidity, postural instability, and dystonia.Bradykinesia is the most typical clinical feature of PD, usually manifested as a slowing of voluntary movements, reduced speed and amplitude of repetitive movements, drooling, masked face, and micrographia.Motor symptoms typically begin on one side of the body and then progress to the other side over several years (Postuma et al., 2015a;Xu et al., 2019).These symptoms significantly impact patients' everyday mobility.Alongside the direct motor symptoms, PD is often accompanied by a range of non-motor symptoms (Zesiewicz, 2019), including cognitive impairment, mild autonomic dysfunction, depression, and anxiety.Some non-motor symptoms and signs have been shown to manifest several years or even decades before a clinical diagnosis of PD (Postuma et al., 2015b;Heinzel et al., 2019).These nonmotor symptoms may indirectly affect patients' motor function by influencing their overall quality of life and activity levels, extending well beyond the impairment of motor function (Lees et al., 2009;de la Riva et al., 2014;LeWitt and Chaudhuri, 2020).Non-motor symptoms persist throughout the course of PD and often manifest as initial clinical presentations (Weintraub et al., 2022).They do not necessarily disappear after the onset of motor symptoms; in fact, they may gradually worsen as the disease progresses.Among these non-motor symptoms, depression is particularly prevalent in PD patients, occurring in approximately 30% of cases (Jellinger, 1999) The onset of depressive symptoms can vary widely, ranging from as little as 1 month to as long as 30 years (Iranzo et al., 2014;Walter et al., 2015).Depression may precede the onset of Parkinson's motor impairments, but in the subsequent course of development, depression and motor dysfunction may mutually influence each other.Currently, research on motor dysfunction in PD primarily focuses on mechanisms, functional assessment, monitoring, and treatment, with relatively few studies investigating the trajectory of motor function in Parkinson's patients.Currently, numerous studies have investigated the relationship between motor and nonmotor functions, including depression (Zahodne et al., 2012), autonomic dysfunction (Qin et al., 2024), cognitive impairment (Moustafa et al., 2016), olfactory dysfunction (Nabizadeh et al., 2022), and sleep disturbances (Bugalho and Viana-Baptista, 2013), all of which have demonstrated close associations with motor function.Among these factors, sleep disturbances are particularly intertwined with motor function, and a bidirectional relationship exists between sleep disturbances and depression, underscoring the importance of further exploring the relationship between depression and motor function (Fang et al., 2019).
Research on the motor function trajectory of PD patients and the relationship between depression and motor trajectory is relatively scarce.Cubo et al. (2000) employed univariate logistic regression analysis to examine the association between motor dysfunction assessed by total the Unified Parkinson's Disease Rating Scale scores and Hoehn-Yahr staging and depression.Darweesh et al. (2017) conducted a nested case-control study, gathering data on daily functional trajectories and motor and non-motor features from 1990 to 2013, spanning 23 years before PD diagnosis.They utilized extended mixed models and found that PD patients increasingly displayed low and slow motor signs throughout the study follow-up compared to controls from 7.5 years before diagnosis, with statistically significant overall differences.However, this study has limitations as motor features were assessed by research nurses subjectively, using subjective ratings rather than specialized quantitative methods.Poonja et al. (2021), aiming to examine and describe the trajectory of Unified Parkinson's Disease Rating Scale motor scores (UPDRS-III) in the late stages of PD, conducted separate modeling of UPDRS-III scores using piecewise linear models and clustered the resulting trajectories based on their characteristics.They concluded that while different populations with varying motor function in PD exhibited different changes, such as continuous deterioration, stable-deterioration, improvement-deterioration, and deterioration-improvement-deterioration, the eventual decline in motor function was universal.This study was a retrospective chart review rather than a data modeling study.Zahodne et al. (2012) obtained data from 186 PD patients (mean duration of disease 8.2 years) from a clinical research database, incorporating 18 months of data to examine the trajectories of motor symptoms and depressive symptoms.Unconditional univariate growth models indicated that over the 18-month observation period, both motor symptoms and apathy exhibited linear deterioration, while depressive symptoms initially showed improvement before worsening, demonstrating quadratic change.The study had a relatively short longitudinal follow-up period, a small sample size, and employed a rather simplistic model.However, the latent growth model (LGM) used in this study typically refers to a statistical approach that allows for consideration of individual differences (Curran et al., 2010), providing a better LGM allows for the estimation of latent variables, revealing the underlying structure and relationships behind observed variables, thus providing a more comprehensive understanding of the data.It includes repeated measurements within individuals or multilevel data, making it more suitable for complex data analysis.This study selected 4-year longitudinal survey data of Parkinson's disease patients from the Parkinson's Progression Markers Initiative (PPMI)database, analyzing the changes in early motor function of PD patients using LGM.Furthermore, the study investigated the predictive level of depression on the developmental trajectory of motor function.Analyzing motor function trajectories can identify trends of motor function degradation in patients, enabling early intervention.Additionally, the association between depression and motor function degradation identified in the study also provides clues for early intervention.Early identification and treatment of depressive symptoms may help delay the progression of motor function degradation, contributing to the development of more comprehensive treatment plans and improving patients' quality of life.

Subjects
We obtained data from the Parkinson's Progression Markers Initiative (PPMI),a publicly available database.In 2010, The Michael J. Fox Foundation and a core group of academic scientists

Measurements
Depression levels were assessed using a condensed version of the Geriatric Depression Scale (GDS), namely the GDS-15.This scale comprises 15 items designed to gauge symptoms of depression, diminished activity, irritability, withdrawal from social engagements, as well as negative ruminations concerning the past, present, and future.Scores on the GDS-15 range from 0 to 15. Scores of 0-4 indicate normalcy; 5-8 suggest mild depression; 9-11 indicate moderate depression; and 12-15 signify severe depression.A score of 5 or higher suggests the likelihood of depression.The GDS-15 exhibits robust reliability and validity when administered to older adults (de Craen et al., 2003).
The Movement Disorder Society-Unified Parkinson's Disease Rating Scale Part III (MDS-UPDRS3) is a scale used to assess motor dysfunction in PD and is one of the standard tools for clinical evaluation of the condition.MDS-UPDRS3 comprises a series of items covering four main aspects of motor function: postural control, bradykinesia, upper limb and trunk function, and gait and balance.These items are graded on a scale from 0 to 4, where 0 represents no symptoms and 4 represents severe impairment.The minimum score is 0, and the maximum score is 132.The total score reflects the severity of motor dysfunction in patients, with higher scores indicating more severe impairment.Widely utilized in both clinical practice and research, MDS-UPDRS3 is considered a reliable and effective tool for assessing motor function in PD patients (Weintraub and Mamikonyan, 2019).

Statistical analysis
Descriptive statistics were performed on the data, with categorical variables analyzed using frequency counts and continuous variables presented as means ± standard deviations.Pearson correlation analysis was conducted to explore the correlation coefficients between variables.The trend of motor dysfunction was observed, and then an appropriate LGM was selected for fitting, based on fit indices.In constructing the LGM, an unconditional model was first established, forming repeated measurements of motor trajectories across five time points.The linear growth model included two latent variables: intercept and slope.The mean and variance parameters of these latent variables were used by the LGM to describe within-group and between-group differences.Specifically, the mean of the intercept factor represented the average initial state, while the variance of the intercept factor indicated the extent of individual differences at specific time points.A greater variance suggests more significant initial differences between individuals.The mean of the slope factor represented the average growth rate between time points, and the variance of the slope factor reflected the magnitude of individual differences in growth rates.The bidirectional arrows between the two factors indicated their correlation.Next, a quadratic growth model was constructed, which added a Quadratic Slope to the linear growth model, representing the quadratic trend of the latent variable, i.e., the variable's quadratic change over time.The quadratic factor described the concave or convex shape of the curve.These factors combined to model the trajectory of latent variables' changes over time.This study examines the trajectory of motor dysfunction change, including significant individual differences in starting level, change trends, and quadratic changes.Additionally, in the model, depression (time-varying covariate) is included to explore the predictive effect of different levels of depression on the trajectory of motor dysfunction change in PD patients.
For the missing data, we use Maximum likelihood (ML) estimation to estimate the parameters of the model.To assess the model fit degree, we rely on Comparative Fit Index (CFI), Tucker-Lewis Index(TLI), Root Mean Square Error of Approximation(RMSEA) and Standardized Root Mean Square Residual(SRMR) values, as the χ2 goodness-of-fit statistic can be overly sensitive for large sample sizes, and therefore, we do not employ it.For CFI and TLI, a value above 0.90 is considered acceptable, and a value exceeding 0.95 indicates a good fit.RMSEA value below 0.05 indicates a good fit to the data, while a value between 0.05 and 0.08 suggests a suitable fit (Hu and Bentler, 1999).SRMR examines the fit of the model by the size of the residue, and its values range from 0 to 1 and indicate a good model fit when the value is less than 0.08.
To implement the necessary model and conduct the analysis, we utilize Mplus 8.9.Descriptive analysis and plotting are performed using SPSS version 25.0 and R (version 4.2.3).The test level is set at a p-value of 0.05.

Preliminary analysis
It can be observed that in our sample (Table 1), there were 244 males (62.7%) and 145 females (37.3%).The onset age was predominantly above 56 years old (71.2%), with the majority being Caucasian (92.3%).Most individuals have had 13-23 years of education (81.3%).The average onset age was 59.08 ± 9.70 years old, and the duration of illness was 1.11 ± 1.44 years.In Table 2, we can observe that the means and standard deviations of both variables are gradually increasing, indicating a progressive worsening of both motor dysfunction and depression levels as the disease progresses.Additionally, at each time point, there is a significant positive correlation between depression levels and motor dysfunction, with correlations being evident (r = 0.104, p < 0.05; r = 0.167, p < 0.01; r = 0.161, p < 0.01; r = 0.136, p < 0.05).

Trajectories of motor dysfunction in Parkinson's patients
Based on the changes observed in the basic statistical descriptions of depression and motor function scores, we proposed two trends: one linear and the other quadratic.The fit indices for the LGM were χ2 = 35.257,df = 10, CFI = 0.966, TLI = 0.966, SRMR = 0.046, and RMSEA = 0.081.For the quadratic LGM, the fit indices were χ2 = 7.419, df = 6, CFI = 0.998, TLI = 0.997, SRMR = 0.019, and RMSEA = 0.025.Comparing the linear LGM and quadratic LGM models, we found that the quadratic LGM model outperformed the linear model, indicating that the quadratic LGM model better fits the trajectory of our data's development.
The quadratic growth LGM model (see Figure 2) reveals that at the starting point of observation, namely the initial state, the average intercept of PD motor function is 21.061 (SE = 0.477, P < 0.001), which significantly differs from 0, indicating an average score of 21.061 for motor dysfunction.The variance is 75.419(SE = 9.845, P < 0.001), suggesting significant individual differences at the initial state, indicating considerable variability among individuals in the initial level of motor dysfunction.Over the course of 4 years, the mean slope of motor function is 3.733 (SE = 0.401, P < 0.001), reflecting a linear growth trend in motor dysfunction throughout the observation period.The variance is 25.759 (SE = 8.106, P = 0.001), indicating significant variability among individuals, even though the mean suggests an overall increasing trend, there are significant differences in the growth rates among individuals.The mean of the quadratic slope is −0.327 (SE = 0.099, P = 0.001), indicating that the growth trend of motor dysfunction is not a simple linear increase, but rather follows a quadratic curve.The variance is 0.965 (SE = 0.420, P = 0.021), indicating significant variability in the quadratic growth trend among individuals.The significant negative quadratic change implies that the increase in the severity of motor dysfunction gradually slows down over time.There is a significant correlation between the slope and quadratic change (r = −4.295,p = 0.013), with the negative correlation indicating a decreasing trend in quadratic change as the slope increases.The correlation between intercept and slope, as well as intercept and quadratic change (r = −7.118,p = 0.375; r = 0.288, p = 0.866), is not significant.

Trajectories of depression and motor dysfunction in Parkinson's patients
The use of covariates in parallel developmental processes' LGM to study their impact on PD patients' motor dysfunction, Unconditional quadratic LGM.
as shown in Figure 4, enables a comprehensive consideration of the developmental trajectories of covariates.This method not only allows for understanding the effects of covariates on the target variable but also facilitates understanding the trends in covariates themselves.Moreover, this approach is flexible in capturing the complex relationships between covariates and the target variable, including nonlinear and asymmetric relationships.
The unconditional LGM with multiple processes fits the data well: χ2 = 29.876,df = 30, CFI = 1.000,TLI = 1.000,SRMR = 0.025, RMSEA < 0.001.In the model, the regression coefficient of depression intercept on motor dysfunction intercept is 0.448 (SE = 0.225; P = 0.046), indicating that individuals with higher levels of depression at the initial moment (intercept) also exhibit higher levels of motor dysfunction.The regression coefficient of depression slope on motor dysfunction slope is 1.316 (SE = 0.437; P = 0.003), suggesting an overall positive correlation between the growth trends of depression and motor dysfunction.This implies that individuals with increasing levels of depression over time are more likely to experience an increase in motor dysfunction levels.The regression coefficient of depression quadratic slope on motor dysfunction quadratic slope is 1.496 (SE = 0.721; P = 0.038), indicating the consideration of depression's quadratic effect in the model, suggesting a positive between the rates of change in depression's growth trend and motor dysfunction's growth trend.Overall, these significant regression coefficients suggest a complex relationship between depression and motor dysfunction, including not only the association at the initial level but also the growth trends over time and potential quadratic effects.This provides some clues for understanding the dynamic relationship between depression and motor dysfunction.

Discussion
First, this study employed a univariate LGM to explore the trajectory of motor dysfunction in PD patients.From the findings, our trajectory appears to follow a quadratic curve.Previous studies have indicated a linear trajectory for motor function (Zahodne et al., 2012;Poonja et al., 2021), but these studies only had data for 18 months.In contrast, we have data spanning 4 years, and from this, it seems that the severity of motor dysfunction gradually increases, albeit at a slowing rate.This may suggest that the progression of the disease slows down or stabilizes at a certain stage.May be due to the motor and depression scales we utilized have potential ceiling effects.The quadratic trend might be partly attributed to the scaling properties of the measurement tools, where scores may plateau, reflecting an upper limit of the scale rather than a true stabilization of symptoms.Nonetheless, the deterioration of motor function is inevitable.Therefore, closely monitoring changes in symptoms, adjusting treatment plans promptly, and taking measures to improve patients' quality of life remain essential.Subsequently, we introduced depression as a covariate into our model, and further incorporated depression as a parallel developing LGM.This approach allows for a comprehensive consideration of the developmental trajectories of covariates, enabling us to understand not only the impact of covariates on the target variables but also the trends in the covariates themselves.Moreover, this method offers greater flexibility in capturing the complex relationships between covariates and target variables.Both models indicate that depression influences the trajectory of motor dysfunction to some extent.Whether depression is considered as a covariate changing over time or as a parallel developing LGM, it demonstrates an impact on motor function.The positive significance of the results suggests that depression be a contributing factor to or exacerbate motor dysfunction to a certain extent.Depression is not merely a short-term effect but may have a sustained impact on motor function throughout the course of the disease.Therefore, in the treatment and management of PD patients, it is essential to consider the influence of depression comprehensively and take measures promptly to manage depressive symptoms in order to improve both motor function and quality of life for patients.
The mechanism underlying motor dysfunction in PD is primarily attributed to biochemical abnormalities in the basal ganglia and disruption of circuit activity.Degeneration of the nigrostriatal dopamine pathway in the substantia nigra to striatum leads to excessive output from the basal ganglia, while excessive inhibition of thalamocortical feedback activity impairs the facilitatory action of cortical motor function, resulting in motor disturbances.The basal ganglia exhibit intricate fiber connections, primarily comprising three crucial neural circuits: the corticocortical circuit involving the cerebral cortex, caudate nucleus, globus pallidus, thalamus, and cerebral cortex; the nigrostriatal circuit connecting the substantia nigra with the striatum and vice versa; and the striatopallidal circuit involving the caudate nucleus, putamen, external segment of the globus pallidus, subthalamic nucleus, and internal segment of the globus pallidus.These circuits form the anatomical basis for the motor regulation function of the basal ganglia, with the balanced activity of these pathways being vital for normal motor function.Treatment of motor disorders in PD, whether pharmacological or surgical, is based on correcting neurotransmitter abnormalities and disrupting circuit activity.Mechanistic studies of depressive symptoms also involve the striatum (Mulders et al., 2022), prefrontal cortical regions (especially the dorsolateral prefrontal cortex), as well as bottomup subcortical networks (including the amygdala) (Almeida Parallel LGM of depression and motor dysfunction.et al., 2009;He et al., 2019), and the hypothalamic-pituitaryadrenal axis (Liu et al., 2021).Additionally, neurodegeneration is widespread in brainstem nuclei, including autonomic areas such as the dorsal motor nucleus of the vagus nerve, motor nuclei like the pontine nuclei, and regulatory nuclei such as the locus coeruleus and raphe nuclei (Seidel et al., 2015).This correlates with both motor dysfunction and depressive symptoms.Therefore, there are some overlaps between the mechanisms of the two conditions, suggesting that depression may exacerbate or worsen motor dysfunction by affecting the neurotransmitter systems and neuronal activity in the brain, thereby influencing the neurotransmitter systems involved in motor control.Our study provides evidence for the association between the two, but further research is needed to elucidate the specific mechanisms between them.Currently, the main treatments for depression include the use of antidepressant medications, electroconvulsive therapy (Xu et al., 2020), repetitive transcranial magnetic stimulation (Dubovsky et al., 2021), etc.Additionally, exercise therapy (Cheng et al., 2016;LaHue et al., 2016) is a non-pharmacological treatment for depression, offering advantages such as affordability, convenience, high compliance, and minimal adverse reactions.Exercise therapy not only improves the worsening of depressive symptoms but also enhances motor function (Kashif et al., 2022).With the continuous advancement of medical technology, more comprehensive and detailed treatment strategies have been proposed, such as comprehensive rehabilitation for elderly patients with PD (Hongxia, 2023).This comprehensive rehabilitation model provides tailored rehabilitation assessments and treatment plans for elderly PD patients at different stages, aiming to comprehensively and scientifically manage the rehabilitation needs of patients.Covering the entire disease process of elderly PD patients, this comprehensive rehabilitation model can effectively improve physical function, mental health, and quality of life, delay disease progression, enhance coping abilities, and is an essential component of integrated treatment.
This study also has some limitations.Firstly, our study participants were mainly Caucasian, which may limit the generalizability of our findings.Secondly, our data were based on a 4-year period for LGM analysis.Although this provides valuable insights, there may be limitations for those interested in longerterm trajectory analyses.Future research could consider continuing follow-ups and increasing the frequency of assessments to obtain more comprehensive and reliable results.Thirdly, regarding the measurement of motor dysfunction, while the scales we used can accurately measure, there are updated techniques available for more precise long-term monitoring of motor symptoms in PD patients throughout the day or over several days to assess the range and severity of motor dysfunction symptoms.Therefore, if feasible, we could adopt more optimized methods in the future.Additionally, classifying PD patients according to different subtypes could provide more detailed insights into the trajectory of motor dysfunction for each subtype, warranting further investigation.Furthermore, exploring the relationship and mutual influence between motor dysfunction and depression in PD patients using other models is also worth considering.

Conclusion
Through the analysis of 4-year longitudinal data from PPMI, the developmental trajectory of motor dysfunction in PD patients was elucidated.Furthermore, the relationship between depression severity and motor dysfunction was revealed using a bivariate latent growth curve model.The findings indicate that the growth trajectory of motor dysfunction follows a quadratic curve rather than a simple linear increase.Specifically, while the severity of motor dysfunction continues to rise, the rate of increase gradually slows down.In the model of depression and motor dysfunction, all regression coefficients were found to be significant and positively correlated, suggesting a complex relationship between depression and motor dysfunction.This relationship encompasses not only the initial level of association but also the growth trend and possible quadratic effects over time.The exacerbation of depressive symptoms may be accompanied by worsening motor function, providing valuable insights into understanding the dynamic relationship between depression and motor dysfunction.Given their interaction, interventions aimed at improving depression symptoms in PD patients may contribute to the rehabilitation of motor dysfunction.

TABLE 1
Demographic information of the study participants (n = 389).
Age at PD symptom onset and duration from PD diagnosis are continuous variables.We describe them using mean ± standard deviation.

TABLE 2
The average level and bivariate correlations of the main study variables.