Predicting Dementia in People with Parkinson's Disease

Predicting Dementia in People with Parkinson’s Disease | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Predicting Dementia in People with Parkinson’s Disease View ORCID Profile Mohamed Aborageh , Tom Hähnel , Patricia Martins Conde , Jochen Klucken , Holger Fröhlich doi: https://doi.org/10.1101/2025.01.27.25321134 Mohamed Aborageh 1 Department of Bioinformatics, Fraunhofer Institute for Algorithms and Scientific Computing (SCAI) , 53757 Sankt Augustin, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Mohamed Aborageh For correspondence: mohamed.aborageh{at}scai.fraunhofer.de holger.froehlich{at}scai.fraunhofer.de Tom Hähnel 1 Department of Bioinformatics, Fraunhofer Institute for Algorithms and Scientific Computing (SCAI) , 53757 Sankt Augustin, Germany 2 Department of Neurology, University Hospital and Faculty of Medicine Carl Gustav Carus, TUD Dresden University of Technology , Dresden, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Patricia Martins Conde 3 Luxembourg Centre for Systems Biomedicine (LCSB), University of Luxembourg , Esch-sur-Alzette, Luxembourg Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jochen Klucken 4 Centre Hospitalier de Luxembourg (CHL) , Luxembourg Find this author on Google Scholar Find this author on PubMed Search for this author on this site Holger Fröhlich 1 Department of Bioinformatics, Fraunhofer Institute for Algorithms and Scientific Computing (SCAI) , 53757 Sankt Augustin, Germany 4 Centre Hospitalier de Luxembourg (CHL) , Luxembourg Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: mohamed.aborageh{at}scai.fraunhofer.de holger.froehlich{at}scai.fraunhofer.de Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Parkinson’s disease (PD) exhibits a variety of symptoms, with approximately 25% of patients experiencing mild cognitive impairment and 45% developing dementia within ten years of diagnosis. Predicting this progression and identifying its causes remains challenging. Our study utilizes machine learning and multimodal data from the UK Biobank to explore the predictability of Parkinson’s dementia (PDD) post-diagnosis, further validated by data from the Parkinson’s Progression Markers Initiative (PPMI) cohort. Using Shapley Additive Explanation (SHAP) and Bayesian Network structure learning, we analyzed interactions among genetic predisposition, comorbidities, lifestyle, and environmental factors. We concluded that genetic predisposition is the dominant factor, with significant influence from comorbidities. Additionally, we employed Mendelian randomization (MR) to establish potential causal links between hypertension, type 2 diabetes, and PDD, suggesting that managing blood pressure and glucose levels in Parkinson’s patients may serve as a preventive strategy. This study identifies risk factors for PDD and proposes avenues for prevention. 2 Introduction Parkinson’s disease (PD) is the fastest growing neurological disease and the second most common neurodegenerative disease [ 1 ]. Although the cause of PD remains unclear in most sporadic cases, mutations in more than 20 genes have been associated with the disease, including genes such as LRRK2 and SNCA [ 2 ]. In addition, environmental risk factors have been linked to the development of PD, which can be triggered by an underlying genetic predisposition [ 3 ]. For example, occupational exposures such as pesticides are associated with an increased risk of PD [ 4 ][ 5 ][ 6 ]. Furthermore, lifestyle factors such as alcohol consumption, physical inactivity and diet [ 7 ], as well as beta-blockers [ 8 ] and comorbidities such as diabetes mellitus and arterial hypertension, have been identified as potential risk factors. Although PD is characterized primarily by three main motor symptoms (bradykinesia, tremor at rest, and rigidity); cognitive impairment is also a frequent problem in patients with PD (PwPD). Approximately 25% of PwPD exhibit mild cognitive impairment and 45% develop dementia within 10 years after diagnosis [ 9 ][ 10 ][ 11 ][ 12 ][ 13 ][ 14 ]. PD dementia (PDD) contributes to an increase in health-related expenditures and a significant decrease in quality of life, underscoring its importance to affected individuals, their caregivers and healthcare systems [ 15 ]. Furthermore, PDD constitutes one of the four milestones that occur on average four years before death, which indicates the onset of the terminal phase of the disease [ 16 ]. The underlying reason why certain PwPD develop cognitive impairment earlier in their disease course is still unknown. Previous genome-wide association studies (GWAS) report possible associations of different genetic risk variants with cognitive impairment in PD, including variants in SNCA [ 17 ][ 18 ], MAPT [ 19 ][ 20 ], TMEM175 [ 21 ] and GBA [ 22 ][ 23 ]. Statistical analysis of a cohort of 827 individuals identified age, duration of the disease, sex, and GBA status as the primary factors associated with cognitive performance and progression to dementia [ 24 ]. [ 25 ] developed machine learning models to predict cognitive outcomes in de novo diagnosed PwPD using a broad spectrum of baseline data (including cognitive test results) from the Parkinson’s Progression Markers Initiative (PPMI) cohort study. However, data from specific cohorts are not necessarily representative of the overall disease population and could therefore differ from real-world data collected in population studies or clinical routine. Moreover, models trained on data collected solely for research, such as DNA methylation, cerebral spinal fluid biomarkers and neuropsychiatric tests, are not easily applicable in clinical practice. Thus, there is a need to investigate models using more routinely collected health data, including comorbidities, lifestyle and environmental factors. The potential impact of such factors on PDD is important for shaping future prevention strategies. Our study used the UK Biobank (UKB) to a) predict dementia in people with Parkinson’s disease (PwPD) using machine learning and b) investigate modifiable risk factors that can causally influence the development of dementia in PwPD, thus informing possible prevention strategies. To our knowledge, no prior research has explored these areas. To achieve our objectives, we developed and evaluated machine learning models to predict PDD based on comorbidities, genetic, environmental, and lifestyle factors. Using Explainable AI (XAI) techniques, we identified the features that significantly affected predictions and used probabilistic graphical models (Bayesian networks) to analyze interactions among the most predictive comorbidities, environmental, lifestyle, and genetic risk factors. Finally, we performed a Mendelian Randomization (MR) analysis to confirm the probable causal relationship between hypertension, type 2 diabetes, and PDD. 3 Methods 3.1 Data Sources The UKB is a large-scale database and research resource, playing an important role in studying a wide spectrum of complex diseases and advancing their diagnosis, prevention, and treatment [ 26 ]. The cohort was released for research in April 2012, and currently includes cross-sectional data for more than 500,000 individuals from the UK. The initial assessments involved the collection of data on demographics, clinical history, and lifestyle. The participating individuals were also genotyped, providing valuable genetic data that can facilitate the study of the genetic architecture of complex diseases. To further validate our findings from UKB, we used data from the PPMI study, which provides multiple data modalities, including demographic and genetic components [ 27 ]. The data set was initially released in 2010, and currently includes data for more than 1500 patients. 3.2 Patients Selection and Variables We selected all UKB patients diagnosed with PD in the hospital according to the ICD-10 code G20 and those with Parkinson’s Dementia (F02.3). This resulted in 3,541 PwPD and 487 patients with PDD. Data were cross-sectional and included past diagnoses, comorbidities, as well as genetic, environmental, and lifestyle factors associated with the risk of PD, as identified in a large meta-analysis [ 28 ]. Following this publication, we constructed variables that reflect known risk factors associated with dementia [ 29 ], including stroke, type 2 diabetes, hypertension, depression, anxiety, hearing loss, visual loss, hypercholesterolemia and obesity. For daytime sleeping/dozing, we used measurements from the UKB field 1220 (daytime dozing / sleeping). All variables are detailed in Supplementary Table 1. For PPMI, we included PwPD and healthy controls, accounting for their cognitive status to include individuals with mild cognitive impairment (MCI) and dementia as recorded in the PPMI cognitive state table (code 2 indicating MCI and code 3 indicating dementia). Together, the data set included a total of 637 PwPD, among which 122 had PDD. 3.3 Use of Genotype Data Genotyping was performed on all UKB participants using UKBiLEVE and UK Biobank Axiom arrays. The UKBiLEVE array was initially designed and ran on approximately 50,000 participants, which was followed by the UK Biobank Axiom array designed to run on the remaining 450,000 participants. Both arrays share over 95% common content. For PPMI, more than 1500 participants were genotyped using the Illumina NeuroX array, which combines approximately 240,000 exome variants and approximately 24,000 custom variants with a focus on neurological diseases. Since PDD is defined as dementia due to PD, we hypothesized that certain genetic variants associated with PD could also contribute to the risk of PDD. To construct variables for our machine learning models, we thus obtained 3189 single nucleotide polymorphisms (SNPs) associated with PD from the ieu-b-7 dataset. In addition, variants associated with all-cause dementia might play a role. Hence, we obtained 924 dementia-associated SNPs from the finn-b-F5 Dementia dataset available on the ieugwas database (8 April 2024) [ 30 ]. We performed SNP clumping to include only highly disease-associated, rather independent SNPs, using thresholds of p < 5 E − 8, r 2 < 0.1 and a physical distance of 250 kb. This process resulted in 10 highly associated SNPs for PD and 13 for dementia. In addition, we calculated two polygenic risk scores that are published in the Polygenic Risk Score Catalog (PGS) [ 31 ], namely PGS4281 for dementia [ 32 ] and PGS777 for PDD [ 33 ]. Finally, we included 3 SNPs that have been associated with PDD from the DisGeNet gene-disease association network [ 34 ]. 3.4 Machine Learning Models Predicting PD Dementia We built three classification models to predict PDD within PwPD. This included gradient boosting [ 35 ], random forests [ 36 ] and elastic net penalized logistic regression [ 37 ]. The aim was to predict PDD using multimodal data, including SNP, polygenic risk scores, demographic characteristics, comorbidities, family history, and environmental and lifestyle-associated factors. The models were trained within a 5-fold nested cross-validation approach which splits data into a training set (80%) and a testing set (20%) while optimizing the model’s hyperparameters within an inner cross-validation on the training set. Details on hyper-parameter optimization are presented in Supplementary Table 2. We evaluated the models’ predictive performance by calculating the area under the receiver operating characteristic curve (AUC) across 20 repetitions of 5-fold nested cross-validation. To address label imbalance, class weights were applied during model training. 3.5 Making Model Predictions Explainable We calculated Shapley Additive Explanations (SHAP values) from the best performing machine learning model to understand the importance of features and the cumulative influence of different data modalities on predictions [ 38 ]. To better understand the interactions between all variables, we subsequently learned a Bayesian network (BN) from the data [ 39 ]. BNs are probabilistic graphical models that represent variables as nodes and conditional probabilistic dependencies between them as edges. We trained a BN on data from all subjects using non-parametric bootstrapping, which randomly selects samples 1000 times and learns a complete BN graph structure within each bootstrap using the tabu metaheuristic search method, which we have previously found to be the most accurate compared to multiple other BN structure learning methods for multimodal clinical data [ 40 ]. Continuous variables were discretized through clustering of k-means before learning the BN structure to account for the non-Gaussian nature of multiple features. BN structure learning was implemented using R-package bnlearn [ 41 ]. As each bootstrap sample resulted in a slightly different BN structure, a consensus structure was subsequently created by applying a conservative bootstrap probability threshold of 0.5, meaning that the edges of the averaged graph were found in at least 50% of all BN. This aims to improve the robustness and reliability of the estimates. Further details of BN learning can be found in Supplementary Table 3. 3.6 Mendelian Randomization to Estimate Causal Effects MR assesses the causal relationship between an exposure (here: a comorbidity) and an outcome (here: PDD) using genetic variants as instrumental variables (IV). The variants used should satisfy three assumptions to be considered as an IV: The variant is associated with the exposure The variant is independent of confounding factors that confound the association of the exposure to the outcome The variant is independent of the outcome given the exposure and the confounding factors In this regard, the genetic variants associated with exposure can be used as proxies to explain how exposure can influence the outcome of interest. For our study, we used MR to assess the putative causal effect of comorbidities on the risk of PDD. We obtained genetic variants for our exposures using the IEU-OpenGWAS API (ieugwasr package [ 42 ]) from summary datasets for stroke (ebi-a-GCST005838), type-2 diabetes (ebi-a-GCST90018926), hypertension (ebi-a-GCST90038604), depression (ukb-b-12064), anxiety (ukb-a-82), hypercholesterolemia (finn-b-E4 HYPERCHOL), hearing loss (finn-b-H8 CONSENHEARINGLOSS), visual loss (finn-b-H7 VISUALDISTBLIND), obesity (finn-b-E4 OBESITY), smoking status (ebi-a-GCST90029014), and daytime sleeping/dozing (ukb-b-5776). The selection of summary datasets was based on the availability of top hits. We conducted one-sample MR by assessing the associations between genetic variants and both the exposures and the outcome at the individual level. We calibrated SNP exposure estimates to our phenotypes and generated SNP outcome estimates through association testing, adjusting for age and sex. The instruments were selected by clumping to ensure that the SNPs were strongly associated with exposures and independent of each other, applying thresholds of p < 5 E − 8, r 2 < 0.1 for linkage disequilibrium and a genomic window of 250 kb. We estimated causal effects using the inverse variance weighted method (IVW) with fixed effects, which calculates the causal effect of exposure X on outcome Y . The causal effect ratio using a genetic variant i is X i /Y i , with the standard error estimated by the delta method as .The IVW estimate combines ratio estimates of the variants in a fixed-effect meta-analysis model. The success of MR depends on the three previously mentioned assumptions, which require a sensitivity analysis for the evaluation of the results [ 43 ]. We utilized MR-Egger and MR-PRESSO for this analysis. The MR-Egger method evaluates directional pleiotropy, assessing whether genetic factors have average pleiotropic effects on the outcome that differ from zero. Pleiotropy in human genetics can manifest in various forms, such as a single variant influencing multiple traits or a causal locus affecting a trait through another one. Horizontal pleiotropy occurs when a variant directly or indirectly impacts the target outcome, often by influencing other traits that causally affect it. In this context, MR-Egger provides a reliable estimate of the causal effect under the assumption of InSIDE (Instrument Strength Independent of Direct Effect) [ 44 ]. In this work, MR-PRESSO was used to investigate horizontal pleiotropy in multi-instrument summary level MR [ 45 ], which involves three steps: a global test for horizontal pleiotropy detection, removal of outliers for correction, and a significance test before and after removal of outliers. 4 Results 4.1 Demographics of Patients and Controls In UKB, PwPD were more likely to be male (61.3%) with a mean age of 62.7 years, consistent with previously reported incidences [ 46 ][ 47 ]. The PDD group showed an even higher proportion of males (69.8%) and a slightly higher mean age of 64.1 years. In PPMI, PwPD showed a proportion of male individuals similar to UKB (60.4%) with a mean age of 61.8 years. The PDD group showed a slightly lower proportion of male individuals than in UKB (61.5%), and the mean age was 67.1 years. Detailed demographic characteristics of the groups can be found in Table 1 . View this table: View inline View popup Download powerpoint Table 1: Demographic characteristics of Parkinson’s disease (PD) and Parkinson’s disease dementia (PDD) groups in the UK Biobank and PPMI datasets. Continuous variables are reported as mean (standard deviation), while categorical variables are presented as count (percentage). 4.2 Predictability of PD Dementia We evaluated penalized logistic regression, Random Forests, and XGBoost to predict PDD. Repetitive nested cross-validation showed an average AUC of 0.62 for Random Forest and logistic regression and 0.61 for XGBoost over 20 repetitions ( Figures 1 and 2 ). We carried out an ablation study to understand the contribution of different data modalities. We found that demographics, comorbidities, and genetics contributed the most, with a statistically significant average of 2% drop in AUC after removing any of them; see Tables 2 , 3 . Download figure Open in new tab Figure 1: Boxplot showing the models’ AUC via repeated nested cross-validation Download figure Open in new tab Figure 2: Variability of mean ROC curve of the Random Forest model for each fold via cross validation View this table: View inline View popup Download powerpoint Table 2: Comparison of the model’s performance showing the median AUC over 20 repeats of nested cross-validation, using all data modalities and each modality separately. A Kruskal-Wallis test was used to determine if there were statistically significant differences in performance after the removal of each modality compared to the full model. View this table: View inline View popup Download powerpoint Table 3: Ablation study results showing the median AUC over 20 repeats of nested crossvalidation after removing each data modality. A Kruskal-Wallis test was used to determine if there were statistically significant differences in performance after the removal of each modality relative to the full model. We repeated our analysis using the PPMI dataset and compared the results with those obtained from the UKB cohort. Due to the absence of several variables in the PPMI dataset that were only available in UKB, we had to take a reduced subset of variables. This subset included only SNPs, PRS, age, and sex. When training the Random Forest model on UKB with this reduced subset of variables, the cross-validated prediction performance of this reduced model was close to the original with an average AUC 0.61 ± 0.01, while the cross-validated prediction performance on PPMI was higher with an average AUC 0.65 ± 0.02 ( Figure 3 ). Download figure Open in new tab Figure 3: Comparison of the Random Forest model trained on the reduced subset of features in PPMI and UK Biobank datasets 4.3 SHAP Analysis For the following analysis, we focused on the best performing model variant after training and tuning on the entire dataset, i.e. the Random Forest model using all data modalities. SHAP analysis revealed that variables from different data modalities contribute significantly to the predictions of the model, with the polygenic risk score PGS4281 being the main predictor, followed by SNP rs769449 , age, SNP rs6859 , diagnosis of depression, sex, BMI and hypercholesterolemia. We also observed that lifestyle variables, such as tea intake and water intake, appeared among the top predictors. Figure 4 shows the impact of the top 20 variables on the model output and Figure 5 shows the relationship between variable values and model predictions using SHAP dependence. We found that higher PGS4281 scores and older age were associated with higher SHAP values, indicating a stronger influence on the predicted probability of PDD. In contrast, a higher BMI was associated with lower SHAP values, i.e. lower predicted likelihood of PDD. For SNP rs769449 , having one or two copies of the effect allele resulted in higher SHAP values. In contrast, for SNPs rs6859 and rs449647 , having more copies of the effect allele was associated with lower SHAP values. Furthermore, a higher predicted likelihood of PDD was associated with being male and having a diagnosis of depression. Download figure Open in new tab Figure 4: Beeswarm plot showing the SHAP values of the top level predictors for the best performing random forest model Download figure Open in new tab Figure 5: A grid showing the SHAP dependence plot of the top 9 features Finally, we analyzed the cumulative influence of different data modalities by summing the SHAP values of their respective variables to understand their overall contribution to model predictions. In agreement with the previously presented ablation study, we found that genetics had the highest influence (49.31%), followed by demographics (24.32%) and comorbidities (15.74%). Figure 6 presents the cumulative influence of all data modalities. Download figure Open in new tab Figure 6: A bar plot illustrating the cumulative influence of different data modalities, with values normalized to a range between 0 and 1 to facilitate comparison. The SHAP analysis of the Random Forest trained on the entire PPMI data set showed similar results with age as the main predictor, followed by PGS4281, SNP rs2927468, rs449647 and rs356219 , while sex and rs769449 had lower SHAP values. A comparison of the beeswarm plots for both datasets, highlighting the SHAP values of the top predictors, is provided in Supplementary Figure 1. 4.4 Understanding the Interactions of Predictors Following our SHAP analysis, we fitted 1000 BNs within a bootstrap procedure to better understand conditional statistical dependencies between all variables in the best performing Random Forest model trained on UKB. All edges discussed below had a bootstrap frequency of (≥ 50%), which means that they were observed in at least 50% of the BNs across all bootstrap samples; thus, these edges had high statistical confidence. A visualization of these high confidence edges as a graph is provided in Supplementary Figure 2. Overall, our analysis revealed expected connections within each data modality, including edges that point from diabetes and obesity to hypertension and hypercholesterolemia, from age to anxiety (which also had an edge toward depression), and from sex to diabetes, smoking, anxiety, hypercholesterolemia, and breastfed. In addition, we observed connections among different SNP groups, including SNPs that map to the genes TMEM175, NSF, LRRC37A2, SNCA and KANSL1 . We also observed connections between genetic and non-genetic variables, namely an edge from SNP rs1372519 located in the SNCA locus towards BMI and from rs2230288 , located in the GBA locus, and rs28399664 , located in the BCAM locus, towards exposure to air pollution. 4.5 Mendelian Randomization Identifies Causal Impact of Hypertension on PDD After calibrating the effects of the SNPs from the summary statistics and applying our instrument selection thresholds, only daytime sleep, diabetes, obesity, smoking, hearing loss, and hypertension had SNPs that met the criteria: 37 for daytime sleep, 293 for diabetes, 3 for obesity, 22 for smoking status, 8 for hearing loss, and 463 for hypertension, which makes them suitable for MR analysis. IVW analysis indicated that hypertension increased the risk of developing Parkinson’s disease (PD) dementia, with a causal estimate of 0.223 [se = 0.06, p = 0.0005], and diabetes also presented a risk with a causal estimate of 0.138 [se = 0.03, p = 0.0004]. Daytime sleep did not show a significant risk of dementia due to PD [p = 0.3], nor did obesity [p = 0.87], hearing loss [p = 0.35] or smoking status [p = 0.27]. The detailed results of our MR analysis are found in Supplementary Tables 4 and 5. MR-Egger sensitivity analysis confirmed a significant causal effect of hypertension on PDD with an estimate of 0.3 [se = 0.19, p = 0.04] but did not reach significance for diabetes [p = 0.22], daytime sleep [p = 0.27], obesity [p = 0.64] or smoking status [p = 0.78]. This may be due to MR-Egger’s lower power compared to IVW analysis due to the inclusion of the intercept term. No comorbidities showed signs of directional pleiotropy, as indicated by insignificant p-values for the MR-Egger intercept. MR-PRESSO analysis indicated horizontal pleiotropy for diabetes [rss = 338.7, p = 0.03], but causal effects remained significant after correction for outliers, with an estimate of 0.18 [se = 0.03, p = 0.000002] and no outlier distortion [p = 0.3]. The global test did not show signs of horizontal pleiotropy for hypertension or daytime sleep. Together, our MR analysis confirmed a potential causal link between hypertension and the risk of dementia from PD. Furthermore, our MR analysis indicates a likely causal impact of type 2 diabetes on the risk of developing PDD. 5 Discussion To our knowledge, our study is the first to explore the predictability of PDD in PwPD and the interplay of various genetic and non-genetic factors in UKB. We developed a machine learning model that predicts PDD in people with Parkinson’s disease (PwPD), achieving an average performance of 0.62 AUC. Although this performance may not be suitable for clinical use, it could be valuable for patient stratification to reduce sample sizes in clinical trials [ 48 ]. Additionally, it allowed us to evaluate the contributions of different data modalities, demonstrating that genetic factors had the most significant cumulative impact on the risk of PDD, followed by demographics and comorbidities. The primary predictor of the model was the polygenic risk score PGS4281, calculated for all-cause dementia using 110 risk variants [ 32 ]. In our study, this score alone allowed us to predict PDD with an AUC of 57.5%. The second most significant predictor was variant rs769449 , located in the Apolipoprotein E APOE locus. Although the APOE ϵ 4 allele is well recognized as a risk factor for sporadic Alzheimer’s disease (AD) [ 49 ], recent research shows that it also impacts α -synuclein pathology and its associated toxicity, which are critical aspects of Parkinson’s disease (PD) pathology [ 50 ]. As expected, age was among the main predictors in this model, as dementia is generally more prevalent in older people, and PwPD can develop dementia years after motor symptoms appear. However, age alone allowed one only to predict PDD slightly above the change level (AUC 56%), highlighting the need to consider different risk factors in combination. The variant rs6859 which is located in the NECTIN2 locus was among the top predictors and has previously been linked to deterioration of cognitive abilities in adults [ 51 ]. We observed a strong influence of sex as a predictor, which is consistent with reported statistics indicating that men are twice as likely as women to develop PDD [ 52 ]. Another relevant predictor was body mass index, which had an inverse relation to the probability for PDD. A high body mass index has previously been associated with a protective effect against fast cognitive decline and dementia development in PwPD [ 53 ][ 54 ][ 55 ]. We identified various comorbidities as significant predictors, including depression, anxiety, hypertension, hypercholesterolemia, and excessive daytime sleepiness. Depression and anxiety are recognized non-motor symptoms of Parkinson’s disease (PD) that can precede motor symptoms [ 56 ]. Excess daytime sleepiness is more prevalent in patients with PD with dementia compared to those with normal cognitive performance [ 57 ]. In addition, both hypertension and hypercholesterolemia are associated with an increased risk of dementia [ 58 ][ 59 ]. Looking at important genetic factors, we observed variants located in the APOE locus, including rs449647 and rs7412 . Furthermore, the variant rs157580 , located in the TOMM40 locus, and rs60049679 , located in the APOC1 locus, which make up the well-known Alzheimer’s disease risk gene cluster, APOE-TOMM40-APOC1 [ 60 ] A notable predictor was tea intake. A previous MR analysis indicated that green tea consumption may slow PD progression and protect against dementia [ 61 ]. In contrast, another MR analysis found that the consumption of more than 13 cups of tea per day was associated with an increased risk of AD dementia [ 62 ], possibly due to pesticide residues [ 63 ]. We examined interactions between phenotype-related variables and genotype-related features using BN structure learning. The expected interactions included the effect of age on hypertension, daily physical activity, and daytime sleepiness. Unexpected connections were also found, such as between maternal smoking and childhood obesity, possibly reflecting social deprivation. Furthermore, we observed links between sex, smoking status, and anxiety, as the prevalence of smoking and anxiety disorders varies between men and women [ 64 ][ 65 ]. Some interactions involved variables from different data modalities. For example, the variant rs1372519 in the SNCA locus was associated with body mass index. Although individuals with Parkinson’s disease (PwPD) generally have a lower BMI [ 66 ], a direct connection between the SNCA mutation and the differences in BMI remains undetermined. Moreover, two edges from rs2230288 in the GBA locus and rs28399664 in the BCAM locus indicated exposure to air pollution. The susceptibility of PwPD to certain gene mutations in regard to air pollution is not well documented, although similar gene-environment interactions have been identified in other diseases, such as chronic obstructive pulmonary disease (COPD) [ 67 ]. Lastly, we used MR analysis to examine the causal effects of type 2 diabetes, hypertension, hypercholesterolemia, obesity, and daytime sleepiness on the risk of developing PDD. MR analysis for other comorbidities was not possible, as none of the SNPs obtained from their GWAS summary statistics passed our instrument selection criteria. Our analysis revealed a putative causal effect of hypertension and type 2 diabetes. Studies have shown that elevated blood pressure in middle-aged populations (40-65 years) may be associated with a higher risk of cognitive impairment, while lower blood pressure in an elderly population (≥65) has been associated with a higher prevalence of dementia [ 68 ]. Another study has shown that hypertension is related to severe basal ganglia dilated perivascular space (BGdPVS), which in turn is associated with white matter hyperintensities (WMH) and cognitive decline in Parkinson’s disease (PD) [ 69 ]. Similarly, previous studies show that type 2 diabetes is associated with faster motor and cognitive impairment in PD [ 70 ], and that long-term variability in blood glucose increases the risk of dementia in PD patients [ 71 ]. Our study has several limitations, including the reliance on ICD coded reports of PD and PDD in UKB data, which may introduce unknown errors, particularly in distinguishing PDD from other types of dementia, such as Lewy Body Disorder. Furthermore, the generalizability of our findings is affected by the fact that most of the participants were white British, older on average, and predominantly male. Furthermore, the sample size for PD and PDD in UKB is still relatively small compared to the general disease population. We were able to partially address this aspect by investigating the replicability of our machine learning model and the SHAP analysis using data from the PPMI study. However, the number of PDD patients with comorbidities investigated in the UKB was inadequate to replicate our MR analysis. 6 Conclusion In conclusion, to our knowledge, our study is the first to explore the predictability of PDD in PwPD based on data from the UKB population study. We demonstrated that prediction performances on PPMI with newly diagnosed PwPD were comparable, and the impact of individual features on model predictions was similar in both PPMI and UKB. Although prediction performance may overall be insufficient for clinical application, our innovative approach combining machine learning, BN structure learning, and MR analysis has shed light on the complex interactions among genetic predisposition, comorbidities, environment, and lifestyle factors contributing to PDD. In particular, we identified potential causal links between hypertension, diabetes, and PDD, suggesting that the management of blood pressure and glucose levels in the blood in individuals with Parkinson’s might be suitable targets for preventive strategies. 8 Data Availability This research has been conducted using the UK Biobank resource under application number 67829. The genetic and phenotype datasets are not publicly available but can be accessed via the UK Biobank data access process. More details are available at http://www.ukbiobank.ac.uk/register-apply/ . PPMI data are publicly available from the Parkinson’s Progression Markers Initiative (PPMI) database www.ppmi-info.org/access-data-specimens/download-data , RRID:SCR 006431. For up-to-date information on the study, visit www.ppmi-info.org . 9 Code Availability The underlying code used for model training, mendelian randomization and learning the bayesian networks in this study will be published on GitHub upon acceptance of the paper under the CC BY-NC-ND 4.0 license and can be accessed via https://github.com/mosala777/Predicting-Dementia-in-PwPD 7 Acknowledgment This research has been conducted using the UK Biobank resource under application 67829. Data used in the preparation of this article was obtained from the Parkinson’s Progression Markers Initiative (PPMI) database www.ppmi-info.org/access-dataspecimens/download-data , RRID:SCR 006431. For up-to-date information on the study, visit www.ppmi-info.org . PPMI – a public-private partnership – is funded by the Michael J. Fox Foundation for Parkinson’s Research and funding partners. A list of names of all the PPMI funding partners can be found at www.ppmi-info.org/about-ppmi/who-we-are/study-sponsors/ . References 1. ↵ Feigin , V. L. et al. Global, regional, and national burden of neurological disorders during 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015 . Lancet Neurol 16 , 877 – 897 . ISSN: 14744422 (11 Nov . 2017 ). OpenUrl CrossRef PubMed 2. ↵ Blauwendraat , C. , Nalls , M. A. & Singleton , A. B. The genetic architecture of Parkinson’s disease . Lancet Neurol 19 , 170 – 178 . ISSN: 14744422 (2 Feb . 2020 ). OpenUrl CrossRef PubMed 3. ↵ Tsalenchuk , M. , Gentleman , S. M. & Marzi , S. J. Linking environmental risk factors with epigenetic mechanisms in Parkinson’s disease . NPJ Parkinsons Dis 9 , 123 . ISSN: 2373-8057 (1 Aug . 2023 ). OpenUrl PubMed 4. ↵ McNaught , K. S. et al. Effects of isoquinoline derivatives structurally related to 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) on mitochondrial respiration . Biochem Pharmacol 51 , 1503 – 1511 . ISSN: 00062952 (11 June 1996 ). OpenUrl CrossRef PubMed 5. ↵ Pouchieu , C. et al. Pesticide use in agriculture and Parkinson’s disease in the AGRICAN cohort study . Int J Epidemiol 47 , 299 – 310 . ISSN: 0300-5771 (1 Feb . 2018 ). OpenUrl CrossRef PubMed 6. ↵ Jo , S. et al. Association of NO 2 and Other Air Pollution Exposures With the Risk of Parkinson Disease . JAMA Neurol 78 , 800 . ISSN: 2168-6149 (7 July 2021 ). OpenUrl PubMed 7. ↵ Hughes , K. C. et al. Intake of dairy foods and risk of Parkinson disease . Neurol 89 , 46 – 52 . ISSN: 0028-3878 (1 July 2017 ). OpenUrl Abstract / FREE Full Text 8. ↵ Belvisi , D. et al. Modifiable risk and protective factors in disease development, progression and clinical subtypes of Parkinson’s disease: What do prospective studies suggest? Neurobiol Dis 134 , 104671 . ISSN: 09699961 ( Feb . 2020 ). OpenUrl CrossRef PubMed 9. ↵ Aarsland , D. et al. Mild cognitive impairment in Parkinson disease . Neurol 75 , 1062 – 1069 . ISSN: 0028-3878 (12 Sept . 2010 ). OpenUrl Abstract / FREE Full Text 10. ↵ Williams-Gray , C. H. et al. The CamPaIGN study of Parkinson’s disease: 10-year outlook in an incident population-based cohort . J Neurol Neurosurg Psychiatry 84 , 1258 – 1264 . ISSN: 0022-3050 (11 Nov . 2013 ). OpenUrl Abstract / FREE Full Text 11. ↵ Anang , J. B. et al. Predictors of dementia in Parkinson disease . Neurol 83 , 1253 – 1260 . ISSN: 0028-3878 (14 Sept . 2014 ). OpenUrl Abstract / FREE Full Text 12. ↵ Amer , H. et al. Genetic Influences on Cognition in Idiopathic Parkinson’s Disease . Neurol Res Int 2018 . ISSN: 2090-1852 ( Aug . 2018 ). 13. ↵ Planas-Ballvé , A. & Vilas , D. Cognitive Impairment in Genetic Parkinson’s Disease . Parkinsons Dis 2021 . ISSN: 2042-0080 ( Dec . 2021 ). 14. ↵ Bäckström , D. et al. Prediction and early biomarkers of cognitive decline in Parkinson disease and atypical parkinsonism: a population-based study . Brain Commun 4 . ISSN: 2632-1297 (2 Mar . 2022 ). 15. ↵ Svenningsson , P. , Westman , E. , Ballard , C. & Aarsland , D. Cognitive impairment in patients with Parkinson’s disease: diagnosis, biomarkers, and treatment . Lancet Neurol 11 , 697 – 707 . ISSN: 14744422 (8 Aug . 2012 ). OpenUrl CrossRef PubMed Web of Science 16. ↵ Kempster , P. A. , O’Sullivan , S. S. , Holton , J. L. , Revesz , T. & Lees , A. J. Relationships between age and late progression of Parkinson’s disease: a clinico-pathological study . Brain 133 , 1755 – 1762 . ISSN: 0006-8950 (6 June 2010 ). OpenUrl CrossRef PubMed Web of Science 17. ↵ Ramezani , M. et al. Investigating the relationship between the SNCA gene and cognitive abilities in idiopathic Parkinson’s disease using machine learning . Sci Rep 11 , 4917 . ISSN: 2045-2322 (1 Mar . 2021 ). OpenUrl PubMed 18. ↵ Kapasi , A. et al. A novel SNCA E83Q mutation in a case of dementia with Lewy bodies and atypical frontotemporal lobar degeneration . Neuropathology 40 , 620 – 626 . ISSN: 0919-6544 (6 Dec . 2020 ). OpenUrl CrossRef PubMed 19. ↵ Williams-Gray , C. H. et al. The distinct cognitive syndromes of Parkinson’s disease: 5 year follow-up of the CamPaIGN cohort . Brain 132 , 2958 – 2969 . ISSN: 1460-2156 (11 Nov . 2009 ). OpenUrl CrossRef PubMed Web of Science 20. ↵ Setó-Salvia , N. et al. Dementia Risk in Parkinson Disease . Arch Neurol 68 . ISSN: 0003-9942 (3 Mar . 2011 ). 21. ↵ Aarsland , D. et al. Parkinson disease-associated cognitive impairment . Nat Rev Dis Primers 7 , 47 . ISSN: 2056-676X (1 July 2021 ). OpenUrl PubMed 22. ↵ Neumann , J. et al. Glucocerebrosidase mutations in clinical and pathologically proven Parkinson’s disease . Brain 132 , 1783 – 1794 . ISSN: 1460-2156 (7 July 2009 ). OpenUrl CrossRef PubMed Web of Science 23. ↵ Setó-Salvia , N. et al. Glucocerebrosidase mutations confer a greater risk of dementia during Parkinson’s disease course . Mov Disord 27 , 393 – 399 . ISSN: 0885-3185 (3 Mar . 2012 ). OpenUrl CrossRef PubMed 24. ↵ Phongpreecha , T. et al. Multivariate prediction of dementia in Parkinson’s disease . NPJ Parkinsons Dis 6 , 20 . ISSN: 2373-8057 (1 Aug . 2020 ). OpenUrl PubMed 25. ↵ Harvey , J. et al. Machine learning-based prediction of cognitive outcomes in de novo Parkinson’s disease . NPJ Parkinsons Dis 8 , 150 . ISSN: 2373-8057 (1 Nov . 2022 ). OpenUrl PubMed 26. ↵ Sudlow , C. et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age . PLoS Med 12 , e1001779 . ISSN: 1549-1676 (3 Mar . 2015 ). OpenUrl CrossRef PubMed 27. ↵ Marek , K. et al. The Parkinson Progression Marker Initiative (PPMI) . Prog Neurobiol 95 , 629 – 635 . ISSN: 03010082 (4 Dec . 2011 ). OpenUrl CrossRef PubMed 28. ↵ Noyce , A. J. et al. Meta-analysis of early nonmotor features and risk factors for Parkinson disease . Ann Neurol 72 . ISSN: 0364-5134 (6 Dec . 2012 ). 29. ↵ Livingston , G. et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission . Lancet 404 , 572 – 628 . ISSN: 01406736 (10452 Aug . 2024 ). OpenUrl CrossRef PubMed 30. ↵ Elsworth , B. et al. The MRC IEU OpenGWAS data infrastructure Aug . 2020 . 31. ↵ Lambert , S. A. et al. The Polygenic Score Catalog: new functionality and tools to enable FAIR research . medRxiv ( 2024 ). 32. ↵ Ohta , R. , Tanigawa , Y. , Suzuki , Y. , Kellis , M. & Morishita , S. A polygenic score method boosted by non-additive models . Nat Commun 15 , 4433 . ISSN: 2041-1723 (1 May 2024 ). OpenUrl PubMed 33. ↵ Liu , G. et al. Genome-wide survival study identifies a novel synaptic locus and polygenic score for cognitive progression in Parkinson’s disease . Nat Genet 53 , 787 – 793 . ISSN: 1061-4036 (6 June 2021 ). OpenUrl CrossRef PubMed 34. ↵ Piñero , J. et al. The DisGeNET knowledge platform for disease genomics: 2019 update . Nucleic Acids Res . ISSN: 0305-1048 ( Nov . 2019 ). 35. ↵ Friedman , J. H. Greedy function approximation: A gradient boosting machine . Ann Stat 29 . ISSN: 0090-5364 (5 Oct . 2001 ). 36. ↵ Breiman , L. Random Forests . Machine Learning 45 , 5 – 32 . ISSN: 08856125 (1 2001 ). OpenUrl CrossRef 37. ↵ Zou , H. & Hastie , T. Regularization and Variable Selection Via the Elastic Net . J R Stat Soc Series B Stat Methodol 67 , 301 – 320 . ISSN: 1369-7412 (2 Apr . 2005 ). OpenUrl CrossRef 38. ↵ Lundberg , S. M. & Lee , S.-I. A unified approach to interpreting model predictions in Proceedings of the 31st International Conference on Neural Information Processing Systems ( Curran Associates Inc., Long Beach , California, USA , 2017 ), 4768 – 4777 . ISBN: 9781510860964 . 39. ↵ Koller , D. & Friedman , N. Probabilistic Graphical Models: Principles and Techniques - Adaptive Computation and Machine Learning ISBN: 0262013193 ( The MIT Press , 2009 ). 40. ↵ Sood , M. et al. Realistic simulation of virtual multi-scale, multi-modal patient trajectories using Bayesian networks and sparse auto-encoders . Sci Rep 10 . ISSN: 2045-2322 (1 July 2020 ). 41. ↵ Scutari , M. & Denis , J.-B. Bayesian Networks with Examples in R 2nd . ISBN 978-0367366513 ( Chapman and Hall , Boca Raton , 2021 ). 42. ↵ Hemani , G. , Elsworth , B. , Palmer , T. & Rasteiro , R. ieugwasr: Interface to the ‘OpenGWAS’ Database API R package version 1.0.1 , https://mrcieu.github.io/ieugwasr/ ( 2024 ). 43. ↵ Zhang , W. & Ghosh , D. A General Approach to Sensitivity Analysis for Mendelian Randomization . Stat Biosci 13 , 34 – 55 . ISSN: 1867-1764 (1 Apr . 2021 ). OpenUrl PubMed 44. ↵ Burgess , S. & Thompson , S. G. Interpreting findings from Mendelian randomization using the MR-Egger method . Eur J Epidemiol 32 , 377 – 389 . ISSN: 0393-2990 (5 May 2017 ). OpenUrl CrossRef PubMed 45. ↵ Verbanck , M. , Chen , C.-Y. , Neale , B. & Do , R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases . Nat Genet 50 , 693 – 698 . ISSN: 1061-4036 (5 May 2018 ). OpenUrl CrossRef PubMed 46. ↵ Wooten , G. F. Are men at greater risk for Parkinson’s disease than women? J Neurol Neurosurg Psychiatry 75 . ISSN: 0022-3050 (4 Apr . 2004 ). 47. ↵ Taylor , K. S. M. , Cook , J. A. & Counsell , C. E. Heterogeneity in male to female risk for Parkinson’s disease . J Neurol Neurosurg Psychiatry 78 . ISSN: 0022-3050 (8 Aug . 2007 ). 48. ↵ Hähnel , T. et al. Progression subtypes in Parkinson’s disease identified by a data-driven multi cohort analysis . NPJ Parkinsons Dis 10 , 95 . ISSN: 2373-8057 (1 May 2024 ). OpenUrl PubMed 49. ↵ Van der Kant , R. , Goldstein , L. S. B. & Ossenkoppele , R. Amyloid-β-independent regulators of tau pathology in Alzheimer disease . Nat Rev Neurosci 21 , 21 – 35 . ISSN: 1471-003X (1 Jan . 2020 ). OpenUrl CrossRef PubMed 50. ↵ Davis , A. A. et al. APOE genotype regulates pathology and disease progression in synucleinopathy . Sci Transl Med 12 . ISSN: 1946-6234 (529 Feb . 2020 ). 51. ↵ Rajendrakumar , A. L. , Arbeev , K. G. , Bagley , O. , Yashin , A. I. & Ukraintseva , S. The SNP rs6859 in NECTIN2 gene is associated with underlying heterogeneous trajectories of cognitive changes in older adults . BMC Neurol 24 , 78 . ISSN: 1471-2377 (1 Feb . 2024 ). OpenUrl CrossRef PubMed 52. ↵ Mouton , A. et al. Sex ratio in dementia with Lewy bodies balanced between Alzheimer’s disease and Parkinson’s disease dementia: a cross-sectional study . Alzheimers Res Ther 10 , 92 . ISSN: 1758-9193 (1 Dec . 2018 ). OpenUrl PubMed 53. ↵ Yoo , H. S. , Chung , S. J. , Lee , P. H. , Sohn , Y. H. & Kang , S. Y. The Influence of Body Mass Index at Diagnosis on Cognitive Decline in Parkinson’s Disease . J Clin Neurol 15 . ISSN: 1738-6586 (4 2019 ). 54. ↵ Rahmani , J. et al. Body mass index and risk of Parkinson, Alzheimer, Dementia, and Dementia mortality: a systematic review and dose–response meta-analysis of cohort studies among 5 million participants . Nutr Neurosci 25 . ISSN: 1028-415X (3 Mar . 2022 ). 55. ↵ Natale , G. , Zhang , Y. , Hanes , D. W. & Clouston , S. A. Obesity in Late-Life as a Protective Factor Against Dementia and Dementia-Related Mortality . Am J Alzheimers Dis Other Demen 38 . ISSN: 1533-3175 ( Jan . 2023 ). 56. ↵ Schrag , A. & Taddei , R. N. in International Review of Neurobiology 623 – 655 ( Elsevier , 2017 ). ISBN: 978-0-12-813708-6 . 57. ↵ Liu , H. et al. Excessive Daytime Sleepiness in Parkinson’s Disease . Nat Sci Sleep Volume 14 . ISSN: 1179-1608 ( Sept . 2022 ). 58. ↵ McGrath , E. R. et al. Blood pressure from mid-to late life and risk of incident dementia . Neurol 89 , 2447 – 2454 . ISSN: 0028-3878 (24 Dec . 2017 ). OpenUrl Abstract / FREE Full Text 59. ↵ Iwagami , M. et al. Blood cholesterol and risk of dementia in more than 1·8 million people over two decades: a retrospective cohort study . Lancet Healthy Longev 2 , e498 – e506 . ISSN: 26667568 (8 Aug . 2021 ). OpenUrl 60. ↵ Kamboh , M. I. et al. Genome-wide association study of Alzheimer’s disease . Transl Psychiatry 2 , e117 – e117 . ISSN: 2158-3188 (5 May 2012 ). OpenUrl 61. ↵ Li , C. , Lin , J. , Yang , T. & Shang , H. Green Tea Intake and Parkinson’s Disease Progression: A Mendelian Randomization Study . Front Nutr 9 . ISSN: 2296-861X ( May 2022 ). 62. ↵ Sun , Y. et al. Extra cup of tea intake associated with increased risk of Alzheimer’s disease: Genetic insights from Mendelian randomization . Front Nutr 10 . ISSN: 2296-861X ( Jan . 2023 ). 63. ↵ De Andrade Arruda Fernandes , I. et al. The bitter side of teas: Pesticide residues and their impact on human health . Food Chem Toxicol 179 , 113955 . ISSN: 02786915 ( Sept . 2023 ). OpenUrl PubMed 64. ↵ Waldron , I. Patterns and causes of gender differences in smoking . Soc Sci Med 32 , 989 – 1005 . ISSN: 02779536 (9 1991 ). OpenUrl CrossRef PubMed Web of Science 65. ↵ McLean , C. P. , Asnaani , A. , Litz , B. T. & Hofmann , S. G. Gender differences in anxiety disorders: Prevalence, course of illness, comorbidity and burden of illness . J Psychiatr Res 45 , 1027 – 1035 . ISSN: 00223956 (8 Aug . 2011 ). OpenUrl CrossRef PubMed 66. ↵ Van der Marck , M. A. et al. Body mass index in Parkinson’s disease: A meta-analysis . Parkinsonism Relat Disord 18 , 263 – 267 . ISSN: 13538020 (3 Mar . 2012 ). OpenUrl CrossRef PubMed 67. ↵ Melbourne , C. A. et al. Genome-wide gene-air pollution interaction analysis of lung function in 300,000 individuals . Environ Int 159 , 107041 . ISSN: 01604120 ( Jan . 2022 ). OpenUrl PubMed 68. ↵ Sierra , C. Hypertension and the Risk of Dementia . Front Cardiovasc Med 7 . ISSN: 2297-055X ( Jan . 2020 ). 69. ↵ Shin , N.-Y. et al. Adverse effects of hypertension, supine hypertension, and perivascular space on cognition and motor function in PD . NPJ Parkinsons Dis 7 , 69 . ISSN: 2373-8057 ( Aug . 2021 ). OpenUrl PubMed 70. ↵ Chohan , H. et al. Type 2 Diabetes as a Determinant of Parkinson’s Disease Risk and Progression . Mov Disord 36 , 1420 – 1429 . ISSN: 0885-3185 (6 June 2021 ). OpenUrl CrossRef PubMed 71. ↵ Kang , S. H. et al. Fasting glucose variability and risk of dementia in Parkinson’s disease: a 9-year longitudinal follow-up study of a nationwide cohort . Front Aging Neurosci 15 . ISSN: 1663-4365 ( Jan . 2024 ). View the discussion thread. Back to top Previous Next Posted January 28, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Predicting Dementia in People with Parkinson’s Disease Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Predicting Dementia in People with Parkinson’s Disease Mohamed Aborageh , Tom Hähnel , Patricia Martins Conde , Jochen Klucken , Holger Fröhlich medRxiv 2025.01.27.25321134; doi: https://doi.org/10.1101/2025.01.27.25321134 Share This Article: Copy Citation Tools Predicting Dementia in People with Parkinson’s Disease Mohamed Aborageh , Tom Hähnel , Patricia Martins Conde , Jochen Klucken , Holger Fröhlich medRxiv 2025.01.27.25321134; doi: https://doi.org/10.1101/2025.01.27.25321134 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Genetic and Genomic Medicine Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (299) Cardiovascular Medicine (4422) Dentistry and Oral Medicine (443) Dermatology (382) Emergency Medicine (607) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1507) Epidemiology (15219) Forensic Medicine (30) Gastroenterology (1123) Genetic and Genomic Medicine (6584) Geriatric Medicine (667) Health Economics (997) Health Informatics (4524) Health Policy (1367) Health Systems and Quality Improvement (1611) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15910) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (145) Nephrology (667) Neurology (6587) Nursing (345) Nutrition (998) Obstetrics and Gynecology (1143) Occupational and Environmental Health (956) Oncology (3331) Ophthalmology (970) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (663) Pediatrics (1690) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5437) Public and Global Health (9218) Radiology and Imaging (2195) Rehabilitation Medicine and Physical Therapy (1369) Respiratory Medicine (1195) Rheumatology (593) Sexual and Reproductive Health (709) Sports Medicine (529) Surgery (709) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ff7acb1ebd409d6',t:'MTc3OTQwOTQzMQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();