Risk factors and predictors for Lewy body dementia: a systematic review

Abstract

Lewy body dementia (LBD), including Parkinson’s disease dementia (PDD) and dementia with Lewy bodies (DLB), is a common and burdensome dementia. Determining risk factors and predictors can provide insights into pathogenesis and guide treatment efforts. In this systematic review, we searched PubMed, Embase, and Web of Science for longitudinal studies assessing risk/prodromal factors; including participants without dementia at baseline; with LBD as the outcome; with good/high quality based on the Newcastle-Ottawa Quality Assessment Scale. Across 167 included studies, more consistently reported factors were older age, male sex, APOEe4, GBA, changes in cognition, mood, behavior, sleep, gait/posture, speech, parkinsonism, smell loss, autonomic dysfunction, white matter disease on MRI, lower CSF amyloid β42 and higher CSF/blood neurofilament light chain. The majority focused on clinical factors preceding PDD with cohorts from North America and Europe, limiting generalizability. Further efforts with more representative cohorts are needed to better identify people at risk for LBD.

Introduction

Lewy body dementia (LBD), including dementia with Lewy bodies (DLB) and Parkinson’s disease dementia (PDD), is the second most common neurodegenerative dementia¹. Parkinson’s disease (PD) is the fastest growing neurological disease globally, with up to 80% of the affected individuals at risk for dementia development during disease course^2,3. Within dementia cases, DLB can account for 7.5% and PDD can account for 3.6%^4,5. Both dementias include a range of cognitive, behavioral, and motor symptoms with substantial burden on the individuals, their loved ones, healthcare, and the society^1,6. Clinically, PD and DLB can be differentiated by the interval between dementia and parkinsonism onset. For PDD, dementia occurs at least a year after the onset of parkinsonism⁷. For DLB, dementia can occur before, at the same time, or within the first year of parkinsonism⁸. Both diseases share many overlapping features and are pathologically diagnosed by alpha-synuclein aggregation with Lewy bodies and Lewy neurites in the brain¹. Accordingly, PDD and DLB are at times grouped together as LBD in research cohorts.

Despite the growing prevalence and significant burden associated with LBD, only symptomatic treatments providing limited relief are available. Disease modification remains the priority in LBD research^9,10. However, these efforts are limited by the diagnostic challenges at the clinic. Clinical heterogeneity for both PDD and DLB underscore the importance of biomarkers to improve the diagnostic accuracy. Fortunately, recent advances in biomarkers are promising¹⁰. In fact, these biomarkers detecting the underlying synucleinopathy and neurodegenerative changes can help identify people even before the development of dementia¹⁰. Identifying people in the prodromal stages can increase the chances of prevention. Furthermore, a better understanding of risk factors can provide more insight into disease mechanisms to help improve both the diagnosis and treatment efforts. The number and strength of predictive models continuously increase in Alzheimer’s disease (AD) and dementia^11,12. While our aim in this review is not to build and validate a risk/predictive model for LBD, a summary of literature for the risk and predictive factors in LBD can lead to future efforts to develop models for clinic and research. To support such efforts, we conducted a systematic review examining the risk factors for LBD. We included studies in LBD, as well as individually in PDD and DLB to provide a more comprehensive review of the literature. As the pathophysiology of LBD likely involves multiple mechanisms, we report a summary of studies on non-modifiable (e.g., age, sex, genetics), modifiable (e.g., education, health conditions, medications, lifestyle factors) and clinical (e.g., clinical signs, imaging, biomarkers) risk and predictive factors.

Results

Longitudinal studies including participants without dementia at baseline with DLB, PDD, or LBD as the outcome assessing a risk, prodromal, or predictive factor were identified through searches of PubMed, Embase, and Web of Science for peer-reviewed articles published by July 19th, 2024. The details of the included 167 studies are summarized in Fig. 1 and Supplementary Tables 1–4. All included studies for data extraction and synthesis had good/high quality based on the Newcastle-Ottawa Quality Assessment Scale. Majority of the studies focused on PDD as the outcome (86.2%, n = 144), with a smaller number of studies for DLB (12.0%, n = 20) and LBD (1.8%, n = 3). The majority were conducted in cohorts from a single country (91.6%, n = 153), with a small number including cohorts from different countries (8.4%, n = 14). Only one study included participants from Africa; three studies included participants from South America; seven included participants from Oceania; 44 included participants from Asia; 52 included participants from North America; and 79 included participants from Europe (Fig. 2). Sample sizes for participants assessed for risk ranged from 15 to 437,045 people.

Fig. 1. Factors associated with Lewy body dementia reported in at least 2 studies after considering contradicting reports and overlapping cohorts.

All factors were reported in Parkinson’s disease dementia (PDD). Factors reported for both PDD and dementia with Lewy bodies are bolded and marked with *.

Fig. 2.

Number of publications from each country.

Two studies did not report age at baseline; two did not report sex ratios; five did not report the clinical diagnostic criteria for dementia. Mean follow-up duration ranged from 1 to 12 years. Mean age at baseline for the participants at risk ranged from 57 to 81 years. One study was conducted only in males, one study was conducted only in females. Out of the rest of the 163 studies, the ratios for females in the cohorts ranged from 11.4% to 72.6%. Out of the 162 studies, the majority (89.5%, n = 145) utilized clinical diagnostic criteria/medical record diagnostic codes, and 10.5% (n = 17) utilized global cognitive screening tools for the diagnosis of dementia. Out of the 167 studies, 31.1% investigated non-modifiable (total n = 52, 33 demographic, 24 genetic; all with PDD; Supplementary Table 1), 34.7% investigated modifiable (total n = 58, 15 medication, 43 health condition, eight social; Supplementary Table 2), and 65.9% investigated clinical factors (total n = 110; 77 clinical signs and symptoms, 30 imaging, 19 fluid biomarkers; Supplementary Tables 3, 4) (Fig. 1 for the summary).

Studies for PDD

For dementia risk in PD cohorts, 52 studies included non-modifiable, 50 included modifiable, and 96 included clinical factors (Fig. 1, Supplementary Tables 1–4). Most used diagnostic criteria (56.3%) to identify PDD in the studies was the MDS PDD criteria⁷. Other diagnostic approaches included Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria, International Classification of Diseases (ICD) codes, and global cognitive screening measures (e.g., Mini Mental State Examination [MMSE], Montreal Cognitive Assessment [MoCA]).

Non-modifiable factors for PDD

Older age^13–37, and male sex^{14,15,36,38–41} were the demographic factors associated with PDD risk. There were also studies contradicting these reports; one study reported younger age was associated with increased dementia⁴², and several studies reported no significant associations for sex^{17,30,42–45}.

One of the most reported genetic risk factors was GBA^40,46–52. Only one study did not identify a significant association between GBA and PDD risk⁵³. Severe GBA1 mutations increased PDD risk more than mild mutations⁴⁷. One study noted a sex interaction; males with GBA mutations had a higher PDD risk than females⁴⁰. Another study noted an interaction with APOEe4; being a carrier of both APOEe4 and GBA mutations increased risk more than only one risk genotype⁵¹. While two studies reported APOE was not significantly associated with risk^40,54, more reported APOEe4 as a risk factor for PDD^{44,45,51–53,55}. There was no sex interaction with APOE for dementia risk⁴⁴. Three studies with a sample size ~100 people with PD, including two studies with the same cohort, reported MAPT H1/H1 genotype as a PDD risk factor^22,35,56. However, a more recent study with a sample size of 1002 people with PD, including the cohort from those two studies with the same cohort and also other cohorts from UK, Sweden, and Norway, did not find MAPT significant⁵¹. High COMT activity haplotypes were associated with higher risk⁵⁷, while two studies did not find significant associations between COMT and PDD risk^35,58. Other reported risk factors included long leukocyte telomere length⁵⁹, PITX3 C allele⁶⁰, SNCA Rep1 263 allele⁶¹, DRD2 ⁹⁵⁷T/T genotype⁵⁸, RIMS2, TMEM108, and WWOX⁵². AQP4 A allele was associated with slower conversion⁶². SNCA (rs356219) was not significantly associated with dementia risk⁵¹. Shorter TOMM40 ‘523’ poly-T repeat was associated with higher risk only for female APOEe3/e3 carriers, with this finding becoming nonsignificant after Bonferroni correction⁶³. Lysosomal pathway-specific polygenic risk score also predicted earlier progression to dementia in people with PD with a low likelihood for AD co-pathology⁶⁴.

Modifiable factors for PDD

Significant risk factors for PDD included lower years of education^28,65, frailty⁶⁶, smoking^67,68, constipation^13,69, insomnia⁷⁰, rapid eye movement sleep behavior disorder (RBD)^{13,14,31,37,45,71–73}, orthostatic hypotension^14,15,71,74, autonomic dysfunction^70,75, hypertension⁷⁶, dyslipidemia⁷⁶, diabetes mellitus¹³, chronic kidney disease⁷⁷, chronic obstructive pulmonary disease⁷⁷, stroke⁷⁷, bipolar spectrum disorder⁷⁸, delirium⁷⁹, abnormal color vision⁷¹, abnormal stereopsis¹³, and olfactory dysfunction^13,80–82. Detrusor overactivity was associated with higher dementia risk in a cohort of only females with PD⁸³. High variability in fasting glucose across visits was associated with PDD risk, even among individuals without diabetes⁸⁴. Among people with metabolic syndrome, those treated for hypertension, diabetes, and hypertriglyceridemia had a lower risk for PDD compared to no treatment⁸⁵. Behavioral problems including impulse control disorders⁷⁴, mood changes (i.e., depression^13,33,48,76, apathy²⁸), psychosis (i.e., hallucinations^{14,43,86–91}, delusions⁸⁹), and hyperactivity increased risk⁹². Lack of phone use, used as a measure of social engagement, predicted PDD²⁰. There were also studies reporting education^30,43, diabetes mellitus^68,93, kidney disease or estimated glomerular filtration rate⁹⁴, olfactory dysfunction⁷¹, depression³⁰, smoking, hypertension, stroke/transient ischemic attack, coronary heart disease, and atrial fibrillation⁹³ were not significant factors. Compared to being under-/normal-weight, overweight/obesity was associated with a lower risk for PDD⁹⁵.

Chinese herbal medicine⁷⁷, and dihydropyridine calcium channel blocker⁹⁶ were associated with lower risk. Statins were associated with lower risk in one study⁷⁷, and higher risk in another study⁹⁷. Although anticholinergics were not individually associated with dementia risk⁹⁸, cumulative dose effect⁹⁸ and exposure for 6 months or more⁹⁹ increased PDD risk. These two studies for anticholinergics were conducted using data from the National Health Insurance Research Database of Taiwan, however the cohorts were not fully overlapping^98,99. Higher levodopa dose^18,28, levodopa induced dyskinesia¹⁰⁰, confusion or psychosis³³ were also associated with higher PDD risk.

Clinical factors for PDD

Subjective cognitive decline⁴⁵, difficulty performing instrumental activities of daily living¹⁰¹ and mild cognitive impairment (MCI)^{14,15,21,23,27,36,71,81,102–109} predicted PDD. Multi-domain MCI was associated with higher risk than single domain^107,109,110 with executive and attention deficits as the stronger predictors¹⁰⁷. For people with single domain MCI, one report showed a higher dementia risk for non-amnestic than amnestic profile¹¹¹, while another report reported the opposite with higher dementia risk for amnestic than non-amnestic profile¹¹². Impairment on global cognitive measures^29,67 including MMSE^{25,45,48,74,102,113}, MoCA¹⁰², and clock drawing¹¹⁴ predicted PDD. Neuropsychological test performances indicating worse function for executive^{30,32,109,110,115–120}, visuospatial^{27,35,80,109,110,115,118,119,121,122}, attention/working memory^{34,109,110,119,122–124}, processing speed^34,114,123, memory^{32,34,43,109,115,119,125,126} and language^{43,109,115,119,122,124,126} were also found to predict PDD.

Older age at PD onset^{39,40,43,48,65,76,118}, more severe motor symptoms^{14,16–19,21,28,30,33,34,36,37,39,43,45,48,67,81,113,125,127,128} and non-tremor-dominant subtype^{37,86,87,91,129,130} were associated with higher PDD risk. Individual motor symptoms associated with risk included changes in gait/posture^{14,34,71,87,131}, speech^67,87,132, rigidity⁸⁷, upper limb bradykinesia⁶⁷, masked face³³. There were a lower number of studies reporting age at PD onset^16,26,27,128, motor symptom severity²⁷, and PD subtype were not significant factors¹²¹. More frequent and severe non-motor symptoms posing more burden on the individual than motor symptoms also increased the PDD risk¹³³.

Frontal and anterior cingulate cortical thinning¹²², more white matter disease burden^{41,108,131,134}, larger choroid plexus volume¹³⁵, larger third ventricle volume³⁴, reduced hippocampal volume¹³⁶, reduced nucleus basalis of Meynert volume¹³⁷, and more heterogeneity in the caudate texture on the less-affected side³² were identified as PDD predictors in magnetic resonance imaging (MRI) studies. Higher scores on the grand total of electroencephalogram (EEG) scale, with more diffuse slowness and focal disturbances were associated with PDD risk¹³⁸. Rapid eye movement sleep and wakefulness slowing ratios in temporal and occipital regions, and dominant occipital frequency predicted PDD¹³⁹. Low cardiac iodine-123-meta-iodobenzylguanidine (MIBG) uptake (H/M ratio < 1.35)²⁴; low beta power on magnetoencephalogram (MEG) combined with impairment on executive tests were significant predictors¹¹⁶. Atypical fludeoxyglucose positron emission tomography (FDG PET) pattern, such as AD-like pattern or DLB-like pattern, characterized by metabolic alterations in the posterior parietal-occipital regions, were associated with higher risk in a PD cohort from Italy with dementia outcome within four- and eight-year time points^140,141. Two ¹⁸F-FP-CIT PET studies with overlapping cohorts from South Korea noted reduced uptake in the anterior putamen as a PDD predictor^41,142, while one study found no significant association for putaminal dopamine uptake¹⁴³. Other significant predictors with ¹⁸F-FP-CIT PET included low cingulate island sign ratio (lower regional uptake in posterior cingulate relative to the precuneus and cuneus compared to mean ratio for controls)¹⁴⁴, lower uptake in posterior putamen¹²⁷, higher cerebral perfusion on early phase scans¹⁴⁵, and reduced uptake in frontal, parietal, temporal, and lateral occipital regions compared to preserved cortical uptake¹⁴⁶. For people with MCI, who subsequently developed PD with half of them converting to PDD during the ~5-year follow-up period, PDD risk was associated with reduced putamen-to-caudate striatal binding ratio in the less affected hemisphere with dopamine transporter single photon emission computed tomography (SPECT), coupled with hypometabolism of temporal lobes on FDG PET¹²⁴.

The most reported cerebrospinal fluid (CSF) biomarker for PDD was lower levels of amyloid β^{23,55,81,122,133,147,148}. Specifically, amyloid β42 was noted as a risk factor^23,55,81,148, while amyloid β40 and amyloid β38 were not associated with risk²³. Higher neurofilament light chain (NfL)^147,149, higher heart fatty acid-binding protein¹⁴⁷, higher glial fibrillary acidic protein¹⁵⁰, and lower glucocerebrosidase activity¹⁵¹ were also associated with risk. Total and phosphorylated tau₁₈₁ (p-tau₁₈₁) levels were not individually associated with risk^23,148, although p-tau₁₈₁ can be helpful for prediction when combined with lower levels of amyloid β42 and higher serum NfL¹⁴⁸. In addition, ratios of CSF biomarkers, such as higher total tau/alpha synuclein and total tau/amyloid β1-42+alpha synuclein, were significant predictors for PDD¹⁵². For blood, higher NfL^53,148,153, lower B12¹⁵⁴, lower zinc (Zn) levels¹⁵⁵, higher levels of a panel including hydroxy-isoleucine, His-Asn-Asp-Ser, Alanyl-alanine, Putrescine [-2H], 3,4-Dihydroxyphenylacetone³⁴ in the serum, and lower epidermal growth factor levels in the plasma¹⁵⁶ were associated with higher PDD risk.

Studies for DLB

For DLB, seven studies investigated modifiable, and 13 investigated clinical factors (Fig. 1, Supplementary Tables 2–4). We did not identify any studies for non-modifiable factors. The majority (75%) were conducted with RBD cohorts. The most used diagnostic approach for DLB was the available McKeith DLB clinical diagnostic criteria at the time of the study (70%)^8,157. Other diagnostic approaches included DSM criteria and ICD codes.

Non-modifiable factors for DLB

Herpes simplex virus, but not varicella zoster virus¹⁵⁸, and adult attention deficit-hyperactivity disorder¹⁵⁹ were reported as risk factors for DLB. In a US cohort of males, Hart and colleagues focused on medications used for benign prostatic hyperplasia comparing α-1 adrenergic receptor antagonists that also bind to and activate an adenosine triphosphate (ATP)-producing enzyme in glycolysis (terazosin, doxazosin, and alfuzosin) and other medications that do not increase ATP (α-1 adrenergic receptor antagonist tamsulosin, and 5α-reductase inhibitor)¹⁶⁰. DLB risk was lower for males on terazosin, doxazosin, or alfuzosin compared to males taking tamsulosin or 5α-reductase inhibitor, with similar risk for tamsulosin and 5α-reductase inhibitor. Two studies noted RBD as a risk factor for DLB^161,162. For people with RBD, cardiovascular disease, hypertension, hypercholesterolemia, and diabetes were not associated with DLB risk¹⁶³. People with RBD with residual injurious symptoms after being treated with clonazepam and/or melatonin had higher DLB risk¹⁶⁴.

Clinical factors for DLB

Conversion to dementia for people with RBD was best predicted by attention, executive, and memory tests including Stroop Color Word Test, Trail Making Test Part B, Color Trails Test, digit span backward, verbal fluency, and word learning tests^165–167. Visuospatial deficits with lower scores on Figure copy^166,167, false noise errors on the noise pareidolia test¹⁶⁸, overall MCI¹⁶⁶ and lower MoCA scores¹⁶⁷ were also predictors for DLB in people with RBD.

For people with MCI, parkinsonism, cognitive fluctuations, RBD, visuospatial deficit, and impaired letter fluency were associated with DLB risk¹⁶⁹. In a retrospective study including people with DLB and controls, the presence of two and more core clinical features (parkinsonism, cognitive fluctuations, RBD, visual hallucinations); or one or more core clinical features combined with apathy, depression, or anxiety differentiated people with prodromal DLB from controls¹⁷⁰.

All the studies investigating imaging and biomarkers for risk included RBD cohorts. Increased hippocampal perfusion on SPECT¹⁷¹, hyperechogenicity of the substantia nigra on transcranial sonography¹⁷², and severe phasic electromyography (EMG) activity compared to mild phasic EMG activity¹⁷³ were reported as DLB predictors. Interhemispheric laterality for striatal dopamine transporter binding on dopamine transporter (DaT) SPECT did not predict DLB¹⁷⁴. Rahayel and colleagues computed a brain-clinical signature combining brain deformation score on MRI and clinical variables¹⁷⁵. The combination of MCI, and akinetic-rigid motor phenotype as the clinical variables with atrophy in the basal ganglia, thalamus, amygdala, frontotemporal gray and white matter, and subarachnoid/ventricular expansion as the deformation variables predicted DLB.

Lower N-acetylneuraminic acid in glycoproteins in the serum¹⁷⁶ and lower CSF amyloid β42 levels¹⁷⁷ predicted DLB. CSF p-tau, CSF total tau and CSF/serum albumin ratio were not significantly associated with risk¹⁷⁷.

Studies for LBD

Two studies focused on modifiable, and one study focused on clinical markers for LBD (Supplementary Tables 2, 4). We did not identify any studies focusing on non-modifiable factors. Diagnostic approaches included McKeith 2017 DLB clinical diagnostic criteria⁸, DSM V criteria, ICD 9 Clinical Modification and Read codes.

Both studies for modifiable factors were conducted retrospectively leveraging large national healthcare databases. Prescription of nonsteroidal anti-inflammatory drugs (NSAIDs) and glucocorticoids over 10 years prior to the dementia diagnosis were associated with higher risk for LBD¹⁷⁸. Cardiovascular diseases treated with anti-hypertensives, cholesterol-lowering agents, and anti-diabetics were associated with lower LBD risk¹⁷⁹.

In the only prospective study for LBD with an RBD cohort, reduced perfusion flow in precuneus, posterior cingulate, and parietal association cortex on the brain perfusion ^99mTc-ECD SPECT was associated with higher risk, without a significant association for the cingulate island sign¹⁸⁰.

Discussion

In this systematic review, we identified studies on a range of non-modifiable, modifiable, and clinical risk factors and predictors for LBD, with the majority including cohorts in North America and Europe and focusing on clinical factors and PDD.

Older age and male sex were risk factors for PDD across several studies, with a lower number of studies suggesting otherwise for age and lack of significance for sex. While both PDD and DLB prevalence typically increases with age, with age at onset in the late fifties to early seventies¹, we did not identify any longitudinal studies focused on demographics for DLB or LBD. Interestingly, Fink and colleagues noted that the sex difference in their analysis for PDD risk disappeared after accounting for sex-specific survival patterns, and modifiable risk factors such as cardiovascular diseases impact PDD risk differently by sex³⁹. Similar to PDD, DLB prevalence is also suggested to be higher for males than females, although this sex difference is not consistent across studies and may disappear with older age¹⁸¹. Impact of genetic variants on PDD risk can also differ by sex as noted for GBA⁴⁰, but not APOE⁴⁴. Thus, the interplay between biological factors needs to be considered for LBD risk.

More frequently studied genetic factors were GBA, APOE, and MAPT. There is an increasing number of studies focused on genetic risk in LBD; however, these studies were mostly cross-sectional and thus not included in our review. The most recent work from the International LBD Genomics Consortium showed GBA, BIN1, TMEM175, SNCA-AS1, and APOE were associated with LBD risk¹⁸². However, this cohort consisted of participants with European ancestry, and these genes were not found to be significant in a smaller study consisting of participants from Japan¹⁸³. This study by Kimura and colleagues identified another gene, CDH23, to be associated with LBD risk¹⁸³. Different findings in these two recent studies underscore the importance of investigating LBD genetic risk factors in different populations. As having both APOEe4 and GBA mutations further enhanced PDD risk⁵¹, utilization of polygenic risk scores to sum the effects of multiple variants can be helpful, although they can fall short if not taking gene-gene interactions into account¹⁸⁴.

Education is frequently associated with better cognitive reserve and lower dementia risk¹⁸⁵. It can impact dementia risk differently based on gender, ethnic, and racial group¹⁸⁶. The number of studies focusing on education for PDD risk were limited in our review. Additionally, these studies examined years of education rather than education quality, which can also play a role in dementia risk¹⁸⁷ and should be investigated for LBD. Lifestyle factors can predict or increase the risk for LBD. Smoking is associated with reduced PD risk¹⁸⁸, and higher dementia risk¹⁸⁵. While there are not many studies focused on smoking in LBD, findings suggest a potential association between smoking and higher PDD risk^67,68. Social isolation is a risk factor for dementia¹⁸⁵, and we identified one study which indicates lack of phone use can signal progression to dementia in people with PD²⁰.

Olfactory dysfunction and RBD are considered predictors for both PDD and DLB¹, as supported by the studies in our review. The risk level for people with RBD can be even higher if they experience autonomic disturbances such as constipation¹³, or sustain injuries during dream enactment despite symptomatic treatment¹⁶⁴. Interestingly, cardiovascular diseases were not consistently associated with higher PDD or DLB risk in the studies included in our review, despite being a common risk factor for dementia¹⁸⁵. Medications may have impacted these findings, as several studies identified a lower LBD risk for people treated for these conditions^77,85,179. As a range of autonomic disturbances were reported as risk factors for PDD, determining the potential impact of treatments for these changes can provide effective prevention strategies for LBD⁷⁵. Psychiatric onset is included as one of the potential clinical profiles in the research diagnostic criteria for prodromal DLB¹⁸⁹. It is characterized by affective symptoms followed by psychosis prior to DLB onset¹⁹⁰. Accordingly, we observed a range of psychiatric symptoms, including depression and hallucination, associated with higher PDD and DLB risk. These conditions occurring during the prodromal PD phase¹⁹¹ can also occur after PD onset as the non-motor symptoms of PD and predict PDD. Studies in large cohorts showing herpes simplex virus as a risk factor in DLB¹⁵⁸ with varicella zoster vaccination as a protective factor for dementia¹⁹²; terazosin, doxazosin, and alfuzosin for benign prostatic hyperplasia¹⁶⁰, NSAIDs, glucocorticoids¹⁷⁸, anti-hypertensives, anti-diabetics, and cholesterol-lowering agents¹⁷⁹ as protective factors can be particularly helpful for disease modification trials in LBD.

Antiparkinsonian medications were also found to be associated with PDD risk. However, the higher total medication dose⁹⁸, longer exposure⁹⁹, and treatment complications³³ were associated with higher risk, instead of individual medications⁹⁸. Considering that more advanced disease is also associated with higher risk, it is likely that disease severity leading to higher medication dose and treatment complications leads to this medication and dementia connection. We noted that various motor symptoms of PD were reported to predict dementia; tremor-dominant subtype was associated with a lower likelihood for dementia development. However, clinical subtypes are unstable and can change over time¹⁹³. Age at onset for PD was amongst the most reported risk factors, although this association may be due to the general effect of age rather than age of disease onset²⁶. Aging related changes coupled with PD pathogenesis increase dementia risk further¹⁹⁴. Risk factors for dementia may differ for people with younger onset compared to older onset.

Subjective cognitive decline and MCI are prodromal phases for dementia and can provide opportunities for intervention to prevent LBD onset. While executive, attention and visuospatial deficits are more typically associated with LBD, people with prodromal LBD can have a wide range of cognitive domains affected¹⁹⁵. MCI can be difficult to identify in the absence of objective cognitive testing¹⁹⁶. Comprehensive neuropsychological assessments can help identify people at risk for LBD. Global cognitive screening tools can also be useful in settings where comprehensive testing in not possible. In addition to cognitive testing, assessing color vision⁷¹, stereopsis¹³, and noise pareidolia test¹⁶⁸ may also provide useful predictive information.

Different imaging modalities including MRI, EEG, PET, SPECT, MIBG, MEG, transcranial sonography and EMG were assessed for LBD prediction. The most consistently reported finding for biomarkers for LBD prediction was lower CSF amyloid β42 levels in people with RBD and PD, underscoring the prevalence and impact of AD co-pathology in LBD¹⁹⁷. However, CSF amyloid β42 cut-off values to detect dementia for AD and PD may differ¹⁹⁸. Recent reports on biomarkers including seed amplification and real-time quaking-induced conversion (RT-QuIC) assays for alpha-synuclein also have promising findings for LBD¹⁹⁹. Clinical profile in the prodromal phase of LBD is heterogeneous and overlaps with prodromal phases of AD and other dementias²⁰⁰. Accordingly, utilizing biomarkers can increase diagnostic accuracy^1,201 and better define the risk and predictive factors for LBD in future longitudinal studies. While we reported factors individually, several studies built statistical models combining different factors (see Supplementary Tables 1–4). Access to imaging and biomarker studies differ across settings, however, combining available biomarkers with clinical findings appears to be a helpful approach for prediction.

Our review showed that the current literature primarily consists of studies focused on predictors and PDD with many populations remaining underrepresented. The majority of the studies included in our review were conducted at single sites and included cohorts from North America or Europe. In their meta-analysis for dementia risk in PD, Gibson and colleagues noted that PDD incidence rate was lowest in Asia and highest in North America with significant heterogeneity and without any studies from Africa or South America²⁰². The meta-analysis by Hogan and colleagues in 2016 noted that DLB accounts for about 5% of all dementia cases with incidence rates ranging from 0.5 to 1.6 per 1000 person-years²⁰³. As the research diagnostic criteria for the MCI with Lewy bodies was published in 2020 and are relatively new^189,204, it is likely that the studies focusing on the prodromal stage of DLB will increase in the upcoming years to better understand the incidence of prodromal DLB in MCI cohorts. For studies included in our review, there were no comparisons for risk across different countries or people from different ethnic and racial groups. This further underscores the need for more diversity in research cohorts and multi-site collaborations for better representation and understanding of the underlying reasons behind different incidence rates. We noted that the PDD and DLB clinical diagnostic criteria were the most used diagnostic approaches. Additionally, 26.3% (n = 44) primarily relied on medical record diagnostic codes and 10.2% (n = 17) only used global cognitive measures (e.g., MMSE, MoCA) with different cut-offs across the studies for diagnosis. Medical record codes have limited diagnostic accuracy for dementia, AD and vascular dementia^205,206. The differences across studies for the diagnostic approach can impact the outcomes. In addition, the wide range for sample size, follow-up duration, age at baseline and sex ratio for cohorts limit the generalizability of findings.

In conclusion, research so far suggest a role for genetics, aging, sex, education, infections, health conditions and medications in the increased risk of developing LBD, although replication of many findings is still needed to determine their applicability. Predictive models combining a range of factors, and biomarkers when available, can help identify people at risk for LBD, ultimately benefitting disease modification efforts. However, identifying causal factors and the interactions between risk factors is required to better understand the pathophysiology and guide clinical trial strategies. With promising advances in the diagnosis and treatment approaches for neurodegenerative diseases, more research should focus on risk factors for LBD, particularly DLB. Longitudinal study design, consistency for diagnostic criteria and assessments, and collaborative multi-site cohorts from diverse populations can provide better insight to support all at risk for and living with LBD.

Methods

This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines²⁰⁷. The review protocol was not pre-registered.

Search strategy

Studies were identified by searches of PubMed, Embase, and Web of Science from inception to July 19th, 2024. The search was limited to original peer-reviewed research articles written in English and including humans rather than animals. Case reports, reviews, meta-analyses, conference abstracts, editorial, opinion papers, book chapters, and preprints were excluded. Search terms were based on the concepts of outcome (dementia, cognitive decline, cognitive impairment) AND disease (Lewy, Parkinson, synuclein) AND risk factors (risk, prediction, prodromal).

Search terms for PubMed were: (“Lewy Bodies”[Mesh] OR “Lewy Body Disease”[Mesh] OR “Parkinson Disease”[Mesh] OR “Synucleinopathies”[Mesh] OR “alpha-Synuclein”[Mesh]) AND (“Dementia”[Mesh] OR “Cognition Disorders”[Mesh]) AND (“Risk”[Mesh] OR “Prodromal Symptoms”[Mesh]). Search terms for Embase were: (‘lewy bodies’:ti,ab,kw OR ‘lewy body disease’:ti,ab,kw OR ‘parkinson’:ti,ab,kw OR ‘synucleins’:ti,ab,kw OR ‘alpha-synuclein’:ti,ab,kw) AND (‘dementia’:ti,ab,kw OR ‘cognitive decline’:ti,ab,kw OR ‘cognitive impairment’:ti,ab,kw OR ‘mci’:ti,ab,kw) AND (‘risk’:ti,ab,kw OR ‘prodromal symptoms’:ti,ab,kw OR ‘prediction’:ti,ab,kw) AND [english]/lim. Search terms for Web of Science were: ((Lewy OR Parkinson OR synuclein) AND (dementia OR cognitive decline OR cognitive impairment OR MCI) AND (risk OR prodromal OR prediction)).

Study selection

To include all published original articles with LBD as an outcome in a longitudinal study, the following inclusion and exclusion criteria were applied.

Inclusion criteria:

1. The study had a longitudinal design

2. For clinically defined cohorts, the study included participants without a dementia diagnosis at baseline develop (a) DLB, (b) PDD, or (c) LBD without specification of DLB or PDD during follow-up.

3. For pathologically defined cohorts, the study included participants without a dementia diagnosis at baseline develop dementia during the study period with underlying Lewy body pathology.

4. A risk factor, predictor or prodromal feature is assessed by statistical analysis.

Exclusion criteria:

1. Study focusing on participants without dementia by the end of the study follow-up period (cognitive impairment without specification of dementia, MCI without progression to dementia)

2. Study focusing on participants with dementia at baseline

3. DLB, PDD or LBD not included as an outcome for risk/prediction analysis

4. DLB, PDD, or LBD not assessed on its own but combined with another type of disease or dementia as an outcome

5. Study without a non-exposed group to include in risk/prediction analysis

The search resulted in 681 studies from PubMed, 2708 studies from Embase and 2365 studies from Web of Science (Fig. 3). After removing the duplicate records, a combination of two reviewers (AR, CJ, EB, SA) independently screened the 4368 studies by title and abstract. The selected 364 studies from this step were sought for full text retrieval. One article was obtained after contact with the corresponding author, and one article could not be accessed. Thus, 363 articles were further screened by full text and a combination of two reviewers (AR, CJ, FD, EB, SA) based on the stated inclusion/exclusion criteria. For any disagreement between the two reviewers, a third reviewer screened the studies to resolve the conflict. Overall, 167 studies were included for data extraction and synthesis.

Fig. 3.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram for studies in the systematic review.

Data extraction and synthesis

For included studies, data including year of publication, country of study, sample size for risk/predictor analysis, length of follow-up, percent females, mean age at baseline, diagnostic criteria for LBD, and findings on the association between LBD and risk factors were extracted. For consistency, median (range) values were transformed to mean (standard deviation). Each data extraction table was completed by one of four reviewers (AR, FD, EB, SA). Risk factors and predictors were grouped as: non-modifiable (demographic, genetic), modifiable (health condition, medication, environment), and clinical (clinical signs and symptoms, imaging, fluid biomarker) factors. The findings were reported using a narrative synthesis. Since sex and gender terms were used interchangeably in the publications, we grouped sex and gender as sex, with female/male categories for brevity.

Quality assessment

The quality and risk of bias for the included studies was evaluated with the Newcastle-Ottawa Quality Assessment Scale including a total of 8 items in the domains of selection, comparability, and outcome²⁰⁸. Selection includes items assessing whether (1) the exposed cohort (DLB, PDD, or LBD for our review) is truly or somewhat representative of the average affected people (DLB, PDD, or LBD for our review) in the community, (2) the non-exposed cohort (controls without DLB, PDD, or LBD for our review) was selected from the same community as the exposed cohort, (3) exposure (DLB, PDD, or LBD diagnosis) was based on secure record or structured interview, and (4) outcome of interest (DLB, PDD, or LBD diagnosis) was not present at start of study. Comparability questions whether the study controls for the most important factor (age for our review) and any additional factors. Outcome includes items on whether (1) the outcome was assessed by independent blind assessment/record linkage, (2) the follow-up was long enough for outcome to occur (one year for our review), and (3) all subjects were accounted for (complete follow-up), the subjects lost to follow-up were not likely to introduce bias or description was provided of those lost. Total scores on the scale range from 0-2 for poor quality, 3–5 for fair quality and 6-9 for good/high quality. Studies with only good/high quality were included.

Author contributions

A.R.: Design of the work; acquisition, analysis of data; have drafted the work F.R.D.: Analysis of data C.J.: Design of the work; acquisition, analysis of data S.A.: Analysis of data K.A.W.: Design of the work; interpretation of data; have substantively revised the work E.B.: Conception and design of the work; acquisition, analysis and interpretation of data; have drafted and substantively revised the work. All authors reviewed the manuscript.

Data availability

No datasets were generated or analysed during the current study.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s44400-025-00022-2.