Diagnostic Accuracy of Biomarkers in CNS-originating Extracellular Vesicles for Parkinsonian Disorders: A meta-analysis

doi:10.21203/rs.3.rs-3161624/v2

Parkinsonian disorders, including Parkinson's disease (PD), multiple system atrophy (MSA), dementia with Lewy bodies (DLB), progressive supranuclear palsy (PSP) and corticobasal syndrome (CBS), exhibit overlapping early-stage symptoms, complicating definitive diagnosis despite heterogeneous cellular and regional pathophysiology. Additionally, the progression and eventual conversion of prodromal conditions such as REM behavior disorder (RBD) to PD, MSA or DLB remains difficult to predict. Extracellular vesicles (EVs) are small, membrane-enclosed structures released by cells, playing a vital role in communicating cell-state-specific messages. Due to their ability to cross the blood-brain-barrier into the peripheral circulation, the measurement of biomarkers in blood-isolated putative CNS-originating EVs has become a popular diagnostic approach. However, replication and independent validation remain challenges in this field. We conducted a PRISMA-guided systematic review and meta-analysis, covering 15 studies with a total of 1,455 patients with PD, 206 MSA, 21 DLB, 172 PSP, 152 CBS, 189 RBD and 1,045 healthy controls (HCs), employing either hierarchical bivariate models or univariate models based on study size. Diagnostic accuracy was moderate for differentiating patients with PD from HCs, but revealed high heterogeneity and significant publication bias, suggesting an inflation of the perceived diagnostic effectiveness. The bias observed indicates that studies with non-significant or lower effect sizes were less likely to be published. Although results for differentiating patients with PD from MSA or PSP and CBS appeared promising, their validity is limited due to the small number of involved studies coming from the same research group. Despite initial reports, our analyses suggest that using CNS-originating EV biomarkers may not reliably differentiate patients with MSA from HCs or patients with RBD from HCs, due to their lesser accuracy and substantial variability among the studies, further complicated by potential publication bias. Our findings underscore the moderate yet unreliable diagnostic accuracy of putative CNS-originating EV biomarkers in differentiating Parkinsonian disorders, highlighting the presence of substantial heterogeneity and significant publication bias. These observations reinforce the need for larger, more standardized, and unbiased studies to validate and enhance the utility of EV biomarkers in the differential diagnosis of these conditions.

Neurology

Cellular & Molecular Neuroscience

Epidemiology

Laboratory Diagnostics

Neurobiology of Disease

Translational Medicine

L1CAM

exosome

movement disorders

diagnosis

α-synuclein

tau

biomarker

differential diagnosis

synucleinopathy

tauopathy

Parkinsonian disorders comprise a group of neurodegenerative conditions marked by motor symptoms such as slow movement (bradykinesia), stiffness (rigidity), and shaking (tremor). Parkinson's disease (PD) is the most common among these conditions [1]. Other less frequent but clinically important parkinsonian disorders include multiple system atrophy (MSA), dementia with Lewy bodies (DLB), progressive supranuclear palsy (PSP), and corticobasal syndrome (CBS) [2]. While these disorders differ in the type of protein, cell type and brain region afflicted, they are often misdiagnosed by neurologists due to symptom overlap, especially in early stages [3-5]. Moreover, currently, there is no concrete method to precisely ascertain the timing, progression, and specific outcomes of prodromal conditions like REM behavior disorder (RBD) pure autonomic failure (PAF) [6, 7].

Misdiagnosis not only negatively impacts patient prognosis, potentially leading to inappropriate treatments and worsening health outcomes, but also exacerbates emotional distress, exacerbating feelings of uncertainty and anxiety about their health conditions and impacts appropriate patient stratification in clinical trials. This lack of reliable diagnostic tools also obstructs our efforts to discover disease-modifying treatments during the prodromal stages, a critical period when neuronal death is believed to predominantly occur [8].

Extracellular vesicles (EVs) are tiny, membrane-enclosed structures released by cells, which play vital roles in facilitating communication between cells and regulating various bodily processes. They contain a diverse array of biomolecules, including proteins, lipids, and nucleic acids, which mirror the condition of the originating cell [9]. Due to their ability to traverse the blood-brain barrier [10, 11], EVs may provide a unique insight into the brain's biochemical processes, enabling the investigation of central nervous system (CNS) functions and the identification of potential biomarkers in neurodegenerative conditions [12].

As potential carriers of cell-state-specific information from the CNS to the peripheral circulation, EVs have emerged as a possible tool for minimally invasive diagnostic and therapeutic strategies in parkinsonian disorders. Many groups have quantified biomarkers in putative CNS-originating EVs for the differential diagnosis of these disorders from one another and/or from healthy controls (HCs) [13]. Despite this, there has been consistent failure in independent validations, replication, and differing outcomes even when the same methodology is employed.

A recent meta-analysis suggested that the concentrations of neuronal and/or oligodendroglial EVs (nEVs and oEVs, respectively) may be higher in patients with PD in comparison to HCs, CBS and PSP [14]. These elevated concentrations could potentially be utilized to assess the precision of a test for differentiating these diseases. However, the meta-analysis did not compare the diagnostic accuracy of tests utilizing biomarkers in putative CNS-originating EVs, which include α-synuclein (α-syn) combined with other biomarkers.

Our goal is to expand upon previous findings by carrying a meta-analysis of diagnostic accuracy using studies attempting to differentiate either prodromal or established parkinsonian disorders from each other or from HCs, using biomarkers in putative CNS-originating EVs.

We performed a systematic review and meta-analysis according to the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols (PRISMA). Our research exclusively utilized anonymized data, with no collection of personal information or involvement of human subjects, thus obviating the need for ethical approval. The study protocol was not registered.

Data sources and search strategy

We performed a thorough search for relevant articles by using specific search terms related to PD and parkinsonian disorders. The search was conducted in two databases (PUBMED and EMBASE) and covered articles published from the inception of the databases until July 8th, 2023. The search terms we used included combinations of “Parkinson's disease OR multiple system atrophy OR Lewy body dementia OR corticobasal syndrome OR progressive supranuclear palsy" AND "Extracellular Vesicle OR exosome” AND “Diagnosis”. We manually examined the reference lists of eligible studies and conducted thorough literature reviews to identify suitable studies for inclusion. Any discrepancies in the selection of articles were resolved through discussions. The comprehensive search strategy can be accessed in Table S1.

Eligibility criteria

The eligible studies included in our analysis focused on assessing biomarkers in putative CNS-originating EVs obtained from cerebrospinal fluid, plasma, serum, urine or saliva in patients with PD along with at least one of the following diseases: MSA, DLB, PSP, CBS, RBD, PAF or HCs. The studies must have included receiver operating characteristic (ROC) analysis and provided sensitivity, specificity, area under curve (AUC) and sample size. We excluded studies that used animals or cell lines, studies that did not include the specified diseases, and studies that did not report the sample size. If sensitivity, specificity, or sample size were not included in the study, we contacted the authors to obtain the missing information. For studies that included longitudinal measurements or treatment interventions, we only considered the baseline assessments.

Risk of bias assessment

The quality and risk of bias of all eligible studies were evaluated using the Quality Assessment for Diagnostic Accuracy Studies (QUADAS-2) criteria [15]. The assessment was carried out by independent researchers (HBT and AB), and any disagreements were resolved through discussion until a consensus was reached. Additional details regarding the quality assessment can be found in Table S2.

Data synthesis and statistics

In this study, we chose the hierarchical bivariate model [16, 17] utilizing a random effect with a restricted maximum likelihood estimation method in analyses where the # of studies >3. This approach allows for a comprehensive assessment of the diagnostic accuracy measures, accounting for both within-study and between-study variability as well as the inherent negative correlation between sensitivities and specificities across studies. In cases where the study # was ≤3, we utilized a univariate model as the parameters in the bivariate model are not recommended when there are only a few studies [18].

In addition, we utilized informative graphical representations, including crosshair plots, which integrate both ROC curves and forest plots means. These visualizations allow us to simultaneously examine the bivariate relationship between sensitivity and false positive rate (FPR or 1-specificity) while assessing the degree of heterogeneity across studies. Notably, wider crosshairs on the plot indicate a larger sample size, reflecting the level of precision and reliability in the estimates. The ROC ellipse plot visually represents the estimated uncertainty of the pair (sensitivity, FPR) in logit ROC space using confidence regions. The ellipses in the plot symbolize the variability of the sensitivity and FPR estimates, providing an indication of their statistical uncertainty.

The summary ROC (SROC) curve utilized both the dotted means obtained from the bivariate model with its corresponding confidence interval as well as the summary line obtained from hierarchical ROC (HSROC) model [19], which describes the relationship between the mean sensitivity and specificity. In this meta-analysis, when significant heterogeneity is present, the summary line provides more informative results compared to the point means of sensitivities and specificities, as it comprehensively takes into account the heterogeneity across the included studies [18]. The accuracy of the test increases as the point summary of sensitivities/specificities and the summary line approach the upper-left corner.

Funnel plots, Begg’s rank correlation [20], Egger’s [21] and Deek’s regression [22] tests as well as the trim-and-fill method [23] were used to evaluate publication bias [24]. In cases where more than 2 ROC models existed, we chose the model with the best AUC for reporting. In three studies [25-27], two models performed similarly, and we include both models. In one study [28] there were two models, but we excluded the one with AUC close to 0.50, indicating no accuracy. For studies including training and validation ROCs, we only considered the validation model.

The systematic and hand search identified 399 articles of which 73 duplicated articles were removed. After title and abstract screening of 326 articles, 63 articles were considered potentially eligible (Fig. 1). After screening of full-text, 48 studies were excluded. Forty-three of those studies did not enrich for putative CNS-originating EVs and are included elsewhere [29]. Four studies enriched for putative CNS-originating EVs [30-33] but did not include information for sensitivity and specificity for the diagnostic test. One article included preliminary data [34]. All authors were contacted to obtain the missing information.

In total, the meta-analysis included 15 studies [11, 25-28, 35-44] with 1,455 patients with PD, 206 with MSA, 21 with DLB, 172 with PSP, 152 with CBS, 189 with RBD and 1,045 HCs (Table 1). Using biomarkers in putative CNS-originating EVs, most studies attempted to differentiate patients with PD from HCs (n=13, 86.7%). Five studies attempted to differentiate patients with PD from MSA [27, 38, 40-42] while two aimed to differentiate patients with PD from PSP and CBS [25, 27]. One study attempted to differentiate patients with PD from frontotemporal dementia (FTD), PSP and CBS [40]. Three studies aimed to differentiate patients with MSA from HCs [38, 41, 42] or patients with RBD from HCs [28, 40, 43].

The studies included in the analysis were generally of high quality as indicated in Table S3. However, there was a lack of clear reporting on the sampling method, which made the assessment of the risk of bias in patient selection unclear. One study excluded participants from their analysis and was deemed to have a high risk of bias [43]. The measurement of biomarkers in nEVs and oEVs was considered to have a low risk of bias in the index test domain, as it is an objective measure unaffected by prior knowledge of the clinical status. While the majority of the articles (66.7%) had a low risk of bias in the Reference Standard domain, four studies using an in-house test and one using western blots were identified as having a high risk of bias [11, 38, 42, 43]. In terms of the Flow and Timing domain, all studies were deemed to have a low risk of bias as the time interval from clinical diagnosis to biomarker measurement could be reliably estimated.

As previously indicated [14, 45], several preanalytical elements can have a substantial effect on the purity, content, dimensions, and amount of EVs. Such elements involve the selection of anticoagulation molecules mixed with plasma, EV isolation methodology, the centrifugation procedure, the transportation characteristics, the frequency of freezing and thawing cycles, the storage parameters, the temperature and the type of tube used for collection [46, 47]. Regrettably, these aspects are not universally standardized across biobanks or methods of clinical lab blood collection. Moreover, the employment of the anti-L1 cell adhesion molecule (L1CAM) antibody clone UJ127 has initiated doubts regarding its possible cross-reactivity with α-syn antibodies [48]. To tackle these concerns, we performed subgroup analyses based on the medium (either plasma or serum) and the type of antibody clone (e.g., L1CAM clone UJ127 or 5G3) for the analyses for patients with PD vs. HCs. We did not perform such analyses for other diseases due to the small number of studies included.

Descriptive statistics of the meta diagnostic analysis including the sensitivity, specificity, FPR, diagnostic odd ratio (DOR), positive likelihood ratio (posLR) and negative likelihood ratio (negLR) for each included analysis are summarized in Table 2.

PD vs. Control

Thirteen studies have attempted to differentiate patients with PD from HCs using biomarkers in nEVs [11, 26-28, 35-37, 39-44] and/or oEVs [41, 42]. The AUC ranged between 0.61-0.89 with the highest AUC obtained in 2021 by Jiang et al. [27], while the sensitivity (Fig. 2A) and specificity (Fig. 2B) ranged between 0.10-0.97 and 0.50-0.92, respectively. The chi-square (χ2) equality test revealed high heterogeneity for sensitivity (χ2 = 313.8, df = 20, p-value < 0.0001) and specificity (χ2 = 301.2, df = 20, p < 0.0001). Both the crosshair and ROC ellipse plots confirmed the heterogeneity present (Fig. S1A-B). Univariate Forest plots of the DOR, posLR and negLR for each individual analysis are shown in Figs. 2C-E.

We conducted a bivariate diagnostic random-effects meta-analysis of the 13 included studies, The estimates for sensitivity and FPR were reported as 0.725 and 0.264, respectively, indicating that the diagnostic test correctly identified the condition approximately 72.5% of the time, while falsely identifying the condition when it's not present in approximately 26.4% of cases. The pooled AUC was reported as 0.79, demonstrating a fair discriminatory ability of the diagnostic test. The partial AUC, focusing on a specific range of FPRs, was slightly lower at 0.687. Combined, the two AUCs suggested that measuring biomarkers in putative CNS-originating EVs was fair in distinguishing patients with PD from HCs.

The heterogeneity (I²) values showed significant variations depending on the approach utilized. Zhou and Dendukuri [49] reported a value of 49.2%, while Holling's sample size unadjusted [50] had values ranging from 88.4% to 94.7%, and the adjusted values ranging between 7.6% to 10.3%. However, all approaches generally indicated substantial heterogeneity across the studies, suggesting that the variability in the results cannot be attributed solely to random chance but rather to differences between the studies themselves, supporting the crosshair and ROC ellipse plots as well as the fact that studies measuring biomarkers in putative CNS-originating EVs generally suffer from failure of independent validation due to methodological and expertise heterogeneities.

The standard deviations for between-study variability in sensitivity (1.052) and FPR (0.754), indicate considerable variability in these parameters across different studies. The correlation between sensitivity and FPR was nearly zero (-0.028), suggesting that studies with a higher sensitivity do not necessarily have a higher FPR, and vice versa.

The summary ROC (SROC) curve for this model is provided in Fig. 2F. Overall, the SROC suggested that there was a great amount of heterogeneity across the studies evidenced by the wide scattering of the individual mean points from each study. The summary line obtained from HSROC model suggested that measurement of biomarkers in putative CNS-originating EVs for distinguishing patients with PD from HCs may not be promising. As the sensitivity increased, the specificity decreased, indicating the presence of the threshold effect, with the line being far away from the upper-left corner. Moreover, while some studies achieved high sensitivity and specificity, the combined mean indicated that this test only achieves a fair distinguishing ability.

Importantly, few studies subdivided patients with PD to early vs. advanced stages [35-37, 43], but only two studies attempted to differentiate patients with early-stage PD from HCs [36, 44] from similar research groups, and unfortunately, one of them [36] had the lowest sensitivity and specificity. This suggests that biomarkers in putative CNS-originating EVs may not be a good way to discriminate early-stage patients with PD from HCs, despite it being the most clinically desired outcome of the test.

All three statistical tests conducted to assess publication bias in our analysis consistently indicated the presence of such bias. Begg’s correlation test revealed a significant positive correlation between lnDOR and its variance (tau = 0.46, p-value = 0.003; Fig. 3A), implying that larger effect sizes were associated with greater variances. Similarly, Egger’s regression test showed a significant positive relationship between the lnDOR and the standard error of the lnDOR (slope = 3.83, SE = 1.66, t = 2.31, p = 0.033; Fig. 3B), suggesting that smaller studies, which tend to have larger standard errors, were reporting larger effect sizes than what would be expected if there was no bias. Finally, Deek’s regression test also indicated potential publication bias, with a significant positive slope (slope = 27.40, SE = 10.58, t = 2.59, p = 0.018; Fig. 3C) showing that studies with smaller effective sample sizes were associated with larger effect sizes.

Further examination of publication bias using Deek’s funnel plot (Fig. 3D) and a bivariate bagplot (Fig. 3E) also suggested the presence of publication bias. Duval and Tweedie's trim and fill method, a non-parametric method of adjusting for publication bias, estimated that there were approximately 4 studies potentially missing from our meta-analysis due to publication bias. These missing studies are hypothesized to be on the left side of the funnel plot (Fig. 3F), indicating smaller studies with lower diagnostic odds ratios, and therefore may explain why they were not published. When these hypothetically missing studies were imputed and included in a random-effects model, the adjusted pooled log diagnostic odds ratio was 1.77 (SE = 0.29, 95% CI: 1.19 to 2.35, z = 6.03, p < 0.0001). This suggests that when adjusting for potential publication bias, the diagnostic effect for patients with PD vs. HCs is much smaller than what is actually reported in the literature.

Collectively, the hierarchical bivariate model revealed moderate diagnostic accuracy of patients with PD from HCs using putative CNS-originating EVs, but with high heterogeneity and unreliability. Publication bias analyses show that smaller studies with non-significant or low effects size results have been less likely to be published. Unsurprisingly, this is to be expected as alluded to previously [14, 45, 51], there has been consistent failure of independent validation across studies using putative CNS-originating EVs, likely due to EVs being very sensitive to various preanalytical factors [45], high complexity of methodologies used to isolate putative CNS-originating EVs as well as user differences in handling. Even though measuring biomarkers in putative CNS-originating EVs for patients with PD vs. HCs has been popular since 2014, only 13 studies currently exist, further indicating that studies with null results might not have been published. When the trim-and-fill method was used to account for the missing studies, the diagnostic effect for patients with PD vs. HCs decreased significantly.

PD vs. Control: analysis by medium and antibody clone

As described above, several preanalytical factors may affect the EV signature obtained from plasma or serum. Recent studies suggested that plasma provides superior accuracy and reliability in comparison to serum for EV biomarker analysis [45, 52], while the antibody clone UJ127 has been reported to cross-react with α-syn proteoforms [48].

In our bivariate diagnostic meta-analysis, we observed distinct differences between studies using plasma and serum. The plasma model yielded an AUC of 0.729 with a sensitivity of 0.689 and FPR of 0.323. In contrast, the serum model yielded an AUC of 0.852, a sensitivity of 0.744, and a FPR of 0.181. Comparison of the hierarchical bivariate SROC (Fig. 4A) obtained from studies using plasma [11, 35-37, 43, 44] or serum [26-28, 39-42] also suggested that the studies using serum had, on average, slightly better accuracy, though there was a decent overlap in the confidence intervals of both models. It should also be noted that out of the seven studies using serum, three [27, 28, 40] and two [41, 42], respectively, originated from the same research group while all studies from plasma originated from unique research groups, suggesting a potential overlap in methodologies which may influence the results. As such, these differences in the number of studies and potential methodological biases do not definitively establish one medium as superior over the other. Further independent studies are needed to draw more conclusive comparisons.

Further comparisons of studies using the anti-L1CAM antibody clone UJ127 vs. 5G3 showed that studies using the 5G3 clone obtained a slightly higher accuracy (Fig. 4B). We did not identify any publication bias differences between the two mediums or antibody clones assessed through Begg’s correlation, Egger’s and Deek’s regression tests and the trim-and-fill method.

PD vs. MSA

Only five studies attempted to differentiate patients with PD from MSA [27, 38, 40-42]. The AUC ranged between 0.709-0.980 while the sensitivity (Fig. 5A) and specificity (Fig. 5B) ranged between 0.53-0.96 and 0.64-0.92, respectively. Similarly, to the above the chi-square (χ2) equality test revealed high heterogeneity for sensitivity (χ2 = 131.63, df = 7, p-value < 0.0001) specificity (χ2 = 57.84, df = 7, p < 0.0001) and both the crosshair and ROC ellipse plots confirmed the heterogeneity present (Fig. S2A-B). Univariate Forest plots of the DOR, posLR and negLR for each individual analysis are shown in Figs. 5C-E.

A bivariate diagnostic random-effects meta-analysis of the 5 included studies comparing patients with PD vs. MSA showed estimates for sensitivity and FPR as 0.849 and 0.167, respectively. The pooled AUC was 0.90, demonstrating relatively good discriminatory ability of the diagnostic test in distinguishing patients with PD from MSA. The partial AUC, focusing on a specific range of FPRs, was found to be 0.862, indicating moderately good discriminatory ability within this range.

The heterogeneity (I²) values exhibited variations based on the approach employed, similar to the previous findings mentioned. The Zhou and Dendukuri approach estimated the heterogeneity at 49.7%. The Holling sample size unadjusted approaches reported higher levels of heterogeneity ranging from 90.9% to 92.2%, while adjusted approaches indicated lower levels of heterogeneity ranging from 8.3% to 12%. These findings suggest substantial heterogeneity across the studies, indicating that the variability in the results may not be due solely to random chance but rather to differences between the studies themselves.

The standard deviations for between-study variability in sensitivity (1.039) and FPR (0.596) indicated notable variability in these parameters across different studies. The correlation between sensitivity and FPR was 0.212 (95% CI: -0.579 to 0.797). This suggested a weak positive relationship between the two measures. However, it's important to note the wide confidence interval and the presence of both positive and negative values, indicating some variability and uncertainty in the correlation estimate.

The SROC curve for this model is provided in Fig. 5F. Similarly, the SROC analysis indicated a significant degree of heterogeneity across the studies. The summary line derived from the HSROC curve analysis was found to be distant from the upper-left corner, suggesting that measurement of biomarkers in putative CNS-originating EVs for distinguishing patients with PD from MSA may not be promising. Moreover, while some studies achieved good sensitivity and specificity, the combined mean for sensitivity and specificity (shown as the circle) indicated that this test.

Publication bias assessment using Begg’s correlation (tau = 0071, p-value = 0.90, Fig. S3A) and Egger’s regression test (slope = -11.91, SE = 10.30, t = -1.16, p = 0.29; Fig. S3B) revealed no publication bias. However, Deek’s regression indicated that there may be some publication bias (slope = -55.9, SE = 8.97, t = -6.22, p = < 0.00118, Fig. S3C). Further examination using Deek’s funnel plot (Fig. S3D), bagplots (Fig. S3E) and the trim-and-fill method (Fig. S3F) suggested no publication bias. Though no publication bias likely exists, the results indicate substantial variability among studies, with a high I² statistic of 87.96% suggesting that a large proportion of the total variability in effect sizes can be attributed to between-study differences rather than chance. This degree of heterogeneity was statistically significant (Q(df = 7) = 69.0077, p < .0001), underscoring the diversity and possible unreliability among the studies included.

PD vs. PSP and CBS

As only two studies attempted to differentiate patients with PD from PSP and CBS [25, 27] and one from FTD, PSP and CBS [40], we use a univariate approach for this analysis. The forest plots sensitivity, specificity, DOR, posLR and negLR are shown in Figs. 6A-E. The model provided an AUC of 0.961 (95% CI: 0.920 – 1.00), indicating high discriminatory ability. The correlation estimates between sensitivity and FPR was -0.185 (95% CI: -0.973-0.944). The wide confidence interval and the presence of both positive and negative values indicated low precision, high variability and uncertainty in the correlation estimate. The coefficient θ of 0.041 (95% CI: -0.0058 – 0.087; plotted as SROC in Fig. 6F) provided support for the utility of this model. The smaller the coefficient θ, the larger the area under the ROC curve, resulting in larger accuracy of the model. Lastly, crosshair (Fig. S4A) and ROC ellipse plots (Fig. S4B) suggested low heterogeneity.

With low heterogeneity (chi-square quality test under heterogeneity: χ2 = 4.11, df = 2, p-value = 0.13), high accuracy and larger overall SMD of biomarkers in patients with PD vs. PSP and CBS [14], measuring biomarkers in putative CNS-originating EVs to differentiate patients with PD from PSP and CBS may be promising. However, as the results came only from three studies, two of which are from the same research group [27, 40], the interpretation and generalizability are limited. A significant challenge in the field arises from the lack of independent validation across studies, and to combat such issue, it is essential to obtain similar results across different laboratories and cohorts.

Assessment of publication bias using Begg’s correlation (tau = 0.67, p-value = 0.33, Fig. S5A), Egger’s regression (slope = -3.98, SE = 20.81, t = -0.19, p-value = 0.88, Fig. S5B), Deek’s regression (slope = -80.7, SE = 21.6, t = -3.72, p-value = 0.16, Fig. S5C) tests, Deek’s funnel plot (Fig. S5D), bagplots (Fig. S5E) and funnel plots using the trim-and-fill method (Fig. S5F) suggested no publication bias.

MSA vs. Control

Three studies attempted to differentiate patients with MSA from HCs [38, 41, 42] and are analyzed here using a univariate approach. The forest plots for sensitivity, specificity, DOR, posLR and negLR are shown in Figs. 7A-E. The coefficient θ of 0.17 (95% CI: -0.55 – 0.89; plotted as SROC in Fig. 7F) indicated that this model is not promising for diagnosing patients with MSA from HCs despite what is reported in the literature [41, 42]. The large coefficient θ suggested smaller AUC and lesser accuracy of this model. Close inspection of the SROC (Fig. 7F) also suggested large variability and heterogeneity, in support of crosshair (Fig. S6A) and ROC ellipse (Fig. S6B) plots.

Assessment of publication bias using Begg’s correlation (tau = 1.0, p-value = 0.083, Fig. S7A), Egger’s regression (slope = 22.9, SE = 3.64, t = 6.28, p-value = 0.024, Fig. S7B), Deek’s regression (slope = 7.74, SE = 210.5, t = 0.037, p-value = 0.97, Fig. S7C) tests, Deek’s funnel plot (Fig. S7D), bagplots (Fig. S7E) and funnel plots using the trim-and-fill method (Fig. S7F) revealed that only Egger’s regression test suspected publication bias.

Synucleinopathy vs. RBD

RBD is recognized as a prodromal disorder that is highly likely to progress and develop into one of the three synucleinopathies [53]. None of the studies that included an RBD cohort [28, 40, 43] in the present meta-analysis provided ROC discriminatory models for the disease against PD or DLB except for MSA [40], precluding our ability to conduct a meta-analysis.

RBD vs. Control

Three studies evaluated biomarkers in nEVs in attempt to differentiate patients with RBD vs. HCs [28, 40, 43] and are analyzed using a univariate approach. The forest plots for sensitivity, specificity, DOR, posLR and negLR are shown in Figs. 8A-E. The large coefficient θ of 0.14 (95% CI: -0.17 – 0.45; plotted as SROC in Fig. 8F) indicated that this model may not be promising in distinguishing patients with RBD from HCs as it suggested smaller AUC and lesser accuracy. Close inspection of the SROC (Fig. 8F) also suggested large variability and heterogeneity, in support of crosshair (Fig. S8A) and ROC ellipse (Fig. S8B) plots. Since the number of studies was small, with one study not reporting any false positives [28], we did not assess publication bias.

The lack of precise and accurate biomarkers for parkinsonian disorders, including PD, MSA, DLB, PSP, and CBS, often leads to misdiagnoses, hampering patients' ability to receive appropriate and timely care. The inability to predict prodromal disease conversion from RBD or PAF to a synucleinopathy further compounds this problem. These challenges are not only distressing for the patients who are left uncertain about their health status and future, but also for the physicians who strive to provide optimal care. Measurement of biomarkers in putative CNS-originating EVs isolated from the blood has been popular due to their hypothesized ability to contain cell-state-specific biomarkers. The current meta-analysis encompassed 15 studies [11, 25-28, 35-44] with 1,455 patients with PD, 206 with MSA, 21 with DLB, 172 with PSP, 152 with CBS, 189 with RBD and 1,045 HCs (Table 1) and aimed to evaluate the diagnostic accuracy of biomarkers in putative CNS-originating EVs for Parkinsonian disorders (Fig. 9).

Studies attempting to differentiate patients with PD from HCs exhibited considerable variability in sensitivity (Fig. 2A) and specificities, (Fig. 2B), indicating potential methodological inconsistencies among them. The analysis showed that while biomarkers in putative CNS-originating EVs achieved a fair ability in distinguishing patients with PD from HCs (Fig. 2F), the results were plagued by high heterogeneity and potential publication bias (Figs 3A-F), casting doubt on the reliability of these findings. Furthermore, our examination suggested that smaller studies with lower or non-significant diagnostic odds ratios have been less likely to be published (Fig. 3F), which could be contributing to an overestimation of the diagnostic utility of putative CNS-originating EVs. Comparing the diagnostic accuracy of biomarkers in putative CNS-originating EVs isolated from the plasma vs. serum, suggested that serum may be superior in accuracy (Fig. 4A). However, three and two studies out of 7 using serum were from the same research group while all 6 studies using plasma were from different research groups, suggesting possible bias. Further comparisons by the anti-L1CAM antibody clone UJ127 vs. 5G3 did not reveal substantial differences with large overlap in the confidence intervals, though studies using the 5G3 obtained a slightly higher accuracy (Fig. 4B).

On the other hand, five studies [27, 38, 40-42] attempted to differentiate patients with PD from MSA and provided mixed results. The wide-ranging values for sensitivity (0.53-0.96, Fig. 5A), specificity (0.64-0.92, Fig. 5B), and diagnostic odds ratio (Fig. 5C) underline the significant variability among these studies. Although the collective AUC was 0.90 (Fig. 5F), suggesting a reasonable diagnostic test's discriminatory capacity, the substantial heterogeneity in the results (I² values ranging from 8.3% to 92.2%) raises concerns about the reliability of the findings.

Only three studies [25, 27, 40] attempted to distinguish patients with PD from those with PSP and CBS. The results, while promising with a AUC (0.961, Fig. 6F), are undermined by wide confidence intervals and both positive and negative values in the correlation estimates between sensitivity and FPR (-0.185, 95% CI: -0.973-0.944). This variability indicated uncertainty in the reliability of these findings. The studies exhibited low heterogeneity (χ2 = 4.11, df = 2, p-value = 0.13), which usually strengthens the findings; however, considering two of the three studies originated from the same research group [27, 40], this limited pool restricted the conclusions' generalizability. More diverse research is required to confirm these results and establish the potential of biomarkers in putative CNS-originating EVs in differentiating patients with PD from PSP and CBS.

Three studies attempted to differentiate patients with MSA from HC, but despite prior reports of successful differentiation [41, 42], our analysis suggested that this approach may not be as promising. A high coefficient θ (0.17, 95% CI: -0.55 – 0.89, Fig. 7F), indicating smaller AUC and lesser accuracy, along with large variability and heterogeneity raise concerns about the reliability of this diagnostic approach.

The prodromal disorder RBD is considered to eventually convert into one of the three synucleinopathies: PD, MSA, or DLB. However, none of the studies that included an RBD cohort [28, 40, 43] provided a ROC discriminatory model for the disease against patients with PD or DLB, except for MSA [36], while no study to date examined biomarkers in putative CNS-originating EVs for the prodromal disorder PAF. The attempt to differentiate patients with RBD from HCs in three studies [28, 40, 43] also appears unpromising, as suggested by a large coefficient θ (0.14, 95% CI: -0.17 – 0.45; Fig. 8F) indicating smaller AUC and lesser accuracy, along with significant variability and heterogeneity.

Notably, one critical challenge is that studies measuring biomarkers in putative CNS-originating EVs suffer from a failure of independent validation due to methodological and expertise heterogeneities. There is also a lack of standardization of preanalytical factors in obtaining putative CNS-originating EVs despite them being highly sensitive to these preanalytical factors [45, 51], which further complicates the generalizability of such a test in the clinic.

Overall, as the search for reliable biomarkers in parkinsonian disorders persists, it becomes evident that a more standardized and rigorous approach is imperative in the field. As we move forward, greater emphasis should be placed on improving study design and minimizing bias, enhancing the comparability and reproducibility of findings, and addressing the heterogeneity in the results. Current efforts by ISEV [54] and others [55, 56] aim toward more rigorous reporting and standardization to enhance accuracy and reproducibility of research utilizing EVs.

Our comprehensive meta-analysis underscores the current limitations and challenges associated with the use of putative CNS-originating EVs as diagnostic biomarkers for Parkinsonian disorders. The significant methodological inconsistencies across studies, combined with high levels of heterogeneity and potential publication bias, considerably undermine the reliability of these findings. Furthermore, the occasional signs of diagnostic promise are frequently offset by the presence of considerable variability, publication bias and the lack of independent validation across different research groups. The absence of standardized protocols for preanalytical factors, which are critical in determining the accuracy of EV-based biomarkers, further compounds these issues. All these aspects culminate in a rather sobering picture, suggesting that this approach may not provide the anticipated breakthrough in the diagnosis of Parkinsonian disorders. As we navigate through the complexities of these debilitating diseases, it is becoming increasingly clear that we may need to re-evaluate our strategies, either by adopting more rigorous standardization and reporting [56] as suggested through current efforts by ISEV [54] and others [55] or exploring alternative avenues for effective biomarker discovery. While the journey ahead may be challenging, our continued pursuit of this endeavor remains crucial in transforming the landscape of discovering biomarkers for Parkinsonian disorders diagnosis and management.

Availability of data and material.

Not applicable.

Ethics approval and consent to participate.

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Funding

None

Authors' contributions

HBT performed literature search, conception, writing of manuscript, and approval of final draft –

HBT and AB performed data collection. All authors reviewed the manuscript.

Acknowledgements

The authors are grateful to Dr. Christie Jeon, ScD, for helpful suggestions.

AUC: area under curve

CBS: corticobasal syndrome

CI: confidence interval

CNS: central nervous system

DLB: dementia with Lewy bodies

DOR: diagnostic odds ratio

EVs: extracellular vesicles

FPR: false positive rate

FTD: frontotemporal dementia

HCs: healthy controls

HSROC: hierarchical summary receiver operating characteristic

I²: heterogeneity statistic

L1CAM: L1 cell adhesion molecule

MSA: multiple system atrophy

nEVs: neuronal extracellular vesicles

oEVs: oligodendroglial extracellular vesicles

PAF: pure autonomic failure

PD: Parkinson’s disease

posLR: positive likelihood ratio

PSP: progressive supranuclear palsy

PRISMA: preferred reporting Items for systematic reviews and meta-Analyses

QUADAS-2: quality assessment for diagnostic accuracy studies

RBD: REM behavior disorder

ROC: receiver operating characteristic

SROC: summary receiver operating characteristics

χ2: chi square

[1] Poewe W, Seppi K, Tanner CM, Halliday GM, Brundin P, Volkmann J, Schrag AE, Lang AE (2017) Parkinson disease. Nat Rev Dis Primers 3, 17013. 10.1038/nrdp.2017.13.

[2] Armstrong MJ, Okun MS (2020) Diagnosis and Treatment of Parkinson Disease: A Review. JAMA 323, 548-560. 10.1001/jama.2019.22360.

[3] Rizzo G, Copetti M, Arcuti S, Martino D, Fontana A, Logroscino G (2016) Accuracy of clinical diagnosis of Parkinson disease: A systematic review and meta-analysis. Neurology 86, 566-576. 10.1212/WNL.0000000000002350.

[4] Schrag A, Ben-Shlomo Y, Quinn N (2002) How valid is the clinical diagnosis of Parkinson's disease in the community? J Neurol Neurosurg Psychiatry 73, 529-534. 10.1136/jnnp.73.5.529.

[5] Baumann CR (2012) Epidemiology, diagnosis and differential diagnosis in Parkinson's disease tremor. Parkinsonism Relat Disord 18 Suppl 1, S90-92. 10.1016/S1353-8020(11)70029-3.

[6] Dauvilliers Y, Schenck CH, Postuma RB, Iranzo A, Luppi PH, Plazzi G, Montplaisir J, Boeve B (2018) REM sleep behaviour disorder. Nat Rev Dis Primers 4, 19. 10.1038/s41572-018-0016-5.

[7] Coon EA, Singer W, Low PA (2019) Pure Autonomic Failure. Mayo Clin Proc 94, 2087-2098. 10.1016/j.mayocp.2019.03.009.

[8] Cheng HC, Ulane CM, Burke RE (2010) Clinical progression in Parkinson disease and the neurobiology of axons. Ann Neurol 67, 715-725. 10.1002/ana.21995.

[9] Dixson AC, Dawson TR, Di Vizio D, Weaver AM (2023) Context-specific regulation of extracellular vesicle biogenesis and cargo selection. Nat Rev Mol Cell Biol. 10.1038/s41580-023-00576-0.

[10] Shi M, Kovac A, Korff A, Cook TJ, Ginghina C, Bullock KM, Yang L, Stewart T, Zheng D, Aro P, Atik A, Kerr KF, Zabetian CP, Peskind ER, Hu SC, Quinn JF, Galasko DR, Montine TJ, Banks WA, Zhang J (2016) CNS tau efflux via exosomes is likely increased in Parkinson's disease but not in Alzheimer's disease. Alzheimers Dement 12, 1125-1131. 10.1016/j.jalz.2016.04.003.

[11] Shi M, Liu C, Cook TJ, Bullock KM, Zhao Y, Ginghina C, Li Y, Aro P, Dator R, He C, Hipp MJ, Zabetian CP, Peskind ER, Hu SC, Quinn JF, Galasko DR, Banks WA, Zhang J (2014) Plasma exosomal α-synuclein is likely CNS-derived and increased in Parkinson's disease. Acta Neuropathol 128, 639-650. 10.1007/s00401-014-1314-y.

[12] Raghav A, Singh M, Jeong GB, Giri R, Agarwal S, Kala S, Gautam KA (2022) Extracellular vesicles in neurodegenerative diseases: A systematic review. Frontiers in Molecular Neuroscience 15. 10.3389/fnmol.2022.1061076.

[13] Dutta S, Hornung S, Taha HB, Bitan G (2023) Biomarkers for parkinsonian disorders in CNS-originating EVs: promise and challenges. Acta Neuropathol, 1-26. 10.1007/s00401-023-02557-1.

[14] Taha HB, Ati SA (2023) Evaluation of a-synuclein in CNS-originating extracellular vesicles for Parkinsonian disorders: A systematic review and meta-analysis. CNS Neurosci Ther, 1-15. https://doi.org/10.1111/cns.14341.

[15] Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, Leeflang MM, Sterne JA, Bossuyt PM, Group Q- (2011) QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med 155, 529-536. 10.7326/0003-4819-155-8-201110180-00009.

[16] Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM, Zwinderman AH (2005) Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol 58, 982-990. 10.1016/j.jclinepi.2005.02.022.

[17] Lee J, Kim KW, Choi SH, Huh J, Park SH (2015) Systematic Review and Meta-Analysis of Studies Evaluating Diagnostic Test Accuracy: A Practical Review for Clinical Researchers-Part II. Statistical Methods of Meta-Analysis. Korean J Radiol 16, 1188-1196. 10.3348/kjr.2015.16.6.1188.

[18] Trikalinos TA, Balion CM, Coleman CI, Griffith L, Santaguida PL, Vandermeer B, Fu R (2012) Chapter 8: meta-analysis of test performance when there is a "gold standard". J Gen Intern Med 27 Suppl 1, S56-66. 10.1007/s11606-012-2029-1.

[19] Rutter CM, Gatsonis CA (2001) A hierarchical regression approach to meta-analysis of diagnostic test accuracy evaluations. Stat Med 20, 2865-2884. 10.1002/sim.942.

[20] Begg CB, Mazumdar M (1994) Operating characteristics of a rank correlation test for publication bias. Biometrics 50, 1088-1101.

[21] Egger M, Davey Smith G, Schneider M, Minder C (1997) Bias in meta-analysis detected by a simple, graphical test. BMJ 315, 629-634. 10.1136/bmj.315.7109.629.

[22] Deeks JJ, Macaskill P, Irwig L (2005) The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol 58, 882-893. 10.1016/j.jclinepi.2005.01.016.

[23] Shi L, Lin L (2019) The trim-and-fill method for publication bias: practical guidelines and recommendations based on a large database of meta-analyses. Medicine (Baltimore) 98, e15987. 10.1097/MD.0000000000015987.

[24] Lin L, Chu H (2018) Quantifying publication bias in meta-analysis. Biometrics 74, 785-794. 10.1111/biom.12817.

[25] Meloni M, Agliardi C, Guerini FR, Zanzottera M, Bolognesi E, Picciolini S, Marano M, Magliozzi A, Di Fonzo A, Arighi A, Fenoglio C, Franco G, Arienti F, Saibene FL, Navarro J, Clerici M (2023) Oligomeric α-synuclein and tau aggregates in NDEVs differentiate Parkinson's disease from atypical parkinsonisms. Neurobiol Dis 176, 105947. 10.1016/j.nbd.2022.105947.

[26] Agliardi C, Meloni M, Guerini FR, Zanzottera M, Bolognesi E, Baglio F, Clerici M (2021) Oligomeric α-Syn and SNARE complex proteins in peripheral extracellular vesicles of neural origin are biomarkers for Parkinson's disease. Neurobiol Dis 148, 105185. 10.1016/j.nbd.2020.105185.

[27] Jiang C, Hopfner F, Berg D, Hu MT, Pilotto A, Borroni B, Davis JJ, Tofaris GK (2021) Validation of α-Synuclein in L1CAM-Immunocaptured Exosomes as a Biomarker for the Stratification of Parkinsonian Syndromes. Mov Disord 36, 2663-2669. 10.1002/mds.28591.

[28] Sharafeldin M, Yan S, Jiang C, Tofaris GK, Davis JJ (2023) Alternating Magnetic Field-Promoted Nanoparticle Mixing: The On-Chip Immunocapture of Serum Neuronal Exosomes for Parkinson's Disease Diagnostics. Anal Chem 95, 7906-7913. 10.1021/acs.analchem.3c00357.

[29] Taha HB, Bogoniewski A (2023) Extracellular vesicles from bodily fluids for the accurate diagnosis of Parkinson’s disease and related disorders: A systematic review and meta-analysis. Research Square. https://doi.org/10.21203/rs.3.rs-3253860/v1.

[30] Ohmichi T, Mitsuhashi M, Tatebe H, Kasai T, Ali El-Agnaf OM, Tokuda T (2019) Quantification of brain-derived extracellular vesicles in plasma as a biomarker to diagnose Parkinson's and related diseases. Parkinsonism Relat Disord 61, 82-87. 10.1016/j.parkreldis.2018.11.021.

[31] Blommer J, Pitcher T, Mustapic M, Eren E, Yao PJ, Vreones MP, Pucha KA, Dalrymple-Alford J, Shoorangiz R, Meissner WG, Anderson T, Kapogiannis D (2023) Extracellular vesicle biomarkers for cognitive impairment in Parkinson's disease. Brain 146, 195-208. 10.1093/brain/awac258.

[32] Kluge A, Bunk J, Schaeffer E, Drobny A, Xiang W, Knacke H, Bub S, Lückstädt W, Arnold P, Lucius R, Berg D, Zunke F (2022) Detection of neuron-derived pathological α-synuclein in blood. Brain 145, 3058-3071. 10.1093/brain/awac115.

[33] Meloni M, Agliardi C, Guerini FR, Saibene FL, Milner AV, Zanzottera M, Bolognesi E, Puligheddu M, Figorilli M, Navarro J, Clerici M (2023) Oligomeric Alpha-Synuclein and STX-1A from Neural-Derived Extracellular Vesicles (NDEVs) as Possible Biomarkers of REM Sleep Behavior Disorder in Parkinson's Disease: A Preliminary Cohort Study. Int J Mol Sci 24. 10.3390/ijms24108839.

[34] Fu Y, Jiang C, Tofaris GK, Davis JJ (2020) Facile Impedimetric Analysis of Neuronal Exosome Markers in Parkinson's Disease Diagnostics. Anal Chem 92, 13647-13651. 10.1021/acs.analchem.0c03092.

[35] Zhao ZH, Chen ZT, Zhou RL, Zhang X, Ye QY, Wang YZ (2019) Increased DJ-1 and α-synuclein in plasma neural-derived exosomes as potential markers for Parkinson's disease. Frontiers in Aging Neuroscience 11. 10.3389/fnagi.2018.00438.

[36] Niu M, Li Y, Li G, Zhou L, Luo N, Yao M, Kang W, Liu J (2020) A longitudinal study on alpha-synuclein in plasma neuronal exosomes as a biomarker for Parkinson's disease development and progression. Movement Disorders 35, S369. 10.1002/mds.28268.

[37] Zou J, Guo Y, Wei L, Yu F, Yu B, Xu A (2020) Long Noncoding RNA POU3F3 and α-Synuclein in Plasma L1CAM Exosomes Combined with β-Glucocerebrosidase Activity: Potential Predictors of Parkinson's Disease. Neurotherapeutics 17, 1104-1119. 10.1007/s13311-020-00842-5.

[38] Yu Z, Shi M, Stewart T, Fernagut PO, Huang Y, Tian C, Dehay B, Atik A, Yang D, De Giorgi F, Ichas F, Canron MH, Ceravolo R, Frosini D, Kim HJ, Feng T, Meissner WG, Zhang J (2020) Reduced oligodendrocyte exosome secretion in multiple system atrophy involves SNARE dysfunction. Brain 143, 1780-1797. 10.1093/brain/awaa110.

[39] Si X, Tian J, Chen Y, Yan Y, Pu J, Zhang B (2019) Central Nervous System-Derived Exosomal Alpha-Synuclein in Serum May Be a Biomarker in Parkinson's Disease. Neuroscience 413, 308-316. 10.1016/j.neuroscience.2019.05.015.

[40] Jiang C, Hopfner F, Katsikoudi A, Hein R, Catli C, Evetts S, Huang Y, Wang H, Ryder JW, Kuhlenbaeumer G, Deuschl G, Padovani A, Berg D, Borroni B, Hu MT, Davis JJ, Tofaris GK (2020) Serum neuronal exosomes predict and differentiate Parkinson's disease from atypical parkinsonism. J Neurol Neurosurg Psychiatry 91, 720-729. 10.1136/jnnp-2019-322588.

[41] Dutta S, Hornung S, Kruayatidee A, Maina KN, Del Rosario I, Paul KC, Wong DY, Duarte Folle A, Markovic D, Palma JA, Serrano GE, Adler CH, Perlman SL, Poon WW, Kang UJ, Alcalay RN, Sklerov M, Gylys KH, Kaufmann H, Fogel BL, Bronstein JM, Ritz B, Bitan G (2021) α-Synuclein in blood exosomes immunoprecipitated using neuronal and oligodendroglial markers distinguishes Parkinson's disease from multiple system atrophy. Acta Neuropathol 142, 495-511. 10.1007/s00401-021-02324-0.

[42] Taha HB, Hornung S, Dutta S, Fenwick L, Lahgui O, Howe K, Elabed N, Del Rosario I, Wong DY, Duarte Folle A, Markovic D, Palma JA, Kang UJ, Alcalay RN, Sklerov M, Kaufmann H, Fogel BL, Bronstein JM, Ritz B, Bitan G (2023) Toward a biomarker panel measured in CNS-originating extracellular vesicles for improved differential diagnosis of Parkinson's disease and multiple system atrophy. Transl Neurodegener 12, 14. 10.1186/s40035-023-00346-0.

[43] Yan YQ, Pu JL, Zheng R, Fang Y, Gu LY, Guo T, Si XL, Zhou C, Chen Y, Liu Y, Guan XJ, Xu XJ, Yan YP, Yin XZ, Zhang MM, Tao ZH, Zhang BR (2022) Different patterns of exosomal α-synuclein between Parkinson's disease and probable rapid eye movement sleep behavior disorder. Eur J Neurol 29, 3590-3599. 10.1111/ene.15537.

[44] Jiao Y, Zhu X, Zhou X, Li Y, Zhou L, Zhao A, Luo N, Niu M, Liu J (2023) Collaborative plasma biomarkers for Parkinson's disease development and progression: a cross-sectional and longitudinal study. Eur J Neurol. 10.1111/ene.15964.

[45] Taha HB (2023) Plasma vs. Serum for Extracellular Vesicle Isolation: A Duel for Reproducibility and Accuracy for CNS-originating Extracellular Vesicles Biomarker Analysis. J Neurosci Res in press.

[46] Lacroix R, Judicone C, Poncelet P, Robert S, Arnaud L, Sampol J, Dignat-George F (2012) Impact of pre-analytical parameters on the measurement of circulating microparticles: towards standardization of protocol. J Thromb Haemost 10, 437-446. 10.1111/j.1538-7836.2011.04610.x.

[47] Gyorgy B, Paloczi K, Kovacs A, Barabas E, Beko G, Varnai K, Pallinger E, Szabo-Taylor K, Szabo TG, Kiss AA, Falus A, Buzas EI (2014) Improved circulating microparticle analysis in acid-citrate dextrose (ACD) anticoagulant tube. Thromb Res 133, 285-292. 10.1016/j.thromres.2013.11.010.

[48] Norman M, Ter-Ovanesyan D, Trieu W, Lazarovits R, Kowal EJK, Lee JH, Chen-Plotkin AS, Regev A, Church GM, Walt DR (2021) L1CAM is not associated with extracellular vesicles in human cerebrospinal fluid or plasma. Nat Methods 18, 631-634. 10.1038/s41592-021-01174-8.

[49] Zhou Y, Dendukuri N (2014) Statistics for quantifying heterogeneity in univariate and bivariate meta-analyses of binary data: the case of meta-analyses of diagnostic accuracy. Stat Med 33, 2701-2717. 10.1002/sim.6115.

[50] Sangnawakij P, Bohning D, Niwitpong SA, Adams S, Stanton M, Holling H (2019) Meta-analysis without study-specific variance information: Heterogeneity case. Stat Methods Med Res 28, 196-210. 10.1177/0962280217718867.

[51] Taha HB (2023) Rethinking the Reliability and Accuracy of Biomarkers in CNS-Originating EVs for Parkinson's Disease and Multiple System Atrophy. Frontiers in Neurology 14. 10.3389/fneur.2023.1192115.

[52] Dhondt B, Pinheiro C, Geeurickx E, Tulkens J, Vergauwen G, Van Der Pol E, Nieuwland R, Decock A, Miinalainen I, Rappu P, Schroth G, Kuersten S, Vandesompele J, Mestdagh P, Lumen N, De Wever O, Hendrix A (2023) Benchmarking blood collection tubes and processing intervals for extracellular vesicle performance metrics. J Extracell Vesicles 12, e12315. 10.1002/jev2.12315.

[53] Postuma RB, Gagnon JF, Bertrand JA, Genier Marchand D, Montplaisir JY (2015) Parkinson risk in idiopathic REM sleep behavior disorder: preparing for neuroprotective trials. Neurology 84, 1104-1113. 10.1212/WNL.0000000000001364.

[54] Thery C, Witwer KW, Aikawa E, Alcaraz MJ, Anderson JD, Andriantsitohaina R, Antoniou A, Arab T, Archer F, Atkin-Smith GK, Ayre DC, Bach JM, Bachurski D, Baharvand H, Balaj L, Baldacchino S, Bauer NN, Baxter AA, Bebawy M, Beckham C, Bedina Zavec A, Benmoussa A, Berardi AC, Bergese P, Bielska E, Blenkiron C, Bobis-Wozowicz S, Boilard E, Boireau W, Bongiovanni A, Borras FE, Bosch S, Boulanger CM, Breakefield X, Breglio AM, Brennan MA, Brigstock DR, Brisson A, Broekman ML, Bromberg JF, Bryl-Gorecka P, Buch S, Buck AH, Burger D, Busatto S, Buschmann D, Bussolati B, Buzas EI, Byrd JB, Camussi G, Carter DR, Caruso S, Chamley LW, Chang YT, Chen C, Chen S, Cheng L, Chin AR, Clayton A, Clerici SP, Cocks A, Cocucci E, Coffey RJ, Cordeiro-da-Silva A, Couch Y, Coumans FA, Coyle B, Crescitelli R, Criado MF, D'Souza-Schorey C, Das S, Datta Chaudhuri A, de Candia P, De Santana EF, De Wever O, Del Portillo HA, Demaret T, Deville S, Devitt A, Dhondt B, Di Vizio D, Dieterich LC, Dolo V, Dominguez Rubio AP, Dominici M, Dourado MR, Driedonks TA, Duarte FV, Duncan HM, Eichenberger RM, Ekstrom K, El Andaloussi S, Elie-Caille C, Erdbrugger U, Falcon-Perez JM, Fatima F, Fish JE, Flores-Bellver M, Forsonits A, Frelet-Barrand A, Fricke F, Fuhrmann G, Gabrielsson S, Gamez-Valero A, Gardiner C, Gartner K, Gaudin R, Gho YS, Giebel B, Gilbert C, Gimona M, Giusti I, Goberdhan DC, Gorgens A, Gorski SM, Greening DW, Gross JC, Gualerzi A, Gupta GN, Gustafson D, Handberg A, Haraszti RA, Harrison P, Hegyesi H, Hendrix A, Hill AF, Hochberg FH, Hoffmann KF, Holder B, Holthofer H, Hosseinkhani B, Hu G, Huang Y, Huber V, Hunt S, Ibrahim AG, Ikezu T, Inal JM, Isin M, Ivanova A, Jackson HK, Jacobsen S, Jay SM, Jayachandran M, Jenster G, Jiang L, Johnson SM, Jones JC, Jong A, Jovanovic-Talisman T, Jung S, Kalluri R, Kano SI, Kaur S, Kawamura Y, Keller ET, Khamari D, Khomyakova E, Khvorova A, Kierulf P, Kim KP, Kislinger T, Klingeborn M, Klinke DJ, 2nd, Kornek M, Kosanovic MM, Kovacs AF, Kramer-Albers EM, Krasemann S, Krause M, Kurochkin IV, Kusuma GD, Kuypers S, Laitinen S, Langevin SM, Languino LR, Lannigan J, Lasser C, Laurent LC, Lavieu G, Lazaro-Ibanez E, Le Lay S, Lee MS, Lee YXF, Lemos DS, Lenassi M, Leszczynska A, Li IT, Liao K, Libregts SF, Ligeti E, Lim R, Lim SK, Line A, Linnemannstons K, Llorente A, Lombard CA, Lorenowicz MJ, Lorincz AM, Lotvall J, Lovett J, Lowry MC, Loyer X, Lu Q, Lukomska B, Lunavat TR, Maas SL, Malhi H, Marcilla A, Mariani J, Mariscal J, Martens-Uzunova ES, Martin-Jaular L, Martinez MC, Martins VR, Mathieu M, Mathivanan S, Maugeri M, McGinnis LK, McVey MJ, Meckes DG, Jr., Meehan KL, Mertens I, Minciacchi VR, Moller A, Moller Jorgensen M, Morales-Kastresana A, Morhayim J, Mullier F, Muraca M, Musante L, Mussack V, Muth DC, Myburgh KH, Najrana T, Nawaz M, Nazarenko I, Nejsum P, Neri C, Neri T, Nieuwland R, Nimrichter L, Nolan JP, Nolte-'t Hoen EN, Noren Hooten N, O'Driscoll L, O'Grady T, O'Loghlen A, Ochiya T, Olivier M, Ortiz A, Ortiz LA, Osteikoetxea X, Ostergaard O, Ostrowski M, Park J, Pegtel DM, Peinado H, Perut F, Pfaffl MW, Phinney DG, Pieters BC, Pink RC, Pisetsky DS, Pogge von Strandmann E, Polakovicova I, Poon IK, Powell BH, Prada I, Pulliam L, Quesenberry P, Radeghieri A, Raffai RL, Raimondo S, Rak J, Ramirez MI, Raposo G, Rayyan MS, Regev-Rudzki N, Ricklefs FL, Robbins PD, Roberts DD, Rodrigues SC, Rohde E, Rome S, Rouschop KM, Rughetti A, Russell AE, Saa P, Sahoo S, Salas-Huenuleo E, Sanchez C, Saugstad JA, Saul MJ, Schiffelers RM, Schneider R, Schoyen TH, Scott A, Shahaj E, Sharma S, Shatnyeva O, Shekari F, Shelke GV, Shetty AK, Shiba K, Siljander PR, Silva AM, Skowronek A, Snyder OL, 2nd, Soares RP, Sodar BW, Soekmadji C, Sotillo J, Stahl PD, Stoorvogel W, Stott SL, Strasser EF, Swift S, Tahara H, Tewari M, Timms K, Tiwari S, Tixeira R, Tkach M, Toh WS, Tomasini R, Torrecilhas AC, Tosar JP, Toxavidis V, Urbanelli L, Vader P, van Balkom BW, van der Grein SG, Van Deun J, van Herwijnen MJ, Van Keuren-Jensen K, van Niel G, van Royen ME, van Wijnen AJ, Vasconcelos MH, Vechetti IJ, Jr., Veit TD, Vella LJ, Velot E, Verweij FJ, Vestad B, Vinas JL, Visnovitz T, Vukman KV, Wahlgren J, Watson DC, Wauben MH, Weaver A, Webber JP, Weber V, Wehman AM, Weiss DJ, Welsh JA, Wendt S, Wheelock AM, Wiener Z, Witte L, Wolfram J, Xagorari A, Xander P, Xu J, Yan X, Yanez-Mo M, Yin H, Yuana Y, Zappulli V, Zarubova J, Zekas V, Zhang JY, Zhao Z, Zheng L, Zheutlin AR, Zickler AM, Zimmermann P, Zivkovic AM, Zocco D, Zuba-Surma EK (2018) Minimal information for studies of extracellular vesicles 2018 (MISEV2018): a position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines. J Extracell Vesicles 7, 1535750. 10.1080/20013078.2018.1535750.

[55] Consortium E-T, Van Deun J, Mestdagh P, Agostinis P, Akay O, Anand S, Anckaert J, Martinez ZA, Baetens T, Beghein E, Bertier L, Berx G, Boere J, Boukouris S, Bremer M, Buschmann D, Byrd JB, Casert C, Cheng L, Cmoch A, Daveloose D, De Smedt E, Demirsoy S, Depoorter V, Dhondt B, Driedonks TA, Dudek A, Elsharawy A, Floris I, Foers AD, Gartner K, Garg AD, Geeurickx E, Gettemans J, Ghazavi F, Giebel B, Kormelink TG, Hancock G, Helsmoortel H, Hill AF, Hyenne V, Kalra H, Kim D, Kowal J, Kraemer S, Leidinger P, Leonelli C, Liang Y, Lippens L, Liu S, Lo Cicero A, Martin S, Mathivanan S, Mathiyalagan P, Matusek T, Milani G, Monguio-Tortajada M, Mus LM, Muth DC, Nemeth A, Nolte-'t Hoen EN, O'Driscoll L, Palmulli R, Pfaffl MW, Primdal-Bengtson B, Romano E, Rousseau Q, Sahoo S, Sampaio N, Samuel M, Scicluna B, Soen B, Steels A, Swinnen JV, Takatalo M, Thaminy S, Thery C, Tulkens J, Van Audenhove I, van der Grein S, Van Goethem A, van Herwijnen MJ, Van Niel G, Van Roy N, Van Vliet AR, Vandamme N, Vanhauwaert S, Vergauwen G, Verweij F, Wallaert A, Wauben M, Witwer KW, Zonneveld MI, De Wever O, Vandesompele J, Hendrix A (2017) EV-TRACK: transparent reporting and centralizing knowledge in extracellular vesicle research. Nat Methods 14, 228-232. 10.1038/nmeth.4185.

[56] Gomes DE, Witwer KW (2022) L1CAM-associated extracellular vesicles: A systematic review of nomenclature, sources, separation, and characterization. J Extracell Biol 1. 10.1002/jex2.35.

TABLE 1. Demographic characteristics of patients with Parkinsonian disorders or healthy controls (HC) included in the meta-analysis

Study (first author, year)

EV isolation method

CNS-EV antibody

EV confirmation method

Quantification method

# PD

# MSA

# DLB

# PSP

# CBS

#

RBD

#

HC

Age (years)

Female (%)

Disease Duration (years)

HY scale

UPDRS III

MMSE

MoCA

Plasma

Shi et al. 2014 (11)

2,000×g for 15 min followed by 12,000×g for 30 min [2] followed by direct IP

L1CAM (clone UJ127)

EM

WB

Luminex (in-house; [3])

267

0

215

PD: 66.3 ± 9.1

HC: 65.7 ± 9.1

PD: 44.6%

HC: 46.0%

9.6 ± 6.6

2.4 ± 0.7

28.4 ± 12.6

28.0 ± 2.6

NA

Zhao et al. 2019 (34)

ExoQuick (Systems Biosciences)

L1CAM (clone 5G3)

TEM

Sandwich ELISA (R&D Systems)

39

Early: 22

Advanced: 17

0

40

PD: 67.5 ± 6.9

Early: 65.2 ± 11.2

Advanced: 67.5 ± 6.8

HC: 66.6 ± 8.8

PD: 41.0%

HC: 57.5%

5.0 ± 3.2

Early: 3.9 ± 2.5

Advanced: 6.4 ± 3.6

NA

48.6 ± 21.0

Early: 37.8 ± 15.2

Advanced: 62.6 ± 19.3

NA

Niu et al. 2020 (35)

2,000×g for 15 min followed by 12,000×g for 30 min [2] followed by direct IP

L1CAM (clone UJ127)

TEM

TRPS

WB

ECLIA (Meso Scale Discovery)

53

Early: 36

Advanced: 17

0

20

21

PD: 65.0 ± 5.3

Early: 64.2 ± 4.9

Advanced: 66.5 ± 5.9

RBD: 63.2 ± 6.0

HC: 64.0 ± 5.4

PD: 53.0%

Early: 50%

Advanced: 59%

RBD: 40.0%

HC: 48.0%

NA

PD: 2.0 ± 0.5

Early: 1.5 ± 05

Advanced: 3.0 ± 0.5

RBD: NA

PD: 22.3 ± 10.3

Early: 18.4 ± 7.5

Advanced: 29.7 ± 11.1

RBD: NA

PD: 27.6 ± 2.6

Early: 28.2 ± 1.6

Advanced: 26.5 ± 3.6

RBD: NA

PD: 23.6 ± 3.6

Early: 24.1 ± 3.3 Advanced: 22.5 ± 4.1

RBD: NA

Zou et al. 2020 (36)

2,000×g for 15 min followed by 12,000×g for 30 min [2] followed by direct IP

L1CAM (clone UJ127)

TEM

NTA

WB

Simoa (Quanterix)

93

Early: 51

Advanced: 42

0

85

PD: 66.9 ± 9.5

Early: 64.7 ± 10.6

Advanced: 67.5 ± 8.15

HC: 66.2 ± 10.3

PD:

43.0%

Early: 43.1%

Advanced: 42.9%

HC: 43.5%

PD: 4.3 ± 2.5

Early: 2.0 ± 3.4

Advanced: 5.3 ± 3.1

PD: 2.8 ± 0.5

Early: 1.5 ± 0.5

Advanced: 3.0 ± 0.5

PD: 28.7 ± 16.0

Early: 22.0 ± 18.5

Advanced: 31.4 ± 15.3

PD: 24.3 ± 2.9

Early: 28.7 ± 2.6

Advanced: 21.9 ± 2.6

NA

Yu et al. 2020 (37)

2,000×g for 15 min followed by 12,000×g for 30 min [2] followed by direct IP

L1CAM (clone UJ127)

CNPase (clone mABcam 44289)

NTA

TEM

WB

Luminex (in-house; [3])

34

32

0

31

PD: 63.6 ± 8.0

MSA: 63.0 ± 6.9

HC: 64.3 ± 7.5

PD: 41.2%

MSA: 40.6%

HC: 51.6%

PD: 4.0 ± 2.2

MSA: 4.0 ± 2.8

NA

PD: 21.4 ± 10.6

MSA (UMSARS): 24.1 ± 10.6

NA

Yan et al. 2022 (42)

ExoQuick (Systems Biosciences)

L1CAM (clone NA)

NTA

TEM

WB

44

Early: 28

Advanced: 16

0

101

48

PD: 64.2 ± 9.6

Early: 63.2 ± 9.8

Advanced: 65.9 ± 9.2

RBD: 61.9 ± 7.9

HC: 61.5 ± 7.1

PD: 56.8%

Early: 53.6%

Advanced: 62.5%

RBD: 56.4%

HC: 54.2%

PD: 3.7 ± 3.8

Early: 2.5 ± 3.0

Advanced: 5.7 ± 4.1

RBD: NA

PD:2.1 ± 1.0

Early: 1.6 ± 0.5

Advanced: 3.1 ± 0.8

RBD: NA

PD: 32.0 ± 19.7

Early: 21.8 ± 10.7

Advanced: 49.7 ± 19.5

RBD: NA

PD: 23.7 ± 6.2

Early: 25.3 ± 4.8

Advanced: 21.1 ± 7.4

RBD: 21.8 ± 5.9

PD: 19.2 ± 7.1

Early: 21.0 ± 5.8

Advanced: 15.9 ± 8.2

RBD: 16.6 ± 6.4

Jiao et al. 2023 (43)

2,000×g for 15 min followed by 12,000×g for 30 min [2] followed by direct IP

L1CAM (clone UJ127)

WB

ECLIA (Meso Scale Discovery

50 (early-stage)

0

50

PD: 64.3 ± 5.6

HC: 64.0 ± 5.8

PD: 56.0%

HC: 50.0%

2.3 ± 1.3

1.6 ± 0.4

21.9 ± 8.6

28.1 ± 1.6

24.3 ± 3.0

Serum

Si et al. 2019 (38)

ExoQuick (Systems Biosciences)

L1CAM (clone UJ127)

EM

WB

Sandwich ELISA (CUSABIO)

38

0

18

PD: 62.4 ± 9.7

HC: 62.7 ± 2.3

TD: 62.7 ± 10.6

NTD: 62.1 ± 10.6

PD: 50%

TD: 45.4%

NTD: 50%

HC: 55.5%

PD: 2.3 ± 1.8

TD: 1.6 ± 1.2

NTD: 3.0 ± 2.5

PD: 1.7 ± 0.6

TD: 1.6 ± 0.60

NTD: 1.7 ± 0.5

PD: 18.6 ± 10.2

TD: 18.3 ± 9.4

NTD: 18.9 ± 10.94

NA

Jiang et al. 2020 (39)

Direct IP

L1CAM (clone UJ127)

NTA

SEM

WB

ECLIA (Meso Scale Discovery)

275

PDD: 45

14

21

35

45

65

144

PD: 68.9 ± 7.1

MSA: 68.1 ± 10.8

DLB: 68.5 ± 4.9

PSP: 68.0 ± 7.5

CBD:61.1 + 7.2

RBD: 64.2 ± 8.3

HC:68.1 ± 10.8

PD: 33.8%

MSA: 40.0%

DLB: 71.4%

PSP: 48.6%

CBS: 40.0%

RBD: 4.6%

HC: 34.7%

PD: 7.5 ± 7.0

MSA: 4.9 ± 2.6

DLB: 3.4 ± 3.0

PSP: 2.8 ± 1.8

CBS: 1.9 ± 1.3

RBD: NA

NA

PD: 32.2 ± NA

MSA: 27.7 ± NA

DLB: 20.9 ± NA

RBD: 5.1 ± NA

PSP: 24.5 ± NA

CBS: 22.5 ± NA

NA

PD: 22.7 ± NA

MSA: 16.9 ± NA

DLB: 16.3 ± NA

PSP: 21.4 ± NA

CBS: 22.3 ± NA

RBD: 25.5 ± NA

Agliardi et al. 2021 (26)

ExoQuick (Systems Biosciences)

L1CAM (clone 5G3)

Exo-Check Antibody Array

NTA

TEM

WB

Sandwich ELISA (SNCOα; MyBiosource)

32

0

40

PD: 69.5 ± 8.6

HC: 57.4 ± 7.6

PD: 34.4%

HC: 47.5%

6.3 ± 3.6

2.0 ± NA

28.5 ± 13.2

NA

24.2 ± 2.5

Jiang et al. 2021 (27)

Direct IP

L1CAM (clone UJ127)

NTA

SEM

WB

ECLIA (Meso Scale Discovery)

290

50

0

116

88

0

191

PD: 65.1 ± 7.8

MSA: 67.1 ± 10.0

PSP: 69.5 ± 2.2

CBD: 64.6 ± 7.2

HC: 64.4 ± 6.8

PD:

MSA:

PSP:

CBD:

HC:

PD: 7.4 ± 3.1

MSA: 5.2 ± 2.7

PSP: 3.5 ± 2.2

CBD: 3.3 ± 2.0

NA

PD: 25.9 ± NA

MSA: 27.7 ± NA

PSP: 31.0 ± NA

CBD: 36.1

NA

PD: 26.8 ± NA

MSA: 26.0 ± NA

PSP: 22.0 ± NA

CBD: 20.9 ± NA

Dutta et al. 2021 (40)

ExoQuick (Systems Biosciences)

L1CAM (clone 5G3)

MOG (clone D-2)

FC

TEM

TRPS

WB

ECLIA (Meso Scale Discovery)

104

80

0

101

PD: 66.8 ± 9.3

MSA: 62.8 ± 8.1

HC: 64.9 ± 10.5

PD: 55.5%

MSA: 51.2%

HC: 55.4%

PD: 7.0 ± 4.4

MSA: 5.0 ± 2.8

PD: 2.4 ± 0.9

MSA: 3.8 ± 1.9

PD: 20.7 ± 14.3

MSA: NA

PD: 27.0 ± 4.2

MSA 27.2 ± 5.5

NA

Meloni et al. 2023 (32)

ExoQuick (Systems Biosciences)

L1CAM (clone 5G3)

Exo-Check Antibody Array

NTA

TEM

WB

Sandwich ELISA (SNCOα; MyBiosource)

70

0

21

19

0

PD: 69.5 ± 7.5

PSP: 72.8 ± 8.5

CBS: 71.9 ± 8.0

PD: 44.2%

PSP: 47.6%

CBS: 57.9%

PD: 7.3 ± 5.6

PSP: 4.0 ± 1.6

CBS: 4.4 ± 3.1

PD: 2.1 ± 0.6

PSP: NA

CBS: NA

PD: 33.0 ± 14.4

PSP: NA

CBD: NA

NA

PD: 25.1 ± 2.7

PSP: 17.8 ± 5.1

CBD: 17.0 ± 8.1

Taha et al. 2023 (41)

ExoQuick (Systems Biosciences)

L1CAM (clone 5G3)

MOG (clone D-2)

FC

TEM

TRPS

WB

ECLIA (in-house; [16])

46

30

0

32

PD: 66.8 ± 11.6

MSA: 62.7 ± 8.2

46.8%

PD: 8.1 ± 5.0

MSA: 62.7 ± 8.2

PD: 2.5 ± 1.0

MSA: 3.8 ± 1.0 (

PD: 25.1 ± 15.6

PD: 26.3 ± 6.4

MSA: 26.5 ± 9.3

NA

Sharafeldin et al. 2023 (28)

On-Chip Immunocapture

L1CAM (clone UJ127)

DLS

FM

NTA

WB

In-house electrochemical assay

20

0

23

29

NA

Abbreviations: CBS – corticobasal syndrome; CFM – confocal fluorescence microscopy; CNPase – 2′,3′-cyclic-nucleotide 3′-phosphodiesterase; DLB – dementia with Lewy body; ECLIA – electrochemilumiscence ELISA; ELISA – Enzyme-linked immunosorbent assay; EM – electron microscopy; EV – extracellular vesicle; FC – flow-cytometry; HC – healthy control; HY – Hoehn and Yahr disease stage scale45; IP – immunoprecipitation; L1CAM – L1 cell adhesion molecule; MCI – mild cognitive impairment; MMSE – Mini-mental state examination; MoCA – Montreal cognitive assessment; MOG – myelin oligodendrocyte glycoprotein; MSA – multiple system atrophy; NC – non-cognitively impaired; NTA – nanoparticle tracking analysis; PD – Parkinson's disease; PDD – PD with dementia; pS129-α-syn – phosphorylated α-syn at Ser 129; PSP – progressive supranuclear palsy; SEM – scanning EM; TEM – transmission EM; TRPS – tunable resistive pulse sensing; UMSARS – unified multiple system atrophy rating scale48; UPDRSIII – Unified Parkinson's disease rating scale.49; WB – Western blot.

TABLE 2. Descriptive statistics of the diagnostic metrics of studies included in the meta-analysis.

	Study	Biomarkers	Sensitivity (95% CI)	Specificity (95% CI)	FPR (95% CI)	DOR (95% CI)	posLR (95% CI)	negLR (95% CI)
PD vs. Control
	Shi et al. 2014 (11)	nEVs α-syn	0.701 (0.659 - 0.740)	0.529 (0.484 - 0.573)	0.471 (0.427 - 0.516)	2.64 (2.02 - 3.44)	1.49 (1.33 - 1.66)	0.565 (0.481 - 0.663)
	Shi et al. 2014 (11)	nEVs α-syn/total α-syn	0.712 (0.670 - 0.750)	0.500 (0.456 - 0.544)	0.500 (0.456 - 0.544)	2.47 (1.89 - 3.22)	1.42 (1.28 - 1.58)	0.577 (0.488 - 0.681)
	Zhao et al. 2019 (34)	nEVs DJ-1/total DJ-1	0.595 (0.485 - 0.696)	0.823 (0.724 - 0.891)	0.177 (0.109 - 0.276)	6.82 (3.28 - 14.17)	3.36 (2.02 - 5.58)	0.492 (0.370 - 0.655)
	Zhao et al. 2019 (34)	nEVs α-syn + DJ-1	0.823 (0.724 - 0.891)	0.519 (0.411 - 0.626)	0.481 (0.374 - 0.589)	5.01 (2.42 - 10.36)	1.71 (1.33 - 2.20)	0.341 (0.203 - 0.575)
	Niu et al. 2020 (35)	nEVs α-syn	0.973 (0.907 - 0.993)	0.541 (0.428 - 0.649)	0.459 (0.351 - 0.572)	42.35 (9.67 - 185.60)	2.12 (1.65 - 2.72)	0.050 (0.013 - 0.199)
	Niu et al. 2020 (early-stage PD vs. HC) (35)	nEVs α-syn	0.095 (0.047 - 0.183)	0.568 (0.454 - 0.674)	0.432 (0.326 - 0.546)	0.14 (0.06 - 0.34)	0.22 (0.10 - 0.46)	1.595 (1.290 - 1.972)
	Zou et al. 2020 (36)	nEVs α-syn, Linc-POU3F3 and plasma GCase activity	0.708 (0.637 - 0.770)	0.831 (0.770 - 0.879)	0.169 (0.121 - 0.230)	11.95 (7.19 - 19.87)	4.20 (2.99 - 5.90)	0.351 (0.277 - 0.446)
	Si et al. 2019 (38)	nEVs α-syn	0.661 (0.530 - 0.771)	0.714 (0.585 - 0.816)	0.286 (0.184 - 0.415)	4.87 (2.19 - 10.85)	2.31 (1.47 - 3.64)	0.475 (0.318 - 0.710)
	Jiang et al. 2020 (39)	nEVs α-syn	0.850 (0.812 - 0.881)	0.740 (0.696 - 0.780)	0.260 (0.220 - 0.304)	16.07 (11.38 - 22.71)	3.27 (2.77 - 3.86)	0.203 (0.161 - 0.257)
	Agliardi et al. 2021 (26)	nEVs α-syn/STX-1A	0.861 (0.763 - 0.923)	0.819 (0.715 - 0.891)	0.181 (0.109 - 0.285)	28.14 (11.46 - 69.08)	4.77 (2.89 - 7.87)	0.169 (0.094 - 0.304)
	Agliardi et al. 2021 (26)	nEVs α-syn/VAMP-2	0.750 (0.639 - 0.836)	0.931 (0.848 - 0.970)	0.069 (0.030 - 0.152)	40.20 (14.02 - 115.30)	10.80 (4.59 - 25.42)	0.269 (0.179 - 0.403)
	Jiang et al. 2021 (27)	nEVs α-syn	0.819 (0.782 - 0.851)	0.740 (0.699 - 0.777)	0.260 (0.223 - 0.301)	12.90 (9.47 - 17.57)	3.15 (2.70 - 3.69)	0.244 (0.201 - 0.298)
	Jiang et al. 2021 (27)	nEVs α-syn/Clusterin	0.861 (0.827 - 0.889)	0.869 (0.836 - 0.896)	0.131 (0.104 - 0.164)	40.99 (28.32 - 59.34)	6.57 (5.21 - 8.30)	0.160 (0.128 - 0.201)
	Dutta et al. 2021 (40)	nEVs α-syn, oEVs:nEVs α-syn, EV concentration	0.712 (0.618 - 0.790)	0.625 (0.529 - 0.712)	0.375 (0.288 - 0.471)	4.11 (2.30 - 7.35)	1.90 (1.44 - 2.50)	0.462 (0.330 - 0.646)
	Yan et al. 2022 (42)	EVs and nEVs α-syn	0.837 (0.748 – 0.899)	0.663 (0.562 - 0.751)	0.337 (0.249 - 0.438)	10.1 (5.01 - 20.38)	2.48 (1.84 - 3.35)	0.246 (0.151 - 0.400)
	Taha et al. 2023 (41)	nEVs α-syn, oEVs:nEVs α-syn, oEVs pS129-α-syn, EV concentration	0.718 (0.610 - 0.806)	0.897 (0.810 - 0.947)	0.103 (0.053 - 0.190)	22.27 (9.22 - 53.82)	7.00 (3.58 - 13.69)	0.314 (0.219 - 0.451)
	Taha et al. 2023 (41)	nEVs α-syn, oEVs:nEVs α-syn, oEV tau, EV concentration	0.419 (0.329 - 0.515)	0.848 (0.767 - 0.904)	0.152 (0.096 - 0.233)	4.01 (2.08 - 7.75)	2.75 (1.66 - 4.55)	0.685 (0.572 - 0.822)
	Sharafeldin et al. 2023 (28)	nEVs α-synuclein	0.650 (0.495 – 0.779)	0.950 (0.835 – 0.986)	0.050 (0.0138 – 0.165)	35.3 (7.39 – 168.4)	13.0 (3.30 – 51.1)	0.368 (0.240 – 0.565)
	Jiao et al. 2023 (43)	nEVs α-syn, plasma CCL2 and CXCL12	0.680 (0.583 – 0.763)	0.940 (0.875 – 0.972)	0.060 (0.0277 – 0.124)	33.3 (13.2 – 84.0)	11.3 (5.16 – 24.9)	0.340 (0.255 – 0.455)
PD vs. MSA
	Yu et al. 2020 (37)	oEVs α-syn	0.621 (0.501 - 0.729)	0.818 (0.709 - 0.893)	0.182 (0.107 - 0.291)	7.38 (3.32 - 16.41)	3.42 (1.98 - 5.89)	0.463 (0.333 - 0.643)
	Yu et al. 2020 (37)	oEVs α-syn/total α-syn	0.530 (0.412 - 0.646)	0.848 (0.743 - 0.916)	0.152 (0.084 - 0.257)	6.32 (2.76 - 14.48)	3.50 (1.89 - 6.47)	0.554 (0.420 - 0.729)
	Jiang et al. 2021 (27)	nEVs α-syn	0.821 (0.776 - 0.858)	0.859 (0.818 - 0.892)	0.141 (0.108 - 0.182)	27.82 (18.42 - 42.02)	5.81 (4.45 - 7.59)	0.209 (0.166 - 0.263)
	Jiang et al. 2021 (27)	nEVs α-syn/Clusterin	0.909 (0.873 - 0.935)	0.641 (0.589 - 0.690)	0.359 (0.310 - 0.411)	17.81 (11.58 - 27.40)	2.53 (2.19 - 2.93)	0.142 (0.101 - 0.201)
	Dutta et al. 2021 (40)	nEVs α-syn, oEVs:nEVs α-syn, EV concentration	0.893 (0.819 - 0.939)	0.864 (0.785 - 0.917)	0.136 (0.083 - 0.215)	53.17 (22.91 - 123.37)	6.57 (4.02 - 10.74)	0.124 (0.070 - 0.217)
	Taha et al. 2023 (41)	nEVs α-syn, oEVs:nEVs α-syn, oEVs pS129-α-syn, EV concentration	0.803 (0.700 - 0.877)	0.895 (0.806 - 0.946)	0.105 (0.054 - 0.194)	34.57 (13.71 - 87.18)	7.63 (3.92 - 14.83)	0.221 (0.139 - 0.349)
	Taha et al. 2023 (41)	nEVs α-syn, oEVs:nEVs α-syn, oEVs tau, EV concentration	0.946 (0.879 - 0.977)	0.707 (0.607 - 0.790)	0.293 (0.210 - 0.393)	41.89 (15.30 - 114.65)	3.22 (2.34 - 4.44)	0.077 (0.032 - 0.182)
PD vs. PSP and CBS
	Jiang et al. 2020 (FTD included) (39)	nEVs α-syn + Clusterin	0.918 (0.888 - 0.941)	0.958 (0.935 - 0.974)	0.042 (0.026 - 0.065)	258.31 (142.92 - 466.86)	22.09 (13.95 - 34.97)	0.086 (0.062 - 0.118)
	Jiang et al. 2021 (27)	nEVs α-syn/Clusterin	0.999 (0.990 - 1.000)	0.948 (0.925 - 0.965)	0.052 (0.035 - 0.075)	18209.24 (1105.43 - 299953.43)	19.39 (13.29 - 28.30)	0.001 (0.000 - 0.017)
	Meloni et al. 2023 (32)	nEVs α-syn/tau	0.941 (0.881 - 0.972)	0.671 (0.579 - 0.752)	0.329 (0.248 - 0.421)	32.82 (13.53 - 79.57)	2.86 (2.19 - 3.75)	0.087 (0.041 - 0.186)
MSA vs. Control
	Yu et al. 2020 (37)	oEVs α-syn	0.836 (0.727 - 0.907)	0.711 (0.590 - 0.808)	0.289 (0.192 - 0.410)	12.53 (5.33 - 29.44)	2.89 (1.94 - 4.31)	0.231 (0.130 - 0.410)
	Yu et al. 2020 (37)	oEVs α-syn/total α-syn	0.523 (0.403 - 0.641)	0.555 (0.433 - 0.670)	0.445 (0.330 - 0.567)	1.37 (0.68 - 2.74)	1.18 (0.82 - 1.68)	0.859 (0.613 - 1.204)
	Dutta et al. 2021 (40)	nEVs α-syn, oEVs:nEVs α-syn, EV concentration	0.956 (0.897 - 0.982)	0.838 (0.755 - 0.897)	0.162 (0.103 - 0.245)	112.27 (38.05 - 331.28)	5.91 (3.79 - 9.21)	0.053 (0.021 - 0.130)
	Taha et al. 2023 (41)	nEVs α-syn, oEVs:nEVs α-syn, oEVs pS129-α-syn, EV concentration	0.992 (0.928 - 0.999)	0.960 (0.880 - 0.988)	0.040 (0.012 - 0.120)	3025.00 (142.28 - 64314.30)	25.00 (7.42 - 84.25)	0.008 (0.001 - 0.131)
	Taha et al. 2023 (41)	nEVs α-syn, oEVs:nEVs α-syn, oEVs tau, EV concentration	0.880 (0.800 - 0.931)	0.922 (0.851 - 0.961)	0.078 (0.039 - 0.149)	86.70 (32.97 - 228.04)	11.27 (5.65 - 22.49)	0.130 (0.075 - 0.224)
RBD vs. Control
	Jiang et al. 2020 (39)	nEVs α-syn	0.610 (0.530 – 0.684)	0.810 (0.742 – 0.865)	0.188 (0.133 – 0.258)	6.67 (4.00 – 11.48)	3.25 (2.27 – 4.64)	0.479 (0.386 – 0.595)
	Yan et al. 2022 (42)	nEVs α-syn	0.936 (0.894 – 0.962)	0.717 (0.652 – 0.773)	0.282 (0.226 – 0.346)	38.33 (20.3 – 72.5)	3.32 (2.67 – 4.13)	0.087 (0.051 – 0.148)
	Sharafeldin et al. 2023 (28)	nEVs α-synuclein	0.682 (0.511 – 0.814)	0.985 (0.870 – 0.998)	0.0151 (0.00158 – 0.129)	139.3 (7.76 – 2500.0)	45.0 (2.84 – 711.4)	0.323 (0.196 – 0.533)

Abbreviations: 95% CI – 95% confidence interval; CCL2 – C-C motif chemokine ligand 2; CBS – corticobasal syndrome; CXCL12 – C-X-C motif chemokine ligand 12; DJ-1 – protein deglycase DJ-1; DOR – diagnostic odds ratio; EV – extracellular vesicle; FPR – false positive rate; FTD – frontotemporal dementia; GCase – glucocerebrosidase; HC – healthy control; Linc-POU3F3 – Long intergenic noncoding RNA POU3F3; MSA –multiple system atrophy; negLR – negative likelihood ratio; nEVs – neuronal extracellular vesicles; oEVs – oligodendroglial extracellular vesicles; PD – Parkinson's disease; posLR – positive likelihood ratio; PSP - progressive supranuclear palsy; pS129-α-syn – phosphorylated Serine 129 α-synuclein; RBD – REM behavior disorder; ; α-synuclein – α-syn; STX-1A – syntaxin-1A; VAMP-2 – vesicle associated membrane protein 2.

Supplementary Tables 1-3 and Supplementary Figures 1-4 are not available with this version.

Diagnostic Accuracy of Biomarkers in CNS-originating Extracellular Vesicles for Parkinsonian Disorders: A meta-analysis

Status:

Journal Publication

Version 2

Abstract

Figures

Background

Methodology

Results

Discussion

Conclusion

Declarations

Abbreviations

References

Tables

Supplementary Files

Status:

Journal Publication

Version 2