Header
Header
Article

A meta-analysis of the diagnostic utility of biomarkers in cerebrospinal fluid in Parkinson’s disease


Literature search findings

We identified 3862 articles after duplicate search removal (Fig. 1). Among these studies, 123 focused on CSF biomarkers and were included in the analysis. Overall, our analysis consisted of 11688 PD patients, 6735 controls (5757 healthy controls [HC], 978 other neurology disorders [OND]), 859 MSA patients, 327 PSP patients, 708 DLB patients, and 32 patients with CBD. The quality assessment and baseline characteristics of our study are reported in Supplementary Fig. 3. Quantitative analysis results for 16 biomarkers from our meta-analysis are provided in Supplementary Table 4. Figures 2 and 3 show the results with P < 0.05. The remaining results are presented in Supplementary Table 5. Forest plots for those biomarkers are available in Supplementary Figs. 610. The results of performance of potential CSF biomarkers for PD and parkinsonism patients and controls were listed in Table 1.

Fig. 1: Flow chart of the systematic article selection strategy.
figure 1

From 19326 studies from Pubmed, and 21319 studies from Science, Cochrane and Embase identified from the search strategy, a total of 123 articles were included for review. The boxes indicate exclusions of articles during each of the following screening stages: duplicate recoeds, unrelated recos and others. Reasons for exclusion are shown inside the right boxes. The box at bottom indicates inclusion articles.

Fig. 2: The number of studies in the meta-analysis for PD versus the parkinsonism and control groups.
figure 2

a The number of included studies for the PD versus parkinsonism groups. The number of studies included for the indicated biomarkers comparing PD patients to those with MSA, PSP, DLB, or CBD is illustrated. b The number of included studies for the PD versus control groups. The number of studies included for the indicated biomarkers comparing PD patients to the control group or the combined HC and OND groups is illustrated. PD Parkinson’s disease, PSP progressive supranuclear palsy, MSA multiple system atrophy, DLB dementia with Lewy bodies, CBD cortico-basal degeneration, HC healthy control, OND other neurodegenerative diseases, Aβ42 the 42-amino-acid form of Aβ, p-tau phosphorylated tau, t-tau total tau, NFL neurofilament light-chain protein, α-syn α-synuclein, IL-6 interleukin-6, CRP C-reactive protein, CHI3L1 YKL-40, chitinase-3-like protein 1, Cu copper, Mn manganese, Fe iron, Zn zinc.

Fig. 3: Meta-analysis for CSF biomarker performance measured with the effect size (ES) and 95% confidence interval (CI) for PD versus control or parkinsonism patients.
figure 3

a The biomarkers illustrated in blue and marked as 1 were significantly different between PD patients and HC. The biomarkers illustrated in green and marked as 2 were significantly different between PD and OND patients. A significant difference was identified between the PD patients and those in the OND and HC groups combined for the orange-labeled biomarkers marked as 3. b The biomarkers illustrated in blue were significantly different between PD and MSA patients. The biomarkers illustrated in green were significantly different between PD and PSP patients. A significant difference was observed between the PD and DLB patients for the orange-labeled biomarkers. The biomarker marked pink was significantly different between the PD and CBD patients. PD Parkinson’s disease, OND other neurodegenerative diseases, HC healthy control, MSA multiple system atrophy, DLB dementia with Lewy bodies, PSP progressive supranuclear palsy, CBD cortico-basal degeneration, α-syn α-synuclein, Aβ42 the 42-amino-acid form of Aβ, t-tau total tau, p-tau phosphorylated tau, NFL neurofilament light-chain protein, YKL-40, CHI3L1 chitinase-3-like protein 1.

PD patients versus controls

A total of 103 studies (9192 PD patients, 5661 controls) assessed the performance of Aβ42, p-tau, t-tau, t-α-syn, p-α-syn, o-α-syn, NFL, Zn, YKL-40, CRP, IL-6, Cu, Fe, Mn, arginine, DJ-1, and citrulline in distinguishing PD patients from controls. Random-effect results showed that PD patients had lower CSF levels for Aβ42 (SMD = −0.239, 95% confidence interval [CI]: −0.309 to −0169), p-tau (SMD = −0.302, 95% CI: −0.376 to −0.228), t-tau (SMD = −0.274, 95% CI: −0.349 to −0.200), t-α-syn (SMD = −0.419, 95% CI: −0.542 to −0.295), Zn (SMD = −0.398, 95% CI: −0.623 to −0.173), DJ-1 (SMD = −0.791, 95% CI: −1.380 to −0.202), and YKL-40 (SMD = −0.322, 95% CI: −0.561 to −0.082) (Fig. 3). In contrast, the CSF levels of o-α-syn (SMD = 1.754, 95% CI: 0.590 to 2.919) and phosphorylated α-syn (p-α-syn) were elevated. No significant differences were found in NFL, CRP, IL-6, Cu, Fe, Mn, arginine, or citrulline concentrations between PD patients and controls.

A high level of heterogeneity was found for DJ-1, and the sensitivity analysis showed that the results for the performance of most CSF biomarkers were not affected by specific studies. Significant bias was not found in most studies, as indicated by Egger’s test and funnel plots. Meta-regression and subgroup analysis were performed to potentially identify the primary source of the heterogeneity. A significant difference was not observed between the subgroup analysis and overall results. However, the random-effect results showed that no significant differences were found for t-tau (SMD = −0.129, 95% CI: −0.269 to 0.012) or p-tau (SMD = −0.13, 95% CI: −0.27 to 0.01) when ONDs were used as controls.

Therefore, DJ-1, t-α-syn, Zn, and YKL-40 significantly distinguished PD from controls with a moderate effect size. However, because of the limited number of studies included in the analysis for DJ-1, Zn, and YKL-40, more studies are needed to further identify their potential diagnostic values. Nevertheless, we believe that t-α-syn will help distinguish PD from controls.

PD versus atypical parkinsonian syndromes

PD versus MSA

Seventeen studies (1,383 PD patients, 503 MSA patients) assessed the capacity of NFL, t-α-syn, t-tau, YKL-40, p-tau, FLT-3, and CRP in differentiating PD and MSA patients. Lower CSF levels of NFL (SMD = −3.609, 95% CI: −4.545 to −1.594), t-tau (SMD = −0.977, 95% CI: −1.520 to −0.434), YKL-40 (SMD = −0.973, 95% CI: −1.292 to −0.655), and CRP (SMD = −0.556, 95% CI: −0.871 to −0.242) were observed in PD patients compared with MSA patients. In contrast, the CSF levels of t-α-syn (SMD = 0.257, 95% CI: 0.055 to 0.459) were elevated. No significant differences were detected for p-tau, Aβ42, or FLT-3 between these two groups.

Therefore, NFL and YKL-40 can significantly distinguish PD from MSA with a large effect size, while t-tau and CRP can only distinguish with a moderate effect size. Because of the small number of studies included for YKL-40 and CRP, they cannot be considered promising biomarkers at this time.

PD Versus PSP

Ten studies (1243 PD patients, 327 PSP patients) assessed the performance of NFL, Aβ42, t-α-syn, and p-tau in distinguishing between PD and PSP patients. Lower CSF levels of NFL (SMD = −1.509, 95% CI: −2.222 to −0.796) and higher Aβ42 levels were observed in PD patients compared with the PSP patients (SMD = 0.561, 95% CI: 0.103 to 1.018). However, no significant differences were detected for t-α-syn or p-tau between these two groups.

In general, NFL can significantly distinguish PD from PSP with a large effect size, while Aβ42 could only distinguish with a moderate effect size.

PD versus DLB

Twenty studies (1,384 PD patients, 708 DLB patients) assessed the performance of p-tau, t-tau, t-α-syn, and Aβ42 in distinguishing PD from DLB. Reduced CSF levels of p-tau (SMD = −0.495, 95% CI: −0.689 to −0.301) and t-tau (SMD = −0.626, 95% CI: −0.888 to −0.365) and higher Aβ42 levels (SMD = 0.775, 95% CI: 0.498 to 1.052) were observed in PD patients relative to DLB patients. However, no significant difference was found for t-α-syn between these two groups.

In general, Aβ42, p-tau, and t-tau were significant for PD with a moderate effect size compared with DLB.

PD versus CBD

Three studies (199 PD patients, 32 CBD patients) evaluated the performance of NFL in differentiating PD and CBD patients. The CSF levels of NFL (SMD = −1.421, 95% CI: −1.928 to −0.914) were reduced in PD patients compared with the CBD patients.

Therefore, NFL can significantly distinguish PD from CBD with a large effect size.

Assessment of heterogeneity

High levels of heterogeneity were found for NFL (PD versus MSA), t-tau and Aβ42 (PD versus DLB), and NFL and Aβ42 (PD versus PSP). Sensitivity analysis showed that the capacity of all biomarkers in CSF was not affected by specific studies. We failed to detect a significant bias in most studies based on funnel plots and Egger’s test.

Subgroup and meta-regression analyses for potential moderators

The influence of potential moderators on expression levels of biomarkers, including age, disease stage, and analytical methods, were performed by subgroups and meta-regression analyses (Table S12). The detection assay did not influence the consistency of our results except for the real-time quaking-induced conversion (RT-QuIC) detection of t-α-syn, which was used to detect t-α-syn aggregates instead of free α-syn. Subgroup analysis for the de novo cohort also confirmed the stability of our results. Age (P = 0 .001) and MMSE (P = 0.004) were significant moderators for t-α-syn and Aβ42, while age (P = 0.008) was a significant moderator for NFL in distinguishing between PD and controls.

Accuracy of biomarkers in PD

For all the biomarkers mentioned above, 18 studies on t-α-syn biomarkers were available to discriminate PD from controls. The pooled sensitivity and specificity were 85% (95% CI = 0.77–0.90) and 74% (95% CI = 0.67–0.80), with an area under the receiver operating characteristic curve (AUC) 0.85 (95% CI = 0.82–0.88) for t-α-syn (Fig. 4). The subgroup analysis of t-α-syn showed that the sensitivity and specificity were 84% (95% CI = 0.66–0.94) and 70% (95% CI = 0.55–0.82) respectively, with AUC 0.81 (95% CI = 0.77–0.84), when other neurological diseases were used as controls. We failed to detect a significant bias in most studies based on funnel plots (P > 0.05).

Fig. 4: SROC with prediction value for α-syn and NFL in two Groups.
figure 4

a SROC with prediction value for PD and control group; b SROC with prediction value for PD and MSA. SROC summary receiver operating characteristic curve, PD Parkinson’s disease, MSA multiple system atrophy, α-syn α-synuclein, NFL neurofilament light-chain protein.

Four studies each on t-α-syn and NFL were available to discriminate PD from MSA. The pooled sensitivity and specificity were 89 % (95% CI = 0.82–0.93) and 66% (95% CI = 0.30–0.90), with AUC 0.89 (95% CI = 0.86–0.92) for t-α-syn. For NFL, the pooled sensitivity and specificity were 98 % (95% CI = 0.89–1.00) and 84% (95% CI = 0.74–0.91), with AUC 0.94 (95% CI = 0.92–0.96) (Fig. 4). Due to the limited number of studies, we failed to conduct a diagnostic meta-analysis on t-α-syn and NFL in PD versus PSP, PD versus DLB, and PD versus CBD groups. Moreover, we failed to conduct a diagnostic analysis for other potential biomarkers in other groups due to the low number of included studies.

However, the HSROC curve of the pooled data showed that the diagnostic effectiveness of t-α-syn was influenced by different cut-off values (P < 0.05); while NFL was not influenced by different cut-off values. We failed to detect a significant bias in most studies based on funnel plots (P > 0.05).

Recommendation for CSF biomarkers

According to the diagnostic utility and included number of studies, we found that t-α-syn showed a good performance in distinguishing between PD and controls. NFL was a good biomarker for the differential diagnosis of PD from all APSs except DLB; however, Aβ42 could bridge the gap. Therefore, we believe that the combination of t-α-syn, NFL, and Aβ42 could be promising for the diagnosis and differential diagnosis of PD (Fig. 3).

Recommendation for CSF biomarkers

According to the diagnostic utility and included number of studies, we found that t-α-syn showed a good performance in distinguishing between PD and controls. NFL was a good biomarker for the differential diagnosis of PD from all APSs except DLB; however, Aβ42 could bridge the gap. Therefore, we believe that the combination of t-α-syn, NFL, and Aβ42 could be promising for the diagnosis and differential diagnosis of PD (Fig. 3).



Source link

Back to top button