590 Shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

ReseaRch papeR

Inflammatory markers in Alzheimer’s disease and mild cognitive impairment: a meta-analysis and systematic review of 170 studiesXue-Ning shen,1 Li-Dong Niu,1 Yan-Jiang Wang,2 Xi-peng cao,3 Qiang Liu,4 Lan Tan,1 can Zhang,5 Jin-Tai Yu6

Cognitive neurology

To cite: shen X-N, Niu L-D, Wang Y-J, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598.

► additional material is published online only. To view please visit the journal online (http:// dx. doi. org/ 10. 1136/ jnnp- 2018- 319148).

For numbered affiliations see end of article.

Correspondence toprofessor Jin-Tai Yu, Department of Neurology, huashan hospital, Fudan University, shanghai 200000, china; yu- jintai@ 163. com

X-Ns and L-DN contributed equally.

Received 2 July 2018Revised 5 November 2018accepted 10 December 2018published Online First 10 January 2019

© author(s) (or their employer(s)) 2019. No commercial re-use. see rights and permissions. published by BMJ.

AbsTrACTObjective Inflammation plays a crucial role in the pathogenesis of mild cognitive impairment (McI) and alzheimer’s disease (aD). Our study aimed to analyse previous inconsistent results of inflammatory markers in aD and McI quantitatively.Methods studies reporting concentrations of peripheral or cerebrospinal fluid (csF) markers were included, and eligible data on aD, McI and control were extracted. pooled hedges’s g was adopted to illustrate comparisons, and various confounding factors were used to explore sources of heterogeneity.results a total of 170 studies were included in the meta-analysis and systematic review, which demonstrated increased peripheral levels of high-sensitivity c reactive protein (hedges’s g 0.281, p<0.05), interleukin-6 (IL-6) (0.429, p<0.005), soluble tumour necrosis factor receptor 1 (sTNFR1) (0.763, p<0.05), soluble tumour necrosis factor receptor 2 (sTNFR2) (0.354, p<0.005), alpha1-antichymotrypsin (α1-acT) (1.217, p<0.005), IL-1β (0.615, p<0.05) and soluble cD40 ligand (0.868, p<0.005), and csF levels of IL-10 (0.434, p<0.05), monocyte chemoattractant protein-1 (Mcp-1) (0.798, p<0.005), transforming growth factor-beta 1 (1.009, p<0.05), soluble triggering receptor expressed on myeloid cells2 (sTReM2) (0.587, p<0.001), YKL-40 (0.849, p<0.001), α1-acT (0.638, p<0.001), nerve growth factor (5.475, p<0.005) and visinin-like protein-1 (VILIp-1) (0.677, p<0.005), in aD compared with the control. higher levels of sTNFR2 (0.265, p<0.05), IL-6 (0.129, p<0.05) and Mcp-1 (0.779, p<0.05) and lower levels of IL-8 (−1.293, p<0.05) in the periphery, as well as elevated concentrations of YKL-40 (0.373, p<0.05), VILIp-1 (0.534, p<0.005) and sTReM2 (0.695, p<0.05) in csF, were shown in McI compared with the control. additionally, increased peripheral sTNFR1 (0.582, p<0.05) and sTNFR2 (0.254, p<0.05) levels were observed in aD compared with McI.Conclusion significantly altered levels of inflammatory markers were verified in comparison between aD, McI and control, supporting the notion that aD and McI are accompanied by inflammatory responses in both the periphery and csF.

InTrOduCTIOnAlzheimer’s disease (AD) is the primary common type of dementia, increasing at an alarming rate in older people.1 It is one of the severest neurode-generative disorders leading to decreased quality of

daily life, disability and eventually death.2 Before patients are diagnosed with dementia, they first go through a stage recognised as mild cognitive impairment (MCI), which is widely considered as the intermediate stage between intact cognition and pathological cognitive ageing.3 Accumulating evidence has suggested that inflammatory responses play a critical role in the neurodegenerative cascades of AD and MCI,4 5 and several markers have some tracing and detecting accuracy for disease severity and progression.6–8

Inflammatory markers such as the tumour necrosis factor α (TNF-α), interleukin-6 (IL-6), chitinase-3-like protein 1 (CHI3L1 or YKL-40) and the acute phase reactant protein C reactive protein (CRP) have been reported as important signalling molecules in inflammation that exert effects on the brains or the periphery of people with dementia.6 9 10 Decreased levels of circulating IL-8 are shown to be involved in AD pathogenesis and may serve as a potential biomarker for monitoring disease progres-sion.7 However, other studies have observed higher peripheral IL-8 levels in patients with AD compared with the healthy elderly.9 11 Inflammatory marker performances in individual studies vary greatly and need comprehensive analyses.12 13

Previous meta-analyses have demonstrated that some inflammatory markers in the peripheral blood or cerebrospinal fluid (CSF) differ between the AD group and the normal controls.6 14 15 But they lack the most recent updated information or have limitations with only peripheral data and only AD comparison with the control. To update this clinical evidence, this meta-analysis and systematic review was conducted to investigate whether the concen-trations of specific inflammatory markers differ quantitatively among patients with AD and MCI and the controls as measured from the peripheral blood and CSF.

MeThOdssearch strategy and selection criteriaThis meta-analysis and systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Original studies reporting measurements of inflammatory markers in the peripheral blood or CSF were searched between 1 January 1960 and 1 October 2018 through PubMed, Embase, Cochrane Library and Web of Science. Search terms

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from

591shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

Cognitive neurology

Figure 1 preferred Reporting Items for systematic Reviews and Meta-analyses flow chart of the literature search. NOs, Newcastle-Ottawa scale.

are presented in online supplementary material 1. The reference lists of relevant articles were also scanned for broader reports.

English-language or Chinese-language publications reporting concentrations of inflammation-associated markers in living human beings were included if they met the following criteria: (1) original studies reported data in at least two of the groups (AD, MCI and control); (2) literature sources and necessary data were available; and (3) the principles that these studies used to diagnose AD and MCI were qualified. The selection criteria for subjects are presented in online supplementary mate-rial 1. Articles were excluded if they measured marker concen-trations in postmortem samples, had sample size less than 5, and reported stimulated levels of markers in vitro or used the samples that overlapped with other studies. We have contacted some corresponding authors via email, and according to their replies prompt and proper adjustments were made to ensure no overlapped research population. Quality assessments of all potentially eligible studies were conducted using the Newcas-tle-Ottawa Scale (NOS). Studies with NOS scores lower than 5 were recognised to be of inferior quality and therefore excluded. Inflammatory markers measured in less than three studies were assessed qualitatively in the systematic review.

data extractionData were taken from cross-sectional studies and baseline of longitudinal studies included. The sample sizes and mean (±SD) concentrations of markers in those eligible literatures were extracted as the primary data, independently by two of the authors (X-NS and L-DN). Marker concentrations that are presented as stratified data or IQR were converted to applicable

data using mathematical formulas for meta-analysis (if requested data from corresponding authors were not available).16 17 Study characteristics (ie, first author, publication year, PubMedUn-iqueIdentifier (PMID)) and information for potential moderator analysis (ie, age, gender, education) were collected (see online supplementary material 2). Any inconsistencies were resolved by consensus.

statistical analysisStatistical analysis was performed using the Comprehensive Meta-Analysis Software (V.2). To adjust for the potential biases resulting from small sample sizes, Hedges’s g was chosen as the final effect size (ES),18 which was converted from standardised mean differences primarily generated from sample size and mean (±SD) marker concentration. The more conservative random-effects model was adopted rather than the fixed-ef-fects model. To circumvent the substantial variability caused by experimental methods and different laboratories, we used ratios between groups (ie, mean marker concentrations of AD group to the control group ratio) to illustrate more intuitionistic compar-isons for consolidated evidence. Every specific ratio was calcu-lated from a single study, and the mean ratio for each marker suggested the combined results of all included studies assessing the same marker.

We conducted sensitivity analyses by excluding one research at a time to assess the robustness of the outcomes. Cochrane Q test was used to calculate between-study heterogeneity, and inconsistency was estimated using the I2 index for the propor-tion of total variability due to heterogeneity. Subgroup analyses were conducted to investigate the possible effect of sample type

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from

592 shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

Cognitive neurology

Figure 2 comparative outcomes of blood and csF markers in the meta-analysis. peripheral (a) and csF (B) inflammatory markers with significant effect sizes (hedges’s g) are displayed in groups of comparisons (aD-cN, McI-cN and aD-McI). Red spots indicate hedges’s g of each marker, and blue bars indicate the number of studies included. α1-acT, alpha1-antichymotrypsin; aD, alzheimer’s disease; cN, controls; cs, cerebrospinal fluid; hscRp, high-sensitivity c reactive protein; IL, interleukin; McI, mild cognitive impairment; Mcp-1, monocyte chemoattractant protein-1; NGF, nerve growth factor; scD40L, soluble cD40 ligand; sTNFR1, soluble tumour necrosis factor receptor 1; sTNFR2, soluble tumour necrosis factor receptor 2; sTReM2, soluble triggering receptor expressed on myeloid cells2; TGF-β1, transforming growth factor-beta; VILIp-1, visinin-likeprotein-1.

and assay type. Meta-regression analyses were used to regress the ES against potential confounding factors (ie, age). Publica-tion bias was inspected by funnel plot, and statistical significance was assessed by Egger’s test, which calculated funnel plot asym-metry. Potential publication biases were further evaluated by the classic fail-safe N method, which analysed the number of missing studies that would lead to a statistically non-significant overall effect. P value <0.05 was accepted to be of statistical significance.

resulTsliterature search findingsThe initial literature search generated altogether 11 067 records (see flow chart in figure 1). A total of 170 studies measuring peripheral or CSF inflammatory markers were ultimately included, combining results for 9842 subjects with AD, 3526 subjects with MCI and 9002 controls (online supplementary materials 2 and 3). Among these, 46 markers were quantitatively analysed in the meta-analysis (online supplementary material 4; results with p<0.05 are shown in figures 2 and 3), while others were presented in the systematic review (online supplementary material 5; figures 4 and 5). Baseline characteristics and quality assessments of the studies are detailed in online supplementary material 2. All official marker names are presented in online

supplementary material 6. Individual marker performances and heterogeneity analyses are shown in online supplementary mate-rial 7.

Comparisons between Ad and control in peripheral marker levelsRandom-effects results demonstrated that patients with AD had higher levels for high-sensitivity CRP (hsCRP) (Hedges’s g 0.281; 95% CI 0.051 to 0.511, p<0.05), IL-6 (Hedges’s g 0.429; 95% CI 0.160 to 0.699, p<0.005), soluble tumour necrosis factor receptor 1 (sTNFR1) (Hedges’s g 0.763; 95% CI 0.153 to 1.373, p<0.05), soluble tumour necrosis factor receptor 2 (sTNFR2) (Hedges’s g 0.354; 95% CI 0.126 to 0.582, p<0.005), alpha1-antichymo-trypsin (α1-ACT) (Hedges’s g 1.217; 95% CI 0.421 to 2.013, p<0.005), IL-1 beta (IL-1β) (Hedges’s g 0.615; 95% CI 0.124 to 1.107, p<0.05) and soluble CD40 ligand (sCD40L) (Hedg-es’s g 0.868; 95% CI 0.289 to 1.448, p<0.005) (figure 2, online supplementary material 4A). The AD/control ratios of markers with significant ES were calculated to further manifest peripheral performances: hsCRP (mean ratio 3.046, p=0.176), IL-1β (2.463, p<0.001), sCD40L (2.368, p=0.007), sTNFR1 (1.962, p=0.036), IL-6 (1.898, p<0.0001), α1-ACT (1.325, p=0.006) and sTNFR2 (1.278, p<0.001) (figure 3).

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from

593shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

Cognitive neurology

Figure 3 Biomarker performance rating of peripheral blood and csF in the meta-analysis. Biomarker performance in serum and plasma (a) and csF (B) based on average ratios. Biomarkers illustrated in green are significant with effect sizes in patients with aD versus cN, in blue significant in patients with McI versus cN, and in pink significant in patients with aD versus McI. α1-acT, alpha1-antichymotrypsin; aD, alzheimer’s disease; cN, controls; csF, cerebrospinal fluid; hscRp, high-sensitivity c reactive protein; IL, interleukin; McI, mild cognitive impairment; Mcp-1, monocyte chemoattractant protein-1; scD40L, soluble cD40 ligand; sTNFR1, soluble tumour necrosis factor receptor 1; sTNFR2, soluble tumour necrosis factor receptor 2.

Comparisons between MCI and control in peripheral marker levelsRandom-effects meta-analyses showed increased peripheral levels of sTNFR2 (Hedges’s g 0.265; 95% CI 0.041 to 0.489, p<0.05), monocyte chemoattractant protein-1 (MCP-1) (Hedg-es’s g 0.779; 95% CI 0.036 to 1.521, p<0.05), IL-6 (Hedges’s g 0.129; 95% CI 0.013 to 0.246, p<0.05) and decreased levels of IL-8 (Hedges’s g −1.293; 95% CI −2.390 to −0.196, p<0.05) in patients with MCI compared with the controls (figure 2, online supplementary material 4B). The MCI to control ratios for IL-6 (1.467, p<0.0001), sTNFR2 (1.074, p<0.001), MCP-1 (1.137, p<0.0001) and IL-8 (0.861, p=0.011) are presented in figure 3.

Comparisons between Ad and MCI in peripheral marker levelsElevated peripheral concentrations of sTNFR1 (Hedges’s g 0.582; 95% CI 0.108 to 1.055, p<0.05) and sTNFR2 (Hedges’s g 0.254; 95% CI 0.072 to 0.436, p<0.05) were detected in AD compared with MCI (figure 2, online supplementary material 4C). Figure 3 also illustrates the AD to MCI ratios for sTNFR1 (1.255, p=0.016) and sTNFR2 (1.122, p<0.001).

Comparisons between Ad and control in CsF marker levelsHigher CSF levels of IL-10 (Hedges’s g 0.434; 95% CI 0.037 to 0.830, p<0.05), MCP-1 (Hedges’s g 0.798; 95% CI 0.293 to 1.304, p<0.005), transforming growth factor-beta 1 (TGF-β1) (Hedges’s g 1.009; 95% CI 0.287 to 1.730, p<0.05), soluble triggering receptor expressed on myeloid cells2 (sTREM2) (Hedges’s g 0.587; 95% CI 0.446 to 0.728, p<0.005), YKL-40 (Hedges’s g 0.849; 95% CI 0.616 to 1.081, p<0.005), α1-ACT (Hedges’s g 0.638; 95% CI 0.328 to 0.947, p<0.005), nerve growth factor (NGF) (Hedges’s g 5.475; 95% CI 2.040 to 8.909, p<0.005) and visinin-like protein-1 (VILIP-1) (Hedges’s g 0.677; 95% CI 0.326 to 1.027, p<0.005) were observed in subjects with AD compared with controls (figure 2, online supple-mentary material 4D). The AD to control ratios were calcu-lated for NGF (3.980, p=0.122), TGF-β1 (3.263, p=0.145),

IL-10 (1.489, p=0.043), α1-ACT (1.424, p=0.0001), MCP-1 (1.361, p<0.0001), VILIP-1 (1.383, p=0.002), YKL-40 (1.359, p<0.0001) and sTREM2 (1.332, p<0.0001) (figure 3).

Comparisons between MCI and control in CsF marker levelsRandom-effects meta-analysis demonstrated only a trend of higher CSF levels of YKL-40 (Hedges’s g 0.373; 95% CI 0.050 to 0.697, p<0.05), VILIP-1 (Hedges’s g 0.534; 95% CI 0.216 to 0.852, p<0.005) and sTREM2 (Hedges’s g 0.695; 95% CI 0.114 to 1.276, p<0.05) in subjects with MCI compared with controls (figure 2, online supplementary material 4E). The CSF MCI to control ratios of YKL-40 (1.202, p<0.0001), VILIP-1 (1.307, p=0.006) and sTREM2 (1.497, p=0.007) were observed slightly above 1.0 (figure 3).

Comparisons between Ad and MCI in CsF marker levelsDifferences in CSF levels of inflammatory markers between subjects with AD and with MCI were also tested (online supple-mentary material 4F). As there were no surprising findings, we did not perform further analysis for comparison between these groups. Markers which did not show significant differences in the comparison could be looked up in online supplementary materials 4 and 7.

Investigation of heterogeneityStrong evidence of heterogeneity was found in most compari-sons (online supplementary material 4). Sensitivity analyses indi-cated that the results of most markers’ performance in peripheral blood and CSF were not unduly influenced by some specific studies (online supplementary material 7). No significant publi-cation bias was detected in most of the studies, as demonstrated by the funnel plots and confirmed by the Egger’s tests (online supplementary materials 4 and 7). Potential publication bias was detected of vascular cell adhesion molecule-1 (VCAM-1) (Egger’s intercept −2.847, p=0.045) in the comparisons of AD with control (online supplementary material 4). For further

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from

594 shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

Cognitive neurology

Figure 4 performance of peripheral blood markers in the systematic review. significant serum and plasma markers of one or two studies are performed in the systematic review. Blue spots indicate hedges’s g of each marker, and pink bars indicate the number of studies included. α1-acT, alpha1-antichymotrypsin; aD, alzheimer’s disease; cFh, complement factor h; cN, controls; cTacK, cutaneous T-cell attracting chemokine; cX3cL-1, cX3 chemokine ligand 1; IL, interleukin; GDF-15, growth differentiation factor-15; KLK6, kallikrein-related peptidase 6; McI, mild cognitive impairment; MIF, macrophage migration inhibitory factor; MIG, c-X-c motif chemokine ligand 9; OpG, osteoprotegerin; scD40, scD40, soluble cD40; scFG-B, stochastic context-free grammar beta; TGF-β1, transforming growth factor-beta; VeGF, vascular endothelial growth factor.

verification, the classic fail-safe N was used to assess publication bias of VCAM-1, of which the results revealed that 0 missing study would be required to make p>0.05.

The results of subgroup analyses for most of the markers did not significantly differ from the overall results (online supple-mentary material 7). However, random-effects results showed that there are some differences in concentrations of TNF-α in serum (Hedges’s g 0.252; 95% CI 0.027 to 0.477, p<0.05) and IL-8 (Hedges’s g −0.967; 95% CI −1.905 to −0.029, p<0.05) in plasma between the AD group and the controls. Subject to the studies’ limited information, meta-regression analyses were mainly conducted for peripheral CRP, TNF-α, IL-6, IL-1β, IL-8, IL-10 and CSF IL-6 and YKL-40 (numbers of relevant studies no less than 10) (online supplementary material 7). A slight correla-tion of Mini-Mental State Exam (MMSE) score against Hedges’s g of peripheral IL-6 level (slope −0.069, p<0.001) was found in the comparison between the AD and control groups. Poten-tial correlations of CSF YKL-40 with age (0.086, p<0.001) and MMSE score (−0.132, p<0.001) in the comparison of AD with MCI and with age (−0.084, p<0.001) in the comparison of MCI with control were also detected. Furthermore, there were no significant outcomes in the remaining analytes.

systematic reviewOf 170 eligible studies, numerous inflammatory markers were qualitatively analysed in the systematic review rather

than included in the meta-analysis. The performances of these markers in the periphery or CSF are presented in figures 4 and 5 and online supplementary material 5, illustrating the comparison outcomes of each marker in one or two qualified studies.

Besides, some of the potential markers for AD were also involved in other neurological diseases, which could complicate the clinical scenario.19 To explore the specificity, performance of those markers in other neurological diseases are further listed in table 1 (online supplementary material 8).

dIsCussIOnInflammatory marker performanceThis meta-analysis analysed and summarised the inconsistent data of previous studies investigating peripheral blood or CSF inflammatory markers in relation to AD and MCI. It demon-strated multiple significant variances of inflammatory marker levels in the comparisons between AD, MCI and control groups. Further rated by ratios of peripheral or CSF concentrations between groups, several markers differentiated disease status with good performance. These findings suggested noteworthy blood and CSF alterations of inflammatory markers in AD and MCI, implying a critical role of inflammation in the pathological process of disease.

Results suggested that publication bias was less likely to be the cause for most of the statistical heterogeneity, while studies

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from

595shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

Cognitive neurology

Figure 5 performance of csF markers in the systematic review. significant csF markers of one or two studies are displayed in the systematic review. Blue spots indicate hedges’s g of each marker, and pink bars indicate the number of studies included. aD, alzheimer’s disease; BDNF, brain-derived neurotrophic factor; cN, controls; csF, cerebrospinal fluid; cX3cL-1, cX3 chemokine ligand 1; GDNF, glial cell line-derived neurotrophic factor; IL, interleukin; IL-2r, soluble IL-2 receptor; Ip, isoprostane; McI, mild cognitive impairment; M-csF, macrophage colony-stimulating factor; OpN, osteopontin; s100a9, s100 calcium-binding protein a9; scD14, soluble cD14; scF, haematopoietic growth factors-like stem cell factors; sp, substance p; sTNFR I, soluble tumour necrosis factor receptor 1; sTNFR II, soluble tumour necrosis factor receptor 2.

that included VCAM-1 had publication bias to some extent. Subgroup analyses showed that sample sources (plasma or serum) and assay types could explain a fraction of the hetero-geneity. Different commercial ELISA systems may also result in different concentrations of the inflammatory markers, due to the sensitivity and specificity of antibodies used. Peripheral IL-6 concentrations proved to be significantly increased in patients with AD compared with controls, and were slightly correlated with MMSE scores by meta-regression analysis. It showed simi-larly prominent ratio in the comparison between MCI and the control, but did not exhibit obvious associations with MMSE or age. These may indicate to some extent that inflammatory responses correlate with disease severity and that the underlying associations might not be apparent in the early stages of the disease.

In the current study, performance of specific inflammatory marker in the periphery was not strictly in conformity with that in CSF. For instance, there were apparent increases in periph-eral IL-6 and IL-1β levels between patients with AD and normal controls, whereas these differences did not prove significant in CSF comparisons. This could be due to diverse methodolo-gies and population heterogeneity among the included studies. Besides, concentrations of some markers remained undetectable in the CSF or blood, and lack of much effective laboratory data may partially account for the inconformity. Previously, Llano and his colleagues20 found that multiple inflammatory markers increased over time in CSF between the AD and the control groups, and such differences peaked at 24 hours after placement of the lumbar catheter. With no changes in plasma during this time, it is suggested that CSF inflammatory responses to lumbar catheter placement vary between the AD and the control groups.20 Analysing the correlation of these marker levels with the CSF to

serum albumin ratio, which is regarded as an index assessing the blood–cerebrospinal fluid barrier (BCSFB) integrity, peripheral plus central inflammatory processes were supported to play a critical role in the early changes of BCSFB function.21 This indi-cated a disruption of the blood–brain barrier (BBB), resulting in alterations of its permeability and transport capacity.21 Based on these, it is reasoned to speculate that the variances of specific inflammatory marker levels in body fluid may be related to that shifting distribution between the periphery and CSF.

Pathophysiological and clinical implicationsEmerging evidence suggests that an early and substantial impact of inflammatory markers is involved in the pathogenesis of AD.4 5 Research showed that peripheral and CSF inflammatory processes could mediate cognitive impairment in the early phase of neurodegeneration, the pathogenic processes which include inhibiting angiogenic, neurotrophic and neuroprotective mech-anisms, inducing neuron loss and activating injury to myelin.21

This meta-analysis verified that expression of proinflamma-tory cytokines and chemokines, such as IL-1β, IL-6 and IL-8, significantly increased in patients with AD, which have been broadly documented to correlate with the presence and metabo-lism of amyloid-beta (Aβ) or tau proteins that subsequently lead to neurodegenerative cascades of AD.22 23 Following Aβ deposi-tion, IL-1β has been reported to bring neuron dysfunction and progressive neurodegeneration in AD.7 22 In accord with previous meta-analyses,24 IL-6 was found to possess potential property that detects the severity of cognitive impairment in patients with AD. It is found that in the presence of Aβ deposition, microglia and astrocytes release IL-6, which could accelerate the patho-logical cascade of AD.4 23 Decreased levels of IL-8 in patients

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from

596 shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

Cognitive neurology

Table 1 Performance of potential inflammatory biomarkers in peripheral blood and CSF of AD and other neurological diseases

Marker Function class systemic actions Cns actions regulation in Ad regulation in other neurological diseases

NGF Neurotrophin Neurotrophic Neurotrophic ↑ CSF ↑ In CSF of major depression (DE) [1]

YKL-40 Protein – Unclear ↑ CSF ↑ In CSF of Huntington’s disease [2]

↓ In CSF of PD and atypical Parkinsonian disorders (P+ diseases) [3]

↑ In CSF of BD [4]

↑ In CSF of MS [5]

IL-6 Cytokine Proinflammatory Proinflammatory ↑ Blood ↑ In plasma of MS [6]

→ CSF ↑ In peripheral blood of PD [7]

↑ In serum of PRES [8]

IL-1β Cytokine Proinflammatory Proinflammatory → CSF ↑ In peripheral blood of PD [7]

↑ Blood

IL-10 Cytokine Anti-inflammatory Unclear ↑ CSF ↑ In peripheral blood of PD [7]

→ Blood ↑ In serum of PRES [8]

sTREM2 Receptor – Neuroprotection ↑ CSF

α1-ACT Protein Proinflammatory ↑ CSF

↑ Blood

IL-8 Chemokine Proinflammatory Proinjury → CSF ↑ In serum of RRMS [9]

↓ Blood

MCP-1 Chemokine Proinflammatory Neurodegeneration ↑ CSF ↑ In serum of RRMS [9]

→ Blood

VILIP-1 Protein – Neurodegeneration ↑ CSF

sTNFR Receptor Proinflammatory – ↑ Blood

sCD40L Protein Proinflammatory – ↑ Blood ↑ In peripheral blood of subarachnoid haemorrhage [10,11]

Prothrombotic

hsCRP Protein Proinflammatory – ↑ Blood ↑ Associated with unfavourable clinical outcomes after acute ischaemic stroke [12]

TGF-β1 Transforming growth factor

Neurotrophic Neuroprotective ↑ CSF

Regulation of these markers in AD demonstrated the results in this meta-analysis. Positive findings of these markers in relation to other neurological disorders are presented in the table. Relative references are provided in online supplementary material 8. Graphic symbols: ↑, increased; →, no changes; ↓, decreased.α1-ACT, alpha1-antichymotrypsin; AD, Alzheimer’s disease; BD, bipolar disorder; CNS, central nervous system; CSF, cerebrospinal fluid; IL, interleukin; MCP-1, monocyte chemoattractant protein-1; MS, multiple sclerosis; NGF, nerve growth factor; PD, Parkinson's disease; PRES, posterior reversible encephalopathy; RRMS, relapsing-remitting MS; TGF-β1, transforming growth factor-beta; VILIP-1, visinin-like protein-1; hsCRP, high-sensitivity C reactive protein; sCD40L, soluble CD40 ligand; sTNFR, soluble tumour necrosis factor receptor; sTREM2, soluble triggering receptor expressed on myeloid cells2.

with AD were linked to descending reparative mechanism in the central nervous system (CNS).7 The proinflammatory cytokine TNF-α and its receptors sTNFR1 and sTNFR2, which regulate numerous physiological processes in CNS, have been detected to exacerbate the main pathological changes of AD (both Aβ and tau pathologies) in vivo.25 Our meta-analyses proved notable vari-ances in the peripheral concentrations of sTNFR1 and sTNFR2 between group comparisons, implying that these markers may be useful in helping to monitor disease progression.

Interestingly, elevated MCP-1 levels were found in the periph-eral blood of patients with MCI and in the CSF of patients with AD compared with controls. Accumulating evidence has revealed that inflammatory responses mediated by MCP-1 were linked to Aβ pathology and microglia accumulations,26 and that peripheral MCP-1 may attract the blood-derived monocytes migrating to CNS.27 Meta-analysed results also demonstrated remarkably increased α1-ACT levels in both the peripheral and CSF of AD compared with the control, which were both in tight associations with dementia progression.28 29 Upregulated in AD, α1-ACT could induce abnormal tau phosphorylation in neurons that leads to neurite loss and eventually apoptosis.29 These provided opportunities to explore the promising therapeutic value of these markers in AD.

Anti-inflammatory strategies are regarded to be beneficial for AD, while our results illustrated elevated CSF levels of anti-in-flammatory marker IL-10 in patients with AD compared with controls. In a transgenic mouse model, researchers also observed unexpected negative effects of IL-10 on Aβ proteostasis, which induced growing plaque burden and consequently cognitive decline.30 Besides, in our results, peripheral sCD40L and CSF NGF concentrations were found to be remarkably elevated in patients with AD compared with controls. Released by activated platelets and T-lymphocytes, sCD40L could interact with CD40 from activated vascular cells and platelets, inducing prothrom-botic and proinflammatory responses.31 This interaction produces angiogenesis-related factors such as the vascular endo-thelial growth factor, correlating with the BBB function which plays a critical role in Aβ homeostasis.31 32 Serving as a neuro-trophic factor, NGF contributes to the neuronal survival, growth and function maintenance, and its deficiency leads to increasing production of Aβ aggregates and neuron death.33

Most of the studies have consistently demonstrated obvi-ously increased CSF levels of YKL-40, VILIP-1 and sTREM2 both in patients with AD and MCI compared with normal controls, and revealed their associations with neurodegeneration biomarkers, such as total tau and phosphorylated tau.34–36 As

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from

597shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

Cognitive neurology

a calcium-mediated neuronal injury protein, VILIP-1 showed higher levels in patients with AD and progressive MCI and was related to accelerating hippocampal atrophy and cognitive decline.36 sTREM2, the soluble form of innate immune receptor, could trigger microglial to express proinflammatory cytokines in the brain.37 It is reported that CSF YKL-40 level was already increased during the prodromal and preclinical phases of AD and could predict conversion from MCI to AD as measured by MMSE scores during the follow-up.38 Collectively, this evidence supported that these proteins could serve as a promising candi-date marker providing fair discrimination in the progression of AD.35 39

limitations and future directionsNoticeable heterogeneity exists in most of the analyses. Meta-re-gression analyses showed minor correlations of CSF YKL-40 alteration with age and cognition levels, yet these correlations were not found in the comparisons of AD with the controls. Some inflammatory markers, hsCRP, NGF and TGF-β1, showed differences in concentrations between the AD and the control groups, but failed to preferably discriminate by ratios. There are also some markers that appear to be positive in individual studies but negative in comprehensive analyses. The foremost reason is lack of quantity in the concerned study and deficiency in research objectives, perspectives and content. Limited by incomplete information of the included studies, it is underpowered to inves-tigate the underlying mechanism of these inflammatory markers in AD and potential correlations with other clinical moderators (ie, education year, ethnicity, medication status, disease dura-tion, APOE gene carriage and level of mental health). Thus, future and modern studies better cover these aspects. Identifying inflammatory biomarkers in AD and MCI holds the promise of removing roadblocks to the therapeutic discovery.40 Further studies are needed to investigate how these peripheral or CSF markers correlate with the established markers such as Aβ and tau. Also, more longitudinal research establishing multiperiod data of inflammatory markers along with the widely acknowl-edged neurodegeneration biomarkers are critical for insights into inflammation in the progression of AD.

COnClusIOnOur meta-analysis demonstrated remarkable alterations in the peripheral levels of IL-1β, IL-6, sTNFR1, sTNFR2, α1-ACT, sCD40L, hsCRP, IL-8 and MCP-1, and CSF concentrations of MCP-1, sTREM2, YKL-40, α1-ACT, NGF, VILIP-1 and IL-10, in pairwise comparisons of AD, MCI and control groups. These findings strengthened the evidence that AD and MCI are accom-panied by systemic and CNS-derived inflammatory processes. For subsequent detection and supervision of MCI stage and progression to AD, these findings need to be validated in larger, multicentred cohort studies.

Author affiliations1Department of Neurology, Qingdao Municipal hospital, Qingdao University, Qingdao, china2Department of Neurology and center for clinical Neuroscience, Daping hospital, Third Military Medical University, chongqing, china3clinical Research center, Qingdao Municipal hospital, Qingdao, china4chinese academy of sciences Key Laboratory of Brain Function and Disease and school of Life sciences, University of science and Technology of china, hefei, china5Genetics and aging Research Unit, MassGeneral Institute for Neurodegenerative Disease (MIND), Department of Neurology, Massachusetts General hospital, harvard Medical school, charlestown, Massachusetts, Usa6Department of Neurology and Institute of Neurology, huashan hospital, shanghai Medical college, Fudan University, shanghai, china

Contributors J-TY conceptualised and designed the study. X-Ns and L-DN conducted the study. X-Ns, L-DN and J-TY analysed and extracted the data. X-Ns, L-DN and J-TY wrote the first draft of the manuscript. all authors reviewed the manuscript.

Funding This work was supported by grants from the National Key R&D program of china (2018YFc1314700), the National Natural science Foundation of china (81471309, 81571245, 81501103 and 81701253), Taishan scholars program of shandong province (ts201511109, tsqn20161078 and tsqn20161079), Qingdao Key health Discipline Development Fund, and shandong provincial collaborative Innovation center for Neurodegenerative Disorders.

Competing interests None declared.

Patient consent Not required.

Provenance and peer review Not commissioned; externally peer reviewed.

RefeRences 1. petersen Rc. Mild cognitive impairment as a diagnostic entity. J Intern Med

2004;256:183–94. 2. Winblad B, palmer K, Kivipelto M, et al. Mild cognitive impairment--beyond

controversies, towards a consensus: report of the International Working Group on Mild cognitive Impairment. J Intern Med 2004;256:240–6.

3. petersen Rc, Lopez O, armstrong MJ, et al. practice guideline update summary: Mild cognitive impairment: Report of the Guideline Development, Dissemination, and Implementation subcommittee of the american academy of Neurology. Neurology 2018;90:126–35.

4. heneka MT, carson MJ, el Khoury J, et al. Neuroinflammation in alzheimer’s disease. Lancet Neurol 2015;14:388–405.

5. Darweesh sKL, Wolters FJ, Ikram Ma, et al. Inflammatory markers and the risk of dementia and alzheimer’s disease: a meta-analysis. Alzheimers Dement 2018;14:1450–9.

6. Olsson B, Lautner R, andreasson U, et al. csF and blood biomarkers for the diagnosis of alzheimer’s disease: a systematic review and meta-analysis. Lancet Neurol 2016;15:673–84.

7. hesse R, Wahler a, Gummert p, et al. Decreased IL-8 levels in csF and serum of aD patients and negative correlation of MMse and IL-1β. BMC Neurol 2016;16:185.

8. hampel h, Toschi N, Baldacci F, et al. alzheimer’s disease biomarker-guided diagnostic workflow using the added value of six combined cerebrospinal fluid candidates: aβ1-

42, total-tau, phosphorylated-tau, NFL, neurogranin, and YKL-40. Alzheimers Dement 2018;14:492–501.

9. Zhu Y, chai YL, hilal s. serum IL-8 is a marker of white-matter hyperintensities in patients with alzheimer’s disease. Alzheimers Dement 2017;7:41–7.

10. O’Bryant se, Lista s, Rissman Ra. comparing biological markers of alzheimer’s disease across blood fraction and platforms: comparing apples to oranges. Alzheimers Dement 2015;3:27–34.

11. Balldin Vh, hall JR, Barber Rc, et al. The Relation between Inflammation and Neuropsychological Test performance. Int J Alzheimers Dis 2012;2012:1–6.

12. Kim sM, song J, Kim s, et al. Identification of peripheral inflammatory markers between normal control and alzheimer’s disease. BMC Neurol 2011;11:51.

13. cestari Ja, Fabri GM, Kalil J, et al. Oral Infections and cytokine Levels in patients with alzheimer’s Disease and Mild cognitive Impairment compared with controls. J Alzheimers Dis 2016;52:1479–85.

14. swardfager W, Lanctôt K, Rothenburg L, et al. a meta-analysis of cytokines in alzheimer’s disease. Biol Psychiatry 2010;68:930–41.

15. Qin XY, cao c, cawley NX, et al. Decreased peripheral brain-derived neurotrophic factor levels in alzheimer’s disease: a meta-analysis study (N=7277). Mol Psychiatry 2017;22:312–20.

16. hozo sp, Djulbegovic B, hozo I. estimating the mean and variance from the median, range, and the size of a sample. BMC Med Res Methodol 2005;5:13.

17. Wan X, Wang W, Liu J, et al. estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol 2014;14:135.

18. Grissom RJ, Kim JJ. Effect sizes for research: A broad practical approach. Mahwah, NJ, Us: Lawrence erlbaum associates publishers, 2005.

19. Zabel M, schrag M, Mueller c, et al. assessing candidate serum biomarkers for alzheimer’s disease: a longitudinal study. J Alzheimers Dis 2012;30:311–21.

20. Llano Da, Li J, Waring JF, et al. cerebrospinal fluid cytokine dynamics differ between alzheimer disease patients and elderly controls. Alzheimer Dis Assoc Disord 2012;26:322–8.

21. Ott BR, Jones RN, Daiello La. Blood-cerebrospinal Fluid Barrier Gradients in Mild cognitive Impairment and alzheimer’s Disease: Relationship to Inflammatory cytokines and chemokines. Nat Commun 2018;10:245.

22. Teixeira aL, Reis hJ, coelho FM, et al. all-or-nothing type biphasic cytokine production of human lymphocytes after exposure to alzheimer’s beta-amyloid peptide. Biol Psychiatry 2008;64:891–5.

23. Uslu s, akarkarasu Ze, Ozbabalik D, et al. Levels of amyloid beta-42, interleukin-6 and tumor necrosis factor-alpha in alzheimer’s disease and vascular dementia. Neurochem Res 2012;37:1554–9.

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from

598 shen X-N, et al. J Neurol Neurosurg Psychiatry 2019;90:590–598. doi:10.1136/jnnp-2018-319148

Cognitive neurology

24. Lai Ksp, Liu cs, Rau a, et al. peripheral inflammatory markers in alzheimer’s disease: a systematic review and meta-analysis of 175 studies. J Neurol Neurosurg Psychiatry 2017;88:876–82.

25. Decourt B, Lahiri DK, sabbagh MN. Targeting Tumor Necrosis Factor alpha for alzheimer’s Disease. Curr Alzheimer Res 2017;14:412–25.

26. Kiyota T, Gendelman he, Weir Ra, et al. ccL2 affects β-amyloidosis and progressive neurocognitive dysfunction in a mouse model of alzheimer’s disease. Neurobiol Aging 2013;34:1060–8.

27. Lee WJ, Liao Yc, Wang YF, et al. plasma Mcp-1 and cognitive Decline in patients with alzheimer’s Disease and Mild cognitive Impairment: a Two-year Follow-up study. Sci Rep 2018;8:1280.

28. DeKosky sT, Ikonomovic MD, Wang X, et al. plasma and cerebrospinal fluid alpha1-antichymotrypsin levels in alzheimer’s disease: correlation with cognitive impairment. Ann Neurol 2003;53:81–90.

29. padmanabhan J, Levy M, Dickson DW, et al. alpha1-antichymotrypsin, an inflammatory protein overexpressed in alzheimer’s disease brain, induces tau phosphorylation in neurons. Brain 2006;129:3020–34.

30. chakrabarty p, Li a, ceballos-Diaz c, et al. IL-10 alters immunoproteostasis in app mice, increasing plaque burden and worsening cognitive behavior. Neuron 2015;85:519–33.

31. Yu s, Liu Y-p, Liu Y-h, et al. Diagnostic utility of VeGF and soluble cD40L levels in serum of alzheimer’s patients. Clinica Chimica Acta 2016;453:154–9.

32. ait-ghezala G, abdullah L, Volmar ch, et al. Diagnostic utility of apOe, soluble cD40, cD40L, and abeta1-40 levels in plasma in alzheimer’s disease. Cytokine 2008;44:283–7.

33. Matrone c, ciotti MT, Mercanti D, et al. NGF and BDNF signaling control amyloidogenic route and a production in hippocampal neurons. Proceedings of the National Academy of Sciences 2008;105:13139–44.

34. alcolea D, Martínez-Lage p, sánchez-Juan p, et al. amyloid precursor protein metabolism and inflammation markers in preclinical alzheimer disease. Neurology 2015;85:626–33.

35. Baldacci F, Toschi N, Lista s, et al. Two-level diagnostic classification using cerebrospinal fluid YKL-40 in alzheimer’s disease. Alzheimers Dement 2017;13:993–1003.

36. Zhang h, Ng Kp, Therriault J, et al. cerebrospinal fluid phosphorylated tau, visinin-like protein-1, and chitinase-3-like protein 1 in mild cognitive impairment and alzheimer’s disease. Transl Neurodegener 2018;7:23.

37. Zhong L, chen X-F, Wang T, et al. soluble TReM2 induces inflammatory responses and enhances microglial survival. J Exp Med 2017;214:597–607.

38. Olsson B, hertze J, Lautner R, et al. Microglial markers are elevated in the prodromal phase of alzheimer’s disease and vascular dementia. J Alzheimers Dis 2013;33:45–53.

39. piccio L, Deming Y, Del-Águila JL, et al. cerebrospinal fluid soluble TReM2 is higher in alzheimer disease and associated with mutation status. Acta Neuropathol 2016;131:925–33.

40. Nakamura a, Kaneko N, Villemagne VL, et al. high performance plasma amyloid-β biomarkers for alzheimer’s disease. Nature 2018;554:249–54.

copyright. on D

ecember 2, 2021 by guest. P

rotected byhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp-2018-319148 on 10 January 2019. D

ownloaded from