Two investigators (WR and ZW) conducted independent searches in the PubMed, Medline, Web of Science, and the Cochrane Library databases for articles published up to 1st May 2025. The main keywords included (YKL40 OR CHI3L1) AND (Alzheimer’s Disease OR Senile Dementia OR AD OR Mild Cognitive Impairment OR MCI OR preclinical AD OR pre-AD) (Supplementary Table S1).

The inclusion criteria were: (1) studies assessing CSF or blood (plasma or serum) YKL40 in AD patients, MCI patients, pre-AD, and HCs. (2) Utilization of international diagnostic criteria for AD, MCI, and pre-AD was reported. (3) Availability of YKL40 data for both disease and control groups. The exclusion criteria included (1) studies lacking precise YKL40 levels, even after contact with the corresponding author; and (2) reviews, abstracts, case reports, letters, and commentaries.

A dual independent review process was implemented following a predefined protocol, wherein two trained investigators systematically identified and screened eligible studies. Data extraction was performed using standardized templates capturing nine critical parameters: the first author’s name, publication year, country of origin, sample size, diagnostic criteria implementation, mean age, gender composition, YKL-40 biomarker concentrations in either CSF or peripheral blood, and assay method. For studies reporting range-based data, parametric approximation methods were applied to derive means and standard deviations (SD) in accordance with Cochrane Handbook guidelines. All extracted variables were systematically archived in standardized spreadsheets to ensure data integrity.

The methodological quality of observational studies was rigorously assessed using the Newcastle-Ottawa Scale (NOS), a validated tool comprising three methodological domains across eight specific items, with a maximum attainable score of 9. Studies achieving ≥7 points were classified as high-quality based on established thresholds (16). Dual independent evaluation was conducted by two trained assessors (WR and ZW) using standardized NOS protocols, with discrepancies resolved through consensus-building consultation with a third senior researcher (PJ).

All statistical analyses were performed using STATA version 18.0 (StataCorp LLC, College Station, TX, United States). The pooled standardized mean difference (SMD) with corresponding 95% confidence intervals (CI) was calculated to compare YKL-40 levels between AD-related cognitive impairment patients and cognitively intact healthy controls. Interstudy heterogeneity was quantified through Higgins I² statistics derived from Cochrane’s Q-test. A random-effects model was employed when I² ≥ 50%, indicating substantial heterogeneity, whereas a fixed-effects model was adopted for lower heterogeneity levels. To investigate potential sources of heterogeneity, predefined subgroup analyses and meta-regression were conducted focusing on three covariates: mean age, diagnostic criteria implementation, and female participant proportion. Sensitivity analysis was performed through iterative exclusion of individual studies to verify the robustness. Publication bias was assessed using Egger’s regression test complemented by funnel plot asymmetry evaluation. Statistical significance threshold was set at p < 0.05 (two-tailed).

Following a systematic database search using predefined search terms, 1,195 potentially relevant articles were initially identified, including 256 in PubMed, 242 in Medline, 43 in Cochrane, and 444 in Web of Science. After rigorous removal of duplicate records through automated and manual verification processes, 362 unique articles remained eligible for evaluation. Subsequent preliminary screening based on titles and abstracts excluded 257 publications that failed to meet thematic relevance criteria, resulting in 105 articles progressing to full-text assessment. After reviewing full-text context of the remaining 105 articles, 75 articles were excluded as follows: 3 articles were removed due to absence of international diagnostic criteria, 33 studies were excluded for inconsistent grouping, 4 articles demonstrated skewed data distribution, 17 studies were eliminated for incomplete datasets, and 18 publications were deemed irrelevant to the research theme. After rigorous screening, 30 articles met all inclusion criteria and were subsequently incorporated into the systematic analysis, which is presented in Figure 1.

The thirty articles (5, 17–45) were all published between 2011 and 2025, with a concentration of publications occurring from 2019 to 2023. The 30 included studies involving 5,815 participants stratified into distinct diagnostic categories: 2,102 AD patients, 1,504 MCI cases, 118 pre-AD individuals, and 2,091 HCs. Biomarker quantification revealed a differential analytical focus across studies, with 8 investigations reporting YKL-40 levels in peripheral blood specimens and 24 studies providing CSF measurements. Regarding methodological approaches, enzyme-linked immunosorbent assay (ELISA) was employed as the primary detection method in 28 studies, while the remaining two studies utilized the Elecsys electrochemiluminescence immunoassay and Simple-Plex microfluidic immunoassay system, respectively. Quality assessment using the NOS revealed scores ranging from 6 to 8, with 28 studies meeting high-quality criteria and 2 studies classified as moderate quality (Supplementary Tables S2, S3).

Sensitivity analysis employing leave-one-out methodology demonstrated negligible fluctuations in pooled effect estimates, confirming the robustness of primary outcomes. Publication bias was systematically assessed across all groups. Egger’s regression test revealed significant publication bias in CSF YKL-40 comparisons between AD patients and HCs (p = 0.017), while no statistically significant bias risk was detected in other subgroups (all p > 0.05), which were corroborated through the funnel plot test (Figure 8).

Previous studies have demonstrated that incorporating CSF inflammatory biomarkers into the AD biomarker panel enhances diagnostic accuracy for clinical staging at baseline and improves the predictive value of AD biomarkers for cognitive decline (46), highlighting the critical role of sensitive inflammatory biomarkers in disease prognostication. YKL-40, a potential biomarker of neuroinflammation and astrocyte activation, exhibits significant disease-promoting effects during the early pathological progression of AD (47). The current study demonstrated statistically significant elevations in CSF YKL-40 levels in AD, MCI, and pre-AD groups compared with HCs (all p < 0.05), with peripheral blood YKL-40 levels also showing significant differences in AD and MCI groups (all p < 0.05). Notably, while CSF YKL-40 levels in AD patients showed an increasing trend compared to MCI patients, this difference did not reach statistical significance (p = 0.53).

These findings are consistent with and extend previous meta-analyses. Earlier studies consistently reported elevated CSF YKL-40 levels in AD, but showed inconsistent or no significant differences in peripheral blood (11). Other meta-analyses, often based on a limited number of studies, reported higher CSF YKL-40 levels in AD than in MCI, with no significant difference between MCI and controls (7). By synthesizing more recent evidence, our updated analysis suggests that YKL-40 elevations may already be present at the MCI and even preclinical stages, thereby observed differences between MCI and AD in earlier reports. Emerging evidence further suggests a stage-dependent discriminative value of YKL-40, particularly in early MCI (48). Despite heterogeneity across studies regarding YKL-40’s efficacy in distinguishing AD stages, our findings support its potential as a predictive biomarker for AD progression.

In subgroup analyses of CSF YKL-40 levels stratified by age, diagnostic criteria, and female proportion, we observed no significant differences in CSF YKL-40 levels between MCI patients aged >70 years and either AD patients or HCs. Comparisons between pre-AD and HCs within the NINCDS-ADRDA diagnostic criteria subgroup and the subgroup with <50% female participants also showed non-significant differences, likely attributable to limited sample sizes. Crucially, age and sex emerge as critical modulators of YKL-40 levels: cumulative evidence demonstrates age-dependent elevation of CSF YKL-40 that correlates with aging progression (49, 50)^. A meta-analysis revealed associations between female proportion and YKL-40 levels in AD/MCI cohorts (51). Mechanistic studies suggest sex-specific interactions, with females exhibiting more pronounced p-tau181 elevation than males at equivalent YKL-40 levels (52). Gene expression profiling further indicates heightened YKL-40 transcription in female AD brains compared to males (53). However, conflicting data exist regarding peripheral blood YKL-40, with a study showing plasma YKL-40 levels independent of age and sex (54).

For peripheral blood YKL-40 analyses, stratification by blood sample type was performed in addition to age, diagnostic criteria, and sex factors. The MCI vs. HC comparison maintained stable age/sex distributions, precluding additional stratification. In AD vs. HC CSF comparisons, the subgroup with <50% female participants showed non-significant differences, potentially reflecting inadequate power. Notably, blood sample stratification revealed distinct patterns: serum YKL-40 showed no group-wise differences in AD vs. HC or MCI vs. HC comparisons, whereas plasma YKL-40 effectively discriminated both AD from HC and MCI from HC. This aligns with previous findings that serum YKL-40 correlates with CSF levels and structural MRI parameters but not with cognitive performance or clinical staging (55). It should be emphasized that the limited sample size of blood YKL-40 studies in our analysis may introduce measurement variability.

Increasing evidence identifies neuroinflammation as a key hallmark of AD pathophysiology. Microglia and astrocytes, acting as the principal cellular mediators of these inflammatory processes, are instrumental in the progression of AD throughout both the preclinical and symptomatic stages (56). Studies on human brain pathology confirm that YKL-40 is synthesized by a subset of astrocytes in AD patients (57). Its colocalization with tau immunoreactivity suggests that YKL-40 levels directly reflect the intensity of the AD pathological cascade, providing a biological basis for its use as a progression biomarker.

YKL-40 is a protein predominantly expressed in astrocytes and encoded by the CHI3L1 gene. CHI3L1 is primarily expressed by astrocytes in the mouse brain, with elevated expression detected in reactive astrocytes in advanced AD, accompanied by increased IBA1-positive microglial activation (58). Emerging evidence demonstrates that astrocyte-specific YKL-40 knockout in symptomatic 5xFAD mice with cognitive deficits effectively decreases amyloid plaque deposition in multiple brain regions, along with reduced glial activation, less neuronal loss, and improved memory function (59–61). These findings are further supported by in vitro evidence, which reveals that pro-inflammatory cytokines such as IL-1, IL-6, and TNF-α, alongside M1-type macrophage secretomes, upregulate CHI3L1 expression in astrocytes (62). These molecular findings confirm the pivotal role of reactive astrocytes in the neuroinflammatory cascade associated with AD, positioning YKL-40 as a key biomarker for reactive astrocytes. They further reinforce the potential application value of YKL-40 as a biomarker for monitoring disease progression in AD.

CSF can effectively reflect pathological changes in brain tissue due to its direct contact. Although CSF is considered an optimal measure for AD, its acquisition requires invasive lumbar puncture, which underlies the urgent need for less invasive diagnostic techniques (63). In recent years, increasing attention has been directed toward blood-based biomarkers as accessible and scalable tools for assessing AD–related pathology. Compared with CSF sampling, plasma biomarkers offer clear practical advantages, including lower invasiveness, improved feasibility for large-scale screening, and suitability for longitudinal monitoring (64). Notably, an umbrella reanalysis of 106 meta-analyses encompassing 52 candidate biomarkers identified blood YKL-40 as the biomarker achieving Class I (convincing) evidence for association with AD (65). Existing evidence confirms that plasma YKL-40 correlates with CSF YKL-40 levels, with correlation coefficients ranging from low (r = 0.24) to moderate (r = 0.40) (66, 67). These findings support the potential of plasma YKL-40 as a peripheral indicator of neuroinflammatory activity related to AD progression, while highlighting its value as a more accessible biomarker for population-based and longitudinal studies.

Specifically, our subgroup analysis demonstrated that plasma YKL-40, rather than serum, serves as a more robust indicator for differentiating AD and MCI from healthy controls. This finding not only aligns with the clinical shift toward non-invasive diagnostics but also provides practical evidence for optimizing sample selection in future clinical applications. Admittedly, individual heterogeneity caused by factors such as age, sex, and comorbidities can affect disease development, and plasma YKL-40 is susceptible to these confounding factors, which may compromise diagnostic specificity when used alone. Therefore, plasma YKL-40 should be considered an adjunctive biomarker rather than a standalone diagnostic tool, and its utility may be enhanced when integrated into multimodal frameworks (e.g., combined with plasma p-tau or MRI parameters) to mitigate the impact of systemic confounders.

Accumulating evidence indicates that CSF YKL-40 levels are positively associated with tau pathology, while showing no significant correlation with amyloid-β (Aβ) burden, demonstrating strong associations with tau pathophysiology even during preclinical AD stages (68). Previous studies have reported that YKL-40 is released into the CSF at a relatively later stage of the AD pathological cascade and may specifically mediate the link between Aβ-induced downstream tau phosphorylation and tau-related neurodegeneration (47). Neuroimaging-pathology correlation studies further reveal spatially specific positive relationships between CSF YKL-40 levels and tau accumulation in AD-vulnerable regions (e.g., hippocampus), independent of Aβ deposition (68). Recent studies have further demonstrated consistent findings showing significant correlations between CSF YKL-40 and both phosphorylated tau (p-tau) and total tau (t-tau), with particularly strong associations observed in A + MCI and T + subgroups (69). Evidence shows that YKL-40 may be involved in the non-amyloid-dependent pathway. Elevated levels of YKL-40 in the cerebrospinal fluid were observed in A + T + and A-T + individuals, whereas such elevation was not detected in A + T- individuals (47, 68, 70, 71). Intriguingly, a structural equation model demonstrates that Aβ fibril aggregates directly trigger astrocytic (47). Quantitative Aβ-PET studies reveal that higher CSF YKL-40 levels correlate with greater global Aβ burden and reduced hippocampal volume, findings replicated across another cohort (52, 72). Crucially, mediation analysis demonstrates these phenotypic associations are fully mediated by p-tau levels.

This study has several methodological limitations: First, although the overall sample size was substantial, the number of studies and participants in certain subgroups, particularly those involving preclinical AD and peripheral blood measurements, remains limited. Thereby constraining the evaluation of YKL-40’s discriminative capacity during the ultra-early transitional stages of AD. Second, key confounders, including comorbid inflammatory conditions, APOE genotypes, and Aβ pathological status, were inadequately documented in most original studies, potentially introducing residual confounding bias. Notably, significant geographic constraints exist—most of the participants originated from Western populations—a sampling bias that may jeopardize the generalizability of conclusions. These limitations underscore the need for future large-scale, longitudinal studies with standardized biomarker measurements to further validate the role of YKL-40 across the Alzheimer’s disease continuum.