Supplemental material: Astrocyte biomarkers in Alzheimer’s disease: a systematic review and meta-analysis
Bellaver, Bruna et al. (2022), Supplemental material: Astrocyte biomarkers in Alzheimer’s disease: a systematic review and meta-analysis, Dryad, Dataset, https://doi.org/10.5061/dryad.5hqbzkh57
Objective: To perform a systematic review and meta-analysis to determine whether fluid and imaging astrocyte biomarkers are altered in Alzheimer's disease (AD).
Methods: PubMed and Web of Science databases were searched for articles reporting fluid or imaging astrocyte biomarkers in AD. Pooled effect sizes were determined with mean differences (SMD) using the Hedge’s G method with random-effects to determine biomarker performance. Adapted questions from QUADAS-2 were applied for quality assessment. A protocol for this study has been previously registered in PROSPERO (registration number: CRD42020192304).
Results: The initial search identified 1,425 articles. After exclusion criteria were applied, 33 articles (a total of 3,204 individuals) measuring levels of GFAP, S100B, YKL-40 and AQP4 in the blood and cerebrospinal fluid (CSF), as well as MAO-B, indexed by positron emission tomography 11C-deuterium-L-deprenyl ([11C]-DED), were included. GFAP (SMD = 0.94; 95% CI = 0.71-1.18) and YKL-40 (SMD = 0.76; CI 95% = 0.63-0.89) levels in the CSF, S100B levels in the blood (SMD = 2.91; CI 95% = 1.01-4.8) were found significantly increased in AD patients.
Conclusions: Despite significant progress, applications of astrocyte biomarkers in AD remain in their early days. The meta-analysis demonstrated that astrocyte biomarkers are consistently altered in AD and supports further investigation for their inclusion in the AD clinical research framework for observational and interventional studies.
Two databases were searched: PubMed and Web of Science. Complete search terms are in Supplemental File 1. No language, study design restrictions or date of publication limit were applied. The search was conducted by one author (BB) in November 17th, 2019. The list of included studies was screened (BB, JPFS and LUDR) in order to identify additional articles for inclusion. An active search was performed by BB in May 2020 in a meta-analysis of CSF and blood biomarkers in AD. Briefly, this meta-analysis evaluated only two astrocyte biomarkers: GFAP (2 cohorts) and YKL-40 (6 cohorts). We also conducted additional search in book chapters related to the theme. No grey literature (i.e. preprint databases) were searched. The authors were contacted in case of absent data or questions about data extraction. If there was no reply after two contacts occurring at least 10 days apart, the study was not included in the meta-analysis.
Data screening, inclusion and exclusion criteria
All articles identified in our search were downloaded into an online software program, Rayyan QCRI. Study inclusion and exclusion were performed with a pre-screening based on title and abstract. If there were no evident exclusion criteria observed (i.e. reviews or editorials, studies not related to AD or conducted using in vitro, animal models or post-mortem analysis), a full-text analysis was performed. Both were conducted by two authors, independently, and blinded to each other's decisions (JPFS and LUDR). Any disagreement was discussed and resolved with two other authors (BB and DTL). All studies reporting astrocyte biomarkers in the blood (serum/plasma), cerebrospinal fluid (CSF) or brain imaging of AD versus CU individuals were included. Astrocyte biomarkers were selected based on a recent review update by Carter and colleagues The following exclusion criteria were applied: studies with less than 10 participants, control group with inflammatory conditions, or with neurologic or psychiatric diagnosis and biomarkers measured by non-quantitative methods. Only studies presenting data as mean and standard deviation (SD) or mean and standard error of the mean (SEM) and using established criteria for AD diagnosis, including clinically defined (NINCDS-ADRDA) and biomarker-defined (NIAA-AA and IWG) criteria, were selected for the meta-analysis. Data were included from cross-sectional or baseline measurements in longitudinal studies.
Data extraction from included articles was performed independently by two authors (JPFS and LUDR). The data extracted were: sample size, gender (percentage of male in each group), age, AD diagnostic criteria, method, imaging or fluid, biomarker mean and SD or SEM. To avoid erroneous data collection, BB checked all the data extracted searching for discrepancies. When data were reported only in graphs, a digital ruler was used to estimate the values from graphs, as previously described. If both methods were not possible, the authors were contacted. Additionally, when two different AD cohorts shared the same CU control group, the number of individuals of the control group was divided by the number of comparisons and rounded down, in order to avoid overestimation of the effect. Genetic background was not considered in this analysis since the majority of articles did not provide genetic information. In particular, out of 33 articles analysed, 8 provided APOEε4 status and only 1 cohort described familial autosomal-dominant AD mutations. Due to the paucity of genetic information, cohorts were classified as early-onset AD (EOAD, patients < 65 years old) and late-onset AD (LOAD, patients > 65 years old). A total of 11 studies did not stratify patients by age, and therefore were identified in this meta-analysis as mixed cohorts.
Risk of bias was assessed by three authors independently (BB, JPFS and LUDR) following the revised Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2), a tool recommended for use in systematic reviews by the Agency for Healthcare Research and Quality, Cochrane Collaboration, and the U.K. National Institute for Health and Clinical Excellence. The 12 questions adapted from QUADAS-2 and applied in this study are listed in Supplemental File 1. Only studies with biomarker-supported diagnosis were considered as low risk of bias in its assessment section. In case of control groups with different ages, only the age-matched group was used in the meta-analysis. If the overall age of controls did not match with AD groups, the study was considered with a high-risk of bias in its specific section of the assessment. Publication bias was tested by visual inspection of funnel plots and by the Egger’s regression test obtained using Stata version 14.
Studies were grouped according to the astrocyte biomarker used - for imaging biomarkers, an additional subdivision was used based on brain regions - and a meta-analysis was conducted for each of them. In order to conduct a meta-analysis, a minimum of 2 cohorts were necessary. To consider intrinsic variability, Hedge´s G method with random-effect models were used to pool effect sizes in order to determine standardized mean differences (SMD) with 95% confidence interval (CI). The significance of pooled effect sizes was measured using the Z-test. False discovery rate (FDR), using a q value of 5%, was applied in order to correct for multiple comparisons and results were considered to be statistically significant if the corrected p-value was < 0.05. Individual study weights were estimated using the inverse of the variance. The heterogeneity among studies was identified using the Chi2 and quantified by I2 [I2 = (Q - df/Q) x 100, where Q = Chi2 test results]. A p-value ≤ 0.1 was considered significant for the Chi2 test. I2 values of 25%, 50%, and 75%, represent low, moderate, and high heterogeneity, respectively. SMD and heterogeneity analysis were obtained using Stata version 14.
Sensitivity analysis was performed to identify whether any specific article or group of articles included in this meta-analysis, in addition to any main methodological decision, might have significantly skewed the analyses performed. For that, the following tests were implemented: (1) the jackknife method, in order to determine the influence of each article on the SMD and heterogeneity; (2) exclusion of studies exhibiting a concerning risk of bias, defined as less than 8 categories presenting low risk of bias in the quality assessment; (3) exclusion of cohorts composed only by EOAD patients and (4) stratifying cohorts by the AD diagnosis used in the study. Sensitivity analyses were performed for biomarkers presenting a minimum of 3 studies.
Conselho Nacional de Desenvolvimento Científico e Tecnológico, Award: 460172/2014-0
Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul, Award: 16/2551-0000475-7
Instituto Nacional de Ciência e Tecnologia para Excitotoxicidade e Neuroproteção, Award: 465671/2014-4
Instituto Serrapilheira, Award: Serra-1912-31365