About AlzBiomarker
Methods
This database contains summary-level data extracted from primary peer-reviewed publications. To identify eligible publications we systematically searched PubMed and the Web of Science database. Search results were evaluated according to predetermined inclusion criteria. Specialist readers extracted data from eligible articles and an independent reader verified the accuracy of the curated data. To facilitate meta-analysis, we attempted to standardize terms pertaining to biomarker names, biomarker quantification methods, and clinical conditions, see glossary.
Search Strategy
Version 1.0, December 2015
The following search terms were used to identify publications possibly containing eligible data:
("alzheimer disease"[MeSH Terms] OR ("alzheimer"[All Fields] AND "disease"[All Fields]) OR "alzheimer disease"[All Fields] OR "alzheimer"[All Fields]) OR ("alzheimer disease"[MeSH Terms] OR ("alzheimer"[All Fields] AND "disease"[All Fields]) OR "alzheimer disease"[All Fields] OR "alzheimer's"[All Fields])) AND (("biological markers"[MeSH Terms] OR ("biological"[All Fields] AND "markers"[All Fields]) OR "biological markers"[All Fields] OR "biomarker"[All Fields]) OR CSF[All Fields] OR ("plasma"[MeSH Terms] OR "plasma"[All Fields]) OR ("serum"[MeSH Terms] OR "serum"[All Fields]) OR cerebrospinal[All Fields]) NOT Review[ptyp] AND ("humans"[MeSH Terms] AND English[lang])
In addition, terms pertaining to a set of candidate biomarkers were included, including: Aβ42 (Aβ-42 OR Abeta42 OR “Abeta 42”), tau (T-tau OR TTau OR P-tau OR Ptau* OR P-tau*), NFL (NEFL OR NFL OR NF-L OR NF68 OR NFH OR NF-H OR NFM OR NEFM OR “NEUROFILAMENT PROTEIN LIGHT POLYPEPTIDE" OR “NEUROFILAMENT PROTEIN LIGHT CHAIN” OR “NEUROFILAMENT PROTEIN HEAVY POLYPEPTIDE” OR “heavy neurofilament subunit” OR “NEUROFILAMENT PROTEIN MEDIUM POLYPEPTIDE“ OR “neurofilament“), VLP-1 (VSNL1 OR VILIP OR VILIP1 OR VILIP-1 OR VLP-1 OR "VISININ-LIKE 1"), hFABP (HFABP OR FABP3 OR MDGI OR "FATTY ACID-BINDING PROTEIN 3" OR "FATTY ACID-BINDING PROTEIN MUSCLE TYPE" OR "MUSCLE TYPE FATTY ACID-BINDING PROTEIN" OR "MUSCLE TYPE FATTY ACID BINDING PROTEIN" OR " MAMMARY-DERIVED GROWTH INHIBITOR" OR "heart fatty acid binding protein"), NSE (ENO2 OR NSE OR “ENOLASE 2" OR “ENOLASE GAMMA” OR “NEURON-SPECIFIC ENOLASE”), GFAP (GFAP OR GFA OR "GLIAL FIBRILLARY ACIDIC PROTEIN" ), YKL-40 (CHI3L1 OR GP39 OR YKL40 OR YKL-40 OR "CHITINASE 3-LIKE 1" OR "CARTILAGE GLYCOPROTEIN 39" OR "CHONDROCYTE PROTEIN YKL40"), MCP-1 (CCL2 OR CCL-2 OR SCYA2 OR MCP1 OR MCP-1 OR MCAF OR "CHEMOKINE CC MOTIF LIGAND 2" OR “SMALL INDUCIBLE CYTOKINE A2” OR “MONOCYTE CHEMOTACTIC PROTEIN 1” OR “MONOCYTE CHEMOTACTIC AND ACTIVATING FACTOR”), albumin ratio (“albumin ratio”), Aβ38 (Aβ38 OR Aβ-38), Aβ40 (Aβ40 OR Aβ-40), and sAPPα and sAPPβ (APP OR APPα OR APPβ OR sAPPα OR sAPPβ OR AAA OR CVAP OR PN2 OR "AMYLOID BETA A4 PRECURSOR PROTEIN" OR "AMYLOID OF AGING AND ALZHEIMER DISEASE" OR "CEREBRAL VASCULAR AMYLOID PEPTIDE" OR "PROTEASE NEXIN II").
Version 1.1, April 2016
Search performed February 7, 2016 using the keywords neurogranin AND Alzheimer*. Search results were evaluated based on the same inclusion criteria as version 1.0 (described below), with the exception that articles were not restricted by publication date.
Version 1.2, June 2016
PubMed search performed April 30, 2016 using the keywords TREM2 AND Alzheimer*. Search results were evaluated based on the same inclusion criteria as version 1.1.
Version 2.0, April 2017
PubMed was searched covering the date range July 1, 2014 to May 31, 2016. The same keywords were included as in Version 1.0 along with neurogranin AND Alzheimer* and TREM2 AND Alzheimer*. Search results were evaluated based on the same inclusion criteria as version 1.1. Forty-two eligible studies were identified.
Version 2.1, June 2018
PubMed was searched covering the date range January 1, 1984 to September 30, 2017. The same keywords were included as in Version 2.0 along with alpha-synuclein. Results were cross referenced with papers that had been previously curated. Search results were evaluated based on the same inclusion criteria as version 1.1. Thirteen new eligible studies were identified for meta-analysis.
Version 3.0, July 2021
PubMed was searched covering the date range January 1, 1984 to April 15, 2020. The same keywords were included as in Version 2.1. along with the MeSH term “Amyloid beta-Peptides.”
Additionally, searches related to the candidate biomarkers in plasma or serum were extended to include papers indexed in PubMed through September 1, 2020. For each candidate biomarker, PubMed was searched using the following terms: (((((((((([synonyms for candidate biomarker listed above]) AND (plasma OR serum OR blood)) NOT "review"[Publication Type]) AND ("2016/01/01"[Date - Publication] : "2020/12/31"[Date - Publication])) AND biomarker) AND (Alzheimer*)))))) AND (("2020/04/15"[Date - Entry] : "2020/09/01"[Date - Entry])).
Finally, searches for tau species and GFAP in blood were further extended to capture papers indexed in PubMed through April 30, 2021.
Results were cross referenced with papers that had been previously curated. Search results were evaluated based on the same inclusion criteria as version 1.1.
We conducted 50 meta-analyses comparing Alzheimer’s disease to controls or comparing MCI that progressed to AD to stable MCI; 1227 comparisons were incldued in these meta-analyses.
We also added 188 meta-analyses comparing biomarker levels in other neurological conditions to levels in AD; 2170 comparisons were included in these meta-analyses.
Inclusion Criteria
Included | Excluded | |
---|---|---|
Study Design |
Biomolecule assessed as a potential biomarker |
Biomolecule assessed during routine clinical testing |
Language of Publication |
English |
Language other than English |
Year of Publication |
1984 - * (see version details) |
Pre-1984 |
Number of Conditions |
≥ 2 eligible conditions |
< 2 eligible conditions |
Conditions |
Study includes AD or MCI-AD |
Study lacks AD or MCI-AD |
Study Subjects |
Human |
Non-human |
Age of Subjects |
Mean age of onset ≥ 18 years |
Mean age of onset < 18 years |
Number of Subjects |
≥ 10 |
< 10 |
Diagnostic Criteria |
Conditions defined by published criteria. Exceptions are controls and certain conditions (e.g., Parkinson's disease and Normal Pressure Hhydrocephalus) |
Conditions not defined by published criteria. Exceptions are controls and certain conditions (e.g., PD and NPH). Heterogeneous groups (e.g., "other dementias") and those lacking published diagnostic criteria |
Source of Biomarker |
CSF, serum, and plasma |
Other body fluid samples (e.g., saliva, urine). Biomolecules measured within blood cells (e.g., lymphocytes) |
Measurement Methods |
Methods that provide quantitative data (e.g., ELISA, xMAP). Selected reaction monitoring (SRM) with an internal standard |
Methods that do not provide quantitative data (e.g., Western blot and exploratory proteomics). Selected Reaction Monitoring (SRM) without an internal standard |
Format of Biomarker Measurement |
Quantitative data reported as mean ± SD or mean ± SEM |
Data reported in formats other than mean ± SD or mean ± SEM (e.g., graphical, ROC). Authors are invited to provide data in eligible formats |
Mild Cognitive Impairment (MCI) Groups |
Groups are considered MCI-Stable if subjects maintained MCI status ≥ 2 years after baseline. MCI groups with subjects converting to AD (i.e., MCI-AD) are included regardless of follow-up time |
MCI without cognitive follow-up. MCI-Stable with mean follow-up of <2 years. Heterogeneous groups at cognitive follow-up (e.g., MCI-other dementias) |
Previously Reported Data |
Original data are only included once to the best of our ability. When multiple papers report the same biomarker in largely overlapping cohorts, the paper with the largest number of subjects is chosen for inclusion in the meta-analyses. Studies with the longest follow-up time in follow-up studies will be included. If the same baseline data of one cohort are re-analyzed in relation to diagnoses at later follow-up time points, the first study is deleted and replaced by a new study. If a later paper reports a new biomarker in the same cohort, only the new biomarker data are added |
|
When a study met all required eligibility criteria with the exception of reporting measurement data in a format other than mean ± SD or mean ± SEM (e.g., median and range) we attempted to obtain data in the format required for meta-analysis. In general, the corresponding author(s) of the report were contacted and asked to provide the data as mean ± SD. When possible, the converted data were included in the database and indicated as such.
Meta-Analysis
Biomarkers with two or more data points were eligible for meta-analysis. For the purposes of meta-analysis, plasma and serum results were meta-analyzed together.
Meta-analyses were performed using ratios of mean biomarker levels: the AD/control ratio, the MCI-AD/stable MCI ratio, or the “other condition”/AD ratio. In studies with more than one cohort per diagnostic group, cohorts were selected as described below. In studies that analyzed eligible biomarkers using more than one assay, only the most common commercial assay was included in the meta-analysis.
AD vs CTRL
In studies with more than one control cohort, only the most cognitively normal and age-appropriate cohort was used. In studies with more than one AD cohort, all AD cohorts were included and divided by the control group to generate multiple ratios per study.
Cross disease
In studies with more than one AD cohort, the cohort with the more confident AD diagnosis was used (i.e., “probable AD” was chosen over “possible AD”). If the AD cohorts were not distinguished on the basis of confidence of diagnosis, the cohort that most closely age-matched the comparison cohort was used.
In studies with more than one cohort for an “other condition,” (e.g., FTD sub-classified into bvFTD, nfvPPA, and svPPA) all “other condition” cohorts were included and divided by the AD group to generate multiple ratios per condition.
The variance of the natural log-transformed ratio of the mean values was estimated using the delta method according to Friedrich et al., 2008. Random effects meta-analysis using the method of DerSimonian and Laird with the estimate of heterogeneity taken from the inverse-variance fixed-effect model (DerSimonian and Laird, 1986) was applied in Stata 13.1 (metan command sbe24_3) (Alzbiomaker versions 1.0-2.1) or R (Alzbiomarker version 3.0). The overall effect size is a weighted average of all individual effect sizes where the inverse of the variances is used as weights.