Salivary beta amyloid 1‐42 (Aβ42) for the diagnosis of Alzheimer’s disease: a systematic review and meta‐analysis

Abstract

AbstractBackgroundAlzheimer’s disease (AD) neuropathology begins many years before dementia symptoms, however validated cerebrospinal fluid and neuroimaging biomarkers are not scalable to widespread use. Higher salivary beta amyloid 1‐42 (sAβ42) is associated with AD, but its clinical utility remains insufficiently evaluated. This systematic review and meta‐analysis quantifies and evaluates the diagnostic performance of sAβ42 in Alzheimer’s disease.MethodSearches were performed in Ovid MEDLINE, Embase, CENTRAL, PubMed and Web of Science for studies published between January 1, 2000, and March 20, 2022, quantifying sAβ42 in adults with and without AD. All AD diagnoses and criteria were included. Studies without separate AD and control groups and quantitative sAβ42 measurements were excluded. Risk of bias and applicability were evaluated using the QUADAS‐2 tool. Diagnostic performance was assessed by a random‐effects meta‐analysis. Standardised mean differences in sAβ42 levels and 95% confidence intervals in AD and control groups are reported.ResultTen of 373 identified articles and 1 study from citation searching were included. Studies unable to quantify sAβ42 (3) and duplicating subject data (2) were removed, leaving 6 studies comprising 264 AD and 463 control subjects for meta‐analysis. Average sAβ42 differentiated AD from controls (pooled weighted SMD 1.83, 95% CI 0.54‐3.12, p = 0.006). However, high sAβ42 variability and substantial heterogeneity (I2 = 97%) limit the reliability of results. Mean sAβ42 levels were higher in AD versus MCI (p<0.0001) and MCI versus controls (p<0.001) but did not differentiate Parkinson’s disease from AD or controls. Age, AD severity and apolipoprotein E status did not reliably correlate with sAβ42 levels.ConclusionAlzheimer’s disease was associated with significantly higher sAβ42 concentrations compared with controls. However, the reliability and applicability of study results are significantly limited by widespread risk of bias and inherent low GRADE of evidence.