Differential detectability of pericyte and blood–brain barrier readouts across the Alzheimer’s disease clinical continuum: a systematic review and meta-analysis
TL;DR
Overall, CSF sPDGFRβ, DCE-MRI Ktrans, and QAlb appear related but non-interchangeable, and the present cross-sectional evidence should not be interpreted as demonstrating temporal priority, within-person progression, or head-to-head biomarker superiority.
Abstract
Background In Alzheimer’s disease (AD), neurovascular unit and blood–brain barrier (BBB) abnormalities are assessed using biologically distinct readouts, often across separate cohorts, complicating interpretation across clinical stages and biomarker domains. Methods We searched major databases (inception to February 7, 2026) for human observational studies quantifying cerebrospinal fluid (CSF) soluble platelet-derived growth factor receptor-β (sPDGFRβ), dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) Ktrans, or the CSF/serum albumin quotient (QAlb) in cognitively normal (CN), mild cognitive impairment (MCI), and AD dementia groups. Primary contrasts were MCI versus CN, AD dementia versus CN, and AD dementia versus MCI. Outcomes were analyzed separately using standardized mean differences (SMDs) in random-effects models with restricted maximum-likelihood estimation and Hartung–Knapp adjustment. Sensitivity analyses incorporated amyloid/AT (N)-informed biologic anchoring and evidence-tier restrictions where possible. Results We included 24 cross-sectional studies (3,644 participants). Compared to CN, MCI showed higher CSF sPDGFRβ (SMD 0.38, 95% CI 0.03–0.73) and DCE-MRI Ktrans (SMD 0.91, 95% CI 0.09–1.73; highly heterogeneous and based on five studies), whereas the difference for QAlb was smaller (SMD 0.19, 95% CI 0.01–0.37). Stricter etiologic restriction weakened the robustness of MCI-stage estimates, particularly for early AD-specific inferences in cohorts without consistent biomarker confirmation. Tier 1-only pooling was not feasible because biomarker-confirmed MCI-stage evidence was sparse. AD dementia versus CN comparisons showed higher values across all readouts: CSF sPDGFRβ (SMD 0.43, 95% CI 0.12–0.73), Ktrans (SMD 0.90, 95% CI 0.56–1.25; based on four studies), and QAlb (SMD 0.28, 95% CI 0.10–0.45). AD dementia versus MCI comparisons showed no significant pooled differences for CSF sPDGFRβ or QAlb; Ktrans pooling was precluded by insufficient data. Conclusion Pericyte- and BBB-related readouts showed readout-specific, cross-sectional detectability patterns across the AD clinical continuum. Ktrans findings should be considered preliminary because of the small evidence base, and MCI-stage estimates remain limited by etiologic uncertainty. Overall, CSF sPDGFRβ, DCE-MRI Ktrans, and QAlb appear related but non-interchangeable, and the present cross-sectional evidence should not be interpreted as demonstrating temporal priority, within-person progression, or head-to-head biomarker superiority. Systematic review registration PROSPERO, CRD420251142518.