Integrating cerebrospinal fluid-
biomarkers into diagnostic workup of patients with
sporadic cerebral amyloid angiopathy may support early and correct identification. We aimed to identify and validate clinical- and cerebrospinal fluid-
biomarkers for in vivo diagnosis of
cerebral amyloid angiopathy. This observational cohort study screened 2795 consecutive patients admitted for cognitive complaints to the academic departments of neurology and psychiatry over a 10-year period (2009-2018). We included 372 patients with available
hemosiderin-sensitive MR imaging and cerebrospinal fluid-based neurochemical
dementia diagnostics, i.e. Aβ40, Aβ42, t-tau, p-tau. We investigated the association of clinical- and cerebrospinal fluid-
biomarkers with the MRI-based diagnosis of
cerebral amyloid angiopathy, applying confounder-adjusted modelling, receiver operating characteristic and unsupervised cluster analyses. We identified 67 patients with
cerebral amyloid angiopathy, 76 patients with
Alzheimer's disease, 75 patients with
mild cognitive impairment due to
Alzheimer's disease, 76 patients with
mild cognitive impairment with unlikely
Alzheimer's disease and 78 healthy controls. Patients with
cerebral amyloid angiopathy showed a specific cerebrospinal fluid pattern: average concentration of Aß40 [13 792 pg/ml (10 081-18 063)] was decreased compared to all controls (P < 0.05); Aß42 [634 pg/ml (492-834)] was comparable to
Alzheimer's disease and
mild cognitive impairment due to
Alzheimer's disease (P = 0.10, P = 0.93) but decreased compared to
mild cognitive impairment and healthy controls (both P < 0.001); p-tau [67.3 pg/ml (42.9-91.9)] and t-tau [468 pg/ml (275-698)] were decreased compared to
Alzheimer's disease (P < 0.001, P = 0.001) and
mild cognitive impairment due to
Alzheimer's disease (P = 0.001, P = 0.07), but elevated compared to
mild cognitive impairment and healthy controls (both P < 0.001). Multivariate modelling validated independent clinical association of
cerebral amyloid angiopathy with older age [odds-ratio: 1.06, 95% confidence interval (1.02-1.10), P < 0.01], prior lobar intracerebral haemorrhage [14.00 (2.64-74.19), P < 0.01], prior
ischaemic stroke [3.36 (1.58-7.11), P < 0.01], transient focal neurologic episodes (TFNEs) [4.19 (1.06-16.64), P = 0.04] and gait disturbance [2.82 (1.11-7.15), P = 0.03]. For cerebrospinal fluid-
biomarkers per 1 pg/ml, both lower Aß40 [0.9999 (0.9998-1.0000), P < 0.01] and lower Aß42 levels [0.9989 (0.9980-0.9998), P = 0.01] provided an independent association with
cerebral amyloid angiopathy controlled for all aforementioned clinical confounders. Both
amyloid biomarkers showed good discrimination for diagnosis of
cerebral amyloid angiopathy among adjusted receiver operating characteristic analyses (area under the receiver operating characteristic curves, Aß40: 0.80 (0.73-0.86), P < 0.001; Aß42: 0.81 (0.75-0.88), P < 0.001). Unsupervised Euclidian clustering of all cerebrospinal fluid-
biomarker-profiles resulted in distinct segregation of
cerebral amyloid angiopathy patients from all controls. Together, we demonstrate that a distinctive set of cerebrospinal fluid-
biomarkers effectively differentiate
cerebral amyloid angiopathy patients from patients with
Alzheimer's disease,
mild cognitive impairment with or without underlying
Alzheimer's disease, and healthy controls. Integrating our findings into a multiparametric approach may facilitate diagnosing
cerebral amyloid angiopathy, and may aid clinical decision-making, but warrants future prospective validation.