Sleep Parameters and Plasma Biomarkers for Cognitive Impairment Evaluation in Patients With Cerebral Small Vessel Disease

Abstract

Objectives: Cognitive impairment caused by cerebrovascular disease accounts for more than half of vascular dementia. However, neuropsychological tests are limited by their subjectivity. Additional effective approaches to evaluate cognitive impairment in patients with cerebrovascular disease are necessary.

Method: One hundred and thirty-two patients with cerebrovascular disease were recruited. One hundred participants met the criteria and completed neuropsychological scales. Sixty-nine participants proceeded with polysomnography, and 63 of them had their peripheral blood biomarkers measured. According to Mini-Mental State Examination scores, patients were divided into cognitively impaired and cognitively normal groups. The differences in biomarkers and sleep parameters between the groups were compared, and decision tree models were constructed to evaluate the evaluation ability of these indicators on cognitive decline.

Results: The integrated decision tree model of sleep parameters yielded an area under curve (AUC) of 0.952 (95% confidence interval [CI]: 0.911-0.993), while that of plasma biomarkers yielded an AUC of 0.872 (95% CI: 0.810-0.935) in the assessment of cognition status. Then the participants were automatically clustered into mild and severe cognitive impairment groups by multiple neuropsychological test results. The integrated plasma biomarker model showed an AUC of 0.928 (95% CI: 0.88-0.977), and the integrated sleep parameter model showed an AUC of 0.851 (95% CI: 0.783-0.919) in the assessment of mild/severe cognitive impairment.

Discussion: Integrated models which consist of sleep parameters and plasma biomarkers can accurately evaluate dementia status and cognitive impairment in patients with cerebral small vessel disease. This innovative study may facilitate drug development, early screening, clinical diagnosis, and prognosis evaluation of the disease.

Keywords: Cognitive evaluation; Machine learning; Vascular dementia.