Integrating Biomarkers for Underlying Alzheimer’s Disease in Mild Cognitive Impairment in Daily Practice: Comparison of a Clinical Decision Support System with Individual Biomarkers

Article type: Research Article

Authors: Rhodius-Meester, Hanneke F.M.^{a; *} | Koikkalainen, Juha^b | Mattila, Jussi^b | Teunissen, Charlotte E.^c | Barkhof, Frederik^d | Lemstra, Afina W.^a | Scheltens, Philip^a | Lötjönen, Jyrki^b | van der Flier, Wiesje M.^{a; e}

Affiliations: [a] Alzheimer Center, Department of Neurology, VU University Medical Centre, Neuroscience Campus Amsterdam, Amsterdam, The Netherlands | [b] VTT Technical Research Centre of Finland, Tampere, Finland | [c] Neurochemistry Lab and Biobank, Department of Clinical Chemistry, VU University Medical Centre, Neuroscience Campus Amsterdam, Amsterdam, The Netherlands | [d] Department of Radiology and Nuclear Medicine, VU University Medical Centre, Neuroscience Campus Amsterdam, Amsterdam, The Netherlands | [e] Department of Epidemiology and Biostatistics, VU University Medical Centre, Neuroscience Campus Amsterdam, Amsterdam, The Netherlands

Correspondence: [*] Correspondence to: Hanneke Frederica Maria Rhodius-Meester, MD, Alzheimer Center, VU University Medical Centre, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands. Tel.: +31 204440816; Fax: +31 204448529; h.rhodius@vumc.nl

Abstract: Background:Recent criteria allow biomarkers to provide evidence of Alzheimer’s disease (AD) pathophysiology. How they should be implemented in daily practice remains unclear, especially in mild cognitive impairment (MCI) patients. Objective:We evaluated how a clinical decision support system such as the PredictAD tool can aid clinicians to integrate biomarker evidence to support AD diagnosis. Methods:With available data on demographics, cerebrospinal fluid (CSF), and MRI, we trained the PredictAD tool on a reference population of 246 controls and 491 AD patients. We then applied the identified algorithm to 211 MCI patients. For comparison, we also classified patients based on individual biomarkers (MRI; CSF) and the NIA-AA criteria. Progression to dementia was used as outcome measure. Results:After a median follow up of 3 years, 72 (34%) MCI patients remained stable and 139 (66%) progressed to AD. The PredictAD tool assigned a likelihood of underlying AD to each patient (AUC 0.82). Excluding patients with missing data resulted in an AUC of 0.87. According to the NIA-AA criteria, half of the MCI patients had uninformative biomarkers, precluding an assignment of AD likelihood. A minority (41%) was assigned to high or low AD likelihood with good predictive value. The individual biomarkers showed best value for CSF total tau (AUC 0.86). Conclusion:The ability of the PredictAD tool to identify AD pathophysiology was comparable to individual biomarkers. The PredictAD tool has the advantage that it assigns likelihood to all patients, regardless of missing or conflicting data, allowing clinicians to integrate biomarker data in daily practice.

Keywords: KeywordsAlzheimer’s disease, clinical decision support system, diagnostic test assessment, mild cognitive impairment, prognosis

DOI: 10.3233/JAD-150548

Journal: Journal of Alzheimer's Disease, vol. 50, no. 1, pp. 261-270, 2016