Biomarker testing in MCI patients: Deciding who to tap

Abstract

AbstractBackgroundIt is unlikely that biomarker testing contributes to a more accurate prognosis in every patient with Mild Cognitive Impairment (MCI). We therefore aimed to derive a stepwise approach to select MCI patients for cerebrospinal fluid (CSF) testing and determine the optimal proportion of patients to receive CSF testing.MethodsWe selected n=402 MCI patients from the Amsterdam Dementia Cohort (ADC) and n=221 from the BioFinder study, with available demographic information, Magnetic Resonance Imaging (MRI) and CSF data. Three‐year progression probabilities to dementia were predicted for all patients, using previously published model (van Maurik et al. 2019, The Lancet Neurology) based on 1) patient characteristics (age, sex, MMSE; demographic model), or 2) additional MRI (hippocampal volume; MRI model). We incrementally augmented the proportion of patients to receive CSF testing, first the most uncertain 10% of patients (percentile 45‐55 of demographic‐ or MRI model derived probabilities), then 20% (percentile 40‐60), and so on. Prognostic performance of this stepwise model was measured with Harrell’s C. The optimal proportion was selected as the point where the stepwise model performed similar to testing all patients. This proportion provides the range of probabilities where additional CSF testing would be indicated (Table 1). To study generalizability, we replicated this approach in BioFINDER.ResultsOn top of demographic characteristics, the stepwise approach showed that 60% of the patients required CSF testing in both cohorts in order to reach optimal prognostic performance (i.e. similar to CSF model; Figure 1A). This proportion corresponded with progression probability between 32%‐56%. This range was similar in BioFINDER (34%‐56%; Table 1). If we took the MRI model as a starting point, the optimal proportion was 60% in the ADC and 70% in BioFINDER (i.e. similar to ATN model; Figure 1B). This corresponded with a probability range of 19%‐40% in the ADC and 19%‐48% in BioFINDER (Table 1).ConclusionWe showed that CSF biomarker testing adds prognostic value on clinical and/or imaging information in 60‐70% of the MCI patients that had less certain probability of progression between 32%‐56% or 19%‐48%. As such, we achieve a generalizable CSF saving recommendation with optimal prognostic accuracy.