Alzheimer’s disease (AD) is the most common form of dementia, however there is currently no cure. Drugs are available to slow down disease progression, proving most effective at the early stages of AD, yet degenerative brain changes may start decades before clinicians are able to diagnose the condition. Cerebrospinal fluid (CSF) biomarkers have been heralded as objective and effective early detectors of AD, although the interpretation of biomarker data can be problematic, with results often presented in a way that is difficult to employ in routine clinical practice. Sylvain Lehmann from CHU Montpellier and the University of Montpellier, France and colleagues, present a new AD prediction model for use with multivariate CSF biomarker data, as published in their recent study in Alzheimer’s Research & Therapy.
Lehmann and colleagues sampled more than 1,000 cognitive disorder patients from six independent memory clinic cohorts in Paris, Lille and Montpellier, and classified them into AD and non-AD (NAD) patients using clinical diagnostic criteria. Concentrations of the CSF AD biomarkers β-amyloid1-42 (Aβ42), total-tau (tau) and phosphorylated-tau (p-tau) were measured from lumbar puncture samples taken from the patients. As expected, differences in individual biomarker concentrations between AD and NAD were apparent. Optimal cut-off values on levels of the three CSF biomarkers were computed, above or below which they are implicated in the disease.
The authors used these data to compare two prediction models: that of logistic regression, already proven to perform highly in AD diagnosis, and a new, simpler scale, the PLM. The PLM scale was based on a straightforward and intuitive rule: class 0 corresponding to no pathologic biomarkers (according to the designated cut-off values); class 1 corresponding to one pathologic biomarker out of three; class 2 corresponding to two pathologic biomarkers out of three; and class 3 with all three biomarkers being pathological. Results revealed that the predictive values from PLM were not significantly different to those of logistic regression, however the difference in patient distribution between the two prediction models was significantly different – notably the PLM scale had more AD patients in class 3 as well as more NAD patients in class 0. The new scale resulted in 23.25 percent more patients better classified than for logistic regression.
The PLM scale not only outperformed the logistic regression but also has the advantage of not requiring complex mathematical adjustments. It potentially provides a simple and effective tool for physicians in memory clinics to determine the probability that a patient has AD, and also has prospective value in grouping patients for clinical research trials.