The dynamics of biomarkers across the clinical spectrum of Alzheimer’s disease

摘要

Example illustrating the procedure for the development of biomarker trajectories. In the first step, the observed data is collected and the characteristics of individuals are considered. In the second step, those individuals that have transitioned to ADem during the study are aligned on a disease progression timeline, where time 0 is the time of the first ADem diagnosis. In the third step, for every individual that has not developed ADem symptoms during the study, the Markov model is used to estimate the expected time (number of years) until ADem; given their demographic and genetic characteristics (here, their age and APOE ε4 status), the expected time to ADem diagnosis is estimated using the formulae derived analytically (Additional file 1: Section S2). The distributions of the 1-year transition probabilities used in the Markov model have been estimated from a generalised linear mixed model (see the ‘Transition probabilities’ section and Additional file 1: Section S1) using the Gibbs sampler (the ‘Statistical analysis’ section). In the last step, a sigmoid function (linear for Aβ1–40, Aβ1–42 and t-tau markers in plasma) is fitted to each biomarker data using non-linear least square estimation. The time point at which the first significant biomarker change occurs (green) is defined as the first point at which the 95% CI of the mean biomarker level does not overlap with the 95% CI of the initial mean biomarker level. The 95% CI of the best fit was estimated using the delta method. The inflection point (purple) is the point at which the maximum biomarker rate of change is reached, that is, the point at which the second derivative of the best fit is equal to 0