Bayes analysis supports null hypothesis of anti‐amyloid beta therapy in Alzheimer's disease
Edo Richard, Melina G.H.E. den Brok, Willem A. van Gool
Abstract
Numerous clinical trials of anti-amyloid beta (Aβ) immunotherapy in Alzheimer's disease have been performed. None of these have provided convincing evidence for beneficial effects. Using traditional frequentist meta-analysis, the conclusion is that there is absence of evidence for a therapeutic effect, with a point estimate effect size of 0.05 (95% confidence interval -0.00 to 0.10, P = .055). In addition, this non-significant effect equates to 0.4 points per year on the cognitive subscale of the Alzheimer's Disease Assessment Scale. This is well below the minimally clinically important difference. Bayesian meta-analysis of these trial data provides strong evidence of absence of a therapeutic effect, with a Bayes factor of 11.27 in favor of the null hypothesis, opposed to a Bayes factor of 0.09 in favor of a treatment effect. Bayesian analysis is particularly valuable in this context of repeatedly reported small, non-significant effect sizes in individual trials. Mechanisms other than removal of Aβ from the brain may be probed to slow progression of Alzheimer's disease.