Empirical assessment of biases in cerebrospinal fluid biomarkers of Alzheimer's disease: an umbrella review and re-analysis of data from meta-analyses.

Abstract

Alzheimer's disease (AD) is a leading cause of years lived with disability in older age, and several cerebrospinal fluid (CSF) markers have been proposed in individual meta-analyses to be associated with AD but field-wide evaluation and scrutiny of the literature is not available. We performed an umbrella review for the reported associations between CSF biomarkers and AD. Data from available meta-analyses were reanalyzed using both random and fixed effects models. We also estimated between-study heterogeneity, small-study effects, excess significance, and prediction interval. A total of 38 meta-analyses on CSF markers from 11 eligible articles were identified and reanalyzed. In 14 (36%) of the meta-analyses, the summary estimate and the results of the largest study showed non-concordant results in terms of statistical significance. Large heterogeneity (I2≥75%) was observed in 73% and small-study effects under Egger's test were shown in 28% of CSF biomarkers. Our results suggest that there is an excess of statistically significant results and significant biases in the literature of CSF biomarkers for AD. Therefore, the results of CSF biomarkers should be interpreted with caution.