Bias in odds ratios by logistic regression modelling and sample size

Abstract

If several small studies are pooled without consideration of the bias introduced by the inherent mathematical properties of the logistic regression model, researchers may be mislead to erroneous interpretation of the results.