You are not comparing like to like. Those are individual experiments; this is a meta-analysis. In fact, the first study you mention was included in the analysis.