Combining heterogenous studies using the random-effects model is a mistake and leads to inconclusive meta-analyses.