Increased risk of type I errors for detecting heterogeneity of treatment effects in cluster-randomized trials using mixed-effect models. — SciRadar