Mathematics anxiety negatively affects teachers by influencing their classroom practice and being transmitted to their students. This issue must be addressed from the earliest stages of teachers’ professional development, making the interventions that are aimed at reducing such anxiety during initial teacher training particularly relevant. This study conducts a meta-analysis of these interventions in order to identify their most effective characteristics, based on effect sizes. The results show that effect sizes are generally moderate in terms of reducing mathematics anxiety but, in addition, suggest an increase in effectiveness during longer interventions. The most effective interventions are those that focus on problem solving, as well as those based on the creation of supportive, welcoming, and non-restrictive learning environments.