Robust confidence intervals for meta-regression with interaction effects.