Comparison of sample size formulae for 2 x 2 cross-over designs applied to bioequivalence studies