Purpose. The FDA Working Group on Highly Variable (HV) Drugs recently presented procedures and conditions for determining the bioequivalence (BE) of HV drug products. They included analysis by the method of scaled average BE (SABE), a switching coefficient of variation of CVS = 30% and a regulatory standardized variation of CV0 = 25% for applying SABE, and the use of a secondary regulatory criterion restricting to 0.80-1.25 the point estimate for the ratio of the estimated geometric means (GMR) of the two formulations. These conditions are scrutinized in the present communication. Methods. 3-period BE studies were simulated with various statistical and regulatory assumptions. Power curves, obtained by gradually increasing the true GMR, compared performances of the methods of SABE, a constrained point estimate of GMR (PE/GMR), and the composite of these two approaches. The consumer risk of each procedure was evaluated. Results. With CV0 = 30% and PE/GMR = 0.80-1.25, the composite criterion of BE relied on the confidence limits of SABE. In contrast, with CV0 = 25% and/or PE/GMR = 0.87-1.15, the composite criterion approached almost completely the features of the GMR point estimate, especially at high within-subject variation. The consumer risk was near 5% with CV0 = 30% but about 15% when CV0 = 25%. Conclusions. A constraint on GMR is difficult to justify scientifically. Still, if it is introduced then the condition of CVS = CV0 = 30% and PE/GMR = 0.80-1.25 is recommended as a composite regulatory criterion. With alternative settings of the conditions, such as the recommended CV0 = 25% and/or PE/GMR = 0.87-1.15, the composite criterion would reflect almost entirely the GMR point estimate. Also, with CV0 = 25%, the BE limits are discontinuous at CVS = 30% and, as consequences, the consumer risk is substantially larger than 5%, and the regulatory uncertainty for making a decision about acceptance or rejection is enhanced. These would be undesirable outcomes.