A pharmacogenomic recommendation advises an altered starting dose for a medicine. Why should a clinician not assume that the dose change can be calculated by simply scaling it in direct proportion to a predicted enzyme activity score?