Sample size recalculation based on the prevalence in a randomized test-treatment study