Bioequivalence and Bioavailability Forum

Dear Pratap!

❝ can you please give us basic information about power, CI…

There are two types of error in statistics, namely in BE:

• Type I (α)
  The risk of coming to the wrong decision of bioequivalence (BE claimed for a ‘true’ bioinequivalent formulation). This is the patient’s risk and is generally kept at 0.05. Since a particular patient may either have too low bioavailability (risk 0.05) or too high bioavailability (risk 0.05), the risk for the population of patients is kept at 2α = 0.10 (therefore, we use a 1–2α = 90% confidence interval). In some legislations (e.g., Brazil) for narrow therapeutic index drugs a 95% confidence interval (patient’s risk 2.5%) must be applied.
  If multiplicity is an issue (e.g., >1 test formulations in a single study, dose proportionality studies, simultaneous evaluation of parent and active metabolite) a Bonferroni-correction* may be necessary in order to keep the overall patient’s risk at ≤0.05.
  
• Type II (β)
  The risk of coming to the wrong decision of bioinequivalence (BE rejected for a ‘true’ bioequivalent formulation). This is the producer's risk and is generally kept at ≤0.2. Power π is the chance of being able to demonstrate BE with the given β, whereas π = 1 – β.

❝ … and application of power and CI in BE studies

With the following parameters you may calculate the sample size:
α (generally set to 5%), β (1–π, generally 10–20%), expected deviation of test from reference (obtained from a reasonably sized pilot study or at least 5%), and intra-subject coefficient of variation (CV_intra from ANOVA).
BE for a metric (e.g., AUC, C_max) is demonstrated if the confidence interval (CI) lies entirely with a predefined acceptance range (AR, e.g., 80–125%).

---

• α_adj = 1 – (1 – 0.05/k)^k, where k is the number of comparisons.