Finite sample corrections for average equivalence testing

Area of Science:

• Biostatistics
• Pharmacometrics
• Statistical Inference

Background:

• Average (bio)equivalence tests determine if a parameter, such as mean difference, falls within predefined equivalence limits.
• The standard Two One-Sided Tests (TOST) procedure is commonly used but can be conservative, leading to power loss with high variability.
• This conservativeness can prevent concluding true equivalence, particularly in pharmaceutical applications with variable drug responses.

Purpose of the Study:

• To introduce a finite sample correction for the TOST procedure, termed the $$ -TOST.
• To guarantee a controlled test size (Type I error rate) of $$ for the corrected procedure.
• To enhance the power of equivalence testing, especially in scenarios with high drug variability.

Main Methods:

• Developed a finite sample and variability correction to the significance level of the TOST procedure.
• The $$ -TOST procedure adjusts the TOST to maintain a specified Type I error rate.
• Evaluated the operating characteristics and power of the $$ -TOST against existing methods.

Main Results:

• The $$ -TOST procedure is uniformly more powerful than the standard TOST.
• The proposed method demonstrates superior operating characteristics compared to competing procedures.
• A case study on econazole nitrate deposition in porcine skin illustrates the practical benefits and advantages of $$ -TOST.

Conclusions:

• The $$ -TOST offers a statistically sound and more powerful alternative for average (bio)equivalence testing.
• This corrected procedure effectively addresses the power loss issues of TOST in the presence of high variability.
• The $$ -TOST is computationally simple and provides improved reliability for concluding equivalence.