Population Bioequivalence

A Shiny app running the FDA's step-wise population bioequivalence (PBE) procedure for comparing a generic and reference drug product from in vitro data alone.

Try it

The app is hosted on Shinyapps.io. Upload a dataset of test and reference in vitro measurements and it will calculate the full step-wise PBE algorithm. The output can be used to determine population bioequivalence

Example data

Click below to download the the worked-example dataset from the FDA's product-specific guidance on Budesonide Inhalation Suspension.

What is population bioequivalence?

Population bioequivalence is a statistical framework used to show a test product performs the same as an already approved reference product using only in vitro measurements. Unlike an ordinary average bioequivalence test, which only compares mean values, this builds a single aggregate criterion out of both the mean difference between products and the ratio of their variances. This ensures the product is consistent.

The FDA step-wise PBE procedure

This app implements the exact procedure recommended in the FDA's product-specific guidance for Budesonide Inhalation Suspension. More specifically:

• Set up the criterion. Bioequivalence requires we test (μ_T−μ_R)² + (σ_T²−σ_R²), divided by either the reference variance σ_R² or a fixed regulatory variance σ_T0², to be less than a regulatory constant θ_p = 2.0891.
• Decompose the variance. For each product, split the total variance into between-container variability (MSB) and within-container variability (MSW), using every batch, container and life stage in the dataset.
• Choose the scaling. If the reference product's variability σ_R is greater than σ_T0 = 0.1, the reference-scaled criterion is used; effectively giving these products a more lenient test.
• Estimate and calculate. A linearised point estimate of the criterion is calculated, along with a 95% upper confidence bound built from the corresponding chi-squared and t-distribution quantities.
• Conclude. The products are declared bioequivalent only if that 95% upper confidence bound is ≤ 0.