About Bioequivalence Assessment

A bioequivalence (BE) study is a critical step in drug development, especially for generic drugs. It aims to demonstrate that a new formulation (Test product) is therapeutically equivalent to an existing one (Reference product). The Bioequivalence Confidence Interval Calculator is an educational tool designed to simplify the core statistical calculation used in this assessment.

What This Calculator Does

This tool performs the standard statistical analysis for average bioequivalence (ABE). It calculates the 90% confidence interval (CI) for the ratio of geometric means (Test/Reference) using summary statistics from a pharmacokinetic (PK) study. Key features include:

• Handling common study designs: 2x2 Crossover, 2-Group Parallel, and Replicate Designs.
• Calculating the 90% CI for key PK parameters like AUC and Cmax.
• Implementing criteria for Scaled Average Bioequivalence (SABE) for highly variable drugs.
• Comparing the calculated CI against standard or scaled acceptance limits to determine if bioequivalence is met.

When to Use It

This calculator is intended for educational and illustrative purposes. It is useful for:

• Pharmacists, students, and researchers learning about bioequivalence statistics.
• Performing quick, preliminary checks on summary data from publications.
• Understanding how changes in sample size, variability (MSE), or geometric mean ratio affect the outcome of a BE study.

It is not validated for use in regulatory submissions to agencies like the FDA or EMA. Official BE analysis requires specialized, validated software (e.g., SAS, R, WinNonlin).

Inputs Explained

Study Design:

The structure of the clinical trial. Crossover: each subject receives both Test and Reference drugs. Parallel: one group receives the Test drug, another receives the Reference. Replicate: a crossover design where at least one product is administered more than once, used for highly variable drugs.

Alpha (α) Level:

The significance level. For a 90% confidence interval, alpha is 0.05. This is the standard for bioequivalence studies, corresponding to two one-sided tests (TOST) at the 5% level.

Subjects in Test/Reference Group (nT / nR):

The number of subjects providing data for each formulation. In a 2x2 crossover, nT must equal nR.

Geometric Mean of Test (%):

The geometric mean of the PK parameter (e.g., AUC) for the Test product, expressed as a percentage relative to the Reference product.

Mean Squared Error (MSE):

Also known as residual variance, this value comes from the Analysis of Variance (ANOVA) on the log-transformed PK data. It represents the unexplained variability in the model and is crucial for calculating the confidence interval.

Intra-subject SD of Reference (sWR):

The within-subject standard deviation of the log-transformed data for the Reference product. This is required only for replicate designs to assess if a drug is "highly variable" (sWR ≥ 0.294), which may trigger scaled acceptance limits (SABE).

Acceptance Bounds (%):

The predefined limits for the 90% CI. The standard limits are 80.00% to 125.00%.

Results Explained

Geometric Mean Ratio (GMR):

The ratio of the geometric mean of the Test product to the Reference product (T/R). A GMR of 100% indicates the central tendency of the two products is identical.

90% Confidence Interval:

The range within which the true GMR is expected to lie with 90% confidence. For bioequivalence to be established, this entire range must fall within the acceptance limits.

Acceptance Limits:

The criteria for passing. This is typically 80.00% - 125.00%. If SABE is applied because the drug is highly variable, these limits are widened based on the observed sWR.

Status (PASS/FAIL):

A clear indication of the outcome. PASS means the 90% CI is fully contained within the acceptance limits. For SABE, the GMR must also be within 80-125%. FAIL means one or both of these conditions are not met.

Formula / Method

The calculation is based on log-transformed pharmacokinetic data, which follows a normal distribution. The steps are:

Step-by-Step Example

Consider a 2x2 crossover study with the following summary statistics for AUC:

• n_T = 24, n_R = 24
• Geometric Mean of Test = 97.5%
• MSE = 0.05
• Alpha = 0.05

Calculation Steps:

1. Difference: μ_diff = ln(0.975) = -0.0253
2. SE: SE = √[ 0.05 × (1/24 + 1/24) ] = √0.004167 = 0.0645
3. df: df = 24 + 24 - 2 = 46
4. t-value: t_{(0.95, 46)} ≈ 1.6787
5. CI (log):
   Lower = -0.0253 - (1.6787 × 0.0645) = -0.1336
   Upper = -0.0253 + (1.6787 × 0.0645) = 0.0830
6. CI (%):
   Lower = e^-0.1336 × 100 = 87.49%
   Upper = e^0.0830 × 100 = 108.65%

Conclusion: The 90% CI is 87.49% - 108.65%. Since this entire range is within the standard 80.00% - 125.00% acceptance limits, the study passes for bioequivalence.

Tips + Common Errors

• Use Geometric Means: Always use geometric means for BE analysis, not arithmetic means. PK parameters like AUC and Cmax are typically log-normally distributed.
• Check the MSE Source: Ensure the MSE value is from the ANOVA on log-transformed data. Using variance or SD from the raw data is a common error.
• Correct Degrees of Freedom: The df calculation depends on the study design and statistical model. The formula (nT + nR - 2) is correct for a simple two-group comparison. Crossover designs use a different df (e.g., n-2 for a 2x2). The calculator correctly uses the total number of subjects minus 2 as a conservative estimate for both parallel and crossover designs.
• SABE is Not Automatic: Scaled Average Bioequivalence (SABE) is only permitted for specific drugs designated as highly variable by regulatory agencies, and it requires a replicate study design. It cannot be applied to a standard 2x2 crossover study, even if the observed variability is high.

Frequently Asked Questions

1. Why is data log-transformed for bioequivalence analysis?

Pharmacokinetic parameters like AUC and Cmax are often skewed and not normally distributed. A logarithmic transformation typically normalizes the data, stabilizing the variance and allowing the use of standard parametric statistical tests (like the t-test via ANOVA).

2. Why is the confidence interval 90% and not 95%?

The 90% confidence interval corresponds to the "two one-sided tests" (TOST) procedure, where we test two separate null hypotheses at a 5% significance level (alpha = 0.05). This is the regulatory standard for bioequivalence.

3. What is the difference between a 2x2 crossover and a parallel design?

In a crossover design, each subject acts as their own control, receiving both the test and reference drugs (in a randomized order). This reduces variability. In a parallel design, one group of subjects receives only the test drug, and a separate group receives only the reference drug.

4. What does "highly variable drug" (HVD) mean?

An HVD is a drug for which the within-subject variability for a PK parameter is high. The regulatory definition is a drug with an intra-subject standard deviation (sWR) of 0.294 or greater, which corresponds to a coefficient of variation (CV%) of approximately 30% or more.

5. When is Scaled Average Bioequivalence (SABE) used?

SABE is an approach used for HVDs. It allows the acceptance limits for BE to be widened based on the observed variability (sWR). This is permitted because for HVDs, it can be unnecessarily difficult to demonstrate BE within the standard 80-125% limits, even if the products are therapeutically equivalent. It requires a replicate design.

6. What if my GMR is 100% but the CI is too wide?

This means the study fails to demonstrate bioequivalence. A GMR of 100% indicates perfect agreement in the central tendency, but a wide CI suggests high variability or an insufficient sample size, leading to low confidence in the result. The entire CI must be within the acceptance limits.

7. How can I reduce the width of the confidence interval?

You can narrow the CI by either (1) increasing the number of subjects in the study (which reduces the SE) or (2) reducing the variability (MSE) through careful study conduct and design (e.g., using a crossover design).

8. Can I use this calculator for food-effect studies?

Yes, the statistical principle is the same. You would analyze the data from the fed state and the fasted state separately, inputting the corresponding summary statistics (GMs, MSE, N) for each analysis.

9. Where do I find the Mean Squared Error (MSE)?

The MSE is a primary output of an Analysis of Variance (ANOVA) table. It is typically labeled as "Residual" or "Error" Mean Square in statistical software output.

References

For more detailed information, consult the official regulatory guidelines:

• FDA Guidance: Bioequivalence Studies with Pharmacokinetic Endpoints for Drugs Submitted Under an ANDA (2013)
• EMA Guideline on the Investigation of Bioequivalence (2010)
• FDA Guidance: Statistical Approaches to Establishing Bioequivalence (2001)
• Chow, S. C., & Liu, J. P. (2008). Design and Analysis of Bioavailability and Bioequivalence Studies. Chapman and Hall/CRC.

Disclaimer