Bioequivalence Study Design: Crossover vs Replicate vs Parallel, Sample Size and Acceptance Criteria

Scientifically reviewed by: Dr. Abdulnaser Alsharaa, PhD Bioanalytical Chemistry, Bioanalysis Department Head, PharmaPrimes

Short answer: For most immediate-release oral generics, the standard bioequivalence (BE) study design is a randomised, single-dose, two-period, two-sequence crossover, usually in healthy subjects. BE is concluded when the 90% confidence interval of the test/reference geometric mean ratio for the primary PK parameters lies within 80.00-125.00%. (Health Canada/FDA presentation on ICH M13A) A parallel design is used when a crossover is impractical, for example for drugs with a long elimination half-life. Replicate designs are used mainly for highly variable drugs, where some regulators allow reference-scaled criteria. The right design depends on the drug, the product and each target regulator.

Choosing a BE study design is not a preference. Half-life, within-subject variability, safety in healthy volunteers and the target regulators decide what is acceptable, and the design then sets sample size, cost and the statistical method. This guide covers each option and when to use it.

The regulatory baseline in 2026

• ICH M13A (BE for immediate-release solid oral dosage forms) was adopted at Step 4 on 23 July 2024, and FDA issued it as guidance in October 2024. (ICH M13A; FDA) A pivotal study needs at least 12 evaluable subjects in a crossover, or at least 12 per group in a parallel design, with a priori power of at least 80%. (M13A Q&A, Q2.1)
• ICH M13B (additional-strength biowaivers) reached Step 4 on 4 August 2026. (ICH M13B)
• ICH M13C will cover highly variable drugs, narrow therapeutic index (NTI) drugs and complex designs such as adaptive designs. (ICH M13B, 1.3) An FDA presentation from September 2025 placed it at Step 1, with Step 2 estimated for June 2027. (FDA) Until M13C is final, regional rules apply to these cases.
• FDA statistical guidance: FDA finalised “Statistical Approaches to Establishing Bioequivalence” on 29 May 2026. It replaces the 2001 guidance and adds reference-scaled analysis for highly variable and NTI drugs, clarified sample size determination, and methods for missing data and outliers. (Federal Register, 29 May 2026)
• Product-specific guidance: FDA publishes product-specific guidances that recommend the design for individual drug products, and these should be checked first. (FDA PSGs)

Comparing the main designs

Replicate design descriptions follow BEBAC and Haidar et al., AAPS J 2008. The parallel design for long half-life drugs is from the M13A presentation.

Bioequivalence study design for highly variable drugs

A drug is generally considered highly variable when its within-subject coefficient of variation for AUC and/or Cmax is 30% or more. (J Pharm Pharm Sci) Reference-scaling needs a replicate design so that the reference product’s variability can be estimated. (BASG)

Sources: EMA approach as described by J Pharm Pharm Sci and BASG; FDA approach as described by BEBAC and Certara. Confirm the details against the primary guidance before use.

Narrow therapeutic index drugs

In the EU, BE for NTI drugs is shown when the 90% CI for AUC, and in some cases Cmax, falls within 90.00-111.11%. Whether a substance counts as NTI is decided case by case. (Pharmaceutics 2022) FDA’s 2026 statistical guidance includes reference-scaled methods for NTI drugs. (Federal Register) For a specific NTI product, follow the product-specific guidance.

What drives sample size

There is no universal correct number of subjects. M13A sets a minimum of 12 evaluable subjects and expects at least 80% power. (M13A Q&A) The actual number depends on:

• Within-subject variability of the primary PK parameters, ideally from a pilot or published study of a comparable product;
• the assumed test/reference ratio: a true ratio away from 1.00 sharply increases the required size;
• the design, because parallel designs depend on total rather than within-subject variability;
• the acceptance method: standard, narrowed for NTI drugs, or reference-scaled;
• target power and significance level: conventionally at least 80% power with two one-sided tests at 5%;
• expected dropouts, added on top of the evaluable number.

The effect of variability is large. The Austrian regulator BASG notes that, under the conventional range, the chance of showing BE for identical products falls markedly as variability rises. (BASG) Sample size should be calculated by a statistician from documented assumptions, not taken from a generic table.

Other design decisions

• Fasting or fed: M13A introduced a risk-based approach to deciding whether fasting, fed or both studies are needed. (FDA presentation on M13A implementation)
• Washout: long enough that drug from one period does not affect the next. Take the pre-dose concentration expectation from the applicable guideline.
• Analyte: generally the parent drug. EMA accepts metabolite data only in exceptional cases. (J Pharm Pharm Sci)
• Bioanalysis: the method is validated under ICH M10, incurred sample reanalysis is expected in all pivotal comparative BA/BE studies, and reanalysis for pharmacokinetic reasons is not acceptable in BE studies. (ICH M10, 3.3.4 and 5)

A practical decision sequence

1. Check for a product-specific guidance from each target authority.
2. Is the product within M13A scope (immediate-release solid oral, systemic action)? If not, identify the applicable guidance.
3. Is the half-life long enough to make a crossover impractical? If yes, consider a parallel design.
4. Is the drug highly variable, or likely to be? If yes, consider a replicate design and check whether each authority permits scaling, and for which parameters.
5. Is the drug NTI in any target market? If yes, apply the stricter regional approach.
6. Are fasting, fed or both studies required under the risk-based approach?
7. Calculate sample size from documented assumptions, including dropouts.

PBPK and virtual BE simulations can help test assumptions about formulation risk before the protocol is finalised. (PharmaPrimes PBPK and virtual BE simulations)

How PharmaPrimes supports BE studies

PharmaPrimes provides integrated clinical and bioanalytical support for BE/BA studies: clinical conduct through the PharmaPrimes Clinical Research Center at Jordan Hospital in Amman, bioanalytical method development and validation according to applicable ICH M10 requirements, clinical sample analysis, PK evaluation and regulatory-ready reporting. See our BE/BA study support.

Frequently asked questions

What is bioequivalence study design?

The plan for comparing a test and a reference product in a BE study: the design (crossover, replicate or parallel), population, dosing conditions, sampling, sample size and statistical criteria.

What are the various types of bioequivalence study designs?

The main PK designs are the 2×2 crossover, partial and full replicate crossovers, and the parallel design. Some products need other approaches, such as steady-state or clinical endpoint studies, as set by the applicable guidance.

What are the FDA guidelines for bioequivalence studies?

For immediate-release oral products, ICH M13A as adopted by FDA, the 2026 statistical guidance, and the product-specific guidance for the drug.

When is a replicate design needed?

Mainly when the drug is highly variable and reference-scaling is planned, because scaling needs an estimate of the reference product’s within-subject variability.

How is sample size calculated for a BE study?

From expected within-subject variability, the assumed test/reference ratio, the design, the acceptance method, target power (at least 80%) and expected dropouts, with at least 12 evaluable subjects under M13A.

Plan your BE study

Discuss your product, reference product and target markets with our clinical and bioanalytical teams at our Amman clinical site. Plan your BE study.

References

1. ICH. M13A, Step 4, 23 July 2024. https://www.pmda.go.jp/files/000269966.pdf
2. FDA. M13A guidance and Q&A, Oct 2024. https://fda.gov/media/165049/download ; https://fda.gov/media/183189/download
3. ICH. M13B, Step 4, 4 Aug 2026. https://database.ich.org/sites/default/files/ICH_M13B%20Final%20Guideline_2026_0717.pdf
4. FDA. Navigating the draft ICH M13B guideline (M13C status), Sept 2025. https://www.fda.gov/media/188796/download
5. Federal Register. Statistical Approaches to Establishing Bioequivalence; final guidance, 29 May 2026. https://www.federalregister.gov/documents/2026/05/29/2026-10705/statistical-approaches-to-establishing-bioequivalence-guidance-for-industry-availability
6. FDA. Product-Specific Guidances for Generic Drug Development. https://www.fda.gov/drugs/guidances-drugs/product-specific-guidances-generic-drug-development
7. ICH M10 (FDA text). https://www.hhs.gov/guidance/sites/default/files/hhs-guidance-documents/FDA/50093493fnl_M10-Bioanalytical-Method-Validation-and-Study-Sample-Analysis_Jan-2025-update.pdf
8. Haidar SH et al. AAPS J. 2008;10(3):450-4. https://pubmed.ncbi.nlm.nih.gov/18726698/