Replication Study Design for Bioequivalence: A Guide

Bioequivalence studies compare a test product to a reference product to determine whether the two formulations deliver the same drug to the bloodstream at the same rate and extent. A replication study design, also called a replicate crossover design, gives each subject the same treatment more than once across multiple periods. This guide explains when a replicate design is appropriate, how it differs from a conventional two-period crossover, and what statistical and operational considerations matter when planning such a study. The intended readers are students, researchers, life-science professionals, and informed general readers who need a practical understanding of replicate designs in bioequivalence assessment.

What Is a Replicate Crossover Design

A conventional bioequivalence study typically uses a two-period, two-sequence crossover design. Each subject receives the test product in one period and the reference product in the other period, with a washout between periods. This design estimates the average difference between formulations but provides limited information about variability within each subject.

A replicate crossover design repeats at least one of the treatments within the same subject. The most common configurations are the full replicate design, where each subject receives both the test and reference products twice, and the partial replicate design, where each subject receives the reference product twice but the test product once. A full replicate design with two treatments, four periods, and two sequences is often written as a 2 by 4 design. A partial replicate design with two treatments, three periods, and two sequences is sometimes called a 2 by 3 design.

The key feature of a replicate design is that it produces repeated measurements of the same formulation within the same person. These repeated measurements allow the study team to separate the total variability into components that a conventional crossover cannot distinguish. Specifically, a replicate design can estimate the within-subject variability for each formulation separately and can detect a subject-by-formulation interaction, which occurs when the difference between the test and reference products varies from one subject to another.

Why Regulatory Agencies Recommend Replicate Designs

Regulatory interest in replicate designs grew during the 1990s when the concept of individual bioequivalence was proposed. Individual bioequivalence was intended to provide assurance that a patient could be switched from the reference product to the test product without a change in efficacy or safety. The US Food and Drug Administration published guidance documents on individual bioequivalence criteria starting in 1997. From a scientific standpoint, the individual bioequivalence criterion offered advantages over the average criterion for some drug products because it allowed comparison of intraindividual variances, scaling the bioequivalence criterion to the reference variability, and detection of an important subject-by-formulation interaction if one exists. Based on these considerations, the FDA recommended replicate study designs for modified release dosage forms and highly variable drug products. The new criterion also promoted inclusion of a heterogeneous population of volunteers in bioequivalence studies. Despite these advantages, questions remained about the optimal use of replicate study designs and the proposed criterion, and the FDA maintained the average bioequivalence criterion while allowing other criteria under certain circumstances. Collection and analysis of bioequivalence data from replicate study designs was expected to permit further assessment and resolution of these questions.

In practice, replicate designs have become particularly important for highly variable drug products. High within-subject variability is usually associated with a coefficient of variation of 30 percent or more. For such products, a conventional two-period crossover may require an impractically large sample size to demonstrate bioequivalence within the standard 80 to 125 percent acceptance range. A partial replicate design with repeating the reference product and scaling the bioequivalence for the reference variability has been proposed for drugs with high within-subject variability. In cases of high variability, some regulatory authorities may accept a two-stage or group-sequential bioequivalence design using appropriately adjusted statistical analysis.

At a Glance: Choosing a Replicate Design

The decision to use a replicate design depends on the expected variability of the drug product, the regulatory context, and the specific questions the study must answer. The table below summarizes the main design options and their practical applications.

The choice among these designs should be made during protocol development and documented with the scientific rationale. A replicate design is not automatically better than a conventional design. It adds operational complexity and subject burden, so the expected benefit must justify the additional cost.

Statistical Principles Behind Replicate Designs

Within-Subject Variability

Within-subject variability refers to the variation in a subject's response to the same formulation when it is administered on separate occasions. This variability arises from biological fluctuations within the person, such as changes in gastrointestinal motility, blood flow, or enzyme activity, as well as from assay and dosing errors. A conventional two-period crossover cannot estimate within-subject variability separately for the test and reference products because each subject receives each formulation only once. A replicate design provides the repeated measurements needed for these separate estimates.

The within-subject variability of the reference product is particularly important for highly variable drugs. Regulatory approaches for highly variable products may use the reference variability to widen the bioequivalence acceptance limits. The European Medicines Agency has proposed methods for average bioequivalence with expanding limits, where the acceptance range widens as the reference within-subject variability increases. Reference datasets have been developed to help companies qualify and validate the software used to evaluate bioequivalence trials in a replicate design intended for average bioequivalence with expanding limits. These datasets were evaluated by seven different software packages according to methods proposed by the European Medicines Agency. For the estimation of the reference within-subject coefficient of variation and one of the methods, all software packages produced results that were in agreement across all datasets. Due to different approximations of the degrees of freedom, slight differences were observed in two software packages for another method in highly incomplete datasets. All software packages were suitable for the estimation of the reference within-subject coefficient of variation and the first method. For the second method, different methods for approximating the denominator degrees of freedom could lead to slight differences, which could eventually lead to contrary decisions in very rare borderline cases.

Subject-by-Formulation Interaction

A subject-by-formulation interaction exists when the relative performance of the test and reference products differs across subjects. In other words, some subjects may absorb the test product better than the reference product, while other subjects show the opposite pattern. This interaction is relevant for switchability, which is the ability of a patient to switch from the reference product to the test product without a change in therapeutic response.

A replicate design can detect a subject-by-formulation interaction because it provides multiple measurements of each formulation within each subject. The individual bioequivalence criterion was designed to capture this interaction along with differences in within-subject variability. However, the practical value of individual bioequivalence has been debated. In one study of methylphenidate immediate-release tablets, a replicated crossover design was employed using 20 subjects. Each subject received a single 20 mg dose of the reference tablet on two occasions and two doses of the test tablet on two occasions. Using an average bioequivalence criterion, the 90 percent confidence limits for the log-transformed maximum concentration and area under the curve fell within the acceptable range of 80 to 125 percent. Using an individual bioequivalence criterion, the test product failed to demonstrate equivalence in maximum concentration to the reference product. The intrasubject variability of the generic product was greater and the subject-by-formulation interaction variance was borderline high. For these reasons, the test tablets were not individually bioequivalent to the reference tablets. This example illustrates that a product can pass average bioequivalence while failing individual bioequivalence, which has implications for regulatory decisions.

Average Bioequivalence Versus Individual Bioequivalence

Average bioequivalence compares the population means of the test and reference products. It is the traditional criterion and remains the primary basis for regulatory approval in most jurisdictions. Individual bioequivalence goes further by considering whether the test product is equivalent to the reference product within individual subjects. The individual bioequivalence criterion allows comparison of intraindividual variances, scaling the bioequivalence criterion to the reference variability, and detection of an important subject-by-formulation interaction if it exists.

In practice, the average bioequivalence criterion remains the standard for most regulatory decisions. The FDA maintains the average bioequivalence criterion while allowing other criteria under certain circumstances. A replicate design can support average bioequivalence assessment while also providing the data needed for individual bioequivalence evaluation if that becomes necessary.

Practical Workflow for Planning a Replicate Study

Step 1: Assess the Expected Variability

The first decision is whether the drug product is likely to have high within-subject variability. This assessment can be based on published pharmacokinetic data, previous bioequivalence studies of the same drug, or the pharmacological properties of the drug. Drugs with high variability often include those with extensive first-pass metabolism, poor or erratic absorption, or narrow therapeutic windows that require precise dosing.

A review of submitted bioequivalence studies at the Food and Drug Administration has examined the mechanisms why drugs and drug products may exhibit large variability. The physiological complexity of the gastrointestinal tract and the interaction with the physicochemical properties of drug substances may contribute to the variation in plasma drug concentration-time profiles of drugs and drug products and to variability between and within subjects. If the expected within-subject coefficient of variation is 30 percent or higher, a replicate design should be considered.

Step 2: Select the Design Structure

For highly variable drug products, a partial replicate design with repeating the reference product and scaling the bioequivalence for the reference variability is a common approach. This design requires three periods instead of four, which reduces the burden on subjects and the risk of dropout. The full replicate design provides more information but requires four periods and may be difficult to complete in some populations.

The choice between partial and full replication depends on the regulatory requirements and the scientific questions. If the study only needs to estimate the reference within-subject variability for scaling purposes, a partial replicate design may be sufficient. If the study needs to estimate within-subject variability for both formulations or to assess individual bioequivalence, a full replicate design is necessary.

Step 3: Determine the Sample Size

Sample size calculation for a replicate design must account for the within-subject variability of the primary pharmacokinetic parameters, the expected geometric mean ratio, and the desired statistical power. The sample size for a replicate design is often smaller than for a conventional crossover when the drug is highly variable, because the replicate design provides more information per subject.

A genotype-based enrichment strategy can further reduce the sample size for drugs with high pharmacokinetic variability due to polymorphic metabolizing enzymes. In a study of tolterodine, a bioequivalence study was performed in a randomized, open-label, two-by-two crossover design. The coefficient of variation was calculated for each cytochrome P450 2D6 genotype group, and the sample size required to meet the power of an equivalence test was estimated based on the two one-sided test in genotype stratified groups as well as in a conventional group. The coefficients of variation of pharmacokinetic parameters in the conventional dataset were much greater than those in the genotype-based stratified groups. While up to 70 subjects were required for statistical power based on the coefficient of variation of the area under the concentration-time curve observed in the conventional dataset, only 26 to 44 subjects in extensive metabolizers and poor metabolizers, respectively, were required for the CYP2D6 genotype groups. This exploration demonstrated that a drug-metabolizing enzyme genotype-based enrichment strategy can be implemented to minimize the sample size in bioequivalence studies of drugs that have high pharmacokinetic variability due to polymorphic metabolizing enzymes.

Step 4: Predefine the Statistical Analysis

The statistical analysis plan must be written before the study begins and must specify the primary endpoints, the statistical model, and the decision rules. The procedure for selecting an appropriate statistical analysis for a replicate study design should be predefined in the study protocol. This is important because the analysis of replicate design data can be complex, and different analysis approaches may produce different results.

In a study of desmopressin nasal spray preparations, a single-dose, replicate study design was used to evaluate bioequivalence of two desmopressin nasal sprays. Thirty-two healthy male volunteers were enrolled in the study and were randomly assigned to receive the test and reference drug on two occasions in a four-period, two-sequence crossover study design. Statistical analysis was initially performed using a complicated mixed-analysis model testing for individual bioequivalence according to recommendations by the Food and Drug Administration. This approach, however, failed to converge with all defined main pharmacokinetic parameters, and a traditional mixed analysis of variance analysis based on population averages was definitely used for testing bioequivalence between study drugs. The procedure of selecting an appropriate statistical analysis for a replicate study design was predefined in the study protocol. This example shows that even a well-planned replicate study may encounter analytical difficulties, and the protocol should anticipate such situations.

Step 5: Plan for Operational Challenges

Replicate designs require more study periods than conventional designs, which increases the duration of the study and the burden on subjects. Dropout rates may be higher, and missing data can complicate the statistical analysis. The study team should plan for these challenges by recruiting an adequate number of subjects, scheduling visits carefully, and implementing procedures to minimize the risk of protocol deviations.

Case Examples of Replicate Designs in Practice

Tacrolimus Capsules

Tacrolimus is marketed for the prophylaxis of organ rejection following allogenic liver or kidney transplantation. A previously conducted, randomized, 24-subject, crossover bioavailability study of 1 and 5 mg capsules, with one period each, failed to demonstrate bioequivalence. A single-dose, four-period, four-sequence, randomized, crossover, replicate study with 32 subjects was therefore used to evaluate the bioequivalence of the marketed 1 and 5 mg capsules in healthy volunteers. Tacrolimus blood concentrations were measured serially over 72 hours using a commercially available ELISA assay. Ninety percent confidence intervals of log-transformed parameter ratios were 90.5 to 101.9, 87.1 to 101.7, and 89.7 to 103.8 for maximum concentration, area under the curve from time zero to the last blood sample collection, and area under the curve from time zero to infinity, respectively. Since all values were within 80 to 125 percent, the capsules were bioequivalent. Based on percent coefficients of variation, intersubject variability was approximately two to three times greater than intrasubject variability. This example demonstrates how a replicate design can resolve a bioequivalence question that a conventional crossover could not answer.

Artesunate and Amodiaquine Fixed-Dose Combination

Artesun-Plus is a fixed-dose combination antimalarial agent containing artesunate and amodiaquine. To overcome the high intrasubject variability of artesunate, a study applied a two-sequence and four-period crossover, replicate study design to assess bioequivalence between the test product and the World Health Organization designated comparator product in 31 healthy male Chinese volunteers under fasting conditions. The results showed that the values of the geometric mean ratios of maximum concentration and area under the curve from time zero to the last blood sample collection for the artesunate component in the test and reference products were 95.9 percent and 93.9 percent, respectively, and the corresponding 90 percent confidence intervals were 84.5 to 108.7 percent and 87.2 to 101.1 percent. The geometric mean ratios for the amodiaquine component in the test and reference products were 95.0 percent and 100.0 percent, respectively, and the corresponding 90 percent confidence intervals were 86.7 to 104.1 percent and 93.5 to 107.0 percent. Bioequivalence between the two products was demonstrated for both components. The study also confirmed high intrasubject variability, especially for artesunate, with coefficients of variation of maximum concentration values for the test and reference products of 39.2 percent and 43.7 percent, respectively, while those for amodiaquine were 30.6 percent and 30.2 percent, respectively.

Desmopressin Nasal Spray

Due to the high variability of plasma pharmacokinetics of intranasally administered peptides like desmopressin, appropriate study designs are required to assess bioequivalence. A single-dose, replicate study design was used to evaluate bioequivalence of two desmopressin nasal sprays. Thirty-two healthy male volunteers were enrolled in the study and were randomly assigned to receive the test and reference drug on two occasions in a four-period, two-sequence crossover study design. Subjects received a single dose of 20 micrograms of desmopressin-acetate per study day separated by wash-out periods of at least 1 week. Desmopressin blood concentrations were measured serially over a 14-hour period using a validated radioimmunoassay method. The 90 percent confidence intervals were calculated for the area under the time-concentration curve, maximum concentration, and the time to reach maximum concentration of test over reference drug ratios for a bioequivalence range from 0.80 to 1.25. The mean test over reference drug ratios were completely within the 90 percent confidence intervals with values of 1.041 and a confidence interval of 0.892 to 1.216.

Tacrolimus Extended-Release Tablets

Two open-label, randomized, two-treatment, four-period, two-sequence, oral single-dose, fully replicate, crossover trials were conducted in healthy subjects under fasting and fed conditions to establish bioequivalence between two doses of a test product of tacrolimus extended-release tablets and a reference product. The study included subjects aged 18 to 45 years with a body mass index of 18.5 to 29.9 kilograms per square meter. Bioequivalence was assessed based on key pharmacokinetic parameters including area under the curve, maximum concentration, time to maximum concentration, and half-life. The analysis set included a total of 85 subjects for the 1 mg fasting study, 83 for the 1 mg fed study, 85 for the 4 mg fasting study, and 85 for the 4 mg fed study. Results showed comparable maximum concentration, time to maximum concentration, area under the curve from time zero to the last blood sample collection, area under the curve from time zero to infinity, and half-life for the 1 mg and 4 mg test product versus the reference product under fasting and fed conditions. The upper limit of 90 percent confidence intervals for within subject standard deviation for test product over within subject standard deviation for reference product, the 95 percent upper confidence bound, and the 90 percent confidence intervals for maximum concentration and area under the curve values were within the corresponding acceptance criteria as per the European Medicines Agency and met all the bioequivalence criteria.

Testosterone Undecanoate Soft Capsules

A randomized, open-label, two-treatment, four-period, single-center, single-dose crossover clinical trial was conducted to evaluate the pharmacokinetics, bioequivalence, and safety of a single postprandial oral dose of testosterone undecanoate and its originator drug in healthy postmenopausal Chinese women. Participants received single oral doses of 40 milligrams of testosterone undecanoate or the originator drug in each period. Serial blood samples were collected from 0 to 24 hours post-dose. The average adjusted geometric mean ratios with 90 percent confidence intervals for the primary pharmacokinetic parameters maximum concentration, area under the curve from time zero to the last blood sample collection, and area under the curve from time zero to infinity were 102.20 percent with a confidence interval of 90.32 to 115.63 percent, 99.85 percent with a confidence interval of 92.82 to 107.41 percent, and 99.79 percent with a confidence interval of 92.90 to 107.20 percent. All 90 percent confidence intervals for maximum concentration and area under the curve values fell within the 80 to 125 percent bioequivalence range. The two drugs showed comparable results for the other pharmacokinetic parameters, indicating that the two drugs were bioequivalent.

Records and Measurements in Replicate Studies

Pharmacokinetic Parameters

The primary measurements in a bioequivalence study are the pharmacokinetic parameters derived from the plasma concentration-time profile. The most important parameters are the maximum observed concentration, the area under the concentration-time curve from time zero to the last measurable concentration, and the area under the concentration-time curve from time zero to infinity. The time to reach maximum concentration is also recorded but is typically analyzed descriptively instead of as a primary bioequivalence endpoint.

In a replicate design, each subject contributes multiple measurements of each parameter for each formulation. These repeated measurements are the basis for estimating within-subject variability. The study team should record all individual concentration-time data, the calculated pharmacokinetic parameters for each period, and the statistical outputs from the mixed model analysis.

Variability Estimates

The key outputs from a replicate study are the within-subject coefficients of variation for the test and reference products and the subject-by-formulation interaction variance. These estimates are used to determine whether the product qualifies for scaled bioequivalence limits and to assess switchability.

The within-subject coefficient of variation for the reference product is particularly important for regulatory decisions. If the reference within-subject variability is high, the acceptance limits may be widened according to the reference-scaled approach. The study report should present the variability estimates with their confidence intervals and should state whether the product meets the criteria for scaled limits.

Protocol Deviations and Missing Data

Replicate studies are more susceptible to missing data than conventional studies because each subject must complete more periods. A subject who drops out after the second period contributes incomplete data, and the statistical analysis must account for the missing observations. The study team should document all protocol deviations, including missed doses, unscheduled visits, and sample collection errors, and should assess the potential impact of these deviations on the study conclusions.

Common Failure Patterns in Replicate Studies

Failure of the Complex Statistical Model to Converge

The analysis of replicate design data often uses mixed-effects models that include random effects for subjects, periods, and formulations. These models can fail to converge, particularly when the data are sparse or the variance components are difficult to estimate. In the desmopressin study, the initial mixed-analysis model testing for individual bioequivalence failed to converge with all defined main pharmacokinetic parameters, and a traditional mixed analysis of variance analysis based on population averages was used instead. The procedure for selecting an appropriate statistical analysis was predefined in the study protocol, which allowed the study team to proceed without post hoc decisions.

Borderline Bioequivalence Decisions

Replicate designs can produce borderline results where the confidence interval is close to the acceptance limit. In such cases, the decision may depend on the statistical method used. Different software packages may produce slightly different results due to different approximations of the degrees of freedom, which could lead to contrary decisions in very rare borderline cases. The study team should use validated software and should confirm the results with an independent analysis if the decision is borderline.

High Dropout Rates

The longer duration of replicate studies increases the risk of subject dropout. Subjects may withdraw due to adverse events, scheduling conflicts, or personal reasons. A high dropout rate can reduce the statistical power and may bias the results if the dropout is related to the treatment. The study team should monitor dropout rates throughout the study and should consider whether the sample size needs to be adjusted.

Carryover Effects

The washout period between doses must be long enough to ensure that the drug from the previous period is completely eliminated before the next dose. If the washout is too short, carryover effects can confound the results. The study protocol should specify the washout duration based on the half-life of the drug, and the study team should verify that the predose concentrations in each period are below the limit of quantification.

Limitations of Replicate Designs

Operational Complexity

Replicate designs require more study periods, more blood samples, and more analytical work than conventional designs. The cost of the study is higher, and the logistics are more complex. The study team must have the resources and expertise to manage the additional workload.

Subject Burden

Subjects in a replicate study must visit the clinic more often and provide more blood samples. This burden can affect recruitment and retention, particularly in studies involving healthy volunteers who may not have a strong motivation to complete the study. The study design should minimize the burden where possible, such as by using limited sampling strategies for drugs with long half-lives.

Statistical Complexity

The statistical analysis of replicate design data is more complex than the analysis of conventional crossover data. The study team must have expertise in mixed-effects modeling and must be familiar with the regulatory requirements for replicate designs. The analysis plan must be written carefully and must anticipate potential problems, such as non-convergence of the model.

Regulatory Uncertainty

Regulatory requirements for replicate designs vary by jurisdiction and have evolved over time. The individual bioequivalence criterion that motivated many replicate designs has not been fully adopted by all regulatory authorities. The study team should confirm the current regulatory requirements with the relevant authority before finalizing the study design.

Safety and Regulatory Context

Subject Safety

Replicate studies expose subjects to the study drug more times than conventional studies. The safety monitoring plan must account for the additional exposures. In the tacrolimus capsule study, the safety of single 5 mg oral tacrolimus doses administered to healthy volunteers at 7-day intervals was ascertained. The study team should monitor adverse events throughout the study and should have clear criteria for stopping the study if safety concerns arise.

Regulatory Acceptance

The acceptability of a replicate design depends on the regulatory authority and the specific product. The European Medicines Agency has proposed methods for average bioequivalence with expanding limits for highly variable products. The FDA maintains the average bioequivalence criterion while allowing other criteria under certain circumstances. The study team should consult the relevant regulatory guidance and should discuss the proposed design with the authority before starting the study.

Model-Integrated Approaches

Advances in quantitative modeling and simulation have expanded the role of model-generated information in generic drug development from a supportive role toward providing critical regulatory evidence. Model-Integrated Bioequivalence represents a focused application of this paradigm in which mechanistic or empirical models are used to directly support bioequivalence determination. While physiologically based pharmacokinetic and physiologically based biopharmaceutics modeling approaches have been widely discussed in the literature, increasing attention is being directed toward population pharmacokinetic modeling for Model-Integrated Bioequivalence implementation, particularly when mechanistic assumptions are uncertain or extensive in vitro characterization is impractical. These approaches may complement or, in some cases, reduce the need for replicate designs, but they are not yet a substitute for well-conducted clinical studies in most regulatory contexts.

Professional Escalation Criteria

A study team should escalate a replicate bioequivalence study to a higher level of review or seek external consultation in the following situations:

• The mixed-effects model fails to converge and the protocol does not specify an alternative analysis approach.
• The within-subject variability estimates are substantially different from the values assumed in the sample size calculation.
• The dropout rate exceeds the level anticipated in the protocol and threatens the statistical power.
• The bioequivalence decision is borderline and depends on the choice of statistical method or software.
• The regulatory authority raises questions about the study design or analysis that the study team cannot resolve internally.
• The safety monitoring identifies a pattern of adverse events that requires further investigation.

In these situations, the study team should document the issue, consult with a biostatistician or regulatory specialist, and consider whether the study protocol needs to be amended or the study needs to be repeated.

Frequently Asked Questions

What is the difference between a full replicate and a partial replicate design?

A full replicate design gives each subject both the test and reference products twice, usually in four periods. A partial replicate design gives each subject the reference product twice and the test product once, usually in three periods. The full replicate design provides within-subject variability estimates for both formulations and can detect a subject-by-formulation interaction. The partial replicate design provides the reference within-subject variability needed for reference-scaled bioequivalence limits but does not provide the test within-subject variability.

When should a replicate design be used instead of a conventional two-period crossover?

A replicate design should be considered when the drug product is expected to have high within-subject variability, typically a coefficient of variation of 30 percent or more. A replicate design may also be appropriate when the regulatory authority requires information about within-subject variability or subject-by-formulation interaction, or when a conventional crossover has failed to demonstrate bioequivalence and the reason is suspected to be high variability.

How does a replicate design help with highly variable drug products?

A replicate design provides an estimate of the reference within-subject variability, which can be used to scale the bioequivalence acceptance limits. For highly variable products, the acceptance range may be widened from the standard 80 to 125 percent to a wider range that depends on the reference variability. This approach can reduce the sample size needed to demonstrate bioequivalence.

What is a subject-by-formulation interaction and why does it matter?

A subject-by-formulation interaction exists when the difference between the test and reference products varies across subjects. Some subjects may absorb the test product better than the reference product, while other subjects show the opposite pattern. This interaction is relevant for switchability because it means that the relative performance of the products is not consistent across the population. A replicate design can detect this interaction because it provides repeated measurements of each formulation within each subject.

Can a product pass average bioequivalence but fail individual bioequivalence?

Yes. A product can have average pharmacokinetic parameters that are within the bioequivalence limits while having greater within-subject variability or a significant subject-by-formulation interaction. In the methylphenidate study, the test product passed the average bioequivalence criterion but failed the individual bioequivalence criterion for maximum concentration because the intrasubject variability of the generic product was greater and the subject-by-formulation interaction variance was borderline high.

What are the main operational challenges of a replicate study?

The main operational challenges are the longer study duration, the higher subject burden, the increased risk of dropout, and the greater cost. Each subject must complete more study periods and provide more blood samples. The study team must plan carefully to minimize the risk of protocol deviations and missing data.

How is the sample size for a replicate study determined?

The sample size is determined based on the expected within-subject variability of the primary pharmacokinetic parameters, the expected geometric mean ratio, and the desired statistical power. For highly variable drugs, a replicate design may require a smaller sample size than a conventional crossover because it provides more information per subject. Genotype-based enrichment strategies can further reduce the sample size for drugs with variability related to polymorphic metabolizing enzymes.

What should be done if the statistical model fails to converge?

The study protocol should predefine the procedure for selecting an appropriate statistical analysis if the primary model fails to converge. In the desmopressin study, the initial model testing for individual bioequivalence failed to converge, and a traditional mixed analysis of variance based on population averages was used instead. The analysis plan should specify alternative approaches and the criteria for choosing among them.

Related Articles

• Choosing a Bioinformatics Master's Program: A Decision Framework
• Amplicon Sequencing for Viral Surveillance: Design and Quality Considerations
• RNA-seq Library Preparation: Study Design and Quality Control
• RNA-seq Library Preparation: Study Design and Quality Control
• RNA-seq Library Preparation: Study Design and Quality Control

References and Further Reading

• Research Data Framework. National Institute of Standards and Technology.
• EQUATOR Network. EQUATOR Network.
• Experimental Design Assistant. NC3Rs.
• NCBI Literature Resources. National Center for Biotechnology Information.
• PubMed. National Library of Medicine.
• A replicate study design for testing bioequivalence: a case study on two desmopressin nasal spray preparations.. European journal of clinical pharmacology, 2003.
• Individual bioequivalence revisited.. Clinical pharmacokinetics, 2001.
• Reference Datasets for Studies in a Replicate Design Intended for Average Bioequivalence with Expanding Limits.. The AAPS journal, 2020.
• Variability and impact on design of bioequivalence studies.. Basic & clinical pharmacology & toxicology, 2010.
• Genotype-based enrichment study design for minimizing the sample size in bioequivalence studies using tolterodine and CYP2D6 genotype
.. International journal of clinical pharmacology and therapeutics, 2019.
• Bioequivalence of methylphenidate immediate-release tablets using a replicated study design to characterize intrasubject variability.. Pharmaceutical research, 2000.
• Bioequivalence of 1 and 5 mg tacrolimus capsules using a replicate study design.. Journal of clinical pharmacology, 1999.
• A replicate designed bioequivalence study to compare two fixed-dose combination products of artesunate and amodiaquine in healthy chinese volunteers.. Antimicrobial agents and chemotherapy, 2014.
• Comparative bioequivalence of 1 mg and 4 mg tacrolimus extended-release tablets: randomized, open label, fully replicate, crossover trials in healthy adult subjects.. 2026.
• Model-Integrated Bioequivalence (MIBE) in Generic Drug Research: Can We Ease the Bioequivalence Burden?. 2026.
• Bioequivalence study of testosterone undecanoate soft capsules in healthy postmenopausal women under fed conditions: a single-center, four-period, repeated crossover trial.. 2026.
• Kullback-leibler Divergence for Evaluating Bioequivalence: Ii. Simulation and Retrospective Assessment of Performance of the Proposed Fda Metric and the Kullback-leibler Divergence in Individual Bioequivalence Assessment Ii. Simulation and Retrospective Assessment of Performance of the Proposed Fda
• Comparison of the reference scaled bioequivalence semi-replicate method with other approaches: focus on human exposure to drugs.. European Journal of Pharmaceutical Sciences, 2009.
• Subject-by-Formulation Interaction in Determinations of Individual Bioequivalence: Bias and Prevalence. Pharmaceutical Research, 2004.
• Alternative Cross-over Designs for Individual Bioequivalence. 2001.
• Triplicate design bioequivalence studies. Acta Pharmaceutica Turcica, 2003.
• A limited sampling approach in bioequivalence studies: Application to long half-life drugs and replicate design studies. International Journal of Clinical Pharmacology and Therapeutics, 1999.
• Non-traditional study designs to demonstrate average bioequivalence for highly variable drug products.. European Journal of Clinical Pharmacology, 2001.

This article is educational and does not replace institutional policy, professional advice, or applicable safety and regulatory requirements.