参比标度平均生物等效性（Reference-scaled Average Bioequivalence, RSABE）方法

Reference-scaled average bioequivalence (RSABE) is a statistical approach increasingly adopted to assess bioequivalence for highly variable drugs (HVDs). These are drug products for which the within-subject (intra-subject) variability in pharmacokinetic measures—AUC and/or C_max—exceeds 30% coefficient of variation (C.V.).

In simpler terms, when the same individual takes the same drug under nearly identical conditions (e.g., same dose, administration route, fasting state, and time of day), the AUC and C_max values are generally expected to be consistent. However, for highly variable drugs, the rate and extent of absorption can fluctuate by more than 30% between administrations—even under controlled conditions—making it challenging to demonstrate bioequivalence using traditional average bioequivalence (ABE) methods.

For highly variable drugs, applying the traditional average bioequivalence (ABE) approach with a standard sample size often fails to demonstrate bioequivalence—even when the test and reference products are genuinely comparable (Drug B in Figure 1). In fact, some HVDs have failed to show bioequivalence even to themselves under standard ABE designs.

This is because high intra-subject variability can obscure real similarities between products, necessitating very large sample sizes to achieve statistical power. Such studies are not only more expensive and time-consuming but also expose more participants to unnecessary risks. As a result, the development and approval of generic versions of HVDs can be significantly delayed or even abandoned, reducing access to affordable alternatives.

The RSABE approach addresses this challenge by scaling the bioequivalence limits according to the within-subject variability of the reference drug. In essence, the permitted acceptance range widens as variability increases, reflecting the inherent pharmacokinetic variability of the drug rather than penalizing it. RSABE can be applied when the within-subject C.V. for the reference product is at least 30%, allowing for a more practical and scientifically justified path to establishing bioequivalence for HVDs.

The implementation of reference-scaled average bioequivalence (RSABE) methodology varies by regulatory agency. However, both the European Medicines Agency (EMA) and the U.S. Food and Drug Administration (FDA) have a common requirement which is the use of replicated crossover study designs in which each subject receives the reference product more than once. This repetition is essential to accurately estimate the within-subject variability of the reference drug.

Examples of common replicated crossover designs include:

• Three-period designs, such as RRT, RTR, or TRR
• Four-period designs, such as RTRT or TRTR

These replicated designs include the reference treatment in at least two study periods, which is essential for accurately estimating the within-subject standard deviation of the reference product (S_WR)—a key requirement for determining whether the RSABE approach can be applied.

RSABE may be used when the within-subject coefficient of variation (CVWR) for the reference product exceeds 30%, which corresponds to a within-subject standard deviation (S_{WR</sub) ≥ 0.294.}

FDA Requirements:

• If S_WR 0.294, standard Average Bioequivalence (ABE) criteria apply.
• RSABE may be applied if S_WR ≥ 0.294 for AUC and/or C_max.
• The 90% confidence interval for the geometric mean ratio may be widened accordingly. However, the point estimate (geometric mean ratio) must still fall within 80–125%.
• Studies must include at least 24 subjects.
• The intention to apply widened bioequivalence limits must be predefined in the protocol before the study is conducted.

EMA Requirements:

• AUC must always be evaluated using traditional ABE with 80–125% confidence interval limits.
• RSABE can be applied to C_max if S_WR ≥ 0.294.
  • In this case, the 90% CI can be widened to a maximum of 70–143%.
  • The point estimate must remain within 80–125%.
  • EMA requires a justification that the observed variability in C_max is reliable (not due to outliers) and that wider differences in C_max are clinically irrelevant.
  • The intention to apply widened bioequivalence acceptance criteria must be prospectively specified in the study protocol

To understand why the RSABE methodology is referred to as “scaled,” it helps to review the equations behind both conventional and scaled bioequivalence assessments.

In Average Bioequivalence (ABE), a drug is considered bioequivalent if the 90% confidence interval for the difference between the logarithmic means of the test and reference products lies within preset regulatory limits (80%–125%). On the log scale, this criterion is expressed as:

OR

These fixed boundaries correspond to approximately ±0.223 on the natural log scale.

For highly variable drugs (HVDs), where within-subject variability (CV_WR) is ≥30%, RSABE allows the acceptance range to be widened proportionally to the variability of the reference product. Both the difference and the limits are scaled. The limits are scaled by a regulatory constant, and the difference is scaled by the within-subject variability for the reference product.

EMA RSABE methodology [1] scales the acceptance range by the regulatory constant of 0.294 (which is the S_WR at 30% CV) and the difference is scaled by the drug’s variability S_WR

which simplifies to:

This means the permissible difference increases with variability, up to a maximum of 69.84%–143.19%.

The FDA methodology [2] applies the same conceptual scaling but uses a different regulatory constant: 0.25. In addition, the FDA uses a more complex statistical test to conclude bioequivalence (not detailed here), but the underlying scaling logic is

which simplifies to:

Therefore, the FDA allows the acceptance range to widen even further than EMA

Using the formula:

we can calculate the resulting acceptance limits determined by

respectively for EMA and FDA, at various levels of within-subject variability:

The RSABE approach has supported several approvals of highly variable generic drug products. It is becoming standard practice in the generic drug industry.

Phoenix WinNonlin can perform RSABE analysis through a series of structured steps. To simplify this process, downloadable project templates are available from the Certara Help Center. While there is no dedicated RSABE module within the software, these templates offer a practical and regulatory-aligned framework for conducting RSABE analyses.

For highly variable drugs, two primary Phoenix project templates are available—one for FDA guidelines and another for EMA. These templates support both partial and full replicate designs commonly used in RSABE studies:

• Partial Replicate Design:
  Only the 3-period design with sequences TRR, RTR, or RRT is accepted.
  (T = Test, R = Reference)
• Full Replicate Design:
  Only the 4-period design with sequences TRTR or RTRT is accepted.

⚠️ These templates are provided as examples and do not follow Certara’s formal software development and validation processes. They should be used at your own discretion and risk, particularly in regulatory submissions.

Additional example templates that use scaled approaches—such as those for narrow therapeutic index (NTI) drugs—are also available in the Certara Help Center.