RSABE divergence post-ICH M13A: a practitioner's map.

Highly variable drugs — within-subject coefficient of variation greater than 30% on AUC or Cmax — fail standard 80–125% bioequivalence with infeasible sample sizes. The regulatory response has been to permit reference-scaled average bioequivalence (RSABE), widening the bound proportional to the within-subject variability of the reference product. The framework was first operationalised by FDA via product-specific guidances starting around 2005, and EMA published its operational rules in the 2010 Guideline on Investigation of Bioequivalence revision. Both agencies built in different scaling philosophies. Both philosophies survived M13A.

/ 01The two rules that did not converge.

FDA: scale on both.

FDA permits reference-scaling on AUC and Cmax. The bounds widen proportionally to the within-subject standard deviation of the reference product, calculated from a partial replicate or full replicate design. The widened bound on Cmax can extend up to roughly 69.84–143.19% in extreme cases. The point-estimate constraint — geometric mean ratio between 80–125% regardless of CI width — applies. The reasoning is that highly variable drugs have biologically wide variability and the regulator should not penalise generic sponsors for the reference's variability.

EMA: scale Cmax only.

EMA permits scaling on Cmax up to a within-subject CV cap of 50% (corresponding to bounds of approximately 69.84–143.19%). EMA does not scale AUC. AUC stays at the standard 80–125% regardless of variability. The reasoning is that AUC variability beyond a threshold raises questions about formulation consistency that scaling masks. Both readings are scientifically defensible. Both readings produce different study designs and different acceptance outcomes.

/ 02What ICH M13A actually harmonised.

M13A — Bioequivalence for Immediate-Release Solid Oral Dosage Forms — reached Step 4 in 2024. Implementation: FDA accepts on a transition-to-binding pathway through 2026; EMA published implementation 2024 with a transition window; PMDA aligned April 2025; ANVISA following ICH cycle. The harmonisation is real and substantial — covering reference-product handling, study design, statistical methods, and reporting expectations. RSABE methodology is referenced. The divergence on AUC versus Cmax-only scaling for HV drugs is not closed.

The reason for the non-closure is documented in the M13A drafting record: ICH M13A Section 2.5 covers HV drug study design generically, with regulatory-region-specific implementation footnoted. The footnote is the divergence. M13A did not eliminate it.

/ 03What this means for a study designed for both regions.

Consider an immediate-release generic of a highly variable drug — within-subject CV on AUC of 38%, on Cmax of 46%. The sponsor wants a single Phase I crossover that supports both ANDA filing in the US and a generic Marketing Authorization Application in the EU. The constraint set:

• EU: AUC must meet standard 80–125% (cannot scale)
• EU: Cmax may scale up to CV 50% cap, widening bound proportionally
• FDA: both AUC and Cmax may scale with reference variability
• FDA: reference-scaled study requires partial replicate or full replicate design
• Both: point-estimate constraint 80–125% on geometric mean ratio

The dominant constraint is the EU AUC requirement. The study must be powered for AUC 80–125% at the observed AUC variability, not at the scaled bound. This typically pushes sample size into the 60–90 subject range for full replicate, depending on observed variability. The same study satisfies FDA scaling for AUC and Cmax automatically because it has more data than FDA needs at the scaled bound. The reverse is not true: a study designed only for FDA RSABE (smaller, scaled on AUC) cannot retroactively satisfy EMA AUC standard 80–125%.

The dominant constraint sets the design. For HV drugs filed in both regions, EMA AUC is the dominant constraint. Design to it. The FDA filing is the byproduct, not the target.

/ 04The ANVISA overlay.

ANVISA's RDC 742/2022 (effective 1 March 2023) is EMA-aligned on the HV-drug scaling philosophy — Cmax scaling permitted, AUC standard. ANVISA-specific overlays remain non-negotiable: the reference product must be locally registered in Brazil; the healthy-volunteer registry mandate (post-2024 Rules) prevents overlapping BE studies in the same volunteer pool; genotoxic impurity records for the BE clinical batch are mandatory. A study designed for EU is closer to ANVISA-acceptable than a study designed for FDA, but the operational overlays still require dedicated Brazilian-side preparation. WHO Prequalification accepts EMA-style or FDA-style scaling depending on the source-country dossier; the WHO assessor applies the stricter of WHO TRS 996 Annex 9 / ICH M13A and the source-country rule.

/ 05The operational stack for sponsors in 2026.

The pragmatic 2026 strategy for HV-drug generic sponsors filing in both FDA and EMA regions:

• Design AUC to standard 80–125% bound — covers EMA, satisfies FDA as a subset
• Scale Cmax under EMA logic up to 50% CV cap — covers EMA, satisfies FDA RSABE methodology
• Use a full replicate design — generates the reference-variability data needed for FDA RSABE while supporting EMA
• Source the reference product per jurisdiction — FDA RLD, EMA-sourced for EU, locally registered for Brazil
• Document study design rationale referencing both M13A and the regulatory-region-specific addendum so the dossier reads consistently to either regulator
• For the Brazilian filing, plan ANVISA volunteer registry pre-clearance and genotoxic impurity records for the BE clinical batch

The cross-link to the bioequivalence library covers the underlying rule set; this note is the practitioner's overlay on what M13A did and did not change.

/ 06Where the convergence goes from here.

The next opportunity for AUC-scaling convergence is the M13B / M13C work on long-acting and modified-release formulations — both at Step 2 currently with Step 4 expected 2027–2028. The political appetite for revisiting the AUC scaling debate within those workstreams is uncertain. Sponsor expectation should be that the divergence persists through at least 2028 and possibly through M13A's first revision cycle around 2030. Plan accordingly.