Purpose <p>U.S. Food and Drug Administration (FDA) recommends reference scaled limits and variability comparison for bioequivalence (BE) demonstration of narrow therapeutic (NTI) drugs while most other regulatory agencies apply direct tightening of BE limits. This study evaluates strengths and limitations of different BE criteria using abbreviated new drug application (ANDA) data received by FDA to support harmonization of NTI drug BE criteria.</p> Methods <p>We analyzed four-way fully-replicated crossover BE study data in NTI ANDAs, applying BE criteria from different agencies and alternative criteria (e.g., Paixão’s criteria, modified Paixão’s and FDA criteria) to compare passing rates.</p> Results <p>Current EMA and FDA criteria seem stringent with NTI drugs having moderate and low within-subject variability (e.g., S_WR), respectively. Capping BE limits at 90.00–111.11% when σ_WR ≤ 0.10 in alternative FDA criteria improved passing rates and better aligns with NTI quality standards. An additional point estimate constraint of 90.00–111.11% enforced geometric mean ratio closer to 1 but reduced passing rates when S_WR was moderate to high. FDA’s regulatory constant resulted in slightly less stringent scaled BE limits than Paixão’s, but may better align with observed S_WR ranges for NTI drugs. Alpha adjustment reduced Type I error but slightly decreased study passing rates.</p> Conclusion <p>Alternative FDA criteria with capping at 90.00–111.11% when σ_WR ≤ 0.10 and applying alpha adjustment provides reasonably stringent standards for BE demonstration of NTI drugs. This in-depth analysis of ANDA BE data will help ICH M13C Expert Working Group make informed decisions about harmonization options for BE demonstration of NTI drugs.</p> Graphical Abstract <p></p>