Editorial: Tenofovir Alafenamide Versus Entecavir for Hepatitis B Surface Antigen Decline—A Step Towards Functional Cure or a Marginal Margin? Authors' Reply

We thank Hsu and Yu [1] for their thoughtful editorial on our work [2] and welcome the opportunity to clarify several methodological points. Propensity score (PS) covariates: Although age, baseline estimated glomerular filtration rate (eGFR), and calendar year were not included in our matching model, age (56 vs. 55 years; p = 0.201) and eGFR (80.3 vs. 79.6 mL/min/1.73 m2; p = 0.787) were well balanced after matching. Calendar year reflects TAF's later Japanese approval (2017) and is intrinsic to any sequential comparison between historically introduced agents. We re-estimated PSs including these three additional covariates, yielding 318 matched pairs; the treatment-by-time interaction remained significant (global p = 0.007), with consistent direction and magnitude (Table 1). Unequal follow-up: The longer ETV exposure reflects earlier market availability rather than informative censoring. Linear mixed-effects models with maximum-likelihood estimation provide unbiased inference under missing-at-random [3]. Restricting the model to 0–3 years, when follow-up was comparable between groups, preserved TAF's advantage (Table 1). We re-fit the primary model with random intercepts and random slopes for time. The slope variance was small and the likelihood-ratio test was not significant (p = 0.21); the fixed-effect treatment-by-time interaction was essentially unchanged. Between-patient heterogeneity in HBsAg trajectories therefore appears modest in this cohort. The cohort was strictly treatment-naïve, and inter-NA switching was an exclusion criterion. Although pill counts were unavailable, closely comparable HBV-DNA suppression and ALT-normalisation trajectories indicate clinically adequate adherence in both arms. We agree that pre-specified interaction tests outperform within-stratum hypothesis testing [4]. Our subgroup analyses were pre-specified; nevertheless, we emphasise their exploratory, hypothesis-generating role. Divergent HBsAg and HBcrAg trajectories may reflect preferential TAF activity against HBsAg derived from integrated HBV DNA [5] or immunomodulation independent of cccDNA suppression [6]. Quantitative HBV RNA, which more directly reflects cccDNA transcription [7], is being collected prospectively in our ongoing cohort. Causal inference: We endorse target-trial emulation with explicit time-zero anchoring, expanded covariate sets and inverse-probability weighting [8] as the appropriate framework for future observational comparisons. An IPTW analysis with stabilised weights yielded concordant results (Table 1). Recent real-world data [9] similarly support TAF's quantitative advantage. As Hsu and Yu acknowledge, randomised trials powered for functional cure are unlikely in the foreseeable future; rigorously designed observational studies must therefore carry the evidential burden. In summary, all sensitivity analyses reinforce our primary conclusion: TAF produces a modestly but consistently steeper HBsAg decline than ETV. Whether this surrogate signal translates into higher functional cure rates—rather than a marginal margin—is the appropriate next question, and one we are actively pursuing through extended follow-up with clinically meaningful endpoints [10]. Takashi Kumada: conceptualization, methodology, writing – review and editing, writing – original draft, data curation. Masanori Atsukawa: conceptualization, methodology, data curation. Taeang Arai: conceptualization, methodology, data curation. Hidenori Toyoda: data curation, methodology, conceptualization. Junko Tanaka: supervision, formal analysis. During manuscript preparation, the author used ‘Claude Opus 4.7’ for language editing and readability improvement. The author reviewed and edited all AI-assisted content and takes full responsibility for the publication. The authors have nothing to report. This article is linked to Kumada et al. papers. To view these articles, visit https://doi.org/10.1111/apt.70704 and https://doi.org/10.1111/apt.70734. The data that support the findings of this study are available from the corresponding author upon reasonable request.