End-of-treatment biomarkers and HBV genotype predict severe biochemical flares after nucleos(t)ide analogue cessation: individual patient level meta-analysis of five prospective studies within the PROSTOP consortium

Abstract

Background and aims: Flares frequently occur after nucleos(t)ide analogue (NUC) cessation, necessitating close follow-up. Identifying end-of-treatment (EOT) predictors could improve safety. We assessed patient characteristics and EOT biomarkers to predict severe biochemical flares (SBFs; ALT >10×ULN). Method: Individual patient-level data from five prospective NUC stop studies within the PROSTOP consortium were pooled. All patients were HBeAg negative, long-term virologically suppressed, and none had advanced liver fibrosis. Multivariate Fine-Gray models for SBFs were fitted with retreatment as competing risk and stratification by study. Age, sex, ethnicity, EOT HBsAg and NUC type were evaluated in the full cohort. HBV genotype (gt) and EOT hepatitis B core related antigen (HBcrAg), hepatitis B core IgG antibodies (anti-HBc) and HBV RNA were evaluated in available subcohorts, adjusting for identified predictors. Predictive performance was assessed by using the time-dependent area under the ROC curve (AUC) at 48 weeks. Results: Among 402 patients, 83 experienced an SBF, corresponding to a cumulative incidence of 18.7% within 48 weeks. During follow-up, 113 (28.1%) patients were retreated, one (0.2%) underwent liver transplantation, and no liver-related deaths occurred. In the full cohort, tenofovir showed a borderline non-significant association with SBF risk (SHR = 1.65, p = 0.0504), with earlier flares compared with entecavir. Male sex was associated with increased SBF-risk (SHR = 2.38, p = 0.002). Overall, log HBsAg levels were not associated with SBF occurrence (p > 0.05), however no patients with HBsAg <10 IU/mL developed a SBF. Ethnicity was not associated with SBF risk. Gt A (SHR = 0.06, p = 0.007) and higher anti-HBc levels (SHR = 0.33, p < 0.001) were protective, while detectable HBcrAg (SHR = 2.67, p = 0.002) and HBV RNA (SHR = 2.2, p = 0.02) increased SBF risk. Prediction based on age, sex, ethnicity, HBsAg and NUC type yielded an AUC of 0.62. Models incorporating gt, HBcrAg, and anti-HBc achieved an AUC of 0.77, increasing to 0.83 with addition of sex, NUC type and HBV RNA. Conclusion: In this pooled analysis of five prospective NUC cessation studies, genotype A and EOT anti-HBc were the strongest protective factors against post-cessation SBFs, whereas EOT HBcrAg and HBV RNA were the most important risk factors. A biomarker-based model substantially improved prediction of SBFs, and may enable safer, individualized NUC cessation strategies.