When patients with chronic hepatitis B (CHB) stop nucleos(t)ide analogue (NA) ther-apy before achieving HBsAg loss, flares often ensue which are challenging to pre-dict early. We determined the incidence, severity, outcome and predictors of flares after NA withdrawal. Forty-five patients enrolled in an RCT were included; 107 pa-tients from an external, prospective cohort were used for validation. Retreatment criteria were pre- defined. Pre- and post-treatment predictors of alanine aminotrans-ferase (ALT) flare (>5× ULN) were evaluated by Cox proportional- hazards regression. Seventy-two weeks after NA withdrawal, 23/45 (51%) patients had developed >5×ULN and 14 (31%) >20× ULN. Median time to develop ALT >5× ULN was 12 weeks after NA withdrawal. Independent predictors of ALT >5× ULN were male sex (HR [95% CI] 3.2 [1.2–8.9]; p= 0.03) and serum HBV DNA (1.2 [1.0–1.8]; p= 0.03) at Week 6 off-therapy. Specifically, week 6 HBV DNA >10,000 IU/ml predicted ALT >5×ULN (3.4 [1.4–8.4]; p= 0.01), which was externally validated. In conclusion, this study on post-treatment flares revealed a high cumulative incidence in CHB. Week 6 HBV DNA >10,000 IU/ml independently predicted flares. The proposed threshold enables prediction of imminent flares in patients who may benefit from closer monitoring and earlier retreatment.
