Stochastic Dynamics of Hepatitis B Virus Infection: Extinction, Persistence, and Ergodicity

Abstract

We develop and analyze a three-compartment stochastic model of within-host hepatitis B virus infection with multiplicative It\^o noise. We establish global well-posedness, strict positivity, and uniform moment bounds, and derive an explicit sufficient criterion for almost sure exponential clearance. Standard extinction arguments based on $\ln(y+z)$ are conservative and fail to recover the deterministic spectral threshold. To overcome this limitation, we optimize over the weighted family $\ln(\alpha y+z)$ and control hepatocyte fluctuations through exact occupation identities for $x+y$, avoiding both equilibrium substitution and an invalid pathwise comparison. The resulting general extinction bound admits a reduced optimized form under an explicit admissibility condition. This reduced bound lies no more than $\tfrac12\max\{\sigma_2^2,\sigma_3^2\}$ below the spectral abscissa $\lambda_{\max}(J)$ of the deterministic infected subsystem and converges to $\lambda_{\max}(J)$ when $\sigma_2,\sigma_3\to0$ with $\sigma_1^2/(\sigma_2^2+\sigma_3^2)\to0$. We further show that a noise-corrected reproduction number above unity guarantees the existence of a unique ergodic invariant measure in the positive orthant, and that the sufficient extinction and persistence regimes are mutually exclusive. Numerical experiments illustrate the theoretical results and quantify the effects of stochastic fluctuations on within-host viral dynamics.