Bayesian Model Ties Liver Follow-Up Intensity to Patient Adherence

Researchers have built a mathematical decision framework to answer a question that has long frustrated hepatologists: how often should patients with mild-to-moderate fibrosis from metabolic dysfunction-associated steatotic liver disease (MASLD) be checked for disease progression? The condition affects more than one-third of the global population, yet follow-up schedules for patients in the F1–F2 fibrosis stage remain undefined, producing both wasted resources and the risk that worsening disease goes undetected.

The new analysis, published in Frontiers in Public Health, models the follow-up decision as a physician–patient interaction under uncertainty about whether the patient is actually adhering to treatment. The team, led by Artem Boltyenkov of Siemens Healthcare Diagnostics with collaborators at Virginia Commonwealth University, Northwell Health, CUNY, and other institutions, compared three strategies: aggressive follow-up every six months (AF), conservative follow-up without mandatory Enhanced Liver Fibrosis (ELF) testing between visits (CF), and conservative follow-up with mandatory between-visit ELF monitoring (CF+ELF).

The framework draws its conceptual structure from game theory, incorporating the ideas of adverse selection, moral hazard, and signaling. In plain terms, physicians cannot directly observe whether a patient is following lifestyle advice or taking medication, so they must make monitoring decisions under uncertainty about patient behavior. The quantitative core, however, is a Bayesian-informed expected net monetary benefit calculation, which weighs the costs and health payoffs of each strategy while updating beliefs about adherence.

Payoffs were grounded in empirical data rather than assumptions alone. Fibrosis transition probabilities came from published clinical trials, health-related quality of life was quantified using US-specific EQ-5D-5L utility values, and costs were anchored to Medicare reimbursement rates. This grounding matters because the value of more frequent monitoring depends on how likely a patient’s fibrosis is to progress silently and how much an earlier detection is worth in both health and money.

The model produced two physician preference thresholds. When the estimated probability of adherence falls below 0.559, aggressive six-monthly follow-up is preferred over conservative follow-up. When that probability exceeds 0.787, conservative follow-up with mandatory between-visit ELF monitoring becomes the preferred strategy. Between these thresholds lies an intermediate zone in which conservative follow-up without mandatory ELF testing is model-preferred, a region the authors suggest may suit future therapies with intermediate adherence profiles.

Applying real-world adherence proxies sharpened the picture. Estimates of roughly 30 percent adherence for lifestyle modification put those patients firmly in the range where aggressive follow-up is model-preferred. By contrast, adherence estimates of 83 to 85 percent for MASLD-targeted pharmacotherapy, including the newer agents resmetirom and semaglutide, place those patients near the upper threshold. For patients on semaglutide or resmetirom, the model favors CF+ELF only if mandatory between-visit ELF testing reduces the probability of progression among non-adherent patients by at least 68 to 77 percent; otherwise, conservative follow-up without mandatory ELF is preferred.

One of the study’s more striking conclusions concerns whose preferences actually count. On health outcome grounds, patients weakly prefer aggressive follow-up, and strictly prefer it when they are non-adherent, because earlier detection mitigates progression. The observed tendency of patients to favor conservative follow-up therefore likely reflects the burden of visits and out-of-pocket costs rather than genuine health preferences, a distinction the authors argue should inform how follow-up policies are designed.

The practical upshot is that follow-up intensity in F1–F2 MASLD should be calibrated to the treatment context and the estimated probability of adherence, rather than set by a one-size-fits-all rule. A patient relying on lifestyle change alone, with adherence hovering near 30 percent, warrants closer surveillance than a patient with high adherence to pharmacotherapy, for whom the marginal value of frequent visits may be small unless between-visit testing meaningfully changes the risk of undetected progression.

Looking forward, the authors propose a hybrid Bayesian-informed decision approach that combines patient signaling with physician oversight informed by non-invasive testing and social determinants of health. Such a framework, they argue, could adapt across both lifestyle-based and pharmacological treatment settings, updating the physician’s belief about adherence as new information arrives and adjusting monitoring accordingly.

The work is a modeling study, and its thresholds depend on the trial-derived transition probabilities, utility values, and reimbursement rates used to populate it; the authors note that all claims are their own and do not necessarily represent their affiliated organizations. Still, by making the adherence uncertainty at the heart of MASLD follow-up explicit and quantifiable, the framework offers clinicians and policymakers a structured way to balance surveillance intensity against cost, and a reminder that the best monitoring schedule may depend as much on patient behavior as on the biology of the disease.

Subject of Research: Bayesian decision modeling of follow-up strategies for mild-to-moderate fibrosis in metabolic dysfunction-associated steatotic liver disease under patient adherence uncertainty

Article Title: A Bayesian decision framework for F1–F2 MASLD follow-up under adherence uncertainty

Article References: Boltyenkov, A. T., Eiswerth, M., Sanelli, P. C., Lazarus, J. V., Wang, J. J., Chen, W., Kao, S.-Y., Chalfin, D. B., Sangha, K., & Sanyal, A. J. (2026). A Bayesian decision framework for F1–F2 MASLD follow-up under adherence uncertainty. Frontiers in Public Health, 14, Article 1930169. https://doi.org/10.3389/fpubh.2026.1930169

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1930169

Keywords: MASLD, fibrosis, Bayesian decision framework, game theory, adherence, ELF score, resmetirom, semaglutide, follow-up monitoring, health economics, expected net monetary benefit, lifestyle modification