Reply Letter to: A Critical Assessment of "Aspirin Use and Risk of HCC and Gastrointestinal Bleeding in Patients With HBV-Related Cirrhosis: A Landmark Analysis".

We appreciate the insightful comments by Mujahid and Haider regarding our article [1]. We found their critical assessment to be exceptionally valuable and insightful, as it addresses several fundamental methodological and clinical considerations that are essential for interpreting the impact of aspirin in patients with hepatitis B virus (HBV)-related cirrhosis. In response to their comments, we provide our perspectives as follows. First, as they mentioned, the potential for exposure misclassification due to the use of over-the-counter (OTC) aspirin is a valid concern. While our study relied on prescription data from the National Health Insurance Service database, which may not capture nonprescription OTC use, long-term low-dose aspirin for prevention purposes is predominantly obtained through prescriptions to ensure insurance coverage and medical supervision in the Korean clinical setting [2, 3]. Furthermore, a previous simulation study demonstrated, under many circumstances, results based on prescription claims data remain valid despite incomplete OTC exposure [4]. In addition, any such misclassification related to unrecorded OTC use would likely bias the results toward the null, suggesting the actual protective effect of aspirin against HCC is likely even more pronounced than observed in our study. We acknowledge that this remains an inherent limitation of claims-based research, and future studies incorporating pharmacy-level data or patient-reported medication use could further refine these estimates. Second, while our analysis adjusted for major clinical characteristics and comorbidities associated with both aspirin use and bleeding risk to avoid overfitting and maintain model stability, we acknowledge that residual confounding related to concomitant medication use, such as proton-pump inhibitors, nonsteroidal anti-inflammatory drugs, and anticoagulants, may remain in claims-based analyses. Further investigations incorporating more detailed pharmacological data may clarify the independent effect of aspirin on gastrointestinal (GI) bleeding risk in patients with cirrhosis. Third, the authors' comment regarding the potential for surveillance bias and the “healthy-user” effect is highly relevant. Although we attempted to minimize these biases by adjusting for baseline health status and health-seeking behaviors, these challenges are inherent to large-scale observational databases. However, because our study population consisted of patients with HBV-related cirrhosis who generally receive regular clinical follow-up, differential health-seeking behavior between aspirin users and nonusers may be less pronounced than in general population cohorts. Nevertheless, we acknowledge that surveillance bias and healthy-user effects cannot be completely excluded in observational studies using administrative data. Future prospective studies with detailed clinical information may further clarify these associations. Fourth, regarding the choice of modeling strategy, we selected a landmark analysis specifically to minimize immortal time bias, which is a critical concern in longitudinal studies evaluating drug effects. This approach allowed us to define exposure status based on a clinically meaningful observation period and to ensure comparability between aspirin users and nonusers after the landmark time point. While this approach does not fully capture changes in aspirin use after the landmark time point, it provides a stable and clinically interpretable exposure definition for evaluating long-term outcomes [5, 6]. Although time-dependent exposure modeling represents an alternative analytical strategy for handling dynamic medication use, the landmark approach remains a well-established method in pharmacoepidemiologic research and was appropriate for the primary objective of our study. We agree that future studies may apply complementary analytic approaches, including time-varying exposure modeling, to further evaluate dynamic patterns of aspirin use. Finally, we agree with the importance of detailed outcome definitions for GI bleeding. As noted, our study utilized ICD-10 codes to identify GI bleeding events, which did not allow differentiation between upper versus lower GI bleeding, or variceal versus non-variceal bleeding. In addition, the administrative nature of our database limited our ability to assess bleeding severity, such as transfusion requirement or clinical management. Nevertheless, the use of standardized diagnostic codes for major clinical events is a common and accepted approach in nationwide claims-based studies. We recognize that more granular clinical data from hospital-based registries may help further characterize the safety profile of aspirin in cirrhosis populations in future research. Although our population-based study has certain limitations, the robustness of our findings is supported by the large nationwide cohort and the use of a landmark design to minimize immortal time bias. We hope that future studies incorporating more detailed clinical information will further clarify the benefit–risk balance of aspirin in patients with HBV-related cirrhosis and help refine hepatocellular carcinoma prevention strategies in this high-risk population.