Bayesian Network Analysis of Adverse Drug Reactions in Adeno-Associated Virus Gene Therapy: Exploring Patient and Serotype Influences

AAV, as a potential new vector in the gene therapy, has been widely used in clinical trials recent years. However, systematic studies that investigate the causality between patient characteristics, different AAV types, and adverse drug reactions (ADRs) are still quite limited. This research seeks to investigate and understand the relationship between the types of ADRs, the severity of these reactions, and various influencing factors. These factors include patient demographics, such as age, sex, race and specific AAV serotypes. To achieve this goal, a Bayesian Network (BN) model was developed. The purpose of this model is to analyze and explain the causal connections between the identified factors and the occurrence of ADRs. Clinical data from a total of 27 different trials were collected and then utilized for the construction of this BN model. The data included variables such as patient age, sex, race, the specific AAV serotype used, dosage levels, and the severity of observed ADRs. These variables were carefully categorized into different layers within the model. The BN model is able to calculate the posterior probability of ADR occurrences based on the given inputs. The findings of this study suggest that children and patients treated with the AAV2 serotype face a higher risk of experiencing ADRs. The BN model is an effective tool that captures complex causal relationships and provides a valuable method for assessing the safety of AAV-based gene therapies. These findings are important as they offer new insights into ADR risks and help improve the safety protocols for future gene therapy applications