Bayesian network modelling study to identify factors influencing the risk of cardiovascular disease in Canadian adults with hepatitis C virus infection

Abstract

Objectives The present study evaluates the extent of association between hepatitis C virus (HCV) infection and cardiovascular disease (CVD) risk and identifies factors mediating this relationship using Bayesian network (BN) analysis. Design and setting A population-based cross-sectional survey in Canada. Participants Adults from the Canadian Health Measures Survey (n=10 115) aged 30 to 74 years. Primary and secondary outcome measures The 10-year risk of CVD was determined using the Framingham Risk Score in HCV-positive and HCV-negative subjects. Using BN analysis, variables were modelled to calculate the probability of CVD risk in HCV infection. Results When the BN is compiled, and no variable has been instantiated, 73%, 17% and 11% of the subjects had low, moderate and high 10-year CVD risk, respectively. The conditional probability of high CVD risk increased to 13.9%±1.6% (p-16) when the HCV variable is instantiated to ‘Present’ state and decreased to 8.6%±0.2% when HCV was instantiated to ‘Absent’ (p-16). HCV cases had 1.6-fold higher prevalence of high-CVD risk compared with non-infected individuals (p=0.038). Analysis of the effect modification of the HCV-CVD relationship (using median Kullback-Leibler divergence; D_KL) showed diabetes as a major effect modifier on the joint probability distribution of HCV infection and CVD risk (D_KL=0.27, IQR: 0.26 to 0.27), followed by hypertension (0.24, IQR: 0.23 to 0.25), age (0.21, IQR: 0.10 to 0.38) and injection drug use (0.19, IQR: 0.06 to 0.59). Conclusions Exploring the relationship between HCV infection and CVD risk using BN modelling analysis revealed that the infection is associated with elevated CVD risk. A number of risk modifiers were identified to play a role in this relationship. Targeting these factors during the course of infection to reduce CVD risk should be studied further.