Impact of liver fibrosis score on the incidence of stroke: A cohort study

摘要

Purpose The purpose was to explore the value of liver fibrosis scores (fibrosis-4, BAAT score and BARD score) for incidence risk of stroke in a cohort study. Methods A total of 9088 participants without stroke history enrolled the follow-up. Three liver fibrosis scores (LFSs) including FIB-4, BARD score and BAAT score were adopted. The end point was stroke. Cox regression analysis was used to calculate hazard ratios and 95 confidence interval. Kaplan-Meier curve was used to show the probability of stoke in different levels of LFSs. Subgroup analysis showed the association between LFSs and stroke under different stratification. Restricted cubic spline could further explore whether there is a linear relationship between LFSs and stroke. Finally, we used C-statistics, Net Reclassification Index (NRI) and Integrated Discrimination Improvement (IDI) to assess the discriminatory power of each LFS for stroke. Results During a median follow-up time of 4.66 years, 272 participants had a stroke. Through the baseline characteristics, we observed that the stroke incidence population tends to be male and older. It was shown by Kaplan-Meier that three LFSs were associated with stroke and high levels of LFSs significantly increase the probability of stroke. In the univariate Cox regression analysis, the HR of stroke risk was 6.04 (4.14-8.18) in FIB-4, 2.10 (1.45-3.04) in BAAT score and 1.81 (1.38-2.38) in BARD score by comparing the high level with the low level at each LFSs. The adjusted HRs for three LFSs were 2.05 (1.33-3.15) in FIB-4, 1.61 (1.10-2.35) in BAAT score and 1.54 (1.17-2.04) in BARD score by comparing the high group with low group. We found that multivariable-adjusted HRs of three LFSs still increased for stroke when stratified by various factors in subgroup analysis. Moreover, after adding LFSs to original risk prediction model which consist of age, sex, drinking, smoking, hypertension, diabetes, low-density lipoprotein cholesterol, total cholesterol and triglycerides, we found that new models have higher C-statistics of stroke. Furthermore, we calculated Net Reclassification Index (NRI) and Integrated Discrimination Improvement (IDI) to show the ability of LFSs to predict stroke. Conclusions Our study showed that three LFSs were associated with stroke amongst middle-aged populations in rural areas of Northeast China. Furthermore, it suggests that LFSs can be used as a risk stratification tool to predict stroke.