#1058 Causal discovery for the volume of white matter in patients undergoing haemodialysis

Abstract Background and Aims White matter hyperintensity is linked to adverse outcomes and a heightened risk of overt stroke and cognitive impairment, with its prevalence being more pronounced in patients undergoing haemodialysis than in the general population. Despite this, the determinants of white matter hyperintensity volume in patients undergoing haemodialysis remain unknown. Here, we aimed to utilise a statistical causal search approach to examine the factors influencing white matter hyperintensity volume, as measured by MRI, in patients undergoing haemodialysis. Method MRI measurements of white matter hyperintensity volumes were taken in 133 patients on outpatient maintenance haemodialysis for >1 year. Factors included in the statistical causal discovery model encompassed sex, underlying diseases that led to dialysis, total brain volume measured using MRI, and age and dialysis vintage at the time of MRI. Additionally, values calculated and averaged over the year before the MRI scan, such as pre-dialysis blood pressure, pulse, weight, blood pressure during dialysis, blood pressure after dialysis completion, the volume of water removed in one session, Kt/V, and URR and occurrences of sudden drops in blood pressure during dialysis, were considered. Missing measurements were supplemented using random forest. A statistical causal search using DirectLiNGAM was conducted, and clear temporal pre- and post-relationships were incorporated into the LiNGAM model as per prior knowledge. To enhance internal validity, 1000 bootstrap samplings were performed, and edges with a frequency of occurrence of ≥50% were considered valid for Directed Acyclic Graphs (DAGs). Factors directly linked to white matter volume were included in the model, and linear regression was executed. Results Patients had a mean age of 69.0 (Standard Deviation: 10.4) years, with 96 (72.2%) being male. The mean history of dialysis was 6.63 [Interquartile Range (IQR): 4.10, 10.78] years, and 40 (39.2%) patients had diabetes as the primary disease. The mean white matter hyperintensity volume was 14334.9 [IQR: 5262.4,30637.2]. Age, pre-dialysis body weight, post-dialysis mean blood pressure, and total brain volume were directly influenced by white matter volume. DAGs displaying causality obtained through bootstrapping are shown in the Fig. The regression model results revealed that white matter hyperintensity increased by 8376.3 [95% Confidence Intervals (CI): 4899.5 to 11853.1] for every 10 years of age, decreased by −2726.0 [95% CI: −4311.5 to −1140.5] with a 5 kg increase in pre-dialysis body weight, increased by 2089.7 [95% CI: 693.3 to 3540.0] with a 5-mmHg rise in post-dialysis mean blood pressure, and a 1 standard deviation increase in total cerebrum volume increased white matter volume by 13057.4 [95% CI: 8958.1 to 17156.8]. Conclusion The causal diagram indicated that age, pre-dialysis weight, post-dialysis mean blood pressure, and total cerebrum volume directly influenced white matter hyper intensity volume. The sudden drop in blood pressure during dialysis impacted white matter hyperintensity volume via post-dialysis mean blood pressure. Furthermore, age, post-dialysis mean blood pressure and total cerebrum volume had a facilitative effect on the increase in white matter hyperintensity volume, while pre-dialysis weight had a protective effect.