Multiparametric models based on variability in lipid tests and pro-inflammatory biomarkers for atrial fibrillation and stroke prediction: The Hong Kong Family Medicine Study

Abstract Objective The objective of this study was to investigate possible associations between measures of variability in lipid tests for atrial fibrillation and stroke prediction, and develop predictive models incorporating these metrics. Methods This was a retrospective population-based cohort study patients attending family medicine clinics from 1st January 2000 to 31st December 2003 with three or more lipid tests in Hong Kong, China, with longitudinal follow-up until 31st Dec 2019. Clinical and medical data including demographics, past comorbidities, blood pressure, medications, and laboratory tests were extracted. The outcomes were atrial fibrillation, transient ischaemic attack/stroke, cardiovascular and all-cause mortality. Cox regression was used to identify significant predictors of the adverse outcomes. Subgroup analyses based on age and gender were conducted according to mean and variability of total cholesterol. Results A total of 104936 patients were initially identified. After exclusion, 86588 patients remained (39% males, median age 63 years old [IQR: 52-72]) with a median follow-up of 6203 days (IQR: 5586-6573). Conclusions Lower mean and higher variability of total cholesterol were significantly predictive with new onset new onset atrial fibrillation, stroke cardiovascular and all-cause mortality. Pro-inflammatory biomarkers provided additional prognostic diagnosis value when combined with lipid profiles and their measures of variability. The predictive values were stronger for older and male patients.