摘要
目的 利用体检人群数据库,通过最小化绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)回归算法筛选与颈动脉斑块形成的关键因子,并探索血红蛋白(hemoglobin,Hb)对颈动脉斑块形成的诊断价值,通过Cox比例风险模型验证HB与颈动脉斑块形成的关联及关联强度.方法 采用海军青岛特勤疗养中心体检中心体检数据,以2014 年体检人群资料为队列随访基线,以 2020 年为随访时间终点,随访事件结局为该人群新发颈动脉斑块,利用LASSO回归算法筛选颈动脉斑块影响因素,并通过Cox比例风险模型分析HB与颈动脉斑块的关联.结果 共纳入 977 名未患有颈动脉斑块的体检人群,经过6 年体检定期随访,随访新发颈动脉斑块326 例,累计患病率为33.4%.LASSO回归筛选变量年龄和吸烟等 9 个变量时,模型均方根误差最小,其对应的λ =0.0113,LASSO回归模型筛选变量通过ROC预测新发颈动脉斑块的曲线下面积AUC为 0.762.通过Cox比例风险模型进一步验证血红蛋白是颈动脉斑块发病的独立危险预测因素.结论 体检人群高水平血红蛋白可能与颈动脉斑块发病相关,应重点关注中老年体检人群血红蛋白水平,提示较高的营养水平的人群是防控颈动脉斑块新发的重点人群.
Abstract
Objective The aim of this study was to investigate the relationship between hemoglobin(HB)levels and carotid plaque formation.Methods The physical examination population database in 2014 was used as the baseline of the cohort,and the end point of the follow-up time was in 2020.The follow-up end-event is the onset of carotid plaque.The influencing factors of carotid plaque are screened by Lasso regression algorithm,The Cox proportional hazard model was used to analyze the association of HB and carotid plaque.Results A total of 977 people without carotid plaque were enrolled.After 6 years follow-up,326 cases of carotid plaque were recorded,and the cumulative incidence was 33.4%.When lasso regression screened 9 variables such as age and smoking,the root mean square error of the model was the smallest λ = 0.011 3.The LASSO regression model screening variables predicted the area under the curve of new carotid plaques by ROC,AUC was 0.762.The Cox proportional hazard model further verified that hemoglobin was an independent risk predictor of carotid plaque.Conclusion The high level of hemoglobin in the physical examination population may be related to the high incidence of carotid plaque.We should focus on the hemoglobin level of the middle-aged and elderly physical examination population,which may suggest that the population with higher nutritional levels is a key group for preventing and controlling new occurrences of carotid artery plaques.
NETL
NSTL
万方数据