| 张红,郑小燕,陈倩立,等.基于 LASSO回归的 ART治疗后 HIV感染者低病毒血症预测模型的建立与验证[J].安徽医药,2025,29(12):2389-2394. |
| 基于 LASSO回归的 ART治疗后 HIV感染者低病毒血症预测模型的建立与验证 |
| Establishment and validation of a LASSO regression model for predicting hypoviremia in HIV-infected patients after ART therapy |
| |
| DOI：10.3969/j.issn.1009-6469.2025.12.012 |
| 中文关键词: 获得性免疫缺陷综合征 人类免疫缺陷病毒 低病毒血症 回归分析 抗逆转录病毒疗法 因果律 |
| 英文关键词: Acquired immunodeficiency syndrome Human immunodeficiency virus Low-level viraemia Regression analysis Antiretroviral therapy Causality |
| 基金项目: |
|
| 摘要点击次数: 665 |
| 全文下载次数: 258 |
| 中文摘要: |
| 目的调查人类免疫缺陷病毒( HIV)感染者低病毒血症( LLV)情况，建立预测模型并进行验证。方法回顾性选择 2021年 1月至 2024年 5月在重庆大学附属三峡医院行抗逆转录病毒疗法( ART)治疗的 339例 HIV感染者资料进行回顾性分析。参照过往研究，以 7∶3比例用 R语言软件中 “caret”包将病人以随机抽签法分为模型组 237例，验证组 102例。收集可能影响 HIV感染者 LLV的相关因素，根据是否存在 LLV将模型组病人分为 LLV组与非 LLV组，比较两组病人基本情况、 HIV感染疾病及治疗相关情况，以 LASSO筛选潜在预测因子后用 logistic回归建立模型，用列线图进行可视化，并对模型进行验证。结果模型组 237例中共有 29.96%(71/237)出现 LLV。LASSO回归基础上行多因素 logisitic回归分析结果显示:年龄、感染确诊至治疗时间、初始治疗方案、治疗方案是否改变基线 CD4+T细胞计数、随访前 7d药物漏服次数为 HIV感染者 LLV的相关因素( P<0.05)。模型组受试者操作特征曲线( ROC曲线)下面积为 0.85，95%CI为( 0.80，0.91)；验证组 ROC曲线下面积为 0.84，95%CI为( 0.79，0.90)；准确度:模型曲线与理想模型曲线基本拟合成对角线。临床有效性分析结果显示当预测概率阈值 0.05~0.81时，使用模型预测 HIV感染者 LLV的净获益最高。结论 HIV感染者 LLV主要受年龄、感染确诊至治疗时间、初始治疗方案等因素的影响，该研究列线图模型用于预测 HIV感染者 LLV风险具有较高的准确性与区分度。 |
| 英文摘要: |
| Objective To investigate low-level viraemia (LLV) in Human immunodeficiency virus (HIV)-infected patients, to develop a predictive model and to validate it. Methods The data of 339 HIV-infected patients who were treated with antiretroviral therapy(ART) in Three Gorges Hospital Affiliated to Chongqing University from January 2021 to May 2024 were selected for retrospective anal-ysis. Referring to previous studies, patients were randomly divided into 237 cases in the model group and 102 cases in the validationgroup in a 7∶3 ratio using the "caret" package in R language software with random lottery. Relevant factors that may affect LLV in HIV-infected patients were collected, and patients in the model group were divided into LLV and non-LLV groups based on the pres-ence of LLV, comparing the basic conditions of the 2 groups, HIV-infected diseases and treatment-related conditions, and the modelwas built by Logistic regression after screening potential predictors by LASSO, and the model was visualised with a column-line dia-gram and validated.Results A total of 29.96% (71/237) of the 237 in the model group developed LLV. The results of multifactorial lo-gisitic regression analysis based on LASSO regression showed that age, time from diagnosis of infection to treatment, initial treatmentregimen, whether the treatment regimen was changed, baseline CD4+ T-cell count and the number of missed doses of medication in the first 7d of follow-up were the correlates of LLV in HIV-infected patients (P<0.05). The area under the receiver characteristic (ROC curve) curve in the model group was 0.85, with a 95%CI of (0.80,0.91); the area under the ROC curve in the validation group was 0.84, with a 95%CI of (0.79,0.90); and as for accuracy, the model curve was basically fitted to a diagonal line with that of the ideal model.The results of clinical validity analysis showed the highest net benefit of predicting LLV in HIV-infected patients using this study's model when the prediction probability threshold was 0.05~0.81.Conclusion LLV in HIV-infected patients is mainly influenced byage, time from diagnosis of infection to treatment, and initial treatment regimen, and the column-line graphical model investigated in this study was used to predict the risk of LLV in HIV-infected patients with a high degree of accuracy and discrimination. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
