Liang Zhihui, Zhao Min, Liu Kaiting, Liang Weican, Luo Shaofang, Guan Jianbin, Zhang Zongmian
The Department of Intensive Care Unit, The Sixth Affiliated Hospital, School of Medicine, South China University of Technology, Foshan, Guangdong Province, China.
Health Management Center, The Third Affiliated Hospital of Sun Yat-Sen University, Guangzhou, Guangdong Province, China.
Shock. 2025 Oct 1;64(4):386-396. doi: 10.1097/SHK.0000000000002653. Epub 2025 Jul 28.
Background: The evolution of lactate levels reflects the complex pathophysiological processes in sepsis. Whether distinct subclusters of sepsis exhibit different lactate trajectories remains unclear. This study aimed to identify novel clusters of sepsis based on lactate trajectories and investigate the association between lactate trajectory and mortality risk and to develop a predictive model for unfavorable lactate trajectories. Methods: Early survivors diagnosed with sepsis were included. A group-based trajectory model was constructed to identify distinct lactate trajectories. Doubly robust estimation models were utilized to assess the association between each cluster and mortality risk. A cross-lagged panel model was applied to examine the temporal causal relationship between lactate levels and Sequential Organ Failure Assessment (SOFA) score. LASSO-logistic regression was used to develop a predictive model for unfavorable lactate trajectories. Results: A total of 4,870 patients from two critical care medicine databases were included. The following 4 lactate trajectory clusters were identified: (1) hyperlactatemia, gradual resolution (cluster 1; 14.0%), (2) consistent near-normal lactate level (cluster 2; 81.5%), (3) extreme hyperlactatemia at admission but with prompt clearance (cluster 3; 2.0%), and (4) consistent hyperlactatemia (cluster 4; 2.5%). Comparisons were conducted using cluster 1 as the reference. Cluster 2 showed reduced 28-day mortality risk (hazard ratio [HR] 0.76; 95% confidence interval [CI] 0.65 to 0.89), while no difference was observed in adjusted mortality hazard risk. Clusters 3 and 4 had higher mortality risks (HR 1.94; 95% CI 1.40 to 2.67 and HR 3.87; 95% CI 2.98 to 5.03 respectively) compared to cluster 1. The cross-lagged panel model analysis showed a bidirectional causal relationship between lactate levels and organ dysfunction (Lactate→SOFA,β = 0.310, P < 0.001 vs. SOFA→Lactate,β = 0.037, P < 0.001). A nomogram with five variables was developed to identify unfavorable lactate trajectories. Conclusion: Lactate trajectories are significantly associated with mortality risk in early-survival patients with sepsis, which provides a valuable framework for risk stratification in sepsis.
乳酸水平的变化反映了脓毒症复杂的病理生理过程。脓毒症不同的亚组是否表现出不同的乳酸变化轨迹尚不清楚。本研究旨在基于乳酸变化轨迹识别脓毒症的新亚组，探讨乳酸变化轨迹与死亡风险之间的关联，并建立不良乳酸变化轨迹的预测模型。方法：纳入早期诊断为脓毒症的幸存者。构建基于组的轨迹模型以识别不同的乳酸变化轨迹。采用双重稳健估计模型评估每个亚组与死亡风险之间的关联。应用交叉滞后面板模型检验乳酸水平与序贯器官衰竭评估（SOFA）评分之间的时间因果关系。使用套索逻辑回归建立不良乳酸变化轨迹的预测模型。结果：纳入了来自两个重症医学数据库的4870例患者。识别出以下4种乳酸变化轨迹亚组：（1）高乳酸血症，逐渐消退（亚组1；14.0%），（2）乳酸水平持续接近正常（亚组2；81.5%），（3）入院时极高高乳酸血症但迅速清除（亚组3；2.0%），以及（4）持续高乳酸血症（亚组4；2.5%）。以亚组1作为对照进行比较。亚组2显示28天死亡风险降低（风险比[HR]0.76；95%置信区间[CI]0.65至0.89），而调整后的死亡风险无差异。与亚组1相比，亚组3和亚组4的死亡风险更高（分别为HR 1.94；95%CI 1.40至2.67和HR 3.87；95%CI 2.98至5.03）。交叉滞后面板模型分析显示乳酸水平与器官功能障碍之间存在双向因果关系（乳酸→SOFA，β = 0.310，P < 0.001 对比 SOFA→乳酸，β = 0.037，P < 0.001）。开发了一个包含五个变量的列线图以识别不良乳酸变化轨迹。结论：乳酸变化轨迹与脓毒症早期存活患者的死亡风险显著相关，这为脓毒症的风险分层提供了一个有价值的框架。
