Issarawattana Thanaphruet, Bhurayanontachai Rungsun
Division of Internal Medicine, Faculty of Medicine, Prince of Songkla University, Hat Yai, Songkhla, Thailand.
Critical Care Medicine Unit, Division of Internal Medicine, Faculty of Medicine, Prince of Songkla University, Hat Yai, Songkhla, Thailand.
Crit Care Res Pract. 2020 Aug 24;2020:5071509. doi: 10.1155/2020/5071509. eCollection 2020.
This retrospective study aimed to determine the correlation of blood glucose and glycemic variability with mortality and to identify the strongest glycemic variability parameter for predicting mortality in critically ill patients.
A total of 528 patients admitted to the medical intensive care unit were included in this study. Blood glucose levels during the first 24 hours of admission were recorded and calculated to determine the glycemic variability. Significant glycemic variability parameters, including the standard deviation, coefficient of variation, maximal blood glucose difference, and J-index, were subsequently compared between intensive care unit survivors and nonsurvivors. A binary logistic regression was performed to identify independent factors associated with mortality. To determine the strongest glycemic variability parameter to predict mortality, the area under the receiver operating characteristic of each glycemic variability parameter was determined, and a pairwise comparison was performed.
Among the 528 patients, 17.8% (96/528) were nonsurvivors. Both survivor and nonsurvivor groups were clinically comparable. However, nonsurvivors had significantly higher median APACHE-II scores (23 [21, 27] vs. 18 [14, 22]; < 0.01) and a higher mechanical ventilator support rate (97.4% vs. 74.9%; < 0.01). The mean blood glucose level and significant glycemic variability parameters were higher in nonsurvivors than in survivors. The maximal blood glucose difference yielded a similar power to the coefficient of variation ( = 0.21) but was significantly stronger than the standard deviation ( = 0.005) and J-index ( = 0.006).
Glycemic variability was independently associated with intensive care unit mortality. Higher glycemic variability was identified in the nonsurvivor group regardless of preexisting diabetes mellitus. The maximal blood glucose difference and coefficient of variation of the blood glucose were the two strongest parameters for predicting intensive care unit mortality in this study.
本回顾性研究旨在确定血糖及血糖变异性与死亡率之间的相关性，并找出预测危重症患者死亡率的最强血糖变异性参数。
本研究纳入了528例入住内科重症监护病房的患者。记录并计算入院后首个24小时内的血糖水平，以确定血糖变异性。随后，对重症监护病房幸存者和非幸存者之间的显著血糖变异性参数，包括标准差、变异系数、最大血糖差值和J指数进行比较。进行二元逻辑回归以确定与死亡率相关的独立因素。为确定预测死亡率的最强血糖变异性参数，计算每个血糖变异性参数的受试者工作特征曲线下面积，并进行两两比较。
在这528例患者中，17.8%（96/528）为非幸存者。幸存者组和非幸存者组在临床特征上具有可比性。然而，非幸存者的APACHE-II评分中位数显著更高（23[21,27]对18[14,22]；P<0.01），机械通气支持率更高（97.4%对74.9%；P<0.01）。非幸存者的平均血糖水平和显著血糖变异性参数高于幸存者。最大血糖差值与变异系数的预测能力相似（P=0.21），但显著强于标准差（P=0.005）和J指数（P=0.006）。
血糖变异性与重症监护病房死亡率独立相关。无论患者既往是否患有糖尿病，非幸存者组的血糖变异性均更高。在本研究中，最大血糖差值和血糖变异系数是预测重症监护病房死亡率的两个最强参数。
