Relationship Between Blood Glucose Levels and Variability and Infections Development in Critically Ill Patient

Overview

Our multicenter prospective observational study aims to show the relationship between blood glucose levels and glycemic variability and the development of infections during the ICU stay and with outcome. Within the secondary endpoints, we will evaluate if a blood glucose range between 70 and 140 mg/dl is associated with an increasing surviving rate in non-diabetic critically ill patients. MATERIALS AND METHODS Multicenter study (ICUs of some Italian University Hospitals). Written informed consent will be request before the inclusion of each patient in the study; if it will not be possible, an informing module will be given to the patient's family and the informed consent will be request to the patients as soon as possible. Inclusion criteria: 300 patients consecutively admitted in each ICU from January 2016 and not later than 31/12/2018. Exclusion criteria: age \< 18, end-stage disease. Data collection An Excel database will be edited with these data about each patient: age, sex, type I or II diabetes, glycated hemoglobin, at-home antidiabetic therapy; admission diagnosis, admission SAPS II score; daily insulin administration (dose and route of administration, time of start, dose at the moment of glycemic measurement and min-max daily range); steroid therapy (molecule, daily dose, date of start and stop); antibiotic therapy (molecule, daily dose, date of start and stop); daily caloric and protein intake and type of nutrition; other therapies; mechanical ventilation (date of start and stop); blood lactates (worst daily value); daily leucocytes and differential white cells count; daily SOFA score; presence of infections (suspected or confirmed; site and microorganism and eventual Multidrug Resistance pattern); presence of sepsis (following SCCM criteria); length of ICU and hospital stay; outcome (ICU and hospital mortality). Every blood glucose level measurement obtained will be registered with date and time. Glycemic variability will be evaluated in terms of: * Standard deviation (SD) * Mean Amplitude of Glycemic Excursions (MAGE); * Coefficient of Variation (CV); * Glycemic Lability Index (GLI). STATISTICAL ANALYSIS Data analysis will be performed with Kolmogorov-Smirnov test; parametric and non-parametric s tests, t-test (or Mann-Whitney test), ROC Curve, binary logistic regression. Subgroups analysis. Statistical significance: p \< 0,05. SAMPLE SIZE 3300 patients.

INTRODUCTION High blood glucose levels and insulin resistance are frequently registered in critically ill patients. In the past, hyperglycemia due to stress was considered as an adaptive body reaction to satisfy higher energetic requests. This theory has been re-discussed after Leuven's study in 2001, which demonstrated that a strict glycemic control, by means of intensive insulin administration, may reduce mortality in critically ill patients. However, more recent randomized controlled studies cannot confirm these results, showing how an intensive insulinic therapy can be associated to an higher risk of hypoglycemia and death. Recently, the attention has been shifted on the issue of glycemic variability in the Intensive Care Units (ICU), that seems to be a better mortality predictor than the simple blood glucose level. A relationship between lower glycemic variability, normal glucose levels (range 72-125 mg/dl) and outcome improving has been investigated. An high glycemic variability it's been associated to an increased in-hospital mortality, also in patients having total parenteral nutrition (TPN) independently from the presence of episodes of hypo or hyperglycemia. Furthermore, an intensive insulinic therapy itself may lead to excessive blood glucose levels fluctuations as it brings to an increased risk of hypoglycemia, usually treated in an "aggressive" way. A recent retrospective study has evaluated the impact of glycemic variability and hypoglycemia on outcome in non critically ill patients (undergone insulin administration during the hospitalization), showing a strong correlation between hypoglycemia and an increasing of mortality rate. A recent work of Arnold and coll. showed that the application of a treatment protocol for hypoglycemia (using administration of 50% dextrose) in critically ill patients was able to obtain a reduction of glycemic variability, remaining safe in avoiding dangerous hypoglycemia. However, the real causal relationship between glycemic variability and mortality has not yet been demonstrated. The increased oxydative stress due to wide glycemic fluctuations seems to play a leading role. The altered glucidic homeostasis in the critically ill patient may also cause a disfunction of immune cells with an increased risk of infections: Hirshberg et al. have found an association between glycemic variability an the risk of nosocomial infections development in a population of pediatric critically ill patients, studied retrospectively. A correlation between high glycemic variability, development of sepsis and risk of nosocomial infections in burn patients has also been demonstrated. Recently, a study by Krinsley and Preiser has highlighted how keeping blood glucose levels in a range between 70 and 140 mg/dl is strongly associated with an increased surviving rate in non diabetic patients, independently from ICU length of stay and from the severity of the clinical conditions. Donati et al. have demonstrated, in a retrospective study, the relationship between glycemic variability and infections; however, it's not yet clear if the glycemic variability is a cause of an effect of infections. Indeed, "relationship" is not equivalent to "causality" and it's yet to be elucidated if an higher glycemic variability just represents a sign of a worse clinical condition or, conversely, really leads to an increased risk of nosocomial infections. In this last case, dedicated protocols should be mandatory, not only to normalize blood glucose levels but also to minimize excessive glycemic fluctuations. STUDY PURPOSE Primary endpoint of this multicenter prospective observational study is to evaluate the relationship between blood glucose levels and glycemic variability and the development of nosocomial infection during ICU stay. Secondary endpoints are: To confirm the correlation between blood glucose levels and outcome (ICU mortality and hospital mortality). To evaluate if keeping the blood glucose level in a range between 70 and 140 mg/dl is associated with an increasing of surviving rate in non diabetic critically ill patients. PRIMARY ENDPOINT: Evaluation of the discriminatory power of Glycemic Lability Index (GLI) on infections development with AUC (ROC curve). SECONDARY ENDPOINTS: Evaluation of the discriminatory power of other indexes of glycemic variability on infections development. Evaluation of the discriminatory power of all the glycemic variability indexes on mortality. Evaluation of the eventual differences between diabetic and non diabetic patients about infections development. Evaluation of the impact of the caloric intake on glycemic variability, infections development and mortality. MATHERIALS AND METHODS The study will take place in ICUs of the participating centers. Written formal informed consent will be request before the inclusion of each patient in the study; if it won't be possible because of particular neurologic conditions of the patient, an informing module will be provided to the patient's family and the informed consent will be request to the patients as soon as possible. Inclusion criteria: 300 patients consecutively admitted in the participating ICUs from January 2016 and not later than 31/12/2018. Exclusion criteria: age \< 18, patients with end-stage disease with life expectancy shorter than 24 hours. Data collection An Excel database will be edited, collecting these data about each patient: age, sex, presence of type I or II diabetes, glycate haemoglobin (if present), at-home antidiabetic therapy (including oral antidiabetics and their dose and insulin); admission diagnosis, admission SAPS II score; for every day during the ICU stay will be registered: insulin administration (if "one-shot" administration, dose and route of administration, if in infusion, time of start, dose at the moment of glycemic measurement and min-max daily range); steroid therapy (molecule, daily dose, date of start and stop); antibiotic therapy (molecule, daily dose, date of start and stop); daily caloric and proteic intake specifying if via enteral or parenteral nutrition of both; other therapies included glucose and propofol; mechanical ventilation (yes/no, date of start and stop); blood lactates (worst daily value); daily leucocytes and differential white cells count; daily SOFA score; presence of infections (suspected or confirmed; if confirmed, site and microorganism and eventual Multidrug Resistance pattern); presence of sepsis (following SCCM criteria); length of ICU and hospital stay; outcome (ICU and hospital mortality). Every blood glucose level measurement obtained from laboratory analysis, bloodgas analysis and Glucostix, will be registered with date and time (hours and minutes). Glycemic variability will be evaluated in terms of these four indexes: Standard deviation (SD) Mean Amplitude of Glycemic excursions - MAGE, calculated as the mean of differences between consecutive values of blood glucose (absolute values) \> 1 SD of the global values; Coefficient of Variation (CV) calculated as SD/mean. Glycemic Lability Index (GLI) calculated following formule (where Gluc n is the number of the values registered on the time H n e N is the total numbers of values registered in a week; measurements with an interval between 1 and 12 hours will be registered; the mean of the GLI in the different weeks will be calculated). GLI ((mmol⁄l\^2 )⁄h\*〖week〗\^(-1) )=∑\_(n=1)\^N▒〖(〖Gluc〗\_n-〖Gluc〗\_(n+1))〗\^2⁄((h\_(n+1)-h\_n)) STATISTYCAL ANALYSIS All patients with an ICU length of stay shorter than 72 hours will be excluded from the statistical analysis for the primary endpoint. Based on an earlier study, about the 40% of the admitted patients have a length of stay shorter than 3 days and the incidence of infections was 30%. For this reason it's calculate to be necessary to include 2928 patients to have an alpha error \< 0,01 and a beta error \< 0,01, excluding every patient with an ICU stay shorter than 72 hours. Considering the variability in the length of ICU stay in the single ICUs and the possibility that on the date of 31/03/2017 some ICUs could not reach the possibility of admitting 300 patients, the theorical sample size is expanded to 3300 patients in total (20% of the theorical sample). The variables distribution will be evaluated with Kolmogorov-Smirnov test; parametric and non parametric statistical tests will be applicated where appropriate. T-Test (or Mann-Whitney test) will be used for the variables comparison between the groups "infections/no infections", "sepsis/no sepsis", "survivors/non survivors". ROC curves (Receiver Operating Characteristics) will be made to evaluate the discriminatory power of blood glucose levels, SD, MAGE, CV and GLI for the infections development, sepsis and outcome. Binary logistic regression will be applicated for analyzing the correlation between glycemic variability, infections development, sepsis and outcome, including APACHE II, SOFA, age, sex, presence of diabetes, blood lactates (at the admission and mean value), insulin administration, type of nutrition, mechanical ventilation and length of ICU stay as covariates. Subgroups analysis will be performed based on admission diagnosis, presence of diabetes, insulin administration on ICU, type of nutrition. Statistical significance will be considered with a p value \< 0,05. SAMPLE SIZE 3300 patients.