In the critically ill, there is a consistent association between each of the three domains of dysglycemia—hypoglycemia, hyperglycemia, and high glycemic variability—and poor outcome, although causal links have not been established (1–3). However, failure to reproduce the beneficial effects reported in the pioneering studies on tight glycemic control in Leuven (4 , 5) has fueled intense debate around the potential benefits and risks associated with intensive insulin therapy (6). Recent data support personalized glycemic control (7).
Current guidelines and expert opinion (8–11) recommend relatively loose glycemic control, with an upper blood glucose (BG) limit set at 180 mg/dL to avoid severe hyperglycemia and to limit glucose variability and hypoglycemia. Indeed, iatrogenic and relative hypoglycemia can occur during any form of insulin therapy and have consistently been associated with an increase in mortality (12 , 13). However, this seemingly easy target is hard to achieve with the currently available equipment, and intermittent glucose monitoring (IGM) is labor intensive (14 , 15). Even in highly motivated centers participating in interventional trials comparing two target ranges of BG, the performance of glycemic control assessed by the proportion of time in range (TIR) calculated from intermittent readings ranged from 31% to 53% (16).
Recent developments in this field include (near)-continuous glucose monitoring (CGM) systems. Glucose can be measured by CGM in whole blood, plasma, interstitial fluid, or microdialysis fluid by glucose oxidase–based techniques, midinfrared spectroscopy, fluorescence or microdialysis, continuously or near continuously (every 5–15 min). However, as reported by a board of experts in a recent opinion article (17) and in recent reviews (18 , 19), clinical experience has demonstrated that the degree of accuracy required for glucose control in critically ill patients can only be achieved by intravascular (IV or intra-arterial) devices, whereas the performance of some point-of-care meters have been found acceptable (20). Conceptually, CGM can decrease the incidence of severe hyperglycemia and hypoglycemia by enabling insulin infusions to be adjusted more rapidly and more appropriately because trends in glucose concentrations can be more readily identified. Marketed CGM devices have been validated in clinical studies by matching of standard requirements for glucose meters. These requirements were designed to assess the accuracy of point-of-care meters in outpatients with diabetes (ISO 15197 2013 and the Food and Drug Administration [FDA] draft guidance of 2014) (21–29) while the standards suggested for CGM were adapted to intensive care medicine (30). However, the effects of a strategy of glycemic control guided by a previously validated CGM device have not been prospectively tested. The present study compared a strategy of glycemic control guided by CGM with the current standard of care based on intermittent BG readings. The quality and safety of glycemic control were assessed using the TIR and the incidence and severity of hypoglycemia.
PATIENTS AND METHODS
Study Design and Conduct
This single-center, cluster-randomized, crossover, prospective, investigator-initiated study was performed in the 35-bed Department of Intensive Care of Erasme Hospital, an 858-bed university hospital in Brussels (Belgium), from July 2014 to June 2015 (Fig. 1). The Department is composed of four separate medical/surgical ICUs positioned in the same area of the hospital. The study was approved by the institutional review board (P2014/162). Signed informed consent was obtained from all patients or their next of kin.
Eligible patients met the following criteria: 1) 18 years old or older 2) critically ill, as reflected by an Acute Physiology and Chronic Health Evaluation (APACHE) II score greater than or equal to 10 (31); 3) expected to stay in the department for at least 3 days; and 4) had persistent hyperglycemia (BG >150 mg/dL) for at least 6 hours after admission and/or were being treated with IV insulin. We excluded pregnant patients and those with any contraindication to heparin, a known history of heparin-induced thrombocytopenia, a documented end-of-life decision to limit therapy, or skin conditions or existing/planned medical instrumentation or dressings that precluded the placement of a peripheral IV catheter (e.g., extensive psoriasis, scarring, tattoos, eczema, dermatitis herpetiformis, burns, surgical dressings). Patients being treated with insulin prior to ICU admission could be included.
After signed informed consent, a peripheral venous catheter (Insyte Autoguard Winged, 20 G, 1.1 × 25 mm; Becton-Dickinson, Franklin Lakes, NJ) was inserted in a forearm vein, after skin disinfection with a chlorhexidine gluconate/isopropyl alcohol solution. Each patient included in the study had a blood sample taken at the time of ICU admission to measure BG using the standard local point-of-care testing device (GEM Premier 4000; Instrumentation Laboratory, Bedford, MA). The CGM sensor (GlucoClear; Edwards Lifesciences, Irvine, CA) was then placed in the dedicated catheter using the manufacturer’s recommended technique and left in place for up to 3 days. Once calibrated and flushed using a heparinized solution, the GlucoClear (Edwards Lifesciences) monitor measured BG every 5 minutes in all patients using a glucose oxidase–based method. No other catheter could be placed in the same forearm.
The four ICUs were randomized in a crossover design (Fig. 1). During the first period, two units were equipped with unblinded CGM monitors, enabling the CGM values to be visualized, and two with blinded monitors, so that the CGM values could not be seen. During the second period, the equipment was switched between units, in order to allocate the unblinded monitors to units previously equipped with blinded monitors and vice versa. Each period lasted up to 28 days, or for a total of 50 patients across all the units, whichever occurred first. All potential users (nursing staff, research associates) were trained in installing, calibrating, and simple troubleshooting of the CGM system.
Each patient was connected to a blinded monitor (IGM group) or an unblinded monitor (CGM group), depending on which unit they were admitted to. Data collection started at the time of admission to the ICU (first BG value available) and finished up to 72 hours later, at the time of discharge from the ICU, or in case of technical failure, whichever came first. At the end of each crossover period, the complete sets of electronic BG data were downloaded from the GlucoClear (Edwards Lifesciences) monitor for analysis. The unblinded monitors had major alarms set at 40 and 220 mg/dL and minor alarms set at 70 and 180 mg/dL. The blinded monitors only had alarms triggered by technical issues, such as open door, inability to measure BG, near-expired cartridge, flushing problem.
Glucose Management Protocol
In all patients, bedside nurses used the standard department dynamic insulin algorithm adapted from Meynaar et al (32), which targets a BG range of 90–150 mg/dL. This algorithm has been used by our department since 2009; about 40% of patients require IV infusion of insulin during the ICU stay. In the IGM group, the rate of insulin infusion was adjusted based on repeated BG values measured using a blood gas analyzer on average four to six times per day. In the CGM group, the BG values displayed on the monitor were introduced manually into the bedside computer every 30–60 minutes to adapt insulin infusion rates. A continuous infusion of regular human insulin (Actrapid HM; Novo Nordisk, Baegsvard, Denmark) with a standard concentration of 50 IU of insulin in 50 mL of 0.9% NaCl was used in both groups.
In the CGM units, additional reference blood samples were taken and checked on the BG analyzer if 1) there was a significant decrease between two consecutive GlucoClear (Edwards Lifesciences) measurements, that is, more than 40 mg/dL without an identified reason; 2) there was a minor or major alarm; 3) there was any doubt concerning the value indicated on the GlucoClear (Edwards Lifesciences) monitor based on a flat trend or any other issue. If the difference between the BG measured by the reference method and the GlucoClear (Edwards Lifesciences) value was greater than 20 mg/dL, the reference BG value was entered into the insulin dynamic protocol, according to the standard of care. When the difference was less than 20 mg/dL, the GlucoClear (Edwards Lifesciences) value was used.
Data were collected from the subject’s medical records and the case report forms. We recorded demographic data, the primary reason for ICU admission, ICU length of stay, and the ICU outcome. The APACHE II score was determined on the day of admission to the ICU. Glucose-related data included BG values measured with the blood gas analyzer and with GlucoClear (Edwards Lifesciences) at least once per hour, technical issues (resiting, flushing, troubleshooting), serial number of each monitor and sensor.
The quality of glycemic control was assessed by the proportion of time the BG was within the target range (TIR90–150). Safety was assessed by the proportion of patients who experienced at least one episode of BG less than 70 mg/dL, the lowest BG values, and the percentage of time the BG was less than 70 mg/dL (TB70). These metrics were calculated from the values recorded by the CGM for each day and for the whole study period and compared between the groups. Glycemic variability was assessed from the values recorded by the CGM using the coefficient of variation (CV, SD/average) and the glycemic lability index (GLI) (33 , 34).
The “down-time” was calculated as the period of time during which there was no BG determination although a display of BG was expected, in relation to monitor-related, cartridge-related, or catheter-related issues. Monitor-related technical issues were reported as the “no-read rate.”
Data are presented as means with SD or medians and interquartile ranges for continuous variables and as numbers and percentages for categorical variables. The Shapiro-Wilk test was used, and histograms and normal quantile plots were examined to verify whether there were significant deviations from the normality assumption of continuous variables. Difference testing between groups was performed using Student t test, Mann-Whitney U test, chi-square test, or Fisher exact test, as appropriate. Statistical analyses were performed using IBM SPSS 24 statistical program for Windows (IBM Corporation, Somers, NY). All statistical tests were two sided, and p values of less than 0.05 were considered statistically significant.
From a total of 3,261 admissions during the study period, a convenience sample of 100 patients was included. In 23 of these patients, there was no peripheral vein suitable for catheter placement (n = 11), IV insulin was no longer required at the time of connection (n = 9), death occurred early (n = 2), or there was a heparin allergy (n = 1). Of the 77 remaining patients who had at least one BG value measured by CGM, 39 were allocated to the CGM group and 38 to the IGM group. The demographic and clinical characteristics were representative of a population of severely ill patients and were similar in the two groups (Table 1).
We recorded a total of 22,721 BG values during a period of 3,012 ± 1,348 minutes in the CGM group and 20,386 during 3,019 ± 1,365 minutes in the IGM group (differences = not significant [NS]), with an overall range of 53–378 mg/dL. No medication used during the study period was known to interfere with the sensor used, according to the manufacturer’s library. A representative graph showing the time course of BG in a patient during monitoring is shown in Figure 2. The down-time averaged 6.4% (range, 1.74–10.8%) and 8.3% (range, 1.4–28.5%) in the CGM and IGM groups, respectively (NS). There were no statistically significant differences in the no-read rate (3.3% in the CGM group and 3.0 % in the IGM group, NS). The average BG was similar in the two groups (121 ± 26 mg/dL in the CGM group and 124 ± 25 in the IGM group, NS), with values ranging from 61 to 368 mg/dL in the CGM group and from 53 to 378 mg/dL in the IGM group. The TIR90–150 and the glycemic variability, assessed by CV or GLI, did not differ between groups (Table 2).
The number of patients who experienced at least one episode of BG less than 70 mg/dL was lower in the CGM than in the IGM group (8/39 [20.5%]) vs 15/38 [39.5%]). The lowest BG values recorded in the CGM and IGM groups were 68 and 53 mg/dL, respectively. No episodes of severe hypoglycemia (BG < 40 mg/dL) occurred in either group. The TB70 was lower in the CGM group than in the IGM group (0.4% ± 0.9% vs 1.6% ± 3.4%; p < 0.05) (Fig. 3).
The most salient findings of this study are the reliability and feasibility of glycemic control using a CGM-guided strategy. Importantly, our results show that the safety of glycemic control was increased when CGM was used to adapt the insulin infusion rate, with decreases in the incidence and severity of hypoglycemia compared with IGM insulin adjustment. This enhanced safety was likely related to more frequent changes in the insulin infusion rates, which were not measured. Admittedly, more frequent BG measurements could potentially lead to overtitration of insulin, but near-continuous monitoring reduces this risk by providing an early warning of impending hypoglycemia enabling insulin doses to be adjusted rapidly and appropriately. In contrast, less frequent BG values, as in the IGM group, increase the risk of hypoglycemia because intermediate BG values that may have enabled the insulin dose to be adjusted are not measured. Other aspect of safety, such as catheter-related issues, was not assessed. Glycemic variability could also be considered as a surrogate marker of safety and was not affected by the use of CGM. The ability to safely achieve a target BG range will enable assessment of the effects of therapeutic strategies targeting narrower or personalized BG ranges. The lack of effect of CGM on the TIR is not surprising, in view of the high TIR (> 70%) in the IGM “standard-of-care” group. This high TIR may be related to the wide target range selected (8–10), the use of an adapted bedside dynamic algorithm, and the motivation and skill of the nursing team (13). We used the TIR calculated from CGM values because it differs from the TIR calculated from intermittent values; in a previous study, we found a 4% difference between the TIR measured using another CGM device and intermittent values (22). The calculation of percentage of time without a dysglycemic event (proportion of time spent without hypoglycemia, hyperglycemia, and/or with low variability) would be desirable for a more accurate evaluation of the effects of the CGM-guided strategy, but it cannot yet be performed because there is no validated reference value for variability using CGM readings.
Importantly, use of the system was facilitated by training sessions delivered to the nursing staff. The unique organization of our department allows comparisons of “clusters” of polyvalent units staffed with a dedicated permanent team of nurses who are used to managing medical and surgical/trauma patients. A crossover design was used to minimize the potential risk of bias related to individual factors or to different levels of knowledge or motivation between the units. The nursing teams expressed no complaints related to the CGM device.
The GlucoClear (Edwards Lifesciences) system has been validated in different settings and meets legal requirements (mean absolute relative difference, ISO 15197, Clarke Error Grid) (24). The accuracy and ease of use of the GlucoClear (Edwards Lifesciences) system may be reduced compared with other CGM devices because of potential physicochemical interference related to the glucose oxidase–based method and by the need for peripheral venous access (impossible in 11% of our patients) and for frequent calibrations. Nevertheless, the “down-time” and the duration of monitoring were similar to those reported for other devices (21–23 , 29). The level of agreement between BG values measured by blood gas analyzers (used as reference method in this study) and in the central laboratories is sufficient for clinical use in ICU, even though it may not meet the stricter requirements of the FDA 2014 draft guidance or 2016 final guidance (35).
Our study has some limitations, including the lack of data on insulin treatment (dose—number of changes), the lack of clinical outcome variables other than ICU mortality, and the absence of a formal evaluation of the feedback from the nursing staff. However, a decrease in nursing workload is likely in view of the reliability of the CGM values, implying a reduction in the need for intermittent BG sampling.
The development of CGM has opened the possibility of using closed-loop systems connected to model-predictive controllers for glycemic control in critically ill patients, an approach already successfully tried in small cohorts of critically ill (11 , 36) or postoperative (37) patients.
In conclusion, in a select group of critically ill adult patients, CGM measurements were associated with safer glycemic control as a result of the decreased incidence and severity of hypoglycemia.
1. Badawi O, Waite MD, Fuhrman SA, et alAssociation between intensive care unit-acquired dysglycemia
and in-hospital mortality. Crit Care Med 2012; 40:3180–3188
2. Falciglia M, Freyberg RW, Almenoff PL, et alHyperglycemia-related mortality in critically ill patients varies with admission diagnosis. Crit Care Med 2009; 37:3001–3009
3. Mesotten D, Preiser JC, Kosiborod MGlucose management in critically ill adults and children. Lancet Diabetes Endocrinol 2015; 3:723–733
4. van den Berghe G, Wouters P, Weekers F, et alIntensive insulin therapy in critically ill patients. N Engl J Med 2001; 345:1359–1367
5. Marik PE, Preiser JCToward understanding tight glycemic control in the ICU: A systematic review and metaanalysis. Chest 2010; 137:544–551
6. Marik PE, Bellomo RStress hyperglycemia
: An essential survival response! Crit Care 2013; 17:305
7. Krinsley JS, Preiser JC, Hirsch IBSafety and efficacy of personalized glycemic control in critically ill patients: A 2-year before and after interventional trial. Endocr Pract 2017; 23:318–330
8. Ichai C, Preiser JCSociété Française d’Anesthésie-Réanimation; Société de Réanimation de langue Française; Experts group: International recommendations for glucose control
in adult non diabetic critically ill patients. Crit Care 2010; 14:R166
9. Moghissi ES, Korytkowski MT, DiNardo M, et alAmerican Association of Clinical Endocrinologists; American Diabetes Association: American Association of Clinical Endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control. Diabetes Care 2009; 32:1119–1131
10. Jacobi J, Bircher N, Krinsley J, et alGuidelines for the use of an insulin infusion for the management of hyperglycemia
in critically ill patients. Crit Care Med 2012; 40:3251–3276
11. Preiser JC, Chase JG, Hovorka R, et alGlucose control in the ICU: A continuing story. J Diabetes Sci Technol 2016; 10:1372–1381
12. Finfer S, Liu B, Chittock DR, et alNICE-SUGAR Study Investigators: Hypoglycemia
and risk of death in critically ill patients. N Engl J Med 2012; 367:1108–1118
13. Di Muzio F, Presello B, Glassford NJ, et alLiberal versus conventional glucose targets in critically ill diabetic patients: An exploratory safety cohort assessment. Crit Care Med 2016; 44:1683–1691
14. Aragon DEvaluation of nursing work effort and perceptions about blood glucose testing in tight glycemic control. Am J Crit Care 2006; 15:370–377
15. Schultz MJ, Harmsen RE, Spronk PEClinical review: Strict or loose glycemic control in critically ill patients–implementing best available evidence from randomized controlled trials. Crit Care 2010; 14:223
16. Krinsley JSGlycemic control in the critically ill: What have we learned since NICE-SUGAR? Hosp Pract (1995) 2015; 43:191–197
17. Wernerman J, Desaive T, Finfer S, et alContinuous glucose control
in the ICU: Report of a 2013 round table meeting. Crit Care 2014; 18:226
18. Shapiro ARFDA approval of nonadjunctive use of continuous glucose monitors for insulin dosing: A potentially risky decision. JAMA 2017; 318:1541–1542
19. Smith JL, Rice MJWhy have so many intravascular glucose monitoring devices failed? J Diabetes Sci Technol 2015; 9:782–791
20. Zhang R, Isakow W, Kollef MH, et alPerformance of a modern glucose meter in ICU and general hospital inpatients: 3 years of real-world paired meter and central laboratory results. Crit Care Med 2017; 45:1509–1514
21. van Hooijdonk RT, Winters T, Fischer JC, et alAccuracy and limitations of continuous glucose monitoring using spectroscopy in critically ill patients. Ann Intensive Care 2014; 4:8
22. Righy Shinotsuka C, Brasseur A, Fagnoul D, et alManual versus Automated moNitoring Accuracy of GlucosE II (MANAGE II). Crit Care 2016; 20:380
23. Bochicchio GV, Nasraway S, Moore L, et alResults of a multicenter prospective pivotal trial of the first inline continuous glucose monitor in critically ill patients. J Trauma Acute Care Surg 2017; 82:1049–1054
24. Foubert LA, Lecomte PV, Nobels FR, et alAccuracy of a feasibility version of an intravenous continuous glucose monitor in volunteers with diabetes and hospitalized patients. Diabetes Technol Ther 2014; 16:858–866
25. Strasma PJ, Finfer S, Flower O, et alUse of an intravascular fluorescent continuous glucose sensor in ICU patients. J Diabetes Sci Technol 2015; 9:762–770
26. Sechterberger MK, van der Voort PH, Strasma PJ, et alAccuracy of intra-arterial and subcutaneous continuous glucose monitoring in postoperative cardiac surgery patients in the ICU. J Diabetes Sci Technol 2015; 9:663–667
27. Crane BC, Barwell NP, Gopal P, et alThe development of a continuous intravascular glucose monitoring sensor. J Diabetes Sci Technol 2015; 9:751–761
28. Schierenbeck F, Öwall A, Franco-Cereceda A, et alEvaluation of a continuous blood glucose monitoring system using a central venous catheter with an integrated microdialysis function. Diabetes Technol Ther 2013; 15:26–31
29. Bochicchio GV, Hipszer BR, Magee MF, et alMulticenter observational study of the first-generation intravenous blood glucose monitoring system in hospitalized patients. J Diabetes Sci Technol 2015; 9:739–750
30. Finfer S, Wernerman J, Preiser JC, et alClinical review: Consensus recommendations on measurement of blood glucose and reporting glycemic control in critically ill adults. Crit Care 2013; 17:229
31. Knaus WA, Draper EA, Wagner DP, et alAPACHE II: A severity of disease classification system. Crit Care Med 1985; 13:818–829
32. Meynaar IA, Dawson L, Tangkau PL, et alIntroduction and evaluation of a computerised insulin protocol. Intensive Care Med 2007; 33:591–596
33. Ali NA, O’Brien JM Jr, Dungan K, et alGlucose variability and mortality in patients with sepsis. Crit Care Med 2008; 36:2316–2321
34. Donati A, Damiani E, Domizi R, et alGlycaemic variability, infections and mortality in a medical-surgical intensive care unit. Crit Care Resusc 2014; 16:13–23
35. Liang Y, Wanderer J, Nichols JH, et alBlood gas analyzer accuracy of glucose measurements. Mayo Clin Proc 2017; 92:1030–1041
36. Leelarathna L, English SW, Thabit H, et alFeasibility of fully automated closed-loop glucose control
using continuous subcutaneous glucose measurements in critical illness: A randomized controlled trial. Crit Care 2013; 17:R159
37. Okabayashi T, Shima Y, Sumiyoshi T, et alIntensive versus intermediate glucose control
in surgical intensive care unit patients. Diabetes Care 2014; 37:1516–1524
Keywords:Copyright © by 2018 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
cluster-randomized study; dysglycemia; glucose control; glycemic variability; hyperglycemia; hypoglycemia