Glycemic targets in critically ill adults:A mini-review

Kay Choong See

Kay Choong See, Division of Respiratory and Critical Care Medicine, Department of Medicine,National University Hospital, Singapore 119228, Singapore

Abstract Illness-induced hyperglycemia impairs neutrophil function, increases pro-inflammatory cytokines, inhibits fibrinolysis, and promotes cellular damage. In turn,these mechanisms lead to pneumonia and surgical site infections, prolonged mechanical ventilation, prolonged hospitalization, and increased mortality. For optimal glucose control, blood glucose measurements need to be done accurately,frequently, and promptly. When choosing glycemic targets, one should keep the glycemic variability ＜ 4 mmol/L and avoid targeting a lower limit of blood glucose ＜ 4.4 mmol/L. The upper limit of blood glucose should be set according to casemix and the quality of glucose control. A lower glycemic target range (i.e.,blood glucose 4.5-7.8 mmol/L) would be favored for patients without diabetes mellitus, with traumatic brain injury, or who are at risk of surgical site infection.To avoid harm from hypoglycemia, strict adherence to glycemic control protocols and timely glucose measurements are required. In contrast, a higher glycemic target range (i.e., blood glucose 7.8-10 mmol/L) would be favored as a default choice for medical-surgical patients and patients with diabetes mellitus. These targets may be modified if technical advances for blood glucose measurement and control can be achieved.

Key Words: Brain injuries; Traumatic; Critical care; Diabetes mellitus; Glycemic control;Insulin infusion systems; Sepsis

INTRODUCTION

Illness-induced hyperglycemia can be a double-edged sword. On the one hand, it may be an adaptive response to provide extra metabolic substrate to organs like the brain and to blood cells[1]. On the other hand, hyperglycemia impairs neutrophil function and innate immunity, increases pro-inflammatory cytokines and oxidative stress[2,3],inhibits fibrinolysis[4], and promotes cellular damage[1]. In addition, hyperglycemia in brain-injured patients can lead to microcirculatory damage, blood-brain barrier disruption, and cellular swelling[5]. These pathological derangements potentially lead to complications such as pneumonia and surgical site infections, prolonged mechanical ventilation, increased intensive care unit (ICU) and hospital lengths of stay, and increased mortality.

Unlike hyperglycemia, hypoglycemia is always harmful. For example, hypoglycemia was independently associated with respiratory complications and prolonged ICU and hospital lengths of stay after cardiac surgery[6]. These adverse events may be mediated by hypoglycemia-related neuronal damage and cardiac arrhythmia[7]. Apart from the clear need to avoid blood glucose extremes, there is also a need to avoid excessive blood glucose fluctuations[8], which can be measured in various ways(Table 1). The simplest measure of blood glucose fluctuation is glycemic variability,which is the difference between the maximum and minimum blood glucose measured over a defined time interval. At the cellular level, glycemic variability has been associated with oxidative stress, endothelial dysfunction, and apoptosis[7]. Clinically,glycemic variability has been linked to increased ICU and hospital mortality[9,10].

Table 1 Types of glycemic targets in intensive care unit

Blood glucose measurements need to be done accurately, frequently, and promptly[11]. Ideally, blood glucose measurements should be done continuously, though continuous glucose monitoring (CGM) for critically ill patients may not be accurate enough, with wide limits of agreement despite small mean bias[12]. CGM appears unreliable when using minimally-invasive subcutaneous devices that assay interstitial glucose measurements[13-15], and does not seem to improve glucose control[16].Although invasive (intravascular) CGM devices may have an acceptable accuracy,some drawbacks include vascular and infectious complications (thrombosis, catheter occlusion, biofilm formation, or intravascular catheter-related infection)[17,18].

Accuracy and variation of glucose measurement methods influence the feasibility and adherence to glycemic targets[19]. In the real world, a variety of blood samples(arterial, venous, and capillary) are assayed intermittently, using both point-of-care and laboratory methods[20,21], and managed using various protocols. Nonetheless,despite such variation, clinical utility of current glucose measurement systems seems adequate, with little evidence of over or under-treatment[22]. Additionally, to achieve optimal clinical outcomes, blood glucose should be lowered if it were to rise too high,blood glucose should not be allowed to dip too low, and blood glucose variability should be constrained.

To determine clinically optimal glycemic targets for critically ill adult patients, the key questions would therefore be as follows:(1) What should thehyperglycemicthreshold be; (2) What should thehypoglycemicthreshold be; and (3) How far apart should these thresholds be? This review aims to integrate empirical evidence to answer these questions, and to suggest practical recommendations for choosing glycemic targets.

EMPIRICAL EVIDENCE FOR GLYCEMIC THRESHOLDS IN ICU

Several trials are inconclusive with respect to intensive (lower)vsconventional(higher) glycemic targets, which may be due to insufficient separation of achieved glucose levels between the intervention and control groups[23-25]. Another reason could be that the impact of glucose control was modified by the main diagnosis (i.e.,casemix). In terms of the hyperglycemic threshold, the blood glucose level beyond which clinical complications occur seems to differ by casemix (Table 2). Patients without diabetes mellitus (DM)[26], patients with traumatic brain injury (TBI), and post-surgical patients at risk of wound infection experience adverse effects of hyperglycemia at a relatively low range, with the threshold set at 6.7-8.3 mmol/L[27-30].

Table 2 Glycemic targets in intensive care unit by casemix and thresholds

DM:Diabetes mellitus; HbA1c:Glycosylated hemoglobin; ICU:Intensive care unit; Obs:Observational study; RCT:Randomized controlled trial; SHR:Stress hyperglycemia ratio; TBI:Traumatic brain injury.

The NICE-SUGAR trial showed that undifferentiated medical-surgical ICU patients had decreased 90-d mortality and incident hypoglycemia when the upper limit of blood glucose was set at 10 mmol/L rather than 6.1 mmol/L[31]. Patients who suffered non-TBI-specific injury[32] or who had post-cardiac arrest[33] also experienced better neurological recovery if blood glucose could be kept below 10 mmol/L.

Patients with prior DM were able to tolerate a higher mean blood glucose level (i.e.,blood glucose level ＞ 10 mmol/L) without excess complications during critical illness,although these patients benefited from lowering blood glucose below 7.8 mmol/L after coronary artery bypass surgery[34]. Chronic hyperglycemia may have compensatory mechanisms in place that provide protection from acute hyperglycemiarelated cellular damage[2]. The upper limit of safety in patients with DM appears to be a blood glucose level of 14 mmol/L[35].

In contrast to the risk of hyperglycemia differing by casemix, the risks of hypoglycemia appear to affect a broad range of patients similarly. Severe hypoglycemia(＜ 2.2 mmol/L), moderate hypoglycemia (＜ 3.3 mmol/L), and even mild hypoglycemia (＜4 mmol/L) have been associated with ICU and hospital mortality[36-39]. Targeting lower blood glucose levels resulted in higher rates of severe hypoglycemia[40,41], and no clinical trial has targeted a lower limit of blood glucose ＜4.4 mmol/L. The NICE-SUGAR trial demonstrated that the risk of hypoglycemia can be mitigated by avoiding targeting blood glucose below 6.1 mmol/L[31]. Nonetheless,if intensive glucose monitoring and management resources are available, and if glycemic control protocols and timely glucose measurements can be strictly adhered to, the Leuven studies demonstrated advantages of targeting blood glucose below 6.1 mmol/L, with surgical patients deriving clearer survival benefit and morbidity reduction compared to medical patients[23,42].

EMPIRICAL EVIDENCE FOR MINIMIZING GLYCEMIC VARIABILITY IN ICU

In a multicenter observational study, Egiet al[43] first showed that ICU non-survivors had a wider spread of glucose values compared to ICU survivors. Specifically, the standard deviation of blood glucose values was 2.3 mmol/L in non-survivors compared to 1.3 mmol/L in survivors. The association between spread of blood glucose with hospital mortality persisted after controlling for confounders (hospitalsite, surgical patients, neurologic diseases, mechanical ventilation, acute physiological and chronic health evaluation II score, age, mean blood glucose level, maximum blood glucose level, and admission blood glucose level).

Subsequently, other observational studies have demonstrated that the difference between maximum and minimum blood glucose levels (i.e., glucose variability) should not exceed 4-6 mmol/L, regardless of casemix[10,44,45] (Table 3). In other words,glycemic target ranges should ideally be ＜ 4 mmol/L in width. Such a narrow range seems to be achievable, given that both single-center and multi-center randomized trials using a variety of protocols have successfully constrained glucose levels within standard deviations of ＜ 2 mmol/L[23,31,42,46].

Table 3 Glycemic targets in intensive care unit by casemix and variability

CHOOSING LOWER VS HIGHER GLYCEMIC TARGET RANGES

To minimize patient harm, empirical evidence suggests that when choosing glycemic targets, one should keep the glycemic variability ＜ 4 mmol/L and avoid targeting a lower limit of blood glucose ＜ 4.4 mmol/L. The upper limit of blood glucose should then be set according to casemix and the quality of glucose control.

A lower glycemic target range (i.e., blood glucose 4.5-7.8 mmol/L) would be favored for patients without DM, with TBI, or who are postoperative and at risk of surgical site infection. Requirements for targeting a lower range and avoiding harm from hypoglycemia would be availability of intensive glucose monitoring and management,strict adherence to glycemic control protocols, and strict adherence to timely glucose measurements (Table 4).

Table 4 Choosing lower vs higher glycemic target ranges

In contrast, a higher glycemic target range (i.e., blood glucose 7.8-10 mmol/L)would be favored as a default choice for medical-surgical patients and patients with DM. Additionally, a higher range would be favored if conditions to avoid hypoglycemia cannot be strictly met,i.e., lack of intensive glucose monitoring and management, less than strict adherence to glycemic control protocols, and less than strict adherence to timely glucose measurements.

This review’s recommendations are in line with current guidelines (Table 5). For hospitalized patients in general, the American Diabetes Association recommends a glycemic target range of 7.8-10 mmol/L[47]. The same glycemic range is recommended for post-resuscitation care of cardiac arrest patients by the European Resuscitation Council[48]. For sepsis patients, the Surviving Sepsis Campaign recommends an upper blood glucose limit of 10 mmol/L[49]. Both the American Diabetes Association and Surviving Sepsis Campaign guidelines mention that lower targets may be appropriate for selected patients if they can be achieved without significant hypoglycemia[47,49].

Table 5 Selected guideline recommendations

Other guidelines have made less definite recommendations. For surgical patients,the World Health Organization recommends glucose control, though no target range was defined[50]. For patients with TBI, the Brain Trauma Foundation does not mention glycemic control[51]. The findings and recommendations from this review can therefore help fill any gaps in these latter guidelines.

FUTURE DIRECTIONS

To increase the safety of lower glycemic targets, technical advances for blood glucose measurement and control would help. Autocorrecting point-of-care glucose measurement devices can adjust for interfering substances (e.g., ascorbic acid and nonglucose sugars) and abnormal hematocrit in critically ill patients[52], enabling these devices to become as accurate as central laboratory plasma glucose measurements.Monte Carlo simulation suggests that glycemic control in critically ill patients is optimal with a blood glucose measurement interval no longer than 1 h, with incremental benefit using shorter measurement intervals of 15 min[53]. This means that devices that can continuously assay blood glucose would be needed. More accurate and frequent blood glucose measurements can feed into automated and closed-loop glycemic control systems[54-62]. For instance, even when targeting a lower range of 4.4-8.3 mmol/L, one such system limited severe hypoglycemic episodes to only 0.01% of all blood glucose measurements and 0.8% of patients[59].

Optimization of glucose control protocols with respect to the following aspects may also be investigated:(1) Addition of bolus insulin "mid-protocol" during an insulin infusion to reduce peak insulin rates for insulin-resistant patients[63]; (2) transition of insulin administration route from intravenous to subcutaneous[64], and (3) use of DMspecific enteral formula for both DM and non-DM patients[65-67].

Given the influence of casemix on the optimal glycemic target range, further work may be done to personalize recommendations for various conditions[68]. For patients with DM, it remains unclear if the upper limit of blood glucose can be safely pushed beyond 10 mmol/L[69], given the risk of ketoacidosis or ketonemia[70]. To address this uncertainty, the LUCID trial will investigate if liberal blood glucose (target 10.0-14.0 mmol/L) will result in less incident hypoglycemia compared to usual care (target 6.0-10.0 mmol/L), while maintaining good clinical outcomes[71].

CONCLUSION

When choosing glycemic targets, one should keep the glycemic variability ＜ 4 mmol/L and avoid targeting a lower limit of blood glucose ＜ 4.4 mmol/L. The upper limit of blood glucose should be set according to casemix and the quality of glucose control. A lower glycemic target range (i.e., blood glucose 4.5-7.8 mmol/L) would be favored for patients without diabetes mellitus, with traumatic brain injury, or who are at risk of surgical site infection. To avoid harm from hypoglycemia, strict adherence to glycemic control protocols and timely glucose measurements are required. In contrast, a higher glycemic target range (i.e., blood glucose 7.8-10 mmol/L) would be favored as a default choice for medical-surgical patients and patients with diabetes mellitus. These targets may be modified if technical advances for blood glucose measurement and control can be achieved.