KDIGO Clinical Practice Guideline for Acute Kidney Injury
Glycemic control in critical illness: Recommendations and Rationale

Chapter 3.3: Glycemic control and nutritional support

Glycemic control in critical illness: renal effects and outcomes

• 3.3.1:In critically ill patients, we suggest insulin therapy targeting plasma glucose 110–149 mg/dl (6.1–8.3 mmol/l). (2C)

Rationale

As outlined in a recent review,¹³⁴ stress hyperglycemia is a distinctive clinical feature of critical illness. Stress mediators, and central and peripheral insulin resistance appears pivotal to the occurrence of stress hyperglycemia. Inflammatory mediators and counter-regulatory hormones have been shown to impede crucial elements of the insulin-signaling pathway. Still, exogenous insulin administration normalizes blood glucose levels in this setting. Insulin treatment may counteract hepatic insulin resistance during acute critical illness. Extensive observational data have shown a consistent, almost linear, relationship between blood glucose levels in patients hospitalized with MI and adverse clinical outcomes, even in patients without established diabetes.^135,136

It has never been entirely clear, however, whether glycemia serves as a mediator of these outcomes or merely as a marker of the sickest patients, who present with the well-known counter-regulatory stress response to illness.¹³⁷ Interestingly, Kosiborod et al.¹³⁵ recently showed, in a population with MI, that while hypoglycemia was associated with increased mortality, this risk was confined to patients who developed spontaneous hypoglycemia. In contrast, iatrogenic hypoglycemia after insulin therapy was not associated with higher mortality risk.

Tight glycemic control is frequently used in patients at risk of AKI, and in the management of those who develop AKI. It has been proposed that tight glycemic control can reduce the incidence and severity of AKI. Since the landmark trial of Van den Berghe et al.,¹³⁸ additional studies provided initial confirmation of the benefits (reduced morbidity and mortality), and some additional mechanistic insights of tight glycemic control in critically ill patients.¹³⁹ Further secondary analysis of the original trial, which was conducted in 1548 mechanically ventilated surgical ICU patients, found that intensive insulin therapy (IIT) target plasma glucose 80–110 mg/dl (4.44–6.11 mmol/l) was associated with substantial cost savings compared to conventional insulin therapy (CIT) target plasma glucose 180–200 mg/dl (9.99– 11.1mmol/l).¹⁴⁰ However, when Van den Berghe et al. repeated their original study in a different population of critically ill patients (medical rather than surgical ICU patients), the primary end-point of in-hospital mortality did not differ between groups (40% CIT group vs. 37.3% IIT group; P = 0.33).¹⁴¹ As in the original surgical ICU study, a variety of secondary end-points were improved in this study, including a lower incidence of AKI and need for RRT. In the original surgical ICU study, severe AKI (peak SCr > 2.5 mg/dl [ > 221 µmol/l]) developed in 7.2% of the IIT group, compared to 11.2% of the CIT group (P = 0.04); the incidence of RRT was also lower in the IIT group than the CIT group (4.8% vs. 8.2%, respectively; P = 0.007).¹³⁸ In the medical ICU study, the IIT group similarly had a significantly lower rate of AKI (doubling of SCr, 5.4%) than the CIT group (8.9%, P = 0.04), although RRT incidence was not decreased.¹⁴¹ In a recent analysis, Schetz et al.¹⁴² combined the renal end-points of both of these trials and used a modified version of the RIFLE classification of AKI to demonstrate that tight glycemic control reduced the incidence of severe AKI (peak SCr increments two- or three-fold increased from baseline) from 7.6% to 4.5% (P = 0.0006) in a combined patient population of 2707. The need for RRT was not decreased in the overall population or the medical ICU population, but was significantly lower in the surgical ICU patients managed with IIT (4% vs. 7.4%, P = 0.008).

Several newer studies have provided additional insight concerning the efficacy and safety of tight glycemic control in critically ill patients^{93,95,143–146}. Thomas et al.¹⁴⁵ conducted a systematic review of randomized trials of tight glycemic control in 2864 critically ill patients, and found a 38% risk reduction of AKI with IIT, and a nonsignificant trend towards less acute dialysis requirement. However, IIT was also associated with a greater than four-fold increase in the risk of hypoglycemia. A body of literature demonstrating that uncontrolled hyperglycemia was associated with increased AKI following cardiac surgery led to the conduct of a 400- patient, single-center RCT of tight vs. conventional intraoperative glucose control.^143,144 The investigators found that this approach did not decrease perioperative morbidity or mortality (included in a composite end-point that included AKI within 30 days of surgery): the composite end-point occurred in 44% of the IIT group vs. 46% of the CIT group. Although the incidence of hypoglycemia was similar in the groups, there was a significantly higher incidence of stroke in the IIT group (4.3%) compared to the CIT group (0.54%), as well as trends towards higher mortality and more postoperative heart block in the IIT group, raising concerns about the safety of this approach.

Further prospective comparison of IIT vs. CIT in critically ill septic patients was provided in the VISEP trial, which also incorporated a comparison on crystalloid vs. colloid infusions in a 2 x 2 factorial design.⁹³ Patients with severe sepsis or septic shock in 18 ICUs were randomized to IIT (target glycemia 80–110 mg/dl [4.44–6.11mmol/l]; n = 247) or CIT (target glycemia 180–200 mg/dl [9.99–11.1mmol/l]; n = 290) (Suppl Tables 2 and 3). There were no significant differences in 28-day or 90-day mortality, Sequential Organ Failure Assessment scores, or AKI rates between the groups. However, hypoglycemia (blood glucose level <40 mg/dl [ < 2.22 mmol/l]) was more frequent in the IIT group (12% vs. 2%; P < 0.001) and led to early termination of the IIT study arm. Following publication of this study, Thomas et al., updated the meta-analysis (discussed above) to include these data, and reported that, with the addition of the VISEP data, the analysis of a 3397-patient group found a 36% risk reduction of AKI with IIT, but this pooled estimate was no longer statistically significant (relative risk [RR] 0.74; 95% CI 0.47–1.17).⁹⁵ In a detailed review of the VISEP trial, Thomas et al., also noted that another multicenter mixed ICU trial of intensive insulin therapy (the GLUCOCONTROL Study: Comparing the effects of two glucose control regimens by insulin in intensive care unit patients; available at: http://www.clinicaltrials.gov/ct/show/NCT00107601) was stopped after 1101 patients were enrolled because of greater rates of hypoglycemia with IIT.⁹⁵ Such data have raised significant concerns regarding the effectiveness and safety of using IIT with tight glycemic control to prevent or ameliorate morbidity and mortality in patients at high risk of AKI and other forms of organ injury.

The recent meta-analysis of IIT vs. CIT by Wiener et al.¹⁴⁶ continued to find a greater incidence of hypoglycemia with IIT, but the balance of evidence now suggests no improvement in survival with this approach. Twenty-nine RCTs totaling 8432 patients contributed data for this meta-analysis. Twenty-seven studies reported no difference in hospital mortality (21.6% in IIT vs 23.3% in CIT) with a pooled RR of 0.93 (95% CI 0.85–1.03; P = NS). Nine studies reported no difference in incidence of new RRT. There was a significant benefit of tight glycemic control in reducing the incidence of septicemia but this was associated with a significantly increased risk of hypoglycemia (blood glucose < 40 mg/dl [ < 2.22 mmol/l]) in patients randomized to IITwith a pooled RR of 5.13 (95% CI 4.09–6.43; P < 0.05).

In summary, pooled analysis of early multicenter studies has failed to confirm the early observations of beneficial effects of IIT on renal function; the risk of hypoglycemia with this approach is significant, and even the survival benefits of IIT are in doubt. More recently, the international Normoglycemia in Intensive Care Evaluation and Survival Using Glucose Algorithm Regulation (NICE-SUGAR) study, with a targeted enrolment of 6100 patients, set out to definitively determine the risk-benefit comparison of tight glycemic control in critically ill patients (Suppl Table 3).^147,148 In this trial, adult patients were randomized within 24 hours after admission to an ICU to receive either intensive glucose control (target blood glucose range of 81–108 mg/dl [4.50– 5.99mmol/l]), or conventional glucose control (target of ≤ 180 mg/dl [ ≤ 9.99mmol/l]).¹⁴⁸ The primary outcome was mortality from any cause within 90 days after randomization. The two groups had similar characteristics at baseline. A total of 829 patients (27.5%) in the intensive-control group and 751 (24.9%) in the conventional-control group died (OR for intensive control, 1.14; 95% CI 1.02–1.28; P = 0.02). The treatment effect did not differ significantly between surgical patients and medical patients. There was no significant difference between the two treatment groups in incidence of new RRT (15.4% vs. 14.5%), respectively. Severe hypoglycemia (blood glucose level ≤ 40 mg/dl [ ≤ 2.22 mmol/l]) was reported in 6.8% in the intensive-control group and in 0.5% in the conventional-control group (P < 0.001). In summary, the largest randomized trial of intensive vs. conventional insulin therapy found that intensive glucose control actually increased mortality among adults in the ICU: a blood glucose target of ≤ 180 mg/dl ( ≤ 9.99 mmol/l) resulted in lower mortality than did a target of 81–108 mg/ dl (4.50–5.99 mmol/l). Furthermore, this trial confirmed the consistent finding of an increased incidence of hypoglycemia associated with IIT, without any proven benefit in reducing mortality, organ dysfunction, or bacteremia.

There were some methodological differences between the Leuven and NICE-SUGAR studies, possibly explaining the different outcomes.¹⁴⁹ These comprised different target ranges for blood glucose in control and intervention groups, different routes for insulin administration and types of infusion pumps, different sampling sites, and different accuracies of glucometers, as well as different nutritional strategies and varying levels of expertise. Finally, Griesdale et al.¹⁵⁰ performed a meta-analysis of trials of intensive vs. conventional glycemic control that included most of the studies in the Wiener meta-analysis, in addition to some newer studies, including data supplied by the NICE-SUGAR investigators. All 26 trials that reported mortality found a pooled RR of death with IIT compared to CIT of 0.93 (95% CI 0.83–1.04). Among the 14 trials reporting hypoglycemia, the pooled RR with IIT was 6.0 (95% CI 4.5–8.0). However, in subset analysis, patients in surgical ICUs appeared to benefit from IIT while patients in the other ICU settings (medical or mixed) did not. Although results from the early trials were better in studies that included surgical¹³⁸ rather than purely medical ICU patients¹⁴¹, and this latest meta-analysis appears to confirm that trend, it should be noted that no such phenomenon was noted in the NICE-SUGAR trial. Overall, the data do not support the use of IIT aiming to control plasma glucose below 110 mg/dl (6.11mmol/l) in critically ill patients, although subset analyses suggest that further trials may disclose benefits in perioperative patients, and perhaps through the use of less-intensive glucose control targets.

Considering the balance between potential benefits and harm (see Suppl Table 2), the Work Group suggests using insulin for preventing severe hyperglycemia in critically ill patients, but in view of the danger of potentially serious hypoglycemia, we recommend that the average blood glucose should not exceed 150 mg/dl (8.33mmol/l), but that insulin therapy should not be used to lower blood glucose to less than 110 mg/dl (6.11mmol/l). The Work Group recognizes that these proposed thresholds have never directly been examined in RCTs but are interpolated from the comparisons tested in the trials so far.