Intensive Versus Intermediate Glucose Control in Surgical Intensive Care Unit Patients

Hyperglycemia is common in acutely ill patients, including those treated in intensive care units (ICUs) (1). Until 2001, neglecting hyperglycemia was standard ICU care because a very impressive large randomized trial involving patients admitted to a surgical ICU showed that intensive insulin therapy (IT), targeting a blood glucose concentration of 4.4–6.1 mmol/L, significantly reduced in-hospital mortality (2). However, trials examining the effects of tight glycemic control (TGC) have had conflicting results (1,3–6). Systematic reviews and meta-analyses have also led to differing conclusions (7,8). The main reason these clinical trials and meta-analyses had negative results for TGC was the high incidence of hypoglycemia (10–17%) induced by intensive IT (7,8).

In recent years, it was reported that the results of the updated meta-analysis including NICE-SUGAR study data also do not support widespread adoption of intensive IT in critically ill patients (9). In this meta-analysis, however, intensive IT may be beneficial to patients admitted to a surgical ICU, although the characteristics of such patients remain to be clearly defined, as do the effect of different blood glucose algorithms, the method of measuring blood glucose, and the influence of nutritional strategies (9). Furthermore, this meta-analysis including NICE-SUGAR study data described that intensive IT significantly increased the risk of hypoglycemia. Effective control of surgical site infection (SSI) can reduce the length of hospitalization and can result in enhanced recovery after surgery. Perioperative hyperglycemia in critically ill surgery patients increases the risk of SSI, which is a common surgical complication. In various surgical settings, it is important to ensure optimal blood glucose levels in order to reduce SSIs. Unfortunately, however, the optimal perioperative blood glucose range to improve surgical outcomes including SSI remains unclear. In the past 10 years, solving the problem of methodology for intensive IT, especially the incidence of hypoglycemia during intensive IT, did not progress at all. The true effect of intensive IT in clinically ill patients cannot be evaluated without solving the occurrence of hypoglycemia during intensive IT. The development of accurate, continuous blood glucose–monitoring devices and closed-loop systems for computer-assisted blood glucose control in the ICU will probably help avoid hypoglycemia in these situations (10).

Based on these findings, we conducted this prospective study to investigate which perioperative glycemic control range was better for reducing SSI—between intensive IT, with target blood glucose range of 4.4–6.1 mmol/L, and intermediate IT, with target blood glucose range of 7.7–10.0 mmol/L—using a closed-loop glycemic control system.

We conducted a parallel-group, randomized, controlled trial involving adult surgical patients admitted to the ICU of our hospital. Patients were randomized at the time of the starting operation. Also, blood glucose control of recruited patients was begun from the start of surgery. All eligible patients were those surgically treated for hepato-biliary-pancreatic diseases and admitted to the surgical ICU overnight until postoperative day 1 after hepatic and/or pancreatic resection to control postoperative blood glucose concentration levels by using closed-loop glycemic control system. Patient exclusion criteria included a body weight loss >10% during the 6 months prior to surgery, the presence of distant metastases, or seriously impaired function of vital organs due to respiratory, renal, or heart disease. Patients were informed of the purpose and details of the study, and written consent was obtained from them prior to enrolment. The study was approved by the local ethics committee at the Kochi Medical School and carried out in accordance with the Declaration of Helsinki. All studies were performed at the Kochi Medical School between August 2008 and August 2012.

A complete physical examination was undertaken and clinical histories were obtained from all patients. Laboratory tests included the measurement of serum levels of albumin, total bilirubin, total cholesterol, and alanine and aspartate aminotransferases, as well as peripheral blood cell counts and prothrombin time. The patients’ clinical history included details of glucose metabolism, including medication for the treatment of diabetes, fasting blood glucose level, and the monitoring of hemoglobin A_1c (HbA_1c) levels in plasma. Of all patients included in this study, diabetes mellitus (DM) status was checked by diabetologist, and DM were all type 2. Operation-related parameters of liver and pancreatic surgery were also evaluated.

In our study, all patients were given parenteral nutrition (PN) postsurgery just after bring admitted to the surgical ICU. The total caloric requirement was calculated according to the Harris-Benedict equation (11). All of the patients who were assigned to this study received intravenous PN and 20% intravenous fat emulsions and 25% glucose solution, which included an adequate amount of protein, glucose, vitamins, minerals, and trace element such as zinc, iron, copper, manganese, and iodine (rate of glucose administration <5 mg/kg/min). Enteral feeding was attempted as soon as possible when the patients were hemodynamically stable. The amount of PN was calculated daily as the difference between the total energy intake that was effectively delivered by enteral nutrition and the calculated caloric goal when postoperative oral intake would be restarted.

The Nikkiso Company (Tokyo, Japan) developed the closed-loop glycemic control system in 1984. The closed-loop glycemic control system is a reliable and accurate device that measures blood glucose concentration compared with the ABL 800FLEX machine (Radiometer Medical ApS, Brønshøj, Denmark) recommended by the National Committee for Clinical Laboratory (12,13). The closed-loop glycemic control system is composed of a glucose sensor, which performs glucose detection/monitoring, and pumps for infusing the appropriate amount of insulin or glucose (14). The insulin and glucose pumps are computer regulated based on a target blood glucose value predefined prior to the system initiation. Peripheral blood for glucose monitoring was sampled continuously at 2 mL/h during at least the first postoperative 18 h in the surgical ICU. Furthermore, the closed-loop artificial pancreas system was used to evaluate the patient’s insulin requirements.

Study participants were randomly assigned to glucose control with one of two target ranges: the intensive (i.e., tight) control target of 4.4–6.1 mmol/L (intensive IT group), based on that used in previous studies (2,15), or an intermediate control target of 7.7–10.0 mmol/L (intermediate IT group), based on that used in a previous study (16). Blood glucose concentration (intermediate vs. intensive) was controlled by a closed-loop glycemic control system from the start of surgery, and either intermediate or intensive IT was performed during the operation. After patients were removed from the closed-loop glycemic control system at postoperative day 1, blood glucose levels were routinely checked by nursing staff and were controlled by the subcutaneous injection of regular human insulin every 2 h: the dose was determined by the commonly used sliding scale, and the target blood glucose level to avoid hypoglycemia was 8.3–11.1 mmol/L (17).

Baseline demographic and clinical characteristics of the patients were well matched between the two study groups (Table 1). Local study coordinators at our institution collected the data; source data were verified by study monitors from our coordinating center at the department of biostatistics. Daily records were kept regarding all intensive care treatments and surgical procedures, new bacterial or fungal infections, the results of blood and urine chemical analyses and hematologic studies, and markers of inflammation. Also recorded were the total energy intake delivered daily by means of enteral and PN, interruptions of delivery of enteral nutrition, and feeding-related complications. In addition, whenever practically feasible, the functional status of patients before hospital discharge was quantified.

Safety end points included vital status (the rates of death in the ICU and the hospital) and especially the rates of hypoglycemia. Hypoglycemia that was resistant to parenteral glucose administration during the intervention window was considered to be a serious adverse event. And a blood glucose level of ≤40 mg/dL was considered a serious adverse event.

The primary end point of this study was to determine whether the incidence of SSI is reduced by perioperative intensive IT. The Centers for Disease Control and Prevention’s National Nosocomial Infections Surveillance has developed standardized surveillance criteria for defining SSI (18). According to these criteria, the infection control team evaluated whether postoperative SSI in patients who underwent liver and pancreatic resection had occurred.

The secondary end point was to evaluate the duration of hospital stay in each patient group. To reduce bias that might result from variability in the availability of beds on regular wards, we defined the time to discharge from our institute as the time by which patients were ready for hospital discharge, according to the suggestion by trained expert staff of the Department of Diagnostic Medicine at our institute and nutritional support team blindly.

Given that the primary outcome of the study is SSI, we aim to detect a clinically relevant difference in 5% of SSI incidence concretely from 10 to 5% of patients during hospitalization. This difference is in accordance with previous reports (19,20) and is considered clinically significant (21). Consequently, the required number to detect this difference with 80% is 219 patients per group. Allowing for attrition of 10% of cases with SSI, we have to recruit 438 patients and randomly assign 219 patients into the intensive IT group and 219 patients into the intermediate IT group (Fisher exact test, two-sided, intention-to-treat basis). Because of the probably rate of dropout that is estimated, ~10% of the patients initially enrolled into the study, additional patients were randomized, resulting in a total subject number of 482 across the two groups. Continuous variables are presented as the mean ± SD. Dichotomous variables are presented as both number and percentage values. P < 0.05 was considered significant. Data were analyzed using the Student t test (two-tailed), with dichotomous variables analyzed by the χ² test (two-tailed) or Fisher exact test (two-tailed), as appropriate.

Subgroup analyses on intent to treat for the primary outcome were based on an unadjusted test of interaction in a logistic model to evaluate which patients would benefit from intensive IT during the surgical ICU to reduce the incidence of postoperative SSI. All analyses were performed using SPSS (SPSS, Chicago, IL). The data were analyzed by an independent statistician of the Department of Biostatistics (Kochi University). The analyses were reviewed by the independent data and safety-monitoring committee, which was charged with recommending that the trial be stopped if there was evidence beyond a reasonable doubt of a difference in the rate of postoperative complications, such as hypoglycemia, hospital death, bile leakage from liver stump after hepatic resection, and pancreatic fistula after pancreatic resection, from any cause between the two treatment groups.

Participants were recruited and had follow-up during the period from August 2008 through August 2012; 502 were randomly assigned to one of the two treatment groups: 252 to intermediate glucose control and 250 to intensive glucose control (Fig. 1). Study treatment was discontinued prematurely in 27 of 252 patients (10.7%) in the intermediate IT group and 28 of 250 patients (11.2%) in the intensive IT group. There were five in-hospital deaths in the intermediate IT group (mortality rate 1%). Four patients underwent hepatectomy for hepatocellular carcinoma. Two of four hepatectomized patients, who were suffering from severe liver cirrhosis, fell into hepatic failure owing to limited liver function of the remnant liver and died after surgery (42 days and 47 days after liver resection for hepatocellular carcinoma). One patient of hepatectomized patients, who had a gastrointestinal bleeding after liver resection, developed sequential hepatic failure and died 63 days after surgery. Another hepatectomized patient, who had an accidentally dissecting aneurysm of aorta, died 2 days after surgery. The remaining patient, who had a 4-cm-sized adenocarcinoma located in the head of the pancreas, with insufficiency of pancreato-enteric anastomosis, was relaparotomized for diffuse peritonitis 27 days after the initial pancreatico-duodenectomy, ran a fatal course due to subsequent intra-abdominal septic complications, and died on the 91st postoperative day. Reasons for discontinuation were withdrawal because of five in-hospital deaths in the intermediate IT group or closed-loop glycemic control system errors in patients who underwent IT during the surgical ICU for 22 of the 252 patients (8.7%) assigned to intermediate IT and 28 of the 250 patients (11.2%) assigned to intensive IT; thus, study data were available for 225 and 222 patients, respectively. Of the 50 patients for whom study data were unavailable, we could recognize 42 cases (8.5%) of an insufficiency for blood sampling from peripheral vessels and 8 cases (1.6%) of trouble with the glucose sensor of the closed-loop glycemic control system itself.

The baseline characteristics of the treatment groups were similar (Table 1). The mean ± SD age was 66.4 ± 10.4 and 66.7 ± 10.1 years in the intermediate IT group and the intensive IT group, respectively; the percentage of male patients 67.1% and 64.0%; and the mean BMI score, 23.1 ± 3.4 and 23.3 ± 3.6. Also, both preoperative liver and renal function were similar in the two groups. Furthermore, the presence of a previous medical history for DM was equally distributed between the two groups. In the current study, because all of the surgeries were scheduled procedures, DM status in patients who were recruited in our study was preoperatively well controlled by the diabetologist; the mean HbA_1c levels were 5.6 ± 1.1% in the intermediate IT group and 5.6 ± 0.9% in the intensive IT group, respectively.

All patients underwent either hepatectomy or pancreatectomy consisting of curative resection of both hepatic and pancreatic tissue for the removal of a tumor. There was no significant predisposition to these operative procedures between the two groups; liver surgery for 142 cases in intermediate IT group and 148 cases in intensive IT group and pancreatic surgery for 83 cases in intermediate IT group and 74 cases in intensive IT group, respectively (Table 2). The operation time and estimated volume of blood loss did not differ significantly between the two groups (operation time 36.2 ± 144.2 min in intermediate IT group and 337.2 ± 142.1 min in intensive IT group; estimated blood loss volume 768.7 ± 764.5 mL in intermediate IT group and 762.1 ± 720.0 mL in intensive IT group) (Table 2).

All patients received PN just after being admitted to the surgical ICU. The mean daily amount of calories administered was 1,479 ± 207 kcal in the intermediate IT group and 1,465 ± 216 kcal in the intensive IT group, respectively. Enteral feeding was attempted as soon as possible when the patients were hemodynamically stable. Postoperative blood glucose levels in the intermediate and intensive IT groups during the first 18 h after surgery are shown in Fig. 2A. The average blood glucose levels in patients from the intermediate IT group were controlled to the target zone (7.7–10.0 mmol/L). The percentage in blood glucose target at intermediate glucose control by a closed-loop system during surgical ICU was 96.8% (Fig. 2A), although blood glucose levels gradually increased during the first 6 h after the hepatectomy and pancreatectomy and reached a plateau of ~13.8 mmol/L between 4 and 8 h by the ordinary blood glucose control methods by using sliding scale and then the concentrations gradually decreased toward 8.3 mmol/L by 18 h after surgery in our previous studies (21,22). Also, the average blood glucose levels in patients from the intensive IT group were well controlled to the target zone (4.4–6.1 mmol/L) but did not decrease to hypoglycemic levels, and the percentage within the blood glucose target for intensive glucose control by a closed-loop system during surgical ICU was 85.8% (Fig. 2A). Patients undergoing intensive glucose control received a larger mean insulin dose (101 ± 88 units [range 8–390] in the surgical ICU vs. 77 ± 72 units [range 0–355] for the intermediate IT group) for 18 h after the hepatic resection (Fig. 2B). This represents a significant difference in total insulin consumption between the two groups (P = 0.015).

The study groups had similar rates of death in the ICU and the hospital. However, mortality occurred solely in the intermediate IT group, although postoperative complications occurred in these patients after surgery. This included four patients in postoperative liver failure after hepatic surgery and one patient in postoperative acute respiratory distress syndrome after pancreatic surgery. No patients in either group became hypoglycemic (<2.2 mmol/L) during their stay in the surgical ICU or during the hospitalization.

The short-term outcomes of patients who underwent hepatic and pancreatic resection are detailed in Table 2. With respect to postoperative characteristics, when blood glucose levels were tightly controlled using the closed-loop glycemic control system, the incidence of SSI in the intensive IT group was significantly lower than that in the intermediate IT group (P = 0.028). There was no readmission of patients who had postoperative SSI in the current study.

Postoperative morbidities were examined by abdominal ultrasonography or computed tomography. There was no significant difference between the two groups in the incidence of bile leakage after liver resection. Interestingly, the incidence of postoperative pancreatic fistula formation after pancreatic surgery in the intensive IT group was significantly lower than that in the intermediate IT group (P = 0.040). It is of note that patients in the intensive IT group required a significantly shorter hospitalization than patients in the intermediate IT group (P = 0.017). There was one operation-related readmission of patients in the intensive IT group due to the abdominal abscess formation for delayed pancreatic fistula in the current study.

In postsubgroup intent-to-treat analyses, we evaluated association between the incidence of postoperative SSI and perioperative predictable factors. Three parameters were identified as independent markers for the occurrence of postoperative SSI by the Spearman correlation analysis: 1) elderly patients (>65 years old) (P = 0.039), preoperative predictable factor; 2) patients with DM (P = 0.036), preoperative predictable factor; and 3) the presence of pancreatic fistula formation after pancreatic resection (P = 0.047), postoperative factor after the surgical management. Considering DM status, which was a preoperatively predictable factor, the rate of SSI was lower in patients without DM (3.7%) than in patients with DM (19.0%, P < 0.001).

We found that there was a significant difference in the incidence of postoperative SSI between intermediate and intensive IT group in the surgical ICU among patients who underwent hepato-biliary-pancreatic surgery. The major strength of this study is the restriction to ICU patients admitted after hepatic and pancreatic surgery. ICU patients comprise a heterogeneous population, and different effects in subgroups of ICU patients may partly explain the conflicting findings in previous trials (1,2,15). The NICE-SUGAR trial was criticized for the overlapping blood glucose levels in the two treatment arms (1). In our study, because the use of a closed-loop glycemic control system was promising for maintaining blood glucose within a reasonable narrow range, 55 (11.0%) of 502 were excluded after randomization to remove the potential impact of any selection bias from dropouts from the study, and we recalculated again the exact number to obtain the adequate results by a biostatistician (23). Stratification by diabetes status is very relevant, as the impact of intensive IT may be different in this subgroup. Intensive IT in the surgical ICU was also associated with a lower incidence of postoperative pancreatic fistula and a shorter duration of hospital stay compared with the intermediate IT group. The current work examined both the safety of intensive IT using the closed-loop glycemic control system and postoperative SSI in an exclusively surgical population. Also, this novel intensive IT using a closed-loop glycemic control system enabled us to maintain stable glycemic control—not only no hypoglycemia but also no hyperglycemia—with less variability of blood glucose concentration. Moreover, our findings support the hypothesis by van den Berghe that accurate continuous blood glucose–monitoring devices and closed-loop systems for computer-assisted blood glucose control in the ICU will likely reduce the incidence of hypoglycemia in intensive IT (10).

Although complication rates in hepato-biliary-pancreatic surgery are improved, wound infections remain a common complication (24). One of the most important risk factors for development of any SSI in patients undergoing general surgery was DM (25,26). Also, our study demonstrates significant associations between the development of SSI and patients who were suffering from DM, and the incidence of SSI was lower in patients without DM than in patients with DM, whereas, in our study, the incidence of postoperative SSI was acceptable in both groups, even if perioperative blood glucose levels were controlled by either intermediate IT or intensive IT. Interestingly, the rate of postoperative SSI in patients with DM was significantly reduced by intensive IT using a closed-loop glycemic control system by subgroup analysis (Fig. 3). The current study suggests that intensive IT might be beneficial to patients with DM admitted to a surgical ICU after hepato-biliary-pancreatic surgery because there were correlative associations between the intensive IT and the reduction of the incidence of postoperative SSI in patients with DM.

Appropriate nutritional support and glycemic control play important roles in the promotion of enhanced recovery after surgery for surgical patients. Various consensus statements and guidelines from international societies recommend the early and adequate initiation of specialized nutrition therapy for critically ill patients admitted for surgical conditions (27,28). Although the formulation and caloric goals of PN solutions have changed in recent years, debate still exists about the timing of the initiation of PN when enteral nutrition either is impossible or does not meet the nutritional goals. Thus, the total caloric energy intake per day during the perisurgical period was assumed to be equal between the two groups. However, there is a dilemma because promotion of nutritional support is likely to result in hyperglycemia, and TGC with the open-loop system always has the risk of hypoglycemia. Indeed, in previous reports of conventional intensive IT without the closed-loop glycemic system, there was risk not only of hypoglycemia associated with TGC but also of insufficient nutritional support mainly because of fear of hyperglycemia (7,9). Therefore, compatibility of sufficient nutritional support and intensive IT seemed to be difficult in critically ill patients. Surprisingly, even in such patients, this study showed that glycemic control using a closed-loop glycemic control system enabled us to complete intensive IT with a target blood glucose range of 4.4–6.1 mmol/L, although postoperative nutritional support in every patient undergoing hepatic resection or pancreatic resection was carried out by calorie intake according to the calculations of the Harris-Benedict equation. Namely, if we used the closed-loop glycemic control system, compatibility of appropriate nutrition support and TGC including intensive IT was possible simultaneously even in critically ill surgical patients. Virtually all large studies comparing an intensive versus an intermediate glucose control range have shown that the incidence of hypoglycemia is higher when targeting a lower goal range of 4.4–6.1 mmol/L. In the past 10 years, the main problem discussed about methodology for intensive IT was the incidence of hypoglycemia during intensive IT; however, that was nonsense. In our study, because we used the closed-loop glycemic control system in the surgical ICU, the true effect of intensive IT in the clinically ill patients could be evaluated without the solution of the occurrence of hypoglycemia during intensive IT. Furthermore, the mean amount of insulin infusion requirement (range 32–71 units/day) was too small to perform the TGC, although the incidence of hypoglycemia differs between surgical and medical populations and patients’ individual insulin resistance should be evaluated (27). At least 70 units i.v. insulin was needed to control the blood glucose levels with the postoperative adequate caloric requirement that was calculated according to the Harris-Benedict equation, even if the targets of blood glycemic concentration were set between 7.7 and 10.0 mmol/L (intermediate IT).

What is the optimal blood glucose range to improve morbidity and mortality in surgical patients? Nobody knows that a glycemic target range of 7.7–10.0 mmol/L (intermediate IT) for a patient in an ICU could be recommended, although neglecting hyperglycemia was standard ICU care (29). In our study, it remained unclear that the advantage of intensive IT for patients who underwent hepato-biliary-pancreatic surgery was an effect in intraoperative IT, postoperative IT, or both. Also, it has not been clarified how long should postoperative blood glucose levels were controlled adequately. In the current study, TGC was performed for 18 h in patients with liver resection at the surgical ICU, and excellent glucose control was successfully observed without hypoglycemia by using a closed-loop glycemic control system. A corollary issue that has arisen from this study is why an abbreviated period of glycemic control had an effect on infection. Note the increase in growth of bacteria, including Escherichia coli, Streptococcus, Proteus, Staphylococcus, and Pseudomonas, displayed for between only 2 h and 18 h at growth curves of aerobic and facultative organisms (30). We believe that the duration of hospitalization after both liver and pancreatic surgery was reduced in patients who received perioperative TGC with intensive IT and that this could be associated with the reduction in postoperative complications due to SSI.

Our results support the conclusions from previous observational studies that intensive IT improves the outcome for critically ill patients, especially in surgical ICU patients (2,9). Because all of the scheduled patients were recruited in the current study, the mortality rate was very low and acceptable compared with previous large studies (25). Of course, limitations of our trial include the use of a subjective criterion the same as previous reviews and meta-analysis describing bias, chance, and atypical clinical practices (7–9). Although our trial was not blinded, which could lead to bias, postoperative calorie administration and surgical managements were matched in both groups by the same staff at our single center. We did not collect specific data to address potential biologic mechanisms of the trial interventions, and on the basis of the results in the predefined pairs of subgroups, we believed that intensive TGC may benefit some surgical patients. Our results suggest that the prospective large randomized control trials should be reprogrammed to evaluate the true effect of TGC for patients in both the surgical and medical ICU for the conclusion of the argument over the incidence of hypoglycemia during intensive IT and propose that the closed-loop glycemic control system is an easy and effective way to control the blood glucose levels and perform insulin treatment in the ICU without occurrence of hypoglycemia during intensive IT.

In conclusion, with use of a closed-loop glycemic control system, perioperative intensive IT was significantly better than intermediate IT to reduce SSI and morbidity, resulting in shorter postoperative hospitalization. An important question that has arisen from this study is how to choose candidates for intensive IT using a closed-loop system. Further studies will be required to address this issue.

Funding. This work was supported by the Kochi Organization for Medical Reformation and Renewal grants.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. T.O. wrote the manuscript and researched data. Y.S. reviewed and edited the manuscript. T.Sum., A.K., T.T., and T.I. researched data. T.Sug., M.K., and M.Y. contributed to discussion and reviewed and edited the manuscript. K.H. contributed to discussion, reviewed and edited the manuscript, researched data, and contributed to discussion. T.O. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.