The trusted source for
healthcare information and
Intensive Insulin Therapy: A Bit More Sugar May Be Nicer
Abstract & Commentary
By Saadia R. Akhtar, MD, MSc, Idaho Pulmonary Associates, Boise, is Associate Editor for Critical Care Alert.
Dr. Akhtar reports no financial relationship to this field of study.
Synopsis: This large multicenter, randomized, controlled parallel-group trial found that intensive glucose management, compared to conventional (target glucose 81-108 mg/dL vs ≤ 180 mg/dL), increases 90-day mortality in both medical and surgical ICU patients.
Source: NICE-SUGAR Study Investigators; Finfer S, et al. Intensive versus conventional glucose control in critically ill patients. N Engl J Med 2009;360:1283-1297.
This study was designed to determine whether intensive glucose control in ICU patients reduces mortality at 90 days. Secondary outcomes included survival time during the 90 days, cause-specific death, duration of mechanical ventilation, need for renal replacement therapy, and hospital and ICU lengths of stay.
Forty-two hospitals (38 academic tertiary care centers and 4 community hospitals in Australia, New Zealand, and North America) took part in this trial of medical and surgical ICU patients. Eligible patients were expected to require ≥ 3 days of ICU care. Subjects were randomized to glucose control that was either intensive (target 81-108 mg/dL) or conventional (target ≤ 180 mg/dL, with insulin stopped once glucose < 144 mg/dL), per predefined treatment algorithms. Randomization was stratified by type of admission (operative or non-operative) and also by region. The target glucose control was continued until the patient was eating or discharged from the ICU, was resumed if the patient returned to the ICU within 90 days of randomization, and was discontinued permanently at 90 days after randomization or death, whichever came first. Severe hypoglycemia was defined by blood glucose ≤ 40 mg/dL. Usual demographic and clinical data were collected, including severity-of-illness scores and information about prior diagnosis of diabetes and recent corticosteroid use. Standard statistical methods and intention-to-treat analysis were utilized. An estimated sample size of 6100 was needed to detect a 3.8% mortality difference with a statistical power of 90% and two-sided alpha of 0.05.
Over 4 years, 6104 subjects were randomized (< 15% from North America). The 2 groups had no significant differences in baseline demographic or clinical characteristics. About one-third of the subjects in each group were operative. Median duration of study treatment was about 4 days, and mean time-weighted glucose levels were 115 mg/dL in the intensive control group vs 144 mg/dL in the conventional group. Study treatment was discontinued early due to adverse events in only 0.4% of subjects in the intensive group vs < 0.1% in the conventional group. More subjects in the intensive control group received corticosteroids.
At 90 days, the absolute difference in mortality between the 2 groups was 2.6% with a statistically significant odds ratio for death in the intensive control group of 1.14 (yielding a number needed to harm of 38); deaths from cardiovascular causes were more common in these patients. These results remained the same after adjustment for 6 predefined variables. There were no significant differences between the 2 groups in other secondary outcomes. As might be expected, severe hypoglycemia occurred more often in the intensive control group (6.8% vs 0.5%).
Observational prospective and retrospective studies over the past decade have clearly shown that severe hyperglycemia is associated with adverse outcomes.1 Conversely, there is some evidence that severe hypoglycemic episodes may lead to worse ICU outcomes, including mortality.2 Thus, although we can agree that glucose control is important, questions remain about what specific glucose level(s) we should be targeting.
Van den Berghe et al's 2001 study of intensive glucose control in a single surgical ICU sparked a huge change in clinical practice when it demonstrated a significant decrease in ICU mortality (as well as hospital mortality, infections, acute renal failure, and polyneuropathy) for patients managed with insulin drips to target glucose 80-110 mg/dL (the study achieved mean glucose of 103 mg/dL).3 The control group's target was glucose between 180-200 mg/dL, and insulin was not started until glucose reached > 215 mg/dL (mean glucose 153 mg/dL in this group). Severe hypoglycemia occurred in about 5% of patients in the intensive glucose control group. The study was criticized for high mortality in the control group (relative to the patients' severity of illness and diagnoses). It was also unclear how the center's early, aggressive use of combined parenteral and enteral nutrition would impact the generalizability of the results. Nevertheless, the findings were striking enough to lead to implementation of tight glucose control via similar protocols in ICUs across the world, often with extrapolation to nonsurgical ICU patients.
Subsequent studies in mixed or medical ICU populations, however, have provided conflicting results (often with high rates of severe hypoglycemia). Van den Berghe's own attempt to replicate the original study in a medical ICU failed to show an overall mortality benefit of intensive glucose management. There were some other improved outcomes (earlier liberation from mechanical ventilation and earlier ICU and hospital discharge). On subgroup analysis, those patients in the intensive glucose control group with ICU length of stay ≥ 3 days did appear to have improved mortality.4 Meta-analyses of studies of intensive glucose control in mixed or medical ICU populations have also not demonstrated improved mortality or overall benefit.5
NICE-SUGAR is a well-designed and well-executed trial that sways the pendulum further away from the tight glucose control recommended by Van den Berghe et al. NICE-SUGAR differs from the Van den Berghe investigations in several ways; perhaps most notably, the target (and actual achieved) glucose in the control group is considerably lower than in the prior studies, suggesting that more modest glucose control may be the true "holy grail" of glycemic management in the ICU. It leaves us to wonder whether there truly is mortality benefit to very tight glucose control in specific ICU populations. If so, how can those patients be identified? How can intensive glucose control be beneficial in post-cardiac surgery ICU patients while at the same time increasing deaths from cardiovascular causes in mixed ICU populations receiving similar glucose management? What is the threshold glucose above and below which harm may occur? What are the potential mechanisms of such harm or benefit?
Although it may raise more questions than it answers, NICE-SUGAR is one of the most noteworthy studies of the year. It has already changed practice in the mixed medical-surgical ICUs where I work and I hope the same is happening across the world. I suggest not initiating insulin therapy in ICU patients until blood glucose exceeds 180 mg/dL; at that point, insulin should be used to target modest glucose control (140-160 mg/dL). Whether intensive glucose control (as defined by Van den Berghe et al) may be beneficial in certain post-surgical ICU patients or in any patient beyond the first 3 days of an ICU stay is uncertain. I believe there are enough questions and enough evidence of potential harm from intensive glucose control (target 80-110 mg/dL) that more modest targets should be utilized until further studies clearly replicate prior positive results.