NICE-SUGAR study-Intensive Blood Glucose control in an ICU might be associated with increased mortality.

Find the original article here

https://www.nejm.org/doi/0full/10.1056/NEJMoa0810625

Author: Peter Ryu Tofts, MD

Editor: Martin M. Cearras, MD

Summary:

This is the largest and most comprehensive study that evaluated insulin control in ICUs to date. They included twice as many subjects as in prior studies (Leuven I & II), included surgical and medical ICUs, academic and community institutions as well as less parenteral feeding, improving its generalizability. This trial found that intensive glucose control in the ICU was in fact harmful to patients raising the mortality with a 2.6% absolute risk increase and a number needed to harm of 38. Even if there are some concerns regarding the fragility index, it seems reasonable to use a less intensive glucose control, even in the case this is a false positive.

PICOTT:

POPULATION:  

N= 6,104 Multiple Hospitals, academic & community 

Inclusion: Adult patients admitted to Medical and Surgical ICUs <24 hours and expected to require 3 days. 

Exclusion: Surgical patients and medical patients able to eat, patients with <3 days expected stay and patients with DNR.  

INTERVENTION: 

Treatment: Rx insulin infusion to maintain blood glucose (BG) 81 to 108 mg/dL  

COMPARATOR:  

Control: Rx insulin infusion when BG >180mg/dL to target 144 to 180 mg/dL 

OUTCOMES:  

• Primary: All-cause mortality at 90 days 

• Secondary: Survival within 90 days, Mechanical Ventilator days, Renal replacement Therapy, LOS ICU, LOS Hospital. 

• Tertiary: All-cause mortality at 28 days, place of death, new organ failure & Positive blood cultures. 

TYPE OF STUDY: 

Randomized Controlled Trial 

TYPE OF QUESTION: 

Therapy

Interpretation of the Study: 

The NICE-SUGAR found that intensive glucose control was in fact harmful to patients, raising the mortality by 2.6% with a NNH of 39. This was compounded by a much higher frequency of hypoglycemic events and the loss of all other markers of benefit such as reduced LOS and mechanical ventilator days noted in the Leuven I & II. While the baseline characteristics and prognosis of the groups were equal, it is noted that the intensive treatment group received more steroids than the other, a variable that remains controversial. The findings were cold water in the face of two large but smaller studies, however NICE-SUGAR remains the more valid study given the wider range of patients included and so it stands. Today we still aim for less intensive glucose targets. 

Primary outcome:  

• All-cause mortality at 90-days: Intensive Rx 829/3010 (27.5%) vs Conventional Rx 751/3012 (24.9%) OR 1.14 CI 1.02 to 1.28; P 0.02 RR 1.1 Relative Risk Increase (RRI) 10% Absolute Risk Increase (RD) 2.6%. NNH 38 

Secondary Outcomes:  

• Severe Hypoglycemia Intensive Rx 206/3016 (6.8%) vs Conventional 15/3014 (0.5%) P <0.001 RRI 1.36 RRI 36% RD 6.3% NNH 15 

• LOS ICU Median (IQR): Intensive Rx 6 vs Conventional Rx 6; P= 0.84 NS

• LOS in Hospital Median (IQR): Intensive Rx 17 vs Conventional 17; P=0.86 NS

• Days of Mechanical Ventilation: Intensive Rx 6.6 +/- 7 vs Conventional Rx 6.6 +/-7; P=0.17 NS

• Renal Replacement Therapy: Intensive Rx 465/3014 (15.4%) vs Conventional 438/3014 (14.5%); P=0.34 NS

  
  

Tertiary Outcomes: 

•  All-cause Mortality at 28 days: Intensive Rx 670/3010 (22.3%) vs Conventional Rx 669/741(20.8%) OR 1.09 CI 0.81 to 1.62; P 0.44 RRI 1.07 RRI 7% RD 1.5% NNH 66 

Overall Risk of bias: Low

The study was well constructed and executed. The baseline characteristics were prognostically similar at the onset but during the study the intensive treatment was noted to have significantly more steroid use (34.6% vs 31.7%; P=0.02) the significance of this is controversial. Sample size was increased from n=4000 to n=6100 based on the Leuven II trial in order to give the study 90% power to detect an absolute mortality difference of 3.8% and a wide range of patients, hospitals and ICUs were included expanding its external validity.

One interesting point is the fact that mortality was statistically significant at 90 days, but not at 28 days, for an intervention that only lasted a median of 4.3 days.

This gives me pause, as the calculated fragility index for the study was 11 (Robust), which indicates that if 11 patients in the control group were "converted" from NOT having the primary endpoint to HAVING the primary endpoint, the study would lose statistical significance.

Unfortunately, due to consent withdrawal, there was a small loss to follow up, 1.4% (82 patients) that could have affected and even changed this result in a best - worse, case scenario, since the absolute risk reduction is also fairly small. If the loss to follow up is larger than the fragility index (82 is greater than 11) the results are a lot less robust. In addition, patients lost to follow up usually have a different prognosis than the others.

There were multiple secondary and even tertiary outcomes, increasing the chance of type I errors (false alarms or false positives)

Teaching Points:

Type I error

Power

Intention to treat

Primary outcomes vs secondary outcomes.

Fragility / robustness Index

Loss to follow up

Verdict:

Mostly Settled. It is unlikely that a larger study will be conducted at this point. But more studies with less loss to follow up could change the current understanding of this topic.