TGC-Fast: Tight Blood-Glucose Control without Early Parenteral Nutrition in the ICU

11 June 2025 Daniel Chung Intensive Care Medicine Leave a comment

Tight Blood-Glucose Control without Early Parenteral Nutrition in the ICU

Gunst et al. N Engl J Med 2023;389:1180-1190 DOI: 10.1056/NEJMoa2304855

Clinical Question

In ICU patients not receiving early parenteral nutrition, does liberal glucose control (initiating insulin when blood glucose level [BGL] >11.9 mmol per liter [>215 mg per deciliter]) or tight glucose control (commencing insulin at BGL 4.4 to 6.1 mmol/L [80 to 110 mg/dL]) increase ICU length of stay or mortality at 90 days?

Background

Optimal blood glucose targets in intensive care patients have been the subject of significant research attention
Hyperglycaemia is associated with poor outcomes in a range of disease states
Earlier trials suggested potential benefits of tight glucose control (TGC) in critically ill populations
- DIGAMI (1995) n=620, multicenter RCT in Sweden demonstrated a mortality benefit (1 year mortality; 18.6% vs 26.1%) in diabetic patients post myocardial infarction with insulin infusion followed by SC insulin with subsequent tighter glycemic control
- Van den Berghe et al. aka Leuven Surgical Trial (2001) n=1548 demonstrated a mortality benefit (4.6% mortality vs 8%) in surgical ICU patients at a single centre with intensive insulin therapy (BGL 4.5-6.5mmol/L) vs conventional therapy (BGL average 8mmol/L)
- Van den Berghe et al. aka Leuven Medical Trial (2006) n=1200 demonstrated a non-statistically significant trend towards reduced mortality and significantly reduced morbidity in a single medical ICU
However, the landmark NICE-Sugar trial (2009), largely ended the practice of tight glycaemic control
- Multicentre, multinational RCT, n=6,104
- Conventional control (BGL <10mmol/L [180mg/dL]) vs intensive control (BGL 4.5-6.0mmol/L [81-108mg/dL])
- Statistically significant difference in 90-day mortality favouring conventional control (27.5% vs 24.9%, odds ratio 1.14 (95% CI 1.02-1.29; p=0.02)
  - This may be mechanistically driven by higher rates of severe hypoglycaemia in the intensive control arm; 6.85% vs 0.5% in conventional-control arm
- Criticisms of NICE-SUGAR include
  - Non-standardised BGL measurements and insulin titration may have increased risk of hypoglycaemia especially in intensive control group who received more insulin
  - High rates of early TPN (as was guideline recommended practice at the time) may have increased insulin utilisation
As such, contemporary guidelines (eg from American Diabetes Association) recommend targets around BGL 5.6–10.0 mmol/L [100–180 mg/dL] in ICU patients

Design

Investigator -initiated prospective multicenter parallel group randomised controlled trial
1:1 randomisation in permuted blocks of 10 via centralised computer system, stratified according to center and diagnostic category
Patients, families and outcome assessors blinded to treatment assignment
Unblinded clinical staff
All analyses performed by intention to treat analysis
9230 patients required to give 80% power to detect a 1 day difference in primary outcome and exclude a 1.5% increase in mortality with the intervention

Setting

11 ICUs at 3 hospitals in Belgium
Recruitment from September 2018 to August 2022
- Brief pause in 2020 due to COVID-19 pandemic mandated by central ethics committee
- Closed COVID-19 units excluded from randomisation thereafter

Population

Inclusion: All adults (≥18 years of age) admitted to participating ICUs
Exclusion:
- Do not resuscitate order
- Moribund patients
- Orally fed – deemed not critically ill
- No arterial or central venous access present or planned
- Planned to receive TPN within 1 week of admission
- Previous inclusion in TGC-Fast or another trial
- DKA/HHS
- Inborn errors of metabolism
- Insulinoma
- Pregnant/lactating
24,195 assessed for eligibility
- 9230 randomised
  - 4622 in liberal group
  - 4608 in tight glucose control group
Pre-specified subgroups
- History of diabetes
- Cardiac surgery
- Non-cardiac surgery/trauma
- Neurologic/neurosurgical diagnosis
- Sepsis
- Predicted ICU length of stay >1 week
Baseline characteristics well matched between liberal and tight group
- Median age 67 vs 67
- 63% male vs 64%
- 21% diabetes vs 20%
- Median APACHE II 21 vs 21 Comparing baseline characteristics of intervention vs. control group

Intervention

Tight glucose control
- Targeted BGL 4.4-6.1mmol/L (80-110mg/dL)
- Guided by LOGIC-Insulin algorithm
  - Computer based system used by nurses
  - Advises insulin dosage, dextrose boluses if needed and timing of blood sampling (1-4 hours)
  - Accounts for nutritional intake, use of drugs eg steroids and trends of BGL and insulin dose
  - Can be overridden at clinician discretion

Control

Liberal glucose control
- Insulin initiated if BGL >11.9mmol/L on 2 consecutive measurements (>215mg/dL) then targeted 10.0-11.9mmol/L (180-215mg/dL)
- Insulin infusion titrated by nursing staff without protocolised advice

Management common to both groups

All blood glucose measured via blood gas analyser until arterial line removed
All insulin administered as IV infusion via central line
Enteral nutrition as soon as possible
Parenteral nutrition initiated only after 1 week in ICU
Intervention ceased when
- Oral diet commenced
- CVC removed
- Discharge from ICU

Outcome

Primary outcome: ICU length of stay/time until readiness for ICU discharge
- No significant difference – hazard ratio for earlier discharge alive with tight glucose control, 1.00; (95% CI, 0.96 to 1.04; p=0.94)
Primary safety outcome: 90-day mortality
- No significant difference Liberal 10.1% vs tight 10.5% Relative Risk 1.04 (95% CI 0.92-1.17 p=0.51)
Separation between groups – intervention effect
- Patients in Tight control group
  - Received more insulin (98.8% of patients vs 45.9%)
  - Lower median
    - Peak BGL: 170 vs 189mg/dL
    - Morning BGL: 107 vs 140mg/dL
  - Higher rate of severe hypoglycaemia 1.0% vs 0.7% in liberal group (RR 1.53, 95% CI 0.97-2.39)
Secondary outcomes: all pre-specified
- Comparing liberal vs. tight glucose control groups
  - No significant difference in
    - Bilirubin, AST or ALT levels
    - New Infection
    - Time to wean from mechanical ventilation
    - Time to wean from haemodynamic support
    - Time to discharge alive from hospital
    - ICU or hospital mortality
  - Significantly less in tight control group
    - Severe acute kidney injury 8.6% vs 7.2% (RR 0.84, 95% CI 0.73-0.97)
    - New renal replacement therapy 5.8% vs 4.7% (RR 0.82, 95% CI 0.68-0.98)
    - Peak GGT and ALP levels
Subgroup analyses – all pre-specified as above
- No significant difference in time to discharge alive form ICU or death at 90 days for any pre-specified subgroup
- Non-statistically significant benefit for death at 90 days favouring tight glucose control in neuro/neurosurgical patients (OR 0.69 95% CI 0.46-1.02)

Authors’ Conclusions

In critically ill patients who were not receiving early parenteral nutrition, hyperglycemia was less severe than that previously reported in patients receiving parenteral nutrition. Further lowering of blood-glucose levels into the normal fasting range, guided by a computer algorithm, avoided iatrogenic hypoglycemia without affecting the length of time that ICU care was needed or mortality

Strengths

Large sample size, multicenter RCT
Heterogenous sample reflects real-word ICU population
LOGIC algorithm was effective in preventing high rates of severe hypoglycaemia seen in the tight glucose control groups in previous large RCTs
- This allows for examination of the effects of glucose targets without being confounded by adverse effects of hypoglyacemia

Weaknesses

Median daily glucose in Tight Glucose Control arm was higher than the target
Non-blinded trial although a blinded trial design would be far less feasible
Sample recruited from a single country and only 3 centers
No correction for multiplicity of secondary outcomes
- Findings of decreased renal failure and hepatic biomarker derangement in Tight Glucose Control group should be interpreted as hypothesis generating only

The Bottom Line

TGC-Fast demonstrates the effective use of a computerised algorithm to achieve tight glucose control in critically ill patients without significant hypoglycaemia as seen in previous studies
However, there is not a convincing signal of a clinical benefit to tight glucose control in these data and the TGC protocol would likely be associated with a higher clinical workload
In the absence of evidence of clinical benefit of tight glucose control, I will continue to target blood glucose levels of <10mmol/L (180mg/dL) in critically ill adults with strict avoidance of hypoglycaemia in line with current guideline recommendations

External Links

Metadata

Summary author: Daniel Chung
Summary date: 01/06/2025
Peer-review editor: David Slessor

Picture by: https://www.pexels.com/@punttim/