Skip to main content

Foundational Trials

Leuven 1

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Intensive insulin therapy in critically ill patients
  • Acronym: Leuven 1
  • Year: 2001
  • Journal published in: The New England Journal of Medicine
  • Citation: Van den Berghe G, Wouters P, Weekers F, Verwaest C, Bruyninckx F, Schetz M, et al. Intensive insulin therapy in critically ill patients. N Engl J Med. 2001;345(19):1359-67.

Context & Rationale

  • Background
    • Stress hyperglycaemia and insulin resistance are common in critical illness, including in patients without pre-existing diabetes.
    • Observational associations linked hyperglycaemia to infections, organ failure, neuromuscular complications, and mortality, but controlled trial evidence in ICU patients was lacking.
    • Surgical ICU populations frequently receive substantial early glucose/caloric exposure (often parenterally), potentially amplifying hyperglycaemia and its downstream effects.
  • Research Question/Hypothesis
    • Does normalising blood glucose (target 80–110 mg/dL) with continuous insulin infusion reduce ICU mortality and major ICU morbidities compared with a conventional, higher-threshold insulin strategy in mechanically ventilated adults admitted to a surgical ICU?
    • Hypothesis: hyperglycaemia and/or relative insulin deficiency causally contributes to severe infections, polyneuropathy, organ failure, and death during critical illness; preventing hyperglycaemia would reduce these outcomes.
  • Why This Matters
    • Unlike many organ-specific ICU interventions, glycaemic management is broadly applicable across diagnostic categories.
    • If a simple, scalable metabolic intervention could materially reduce sepsis-related multiple organ failure, renal failure, prolonged ventilation, and ICU mortality, it would reshape ICU care pathways and staffing/practice norms.

Design & Methods

  • Research Question: In mechanically ventilated adult surgical ICU patients, does intensive insulin therapy targeting blood glucose 80–110 mg/dL reduce death during ICU stay and major ICU morbidities compared with conventional insulin therapy initiated only for marked hyperglycaemia and targeting 180–200 mg/dL?
  • Study Type: Prospective, randomised, controlled, single-centre, parallel-group trial in a surgical intensive care unit (Leuven, Belgium); open-label delivery with blinded analysis of key outcomes.
  • Population:
    • Setting/timeframe: Surgical ICU (predominantly but not exclusively surgical), February 2, 2000 to January 18, 2001.
    • Inclusion: All adults receiving mechanical ventilation admitted to the ICU during the enrolment period, with written informed consent from the closest family member.
    • Exclusions (pre-randomisation): 14 total (5 participating in other trials; 9 moribund or with do-not-resuscitate orders).
    • Key baseline features: 13% history of diabetes; 75% had admission glucose >110 mg/dL; 12% had admission glucose >200 mg/dL; 59% of cardiac surgery admissions underwent CABG, 27% valve replacement, 14% combined.
  • Intervention:
    • Insulin strategy: Continuous insulin infusion started if glucose >110 mg/dL, titrated to maintain 80–110 mg/dL (4.4–6.1 mmol/L); maximum dose set at 50 IU/hour.
    • Monitoring/delivery: Whole-blood glucose from undiluted arterial samples measured every 1–4 hours using a glucose analyser; dose titration by ICU nurses following a strict algorithm, assisted by a study physician not involved in clinical care decisions.
    • Nutrition co-intervention (standardised for both groups): Intravenous glucose 200–300 g/24 hours on admission; from day 2, standardised schedule of total parenteral, combined parenteral/enteral, or total enteral feeding (20–30 non-protein kcal/kg/24 hours; 0.13–0.26 g nitrogen/kg/24 hours; 20–40% of non-protein calories as lipids), with enteral feeding attempted as early as possible.
    • Post-ICU: On ICU discharge, a conventional glucose approach (180–200 mg/dL) was adopted.
  • Comparison:
    • Insulin strategy: Continuous insulin infusion started only if glucose >215 mg/dL (11.9 mmol/L), titrated to maintain 180–200 mg/dL (10.0–11.1 mmol/L).
    • Monitoring and nutrition: Same sampling frequency framework and the same standardised nutritional schedule as the intervention arm.
  • Blinding: Not blinded (insulin titration requires frequent glucose monitoring); insulin dosing handled by a dedicated nurse team with a study physician not participating in clinical decisions; key outcome measures underwent blinded analysis.
  • Statistics: Planned enrolment 2500 to detect a 5% absolute mortality difference among patients staying >5 days in ICU and a 2% difference among all ICU patients, with two-sided alpha <0.05; beta/power not reported. Interim analyses of ICU mortality were performed every 3 months with stopping boundaries (two-sided alpha <0.01); sequential adjustment for the primary outcome used the Lan–DeMets approach; analyses were intention-to-treat.
  • Follow-Up Period: Through ICU stay and hospital stay (patients discharged alive from hospital considered survivors); survival curves presented out to ~160 days (ICU survival) and ~250 days (in-hospital survival).

Key Results

This trial was stopped early. It was stopped after the fourth pre-specified interim analysis (3-monthly) at 12 months (1548 patients enrolled) because conventional treatment was inferior for the primary endpoint (death during intensive care).

Outcome Intensive insulin (80–110 mg/dL) Conventional insulin (start >215; target 180–200 mg/dL) Effect p value / 95% CI Notes
Death during ICU (primary) 35/765 (4.6%) 63/783 (8.0%) Apparent risk reduction 42% Adjusted reduction 32%; 95% CI 2 to 55; P<0.04 (sequentially adjusted); nominal P=0.005 Primary endpoint; interim monitoring with stopping boundary; sequential adjustment applied only to this endpoint.
In-hospital death 55/765 (7.2%) 85/783 (10.9%) Relative reduction 34% P=0.01; 95% CI not reported Nominal (unadjusted) P value reported for morbidity/non-primary outcomes.
Death during ICU, ICU stay >5 days 22/208 (10.6%) 49/243 (20.2%) Not reported P=0.005; 95% CI not reported Deaths during first 5 ICU days were similar between groups (13/557 vs 14/540).
Bloodstream infection (septicemia) 33/765 (4.2%) 61/783 (7.8%) Relative reduction 46% 95% CI 25 to 67; P=0.003 Nominal P values (no sequential adjustment for morbidity endpoints).
Renal replacement therapy (dialysis or haemofiltration) 37/765 (4.8%) 65/783 (8.3%) Relative reduction 41% P=0.007; 95% CI not reported Peak creatinine and urea nitrogen were also lower in the intensive group (medians reported in Table 4).
Critical-illness polyneuropathy (EMG evidence; screened subgroup) 56/195 (28.7%) 106/206 (51.4%) Relative reduction 44% P<0.001; 95% CI not reported Fewer patients were screened in the intensive group (20.5% vs 26.3%) because ICU stay was shorter among long-stay patients.
Prolonged mechanical ventilation (>14 days) 57/765 (7.5%) 93/783 (11.9%) Not reported P=0.003; 95% CI not reported Median ventilation duration in ICU stay >5 days: 10 (5–18) vs 12 (5–22) days (P=0.006).
Hypoglycaemia (glucose ≤40 mg/dL) 39 patients 6 patients Not reported P value / 95% CI not reported Two symptomatic events (sweating/agitation) in intensive group; no haemodynamic deterioration or convulsions reported.
ICU length of stay, ICU stay >5 days (median, IQR) 12 (8–20) days 15 (9–27) days Not reported P=0.003 All-patient median ICU duration: 3 (2–6) vs 3 (2–9) days (P=0.2).
Cumulative ICU resource use (TISS-28), ICU stay >5 days (median, IQR) 431 (271–670) 563 (329–956) Not reported P<0.001 TISS-28 on last ICU day median 30 in both groups (reported in text).
  • Clear biochemical separation was achieved: morning blood glucose 103±19 mg/dL (intensive) vs 153±33 mg/dL (conventional), with insulin administered to 98.7% vs 39.2% of patients.
  • The mortality signal was concentrated among patients requiring prolonged ICU care (>5 days), with similar early ICU deaths in both groups.
  • Major morbidity reductions were reported alongside a substantially higher frequency of hypoglycaemia (39 vs 6 patients).

Internal Validity

  • Randomisation and Allocation:
    • Randomisation at ICU admission; sealed envelopes; permuted blocks of 20; stratified by diagnostic category; balancing for sex and history of diabetes.
    • Very limited pre-randomisation exclusions (14 patients total), reducing risk of major selection filtering beyond the ventilated surgical ICU population.
  • Drop out or exclusions (post-randomisation):
    • Not reported for the primary mortality outcomes (ICU and in-hospital mortality are presented on an intention-to-treat basis for all randomised patients).
    • Some morbidity assessments were necessarily conditional on survival and duration of ICU stay (e.g., electromyography screening for polyneuropathy among longer-stay survivors).
  • Performance/Detection Bias:
    • Intervention delivery was necessarily unblinded due to frequent glucose measurement and insulin titration.
    • Insulin adjustment was separated from clinical decision-making via a nurse-led protocol and a study physician not involved in clinical care, with blinded analysis of key outcomes reported.
  • Protocol Adherence:
    • High adherence to insulin delivery in the intensive arm (755/765; 98.7%) versus conventional (307/783; 39.2%).
    • Insulin exposure differed materially: median insulin dose on insulin-days 71 (IQR 48–100) vs 33 (17–56) IU/day; insulin administered for 100% vs 67% (IQR 40–100) of ICU stay.
  • Baseline Characteristics:
    • Clinical and demographic characteristics were reported as similar at randomisation, including diabetes prevalence and admission hyperglycaemia rates.
    • Severity scoring considerations: APACHE II scores at ICU admission could be artificially lowered for patients admitted after resuscitation outside ICU and for sedated patients.
  • Heterogeneity:
    • Single-centre surgical ICU with a large proportion of post-cardiac surgery patients; subgroup analyses by severity (APACHE II, TISS-28) were presented, with reported consistency across cardiac vs non-cardiac surgery subgroups.
  • Timing:
    • Insulin protocol initiated at ICU admission; glucose monitored every 1–4 hours using arterial samples.
  • Dose:
    • Intensive arm targeted 80–110 mg/dL with a maximum insulin infusion rate capped at 50 IU/hour.
    • Conventional arm initiated insulin only for glucose >215 mg/dL and targeted 180–200 mg/dL.
  • Separation of the Variable of Interest:
    • Morning blood glucose: 103±19 mg/dL (intensive) vs 153±33 mg/dL (conventional), P<0.001.
    • Among conventional-group patients receiving insulin, mean glucose was 173±33 mg/dL versus 103±18 mg/dL in the intensive group, P<0.001.
    • Hypoglycaemia ≤40 mg/dL: 39 vs 6 patients (clinical sequelae not reported beyond two symptomatic episodes).
  • Key Delivery Aspects:
    • Standardised, protocolised nutrition was implemented, including early intravenous glucose 200–300 g/24 hours, then scheduled parenteral/enteral feeding.
    • Non-protein calorie intake was similar between groups: 19.1±7.1 vs 18.5±7.5 kcal/kg/24 hours.
  • Outcome Assessment:
    • Primary outcome (death during ICU stay) is objective; secondary outcomes included culture-defined septicemia, renal replacement therapy, transfusions, and electrophysiologically defined polyneuropathy in screened patients.
    • Sequential adjustment was applied to the primary endpoint; nominal P values were reported for morbidity endpoints.
  • Statistical Rigor:
    • Intention-to-treat analysis stated; sequential interim monitoring applied to the primary endpoint with Lan–DeMets adjustment.
    • Multivariable logistic regression and Kaplan–Meier/log-rank analyses were reported for mortality/time-to-event assessments.

Conclusion on Internal Validity: Overall, internal validity appears moderate: randomisation, substantial separation of glycaemic exposure, objective mortality endpoints, and intention-to-treat analyses support causal inference, but open-label delivery and early stopping complicate confidence in the precision of the effect size and the interpretation of multiple secondary endpoints.

External Validity

  • Population Representativeness:
    • Predominantly surgical ICU population with a large post-cardiac surgery component; all participants were mechanically ventilated adults.
    • Admission hyperglycaemia was common but marked diabetic-range hyperglycaemia at admission was relatively infrequent (12% >200 mg/dL).
  • Applicability:
    • Protocol required 1–4 hourly arterial glucose measurements and nurse-driven insulin titration, implying substantial monitoring capability and staffing infrastructure.
    • Standardised nutrition included early high intravenous glucose exposure (200–300 g/24 hours), which may differ from modern nutritional strategies in other settings.

Conclusion on External Validity: Generalisability appears limited-to-moderate, most directly applicable to well-resourced surgical ICUs with similar nutritional practices and intensive monitoring capacity; extrapolation to medical ICUs, resource-limited environments, or different feeding paradigms is uncertain.

Strengths & Limitations

  • Strengths:
    • Large randomised trial for its era in ICU metabolic management, with broad inclusion within the ventilated surgical ICU population.
    • Marked, quantifiable separation in glycaemic exposure and insulin delivery between groups (morning glucose 103±19 vs 153±33 mg/dL; insulin use 98.7% vs 39.2%).
    • Clinically meaningful endpoints including ICU and in-hospital mortality, renal replacement therapy, septicemia, ventilation duration, and electrophysiologically defined polyneuropathy.
    • Pre-specified interim monitoring framework and stated intention-to-treat analysis.
  • Limitations:
    • Single-centre design in a predominantly post-operative surgical ICU population.
    • Unblinded delivery (inherent to insulin titration); potential for care-process differences despite mitigation strategies.
    • Early stopping after interim analysis, and nominal (unadjusted) P values for morbidity endpoints.
    • Secondary outcomes such as polyneuropathy relied on screening within a subset, with differential screening proportions between groups (20.5% vs 26.3%).
    • Substantial increase in hypoglycaemia events (39 vs 6 patients), an important safety consideration for translation.

Interpretation & Why It Matters

  • Clinical signal
    • In this surgical ICU cohort, intensive insulin therapy was associated with lower ICU mortality (4.6% vs 8.0%) and lower in-hospital mortality (7.2% vs 10.9%).
    • The mortality difference was largely attributable to outcomes among longer-stay patients (>5 ICU days: 10.6% vs 20.2%).
  • Breadth of morbidity effect
    • Lower rates of septicemia (4.2% vs 7.8%), renal replacement therapy (4.8% vs 8.3%), prolonged ventilation (>14 days: 7.5% vs 11.9%), and polyneuropathy in screened survivors (28.7% vs 51.4%).
    • Reduced ICU resource utilisation among longer-stay patients (cumulative TISS-28 median 431 vs 563) consistent with fewer/shorter high-intensity interventions.
  • Safety/feasibility
    • Hypoglycaemia was substantially more frequent (39 vs 6 patients), highlighting the tension between tight targets and safety, particularly outside highly monitored environments.

Controversies & Subsequent Evidence

  • Early stopping and confirmation needs:
    • The study was stopped early after an interim analysis identified inferiority of conventional therapy, a design feature that can amplify apparent treatment effects relative to fully accrued trials.
    • The accompanying editorial positioned the result as potentially practice-changing but emphasised the need for confirmation in other ICU populations and settings.1
  • Reproducibility across ICU populations:
    • A subsequent Leuven trial in a medical ICU population did not demonstrate the same uniform mortality benefit, with treatment effects varying by ICU length of stay.2
    • The large multicentre NICE-SUGAR trial comparing similar tight targets to more liberal control found worse survival and substantially more hypoglycaemia with intensive control, shifting the centre of gravity away from 80–110 mg/dL targets in general ICU care.3
  • Safety signals and trial discontinuation in sepsis and mixed ICUs:
    • In severe sepsis, VISEP reported major hypoglycaemia concerns and did not establish a survival advantage for intensive insulin therapy.4
    • The multicentre GLUCONTROL trial did not confirm a clear mortality benefit and contributed additional data on feasibility and hypoglycaemia risk in broader ICU settings.5
  • Evidence synthesis:
    • Meta-analytic evidence integrating early ICU trials found no consistent mortality benefit with tight glucose control and highlighted increased hypoglycaemia risk, supporting more moderate targets in routine practice.6
  • Guideline convergence (current practice):
    • Contemporary critical care guidelines generally recommend initiating insulin for persistent hyperglycaemia around 10 mmol/L (180 mg/dL) and targeting a moderate range rather than strict normoglycaemia, reflecting subsequent multicentre evidence and safety considerations.789

Summary

  • Single-centre surgical ICU RCT (Leuven 1) testing intensive insulin therapy to maintain glucose 80–110 mg/dL versus conventional insulin initiated only for glucose >215 mg/dL with targets 180–200 mg/dL.
  • Stopped early at 12 months after the fourth interim analysis (1548 patients) due to inferiority of conventional treatment for ICU mortality.
  • Lower ICU mortality (4.6% vs 8.0%; sequentially adjusted P<0.04) and lower in-hospital mortality (7.2% vs 10.9%; P=0.01), with mortality benefit concentrated in patients staying >5 ICU days.
  • Lower septicemia, renal replacement therapy, prolonged ventilation, transfusion burden, and polyneuropathy in screened longer-stay survivors, alongside reduced ICU resource use in long-stay patients.
  • Hypoglycaemia was substantially more frequent with intensive insulin therapy (39 vs 6 patients), underscoring a key safety trade-off.

Notes

  • Unit conversion: mg/dL to mmol/L for glucose uses a multiplier of 0.05551; the trial’s intensive target (80–110 mg/dL) corresponds to ~4.4–6.1 mmol/L.

Overall Takeaway

Leuven 1 was a landmark ICU trial because it provided early randomised evidence that targeting near-normoglycaemia with intensive insulin therapy could reduce ICU mortality and several major morbidities in a surgical ICU cohort. Its striking signal catalysed widespread adoption of tight glycaemic control, but subsequent multicentre trials and evidence synthesis shifted practice towards safer, more moderate glycaemic targets while preserving the principle that severe hyperglycaemia should be actively treated.

Overall Summary

  • Early-stopped single-centre surgical ICU RCT showing lower ICU mortality with glucose 80–110 mg/dL versus a higher-threshold conventional strategy.
  • Benefits concentrated in longer-stay patients and accompanied by reduced infections, renal replacement therapy, and neuromuscular complications.
  • Hypoglycaemia was substantially more frequent with intensive insulin, and later multicentre evidence shifted guidelines towards moderate glycaemic targets.

Bibliography