Skip to main content

Foundational Trials

EFFORT Protein

Image: Shutterstock / luchschenF

Publication

  • Title: The effect of higher protein dosing in critically ill patients with high nutritional risk (EFFORT Protein): an international, multicentre, pragmatic, registry-based randomised trial
  • Acronym: EFFORT Protein
  • Year: 2023
  • Journal published in: The Lancet
  • Citation: Heyland DK, Patel J, Compher C, Rice TW, Bear DE, Lee ZY, et al. The effect of higher protein dosing in critically ill patients with high nutritional risk (EFFORT Protein): an international, multicentre, pragmatic, registry-based randomised trial. Lancet. 2023;401:568-76.

Context & Rationale

  • Background
    • Critical illness is characterised by rapid skeletal muscle loss and negative nitrogen balance; protein provision is a plausible, modifiable determinant of recovery.
    • Protein-dose recommendations in ICU nutrition guidelines historically spanned a broad range (commonly ~1.2–2.0 g/kg/day), reflecting uncertainty and generating major between-unit practice variation.
    • Pre-trial randomised evidence was limited by small sample sizes, heterogeneity of populations/interventions, and inconsistent separation in delivered protein.
    • Mechanistic equipoise existed: higher protein might attenuate catabolism, but could also increase nitrogenous waste and be harmful in some phenotypes (eg acute kidney injury).
  • Research Question/Hypothesis
    • In mechanically ventilated ICU patients at high nutritional risk, does prescribing higher protein (≥2.2 g/kg/day) versus usual protein dosing (<1.2 g/kg/day) improve clinically important outcomes?
    • Hypothesis: higher protein dosing would improve survival and/or accelerate recovery (shorter time-to-discharge-alive from hospital).
  • Why This Matters
    • Protein targets drive ICU nutrition prescriptions (formula selection, modular protein use, parenteral amino acids), with system-level consequences for workload, costs, and monitoring.
    • A clear benefit would support aggressive early protein strategies; neutrality or harm would justify more conservative and individualised dosing, especially in organ dysfunction phenotypes.

Design & Methods

  • Research Question: Among mechanically ventilated ICU patients at high nutritional risk, does a higher protein prescription (≥2.2 g/kg/day) compared with usual protein prescription (<1.2 g/kg/day) improve patient-centred outcomes (time-to-discharge-alive and mortality)?
  • Study Type: International, multicentre, pragmatic, registry-based, single-blinded (patient) parallel-group randomised trial in adult ICUs; protocol published before trial completion1
  • Population:
    • Setting: 85 adult ICUs across 15 countries (international, multicentre).
    • Key inclusion criteria: adults ≥18 years; enrolled within 96 h of ICU admission; expected to remain invasively mechanically ventilated ≥48 h from screening; ≥1 high nutritional risk feature (BMI ≤25 kg/m2, BMI ≥35 kg/m2, Clinical Frailty Scale ≥5, SARC-F ≥4, malnourished by local assessment, or projected duration of mechanical ventilation >4 days).
    • Key exclusion criteria: invasively mechanically ventilated >96 h before screening; expected death or withdrawal of life-sustaining treatments within 7 days; pregnancy; absence of clinician equipoise regarding protein dose; requirement for parenteral nutrition only when site lacked products to meet the high-protein target.
  • Intervention:
    • Protein prescription target: ≥2.2 g/kg/day.
    • Dosing weight: pre-ICU actual “dry” weight; if BMI >30 kg/m2, ideal bodyweight based on BMI 25 kg/m2.
    • Delivery: any combination of enteral nutrition, parenteral nutrition, intravenous amino acids, and/or enteral protein supplements; clinicians encouraged to achieve ≥80% of prescribed protein.
    • Duration: continued up to 28 days post-randomisation, ICU discharge, death, or full oral intake; if readmitted within 28 days, study strategy resumed.
  • Comparison:
    • Protein prescription target: <1.2 g/kg/day (lower end of guideline-based usual care range).
    • Energy and co-interventions: energy targets and other ICU care were not protocolised beyond clinician encouragement to avoid overfeeding; route and supportive care at clinician discretion.
  • Blinding: Single-blind (patients not informed of allocation); clinicians not blinded.
  • Statistics: A total of 4000 patients were planned to detect a 4% absolute reduction in 60-day mortality (from 30% to 26%) with 80% power at the 5% (two-sided) significance level; COVID-19 feasibility constraints led to curtailed recruitment (~1200 planned) and switching of the primary outcome to time-to-discharge-alive while allocation remained blinded2. Primary analyses used a modified intention-to-treat approach (randomised participants who received study intervention and had useable data) with competing-risk methods (Fine–Gray) for time-to-discharge-alive.
  • Follow-Up Period: Primary outcome assessed up to 60 days after ICU admission; nutrition intervention delivered up to 28 days (or until ICU discharge/death/full oral intake).

Key Results

This trial was stopped early. Recruitment was curtailed (planned n=4000; randomised n=1329) due to COVID-19 feasibility constraints; no interim analyses were undertaken, and the primary outcome was switched from 60-day mortality to time-to-discharge-alive while allocation remained blinded2.

Outcome High-dose protein Usual-dose protein Effect p value / 95% CI Notes
Time-to-discharge-alive from hospital (day 60 cumulative incidence) 46.1% (95% CI 42.0 to 50.1) 50.2% (95% CI 46.0 to 54.3) sHR 0.91 95% CI 0.77 to 1.07; P=0.27 Primary outcome; Fine–Gray competing-risk model (death competing event).
60-day mortality 222/642 (34.6%) 208/648 (32.1%) RR 1.08 95% CI 0.92 to 1.26; P=Not reported Secondary outcome; denominators differ from randomised due to withdrawal/loss to follow-up.
ICU length of stay (days) 10.0 (5.6 to 18.2) 9.4 (5.1 to 19.2) Not reported Not reported Median (Q1 to Q3)2.
Duration of mechanical ventilation (days) 6.1 (3.0 to 13.8) 6.1 (2.8 to 12.8) Not reported Not reported Median (Q1 to Q3)2.
Highest urea (mmol/L) 14.0 ± 8.5 11.9 ± 7.2 Not reported P<0.0001 Metabolic tolerance (within-patient averages); mean ± SD2.
Urea ≥30 mmol/L 43/645 (6.7%) 20/656 (3.0%) Not reported P=0.002 Metabolic tolerance; crude proportions2.
  • Despite a substantially higher protein prescription (2.2 ± 0.1 vs 1.1 ± 0.2 g/kg/day), achieved intake averaged 1.6 ± 0.5 vs 0.9 ± 0.3 g/kg/day, with broadly similar energy delivery (14.7 ± 5.5 vs 13.2 ± 4.3 kcal/kg/day)2.
  • Higher protein dosing did not improve time-to-discharge-alive (46.1% vs 50.2%; sHR 0.91; 95% CI 0.77 to 1.07; P=0.27) and did not reduce 60-day mortality (RR 1.08; 95% CI 0.92 to 1.26).
  • Exploratory subgroup analyses suggested effect modification in acute kidney injury: time-to-discharge-alive sHR 0.5 (95% CI 0.4 to 0.8; interaction P=0.001) and 60-day mortality RR 1.4 (95% CI 1.1 to 1.8; interaction P=0.02); these findings were not adjusted for multiplicity and should be considered hypothesis-generating2.

Internal Validity

  • Randomisation and allocation: Central web-based randomisation with permuted blocks (2, 4, 8) and stratification by site; allocation concealed until assignment.
  • Dropout/exclusions and missingness: 1329 randomised; 28 excluded from all analyses post-randomisation (14 per arm) due to leaving before receiving intervention, withdrawal of consent for data use, or inability to obtain useable data; primary time-to-event analysis included 1297 participants due to 4 missing hospital outcomes; 60-day mortality denominators were 642 and 648, reflecting additional withdrawals/loss to follow-up.
  • Performance/detection bias: Clinicians were unblinded; delivery of nutrition and some downstream decisions (eg discharge planning) could be influenced by knowledge of allocation; key outcomes were largely objective, but discharge timing is susceptible to system-level and clinician-level influences.
  • Protocol adherence:
    • Protein prescribed: 2.2 ± 0.1 vs 1.1 ± 0.2 g/kg/day.
    • Protein delivered: 1.6 ± 0.5 vs 0.9 ± 0.3 g/kg/day (median 1.7 [IQR 1.3 to 2.0; range 0.1 to 3.2] vs 1.0 [0.8 to 1.1; 0.1 to 1.9]).
    • Protein adequacy (delivered/prescribed): 72.6% ± 24.0 vs 75.4% ± 25.1; ≥80% achieved in 45.0% vs 47.1%.
    • Energy delivered: 14.7 ± 5.5 vs 13.2 ± 4.3 kcal/kg/day; energy adequacy 64.1% ± 21.9 vs 60.7% ± 20.9 (≥80% achieved in 16.8% vs 11.3%).
    • Route of nutrition was similar (enteral-only 92.6% vs 91.3%; parenteral-only 0.5% vs 2.0%; combined enteral+parenteral 7.0% vs 6.7%).
    2
  • Timing:
    • Time from ICU admission to randomisation: 38.5 ± 26.4 vs 39.8 ± 27.3 h.
    • Time from ICU admission to start of enteral nutrition: 16.0 ± 18.2 vs 16.1 ± 18.1 h.
    2
  • Baseline characteristics: Groups were broadly comparable at enrolment, including illness severity and organ support (eg APACHE II 21 [16–27] vs 21 [15–26]; SOFA 9 [6–11] vs 9 [6–11]; vasopressors 39% vs 41%; baseline acute kidney injury 25% vs 23%).
  • Heterogeneity: Pragmatic delivery across 85 ICUs and varied diagnoses improves ecological validity but increases clinical and practice heterogeneity; site stratification mitigates centre effects but does not eliminate effect dilution in a biologically diverse syndrome.
  • Statistical rigour: Competing-risk methods were appropriate for the discharge-alive endpoint; early termination and the mid-trial primary outcome switch reduced interpretability of mortality effects and placed greater weight on assumptions underpinning the revised power strategy.

Conclusion on Internal Validity: Moderate — randomisation and balanced baseline risk support internal validity, but open-label delivery, post-randomisation exclusions, incomplete achievement of the high-dose target, and early outcome switching/curtailed recruitment constrain confidence (particularly for subgroup inferences).

External Validity

  • Population representativeness: Enrolled mechanically ventilated adults at “high nutritional risk” across diverse countries and ICU types, increasing generalisability to similar ventilated ICU cohorts.
  • Key exclusions and selection: Exclusion of those without clinician equipoise and those expected to die/withdraw life-sustaining treatment within 7 days limits applicability to the most imminently dying; consent processes (non-approach and non-consent among otherwise eligible patients) may introduce selection effects.
  • Intervention feasibility: Findings are most applicable to ICUs with established dietetic input and access to high-protein enteral feeds, modular protein, and/or parenteral amino acids; settings without these resources may not achieve similar separation.
  • Clinical scope: Applicability to non-ventilated ICU patients, lower nutritional risk patients, and prolonged critical illness beyond the 28-day intervention window is uncertain.

Conclusion on External Validity: Generalisability is good for mechanically ventilated, nutritionally high-risk ICU patients in resource-equipped ICUs internationally, but is more limited for low-risk ICU populations, settings without advanced nutrition delivery options, and patients with imminent end-of-life trajectories.

Strengths & Limitations

  • Strengths:
    • Large, international, pragmatic, registry-based randomised trial across 85 ICUs.
    • Enrichment for high nutritional risk patients most likely to be protein-responsive.
    • Meaningful achieved separation in protein delivery (1.6 ± 0.5 vs 0.9 ± 0.3 g/kg/day) with broadly similar energy delivery.
    • Appropriate competing-risk methods for the discharge-alive endpoint.
  • Limitations:
    • Recruitment curtailed with a mid-trial primary outcome switch, reducing power for mortality and complicating interpretation2.
    • Open-label delivery with potential performance bias and discharge-practice influence.
    • Modified intention-to-treat analysis with post-randomisation exclusions and small amounts of missing follow-up data.
    • High-dose target (≥2.2 g/kg/day) was not achieved on average (mean delivered 1.6 g/kg/day), increasing risk of a “dose dilution” bias towards null for benefit (and possibly for harm).
    • No pre-specified functional recovery endpoints (eg strength, muscle mass, long-term quality of life), despite mechanistic rationale focused on muscle preservation.

Interpretation & Why It Matters

  • Practice signal
    In nutritionally high-risk ventilated ICU patients, a strategy of very high protein prescription (≥2.2 g/kg/day; achieved ~1.6 g/kg/day) did not improve discharge-alive or 60-day mortality compared with usual protein dosing.
  • Safety and phenotype
    Higher protein dosing increased azotaemia (higher urea; more urea ≥30 mmol/L) and showed an exploratory signal of harm in acute kidney injury, supporting a cautious, phenotype-aware approach rather than indiscriminate escalation.
  • Implementation reality
    Even in motivated trial settings, achieving a sustained delivered intake of ≥2.2 g/kg/day was challenging; practice strategies should account for feasibility, monitoring burden, and diminishing marginal returns at higher doses.

Controversies & Other Evidence

  • Early stopping and outcome switching:
    • Stopping early (planned n=4000; randomised n=1329) and switching the primary endpoint to time-to-discharge-alive increases susceptibility to outcome switching concerns and limits precision for mortality effects, even when changes are made while blinded2.
    • Time-to-discharge-alive is clinically relevant but partially health-system-dependent (bed capacity, step-down availability, discharge practices), and might be less directly mechanistic than functional recovery outcomes.
  • “Dose dilution” and overlap between groups:
    • The achieved difference (1.6 vs 0.9 g/kg/day) was meaningful but below the prescribed high-dose target (≥2.2 g/kg/day), with overlap at the tails (control maximum 1.9 g/kg/day; intervention minimum 0.1 g/kg/day), which can attenuate true effects2.
  • Renal phenotype signal (acute kidney injury):
    • Exploratory subgroup analyses suggested possible harm in acute kidney injury (time-to-discharge-alive sHR 0.5; 95% CI 0.4 to 0.8; interaction P=0.001; 60-day mortality RR 1.4; 95% CI 1.1 to 1.8; interaction P=0.02)2.
    • A subsequent post hoc analysis further examined kidney injury and suggested differential effects according to renal replacement therapy exposure, reinforcing caution with high protein in acute kidney injury phenotypes5.
    • Increased urea and urea ≥30 mmol/L provide biological coherence for renal solute burden as a potential mediator or marker of intolerance2.
  • Editorial and correspondence critique:
    • Contemporary commentary emphasised that EFFORT Protein challenges the assumption that “more protein is always better” early in critical illness and highlighted the need for stage-based and individualised protein strategies (including caution in organ failure phenotypes)3.
    • Correspondence reinforced uncertainty regarding optimal timing and patient selection and supported a cautious approach to very high early protein, particularly in acute kidney injury4.
  • Subsequent trials, syntheses, and guidelines:
    • PRECISe (protocolised, blinded trial in mechanically ventilated ICU patients) reported no functional recovery advantage and signalled worse health-related quality of life and longer time-to-discharge-alive with higher protein provision; it did not reproduce the EFFORT Protein harm signal in acute kidney injury or high SOFA subgroups6.
    • An updated systematic review/meta-analysis incorporating newer trials reported no overall short-term mortality benefit (RR 0.99; 95% CI 0.90 to 1.08) and identified a signal of increased mortality with higher protein among patients with acute kidney injury (RR 1.42; 95% CI 1.11 to 1.82), alongside higher urea (mean difference 2.31 mmol/L; 95% CI 1.85 to 2.77)7.
    • Recent formalised expert recommendations propose lower protein (0.2–0.9 g/kg/day) during the first ICU week rather than “standard” 1.0–1.3 g/kg/day, with escalation back to 1.0–1.3 g/kg/day by the end of week 1, reflecting incorporation of contemporary high-protein trial evidence8.

Summary

  • International pragmatic registry-based RCT in mechanically ventilated ICU patients with high nutritional risk (1329 randomised; 1301 analysed).
  • Stopped early (planned n=4000) with a blinded protocol amendment switching the primary outcome to time-to-discharge-alive.
  • Achieved protein delivery was higher with the intervention (1.6 ± 0.5 vs 0.9 ± 0.3 g/kg/day) with broadly similar energy delivery.
  • No improvement in time-to-discharge-alive (sHR 0.91; 95% CI 0.77 to 1.07; P=0.27) and no reduction in 60-day mortality (RR 1.08; 95% CI 0.92 to 1.26).
  • Higher protein increased azotaemia and showed an exploratory harm signal in acute kidney injury; subsequent trials, meta-analyses, and guideline recommendations support caution with very high early protein and favour progressive individualisation.

Overall Takeaway

EFFORT Protein is a landmark pragmatic RCT because it directly tested a widely promoted “high protein early” strategy at scale in nutritionally high-risk ventilated ICU patients and found no overall clinical benefit. Its most practice-shaping contribution is the combination of neutrality for core outcomes with a coherent signal of intolerance/possible harm in acute kidney injury phenotypes, pushing contemporary practice towards moderate, progressive, and individualised protein dosing rather than universal escalation.

Overall Summary

  • In high nutritional risk ventilated ICU patients, prescribing very high protein (achieved ~1.6 g/kg/day) did not improve discharge-alive or mortality, increased urea burden, and showed an exploratory harm signal in acute kidney injury.

Bibliography