The Effect of High Protein Dosing in Critically Ill Patients: An Exploratory, Secondary Bayesian Analysis of the EFFORT Protein Trial

Academic Background

In the field of critical care medicine, nutritional support is a crucial component of treating critically ill patients. Protein, as a fundamental requirement for cellular function and survival, plays a vital role in patient recovery. However, the optimal protein intake for critically ill patients remains uncertain. While insufficient protein intake may lead to muscle atrophy and immune dysfunction, whether excessive protein intake has adverse effects on critically ill patients, especially those with organ failure, remains an unresolved question.

The EFFORT Protein Trial was a multicenter, randomized controlled trial designed to evaluate the impact of high protein intake (≥2.2 g/kg/day) versus standard protein intake (≤1.2 g/kg/day) on clinical outcomes in critically ill patients. The primary results of the trial showed that high protein intake did not significantly reduce 60-day all-cause mortality or shorten hospital stay. However, predefined subgroup analyses suggested that high protein intake might have adverse effects on patients with acute kidney injury (AKI) and those with higher organ failure scores.

To further explore the potential harms of high protein intake and assess heterogeneity of treatment effects (HTE), the research team conducted a secondary Bayesian analysis of the EFFORT Protein Trial. Bayesian analysis provides a more intuitive probabilistic interpretation, helping clinicians and researchers quantify uncertainties and refine intervention strategies for future clinical trials.

Source of the Paper

The study was conducted by Ryan W. Haines, Anders Granholm, Zudin Puthucheary, and other scholars from renowned institutions such as the Royal London Hospital, Copenhagen University Hospital, and Kingston Health Sciences Centre. The paper was published online on October 24, 2024, in the British Journal of Anaesthesia, titled The Effect of High Protein Dosing in Critically Ill Patients: An Exploratory, Secondary Bayesian Analysis of the EFFORT Protein Trial.

Study Design and Methods

Study Design

This study is a secondary Bayesian analysis of the EFFORT Protein Trial, aiming to reinterpret the trial data using Bayesian models. The study adhered to the STROBE statement (Strengthening the Reporting of Observational Studies in Epidemiology) and the ROBUST guidelines (Reporting of Bayes Used in Clinical Studies). The research team developed an internal protocol and statistical analysis plan after the results of the EFFORT Protein Trial were published.

The EFFORT Protein Trial was a multicenter, pragmatic, volunteer-driven, registry-based randomized controlled trial involving 1,301 critically ill adult patients requiring mechanical ventilation for at least 48 hours. Patients were randomized to either a high-protein group (≥2.2 g/kg/day) or a standard-protein group (≤1.2 g/kg/day), with the primary outcomes being 60-day all-cause mortality and time to discharge alive.

Statistical Analysis Methods

The study used R (version 4.1.2) and Stan (via the brms package) for Bayesian analysis. All analyses were based on the modified intention-to-treat population (excluding 28 patients who did not receive the intervention). Bayesian models combined prior probability distributions with trial data to estimate posterior probability distributions of treatment effects. Three prior distributions were used: a neutral skeptical prior, an optimistic prior (favoring the high-protein group), and a pessimistic prior (favoring the standard-protein group).

Primary and Secondary Outcomes

The primary outcome was 60-day all-cause mortality, and the secondary outcome was time to discharge alive within 60 days. The study used hierarchical Bayesian logistic regression models to analyze mortality and Bayesian Cox regression models to analyze time to discharge alive. The study also assessed heterogeneity of treatment effects (HTE), particularly the impact of disease severity (SOFA score), acute kidney injury (AKI), and baseline serum creatinine levels on mortality.

Results

Primary Outcome

In the analysis of 60-day all-cause mortality, Bayesian models using a neutral skeptical prior showed an absolute risk difference (RD) of 2.5% (95% credible interval: -6.9% to 12.4%) between the high-protein and standard-protein groups, with a relative risk ratio (RR) of 1.08 (95% credible interval: 0.82 to 1.44). There was a 72% probability of any harm (RD > 0%) and a 54% probability of clinically significant harm (RD ≥ 2%) in the high-protein group.

Secondary Outcome

In the analysis of time to discharge alive, the hazard ratio (HR) for the high-protein group was 0.91 (95% credible interval: 0.80 to 1.04), with a 92% probability of harm (HR < 1). Sensitivity analyses showed that the probability of harm in the high-protein group was 89% and 94% when using optimistic and pessimistic priors, respectively.

Heterogeneity of Treatment Effects (HTE)

The study evaluated the impact of different baseline characteristics on treatment effects. The results showed a 97% probability of a positive interaction between high protein intervention and baseline serum creatinine levels, indicating that patients with higher baseline serum creatinine levels may face a higher risk of mortality with high protein intervention. Additionally, there was a 95% probability of a positive interaction between high protein intervention and baseline SOFA scores, suggesting that patients with higher disease severity may experience worse outcomes with high protein intervention.

Conclusions

This Bayesian analysis indicates that high protein intake poses moderate to high probabilities of harm for critically ill patients, particularly those with baseline renal dysfunction or higher disease severity. The findings support the primary results of the EFFORT Protein Trial and further quantify the potential risks of high protein intake.

Highlights of the Study

1. Potential Harms of High Protein Intake: The study quantified the probability of harm associated with high protein intake in critically ill patients, providing important insights for clinical practice.
2. Heterogeneity of Treatment Effects: The study revealed that high protein intake may lead to worse outcomes in patients with renal dysfunction or higher disease severity, offering new directions for future clinical trial designs.
3. Application of Bayesian Analysis: The study demonstrated the advantages of Bayesian analysis in clinical trials, providing more intuitive probabilistic interpretations and quantification of uncertainties.

Significance and Value of the Study

This study provides critical clinical evidence for nutritional support in critically ill patients, particularly regarding the safety of high protein intake. The findings suggest that high protein intake may have adverse effects on certain patients, especially those with renal dysfunction or higher disease severity. This has important implications for clinical practice, indicating that clinicians should consider baseline patient characteristics when designing nutritional support strategies to avoid excessive protein intake.

The study highlights the value of Bayesian analysis in clinical trials, offering a more comprehensive probabilistic interpretation for clinical decision-making. Future research could further explore the optimal protein intake for different patient subgroups to refine nutritional support strategies for critically ill patients.