What a negative trial does not tell us: Bayesian reanalysis of the enteral citrulline supplementation versus placebo on SOFA score on day 7 in mechanically ventilated critically ill patients randomized clinical trial.

BACKGROUND

The enteral citrulline supplementation versus placebo on SOFA score on day 7 in mechanically ventilated critically ill patients (IMMUNOCITRE) randomized controlled trial, evaluating the immune effect of restoring arginine deficiency through enteral administration of citrulline in critically ill patients, found no statistically significant difference in Day-7 SOFA score between enteral l-citrulline and placebo in mechanically ventilated ICU patients (p = 0.9). The trial was likely underpowered due to lower-than-expected baseline SOFA scores. We conducted a post hoc Bayesian reanalysis to quantify the posterior probability of benefit and characterize the residual uncertainty beyond the binary frequentist result.

METHODS

We performed a Bayesian reanalysis of all patients from the IMMUNOCITRE trial that included 120 mechanically ventilated ICU patients without sepsis or septic shock from four French intensive care units (ICU). Patients received either enteral l-citrulline (5 g twice daily for 5 days) or isonitrogenous, isocaloric placebo. Three prior distributions were specified to reflect optimistic, neutral, and pessimistic pre-existing beliefs about treatment benefit. The primary outcome (Day-7 SOFA score) was modelled as an ordinal variable using a Bayesian cumulative proportional-odds model adjusted for baseline Day-1 SOFA score. To account for competing terminal events without introducing survivor bias, patients who died before Day 7 were assigned a penalized score of 25. ICU mortality and ≥2-point SOFA improvement were pre-specified secondary endpoints, modelled by baseline-adjusted Bayesian logistic regression under the neutral prior only.

RESULTS

Under the neutral prior, the posterior median odds-ratio (OR) for Day-7 SOFA was 1.01 (95% credible intervals (95% CrI) 0.63-1.64), with a posterior probability of any benefit of 48.6%. Under the optimistic prior, this probability reached 73.6% (median OR 0.85, 95% CrI 0.52-1.39). The probability of a clinically meaningful effect (OR < 0.9) was 32.5% and the probability of a substantial effect (OR < 0.8) was 17.2% under the neutral prior. For ICU mortality, the posterior probability of benefit was 41.1% under the neutral prior (OR 1.07, 95% CrI 0.61-1.86). Credible intervals were wide across all analyses, reflecting the limited statistical power of the original trial. Sensitivity analyses yielded directionally consistent results.

CONCLUSION

This Bayesian reanalysis demonstrates that the IMMUNOCITRE trial was insufficiently powered to characterize the effect of l-citrulline on organ dysfunction. In this unselected cohort with low-to-moderate initial severity, the probability of achieving a large, clinically meaningful benefit remains limited. These findings do not support immediate clinical implementation. Future multicenter randomized controlled trial is highly warranted but should specifically target patients with higher baseline organ dysfunction and documented arginine deficiency.

TRIAL REGISTRATION: This analysis was pre-registered on the Open Science Framework (https://doi.org/10.17605/osf.io/ytgba). The IMMUNOCITRE trial was registered on ClinicalTrials.gov Identifier NCT02864017 (date of registration: 11 August 2016).