The Effect of Liberal versus Restrictive Haemoglobin Thresholds on Cerebral Tissue Oxygenation in Preterm Infants: A Systematic Review and Meta-Analysis.

INTRODUCTION

Near-infrared spectroscopy (NIRS) is emerging as a promising method of assessing individualized cerebral oxygenation responses to red blood cell (RBC) transfusion. However, the impact of liberal (high) versus restrictive (low) haemoglobin thresholds on transfusion-related changes in cerebral tissue oxygenation (crSO2) remains unclear. This meta-analysis compares cerebral oxygenation changes in preterm neonates following transfusion with liberal compared to restrictive haemoglobin thresholds.

METHODS

Data were extracted (PubMed/MEDLINE, Embase, Web of Science, and Cochrane Database of Systematic Reviews) and risk of bias and certainty of evidence reviewed. Included studies reported changes in cerebral oxygenation in preterm infants receiving RBCs in response to reaching either a liberal or restrictive transfusion threshold. A random-effects model was used to determine effect size and 95% confidence for the primary outcome of change in crSO2.

RESULTS

Forty-three full-text articles were assessed for eligibility with 7 studies included for meta-analysis (restrictive threshold: n = 357, liberal threshold: n = 220). crSO2 increased following transfusion in both the restrictive (6.40% 95% CI [3.85, 8.95], p < 0.001) and liberal (2.75% 95% CI [0.35, 5.14], p = 0.03) groups. However, the magnitude of change was greater for those transfused at a restrictive threshold (Q = 4.19, df = 1, p = 0.04). The restrictive group also demonstrated a greater transfusion-related decrease in cerebral fractional oxygen extraction than the liberal group (Q = 8.30, df = 1, p < 0.001).

CONCLUSION

Greater improvements to cerebral oxygenation occurred in preterm neonates with a restrictive threshold, suggesting enhanced physiological benefit in response to transfusion. However, meaningful clinical conclusions are limited due to low certainty of evidence and considerable inconsistencies across reporting.