Highlight
– Restrictive (lower) hemoglobin thresholds result in about one fewer transfusion per infant during the primary hospital stay compared with liberal thresholds.
– High-certainty evidence from pooled randomized trials shows no important difference in the composite outcome of death or neurodevelopmental impairment at 18–26 months between restrictive and liberal strategies.
– Mortality at follow-up was also similar; the safety of hemoglobin levels below those tested remains unproven outside trials.
Background and clinical context
Infants of very low birthweight (VLBW; typically defined as birthweight <1500 g) are among the highest users of packed red blood cell transfusions in modern neonatal intensive care units (NICUs). Transfusions are given to treat anemia of prematurity and to support oxygen delivery during critical illness, but transfusion exposure has been associated observationally with complications such as necrotizing enterocolitis, bronchopulmonary dysplasia, and infection—although causality is uncertain.
Clinicians balance two competing concerns: the potential harms of anemia in the developing brain and other organs versus the direct and indirect risks of transfusion. Historically, transfusion practice in preterm infants has been guided by prespecified hemoglobin (or hematocrit) thresholds that vary by postnatal age and clinical status. Over the last two decades randomized trials have compared restrictive (lower) and liberal (higher) transfusion thresholds to determine whether lower thresholds could safely reduce transfusions without compromising survival or neurodevelopment.
Study design and scope of the evidence
This article interprets the updated Cochrane systematic review and meta-analysis (Andersen et al., 2025) that pooled randomized controlled trials (RCTs) comparing lower (restrictive) versus higher (liberal) hemoglobin or hematocrit thresholds for transfusion in VLBW infants. The review searched multiple databases through January 2024 and included six RCTs enrolling 3,451 infants. Trials variably adjusted thresholds by postnatal age and illness severity and used different transfusion volumes and algorithms but were sufficiently similar to permit pooled analysis for key outcomes.
Main outcomes prespecified were the composite of death or neurodevelopmental impairment (NDI) at approximately 18–26 months postmenstrual age, all-cause mortality, and the number of transfusions during the primary hospitalization. The review used standard Cochrane methods and the GRADE framework to assess certainty of evidence.
Key findings
Death or neurodevelopmental impairment (primary composite)
Three large trials reporting later neurosensory follow-up (total 3,041 infants) were pooled for the composite outcome of death or NDI at 18–26 months. The pooled risk ratio (RR) was 1.02 (95% CI 0.95 to 1.09), with a risk difference (RD) of 0.01 (95% CI -0.03 to 0.04). Heterogeneity was moderate (I2 = 55%). Using GRADE the evidence was rated as high certainty. This indicates little to no difference between restrictive and liberal thresholds for the combined outcome of death or NDI in the time window studied.
Mortality at follow-up
All-cause mortality at 18–26 months (three studies, 3,186 infants) showed no important difference: RR 0.99 (95% CI 0.83 to 1.17); RD -0.00 (95% CI -0.03 to 0.02); I2 = 0%. Certainty of evidence was rated high. In-hospital mortality was also not meaningfully different across arms in the included trials.
Transfusion exposure
Across all six trials (3,451 infants), infants assigned to the restrictive transfusion thresholds received fewer transfusions during the primary hospital stay: mean difference -1.05 transfusions per infant (95% CI -1.26 to -0.84). Heterogeneity here was high (I2 = 84%), and GRADE rated the certainty of evidence as low. The modest reduction in transfusion numbers reflects consistent directionality across trials but variable magnitude related to trial protocols, baseline transfusion practice, and included populations.
Other morbidity outcomes
Individual trials reported in-hospital outcomes including rates of intraventricular hemorrhage (IVH), bronchopulmonary dysplasia (BPD), necrotizing enterocolitis (NEC), sepsis, and retinopathy of prematurity. No consistent pattern emerged to suggest clinically important differences favoring one strategy across these morbidities when pooled, though small differences in specific endpoints in single trials have been observed. Importantly, none of the trials showed a signal of large, consistent harm from restrictive thresholds within the tested ranges.
Subgroups and contextual factors
Trials differed in the exact hemoglobin thresholds used, whether thresholds varied with postnatal age or illness severity, and transfusion volumes. Most participants were extremely preterm infants cared for in high-resource NICUs; infants with major congenital anomalies or severe early illness were variably included or excluded. The effects in subpopulations such as surgical neonates, those with significant cardiac disease, or infants in low-resource settings remain less certain.
Clinical interpretation and implications
The pooled evidence supports that adopting restrictive hemoglobin thresholds of the type tested in RCTs leads to modest reductions in transfusion exposure without increasing the risk of death or neurodevelopmental impairment at about 18–26 months’ corrected age. For clinicians and NICU programs aiming to reduce donor exposure, transfusion-related costs, and resource use, a restrictive strategy consistent with the trial protocols appears safe for most VLBW infants represented in those studies.
Important caveats:
- The trials evaluated specific lower thresholds; the safety of hemoglobin concentrations below those lower limits has not been tested and should not be assumed.
- Follow-up to 18–26 months captures early neurodevelopmental milestones but does not exclude later-emerging cognitive, behavioral, or educational outcomes; longer-term follow-up is desirable.
- Heterogeneity in trial algorithms means local implementation should adapt protocols to clinical contexts (e.g., adjustment for critical illness, individualized assessment for oxygenation and hemodynamics).
- Transfusion strategy is one part of broader transfusion stewardship; complementary measures—delayed cord clamping, minimizing iatrogenic blood loss, judicious laboratory testing, and considering erythropoiesis-stimulating agents where appropriate—remain important.
Mechanistic and translational considerations
Anemia may theoretically compromise tissue oxygen delivery; however, compensatory mechanisms (increased cardiac output, altered oxygen extraction) may mitigate risk in many infants. Conversely, transfusions can cause immunomodulation, volume overload, and oxidative stress—mechanisms implicated in observational associations between transfusion and morbidities. Randomized trial data suggest that within the studied thresholds the balance favors restrictive strategies for reducing exposure without detectable clinical harm by 18–26 months.
Limitations of the evidence
– Heterogeneity: trials used different threshold algorithms and transfusion volumes, contributing to statistical heterogeneity for transfusion number outcomes.
– Follow-up duration: neurodevelopmental assessment at 18–26 months is an important endpoint but may miss late-emerging deficits.
– Generalizability: most trials were conducted in high-income-country NICUs; applicability to low-resource settings or to infants with complex comorbidities is less certain.
– Unstudied lower limits: hemoglobin concentrations below the lowest trial thresholds remain untested and should not be adopted outside clinical trials.
Expert commentary and guideline context
Recent large randomized trials and systematic reviews have shifted the balance of evidence toward restricted transfusion strategies in VLBW infants. Professional guidance increasingly supports contextualized restrictive thresholds while emphasizing individualized care and monitoring. Implementation requires clear protocols that define thresholds by postnatal age and clinical status, staff education, and rigorous monitoring of outcomes including neurodevelopmental follow-up.
Practical takeaways for clinicians
- Consider implementing restrictive transfusion thresholds consistent with contemporary trial protocols for VLBW infants to reduce transfusion exposure without observable harm by early childhood.
- Do not adopt hemoglobin targets below the lowest limits tested in trials outside of research settings.
- Integrate transfusion protocols with broader blood-sparing strategies (delayed cord clamping, reduced phlebotomy, clinical monitoring).
- Commit to structured neurodevelopmental follow-up and local audit when changing transfusion practice.
Conclusion and recommendations for practice
High-certainty evidence from randomized trials indicates that restrictive hemoglobin transfusion thresholds, of the type tested in contemporary NICU trials, reduce transfusion exposure modestly and do not increase the risk of death or neurodevelopmental impairment at roughly two years’ corrected age. Clinicians may consider adopting restrictive protocols aligned to the published trial algorithms while ensuring close clinical monitoring and avoiding hemoglobin levels below those studied. Ongoing data collection and longer-term neurodevelopmental follow-up are prudent, and further research should examine very low thresholds, specific vulnerable subgroups, and outcomes beyond early childhood.
Funding and trial registration
The Cochrane review (Andersen et al., 2025) synthesizes trials that were individually funded and registered; prominent trials in this field include the TOP and ETTNO trials (see trial publications for specific funding and registration details). Individual trial publications and trial registries (ClinicalTrials.gov and equivalent registries) should be consulted for exact funding sources and registration numbers.
Selected references
1. Andersen C, Stark MJ, Crawford T, Whyte RK, Franz AR, Soll RF, Kirpalani H. Low versus high haemoglobin concentration threshold for blood transfusion for preventing morbidity and mortality in very low birthweight infants. Cochrane Database Syst Rev. 2025 Dec 11;12(12):CD000512. doi:10.1002/14651858.CD000512.pub3 IF: 9.4 Q1 .
2. Kirpalani H, Whyte RK, Andersen C, et al. (TOP trial and related publications). Randomized trials comparing restrictive and liberal transfusion thresholds in preterm infants. See primary trial publications (e.g., Kirpalani et al., 2020–2021) for detailed methodology and results.
3. Whyte RK, Franz AR, et al. (ETTNO trial and related publications). See primary trial publications for trial registration and funding details.
4. British Committee for Standards in Haematology (BCSH). Guidelines on the transfusion of red cells in neonates and children. (Consult the relevant BCSH guideline for practical transfusion thresholds and recommendations.)
