The tangled duo of anaemia and transfusion travel together, that is the problem

In 1981, Dallman recognised that ‘rapid developmental changes and complex interactions for oxygen delivery (prevent us from) developing clear cut criteria for transfusion. Consequently clinical practices vary widely’ 1. This confusion remains, as an international survey shows 2. Making aetiological links to neonatal diseases and levels of haemoglobin prompting transfusions are fraught by the relationship between the two. Lundgren and coworkers in this issue provide another facet of potential clinical relevance, this time to retinopathy of prematurity (ROP) 3. Because of the central role of oxygen in ROP 4, 5, a potential aetiological role of transfusions was logically explored. A link to oxidative iron reactions was proposed in 1941 6. When VEGF was found to play a central role in the development of ROP 7, it was also found to have time-sensitive effects. Understanding the time course of VEGF in terms of postnatal age could resolve confusing disagreements between clinical studies. In the modern era, studies examining the relationship between transfusion and ROP have evolved in three historical components. First, early cohort retrospective observations proposed an aetiology for red blood cell transfusions in ROP. These studies were largely small and single site 7-12. A second wave of studies used prospective observations to examine the covariables of transfusion from blood volumes, numbers of transfusions, potential iron loading and degree and duration of anaemia 13-17. These prospective studies corroborated a dose-dependent relationship of ROP to blood transfusions, including in the first week of life but did not resolve the relationship with iron loading. None could, limited by design, resolve the interdependence of anaemia and transfusion itself. In this edition of Acta Paediatrica, Lundgren et al. again address the relationship between anaemia, transfusions and retinopathy warranting treatment (TW-ROP) 2. This retrospective cohort study of infants <28 weeks gestation relies on an apparent unit policy to obtain a daily routine haemoglobin. The investigators defined anaemia as <110 g/L and severe anaemia as <80 g/L. They collected the interdependent variables haemoglobin, days of anaemia and of severe anaemia, blood transfusion, sepsis, ventilation and duration of oxygen requirement, each over four overlapping epochs of life. The epochs range from first week of life to up to 35 weeks PMA. In the survivors to 40 weeks of postmenstrual age, 25 of 227 (11%) suffered TW-ROP. Not surprisingly, infants with TW-ROP when compared with those without were of significantly smaller birthweight (694 vs 908 g), lower gestational age (24 vs 26 weeks) and of lower haemoglobin in the first week of life (125 vs 138 g/L). Severe anaemia was also significantly more common (57% vs 53%), as was sepsis (26% vs 24%). After adjusting for gestational and postnatal age, the development of TW-ROP was significantly associated with anaemia in the first week of life, blood transfusion in the first four weeks of life, sepsis in the first four weeks of life and ventilator support in the first four weeks of life. Of importance, measures of intrauterine or postnatal weight gain were not included. After these variables were entered into a stepwise forward multivariate regression model, only three remained statistically significant (OR, 95% CI): episodes of sepsis at one week to four weeks 3.1 (1.1–9.0), days of anaemia <110 g/L OR: 1.46 (1.16–1.83) and days of ventilation from birth to 35 weeks PMA 1.03 (1.01–1.06). Clearly, there is an association between episodes of sepsis, early anaemia, days of ventilation and retinopathy of prematurity. But the relationship between TW-ROP, anaemia and transfusions could not be addressed, because anaemia and blood transfusion are highly correlated and not separable in this analysis. Here, daily measurements of haemoglobin may simply be a strong predictor of blood transfusion, in which case the transfusion variable makes no further contribution to the model. For a similar reason, a much larger Danish national study of similar design 18 identified blood transfusion rather than neonatal anaemia as a risk factor for TW-ROP; as anaemia was identified only by its ICD-10 code, it is likely that blood transfusion was representing a better estimate of anaemia in the model. The third phase of relevant studies includes trials of therapy (primarily transfusion) and not, directly, aetiology (anaemia). A Cochrane review identified four RCTS where transfusion was triggered by anaemia and found no association between the use of restrictive and liberal transfusion haemoglobin thresholds and ROP 19. The difference between randomized groups in transfusions received was modest (overall mean 1.1 transfusions per infant), as were differences in mean haemoglobin levels throughout the trials, the largest difference being 11 g/L in PINT. The trials, even when combined, were not sufficiently powered to address the effects of either anaemia or transfusion on TW-ROP (restrictive vs liberal RR: 1.04, 0.68–1.58). The review did not address a small ‘symptom-based’ trial by Brooks of late transfusion policy of infants of less than 1251 g birthweight and greater than four weeks of age 20. This design resulted in much greater differences in transfusion rates (2.8 vs 5.7 transfusion per infant) and therefore in haemoglobin (108 vs 132 g/L). No significant differences were found in any category of ROP (restrictive vs liberal 83% vs 73%); this study was also underpowered. If, in clinical practice, the point of aetiology is primarily to direct therapy, the ultimate test will be the randomized trial, even when, as in this example, the therapy (transfusion) has complex effects that go beyond simply correcting the aetiology (anaemia). Lundgren et al. contribute to the message that anaemia is strongly associated with TW-ROP 3. The relationship between TW-ROP and blood transfusion remains, as before, unaddressed.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.