NICU transfusion guidelines and strategies to minimize transfusions

Introduction

Pathologic anemia is defined by a red blood cell (RBC) mass inadequate to meet the oxygen needs of tissues. Neonates often require red cell transfusions due to a decreased red cell mass brought about by acute blood loss. Preterm infants frequently experience a decline in red cell mass due to phlebotomy, a shortened red cell lifespan, and lack of erythropoietic response to a dropping hematocrit due to lack of erythropoietin (Epo) production, termed the anemia of prematurity. Target hemoglobin and hematocrit have been used as an indicator for a red cell transfusion in preterm neonates. Recently published data from two large multicentered trials in Europe and the United States showed that a restrictive approach (using a lower hematocrit to trigger a transfusion) resulted in fewer transfusions without an increase in death or serious neurodevelopmental impairment. The identification of individualized markers for transfusion need, beyond hematocrit triggers, that optimally balance the risk and benefits of transfusion, coupled with strategies to increase red cell mass to minimize transfusions, will serve to improve “transfusion stewardship,” the appropriate and judicious use of transfusions in the neonatal intensive care unit (NICU).

The development of neonatal transfusion guidelines

Approaches to RBC transfusions in neonates have changed significantly over the last 40 years. In the 1970s and 1980s, phlebotomy losses in NICU patients were carefully recorded, and losses replaced when they reached 10 mL/kg. Transfusions were generally administered when the hematocrit dropped below 40%, and top-off transfusions (5–10 mL/kg volumes) were common in the later weeks of NICU hospitalization. Signs ascribed to anemia in preterm infants—apnea, poor feeding, tachycardia, and poor weight gain—would result in frequent transfusions over a wide range of hematocrits, without clear evidence of benefit. A focused approach to developing transfusion guidelines was initiated in Canada in the 1990s after thousands of Canadians were exposed to hepatitis C and human immunodeficiency virus from contaminated blood. Major Canadian multicentered trials evaluating restrictive vs. liberal transfusion triggers in adult (TRICC) and pediatric critical care patients (TRIPICU) both showed that a hemoglobin threshold of 7 g/dL for red cell transfusions decreased the number of transfusions received without increasing adverse outcomes. ^,

Since the early 2000s a number of studies have been performed evaluating hematocrit thresholds for transfusion in preterm infants. Bell et al. randomized 100 preterm infants 500 to 1300 g birth weight to a liberal (higher hematocrit) or restrictive (lower hematocrit) transfusion threshold strategy and compared clinical outcomes. Infants in the liberal-transfusion group received more RBC transfusions (5.2±4.5) compared to the restrictive-transfusion group (3.3±2.9). There were no differences in pretransfusion cardiac output, hospital days, or survival to discharge. Investigators found an increase in grade 3 or 4 intraventricular hemorrhage (IVH) or periventricular leukomalacia (PVL) in the restrictive group (6/28 vs. 0/24 in the liberal group). They concluded that the more frequent major adverse neurologic imaging findings in the restrictive group suggested restrictive transfusions might be harmful. Long-term follow-up of infants enrolled in the Iowa trial was performed on a subset of children available for evaluation at 12 years. Both cognitive function (based on developmental assessment) and magnetic resonance imaging (MRI) outcome were actually better in the restrictive group, lessening the initial concerns that a restrictive strategy would impair long-term neurodevelopment. ^,

Canadian investigators for the Premature Infants in Need of Transfusion (PINT) Study randomized 451 extremely low birth weight (ELBW) infants to a high or low threshold transfusion strategy within 48 hours of birth to determine effects of transfusion thresholds on hospital outcomes. Infants in the low threshold group received fewer transfusions and were transfused at a later age. There were no differences in morbidities or mortality between the low and high hemoglobin threshold groups, resulting in no difference in the composite outcome of death and serious morbidity at the time of discharge (bronchopulmonary dysplasia [BPD], severe retinopathy of prematurity [ROP], or, importantly, brain injury identified on ultrasound). At 18 to 21 months corrected age there were no differences between the two groups in the composite outcome of death or neurodevelopment impairment, defined as cerebral palsy (CP), significant visual or hearing impairment, or a Bayley Scales of Infant Development II (BSID II) Mental Development Index (MDI) score below 70. In post hoc analyses using an MDI score <85 instead of <70, the primary outcome of death or neurodevelopmental impairment (NDI) was more likely in the low hematocrit group (45%) compared to the high hematocrit group (34%), leading investigators to hypothesize that maintaining a higher hematocrit would decrease the incidence of neurodevelopmental impairment in preterm infants.

A 2012 meta-analysis included both the Iowa and PINT studies and reported no difference in morbidities or mortality rates between high and low threshold transfusion groups. Similar to adult and pediatric critical care populations, a restrictive (low) hematocrit threshold compared with a liberal (high) threshold (hematocrit 35–40%) resulted in fewer transfusions with no increase in mortality or serious morbidity. ^,

Recently, two similarly designed, large multicenter randomized trials were performed to confirm previous findings and determine effects of transfusion strategies on long-term outcomes. The Transfusion of Prematures (TOP) trial was designed by the principal investigators of the PINT and Iowa studies to test the hypothesis that maintaining a higher hematocrit would result in decreased NDI, by comparing neurodevelopmental outcomes of ELBW infants randomized to a high or low hematocrit threshold for transfusion. This hypothesis was based on concerns over the increased IVH identified in the Iowa study, and the post hoc finding of increased percentage of infants in the low hematocrit group scoring below 85 in the PINT trial. Similarly, European investigators performing the Effects of Liberal vs. Restrictive Transfusion Thresholds on Survival and Neurocognitive Outcomes (ETTNO) study evaluated higher and lower hematocrit strategies for red cell transfusions, testing the hypothesis that the lower hematocrit strategy would lead to an increase in the primary outcome of death or neurodevelopmental disability.

For the TOP trial, neonatologists were surveyed to determine an acceptable range of hematocrits that could be used to trigger transfusions, and triggers were chosen to achieve a statistical difference in hemoglobin of 2 to 2.5 g/dL between groups (hematocrit 5–6%). In the ETTNO study, transfusion triggers were guided by current clinical practice in Germany. The high and low thresholds chosen also aimed to produce a clinically relevant difference in mean hemoglobin concentrations between treatment groups of about 2 g/dL, to improve recognition of any effect of hemoglobin thresholds on neurocognitive outcome compared with the PINT trial, where differences between high and low thresholds were narrower.

Enrollment criteria and study methods for the two studies are shown in Table 10.1 . In the ETTNO trial, transfusions prior to enrollment did not preclude participation in the study, and approximately 25% of infants received least one transfusion prior to randomization (24% vs. 25%, low vs. high). In TOP, transfusions could be given emergently prior to 6 hours of age (this occurred in 5% of the high group and 4% of the low group), but infants were ineligible if they received a transfusion after 6 hours of age. Transfusion triggers for each study are shown in Table 10.2 .

The primary outcome for both trials was the combined outcome of NDI or death. NDI was defined in similar fashion in both trials: cognitive score less than 85 (on the composite cognitive score of the BSID III for TOP; BSID II MDI for ETTNO), moderate or severe CP (gross motor function classification system [GMFCS] ≥2 in TOP; Surveillance of Cerebral Palsy in Europe network definition for ETTNO), severe vision impairment, or severe hearing impairment.

Outcomes for both trials are shown in Table 10.3 . Both trials found no differences between high and low groups in the primary outcome. In fact, outcomes were basically identical between groups for both studies. There were also no differences between groups in the individual components (death or NDI) of the primary outcome. Importantly, the percentage of infants with BSID III composite cognitive scores less than 85 was similar between groups: 269/695 (38.7%) in the high groups compared with 270/712 (37.9%) in the low group, with an adjusted relative risk (RR) of 1.04 (95% confidence interval [CI] 0.91–1.18). Concerns identified in the post hoc analyses performed in the PINT trial were alleviated by this result. Cognitive delay was the primary factor determining NDI and was identified in 97% of the infants in the high groups and 91% of the infants in the low group who were designated as neurodevelopmentally impaired. No differences between groups were identified in common neonatal hospital morbidities (BPD, ROP, grade 3–4 IVH or PVL, or necrotizing enterocolitis [NEC]) (see Table 10.3 ). Metrics associated with severity of illness such as length of stay, time to full feeds, length of time on a ventilator, and duration of caffeine treatment were similar between low and high groups.

Similar to TOP, there were no differences between groups in the individual components (death or NDI) of the primary outcome for the ETTNO study. Importantly, the percentage of infants with BSID II MDI scores below 85 was similar between groups (37.8% high vs. 35.9% low), with an adjusted RR of 1.09 (95% CI 0.81–1.46). Cognitive delay (BSID II <85) was the primary factor determining NDI and was identified in 88% of the infants in the high group and 86% of the infants in the low group who were designated as neurodevelopmentally impaired.

The number of transfusions administered were significantly lower for both studies in the restrictive group compared to the liberal group: 4.4±4.0 vs. 6.2±4.3 transfusions in TOP; 1.7 vs. 2.6 transfusions in ETTNO. Moreover, a greater number of infants in the low threshold groups remained untransfused (see Table 10.3 ).

The results of these two well-designed, well-performed multicenter trials prove conclusively that transfusing critically ill ELBW infants at lower hematocrits did not result in adverse outcomes, while infants transfused at higher hematocrits did not do worse. Was there evidence that infants in either arm actually required a transfusion and improved following the transfusion? Aside from change in hematocrit, efficacy data were not collected. Because both studies relied on consensus in determining hematocrit thresholds, both studies likely ended up measuring what infants received, rather than what they needed. In promoting transfusion stewardship, documenting evidence of benefit should be a part of future studies.

Strategies to minimize transfusions

Prevalence of anemia

Although a physiologic decrease in hematocrit occurs universally following birth, identifying a true incidence of clinically significant anemia at various gestational ages is difficult. A variety of cutoff thresholds for anemia, inclusion or exclusion of infants receiving transfusion, and varying thresholds for transfusion makes identifying this incidence difficult. However, in a landmark study, Dallman published trends in anemia of prematurity. In this study, he demonstrated that in term infants weighing more than 3000 g, the physiologic nadir of hemoglobin levels occurred at about 2 months of life, reaching 11.0 g/dL. The physiologic hemoglobin nadir was more pronounced and occurred slightly earlier in premature infants, with hemoglobin levels decreasing to 9.5 g/dL for infants 1500 to 2000 g and 9.0 g/dL for infants less than 1500 g.

The physiologic decrease in hemoglobin and hematocrit is largely due to a rapid decrease in endogenous erythropoietin levels just after birth. Following birth, the P _a O ₂ (generally ~27 mm Hg before delivery) increases dramatically once the newborn infant has taken first breaths. With increased oxygenation there is a resultant rapid decrease in endogenous erythropoietin production, which is followed in 5 to 7 days by a decrease in reticulocyte count. These factors contribute to the physiologic anemia that universally occurs following birth.

Christensen et al. reported the trend in hematocrit and hemoglobin over the first 28 days of life among infants born at 35 to 42 weeks and 29 to 34 weeks of gestation. This study included nearly 42,000 patients born at 35 to 42 weeks of gestation and nearly 40,000 patients born at 29 to 34 weeks of gestation ( Fig. 10.1 ). Infants were excluded when their diagnosis included abruption, placenta previa, fetal anemia, or when a blood transfusion was given. In the 35- to 42-week cohort, the hemoglobin decreased from 18 g/dL to 13 g/dL, a decrease of 5 g/dL. Among the 29- to 34-week cohort, the initial hemoglobin decreased from approximately 17 g/dL to 11 g/dL.

Infants born more premature, smaller, and more critically ill undergo greater phlebotomy losses (both in number of times blood is sampled and in total volume/kg) to provide care. These phlebotomy losses exacerbate the likelihood of anemia. Estimates of NICU laboratory blood draws primarily in the first 2 weeks of life among premature VLBW infants are higher than the remainder of a neonate’s hospitalization. During the first 6 weeks of life laboratory blood losses have been reported to range from 11 to 22 mL/kg/week, which is equivalent to 15% to 30% of the circulating blood volume. There is a direct relationship between the volume of blood removed for laboratory testing and the volume of blood transfused from VLBW infants ( Fig. 10.2 ). In several studies, highly significant direct associations have been observed between volume of blood removed with that transfused (i.e., correlation coefficients of 0.75–0.90). As described, phlebotomy loss is a significant contributor to anemia in ELBW and VLBW infants. Despite widespread acceptance of this practice, marked variation in blood lost due to laboratory testing has been noted. For example, in comparing laboratory blood loss over the first 2 weeks of life in infants weighing less than 1500 g, Ringer et al. reported that one NICU drew 17.5 mL/kg while another drew nearly twice as much (34.1 mL/kg). After statistical adjustment for birth weight, gestational age, and severity of illness, laboratory blood loss remained 10.7 mL/kg per patient different between the two NICUs. Obladen et al. in the late 1980s reported blood loss of 24, 60, and 67 mL/kg for increasingly sick infants in their cohort. In addition to reporting significant variation in phlebotomy loss (7–51 mL/kg) Nexo et al. reported that among 20 VLBW infants that 25% of the phlebotomy losses were in excess of the need for analytical procedures. While variation in phlebotomy loss may in part be explained by differences in gestational age, birth weight, severity of illness, and other factors, the wide variation observed indicates considerable opportunity to limit phlebotomy loss.

Rates of phlebotomy vary greatly between hospitals. Rosebraugh et al. conducted a prospective study measuring the actual phlebotomy and “hidden” phlebotomy loss among ELBW infants in the first 28 days of life and reported phlebotomy losses among VLBW infants in the first month of life of greater than 60 mL/kg, with blood loss of 10 mL/kg on the first day of life. A more recent study among 20 infants with an average birth weight of 836 g reported approximately 30 mL/kg cumulative iatrogenic blood loss in the first 28 days of life, with 2.7 mL/kg blood loss on the first day of life. This institution routinely uses inline point of care for analysis of blood gas, glucose, and bilirubin.