Introduction

The most impactful NICU platelet transfusion threshold study published to date is the randomized clinical trial known as PlaNet-2 (Platelets for Neonatal Transfusion – study 2) [1]. That report, followed by additional analyses of the data [2, 3] and a two-year follow-up study [4], offers high quality evidence that liberal platelet transfusion practices cause short- and long-term harm to premature infants.

The PlaNet-2 study randomized neonates born less than 34 weeks gestation to receive prophylactic platelet transfusions if their platelet count fell below either a low threshold (25,000/µL) or a higher threshold (50,000/µL). Among 660 study infants, 53% of those in the low-threshold group received at least one platelet transfusion vs. 90% of those in the high-threshold group. A new major bleeding episode or death occurred in 19% of the low-threshold vs. 26% of the high-threshold infants (odds ratio, 1.57; 95% CI, 1.06 to 2.32; P = 0.02). On analysis of secondary outcomes, the percentage of infants surviving with bronchopulmonary dysplasia was 54% in the low-threshold vs. 63% in the high-threshold group (odds ratio, 1.54; 95% CI, 1.03 to 2.30). The authors concluded that the lower transfusion threshold was a safer practice, resulting in fewer transfusions and a lower risk of major bleeding, death, or BPD [1,2,3,4].

Publication of high-level evidence showing that one practice is more beneficial than another is not always followed by rapid widespread adoption of that improvement. In fact Rubin recently determined it takes an average of a decade and a half to accomplish this [5]. It is clear that NICU guidelines for platelet transfusions should incorporate the evidence of the PlaNet-2 Trial. Specifically, the platelet count threshold, below which a prophylactic platelet transfusion could be considered for most non-bleeding NICU patients, should not be set at 50,000/µL but at 25,000/µL. It should not take us 17 years to accomplish this practice change.

Sometimes a single hospital can expeditiously implement a new evidence-based practice, whereas implementation in a large multihospital health system, and in the majority of real-world practice sites, typically requires years of education, accountability-tracking, and identifying and removing the barriers to implementation. An example of rapid adoption of restrictive platelet transfusion guidelines by a single NICU occurred at Boston Children’s Hospital [6]. Restrictive guidelines, with a threshold <25,000/µL for most patients, consistent with PlaNet-2 conclusions, were proposed by a multidisciplinary team, with the outcome measure of reducing “non-indicated” NICU platelet transfusions (Table 1). Over the 12-months following adoption of the new restrictive guidelines, the rate of non-indicated platelet transfusions per 100 patient admissions decreased from 12.5 to 2.9, while rates of major bleeding remained stable.

Table 1 Neonatal platelet transfusion guidelines at Boston Children’s Hospital before (version 1) and after (version 2) January 2019 [6]. (Modified from Davenport et al. J Perinatology 2021: 42; 1487-94).
Full size table

Simultaneous attempts to institute basically the same restrictive guidelines in the multihospital Intermountain Health system were less successful. Three years after changing the guidelines to the <25,000/µL threshold for most NICU patients, a system-wide pre- vs. post-practice change analysis postulated a significant reduction in platelet transfusions and improved outcomes, but neither was found [7].

The purpose of creating this review article was to facilitate efforts to adopt evidence-based platelet transfusion guidelines in real world neonatology practice. Our intent is to accomplish this by heightening general awareness of; (1) the adverse effects of platelet transfusions to neonates, (2) the association between platelet transfusion “refractoriness” in neonates and adverse outcomes, and (3) the need to discover alternatives to adult donor platelet transfusions for thrombocytopenic NICU patients.

Adverse effects of neonatal platelet transfusions

Liberal NICU platelet transfusion practices of the past (such as using a 50,000/µL or even a 100,000/µL threshold for prophylactic transfusions of NICU patients) were based on undefined transfusion risks, and by a fear of significant bleeding. For instance, if a stable preterm infant had a platelet count of 50,000/µL, it might have been less anxiety-provoking to order a platelet transfusion than to withhold the transfusion and worry about the potential occurrence of a significant bleeding event, such as intraventricular hemorrhage.

Some of the early evidence that NICU platelet transfusions carried significant risks arose from studies at the University of Florida in 2001, where it was reported that neonates who received one platelet transfusion had a relative risk of death 10.4 times that of neonates who received none (p = 0.0001), and those who received >4 platelet transfusions had a risk of death 29.9 times that of those who received none (p = 0.0001) (Table 2) [8]. Subsequently, two additional studies found an association between the number of platelet transfusions administered during NICU hospitalization and mortality (Figs. 1 and 2) [9, 10]. Associations like this do not prove causality, because the number of platelet transfusions received might be a co-variant of severity of illness. However, after repeated studies revealed significant statistical associations between number of platelet transfusions and mortality rate [8, 11], statistical evaluations were performed including sensitivity analyses, which showed that the platelet transfusions were likely responsible for some fraction of the increasing mortality rate as transfusion numbers increased [9]. Finally, with the publication of PlaNet-2, it has become clear that neonatal platelet transfusions can result in harm to preterm infants, regardless of severity of illness, validating the previously published work by our groups and others.

Table 2 Relative risk of death of patients in the NICU at Shands Hospital, University of Florida, according to the number of platelet transfusions received* [8]. (Modified from DelVecchio et al. Transfusion 2001; 41;6: 803-8).
Full size table
Fig. 1: Mortality rate of 1600 NICU patients with a platelet count <150,000/µL, according to the number of platelet transfusions received [9].
figure 1

(From Baer VL et al. J Perinatology 2007; 27: 790–6).

Full size image
Fig. 2: Mortality rate of 273 NICU patients with a platelet count <50,000/µL, according to the number of platelet transfusions received [10].
figure 2

(Modified from Baer VL et al. Pediatrics 2009; 124; e1095–1100).

Full size image

Long-term follow up of children from the PlaNet-2 Trial at two-years’ corrected age found an increased rate of death or neurodevelopmental impairment (NDI) in the high- compared to the low- threshold group (50% versus 39%, respectively, OR 1.54, 95% CI, 1.09–2.17) [4]. These findings were corroborated by a secondary analysis of the Preterm Erythropoietin Neuroprotection Trial (PENUT), which found an increased incidence of death or severe NDI at two-years’ corrected age in infants exposed to platelet transfusions during their NICU admission after adjusting for propensity score, gestational age, and trial treatment group (46.5% versus 13.9% OR 2.43 95% CI 1.24–4.76) [12]. Further, secondary analysis of the specific components of NDI found an association between platelet transfusion and lower mean motor scores (measured with the Bayley Scales of Infant Development—Third edition), with each additional platelet transfusion being associated with a 1.1-point decrease in mean motor score.

In addition to their association with death and NDI, neonatal platelet transfusions have a significant association with the incidence and severity of bronchopulmonary dysplasia (BPD). As early as 2009 our group described the association between very high users of platelet transfusion and the subsequent development of severe BPD [13]. The PlaNet-2 Trial found an increased incidence of BPD in infants randomized to the high-compared to low-threshold (as described above). Strikingly, in the follow up of the PlaNet-2 Trial, there was an increased need for supplemental oxygen or respiratory support at two-years’ corrected age in infants randomized to the high- compared to low-threshold group (11% versus 4%, respectively, OR 2.86, 95% CI, 1.25–6.51) [4]. Most recently, in a case series of 10+ years’ experience in the Intermountain Health NICUs, we found that all neonates who received 25 or more platelet transfusions developed severe BPD or died [13, 14].

The mechanisms underlying the harm caused by platelet transfusions are unknown, but pre-clinical and clinical research in this field is active and ongoing. It has been hypothesized that at least part of the harm is due to a developmental mismatch that occurs when neonates receive a transfusion of comparatively hyper-reactive adult platelets [6, 15, 16]. Importantly, the hypo-reactivity of neonatal platelets is not a deficiency, but rather an integral part of a unique, well-balanced, and effective neonatal hemostatic system [17]. However, it is critical that clinicians consider the immune functions of platelets, in addition to their classic role in hemostasis, when deciding to order a platelet transfusion. Platelets directly interact with immune cells and regulate critical immune and inflammatory responses through variable mechanisms, to the point that many experts consider them de facto immune cells [18]. Findings from a recent study comparing adult and cord blood platelets suggested that adult platelets may be more pro-inflammatory than neonatal platelets [19]. Pre-clinical studies transfusing adult platelets to newborn mice have also shown that these transfusions trigger dysregulated immune responses in the recipient neonates, generating increased levels of pro-inflammatory cytokines and making the neonatal immune cells more able to migrate to sites of inflammation [20]. Thus, it is quite possible that the inflammatory, rather than the hemostatic, effects of transfused adult platelets are responsible for the associations with pulmonary inflammation [21], chronic lung disease [4, 22], neurodevelopmental impairment and death [12].

Platelet transfusion refractoriness and adverse outcomes

For better patient outcomes, and to conserve resources, it is highly desirable to avoid multiple platelet transfusions. One group of thrombocytopenic neonates likely to receive multiple platelet transfusions is those developing platelet transfusion refractoriness. If a platelet transfusion fails to increase the platelet count “adequately,” a repeat platelet transfusion is typically ordered. If that next platelet transfusion also fails to generate an adequate rise, yet another platelet transfusion is very likely to be ordered, and so on. In this way, thrombocytopenic neonates can receive one or more platelet transfusions daily, for many days or weeks. It is our experience that virtually all such heavily transfused neonates (e.g., 20 to 25 or more platelet transfusions) develop severe BPD or die from respiratory failure [13, 23].

The diagnosis of platelet transfusion refractoriness in neonates is made when a transfusion dose of 10 mL/kg fails to increase the platelet count by an increment of at least 5,000/µL [23, 24]. The “increment” is defined as the platelet count 1 h after finishing the transfusion minus the count before the transfusion [23]. Among eight neonates we cared for, who received 29–52 platelet transfusions each, all had many refractory transfusions. In fact 19–73% of the multiple transfusions given to these eight failed to increase their platelet count by ≥5000/µL/10 mL/kg of platelets transfused [14]. Three of the eight had late NICU deaths related to respiratory failure; all five survivors had severe bronchopulmonary dysplasia requiring tracheostomy for prolonged ventilator support for at least one year. Platelet transfusion refractoriness in adult transfusion recipients have uniquely different causes than in neonates. In adults, antibodies causing immune destruction of platelets are typically implicated. These can be antibodies to HLA antigens or human platelet antigens (HPA) [24].

To better inform the issue of platelet transfusion refractoriness in thrombocytopenic neonates, we conducted studies assessing factors associated with the magnitude of post-transfusion increment. In a dataset of nearly 2000 NICU platelet transfusions, we found an average increment of 35,000/µL/10 mL/kg transfusion, but with marked variability [23]. Table 3 reviews the variables we found to be associated with the magnitude of post-platelet transfusion increment. To standardize reporting we advocate obtaining a platelet count one hour after the platelet transfusion finishes. Some variation in increment is the result of variation in the number of platelets transfused. The American Red Cross minimum standard for platelet concentration in an apheresis unit is 3 ×1011 platelets, in a volume of generally about 300 mL. However, in actual practice this might be 2-fold higher, which can be a factor in variability of increment between transfusions [23, 25].

Table 3 Determinants of the post-transfusion platelet count increment.
Full size table

The underlying cause of the thrombocytopenia is an important determinant of the post-transfusion increment. Some of the causes of thrombocytopenia due to platelet destruction (i.e., disseminated intravascular coagulopathy, sepsis, or necrotizing enterocolitis) adversely affect the donor platelets as well, leading to a smaller increment. The important safety process of pathogen reduction can also reduce the increment, but rather modestly; an increment of about 10,000/µL less for every 10 mL/kg platelets compared. The oldest platelets in the blood bank, those about to expire, will typically give a lower increment than the freshest platelets. Also, ABO mismatched platelets are more likely to give a lower increment than ABO identical (data supporting each conclusion found in reference) [23].

Cognizance of these determinants of the post-transfusion increment suggests methods to reduce the number of platelet transfusions given to infants with refractoriness. Table 4 lists some of these methods. For instance, if a thrombocytopenic infant fails to achieve an increment >5000/µL, neonatologists can consider reducing the transfusion threshold (for instance from 25,000 to 20,000 or 15,000/µL). However, it is important to recognize that the safety of using thresholds lower than 25,000/µL has not been established in neonates. Also, for subsequent platelet transfusions of an infant who is refractory, consult with the blood bank’s transfusion medicine physician or pathologist about freshest-possible or ABO identical platelets, and consider medications the infant is receiving that have been shown to elicit drug-dependent platelet-reactive antibodies. Although we judge this latter to be uncommon in infants, it has been reported, and the testing can be done by contacting the blood bank [26, 27].

Table 4 Recommendations for avoiding multiple platelet transfusion among thrombocytopenic infants who develop platelet transfusion refractoriness.
Full size table

Unless an infant’s platelet count is extremely low, for instance under 10,000/µL, the platelet count by itself is not an accurate measure of that infant’s bleeding risk [28,29,30]. Table 5 lists additional tests that could assist in informing the significance of any given platelet count to predict the bleeding risk. An elevated immature platelet fraction (analogous to an elevated reticulocyte count) suggests increased platelet production; thus a larger fraction of circulating platelet are younger, larger, and more hemostatically capable at any given platelet count [31,32,33,34,35]. Similarly an elevated mean platelet volume (MPV) in a thrombocytopenic neonate suggests younger, larger and more hemostatically capable platelets, at any given platelet count [23, 35,36,37,38,39,40,41], although it is important to distinguish this presentation from that of platelet function disorders characterized by thrombocytopenia and large platelets (like Bernard-Soulier or gray platelet syndrome). The template bleeding time measures the time to hemostasis of a standardized cutaneous incision, but it is highly operator dependent and is rarely used outside of research applications. Also the template bleeding time is not recommended in cases of severe thrombocytopenia [42,43,44,45,46]. The PFA-100 (Dade Behring, Marburg, Germany) is an in vitro method for quantifying platelet adhesion, activation, and aggregation in whole blood. It has advantages over the template bleeding time, including that it is far less operator-dependent and probably more reproducible [45,46,47,48,49,50,51,52]. Limitations include the inability to run the PFA-100 test if the hematocrit is <20% and/or the platelet count is <50,000/µL. Also, the amount of blood required for the test likely would limit its widespread use.

Table 5 Suggested adjunctive laboratory tests that might be used in addition to the platelet count to better inform the bleeding risk of thrombocytopenic neonates.
Full size table

Potential alternatives to transfusing adult donor platelets

In the future, transfusing neonates using platelets obtained from adult donors might not be our only therapeutic option. As listed in Table 6, other potential therapies under study include; (1) transfusing platelets obtained from otherwise discarded healthy term umbilical cord blood, (2) transfusing in vitro-generated platelets, (3) stimulating endogenous platelet production with thrombopoietin receptor agonists, and (4) thrombosomes, fragments, synthetic platelets and nanoparticles.

Table 6 Potential alternatives to transfusing adult donor-derived platelets, for thrombocytopenic neonates who qualify for a platelet transfusion.
Full size table

Apheresis platelet units contain at least 3 × 1011 platelets in a volume of approximately 300 mL, or a concentration of about 1 × 109 platelets/mL [23, 25]. Could a comparable concentration of platelets (1 × 109 platelets/mL) in a much smaller volume (perhaps 10 mL) be harvested from otherwise discarded umbilical cord blood after healthy term deliveries? The average platelet count in term umbilical cord blood is 250,000/µL (95% CI, 100,000 to 400,000/µL) [38]. Thus 50 mL of cord blood contain an average of 12.5 × 109 platelets (95% CI, 5 × 109 to 20 × 109). If a yield of only half of these platelets could be harvested by centrifugation, and suspended in a volume of 10 mL, that 10 mL transfusion product would contain 0.63 × 109 platelets/mL (95% CI, 0.25 to 1.0 × 109/mL). Therefore, the number of platelets that might be harvested from otherwise discarded term umbilical cord blood could be sufficient for transfusing a preterm neonate (10–15 mL/kg). Figure 3 illustrates the schema we are testing to harvest healthy, term, otherwise discarded umbilical cord blood, from which fetal platelets could be harvested and prepared as a transfusion product for thrombocytopenic neonates. It is doubtful that umbilical cord blood would yield sufficient platelets to meet the transfusion needs of a term thrombocytopenic neonate. Also, the short shelf-life (5–7 days) of harvested platelets would be a barrier to implementing a baby-platelets-for-baby-transfusion program. Obviously, significant experimental testing and FDA approval would be needed before any such platelet transfusion product could be tested in randomized trials. Until such work is done, it will not be clear whether platelets harvested from term umbilical cord blood would hold any advantage over adult donor platelets, for treating thrombocytopenic neonates who qualify for a platelet transfusion [53,54,55,56,57,58].

Fig. 3: Schematic representation of umbilical cord collection to produce a platelet transfusion product.
figure 3

After the birth of a healthy term neonate, blood remaining in the cord and placenta (after delayed cord clamping) is drawn by gravity-drainage using American Red Cross donor blood drawing kits. Collected cord blood bags are transported to the Red Cross Cord Blood Processing Laboratory where all FDA required testing and processing to platelet packs are conducted using approved protocols. Mini-bags suitable for transfusion are transferred to the hospital transfusion service for treatment of thrombocytopenic neonates enrolled in a study.

Full size image

Hematopoietic progenitors can be induced in vitro to differentiate into megakaryocyte progenitors and expanded as a potential source of platelets for transfusions [59,60,61,62,63,64,65,66]. Whether in vitro produced platelets would be a safer source of platelets for thrombocytopenic neonates will require considerable experimental and clinical trials effort.

For highly selected neonates with chronic prolonged thrombocytopenia, administration of a thrombopoietin receptor agonist (i.e., Romiplostim, Amgen, Inc.) could be advantageous to stimulate the production of platelets from megakaryocytes [67,68,69]. The FDA has approved this treatment for thrombocytopenic children as young as one year of age, but new studies would be needed to gain approval for infants. Case reports appear to show benefit among four neonates with severe and prolonged thrombocytopenia, each of whom had received multiple transfusions before the weekly subcutaneous dosing of Romiplostim was associated with rising platelet counts and cessation of platelet transfusion needs [14, 70,71,72]. One limitation of platelet agonists is that they take five days to begin to increase the platelet count, with peak responses observed two weeks after administration. Thus, their use would be reserved for neonates in whom we can predict, based on their underlying condition, a long duration of severe thrombocytopenia.

Conclusions

Platelet transfusions have a definite place in the practice of Neonatology and continue to be lifesaving interventions when given to the right patients. However, it is clear that our historical practice of transfusing non-bleeding neonates at high transfusion thresholds in the hope of preventing bleeding is wrong and can cause harm. Therefore, strategies are needed to implement the evidence-based restrictive guidelines informed by the PlaNet-2 Trial. Practice change, especially for neonatal platelet transfusions that have been a cornerstone of neonatal intensive care for decades, will require educational efforts at all levels, and collaboration with our colleagues who specialize in quality improvement/implementation science to ensure success.

Many questions about neonatal transfusion medicine remain unanswered and will require considerable collaborative effort to arrive at evidence-based answers. Here we will mention two issues that require evidence, but in the meantime perhaps we can come to a consensus and a consistent approach. One question is whether it is ever appropriate to administer a platelet transfusion to a neonate who has a platelet count over 25,000/µL. We judge that it probably is sometimes appropriate. A 25,000/µL threshold might be too low for patients on ECMO, patients undergoing lumbar puncture, patients immediately before, during or after surgery, or patients who very recently experienced a major bleed. For such patients perhaps a threshold of 50,000/µL is better. Until experimental evidence indicates otherwise, we could advocate for a 50,000/µL threshold for such patients (Table 1). Because nearly 40% of neonates enrolled in the PlaNeT-2 trial received one or more platelet transfusions before randomization, and because the study required a point-of-care ultrasound at the time of randomization to rule out a major IVH, there is also lingering uncertainty of the best transfusion threshold for neonates <27 weeks gestation in the first week of life; the highest risk period for IVH (See Table 1). A new trial aimed at addressing this question will be starting soon and we hope it will provide additional data to guide clinical practice in this high-risk population.

A second unanswered question is, it ever appropriate to administer a platelet transfusion to a neonate who has a platelet count over 100,000/µL? Probably not. In one of our recent studies, over five percent of platelet transfusions were given to neonates with counts >100,000/µL, yet in none could we find justification [23]. With the exception of actively bleeding patients, ordering a platelet transfusion when the platelet count is over 50,000/µL should not be considered best practice, unless future evidence indicates that a higher threshold is superior.

We hope this review conveys appreciation for the risks of liberal platelet transfusions to neonates, particularly the increased risk of death, major bleeding, NDI, and severe BPD after multiple platelet transfusions. We also hope our review of platelet transfusion refractoriness, and suggestions regarding how it can be mitigated without resorting to multiple platelet transfusions, will be useful toward the goal of reducing platelet transfusions. We judge that the most crucial step toward reducing platelet transfusions is changing the threshold for considering a prophylactic platelet transfusion, for most neonates, from 50,000/µL to 25,000/µL.

Because of the high rate of adverse effects following multiple platelet transfusions, and because of the subsequent financial and personal consequences to neonates, families, and society, we maintain that efforts to implement evidence-based platelet transfusion guidelines, and to identify safe and effective alternative therapies for thrombocytopenic neonates, have considerable merit. We encourage transfusion medicine specialists, basic and clinical researchers, hematologists, and neonatologists globally to support, advocate for, and actively engage in these efforts.