Parents and guardians of infants who had a platelet count of less than 100,000 per cubic millimeter were identified. Randomization occurred after written informed consent was received if the infant was receiving care in a participating neonatal intensive care unit and the following criteria were met: a gestational age at birth of less than 34 weeks, a platelet count of less than 50,000 per cubic millimeter, and cranial ultrasonography performed within 6 hours before randomization that did not show a major intraventricular hemorrhage. Exclusion criteria were a major or life-threatening congenital malformation, major bleeding within the previous 72 hours, fetal intracranial hemorrhage, immune thrombocytopenia, no administration of parenteral vitamin K, or a low probability of survival beyond several hours. Preterm infants with major bleeding became eligible for randomization 72 hours later provided there was no further major bleeding.
Infants were randomly assigned to receive a platelet transfusion (at a dose of 15 ml per kilogram of body weight) when the platelet count was less than 25,000 per cubic millimeter (the low-threshold group) or less than 50,000 per cubic millimeter (the high-threshold group). One Dutch trial site administered platelet hyperconcentrates in a dosage of 10,000 to 20,000 × 106 platelets per kilogram. Platelet products conformed to national standards in the United Kingdom, Ireland, and the Netherlands. The protocol (available with the full text of this article at NEJM.org) permitted additional platelet transfusions for clinically significant bleeding or surgery or invasive procedures. Treating clinicians and parents and guardians were aware of the treatment assignments, but neonatologists adjudicating the outcomes were unaware of these assignments.
The primary outcome was a composite of death or major bleeding up to and including day 28. Prespecified secondary outcomes were the following: survival up to day 28 after a major bleeding episode, death up to day 28, the rate and time from randomization to major bleeding up to day 28, at least one minor bleeding episode up to day 14, at least one moderate bleeding episode up to day 14, a major bleeding episode after red-cell transfusion, chronic lung disease (dependency on oxygen or respiratory support at >36 weeks of postmenstrual age) up to the end of the trial, stage 2 retinopathy of prematurity (unilateral or bilateral) up to 38 weeks of corrected gestational age, retinopathy of prematurity leading to laser or bevacizumab therapy up to 38 weeks of corrected gestational age, discharge by 38 weeks of corrected gestational age, the number of platelet-transfusion episodes per participant up to day 28, receipt of at least one platelet transfusion, the median platelet-count nadir before a major bleeding episode, the median platelet count closest to that of a major bleeding episode, a new sepsis event up to end of the trial, a new necrotizing enterocolitis event up to the end of the trial, a serious adverse event up to the end of the trial, and platelet transfusion–related adverse events up to the end of the trial. An additional secondary outcome was the neurodevelopmental outcome at 2 years; these data were not available as of this writing.
A bleeding-assessment tool was designed and validated for use in this trial.12 Grading of bleeding as minor, moderate, major, or severe was assigned centrally with the use of a computer algorithm and based on a modified version of the World Health Organization grading system used in other platelet-transfusion trials12-14 (see the Supplementary Appendix, available at NEJM.org). The outcome of “major bleeding” included intracranial hemorrhage (leading to neurosurgical intervention or radiologic imaging showing midline shift), intraventricular hemorrhage filling 50% or more of the cerebral ventricle, pulmonary hemorrhage (fresh bleeding through an endotracheal tube with increased ventilatory requirements), frank rectal bleeding, and severe bleeding (fatal bleeding, life-threatening bleeding associated with shock, or bleeding requiring fluid boluses or red-cell transfusion). Our definition of rectal bleeding was pragmatic; we defined any amount of fresh visible blood as rectal bleeding; we also performed a prespecified sensitivity analysis excluding rectal-only bleeding as a component of the primary outcome.
A bleeding-assessment form was completed daily for 14 days after randomization. Thereafter, data on infants who were not discharged or transferred were collected weekly. For infants transferred to another hospital before day 28, primary outcome data up to and including day 28, as a minimum, were documented. Reporting of safety outcomes was mandatory only at participating hospitals.
A Web-based service (www.sealedenvelope.com) assigned infants randomly in a 1:1 ratio. Minimization was used for the factors of intrauterine growth restriction and gestational age, with a 60% chance of simple random assignment. Intrauterine growth restriction was defined as birth weight in less than the 10th percentile in conjunction with an estimated fetal weight crossing percentiles downward during pregnancy, ultrasonographic evidence of uteroplacental insufficiency, or both.
This trial was undertaken in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. The protocol was approved by independent ethics committees in the United Kingdom, the Netherlands, and Ireland. An independent data and safety monitoring committee reviewed the interim data analysis and monitored patient safety in 6 monthly data review sessions. The first two authors take responsibility for the accuracy and completeness of data and the fidelity of the trial to the protocol. The statistical analysis plan is available with the protocol at NEJM.org.
Data on the frequency of bleeding outcomes in infants with severe thrombocytopenia were available from the observational Platelets for Neonatal Transfusion–Study 1 (PlaNeT-1),6 in which 30 of 169 infants (18%) died or had a major bleeding episode. However, that study included term infants, and therefore overall mortality and the incidence of new major bleeding episodes were expected to be higher in our trial. We estimated that the assignment of 329 infants to each group would provide 80% power (with the use of a two-sided test, at a 5% significance level), assuming an event rate for the primary outcome of 20% in the low-threshold group, to detect a difference of 8 percentage points (assuming that the event rate in the high-threshold group was 12%). This number of infants was rounded to a total of 660. At the time of the preplanned interim analysis after 87 infants were enrolled, the event rate in the low-threshold group was 18%, suggesting that the sample-size calculation was adequate.
All analyses were performed according to the intention-to-treat principle with adjustments for trial site as a random effect, gestational age, and intrauterine growth restriction. All tests were two-sided; a P value of less than 0.05 was considered to indicate statistical significance. SAS software, version 9.4 (SAS Institute), was used to conduct the analyses. All odds ratios and hazard ratios are presented as the high-threshold group as compared with the low-threshold group.
The primary outcome was analyzed with the use of a mixed logistic-regression model and was compared with the use of cranial imaging performed within the period from day 23 to day 38 (5 days before to 10 days after day 28), and details of any major bleeding were reported up to day 28. Missing primary outcomes were inferred by investigators who were unaware of the treatment assignments. For example, if an infant was discharged before day 28 and was not readmitted within the 28-day period, we assumed that no major bleeding had occurred (details are provided in the statistical analysis plan and the Supplementary Appendix).
Further prespecified sensitivity analyses were performed, including a per-protocol analysis, an analysis that excluded rectal bleeding, and an analysis that assessed sensitivity to missing primary outcome data. Binary outcomes were analyzed with the use of logistic regression, and count variables were analyzed with the use of negative binomial regression with an offset to account for the number of days of follow-up. Several secondary outcomes were analyzed with the use of Cox proportional-hazards regression to allow for differing numbers of days of follow-up. Data on infants who were transferred to nonparticipating hospitals and who had not had an event were censored at the time of transfer. The number of platelet transfusions administered to each group and platelet counts at which transfusions were administered were analyzed to assess adherence to the protocol.