 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
In Reply.—
Drs Boedy and Mathew doubt that the association of restrictive-transfusion practice with adverse neurologic events found in our study is real. Such is their right, as it is the right and responsibility of every reader to form his or her own conclusions when presented with the methods and results of a clinical study.
We will address the most important of the many points raised in their letter. First, regarding the difference in grade 4 intraventricular hemorrhage (IVH) between groups, 4 of 49 infants in the restrictive-transfusion group had grade 4 hemorrhage compared with none of 51 infants in the liberal-transfusion group (P = .054, 2-tailed). These hemorrhages all developed by 9 days of age. Infants in the liberal-transfusion group were more likely to have received transfusions in the first week. We speculate that transfusion confers some protection from the most severe hemorrhages in the first week of life, possibly because preterm infants are often born hypovolemic, having been delivered quickly with little time for placental transfusion. Boedy and Mathew state that this difference is not statistically significant. Of course, this depends on one's definition of statistical significance. Level of significance is a continuum; using .05 as a break point is artificial. Whether P is .054 or .049, there is only a 5% probability that the difference in incidence of grade 4 hemorrhage occurred purely by chance.
Regarding the incidence of periventricular leukomalacia (PVL), 4 of 28 infants in the restrictive-transfusion group who had late head ultrasound examinations were found to have PVL, whereas none of 24 infants with late head ultrasounds in the liberal-transfusion group developed this complication (P = .115, 2-tailed). This means that there is only an 11% probability that the difference in incidence of PVL occurred by chance. In each of the 4 cases, the PVL was detected by 2 weeks of age, and in all cases it was already cystic in nature at 2 weeks or subsequently evolved to cystic PVL. In calculating the statistical significance of this difference, we made no assumption about the incidence of PVL in the infants who did not have late head ultrasound examinations. Had we assumed no PVL in any of the infants in either group who did not have late head ultrasounds, the significance level would have been .054. Moreover, the study was not powered to detect a difference in IVH or PVL. If the study had been done with a larger sample size, it may well have shown the differences in severe IVH and PVL to be more highly significant.
The analyses of IVH by grade and of PVL were planned before the study was begun. We did not plan to examine the composite outcome or grade 4 IVH of PVL until we saw that both were more common in the restrictive-transfusion group. We felt justified in combining these outcomes for statistical analysis, because they are the 2 most serious neurologic events that can be detected in the neonatal period.
The grade 4 hemorrhages were detected on days 4, 5, 7, and 9. The PVL was detected on days 4, 7, 14, and 14. Of the 6 infants, 3 had received 1 or no transfusion before the ultrasound abnormality was detected, and the other 3 had received >1 transfusion. The numbers were too small to permit analysis of the timing of transfusion relative to the onset of IVH or PVL. This analysis is also hindered by the fact that ultrasound examinations were not performed on a daily basis. In the liberal-transfusion group, 28 of 51 infants were received transfusions in the first week, most of them several times; the largest number of first-week transfusions was 6.
The infants with grade 4 IVH or PVL all had multiple apnea episodes beginning in the first week of life, but the numbers are too small to test for an association of apnea frequency or severity with adverse neurologic events.
We cannot rule out the possibility of 
1 unidentified confounders explaining the apparent excess of grade 4 IVH and PVL in the restrictive-transfusion group. However, the randomized clinical trial design minimizes the chance of this occurring.
Boedy and Mathew pose a number of questions in their penultimate paragraph. Again, with only 4 infants with PVL, we cannot meaningfully analyze the possible association with patent ductus arteriosus. Of the 4 infants with PVL, 3 had patent ductus arteriosus; all 3 received indomethacin, and 1 required surgical closure of the ductus arteriosus. Information on the use of inotropic drugs is not immediately available, but inotropic drugs are rarely given to young preterm infants in our NICU; the same was true during the years of this study. During these years, we routinely practiced permissive hypercapnia in our preterm patients; thus, few if any ever had carbon dioxide tension low enough to affect brain perfusion.
We take issue with the final remark of Boedy and Mathew, stating that we did a disservice to our readers by suggesting that use of restrictive-transfusion guidelines may have caused the increase in severe IVH and PVL seen in our patients. We choose to believe that we have performed a service for neonatologists and their patients by providing some guidance in an area in which none existed before: the choice of hematocrit or hemoglobin threshold used to trigger red blood cell transfusion. We believe readers would be well advised not to dismiss our findings but to carefully consider the possibility that the current widespread effort to avoid red blood cell transfusions in critically ill patients may be harmful.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2005 American Academy of Pediatrics. All rights reserved. |
||
|
||
