Interpreting risks of steroids for preterm infants

Abstract

Kopelman et al1Kopelman AE Moise AA Holbert D Hegemier SE A single very early dexamethasone dose improves respiratory and cardiovascular adaptation in preterm infants.J Pediatr. 1999; 135: 345-350Abstract Full Text Full Text PDF PubMed Scopus (53) Google Scholar are to be applauded for their contribution to neonatal science. In their prospective, blinded, placebo-controlled study of early (before 2 hours of life) dexamethasone (0.2 mg/kg) to improve pulmonary and cardiovascular adaptation, they carefully collect and report a large amount of data, albeit on a relatively small number of patients. Our issue is that we find the title and the conclusion of the abstract misleading. Both the title and abstract suggest that early administration of dexamethasone may have some benefit. The authors show that dexamethasone improves respiratory and cardiovascular adaptation as assessed by physiologic changes in gas exchange and blood pressure. What is left out of the title and abstract (and the authors clearly note in the discussion) is that dexamethasone was associated with a higher mortality rate (27% vs 15%, P =.23), more grade 3,4 intraventricular hemorrhages (22% vs 12%, P =.26) and more necrotizing enterocolitis (14% vs 9%). We suspect that these observations are given a lower priority because they are not statistically significant. We would argue that these observations are the key finding because of their clinical importance. To avoid any misunderstanding, we agree fully with the authors’ last paragraph and careful closing comments. We too would discourage routine early use of dexa-methasone. We would go further and suggest that the evolving data on “early routine steroid use” warrant a clinical alert against its use except under careful study.2Papile LA Tyson JE Stoll BJ et al.A multicenter trial of two dexamethasone regimens in ventilator-dependent premature infants.Pediatrics. 1999; 104: 22-27Crossref PubMed Scopus (84) Google Scholar, 3O’Shea TM Kothadia JM Klinepeter KL et al.Randomized placebo-controlled trial of a 42-day tapering course of dexamethasone to reduce the duration of ventilator dependency in very low birth weight infants: outcome of study participants at l-year adjusted age.Arch Dis Child Fetal Neonatal Ed. 1998; 79: F26-F33Crossref PubMed Scopus (123) Google Scholar, 4Yeh TF Lin YJ Huang CC et al.Early dexamethasone therapy in preterm infants: a follow-up study.Pediatrics. 1999; 104: 15-21Crossref PubMed Scopus (311) Google Scholar This study also emphasizes an important clinical truth. Improvements in physiology are not surrogate markers for improved health outcomes. It is easy to argue that reduced ventilator requirements and increased blood pressure are good things that should increase the likelihood that a neonate will go home healthy. The flaw in this reasoning is the failure to ask the question, “How was the physiology improved and what was the health consequence and cost of the intervention used to improve blood pressure and gas exchange?” If, even remotely, the health consequence of early use of steroids to improve physiology is death and brain injury, then the cost is too high. Remember a P value of.23 means there is a 77% chance that steroids increased the mortality rate, and although this does not meet the standard confidence of 95% certainty, it is mighty worrisome. Although we empathize with the authors’ approach and understand the limitations of a small clinical trial, we believe it is time for neonatologists to quit measuring and reporting studies of changes in physiology and look at more important outcomes. To that end, we have the following questions:Was intact survival evaluated (ie, were the 2 groups compared for discharge home without chronic lung disease, grade 3,4 intraventricular hemorrhage, or necrotizing enterocolitis)?Did the authors do any logistic modeling of their data to determine whether steroids had independent effect on mortality or grade 3,4 intraventricular hemorrhage? The small numbers make this hard, but we suggest a simple model including gestational age, treatment group, and sex to determine whether there are any hidden interactions that might be important in designs of future trials. 9/35/107616