Individual Patient Meta-analysis in Pediatrics

In this issue of Pediatrics, Askie et al1,2 present their report of an individual patient data (IPD) meta-analysis of randomized controlled trials of inhaled nitric oxide (iNO) in preterm infants. This article is of great interest regarding what it tells us about both IPD meta-analysis and the use of iNO in preterm infants.By now, everyone in the pediatric community is familiar with the idea of systematic reviews and meta-analyses. Most meta-analysis reports that pediatricians would have read involve aggregate data analyzed at the trials level. Meta-analyses of aggregate data are typical of Cochrane reviews. Data from these types of analyses enable us to gain greater precision regarding clinically important outcomes. Aggregate-data meta-analyses at the trials level have been used in creating policies and guidelines. Systematic reviews and meta-analyses of antenatal3 and postnatal4,5 corticosteroids have greatly influenced practice in neonatal-perinatal medicine. However, there are limitations to this analytic approach. There are obvious differences between the studies included in a meta-analysis, from both a clinical and methodologic perspective.6 Clinical heterogeneity can occur at the trial level (which applies to all of the patients in each individual trial) or at the patient level (where the characteristic varies both within and between trials).7 Typical aggregate-data meta-analyses can explore heterogeneity at the trial level through subgroup analyses or meta-regression but cannot adequately address the heterogeneity that occurs at the patient level within the trial. Examining individual patient data can help researchers avoid this pitfall in meta-analysis. IPD meta-analysis is based on a re-analysis of treatment effects and potential confounders using data from each patient.7IPD meta-analysis is a demanding exercise that requires active collaboration of all of the original investigators and great organizational skill. Askie et al1,2 performed IPD meta-analyses on 3298 infants from 11 of the 13 trials. They found no statistically significant effect of iNO on death or chronic lung disease (relative risk: 0.96 [95% confidence interval: 0.92–1.01]) or severe neurologic events or imaging (relative risk: 1.12 [95% confidence interval: 0.98–1.28]). There were no statistically significant differences in iNO effect according to any of the patient-level characteristics tested, including gestational age, oxygenation index, evidence of pulmonary hypertension, and mode of ventilation. These results are consistent with the trial-level meta-analysis conducted by Barrington and Finer.8 Many people will be disappointed in this result. iNO has been shown to be effective in term infants with persistent pulmonary hypertension and has shown great promise in premature infants.9 Studies in a variety of animal models that have addressed the effects and mechanisms of iNO in lung development and lung injury have noted a decrease in airway resistance, which translates into a decreased need for supplemental oxygen and ventilator support, as well as normalization of alveolarization.10 In the trial-level Cochrane review of iNO in preterm infants, the heterogeneous nature of the studies was noted and led to the analysis of trials based on 3 subgroups: trials that attempted to prevent chronic lung disease in at-risk infants; trials that treated preterm infants with severe hypoxic respiratory failure in the first few days of life; and trials that treated infants later in the course of their respiratory distress as it evolved to early chronic lung disease.9 Authors of the trial-level Cochrane review noted a trend toward a reduction in death or chronic lung disease at 36 weeks' postmenstrual age in studies with routine use of iNO in intubated preterm infants. However, trials of early treatment of infants based on oxygenation criteria or of later treatment based on the risk of chronic lung disease did not find significant benefits of iNO on the primary end point of death or chronic lung disease at 36 weeks' postmenstrual age. This meta-analysis has been criticized because of its significant clinical heterogeneity, and it was hoped that a study such as the IPD meta-analysis conducted by Askie et al might help clarify the possible effects of iNO. However, no new signal has emerged to make us believe that we understand under which conditions iNO will be effective in preterm infants. The conclusion of the National Institutes of Health Consensus Development Conference that "the available evidence does not support use of inhaled nitric oxide in early routine, early rescue, or later rescue regimens in the care of premature infants less than 34 weeks gestation" stands.11