study comparing 2 treatment outcomes than to compare a single treatment to placebo, which makes enrollment more difficult. Inclusion criteria and the consent process may result in a different set of baseline characteristics between the trial cohort and the general population. This was noted most recently during enrollment for the SUPPORT trial, which noted differences between enrolled and nonenrolled patients by receipt of prenatal antibiotics, antenatal steroids, delivery room interventions, and outcomes including mortality, bronchopulmonary dysplasia (BPD) and severe intraventricular hemorrhage.35,36 As a result, interpreting the study findings of higher mortality in the lower oxygen saturation target group has been difficult, because both groups had lower mortality than nonenrolled patients.21,37,38 Measuring the effect of an intervention versus a placebo may require changes to “standard care” by altering the timing of the primary intervention and its associated care. Placebo effects may change the outcome in either a treatment or a control group. The process of selecting measurable outcome criteria with a limited sample size may or may not result in trial results that reflect the key factors used by a clinician in weighing treatment options.39 One criticism of RCTs in CER is that enrolled patients may be a narrow group without comorbidities. In neonatology, clinical trials often enroll patients in a stratified fashion based on gestational age (for diseases of prematurity) rather than excluding patients with comorbid conditions. Gestational age or birth weight groups are often the clinically relevant subgroup analysis. This was demonstrated in initial comparisons of prophylactic surfactant versus early NCPAP, which suggested that infants born at earlier gestational ages treated with early NCPAP may have higher incidence of pneumothorax,40 which prompted more recent studies.24 Another potential limitation of RCTs for the purposes of CER is the site of care. Most patients in RCTs are treated in academic medical centers, and may have different baseline characteristics than patients treated in other settings. This may be less of an issue in neonatology than in other areas of medicine, such as primary care, because patients with high-risk conditions such as extreme prematurity or conditions requiring mechanical ventilation or surgical intervention are concentrated in level III/IV units. Because the enrollment criteria for many RCTs are based on gestational age, the study populations in academic and nonacademic centers are often comparable.33,41 For patients with uncommon conditions requiring multiple subspecialty care, the only (and therefore typical practice environment) is an academic center. Although most RCTs are conducted at academic-affiliated centers, the Vermont Oxford Network has conducted trials that enroll patients within private centers, such as comparisons of ventilatory and heat loss prevention strategies during resuscitation of preterm infants.30,42 For patients who do not require level III care, it would be of interest to compare outcomes for infants treated in lower acuity centers to be able to generalize results to the setting under which most of those infants are treated. As 836 Lagatta et al with other fields of medicine, most lower acuity centers do not have the infrastructure or the volume to support prospective clinical trials. Systematic Review and Metaanalysis Systematic reviews and metaanalyses are overlapping methods of evaluating existing evidence. Systematic reviews use prespecified search methods to evaluate and synthesize eligible studies on a specific clinical question. Metaanalysis refers to a quantitative re-analysis of pooled data from individual studies.43 These techniques allow results from multiple independent studies to be combined into a quantitative estimate of effect, such as combining results of multiple RCTs or epidemiologic studies. Neonatology has a strong history of systematic review, beginning with the Oxford Database of Perinatal Trials in the 1980s and continuing as the Cochrane Neonatal Group.44,45 Numerous systematic reviews and metaanalyses are updated through the Cochrane Neonatal Reviews, which serve different goals within the framework of CER.46,47 Many analyze direct comparisons of alternative potentially standard therapies, such as dopamine versus dobutamine for hypotensive preterm infants.48 In addition, the synthesis of multiple similar studies can increase power to detect a treatment effect when not all individual studies have found statistical significance. They are advantageous in studying rare or adverse events, and can also highlight effects in relevant subgroups, which may be too small in single RCTs. The Cochrane review of ibuprofen for the treatment of patent ductus arteriosus provides an example of several of these advantages. Treatment with ibuprofen versus indomethacin showed equivalent effectiveness in patent ductus arteriosus closure, but less risk of necrotizing enterocolitis in pooled estimates, although no single trial showed a significant difference.49 Although not strictly CER because it reviewed comparisons of an intervention with placebo, a significant example of neonatal research metaanalysis showing subgroup benefits was in the use of antenatal corticosteroids for prevention of respiratory distress syndrome. Crowley identified RDS reduction in