As new therapies are integrated into practice, health care providers must continue to increase existing knowledge of the health and health problems of newborns. Providers must learn to overcome impatience with asking specific questions about the quality of evidence regarding risks and benefits of new practices. Clinical questions arise through practice and careful reading of research literature. Evidence-based practice begins with questions that can best be answered by the careful design and conduct of clinical trials.

It is not the purpose of this chapter to provide a detailed review of the various research designs that permit reliable scientific inference. Rather, our purpose is to promote the propositions that (1) challenge clinical observations and wisdom by subjecting them to systematic study and (2) encourage careful assessment and critique of research that supports or challenges the use of new and established clinical practices.

Clinical observations, although valuable in shaping research questions, are limited by selective perception—a desire to see a strategy work or fail to work. At times, a single case or case study may prompt us to question whether we should consider changing current practice. In some situations, much can be learned from carefully maintained databases. However, such knowledge is gained only when we have formed databases with clear intentions and have collected the necessary data.

Sinclair and Bracken⁴³ described four levels of clinical research used to evaluate safety and efficacy, based on their ability to provide an unbiased answer. In ascending order, these are (1) single case or case series reports without controls, (2) nonrandomized studies with historical controls, (3) nonrandomized studies with concurrent controls, and (4) randomized controlled trials (RCTs). For a variety of reasons, case-controlled and prospective studies with historic controls are sometimes the only feasible way to approach clinical questions. ^21,25RCTs have been identified as the strongest design for evaluating the effects of therapy. RCTs test hypotheses by using randomly assigned treatment and control groups of adequate size to examine the efficacy and safety of a new therapy. In theory, random assignment of the treatment balances unknown or unmeasured factors that might otherwise bias the outcome of the trial. A meta-analysis combines and reports the results of several trials. Tyson⁴⁷ has suggested criteria for identifying proven therapies in current literature (Box 1-1). Although conclusions drawn from systematic reviews of RCTs have been regarded as the strongest level of evidence, evidence from descriptive and qualitative studies should be factored into clinical decisions when either randomized trials are not available or additional information is sought related to personal experiences or perspectives on care. Evidence-based practice, using knowledge gained from qualitative studies, recognizes the expertise of individual clinicians and parents in evaluating health care provision. Price,³⁶using a qualitative methodology, explored the experience of parents with a child in the neonatal intensive care unit (NICU) and revealed how non-technical aspects of care, such as comforting infants after painful procedures, were as important to parents as the “technical aspects.” Wigert⁵¹ used a phenomenological hermeneutic design to elicit the experiences of having an infant in the NICU. Parents experienced a tension between exclusion and participation in their infant’s care with an emphasis on exclusion. In a qualitative study, Charchuk and Simpson¹² described a similar situation of parents experiencing a lack of disclosure of their infant’s condition and a lack of control. Raines³⁷ used a qualitative design to describe nurses’ expertise in the NICU. Through analysis, Raines described how neonatal nurses, after learning technical skills, could focus on parental needs.

Although RCTs are cited as providing the best evidence for guiding clinical decisions, they take time, and it is difficult to delay introduction of promising therapies. The pressure to intervene and change practice before studies are conducted or completed has many sources. Ian Chalmers, ⁹ in considering the struggle to gather scientific evidence in a climate of firmly held beliefs, has illustrated some of the pressures facing the scientific investigator (Figure 1-1). Bryce and Enkin⁸ discussed myths about RCTs that lead to reasons to not conduct them. One myth is that randomization is unethical. This might be true in the rare instance that an intervention is dramatically effective and lifesaving. The more common situation is one in which there is relatively poor evidence for both the current and the alternative strategy. To compare an established strategy whose benefit is not scientifically supported (even though widely acclaimed) with one that is openly “experimental” is not unethical; in fact, both are “experimental.” It is continued use of an unstudied practice that might more readily be labeled “unethical.”

Pressure to intervene is, however, often overpowering. Believing that an infant is in trouble, interventions occur through a cascade of interventions, ³² one leading to the next and each carrying risk. One of the most frequently cited examples is the epidemic of blindness associated with use of oxygen in newborns. ^40,41 Oxygen, used since the early 1900s for resuscitation and treatment of cyanotic episodes, was noted in the 1940s to “correct” periodic breathing in premature infants. After World War II and introduction of new gas-tight incubators, an epidemic of blindness occurred, resulting from retrolental fibroplasia (RLF). Silverman pointed out that although many causes were suspected, it was not until 1954 that a multicenter, controlled trial confirmed the association between high oxygen concentrations and RLF. ⁴⁰ Frequently forgotten, however, is that in subsequent years, mortality was increased in infants cared for with an equally experimental regimen of strict restriction of oxygen administration and many survivors had spastic diplegia. In the 1960s, introduction of microtechniques for measuring arterial oxygen tension permitted better monitoring of oxygen therapy, with a reduction in mortality, spastic diplegia, and RLF, now called retinopathy of prematurity (ROP). ROP is currently limited to extremely low-birth-weight (ELBW) infants; ⁴⁰ research continues to explore causes, preventive measures, and treatments (see Chapter 23).

The desire to see an intervention “work” encourages practitioners and investigators to seek early signs of benefit. Long-term effects are frequently overlooked. One reason is that they may not be foreseen. Consider the example of diethylstilbestrol (DES). DES administration to pregnant women was introduced in 1947 without clinical trials to prevent miscarriage, fetal death, and preterm delivery. ^8,22 It was thought to be effective after uncontrolled studies despite controlled trials summarized in an overview (meta-analysis) by Goldstein et al²³ (Table 1-1) that showed the opposite. Clearly, DES was not effective, but it continued to be used until the 1970s, when the Food and Drug Administration (FDA) finally disapproved its use. The unforeseen result was that female children born to mothers who were given DES had structural abnormalities of the genital tract, pregnancy complications, decreased fertility, and an increased risk for vaginal adenocarcinoma in young women. Male children had epididymal cysts. The Centers for Disease Control and Prevention (CDC) has even considered the possibility of effects in grandchildren of exposed women. ¹⁷ This is not the only example of physicians continuing to use therapies that have been shown in RCTs to be of no benefit. ¹¹

The costs of long-term studies and follow-up surveillance are numerous. However, when effects are measurable later in life (e.g., psychological problems, ability to function in school), cost should not determine study design. Even when randomized trials are conclusive, unanswered questions remain: Will a technology or treatment have the same effect in all settings? Has an “appropriate” target population been selected? Are there long-term unforeseeable consequences?

The major cause of death in premature infants is respiratory failure from respiratory distress syndrome (RDS) (see Chapter 23). Previously called hyaline membrane disease (HMD), this syndrome of expiratory grunting, nasal flaring, chest wall retractions, and cyanosis unresponsive to high oxygen concentrations was a mystery until the 1950s. ⁴¹

The evaluation of various therapies for RDS contrasts the value of controlled and uncontrolled trials. Sinclair⁴² noted that uncontrolled studies were more likely to show benefit than controlled trials. In 19 uncontrolled studies, 17 popular therapies showed “benefit.” In 18 controlled studies, only 9 demonstrated benefit. An untrained reviewer of the research might base clinical practice on faulty conclusions of uncontrolled trials

Evidence-based medicine (EBM) has been defined as the “conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients”³⁵ Examples from the literature, such as those cited in the preceding sections, illustrate how the application of the principles of evidence-based medicine offers a strong argument countering those who assert that EBM is nothing more than “typical practice using good clinical judgment.” Proponents of EBM argue that the principal four steps of evidence-based practice—formulating a clinical question, retrieving relevant information, critically appraising the relevant information, and applying the evidence to patient care—provide a foundation for practice that leads to improved newborn outcomes and avoidance of repeating medical disasters. It is interesting to note that the roots of the EBM movement can be found in perinatal medicine.

Believing that the results of perinatal controlled trials had to be summarized in a manner useful to practitioners, Chalmers et al¹¹ and other perinatal professionals from various countries developed a registry of RCTs. They reviewed a vast amount of literature from published trials, sought out unpublished trials, and encouraged those who had begun, but not completed, studies to make them known to the registry. Once gathered, the studies’ findings were summarized in “overviews.”

A systematic review (or meta-analysis) pools the results of independently conducted RCTs whose study methods are reasonably similar, both in the selection and characteristics of participants and in the treatments that are offered. The results produce unbiased estimates of the effect of an intervention on clinical outcomes and are distinguished from nonsystematic reviews in which author opinions often are reported along with the evidence. Tables 1-1 and 1-2 were developed after pooling the results of different studies.

From these systematic reviews, practitioners can learn the strengths or weakness of clinical trials and evaluate the claims of benefit for implementing a strategy. The result of the efforts of Chalmers et al was the 1989 publication of a remarkably useful book, Effective Care in Pregnancy and Childbirth.¹¹ At the end of their book, they reported their own views of the reviewed treatments based on conclusions formed in the preceding articles. They found that although some strategies and forms of care were useful, others were questionable. Some interventions believed to be useful were not useful, of little benefit, or in fact, even harmful. In 1991, a companion publication, Effective Care of the Newborn Infant,⁴³ compiled and reviewed neonatal RCTs.