Infanticide has a long and infamous history in many cultures but it has been practiced for social, cultural, economic, and criminal reasons more than for medical ones.³ There is an important distinction between infanticide and withholding medical treatment to allow a seriously ill infant to die. Infanticide is the purposeful killing or abandoning of an infant, usually with malicious intent, in which the neonate’s state of health is irrelevant because the act alone is sufficient to produce death. By contrast, when LST is withdrawn or withheld from a severely ill neonate, the underlying medical condition alone is sufficient to cause death. Rather than being killed, infants are permitted to die of their underlying congenital disorders, which are terminal illnesses in the absence of LST. This accepted medical practice is not classified as infanticide or active euthanasia.⁴

In contemporary neonatology, the practice of selective treatment of neurologically damaged and other severely malformed neonates was highlighted in the early 1970s by the provocative reports of John Lorber in England and Raymond Duff and A.G.M. Campbell in the United States. Selective treatment was practiced earlier but not explicitly reported.⁵ In a 1971 article on treatment efficacy in infants with meningomyelocele, Lorber recommended that physicians select the most severely affected babies and allow them to die by purposely not treating them.⁶ In 1973, Duff and Campbell reported a consecutive series of 299 neonatal deaths in the nursery at the Yale-New Haven Medical Center in which 14% (43 infants) died because they were purposefully not treated.⁷ They explained that the treatment decisions were made by physicians and parents because the baby’s prognosis “for meaningful life was extremely poor or hopeless.” The diagnoses of the 43 infants who were purposely permitted to die, in order of frequency, were: multiple anomalies, trisomy, cardiopulmonary disease, meningomyelocele, other central nervous system (CNS) disorders, and short bowel syndrome. The following year, the Jesuit theologian Richard McCormick published an influential article in JAMA identifying and analyzing the moral issues in these cases and emphasizing the need to establish treatment guidelines.⁸

By the 1980s, the percentages of neonates in NICUs who died as a result of selective treatment were even higher. A figure of 30% was quoted at Hammersmith Hospital in London⁹ and a figure of 21% was reported from Aberdeen Maternity Hospital in Scotland.¹⁰ In a NICU study from the University of California at San Francisco during 1989-1992, 73% of the 165 infant deaths resulted from withholding or withdrawal of LST based on a poor prognosis. Diagnoses of the neonates were: intracerebral hemorrhage (39%), necrotizing enterocolitis (29%), extremely low birth weight (28%), congenital abnormalities (24%), and respiratory failure (17%).¹¹

Over the past 20 years, there has been a striking evolution of neonatology practices away from selective treatment.¹² In the 1970s, neonates with Down syndrome commonly did not receive standard medical and surgical treatments for curable complications.¹³ In 1983, surgery for meningomyelocele was withheld from half of affected infants routinely.¹⁴ But by the 1990s, as a result of federal laws in the United States, such as the “Baby Doe” regulations which mandated treatment and the use of infant care review committees, normative practices had evolved to aggressively treat all but the most severely neurologically damaged, exceedingly premature, and incipiently dying neonates. In summarizing this remarkable practice transformation, Norman Fost observed “A prolonged history of what is now
perceived as serious undertreatment of infants with reasonable prospects for living a meaningful life was replaced by an era of serious overtreatment. One form of child abuse, neglect, was replaced by a form of medical battering. In both cases, the interests of the patient seemed to be a casualty.”¹⁵

The resuscitation and aggressive treatment of neonates born with extreme prematurity has created an additional and growing ethical controversy in contemporary American neonatology. It is now commonplace for infants born at 23 to 25 weeks of gestation—and even some at 22 weeks—to receive aggressive support in NICUs, often at the insistence of their parents. Media publicity of cases highlighting the successful treatment of some of these babies (often without also noting the frequent treatment failures and subsequent disabilities) has led to high expectations from parents and to disputes over treatment with neonatologists.¹⁶ There are also poignant cases in which parents refuse resuscitation and aggressive treatment that has been advocated by neonatologists.¹⁷ These disputes between neonatologists and parents over the advisability of neonatal resuscitation have been called the tension between “medical authority and parental autonomy.”¹⁸

Neonatal resuscitation and intensive treatment have advanced technically, but the appropriate application of these advances remains to be determined because of the statistical uncertainty about prognosis. Ethical questions arising in the care of extremely premature neonates now are a more common problem that those arising in the care of congenitally malformed neonates. The American Academy of Pediatrics (AAP) Committee on Fetus and Newborn recently published guidelines for neonatologists. The AAP acknowledges the important role of parents in decision making but states that “the physician’s first responsibility is to the patient.” They further advise that “The physician is not obligated to provide inappropriate treatment or to withhold beneficial treatment at the request of the parents. Treatment that is harmful, of no benefit, or futile and merely prolonging dying should be considered inappropriate. The physician must ensure that the chosen treatment … is consistent with the best interest of the infant.”¹⁹

Premature babies commonly are classified into three groups: (1) low-birth-weight (LBW) neonates weighing less than 2,500 grams; (2) very-low-birth-weight (VLBW) neonates weighing less than 1,500 grams; and (3) extremely-low-birth-weight (ELBW) neonates weighing less than 750 grams.²⁰ Advances in NICU technology producing successful outcomes have led neonatologists to treat LBW and VLBW babies routinely. The most vexing ethical issue is whether to treat infants in the ELBW category aggressively or allow them to die.

There is a growing evidence base recording the neurological and developmental outcomes of infants after extremely preterm births. In one study, approximately half of children born at 25 or fewer weeks of gestation have disabilities when tested at 30 months of age, and approximately half of those disabilities are severe.²¹ A study of the outcomes at age 6 of extremely preterm births showed that survivors had a 86% rate of moderate-to-severe disability.²² Another study showed an 86% mortality rate of infants weighing under 600 grams and a 44% mortality rate of infants weighing 600 to 750 grams. Over two-thirds of surviving infants tested at 30 months had impaired mental development.²³ In a study of over 4,000 ELBW infants of birth weight 401 to 500 grams, only 17% survived to discharge and most had serious comorbidities.²⁴ These data should be shared and discussed with parents when they consider consenting for neonatal resuscitation and other aggressive treatment for ELBW babies.²⁵ Jonathan Muraskas and colleagues produced an algorithm for decision making in ELBW patients, considering gestational age, birth weight, and parental decision.²⁶

As is true with clinical decision making in older patients, a clear diagnosis and prognosis are prerequisites for treatment decisions in neonates. Robert Weir and James Bale believe that neonates born with severe neurological
and other profound congenital anomalies can be grouped into three general diagnostic and prognostic categories on the basis of the severity of their anomalies and their prognoses. This categorization by diagnosis and prognosis permits an ethical analysis of similar cases within each group and establishes a procedure for decision making in novel cases.²⁷

The first category consists of those neonates with lethal anomalies so profound that it is beyond the best efforts of physicians and current technology to sustain their lives. Neonates in the second category have profound anomalies, but some may survive with aggressive treatment. Infants whose disorders are serious but not profound and whose lives generally can be saved with treatment comprise the third category.²⁸

In a recent article, Tracy Koogler and colleagues raised an important caveat: the commonly used prognostic term “lethal anomalies” is misleading. Many obstetricians, neonatologists, and geneticists use the term “lethal anomalies” to describe infants with profound neurological compromise, structural abnormalities, or functional disabilities (such as anencephaly, trisomy 13, and trisomy 18), that, if untreated, cause death within a few months, The authors pointed out that what makes these conditions lethal is a medical decision not to treat them aggressively in light of their poor neurological prognosis.²⁹ This caveat should be kept in mind in the following analysis by Robert Weir and James Bale.

According to the Weir and Bale argument, clinicians have no ethical duty to attempt to treat infants in the first group because they cannot be helped by any therapy. For example, anencephalics do not survive beyond infancy, irrespective of treatment.³⁰ The current debate about selective treatment principally involves the second group. The survival data on infants in group two are more ambiguous than that for those in group one, but, generally, group two infants have a poor prognosis for survival or cognitive life. Nearly all neonatologists believe that infants in group three should be treated aggressively because their prognosis for survival is good and their prognosis for developing the capacity for intellectual life is at least fair.

The diagnoses in infants in group one (disorders so profound that prolonged survival is impossible or highly improbable) include anencephaly, multiple severe congenital anomalies requiring repetitive efforts to resuscitate, and profound prematurity under 22 weeks of gestation. Even the most conservative, “protreatment” neonatologists generally do not advocate treatment of profoundly malformed or extremely premature neonates because treatment attempts are doomed to failure.³¹

The second group of neonatal neurological disorders that many believe justifies selective nontreatment includes hydranencephaly, schizencephaly, holoprosencephaly, trisomy 13, trisomy 18, other severe chromosomal disorders (such as triploidy syndrome and 4p-syndrome), infantile Werdnig-Hoffman disease, X-linked myotubular myopathy, severe perinatal trauma, high myelomeningocele coupled with other major congenital abnormalities, severe high-cervical spinal cord injury with ventilator dependency, and large vein of Galen malformation associated with neonatal congestive heart failure. Also in the second group are disorders that present in infancy but not in the neonatal period: Canavan’s disease, Alexander’s disease, metabolic-storage diseases (such as Lesch-Nyan syndrome, Tay-Sachs disease, Menkes’ disease, and Krabbe’s disease), and infantile ceroid lipofuscinosis. The neurological conditions classified in the second group account for less than 1% of all births.³² Premature babies of 22 to 25 weeks gestation may be added to this list because some can survive with aggressive treatment.³³

The third group for whom aggressive treatment routinely should be ordered represents the majority of neurologically imperfect newborns. These neonates usually survive in a state of partial disability. Disorders in this group include prematurity (but greater than 25 weeks of gestation), hydrocephalus, most cases of meningomyelocele, most cases of neonatal intraventricular hemorrhage, trisomy 21, many metabolic conditions, such as phenylketonuria, maple syrup urine disease, organic acidemias, and congenital hypothyroidism, and other major congenital central nervous system malformations, such as porencephaly, agenesis of
the corpus callosum, and neuronal migration disorders.³⁴

Neonatologists first should make the correct diagnosis and assign neurologically impaired infants to the correct group. Usually, this task can be accomplished successfully. Generally, making the diagnosis is easier than making the prognosis with an equivalent level of certainty. Clinical inspection and neuroimaging procedures can diagnose anencephaly and major cranial and spinal malformations with a high degree of certainty. Many genetic disorders can be diagnosed confidently with chromosomal or DNA analysis. Many metabolic disorders can be diagnosed with a high level of certainly by chemical tests.

Clinicians can confidently state the outcome for survival and neurological functioning only for the first group of disorders. In the second and third groups, an inescapable degree of uncertainty exists because of the probabilistic nature of studies on the outcomes of infants with these disorders who were treated and those who were not. A statistical prognosis can be based on clinical findings, such as diagnosis, severity of illness, and presence of comorbid medical problems. This prognosis will be accurate when the outcomes of sizable numbers of similarly affected infants are averaged. However, because an individual case may fall anywhere within the statistical range, only a probabilistic statement can be issued concerning the prognosis in an individual case.³⁵ Wise neonatologists have adopted the position of “prognostic humility.”

Neonatologists have responded to this irreducible uncertainty by adopting one of three general decision-making strategies. The first strategy is a “statistical” approach: generate practice criteria purely from statistics on outcome data that are applied to every subsequent case. For example, a statistical approach would rule that no infant born before 22 weeks gestation would be treated, because outcome data show that treatment is futile in this group.³⁶ The second strategy is a “wait until near certainty” approach in which all severely ill neonates in categories two and three are treated aggressively until the prognosis becomes certain. This practice is becoming increasingly common in the United States. The third strategy is an “individualized approach” to initiate treatment on every infant in categories two and three and then to permit the parents the option to terminate treatment before the prognosis has become definite. This approach is more common in Great Britain.³⁷

Knotty problems result from each approach. The statistical approach assumes that the outcome data on which it is based are sufficiently robust to justify the application of binding treatment rules. Although some outcome data achieve this level of clarity, other data show significant rates of false-negative and false-positive errors. With false-positive errors, viable neonates will be permitted to die unjustifiably. With false-negative errors, nonviable neonates will be treated aggressively, causing unjustified suffering to the patient and family.

One logistical problem limiting the usefulness of outcome studies is that to achieve statistical validity, outcome studies generally require that infants in each class receive uniform treatment. While such a study would comprise good science, it is unrealistic in practice because uniform treatment may be inappropriate in many cases. For example, measuring the outcomes of infants with selected fetal defects solely by seeing how long they live assumes that each infant is treated with uniform aggressiveness. Often, this assumption is erroneous because parents and physicians may feel that aggressive medical treatment is contraindicated in a particular case. Systematically incorporating these outcome data into studies as if they represented the natural history of the treated disease may be misleading and may produce the fallacy of the self-fulfilling prophecy. (See chapter 10.)

Irreducible uncertainty is another limitation of outcome studies. Alan Shewmon has demonstrated the mathematical limits to the certainty of a prognosis of “irreversible” brain damage based on experimental data from outcome studies. The irreducible degree of uncertainty results from the relatively small number of patients available for recruitment into studies from even large institutions or into inter-institutional cooperative outcome studies. When these studies are published, the investigators should also report the statistical confidence intervals to acknowledge the degree
of uncertainty resulting from a small sample size.³⁸

For example, John Lantos and colleagues studied the outcome of cardiopulmonary resuscitation (CPR) performed on VLBW babies (below 1,500 grams). They found that none of the babies who underwent CPR in the first three days of life survived, and only 36% of the babies who underwent CPR thereafter survived, of whom three-fourths had residual neurological deficits. They concluded that CPR was futile therapy in this group.³⁹ However, the number of patients they were able to study was necessarily small because of the rarity of performing CPR on VLBW babies. In fact, the reported “75% of patients” with residual neurological deficits actually represented three infants of the four infants studied. The confidence intervals were therefore necessarily very low given the small number. Anyone who attempts to develop practice guidelines from these data must take into account the statistical uncertainty resulting from the necessarily small numbers of patients studied.

The clinician who adopts the wait-until-near-certainty approach inevitably will treat some neonates who should have been permitted to die earlier. Treatment of these nonviable neonates causes them and their families to suffer, places an expensive burden on society, and postpones but does not obviate the need to terminate treatment. In some cases, this approach simply represents unwillingness by the physician or parents to make a psychologically difficult decision to terminate treatment in a timely manner.

The individualized approach permits frequent reassessments and allows the parents to see that the baby will not survive or develop the capacity for any more than a minimally cognitive existence. With this approach, both overtreatment and undertreatment errors can result, but both these errors are less likely to result than errors resulting from the first two approaches.⁴⁰ The American Medical Association Council on Ethical and Judicial Affairs opined: “It is not necessary to attain absolute or near-absolute certainty before life-sustaining treatment is withdrawn, since this goal is often unattainable and risks unnecessarily prolonging the infant’s suffering.”⁴¹

Cultural factors mold attitudes and influence behaviors of neonatologists when they decide which infants should be treated. In the EURONIC Study Group (European Project on Parents’ Information and Ethical Decision Making in Neonatal Intensive Care Units), neonatologists from 10 European countries were surveyed about their attitudes and self-reported practices. Neonatologists’ attitudes and practices varied by nationality, especially when comparing the question of preservation of life at any cost versus concerns about the infants’ quality of life. Self-reported practices of aggressively treating versus allowing to die followed the reported attitudes.⁴² In other published reports from this study, neonatologists’ practices correlated positively to the laws of their countries of origin,⁴³ the extent of participation of parents in the decision making,⁴⁴ and the religious and moral beliefs of physicians in different countries.⁴⁵

The principal ethical issues involved in withholding or withdrawing LST of neurologically ill neonates may be formulated as three questions: (1) What is the right decision? (2) Who should make the decision? and (3) By what standards should the decision be made?