*Correspondence: Professor R. M. Ward, Department of Pediatrics, University of Utah, University Medical Center 2A122, 50N Medical Drive, Salt Lake City, UT 84132, USA.
Improvements in neonatal intensive care during the last 20 years have increased the survival of the most immature newborns at 23 weeks from 0% to 65% at some centres, although rates vary widely among neonatal care centres. University of Utah, USA data show that each week in utero after week 23 raises survival by 6–9%, to 90% by 27–28 weeks and 95% by 33 weeks. Provision of care in specialised centres to provide high-risk obstetric and neonatal intensive care, prenatal treatment with corticosteroids, postnatal treatment with surfactant and nitric oxide, and improvements in respirators and equipment to care for extremely immature infants all contribute to these changes. The increased rate of survival for extremely premature newborns has not been accompanied by an increased rate of severe intraventricular haemorrhage or neurological impairment, such as cerebral palsy. Regardless, intraventricular haemorrhage remains a significant problem, especially if associated with post-haemorrhagic hydrocephalus, leading to long-term neurological impairment and decreased survival. Necrotising enterocolitis (NEC) is more common in premature than in term newborns and is the most frequent cause of short bowel syndrome in infancy. Survival after surgery for NEC has improved during the last two decades, but complications of nutritional support produce many long-term problems. Retinopathy of prematurity (ROP) remains a frequent cause of neurosensory impairment for extremely premature newborns. Laser photocoagulation for advanced ROP is more effective than cryotherapy for preventing retinal detachment and improving visual outcomes. Despite prenatal corticosteroid treatment and postnatal surfactant administration, many extremely premature newborns still develop bronchopulmonary dysplasia. Abnormal pulmonary function may persist into adulthood, but newer ventilators and management schemes appear to be reducing this long-term morbidity. Many changes in neonatal care occur each year, but carefully controlled outcome studies are needed to evaluate the effectiveness of these newer styles of neonatal intensive care.
Survival for extremely premature infants has increased significantly during the last two decades. Complications of prematurity are becoming more common as more survivors are spending time in newborn intensive care units (NICUs). Most premature infants born at <32 weeks gestation will remain in the NICU until close to term to allow for sufficient organ maturation so that the infant can be cared for independent of intensive care. Immaturity of multiple organ systems places them at high risk for a variety of complications during these prolonged hospital stays. A comprehensive discussion of all potential complications of preterm birth is beyond the scope of the paper, but some of the more frequent and significant ones are listed in Table 1. Specific complications (CNS haemorrhage and/or ischaemia, necrotising entercolitis (NEC), chronic lung disease, and retinopathy of prematurity [ROP]) were selected, as they have a significant impact on long-term development and outcome.
Table 1. Selected neonatal complications following preterm birth.
• Catheter complications
• Chronic lung disease
• Developmental delay
• Growth reduction
• Hearing impairment
• Intraventricular haemorrhage
• Necrotising enterocolitis ± perforation
• Neonatal abstinence
• Nosocomial infections
• Patent ductus arteriosus
• Periventricular leucomalacia
• Pulmonary barotraumas
• Respiratory distress syndrome
• Retinopathy of prematurity
Databases of Neonatal Outcomes
Several large collaborative networks have been established to evaluate immediate and long-term outcomes of premature infants. Although gestational age is regarded as the best predictor of maturation, physiological function and subsequent complications, it cannot be measured as accurately as birthweight. Thus, most networks report data in terms of birthweight, which is easily determined but does not take into account infants who are large or small for gestational age. To illustrate this, the National Institute of Child Health and Development (NICHD) Neonatal Network classified infants according to weight by 100 g increments. The average gestational age was the same (25 weeks) for infants in birthweight categories 401–500, 501–600, or 601–700 g at birth1. The Vermont-Oxford Network includes data from 352 NICUs in 17 countries2. In 2000, they summarised data from 29,177 newborns with birthweight <1500 g in the Vermont-Oxford Neonatal Network, and those data will be cited for the incidence of specific complications.
Increased survival for extremely premature newborns during the last two decades is illustrated with data from the University of Utah, USA in five-year intervals from 1981 to 2000. Survival increases with each completed week in utero for newborns delivering at gestations of 22 to 28 weeks (Fig. 1). At 23 weeks, survival increased from 0 to almost 50% while two weeks later, at 25 completed weeks, survival rose from approximately 20% to nearly 90%. Factors contributing to this improvement are multiple, but are thought to reflect increased delivery of premature newborns in perinatal care centres specialising in high-risk obstetric care and neonatal intensive care, increased treatment with corticosteroids prenatally and surfactant postnatally, and improvements in the equipment for provision of neonatal intensive care, such as incubators, overhead warmers, and respirators.
Central Nervous System Complications
Several brain lesions are associated with adverse neonatal outcomes: intraventricular haemorrhage (IVH), periventricular haemorrhagic infarction (PVHI), and periventricular leucomalacia (PVL). Data from the Vermont Oxford Network database for 2000 show that the incidence of IVH and PVL are inversely related to gestational age (Fig. 2). Intraventricular haemorrhage is thought to originate in the microcirculation of the germinal matrix and is graded 1–4 based on the extent of haemorrhage as seen on ultrasound3. Altered cerebral blood flow secondary to poor cerebral autoregulation or systemic hypo- or hypertension, platelet and coagulation disturbances as seen in infection, and decreased capillary integrity and vascular support have been implicated in the pathogenesis of IVH3. In grade I IVH, the haemorrhage is confined to the germinal matrix. In grade II IVH, blood is present in the germinal matrix and a small amount of blood is also present in the ventricles. Grade III IVH occurs when the ventricles are >50% filled with blood and dilated. About 15% of infants with grades I–III IVH progress on to develop grade IV IVH or PVHI. In grade IV IVH or PVHI, blood extends into the brain parenchyma, due to venous congestion of the terminal veins that border the lateral ventricles, which leads to white matter necrosis4. These areas of necrosis are often large, asymmetric, and increase with increasing severity of IVH. Grade III IVH and PVHI are associated with spastic haemiplegia, with the lower extremities more affected than the upper extremities due to the proximity of the descending motor fibres, and they may also affect intellect. Smaller haemorrhages, grades I and II IVH, are not associated with an increase in abnormalities compared with infants without an intraventricular haemorrhage. From the early 1980's to 1993, the frequency of severe IVH has decreased from 14.9% to 10.5% in the United Kingdom5 and from 30% to 8% in Canada6.
Hydrocephalus develops in approximately one quarter to one third of infants with IVH, and one third of these infants require the placement of a shunt to provide an alternate route for CSF reabsorption7. Infants with post-haemorrhagic hydrocephalus who require a shunt have a higher mortality and worse developmental outcome. Worse outcomes correlated with increased severity of the IVH.
Periventricular leucomalacia (PVL) is necrosis of the white matter due to arterial ischaemia in the cerebral artery watershed area as opposed to venous congestion seen in PVHI4. Periventricular leucomalacia is expressed as either focal periventricular necrosis with subsequent development of cysts or diffuse cerebral white matter injury, which is more prevalent in extremely premature infants. Clinically, infants with PVL usually exhibit spastic diplegia, and this lesion is often symmetrical. Intellect and cognitive function are frequently affected due to interference with normal cerebral cortical organisation. Cysts may be too small to be visualised by ultrasound. In addition, the timing of the cranial ultrasound is critical, as cysts develop 3–4 weeks after the ischaemic insult and often involute with time. Thus, long-term deficits may be due to unrecognised PVL in infants with an apparently ‘normal’ cranial ultrasound4.
Necrotising enterocolitis is an inflammatory disorder of the gastrointestinal tract that occurs 100 times more frequently in premature than in term neonates8. The intestinal lesions of NEC are characterised by coagulation necrosis ranging from superficial mucosal erosions accompanied by eosinophilic infiltrations to transmural bowel necrosis with perforation9. Necrotising enterocolitis is thought to develop following injury to the gastrointestinal tract followed by bacterial invasion frequently starting in the terminal ileum, but ranging from the stomach to the colon. Invading bacteria produce fermentation gases, such as hydrogen, methane and carbon dioxide, which create intramural gas collections known as pneumatosis intestinalis, the pathognomonic radiographical sign of NEC. These intramural gas collections are often visible on inspection of resected bowel and represent the pathological correlate of the radiographical finding of pneumatosis intestinalis.
To improve comparison of outcomes and management, NEC has been graded in severity by Walsh and Kliegman: stage 1, suspected NEC; stage 2, definite NEC with pneumatosis intestinalis; stage 3, definite NEC with severe disease, including hypotension, metabolic acidosis, disseminated intravascular coagulopathy, neutropenia with or without perforation10. Although the frequency of NEC among all NICU admissions varies from 1.2% to 4.7%11, the frequency, severity and age of onset of NEC increase with decreasing birthweight and greater degrees of immaturity (Fig. 3). In the smallest premature neonates with birthweights of 500–750 g reported in the Vermont-Oxford Network, the frequency averaged 9%, but may reach 19% in specific NICUs, which illustrates the wide variation among centres12.
The exact aetiology of NEC remains uncertain, however, the majority of cases of NEC occur in formula-fed, premature neonates with the age of onset inversely related to gestational age, frequently at 2–3 weeks after birth for infants with gestational ages of 32 weeks or less9,11. Necrotising enterocolitis in term infants presents at a median age of 2 days with a rate that is 1/100th of that in premature newborns8. Several viral and bacterial organisms have been associated with NEC that may occur sporadically or in clusters of cases, but no single organism is thought to be the cause. Early reports correlated NEC with disorders hypothesised to cause bowel ischaemia, such as umbilical catheterisation, asphyxia/low Apgar scores and in utero reduction of umbilical vessel flows. Case-control studies have not confirmed those associations in premature neonates, but rather revealed that these are markers of sick, premature neonates rather than factors that lead to NEC11. In a review of 338 cases of NEC in term newborns, Wiswell et al. found that NEC correlated with congenital heart disease, intrauterine growth retardation, asphyxia and exchange transfusion8. Recent reports and animal studies point to a complex interplay between ischaemia, infection, intraluminal content and immature host defenses of premature newborns. NEC may represent another form of the systemic inflammatory response syndrome in neonates that is a result of an immature intestinal response to injury involving a variety of factors, including platelet activating factor, tumour necrosis factor, and nitric oxide13.
The frequency of surgical intervention ranges from 27% to 63% (average 48%) of cases14. The indications for surgical intervention for NEC are hard to define beyond pneumoperitoneum9. Loops of dead bowel may develop without evidence of perforation and may present as an immobile dilated bowel loop on serial radiographs. Surgical interventions range from placement of peritoneal drains to bowel resection with formation of multiple ostomies. Overall survival has increased from 51% in the 1970s to 70% in the 1980s15. Long-term morbidity following NEC is related to complications associated with strictures, malnutrition from limited intestinal absorptive surface, and nutritional support requiring TPN with associated catheter-related sepsis in 54%, anaemia in 50%, biliary dysfunction in 40%, rickets in 6%, cholelithiasis in 6%, and endocarditis in 4%14. In cases that require surgery, NEC leads to short bowel syndrome in 22–50% of cases, making it the leading cause of short bowel syndrome in infancy15,16.
The frequency of respiratory distress syndrome (RDS) can be decreased through prenatal treatment with the potent, halogenated corticosteroids, betamethasone and dexamethasone, but only by about 50%17. Severe neonatal lung disease from surfactant deficiency, structural immaturity, and infection remains a frequent neonatal problem (Fig. 4). Postnatal treatment of RDS with a variety of surfactant preparations reduces the frequency of barotrauma that may lead to pneumothorax and interstitial emphysema and improves lung function acutely in many, but not all, preterm newborns18. Unfortunately, surfactant treatment administered immediately after birth as prophylaxis for RDS, or as rescue treatment for established RDS, does not decrease significantly the frequency of chronic lung disease in premature newborns, bronchopulmonary dysplasia (BPD)18.
Both the nature and the definition of BPD have changed with increasing survival of extremely low birthweight neonates. Bronchopulmonary dysplasia was originally defined by Northway et al. as a lung disease of premature infants who had required mechanical ventilation for at least one week, had persistent respiratory distress symptoms with a requirement for oxygen at four weeks after birth, and had an abnormal chest radiograph19. At that time, however, the average birthweight and gestational age of survivors was 34 weeks and 2234 g20. In these larger infants, BPD often followed barotrauma in the form of a pneumothorax or interstitital emphysema and included histopathological changes of airway inflammation, fibrosis, and atelectasis alternating with hyperinflation20. Increased survival of extremely low birthweight neonates has led to large numbers of preterm neonates who require mechanical ventilation and oxygen support for several weeks after birth due to mild lung disease superimposed on their small size, but who do not have the same long term respiratory problems of the infants described by Northway. At 24–28 weeks gestation, the distal saccules are just beginning to alveolarise, and mechanical ventilation arrests normal lung development at this early stage21. In these very immature infants, as well as animal models, BPD develops with less inflammation and fibrosis, but with fewer alveoli that are larger and have fewer septae21,22. Because more immature infants (especially <32 weeks gestation) can be classified as having BPD by the Northway criteria, but do not demonstrate the same long-term lung disease, investigators and clinicians have redefined BPD as an oxygen requirement at 36 weeks post conceptional age with an abnormal chest radiograph following mechanical ventilation at birth23,24. Based on the original Northway criteria, more than 75% of infants with birthweights <1000 g have BPD at 28 days after birth20, yet only about 30% of infants surviving to 36 weeks meet the new criteria for BPD24. To insure identification of infants with long-term lung disease, the Vermont-Oxford Network tracked infants who continue to require oxygen at discharge (Fig. 4). Bronchopulmonary dysplasia has now become the leading cause of chronic lung disease in infants in the USA as well as contributing to pulmonary abnormalities in adulthood25.
Long-term evaluations of earlier cases of BPD show that pulmonary problems related to BPD may persist into adolescence and young adulthood. Northway compared pulmonary function in 24 of 26 adolescents and young adults diagnosed with BPD following NICU treatment from 1964 to 1973 with a cohort born at similar gestational ages, but without BPD, as well as to normal subjects born at term25. In the BPD group, 76% had abnormalities of pulmonary function in the form of airway obstruction, airway hyperreactivity and hyperinflation. They demonstrated significant decreases in peak expiratory flow rate, forced vital capacity, FEV1, forced expiratory flow between 25% and 75% of vital capacity, and maximal expiratory flow velocity at 50% of vital capacity. Similar abnormalities of pulmonary function were observed at seven years of age in children with BPD born in the pre-surfactant era of 1985–198626. Lung function in children with BPD may improve during growth27. Blayney et al. evaluated the change in pulmonary function in 32 children with BPD between 7 and 10 years of age after preterm births between 1977 and 198027. At 7 years, functional residual capacity and total lung capacity were normal and remained normal at 10 years. FEV1 corrected for height and gender, however, improved from 63 ± 11% to 72 ± 16% predicted during the same interval (P < 0.01).
Long term studies of BPD described above reflect ventilator equipment and styles of ventilator management that are currently not used and treatment prior to widespread use of corticosteroids prenatally and surfactant postnatally. A recent study compared pulmonary function at 6.4 years in 69 children born prematurely at an average gestation of 30–31 weeks, treated postnatally with surfactant, and randomised to treatment with conventional ventilation (CV) or high frequency oscillator ventilation (HFOV)28. The HFOV group had less chronic lung disease and better pulmonary function at 6.4 years. The CV group, although not diagnosed with BPD, had pulmonary function changes consistent with obstructive airway disease similar to that found in subjects with BPD. Thus, newer pulmonary treatment strategies may reduce the lung damage found to accompany earlier neonatal care.
Retinopathy of Prematurity
Retinopathy of prematurity (ROP) is a process of abnormal neovascularisation due to extreme prematurity, hyperoxia and retinal ischaemia. The incidence and severity of ROP is inversely related to gestational age (Fig. 5). Retinopathy of prematurity is rarely identified in infants with birthweight >1500 g or >32 weeks gestational age. Severe ROP leading to retinal detachment and blindness has a much higher incidence in infants with birthweight <1000 g29. Compared to fetal oxygen levels, the extrauterine environment is relatively hyperoxic to the premature infant even if no supplemental oxygen is required. The choroidal circulation, unlike the retinal circulation, does not autoregulate in response to changes in oxygen tension. Thus, under conditions of hyperoxia, excess blood and oxygen move from the choroidal to the retinal circulation. This increase in oxygen can cause the retinal vessels to constrict to the point of obliteration. In addition, oxygen promotes creation of free radicals that can overwhelm the available antioxidants. Genetic factors also play a role in the incidence of severe ROP, as African-American preterm infants are less likely to develop ROP than Caucasian preterm infants, controlling for gestational age and severity of illness29.
Retinopathy of prematurity can be diagnosed beginning at 32–34 weeks postconceptional age regardless of the gestational age at the time of delivery. The international classification of ROP describes the location of the lesion in concentric rings relative to the optic nerve (zones I, II, III), the degree of abnormality (stage), the extent of the developing blood vessels (clock hours), and the presence of engorged and tortuous vessels (plus disease). Retinal blood vessels develop from the optic nerve to the periphery. Zone I is the area immediately surrounding the optic nerve and macula (Fig. 6). Retinopathy of prematurity in zone I is the most concerning as progression leads to scar formation, visual impairment, and retinal detachment. Lesions in zone III do not usually lead to severe visual impairment.
ROP is classified into five stages. In stage 1, there is a line of demarcation between the vascular and avascular retina. In stage 2, the line of demarcation develops into a rolled ridge of scar tissue. Stage 3 is characterised by the development of extraretinal blood vessels and fibrous tissue. In stage 4, the retina is partially detached due to scar tissue pulling the retina away from the orbit. Total retinal detachment is defined as stage 5. The number of clock hours affected determines the extent of ROP (Fig. 6). The more clock hours involved, the more likely that the lesion will progress to retinal detachment and will not spontaneously resolve. Threshold disease is defined as stage 3 disease in zone I or II with five contiguous or eight non-contiguous clock hours. Greater than 50% of eyes with threshold ROP will progress to retinal detachment. Stage 3 blood vessels that are tortuous and engorged (plus disease) often progress to scarring and retinal detachment29.
In order to prevent subsequent retinal detachment, threshold ROP is treated with either cryotherapy or more recently by laser ablation to destroy abnormal retinal tissue, eliminate growth of abnormal blood vessels, and hopefully end the progression of scar tissue formation. In a large multicentre trial that enrolled 4,099 infants with birthweight <1251 g, 291 infants (7.1%) developed threshold ROP30. Infants were randomly selected to receive cryotherapy or no therapy. Even though cryotherapy did improve outcome when compared with control infants, 44% of eyes treated with cryotherapy had an unfavourable functional outcome30.
Recently, laser photocoagulation has replaced cryotherapy in the treatment of ROP because cryotherapy is associated with more side effects (i.e. thermoregulation, apnea, conjunctival laceration, vitreous haemorrhage, and constricted visual fields)31. In a smaller study involving only infants with bilateral and symmetric threshold ROP, one eye received cryotherapy and the other eye received laser photocoagulation. In both the 5 and 10 year follow-up, the laser treated eye demonstrated significantly better visual acuity due to decreased retinal dragging from scar tissue and decreased myopia compared with the cryotherapy-treated eyes32–34.
Retinopathy of prematurity may regress without treatment at all stages, but vision may still be impaired. Even stage 1 to 3 mild ROP may resolve and leave no scar tissue, but the infants are still at risk for reduced visual acuity. Pre-threshold and threshold ROP can also regress spontaneously, but the remaining scar tissue can lead to retinal detachment at a much later age, due to traction of the retinal scar as the eye grows. Other complications of ROP in addition to altered visual acuity, retinal detachment, and blindness include strabismus, amblyopia, myopia and retarded ocular growth.
The most important measure of neonatal intensive care is the long-term, developmental outcome of the survivors. Many physicians and families have been concerned that the increasing survival of extremely premature newborns will increase the percentage of neurologically impaired children. Outcome studies are hampered by difficulties in longitudinally tracking a large cohort of infants over time and the limitation that they do not reflect newer treatments, such as prenatal corticosteroids, surfactant, high-frequency ventilation, laser photocoagulation treatment for ROP, and individualised developmental care.
Several studies have investigated the change in long-term outcomes of infants with birthweight <1000 g born in discrete time intervals. O'Shea et al. examined infants from North Carolina born during three five-year intervals from 1979 to 199435. This study reported evaluations at one year of age of 209 (97%) of the 216 survivors from 513 live births with birthweights of 501–800 g. During the three epochs, survival increased progressively from 20% to 59%. Major neurosensory impairment, defined as cerebral palsy, blindness, or Bayley MDI < 68, did not vary significantly over time at 25%, 28%, and 21%, sequentially. Cerebral palsy varied from 13% to 19% to 7% in the latest epoch. The major risk factors for neurosensory impairment were younger gestational age and major intracranial ultrasound abnormalities. Increased maternal education displayed a protective effect. Important confounding factors in this study must be noted. During the first five-year epoch, approximately 75% of the infants were small for gestational age and during the last epoch, one third of the infants were treated with dexamethasone for chronic lung disease, which has been implicated in poorer developmental outcomes36.
The Victorian Infant Collaborative Study Group also investigated outcomes for infants with birthweights of 500–999 g born in three discrete time periods between 1979 and 1992 in Australia. Infants were evaluated at two years of age37. During this time, cerebral palsy decreased from 13.5% to 9.3% and moderate disability (blindness, severe cerebral palsy, and developmental quotient >2 standard deviations below the norm) decreased from 21.3% to 13.5%. Smaller infants with birthweights of 500–749 g had worse outcomes at each time period compared to infants with birthweights of 750–999 g. During the last epoch of 1991–1992, moderate to severe disability for a control full-term cohort was 3.3%. Since very low birthweight infants are only a small fraction of total births, the actual number of moderate to severely impaired preterm infants is much smaller than the number of similarly impaired term infants.
In 1993–1994, the NICHD Neonatal Network enrolled 2,498 infants with birthweight 400–1000 g of which 1,527 infants survived1. At 18 months, 1,151 infants were evaluated, reduced from the original population by 47 deaths after discharge and the rest (21%) lost to follow-up. Overall, 25% of the infants had an abnormal neurological exam with cerebral palsy present in 17%, blindness in 2%, and hearing impairment in 11%. The highest incidence of blindness and cerebral palsy occurred in the 400–500 g birthweight group, but the number of infants was small (n= 14). For infants with abnormal neurological examinations, 69% and 73% exhibited a MDI or PDI < 70, respectively. Most concerning, 26% of the children with a normal neurological examination had a MDI < 70. Again in this population, 44% of the infants were treated for chronic lung disease with dexamethasone.
In conclusion, increasing survival of extremely low birthweight newborns has not been accompanied by an increasing incidence of neurosensory impairment. In some areas, the increasing rate of survival may increase the absolute number (not the percentage) of survivors with significant impairment. This increased survival of infants who are extremely immature at birth has changed the nature of BPD from severe pulmonary inflammation and fibrosis to mild inflammation and hypoplasia, but it still may cause severe, long-term, pulmonary dysfunction. Newer ventilator management offers hope for a reduction in the impairment of lung development currently observed. While the incidence of severe IVH and ROP has decreased significantly, they remain problems, having a significant adverse impact on long-term neurosensory function for extremely premature newborns. Current changes in neonatal care warrant careful ongoing evaluation during the coming years.