Main

The increased survival of very low birth weight (VLBW) infants (<1.5 kg) has resulted in a marked change in the epidemiology of cerebral palsy and related forms of neurologic impairment (NI). Surviving VLBW infants currently account for 25–50% of all cases (1, 2). Several recent studies have addressed perinatal risk factors for NI in these infants. Parenchymal brain lesions such as severe intraventricular hemorrhage with periventricular infarction and periventricular leukomalacia detected by neonatal imaging studies are the strongest correlates (3–5). Other important risk factors for NI include amniotic fluid infections, hypocarbia, hypothyroxinemia, prolonged rupture of membranes, and the development of chronic lung disease (6–11).

In a case control study of VLBW infants delivered and followed at Rainbow Babies and Children's Hospital over the period 1983–1991 we identified clinical risk factors associated with NI at 20-mo corrected gestational age (12). We subsequently described the placental pathology of these infants (13). Placental lesions may act directly to cause brain injury or indirectly by increasing the incidence of neonatal complications. To distinguish between these alternatives we analyzed the interrelationships between placental and clinical risk factors and assessed their relative and temporal contributions to NI. In view of recent interest in the role of amniotic fluid infections in brain injury, we also explored the relationship between histologic CA and clinical CA, their associations with neonatal sepsis, and the mechanisms by which each contributes to NI (14, 15).

MATERIALS AND METHODS

Study population.

This was a nested case control analysis of placental and clinical risk factors associated with NI at 20-mo corrected age in singleton infants without major congenital anomalies using data collected in a prospective study of VLBW (<1500 g) infants born between 1983 and 1991 at our institution and followed in a high-risk follow-up program at Rainbow Babies and Children's Hospital. In a previous study, 72 with major NI at 20-mo corrected age (50 infants with strictly defined cerebral palsy plus 22 infants with other neurologic abnormalities that overlap with the diagnosis of cerebral palsy—hypertonia, defined as increased muscle tone, usually in the lower extremities with normal truncal tone; hypotonia, defined as decreased muscle tone involving the trunk and/or extremities; and shunt dependent hydrocephalus without abnormal neurologic signs—were compared with 72 controls with a normal neurologic examination at 20-mo corrected age selected among infants born in the same year and matched for weight (±250 g), gestational age (+ 2 wk), and race (12).

The population evaluated in this study was the subset of cases and controls from the original sample in whom placentas were submitted to pathology (13). Placental reports and slides were available in 119 of the 144 original study patients [strict cerebral palsy—42/50 (84%), strict cerebral palsy controls—43/50 (86%), other neurologic abnormalities—18/22 (82%), and other neurologic abnormalities controls—16/22 (73%)]. Comparisons among the original matching variables in patients with placentas available for study were as follows: African-American mothers—57% for cases versus 59% for controls; mean gestational age—27.3 ± 2.3 wk for cases and 27.5 ± 2.2 wk for controls; and mean birth weight—978 ± 228 g for cases versus 1012 ± 226 g for controls. None of the above differences were statistically significant.

Clinical findings.

The significance of clinical determinants of NI from our previous study was recalculated for the subsample with placental pathology. Of the determinants identified as associated with NI in the previous study, only hypothyroxinemia had a p value >0.10 in the subsample. The definitions of the clinical findings evaluated in this report have been previously described and are briefly summarized below (12). Clinical CA was defined as premature rupture of membranes with a temperature of either >37.8°C on two occasions, at least 1 h apart, or >38.3°C on one occasion with no other known sources for the fever and one of the following: maternal tachycardia (≥100 beats per min), fetal tachycardia (≥160 beats per min), maternal leukocytosis (>11 000/mm white blood cells), or foul-smelling amniotic fluid. Chronic lung disease was defined by oxygen dependence at 36-wk postconceptional age. Sepsis was defined by at least one positive blood culture accompanied by clinical signs suggestive of infection. Sepsis was subdivided into early sepsis (<7 d after birth) and late sepsis (≥7 d after birth) (16). Hyperbilirubinemia was defined as a maximum total bilirubin level >10 mg/dL. Severe cranial ultrasound abnormality was defined as the diagnosis of one or more of the following at any time during admission: grade 3–4 periventricular hemorrhage, parenchymal infarction, or periventricular leukomalacia, or the presence of ventricular dilatation at discharge.

Placental pathology.

All placentas were examined and processed using a standard protocol (17). Details of the examination and classification scheme have previously been described (13) and conform to published consensus criteria (18). Chorionic vessel thrombi were defined by laminated fibrin adherent to the wall of chorionic vessels. Severe (grade 3) villous edema was defined as large vacuoles in the stroma affecting >19% of villi. Histologic CA was separated into maternal and fetal inflammatory components and graded as follows: Maternal grade, based on the worst area of the chorionic plate: 1—rare neutrophils [<10/high-powered field (HPF) ], 2—intermediate neutrophils (11–30/HPF), and 3—abundant neutrophils (>30/HPF). Fetal grade, based on the most severely affected chorionic plate vessel: 1—rare neutrophils, 2—intermediate neutrophils, and 3—abundant neutrophils. The category of multiple placental lesions was defined by the presence of two or more of the following lesions: retroplacental hematoma with indentation, multiple villous infarcts, diffuse microinfarction, increased nucleated red blood cells, and grade 2–3 increased basal perivillous fibrin.

Statistical analysis.

χ2 and Fisher exact tests were used for nominal data. To test the independent effects of placental and other perinatal risk factors, unconditional logistic regression was conducted. Due to limited sample size we included only variables with univariate associations at the p < 0.10 level. Variables were separated into three groups: prenatal, neonatal, and sonographic. To examine the effects of these three groups of findings, each group of variables was added by block entry as a separate step in the logistic regression analysis. Finally, we used methodology drawn from the principles of path analysis to temporally examine the relationships between antecedents of NI (19). A series of logistic regression analyses was conducted, with the dependent variable changing as we moved back along the causal chain, including all variables that could be considered as antecedent to that dependent variable. The groupings for this analysis were findings indicating antenatal infection, placental lesions, and neonatal variables. The purpose of these analyses was not to calculate indirect or total effects as done in linear-based path analysis, but rather to provide a conceptual framework for exploring interrelationships.

RESULTS

The frequencies of all placental and risk factors significant at the p > 0.10 level in VLBW infants with and without NI at 20-mo corrected age is shown in Table 1 with their associated odds ratios and 95% confidence intervals. Placental abnormalities that increased in infants with NI were chorionic vessel thrombi, severe (grade 3) villous edema, severe fetal (grade 3) histologic CA, and multiple placental lesions. Clinical findings that increased in infants with NI were clinical CA, oxygen dependence at 36-wk postconceptional age, sepsis, hyperbilirubinemia (>10 mg/dL), and severe cranial ultrasound abnormalities.

Table 1 Univariate associations (p < 0.10) of placental lesions and clinical findings with neurologic impairment * Odds ratio (95% confidence intervals). † Number positive (% positive). ‡ See materials and methods.
Full size table

Interrelationships among risk factors indicative of perinatal infection (histologic CA, clinical CA, and sepsis) are illustrated for descriptive purposes in Table 2. Four histologic subgroups were distinguished: 1—no histologic CA, 2—histologic CA without fetal inflammation (maternal inflammation only), 3—histologic CA with mild-to-moderate (grades 1–2) fetal inflammation, and 4—histologic CA with severe (grade 3) fetal inflammation. Groups 1 and 2 (no histologic CA and histologic CA without fetal inflammation) had equivalent low levels of concordance with clinical CA (9% and 7%, respectively). Group 3 (histologic CA with mild-to-moderate fetal inflammation) showed only slightly better concordance (15%). Only Group 4 (histologic CA with severe fetal inflammation) had a high level of concordance (77%) with clinical CA. Early sepsis (<7 d after birth) was uncommon (6% of total cases) and occurred only in infants whose placentas had histologic CA with fetal inflammation (groups 3 and 4). Late sepsis (≥7 d after birth) was unrelated to histologic CA. The relative effects of clinical CA and/or histologic grade 3 CA on the risk of NI are summarized at the bottom of Table 2. While fetal grade 3 histologic CA was associated with clinical CA as described above, more infants (54%) had one or the other finding than both. Odds ratios for NI were higher with clinical CA alone (OR 6.0) and fetal grade 3 histologic CA (OR 5.3) than they were when both were present (OR 2.1). Because of the small numbers involved and the retrospective nature of the data, no statistical inferences were drawn from this unexpected finding.

Table 2 Histologic chorioamnionitis and clinical infection in infants with and without neurologic impairment * Infants with NI. † Infants with no NI. ‡ Number positive (% positive). § Assumes one case among controls to avoid incalculable odds ratio.
Full size table

To determine independent predictors of NI, logistic regression with block entry of prenatal, postnatal, and sonographic risk factors was performed (Table 3). In the full analysis including all three groups of variables (Model 3), three risk factors—multiple placental lesions, oxygen dependence at 36-wk postconceptional age, and severe cranial ultrasound abnormalities—were found to be independent predictors of NI. After exclusion of findings obtained by direct imaging of the CNS (Model 2), grade 3 villous edema achieved significance. Chorionic vessel thrombi, while not significant, maintained adjusted odds ratios (3.5–3.8) similar to the unadjusted odds ratio (3.7) shown in Table 1 in all three models. Clinical CA, fetal grade 3 histologic CA, hyperbilirubinemia, and sepsis were not independent predictors of NI in any of the models tested.

Table 3 Logistic regression analysis of placental and clinical risk factors for neurologic impairment* * Total = 119 patients, 60 cases, and 59 controls. † CV thrombi, chorionic vessel thrombi; Gr3 edema, grade 3 villous edema; FGr3 HCA, fetal grade 3 histologic chorioamnionitis; Multiple, multiple placental lesions; Clin CA, clinical chorioamnionitis; Hyperbili, hyperbilirubinemia (>10 mg/dL); O2 36 wks, oxygen dependence at 36 wk corrected gestational age; Sev U/S, severe ultrasound abnormality. (See Materials and Methods for detailed definitions.)
Full size table

Finally, interrelationships among risk factors operating at different stages were explored. A series of logistic regressions was conducted with the dependent variable changing as we moved back along the causal chain, considering all variables that might be considered antecedent to that dependent variable. The temporal groupings for this analysis were 1—findings indicative of antenatal infection, 2—placental lesions, and 3—neonatal variables. The rationale for the ordering of groups 1 and 2 is that, although antenatal infection could be an antecedent for some of the placental lesions in group 2, the converse is highly unlikely. Variables in groups 1 and 2, occurring before birth, were clearly antecedent to neonatal variables in Group 3. Figure 1 illustrates the groupings and the relationships between variables with their associated odds ratios and 95% confidence intervals. Only relationships significant at p < 0.45 are shown. In addition to the predictors of NI described above, four significant associations between intermediary variables were found: grade 3 villous edema with severe ultrasound abnormalities, clinical CA with grade 3 villous edema, fetal grade 3 histologic CA with chorionic vessel thrombi, and fetal grade 3 histologic CA with clinical CA. Oxygen dependence at 36-wk postconceptional age showed weak nonsignificant relationships to two other variables—grade 3 villous edema and clinical CA.

Figure 1
figure 1

Direct and mediating effects of risk factors for NI as determined by serial logistic regressions (see “Materials and Methods”). Variables were considered in three groups: indicators of antenatal infection, placental abnormalities not directly attributable to bacterial infection, and neonatal factors. Significant interrelationships are indicated by solid lines with odds ratios and 95% confidence intervals. All nonsignificant interrelationships at the p < 0.450 level are also shown (dashed lines). Abbreviations are as follows (see “Materials and Methods” for definitions): NI, neurologic impairment; Sev U/S, severe cranial ultrasound abnormalities; O236, oxygen dependence at 36-wk postconceptional age; Gr3 Edema, grade 3 villous edema; CV Thrombi, chorionic vessel thrombi; Multiple, multiple placental abnormalities; Clin CA, clinically diagnosed chorioamnionitis; FGr3 HCA, fetal grade 3 histologic chorioamnionitis.

Full size image

DISCUSSION

Identifying the causes of cerebral palsy and related forms of NI in VLBW infants is complicated by several inherent difficulties (20). First, cerebral palsy is an operational definition combining an etiologically diverse group of children with nonprogressive motor disorders of central origin. Second, delayed ascertainment, the inability to diagnose cerebral palsy until long after the time of the actual injury, is particularly problematic for the VLBW infant where the diagnosis often follows a prolonged hospitalization with multisystem disease. To overcome this second difficulty, recent studies have focused on early objective parameters of brain injury such as severe ultrasound abnormalities (21, 22). However, these are surrogate outcomes not always corresponding to the population with long-term NI. Patients with severe ultrasound abnormalities may either be neurologically normal, die, or have severe multisystem disease removing them from the group of children eventually classified with NI. In this study we used a comprehensive assessment of neurologic motor function at 20-mo corrected age to define outcome.

It is widely acknowledged that NI reflects an interplay between antenatal, intrapartum, and neonatal processes (23, 24). Antenatal processes are particularly difficult to study because they are poorly understood and often estimated by proxies such as rupture of membranes or maternal fever rather than by specific processes affecting the fetoplacental unit. Evaluation of the placenta has the potential to provide insight into these antenatal processes by demonstrating anatomic lesions associated with decreased maternal substrate delivery, altered transport function, compromise of the fetal vasculature, or unrecognized inflammatory processes. Few studies have directly measured the impact of placental lesions on neurologic outcome and much of the existing data are limited by one or more of the following problems: data collection before modern intensive care (25, 26), analysis of surrogate outcomes such as ultrasound abnormalities or neuropathologic lesions (21, 27–29), and restricted emphasis on a single placental lesion (29, 30). Very few previous studies have attempted to separate direct from indirect effects and study interactions between placental and other perinatal risk factors (21, 25).

We studied a cohort of VLBW infants derived from a single institution born over a relatively confined time period and followed up in a comprehensive program to determine long-term outcome. In two previous papers we reported individual clinical and pathologic risk factors associated with NI (12, 13). In this paper we have studied the interrelationships between clinical and placental risk factors to determine independent predictors of NI and their antecedents. Weaknesses of the present study include retrospective analysis of data, the relatively small number of cases, the low incidence of some pathologic lesions, and the 8 y time period over which our cases were collected. Three independent risk factors for NI were identified: multiple placental lesions, severe cranial ultrasound abnormalities, and chronic lung disease as estimated by oxygen dependence at 36-wk postconceptional age.

The high rate of NI observed in infants having placentas with multiple placental lesions supports the concept that risk factors may act in an additive or synergistic fashion. Lesions in this category included two indicators of severely decreased uteroplacental perfusion (multiple infarcts and diffuse microinfarction), an idiopathic pregnancy disorder (increased basal perivillous fibrin, also known as “maternal floor infarction”), an indicator of premature placental separation (retroplacental hematoma with indentation), and a measure of subacute fetal hypoxia (increased nucleated red blood cells) (18). Similar additive or synergistic interactions between different types of lesions have been reported for ultrasound abnormalities associated with NI and placental lesions associated with stillbirth (31, 32). The relative importance of interactions among specific placental lesions is likely to depend on time of onset, duration, severity, and host susceptibility and requires larger studies to resolve.

The association of severe cranial ultrasound abnormalities (periventricular white matter degeneration, periventricular hemorrhagic infarction, and/or posthemorrhagic hydrocephalus) with NI has been well documented and thoroughly discussed in previous studies (3, 4, 31, 33). In some cases the severity of the ultrasound lesion is such that it represents a proxy for the outcome. Nevertheless, some patients with cranial severe ultrasound abnormalities (18% in our study) did not have NI at 20 mo and many patients with NI at 20 mo did not have these sonographic findings (40% in our study). Further data regarding the exact nature and timing of ultrasound abnormalities and/or more sophisticated imaging techniques could potentially improve the predictive power of neuroradiologic findings.

Severe villous edema showed a strong and somewhat unexpected relationship to cranial ultrasound abnormalities in our study. Villous edema was originally described by Naeye and further characterized by Kliman (34, 35). Previous data have suggested that this lesion is associated with poor Apgar scores, acidosis, antenatal and neonatal death, and neuropsychological abnormalities (36). While the exact pathogenesis of villous edema is not clear, it appears to reflect a fetal circulatory abnormality, to have a relatively rapid onset, and to resolve slowly. The lesion has been suggested to exert its effect by compression of villous arterioles (35). An alternative view is that villous edema is a marker for a more generalized loss of fetal circulatory integrity (37).

Oxygen dependence at 36-wk postconceptual age, an indicator of chronic lung disease, has been identified as a risk factor for NI in previous studies (3, 11). While none of the antenatal variables studied were significant predictors of chronic lung disease when the latter was treated as a dependent variable, two processes showed borderline significance—clinical CA (p = 0.113) and grade 3 villous edema (p = 0.090). Several studies have suggested that chorioamnionitis may be a predisposing factor for chronic lung disease (38–40). Possible roles for cytokines, proteases, and reactive oxygen intermediates produced by leukocytes aspirated into the lungs during the course of infection have been suggested as possible mechanisms (41, 42). In view of the suggestive nature of the available evidence and the emergence of new patterns of chronic lung disease owing to smaller surviving babies and new management strategies a larger study using more recent cases would be informative.

A number of studies have identified clinical CA as a risk factor for NI (6, 7, 12, 14, 25, 43, 44). An association between clinical and histologic CA in our study was found only for the subgroup of histologic CA with severe (grade 3) fetal inflammation. While there was considerable overlap between clinical CA and fetal grade 3 histologic CA, it was the cases with one or the other, but not both, that contributed most strongly to NI. The distinction between these two risk factors was supported by different associations with other placental lesions: fetal grade 3 CA with chorionic vessel thrombi and clinical CA with villous edema. One possible explanation might be that fetal grade 3 CA selects for longstanding amniotic fluid infections with intact membranes whereas clinical CA required ruptured membranes by definition. Interestingly, both short (<1 h) and prolonged (> 24 h) duration of membrane rupture have been associated with NI in previous studies (9, 14). Misclassification of extrauterine infections, which have also been associated with NI, as clinical CA is another possible explanation (45, 46).

Our finding that fetal as opposed to maternal inflammation is the critical feature linking histologic CA with NI, clinical CA, and early neonatal sepsis is in agreement with other recent studies and emphasizes the importance of the recently described fetal inflammatory response syndrome (47, 48). This syndrome was introduced to emphasize the deleterious effects of acute phase reactants, circulating cytokines, and other procoagulant molecules released into the fetal circulation during amniotic fluid infection. Two of these deleterious effects may be chorionic vessel thrombi and villous edema, as suggested by this study.

In summary, our retrospective study has identified multiple placental lesions, severe cranial ultrasound abnormalities, and chronic lung disease as independent predictors of NI in VLBW infants. We have also delineated patterns of placental injury that interact both with one another and with neonatal processes to increase the risk of NI. Confirmation of these relationships and extension to more specific subgroups of NI requires further study.