toregulation.15'16 PVL has been reported to occur in
5-17% of all infants born weighing <1500 g.17~19 Between 66-100% of infants that develop PVL will develop CP.16-19 Hence, the incidence of CP is 30-50
per 1000/live births <1500 g20-23 compared to 3 per
1000 in all births.24"26 The characteristic type of CP
preterm infants develop is spastic diplegia.2127 The
spastic diplegic form occurs in up to 80% of
preterm infants that develop CP28 compared to
16% of all children with CP.29
Nelson and Grether30 recently reported that
antenatal exposure to magnesium sulfate (MgSO4)
is associated with a reduced risk of developing cerebral palsy in very low-birth-weight infants. It was interesting to note that of the three preterm infants
that developed CP in thier MgSO4 exposed group,
two had the hemiparetic type that is typically not related to the cPVL lesion and 1 (33%) had the spastic diplegic type. In comparison, 23 of the 39 (59%)
infants found to have CP that were not exposed to
MgSO4 had the spastic diplegic form. Because there
were not cranial ultrasounds reported in Nelson
and Grether's study it is not possible to explore the
relationship between MgSO4, cPVL, and CP. However, these data are suggestive that MgSO4 may be
more specifically protective against spastic diplegia
and its antecedent cPVL. In utero, fetal exposure to
MgSO4 occurs in the treatment of maternal preeclampsia and when the agent is used as a tocolytic
to delay labor.
In an effort to determine if antenatal exposure
to MgSO4 is associated with any effect on the incidence or relative risk of developing cPVL in the
population of preterm infants treated at our institutions, we performed the following retrospective casecontrol study. Periventricular echodensities were
not included in this analysis because their significance are not well understood.
The study period covered the years 1992 to
1994. During this period all infants born at three
teaching institutions (NYU Medical Center, Morristown Memorial Hospital, and Bellevue Hospital)
were reviewed for entrance criteria for the study.
Criteria included birth weights less than 1750 g, survival for more than 7 days, at least one cranial ultrasound after 7 days of life and must have had all of
the following information in the infant's chart or
the labor and delivery room records: gestational
age; history of maternal complications and reason
for prematurity; length of rupture of membranes;
whether antenatal steroids, MgSO4, or other tocolytic agents were administered; mode of delivery;
evidence of pre-eclampsia; Apgar scores; and number of days intubated. Birth weights were recorded
and categorized as between 500-700, 751-1000,
1001-1250, 1251-1500, or 1501-1750 g.
All infants received care in one of the three tertiary level neonatal intensive care units (NICU).
304 Each center was equipped with a portable neonatal
cranial ultrasound machine with a 7.5-MHz transducer producing transfontanel coronal and sagittal
scans. One institution reproduced ultrasound images on paper. Cranial ultrasounds from all infants
were collected and re-read by a single pediatric radiologist experienced in newborn diagnostic imaging. Cranial ultrasounds were classified as normal,
grade I through III intraventricular hemorrhage,
intraparenchymal hemorrhage, and cPVL. The radiologist was blinded to all clinical parameters. Ultrasound quality was judged on a 4-point grading
system. A quality grade 1 study was a technically
good study and a grade 4 was a poor-quality study.
Controls were selected from a pool of infants
with a normal cranial ultrasound or grade I intraventricular hemorrhage. Infants in the control pool
with birth weights outside the range found in the
cPVL group were removed to make the control pool
similar to the cPVL group. From this control pool,
two controls were randomly selected for each cPVL
Chi-square analyses were used to compare incidence of cPVL in infants that received MgSO4 in
utero with the control group. Odds ratios (OR)
were calculated to estimate the relative risk of developing cPVL in the two groups. An OR <1 represents
a reduction of a disease outcome related to the intervention. We used a 95% confidence interval (CI)
as a measure of statistical precision for each OR.
To determine if the cPVL and control groups
had a similar demographic make-up, two separate
analyses were carried out. First, a multivariate test
for group differences was carried out on a number
of noncategorical variables. Specifically, a one-way
multivariate analysis of variance (MANOVA) by
groups (cPVL vs. control) and a univariate analysis
was conducted. We also analyzed the data with logistic regression in an effort to identify whether or not
the variables in the study could predict cPVL. In
particular, we were interested in the nature of
MgSO4 in the prediction model. Criteria used for
including variables in the prediction model were
(a) there was a statistically significant logistic regression, and (b) the variables in the model served
to maximize correct prediction of cases in the sample. Because this is a preliminary investigation and
there were no clear a priori reasons to identify primacy among variables, hierarchical or stepwise procedures were not used. Second, categorical variables were analyzed by Chi-square analysis.
This protocol was approved by the institutional
review boards of the participating hospitals.
During the 2-year study period there were
23,382 infants born at the three institutions. There
were 561 (2.4%) infants admitted to one of the
three NICUs with a birth weight of less than 1750 g.
Of these infants, 492 met the entrance criteria of
the study. Seventeen infants died prior to 7 days of
life, 23 infants did not receive cranial ultrasounds
Downloaded by: NYU. Copyrighted material.
AMERICAN JOURNAL OF PERINATOLOGY/VOLUME 14, NUMBER 5
MAGNESIUM SULFATE A N D CYSTIC PERIVENTRICULAR LEUKOMALACIA I N PRETERM INFANTS/FineSmith et al
not a statistically significant increase in the risk of
developing cPVL associated with any of these variables except magnesium. The Chi-square tests were
nonsignificant suggesting a general similarity between the experimental and control groups except
for magnesium, the variable of interest. See Tables
1 and 2.
The results of the logistic regression showed
the solution that best satisfied the criteria of statistical significance and maximum predictability included six variables: gestational weight, antenatal
exposure to MgSO4, 1- and 5-min Apgar scores,
number of days intubated, reason for prematurity,
and type of delivery. The logistic solution including
these variables was significant (Chi-square = 23.4,
df= 12, p = 0.014). Of all the variables in the logistic
equation, MgSO4 had the highest value of the R statistic, which in the present analysis is a measure of
the partial correlation of the variables. The R statistic for MgSO4 was significant (p < 0.03, df= 1). The logistic regression model was able to correctly predict
67% (12 of 18) of the cPVL cases and 92% (33 of
36) of the controls. When this same logistic solution
was calculated without MgSO4, prediction in the
cPVL group dropped to 44%, less than chance, and
the regression was no longer statistically significant.
Therefore, inclusion of MgSO4 resulted in a 22% increment in prediction and increased prediction
17% above chance levels. This bears out the findings in the analyses above that MgSO4 plays a role in
the prevention of cPVL.
Downloaded by: NYU. Copyrighted material.
after 7 days of life, and 29 infants had insufficient
data recorded in their charts. There were no cases
of cPVL detected in any of the cases excluded from
the study. A total of 857 ultrasound studies were reviewed. There were 12 quality grade 2 and 2 quality
grade 3 ultrasound studies. The remainder were
grade 1 studies.
There were 18 cases of cPVL that formed the
experimental group. Two infants in this group
(11.1%) were exposed to MgSO4 in utero compared
to 14 of the 36 (38.8%). Table 1 shows the OR, significance and 95% CI for MgSO4 exposure. Because
earlier studies have examined the relationship between preeclampsia and C.P. (see, e.g., Ref. 30), we
specifically examined our data from this perspective. Preeclampsia was diagnosed in 5.5% (1 out of
18) of the mothers with children with cPVL compared to 14.7% (6 out of 36) of the controls. To
evaluate the possible effect of preeclampsia on
cPVL, the number of infants born with cPVL to
mothers with preeclampsia were compared between the two groups with results shown in Table 1.
We also compared the nonpreeclamptic infants exposed to MgSO4 to the nonpreeclamptic controls.
In the cPVL group, 2 of the 17 nonpre-eclamptic infants were exposed and 9 of the 30 nonpre-eclamptics in the control group were exposed (11.8 vs.
The PVL and control group were found to be
equal in all variables except magnesium exposure.
Analysis was conducted on the continuous and categorical variables. The continuous variables (gestational age, number of days intubated and 1- and 5min Apgar scores) were found to be nonsignificant
(p = 0.53) although days of intubation was found to
covary with each of the variables tested. In part this
is due to the fact that one infant in the cPVL group
(mean = 16 days) was intubated for 100 days. Without this one case, the mean for the group would
have been 11 days and the group difference would
not have been statistically significant (t= 0.53, p = 0.14).
Second, several categorical variables were available
in the records of the infants. These variables included gestational weight, time of rupture of membranes, reason for prematurity, type of delivery,
whether or not the mother received terbutaline, ritodrine, steroids, or magnesium, and the number
of infants in each group that survived. There was
In this study of preterm infants with a birth
weight of <1750 g, we found that those infants with
cPVL were less likely to have been exposed to antenatal MgSO4. The groups were shown to be statistically similar in all other variables recorded. These
observations are consistent with those of Nelson
and Grether30 who noted a reduction in the risk of
developing CP in the group exposed to antenatal
MgSO4. Our findings suggest that the reduction in
CP may be related to a reduction in cPVL. However,
a larger prospective randomized study is needed to
substantiate these findings.
Table 1 . Maternal and Fetal Variables and Risk of Developing cPVL
MgSO 4 exposed
AMERICAN JOURNAL OF PERINATOLOGY/VOLUME 14, NUMBER 5 May 1997
Table 2. Comparison of Neonatal Variables in cPVL and Control Group
# days intu bated
Apgar at 1 min
Apgar at 5 min
Since pre-eclampsia is a common indication
for antenatal exposure to MgSO4, it would be important to consider the possibility that preeclampsia itself may confer some neuroprotection. However, when we looked only at those infants born to
preeclamptic mothers, we were unable to find any
evidence of such a primary protective effect with regard to cPVL. Conversely, the association between
antenatal MgSO4 and the reduction in cPVL persisted in infants born to nonpreeclamptic mothers
who received MgSO4 solely for tocolysis. This further supports the notion that it is the MgSO4 and
not preeclampsia that is protective.
There is convincing evidence in studies at the
cellular level 31 as well as in ischemic animal models
that magnesium has neuroprotective properties.32"35
Magnesium markedly reduces infarct size in perinatal rats. The mechanism of action may be related to
the inhibition of the iV-methyl D-aspartate (NMDA)
receptor.36 The NMDA receptor is stimulated by excitatory amino acids (EAA) in hypoxic conditions
and this results in a massive inflow of calcium ions
into the neuron. This has been described as the initial phase in the cascade of neuronal cell death.
Magnesium blocks the ion pore in the NMDA receptor in a voltage-dependent manner. Even when
the NMDA receptor is stimulated by EAAs, calcium
cannot enter the cell until the magnesium ion is extruded, thereby preventing the initiation of the cell
Other postulated mechanisms may be related
to the overall metabolic and functional depressive
effects of magnesium on the brain. Magnesium may
also act as an antagonist to the presynaptic adenosine a3 receptor and inhibit EAA release.
This study is limited by the small number of follow-up ultrasounds in those infants found to have a
normal ultrasound early in their course. Also, in
this retrospective study we were unable to document the dose of medication, duration of in-utero
exposure, and serum levels of magnesium in infants
exposed to MgSO4. Additionally, the number of
mothers included in the preeclampsia comparisons
was small making it difficult to demonstrate a statistical relationship. As noted, the wide discrepancy in
the number of days intubated between the cPVL
group (0 to 100 days) and the controls (0 to 20
days) made it difficult to determine significance.
The logistic regression analysis in this study, while
significant and showing improved predictability,
contained more variables than may be practical to
use in clinical settings. When we removed some of
these variables from the regression equation, however, predictability suffered, thus leading us to the
conclusion that there are complex interactions in
these variables that we may not fully understand.
Such interactions should be the focus of future research. Further, the number of subjects in this study
is limited and it may be that a larger sample, with
more stable parameters, may yield more parsimonious regression solutions. Finally, in this preliminary analysis, the method of the regression was simply to enter all the variables into the equation at the
same time. Hierarchical and stepwise methods may
be able to refine models in future studies.
Neonatal resuscitation has been greatly improved resulting in more timely restoration of cerebral oxygenation and perfusion in preterm infants
with evidence of hypoxic-ischemic disease (HID).
Unfortunately, HID may occur pre- or perinatally
when the neonate is not accessible to resuscitation.
However, the fetus is accessible to in-utero pharmacological therapy. The occurrence of the HID is
much greater in the lower birth-weight and gestational age infants.37 Many of these infants have transient respiratory and cardiac complications that
either resolve or are amenable to medical management. However, these nonpermanent cardiorespiratory deficiencies may result in permanent brain
damage. The outcome and quality of life of these
infants is often dramatically altered.
The etiology of CP is clearly heterogeneous,
but two prominently identifiable and frequently associated risk factors are preterm birth and HID.38 It
is therefore pertinent that the first step in prevention be aimed at the treatment of those infants that
are either at greatest risk of, or that have already
suffered HID. Comprehensive prenatal care, fetal
monitoring, and improved neonatal resuscitation
are currently utilized but this has not been enough
to improve the incidence of CP in the preterm infant. Therefore, the next frontier is to learn to protect the neonates brain against cell death and injury
in the face of the HID. With appropriate treatment,
and ultimately prophylactic treatment of preterm
neonates, the occurrence of neurological sequela
and the quality of life of many thousands of individuals could be vastly improved.
Downloaded by: NYU. Copyrighted material.
NS-not statistically significant.
MAGNESIUM SULFATE AND CYSTIC PERIVENTRICULAR LEUKOMALACIA IN PRETERM INFANTS/FineSmith et al
Cerebral palsy of cystic periventricular leukomalacia in lowbirth-weight infants. Acta Paediatrica 1994;83:397-401
20. Pidcock FS, Graziani LJ, Stanley C, Mitchell DG, Merton P.
1. Guyer B, Strobino DM, Venttura SJ, Singh GK. Annual sumNeurosonographic features of periventricular echodensities
mary of vital statistics—1994. Pediatrics 1995;96:1029-39
associated with cerebral palsy in preterm infants. J Pediatr
2. Wagner ME. Annual summary of statistics—1993. Pediatrics
21. Powell TG, Pharoah PD, Cooke RI, Rosenbloom L. Cerebral
3. Horbar JD, McHuliffe TL, Adler SM, Albersheim S, Cassady
palsy in low-birthweight infants. II. Spastic diplegia: AssociaG, Edwards W, Jones R, Kaftwinkel J, Kraybill EN, Krishnan
tions with intrapartum stress. Dev Med Child Neurol 1988;
V, Raschko P, Wilkinson AR. Variability in 28 day outcomes
for very low birthweight infants: An analysis of 11 neonatal
22. Powell TG, Pharoah PD, Cooke, RI, Rosenbloom L. Cerebral
intensive care units. Pediatrics 1988;82:554-59
palsy in low-birthweight infants. I. Spastic hemiplegia: Asso4. Hagberg B, Hagberg G, Olow I. The changing panorama of
ciations with intrapartum stress. Dev Med Child Neurol
cerebral palsy in Sweden. Acta Paediatr Scand 1984;73:
23. Nelson KB, Ellenberg JH. Neonatal signs as predictors of
5. Pharoah PD, Cooke T, Rosenbloom I, Cooke RI. Trends in
cerebral palsy. Pediatrics 1979;64:225-32
birth prevalence of cerebral palsy. Arch Dis Child 1987;2:
24. Naeye RL, Peters EC, Bartholomew M, Landis B. Origins of
cerebral palsy. AmJ Dis Child 1989; 143:1154-61
6. Hagberg B, Hagberg G, Zetterstrom R. Decreasing perinatal
25. Boyle CA, Decoufle P, Yeargin-Allsopp M. Prevalence and
mortality—Increase in cerebral palsy morbidity. Acta Paedihealth impact of developmental disabilities in US children.
atr Scand 1989;78:664
7. Krudrjavcev T, Schoenberg BS, Kurland LT, Groover, RV.
26. Murphy CC, Yeargin-Allsopp M, Decoufle P, Drews CD.
Cerebral palsy—Trends in incidence and changes in concurPrevalence of cerebral palsy among ten-year-old children in
rent neonatal mortality: Rochester, MN, 1950-1976. Neurolmetropolitan Atlanta, 1985 through 1987. J Pediatr 1993;
8. Naulty C, Long L, Pettett G. Prevalence of prematurity, low
27. Stanley FJ. Spastic cerebral palsy: Changes in birthweight
birth weight, and asphyxia as perinatal risk factors in a curand gestational age. Early Hum Dev 1981;51:167-78
rent population of children with cerebral palsy. Am J Perina28. McDonald AD. Cerebral palsy in children of very low birth
tol 1994; 11:377-81
weight. Arch Dis Child 1963;38:579
9. Nelson KB, Ellenberg JH. Antecedents of cerebral palsy. I.
29. Jarvis S, Holloway J, Hey E. Increase in cerebral palsy in norUnivariate analysis of risk. AmJ Dis Child 1985;139:1031-8
mal birthweight babies. Arch Dis Child 1985;60:1113-21
10. Banker BQ, LarrocheJC. Periventricular leukomalacia of in30. Nelson KB, Grether JK. Can magnsium sulfate reduce the
fancy: A form of neonatal anoxic encephalopathy. Arch Neurisk of cerebral palsy in very low birthweight infants? Pedirol 1962;7:386-410
11. de Vries L, Levene M. Cerebral ischemic lesions. In Levene
31. Ames A, Nesbett AM. Pathophysiology of ischemic cell
M, Lilford R, eds. Fetal and Neonatal Neurology and Neurosurdeath: I. Time of onset of irreversible damage; Importance
gery. London: Churchill Livingstone; 1995:323
of the different components of the ischemic insult. Stroke
12. Young RS, Hernandez MJ, Yagel SK. Selective reduction of
blood flow to white matter during hypotension in newborn
32. Izumi Y, Roussel S, Pinard E, Seylaz J. Reduction of infarct
dogs: A possible mechanism of periventricular leukomalacia.
volume by magnesium sulfate after middle cerebral artery
Ann Neurol 1982;12:445-8
occlusion in rats. J Cereb Blood Flow Metab 1991 ;11:1025-30
13. Duffy TE, Cavazzuti M, Cruz NF, Sokoloff L. Local cerebral
33. Schmitt H, Barth G, Thierauf P. Neuronal protection by ischglucose metabolism in newborn dogs: Effects of hypoxia and
emic brain perfusion: An electron microscopy study in the
halothane anesthesia. Ann Neurol 1982;11:233-46
rat. J Neurosurg Anesthesiol 1994;6:265-74
14. Cavazzuti M, Duffy TE. Regulation of local cerebral blood
34. Marret S, Gressens P Gadisseux J, Evrard P. Prevention by
flow in normal and hypoxic newborn dogs. Ann Neurol
magnesium of exitotoxic neuronal death in the developing
brain: An animal model for clinical intervention studies. Dev
15. Takashima S, Tanaka K. Development of cerebrovascular arMed Child Neurol 1995;37:473-84
chitecture and its relationship to periventricular leukomala35. Thordstein M, Bagenholm R, Thringer K, Kjellmer I. Scavcia. Arch Neurol 1978;35:11-16
engers of free oxygen radicals in combination with magne16. Volpe, JJ. Neurology of the Newborn. Philadelphia: Saunders;
sium ameliorate perinatal hypoxic-ischemic brain damage in
the rat. Pediatr Res 1993;34:23-6
17. de Vries LS, Eken P, Groendaal F, van Haastert IC, Meiners
36. McDonald JW, Silverstein FS, Johnston MV. Magnesium reLC. Correlation between the degree of periventricular
duces iV-methyl-D-aspartate (NMDA)-mediated brain injury
leukomalacia diagnosed using cranial ultrasound and MRI
in perinatal rats. Neurosci Lett 1990;109:234-8
later in infancy in children with cerebral palsy. Neuropedi37. McDonald HM, Mulligan, JC, Allen AC, Taylor PM. Neonatal
asphyxia. I. Relationship of obstetric and neonatal complica18. Rodgers B, Msall M, Owens T, Guernsey K, Brody K, Buck G,
tions to neonatal mortality in 38,405 consecutive deliveries. J
Hudak M. Cystic periventricular leukomalacia and type of
cerebral palsy in preterm infants. J Pediatr 1994;124:Sl-8
38. Nelson KB, Ellenberg, JH. Antecedents of cerebral palsy.
19. Fujimoto S, Yamaguchi N, Togari H, Wada Y, Yokochi K.
Multivariate analysis of risk. NEJM 1986;315:81-6
Downloaded by: NYU. Copyrighted material.