Professional Documents
Culture Documents
The aim of this study was to define the incidence, Bassan H, Feldman HA, Limperopoulos C, Benson CB,
clinical associations, and short-term outcome of Ringer SA, Veracruz E, Soul JS, Volpe JJ, du Plessis AJ.
periventricular hemorrhagic infarction in the modern Periventricular hemorrhagic infarction: Risk factors and neo-
neonatal intensive care unit. From 5774 infants (birth natal outcome. Pediatr Neurol 2006;35:85-92.
weight <2500 gm), periventricular hemorrhagic in-
farction diagnosed by cranial ultrasound was identified
and confirmed. gestational age-matched control infants
were identified with normal cranial ultrasounds and Introduction
detailed clinical data were obtained in both groups.
Periventricular hemorrhagic infarction was confirmed in Periventricular hemorrhagic infarction is a serious com-
58 infants. Incidence was 0.1% (1500-2500 gm), 2.2% plication of germinal matrix-intraventricular hemorrhage,
(750-1500 gm), and 10% (<750 gm). Data across 6 study with adverse neurodevelopmental sequelae approaching
90% in earlier reports [1]. The impact of this lesion is best
years reveal increased incidence in infants <750 gm.
appreciated when one considers that approximately 20%
Compared with control infants, infants with periventricu-
of the 55,000 premature infants born with birth weight
lar hemorrhagic infarction had significantly greater as-
⬍1500 gm in the United States each year develop germi-
sociation with assisted conception, intrapartum factors
nal matrix-intraventricular hemorrhage [2]. Periventricular
(emergency cesarean section, low Apgar scores), early
hemorrhagic infarction is usually accompanied by a large
neonatal complications (patent ductus arteriosus, pneu-
germinal matrix-intraventricular hemorrhage and was pre-
mothorax, pulmonary hemorrhage), blood gas distur- viously viewed as a simple “extension” of intraventricular
bances, and need for pressor, volume infusion, and respi- blood into the parenchyma, hence its earlier designation as
ratory support. Neonatal mortality of this group was grade IV germinal matrix-intraventricular hemorrhage.
40% (n ⴝ 23). Survivors had longer duration of mechan- Our current understanding is that periventricular hemor-
ical ventilation and critical care stay compared with rhagic infarction is a venous hemorrhagic infarct in the
control subjects. Thirty-seven percent of survivors re- drainage area of the periventricular terminal vein [2].
quired cerebrospinal fluid shunt placement. Periventricu- Prior reports addressing periventricular hemorrhagic
lar hemorrhagic infarction remains an important neuro- infarction as an extension of germinal matrix-intraventric-
logic complication of prematurity. A growing population ular hemorrhage have had two important limitations.
of survivors is apparent among infants with birth weight Firstly, data from infants with grade III germinal matrix-
<750 gm. Multiple hemodynamic factors associated with intraventricular hemorrhage (without parenchymal in-
periventricular hemorrhagic infarction cluster in the volvement) often were combined with data from infants
intrapartum and early neonatal periods. © 2006 by who had periventricular hemorrhagic infarction into de-
Elsevier Inc. All rights reserved. scriptions of a single entity, usually “severe germinal
From *Department of Neurology, Neonatal Neurology Research Communications should be addressed to:
Group, Children’s Hospital Boston and Harvard Medical School; Dr. du Plessis; Department of Neurology, Fegan 11; Children’s
†
Clinical Research Program, Children’s Hospital Boston and Harvard Hospital; 300 Longwood Avenue; Boston, MA 02115.
Medical School; ‡Department of Radiology, Brigham and Women’s E-mail: adre.duplessis@childrens.harvard.edu
Hospital and Harvard Medical School; §Department of Neonatology, Received June 21, 2005; accepted March 7, 2006.
Brigham and Women’s Hospital and Harvard Medical School, Boston
Massachusetts.
© 2006 by Elsevier Inc. All rights reserved. Bassan et al: PVHI in Preterm Infants 85
doi:10.1016/j.pediatrneurol.2006.03.005 ● 0887-8994/06/$—see front matter
matrix-intraventricular hemorrhage” [3-5]. This clustering ultrasound throughout their neonatal intensive care unit
occurred despite the two entities’ differing pathophysiol- stay. We then matched to each study infant one control
ogies, periventricular hemorrhagic infarction being a com- infant according to gestational age at birth (⫾4 days) and
plication of germinal matrix-intraventricular hemorrhage institution of birth. Gestational age was defined according
and clearly differing in outcome; e.g., neurologic morbid- to the first day of the mother’s last normal menstrual
ity approximates 35% in grade III germinal matrix-intra- period. This study was approved by the Institutional
ventricular hemorrhage but can reach 90% in periventricu- Review Board of Brigham and Women’s Hospital, Bos-
lar hemorrhagic infarction [2,6]. Secondly, many of these ton, Massachusetts.
previous studies did not adequately control for gestational
age, thus precluding an accurate understanding of the risk
factors associated with periventricular hemorrhagic infarc- Ultrasonographic Criteria
tion.
The cranial ultrasound protocol in our neonatal inten-
In addition, with advances in modern neonatal intensive
sive care unit remained consistent over the study period,
care such as the advent of surfactant, antenatal steroids,
with a range of 1-4 studies performed during the first week
and new ventilatory support techniques, there has been a
of life. Periventricular hemorrhagic infarction was defined
concurrent decrease in the overall incidence of germinal
as an echodense lesion in the periventricular white matter
matrix-intraventricular hemorrhage and an increase in the
which is unilateral or, if bilateral, obviously asymmetric,
survival rate of small premature newborns [2]. The effect
and associated with a germinal matrix-intraventricular
of these trends on the incidence, risk factors, and outcome
hemorrhage lesion which is usually ipsilateral or larger on
of periventricular hemorrhagic infarction in premature
the ipsilateral side [2,7-9]. Cases that met criteria for
infants in the modern neonatal intensive care unit has not
isolated periventricular leukomalacia (i.e., bilateral, usu-
been defined clearly. For these reasons, the aim of the
ally symmetric echodensities in the white matter dorsolat-
present study was to define further the current character-
eral to the lateral ventricles and sometimes but not
istics of periventricular hemorrhagic infarction. The goals
invariably associated with evolving germinal matrix-intra-
were (1) to determine the current incidence of periven-
ventricular hemorrhage) were excluded [2]. Also excluded
tricular hemorrhagic infarction diagnosed by neonatal
were cases with transient echodensities (transient features
cranial ultrasound in a modern neonatal intensive care
of periventricular hemorrhagic infarction) that resolved
unit, (2) to identify and characterize perinatal factors of
without cystic residua and cases of congenital porence-
periventricular hemorrhagic infarction in the current era,
phalic cyst (periventricular cysts present within the first
and (3) to determine the short-term outcome of infants
days of life), owing to uncertainty regarding the precise
with periventricular hemorrhagic infarction.
etiologic nature of these pathologies.
Study Design
Clinical Data Collection
In this retrospective, case-controlled study, an elec-
tronic database search of all neonatal cranial ultrasound From the clinical records of all identified infants,
reports at Brigham and Women’s Hospital from January specific demographic, maternal and antenatal, labor and
1997 through December 2002 was performed, using as key intrapartum, and early postnatal data from the first 5 days
words periventricular hemorrhagic infarction, grade IV- of life were abstracted. Data were collected from the
IVH, parenchymal hemorrhage/echodensity, venous in- placental pathology reports for the presence of chorioam-
farction, and hemorrhagic infarction. For each identified nionitis, placental abruption, or placental infarction. Intra-
candidate, all neonatal cranial ultrasound studies were uterine growth retardation was defined as birth weight
reviewed simultaneously by two of the co-investigators under the 10th percentile for gestational age. Fetal heart
(C.B. and H.B.), both of whom were unaware of the rate abnormalities were defined as sustained fetal brady-
clinical course and outcome of these infants. In five cases cardia ⬍100 beats/minute, decreased variability, and late
of disparity between these two reviewers, a third reviewer decelerations as noted in maternal records. Maternal fever
(A.J.d.P.) was used as a tie-breaker. Inclusion criteria were was defined as temperature ⬎38oC within 72 hours before
all preterm infants ⬍2500 gm admitted to the neonatal delivery. Patent ductus arteriosus was included when
intensive care unit over the 6-year study period with the confirmed by echocardiogram before day of life 5 or if a
diagnosis of periventricular hemorrhagic infarction (see clinical diagnosis before day of life 5 was confirmed by
diagnostic criteria below). Exclusion criteria were term subsequent echocardiogram. Infection was defined as
infants and preterm infants with known or suspected brain positive blood and/or urine culture, or if clinical suspicion
malformation, dysmorphic features or congenital anoma- led to antibiotic treatment beyond the 48 hours routinely
lies suggestive of a genetic syndrome, metabolic disorders, used in our preterm infants. Next, survival rate and the
chromosomal abnormality, or confirmed central nervous neurologic, gastrointestinal, ophthalmologic, and cardio-
system infection within the first 6 days of life. The control respiratory complications of all survivors were docu-
group was defined as infants with normal neonatal cranial mented.
Statistical Analysis excluded for the following reasons: criteria met for
periventricular leukomalacia (n ⫽ 2); meningitis (n ⫽ 1);
The trend in the rate of incidence of periventricular congenital periventricular cysts (n ⫽ 2); and transient
hemorrhagic infarction was analyzed using logistic regres- echodensities without evolution to cystic residua (n ⫽ 5).
sion analysis. Simple comparisons between cases and Fifty-eight infants (1%) met the diagnostic criteria
control infants were made for continuous variables by for periventricular hemorrhagic infarction. The compar-
Student paired t test and for dichotomous variables by the ison group included 58 gestational age-matched control
McNemar chi-square test. To determine the independent infants with normal neonatal cranial ultrasound. Demo-
effects of the several variables associated with periven- graphic and biometric measures were comparable be-
tricular hemorrhagic infarction, multivariable models were
tween the study and control groups (Table 1). Incidence
constructed by conditional logistic regression analysis,
of periventricular hemorrhagic infarction as a function
using case-control status as the dichotomous outcome.
of gestational age and birth weight is presented in Table
Collinearity among groups of variables associated with
2 and Figure 1. A significant stepwise increase was
periventricular hemorrhagic infarction was investigated by
observed in frequency of periventricular hemorrhagic
use of principal-components analysis. SAS (SAS Inc.,
infarction with decreasing birth weight: 0.1%, 2.2%,
Cary, NC) and SPSS (SPSS Inc., Chicago, IL) software
10% for birth weights between 1500-2500 gm, 750-
were used for all computations.
1500 gm, and ⬍750 gm respectively (P ⬍ 0.001). Thus
incidence of periventricular hemorrhagic infarction
Results
among infants ⬍2500 gm was 1% (0.4-1.6%; range
Patient Demographics and Incidence over 6 years) and for infants ⬍1500 gm was 4%
(1.3-7.7%; range over 6 years). Incidence revealed a
A total of 5774 premature infants ⬍2500 gm were born significant upward trend across the 6-year study period
at the study site in the period 1997-2002. The initial only for infants ⬍750 gm (P ⫽ 0.018). For all other
electronic database search identified a total of 68 infants birth weight categories, the incidence of periventricular
with possible periventricular hemorrhagic infarction. By hemorrhagic infarction tended to fluctuate between the
the selection criteria described above, 10 infants were years with no clear significant trend over the study
Table 2. Annual incidence of periventricular hemorrhagic infarction over study period by birth weight category
Birth Weight <750 gm Birth Weight 750–1500 gm Birth Weight 1500–2500 gm Birth Weight <2500 gm
Live Live Live Live
Year Births PVHI % Births PVHI % Births PVHI % Births PVHI %
Abbreviation:
PVHI ⫽ Periventricular hemorrhagic infarction
* Odds ratio per calendar year (95% confidence interval). P value tests for significant trend.
†
Too few events to test for trend.
interval: 1.3-163.1, P ⫽ 0.03) and patent ductus arteriosus control infants. Clinical neonatal seizures occurred in 19%
(odds ratio ⫽ 34.6, 95% confidence interval: 1.6-730.7, P (n ⫽ 11) of all infants with periventricular hemorrhagic
⫽ 0.02) were observed. The remaining predictor was infarction. Bronchopulmonary dysplasia, grade III retinop-
high-frequency ventilation (odds ratio ⫽ 10.2, 95% con- athy of prematurity, necrotizing enterocolitis, and patent
fidence interval: 1.1-97, P ⫽ 0.04). Then, two reduced ductus arteriosus ligation were more common in infants
models were generated (A: assisted conception, patent with periventricular hemorrhagic infarction but were be-
ductus arteriosus, abnormal fetal heart rate, and early low the threshold for statistical significance for all vari-
minimum pH; B: assisted conception, patent ductus arte- ables (Table 5).
riosus, high-frequency ventilation) in which both assisted
conception and patent ductus arteriosus again revealed Discussion
strong and independent contributions. In these reduced
models, high-frequency ventilation could be confounded This study sought to determine the impact of recent
by abnormal fetal heart rate and minimum pH over days advances in neonatal intensive care on the incidence and
1-2. Controlling for multiple pregnancy or substituting early outcome of periventricular hemorrhagic infarction.
multiple pregnancy for assisted conception in the model We hypothesized that modern perinatal care would be
did not affect the statistical significance or odds ratio for associated with decreasing incidence and improved sur-
assisted conception. vival of infants with periventricular hemorrhagic infarc-
tion since the first major studies of this problem almost 20
Outcome in the Neonatal Intensive Care Unit years ago [1,10].
The data presented in this report suggest a decrease in
Neonatal mortality was 40% (n ⫽ 23) and 5% (n ⫽ 3) the incidence of periventricular hemorrhagic infarction
for infants in the periventricular hemorrhagic infarction when compared with studies from the previous two de-
and control groups, respectively. Twenty-two of 23 deaths cades [1,7,11-15]. However, over each of the study years
in the periventricular hemorrhagic infarction group, and 1997-2002, the incidence of periventricular hemorrhagic
two of three deaths in the control group followed a infarction for infants ⬍2500 gm fluctuated but did not
decision to withdraw life support. Death was not associ- significantly change (P ⫽ 0.23), similar to the findings of
ated with sex (14 males, 9 females; P ⫽ 0.78). Median age other recent studies [16,17]. Whether the incidence of
at death was 2.4 days (0.5-16 days, periventricular hem- periventricular hemorrhagic infarction has reached a pla-
orrhagic infarction group) vs 19.5 days (13-26 days, teau or there continues to be a decreasing incidence is yet
control group) (P ⫽ 0.054). Survivors with periventricular unclear; multicenter epidemiologic studies are required to
hemorrhagic infarction required longer periods of mechan- address this issue. Strikingly, across the 6 years of the
ical ventilation and had significantly longer length of present study there was a significant increase in the
neonatal intensive care unit stay compared with their incidence of periventricular hemorrhagic infarction in
infants with birth weight below ⬍750 gm. This trend requiring cardiac resuscitation to be at greater risk for
likely reflects the decrease in mortality among the sickest periventricular hemorrhagic infarction, although our sta-
and most immature infants, i.e., those at greatest risk for tistical power was relatively low to draw significant
periventricular hemorrhagic infarction [2]. conclusions.
The second major objective of the present study was to Among the postnatal factors, pneumothorax, pulmo-
determine whether associated factors for periventricular nary hemorrhage, and patent ductus arteriosus, all of
hemorrhagic infarction had changed given advances in which were associated with periventricular hemorrhagic
neonatal care in recent decades. A number of previous infarction in this study, have previously been implicated as
studies [4,18,19] have described certain prenatal factors causing pronounced disturbances in cerebral perfusion and
as protective against severe germinal matrix-intraventric- oxygenation [4,13,26]. Patent ductus arteriosus, an inde-
ular hemorrhage (including periventricular hemorrhagic pendent factor in the multivariate model, is associated with
infarction), such as pregnancy-induced hypertension and decreased diastolic blood flow [27] and low superior vena
administration of antenatal steroids, magnesium sulfate, cava flow [28]. Premature infants receiving high-fre-
and beta-agonists. In this study, none of these factors was quency ventilation were at increased risk for periventricu-
significantly different between case and control infants, lar hemorrhagic infarction in the multivariate analysis,
nor was administration of antenatal indomethacin, a med- confirming a previous report [29]. Although the require-
ication previously identified as increasing the risk for ment for high-frequency ventilation may reflect the sever-
severe germinal matrix-intraventricular hemorrhage [20]. ity of illness of these infants, it might also be associated
In the present study, assisted conception was found to be with an elevation of intrathoracic and hence central venous
an independent factor for periventricular hemorrhagic pressure, thereby increasing cerebral venous pressure.
infarction in the multivariate analysis, as previously sug- Cerebral venous pressure could also be increased by
gested [21,22]. The reason for this association remains pneumothorax and pulmonary hemorrhage.
unclear. Finally, contrary to other studies [22,23], this In addition, acidosis and hypercarbia, identified as
study found no association between maternal fever, doc- factors associated with periventricular hemorrhagic infarc-
umented maternal infection, pathologically confirmed tion in our and other studies [11], are known to impair
chorioamnionitis or neonatal infection and the develop- cerebral pressure autoregulation [30], rendering the brain
ment of periventricular hemorrhagic infarction. more vulnerable to hemodynamic fluctuations. This obser-
The intrapartum factors for periventricular hemorrhagic vation is particularly relevant because the need for both
infarction identified in the present study support fetal inotropic pressor support and intravascular volume ex-
distress as a common mechanism for this lesion. Specifi- panders was significantly associated with periventricular
cally, emergent cesarean section, low Apgar scores, and hemorrhagic infarction in the present study as in others
the need for respiratory resuscitation were significantly [14]. In fact, nearly all our infants with periventricular
more common in infants with periventricular hemorrhagic hemorrhagic infarction required these two interventions
infarction. These findings corroborate those from several The third objective of this study was to determine the
previous studies [13,14,24]. However, unlike others short-term outcome of periventricular hemorrhagic infarc-
[24,25] we did not find infants delivered vaginally or those tion. Although lower than the 60% mortality reported