-
7/28/2019 Umbilical Artery Pulsatility Index and Fetal
Abdominal
1/5
Ultrasound Obstet Gynecol2011; 38: 538542Published online in
Wiley Online Library (wileyonlinelibrary.com). DOI:
10.1002/uog.8947
Umbilical artery pulsatility index and fetal
abdominalcircumference in isolated gastroschisis
U. HUSSAIN*, A. DAEMEN, H. MISSFELDER-LOBOS*, B. DE MOOR, D.
TIMMERMAN,
T. BOURNE and C. LEES**Division of Fetal Medicine, Addenbrookes
Hospital, Cambridge, UK; Department of Electrical Engineering
(ESAT), KatholiekeUniversiteit, Leuven, Belgium; Department of
Obstetrics and Gynecology, University Hospitals, KU Leuven, Leuven,
Belgium; Instituteof Development and Reproductive Biology (IRDB),
Imperial College London, Hammersmith Campus, London, UK
K E Y W O R D S: abdominal circumference; FLDA; gastroschisis;
umbilical artery Doppler
ABSTRACT
Objectives To investigate changes in abdominal circum-
ference (AC) and umbilical artery pulsatility index (UA-PI) with
gestation in fetuses with isolated gastroschisis,and to determine
whether a relationship exists betweenUA-PI and fetal AC.
Methods Data from 58 pregnancies with isolated gas-troschisis
diagnosed at between 24 and 36 weeks
gestation were included in the study. Z-scores werecalculated
with respect to expected UA-PI values in nor-mal pregnancies after
log-transformation. AC-Z-scoreswere calculated with respect to
expected size in normal
pregnancies according to a standard chart. Functional lin-ear
discriminant analysis (FLDA) was applied to generate
50th, 5th and 95th percentile curves for changes in bothAC and
UA-PI with gestational age in fetuses with gas-troschisis. These
curves were compared with the standardcurves, as were the means.
UA-PI was also plotted againstAC. For this relationship, a robust
Spearman correlationcoefficient was obtained with FLDA.
Results In fetuses with gastroschisis, there was a
highlysignificant negative correlation between UA-PI and
AC,normalized for gestation using Z-scores (median corre-lation
coefficient, 0.289; median P = 0.000023). More-over, compared with
standard curves AC was lower andUA-PI higher in the gestational-age
range studied. Boththe AC and UA-PI curves showed a significantly
differentrate of change with gestation compared with the
normalranges. The mean values for fetuses with
gastroschisiscompared with the standard AC and UA-PI range
curveswere significantly different for AC throughout gestation,and
for UA-PI from 32 weeks gestation.
Conclusions In fetal gastroschisis, it is well known thatAC
tends to be smaller, though UA-PI has not been
reported to be abnormal in any consistent way. There isa clear
relationship between the fetuss AC for gestationand UA-PI, which is
not the case for normally grown
fetuses. The data suggest that the growth restriction seenin
gastroschisis may be explained by hypoxia, and notsimply by the
classical explanation of extra-abdominaldisplacement of the
abdominal viscera. Copyright 2011ISUOG. Published by John Wiley
& Sons, Ltd.
INTRODUCTION
Gastroschisis is an uncommon condition, although itsprevalence
is rising and it is now thought to affect 1 in3000 pregnancies1.
The condition is characterized by fail-ure of the anterior
abdominal wall to close; the intestineand, occasionally, other
abdominal viscera are situatedoutside the fetal abdomen, exposed to
amniotic fluid. Thecondition is not normally associated with other
congen-ital defects or aneuploidy, and the outlook after
surgicalrepair is generally good.
It has been known for many years that gastroschi-sis affects
fetal growth, many publications reporting anassociation of
gastroschisis with small-for-gestational-ageinfants. A recent study
reported that 38% of affectedbabies were born weighing less than
the 10th percentilefor gestation2. The smallness of these babies
has gen-erally been attributed to the abdomen growing poorly
on account of the lack of distension by extra-abdominalcontents.
More recently, the growth parameters biparietaldiameter and head
circumference have been shown to besymmetrically small by
derivation of growth charts forbabies with gastroschisis3.
In our clinical practice, we routinely record umbil-ical artery
(UA) Doppler parameters for babies withgastroschisis, and noticed
that the UA pulsatility index(UA-PI) was frequently high in these
babies. We therefore
Correspondence to: Dr C. Lees, Division of Fetal Medicine,
Addenbrookes Hospital, Cambridge, CB2 2QQ, UK
(e-mail: [email protected])
Accepted: 12 January 2011
Copyright 2011 ISUOG. Published by John Wiley & Sons, Ltd.
ORIGINAL PAPER
-
7/28/2019 Umbilical Artery Pulsatility Index and Fetal
Abdominal
2/5
UA-PI and fetal AC in isolated gastroschisis 539
undertook to investigate the relationship between UA-PIand
change in fetal abdominal circumference (AC) inpregnant women
referred for ongoing care for apparentlyisolated fetal
gastroschisis. To do this, we normalized ACand UA-PI data for
gestational age. In order to correctfor codependency of repeat
measurements in the samesubject, we used functional linear
discriminant analysis
(FLDA), a statistical technique that has recently beenapplied to
biomedicine in the context of first-trimesterfetal growth4,5.
METHODS
All women in the study had a viable ongoing pregnancyaffected by
fetal gastroschisis with no other prenatallyidentified abnormality,
and were referred for ongoingcare and delivery at Addenbrookes
Hospital at a varietyof gestational ages. The data were recorded
prospec-tively as part of an ongoing trust-registered clinical
audit relating to fetal growth and outcome in antena-tally
diagnosed abdominal-wall abnormalities, and wereanonymized before
analysis.
Fifty-eight women were included in the study, of whom52 had two
or more scans with complete AC and UA-PIdata: 10 had 2 scans, 9 had
3, 17 had 4, 10 had 5, 4 had6, 1 had 7 and 1 had 8.
Change of AC or UA-PI with gestation
Firstly, and taking no account of data dependency, wederived
scatterplots to show the relationship between AC
and UA-PI and gestational age. These plots included alldata
points. To take the codependency between scans ofthe same subject
into account without diminishing thestrength of our data set and
without having to excludedependent data points, we subsequently
applied FLDA4.With this technique a mean coefficient vector or
curveis obtained for the considered relationship between sizeand
Doppler index. The reliable portion of the FLDAcurves was defined
using the concept of robust bivariateboxplots6.
To compare these FLDA curves with the AC and UA-PIchanges in
normal pregnancies, 1000 data points were
randomly sampled from each FLDA curve based on itsmean and SD.
This means that a gestational-age valuewithin the range 154 266
days was randomly selected,repeated 1000 times. At each
gestational-age value a cor-responding AC or UA-PI value was then
selected froma normal distribution with mean and SD specific for
thecurrent gestational age. These data points were used forthe
construction of the regression curve:
y = a1 + a2 GA+ a3 group+ a4 groupGA,
where group is a binary variable equal to 0 for the normalcurve
and 1 for the gastroschisis curve, and group GAan interaction term
between group and gestational age totest for a difference in slope
between both groups.
To confirm the results obtained by comparing thecurves, a second
approach was used for comparison.For the normal curve and the FLDA
gastroschisis curve,a normal distribution with known mean and SD
wasconstructed at four gestational ages (24, 28, 32 and36 weeks),
followed by a random sampling of 100 pointsfrom each distribution.
The two-sample t-test was per-
formed at each gestational age to verify whether thenormal
distribution and the distribution for gastroschisishad equal mean.
The construction of the distributions,data sampling and t-test were
repeated 1000 times.
Relationship between AC and UA-PI
The Spearman correlation coefficient was calculated forthe
relationship between AC and UA-PI. However, forthe majority of
women, data from multiple scans wereavailable. In the first
instance we included all data,neglecting possible codependencies
between data obtained
from the same woman at multiple time points duringher pregnancy.
A correlation coefficient between AC andUA-PI was calculated after
transformation of the datato Z-scores. For the Doppler indices, the
Z-scores werecalculated with respect to the expected Doppler
valuesin normal pregnancies after log-transformation of
theindices7. For AC, Z-scores were calculated with respectto
expected size in normal pregnancies according to themethod of
Chitty et al.8.
We then converted the relational curve between AC andUA-PI
within the 95%-range into the FLDA coefficient,definedas a
correlation coefficient based on all the individ-
ual curves of the subjects. First we applied FLDA resultsin
spline curves fitted through all data points belongingto each
individual subject. The median number of datapoints per subject was
calculated (x) and subsequently xdata points were randomly selected
from each individualcurve, resulting in x times n data points,
where n is thenumber of subjects in the data set. Finally, the
Spear-man correlation coefficient was calculated for these
datapoints. The random selection of x data points from eachcurve
was repeated 1000 times due to the variance of thecorrelation
coefficient when different data points alongthe curves are chosen.
Median correlation coefficients andP are reported.
R E S U L T S
Fifty-eight women were included in the study, though theFLDA
dataset totalled 52 with repeat scans, in whom 204scans were
carried out. There was one intrauterine deathand one termination of
pregnancy; for the live births(n = 49) the mean birth weight was
2477 g and meangestational age at delivery was 36 + 3 weeks.
Scatterplots of AC and UA-PI in relation to gesta-tional age are
shown in Figures 1 and 2, superimposedon the normal charts of
Chitty et al.8 for AC and ofParra-Cordero et al.7 for the Doppler
indices, all shownas mean, 5th and 95th percentiles.
Copyright 2011 ISUOG. Published by John Wiley & Sons, Ltd.
Ultrasound Obstet Gynecol2011; 38: 538542.
-
7/28/2019 Umbilical Artery Pulsatility Index and Fetal
Abdominal
3/5
540 Hussain et al.
400
350
300
250
200
150
100
50
0
Abdominalcirc
umference(mm)
10 15 20 25 30 35 40 45
Gestational age (weeks)
Figure 1 Scatterplot of abdominal circumference (AC)
againstgestational age in fetuses with gastroschisis, superimposed
on thenormal AC chart (5th, 50th and 95th centiles) of Chitty et
al.8.
2.4
2.2
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
Umbilicalarterypulsatilityindex
18 20 22 24 26 28 30 32 34 36 38 40
Gestational age (weeks)
Figure 2 Scatterplot of umbilical artery pulsatility index
(UA-PI)against gestational age in fetuses with gastroschisis,
superimposedon the normal UA-PI chart (5th, 50th and 95th centiles)
ofParra-Cordero et al.7.
The 5th, 50th and 95th percentile curves obtained withFLDA are
shown in Figures 3 and 4 and are superim-
posed on standard graphs in normal usage. As statisticallyproven
below, the percentile lines in all cases fall sig-nificantly below
those for the corresponding normalpopulation for AC, and
significantly above for UA-PIfrom 32 weeks onwards.
Abdominal circumference
For AC, the interaction term between group and GAwas significant
(P = 0.0006), indicating a significantlyslower growth in the
gastroschisis group. Significanceremained with a decrease in the
number of sampleddata points to 100. For a direct comparison of
ACin normal pregnancies and gastroschisis at the fourgestational
age points 24, 28, 32 and 36 weeks, the
400
350
300
250
200
150
100
Abdominalcirc
umference(mm)
160 180 200 220 240 260 280
Gestational age (days)
Figure 3 Functional linear discriminant analysis curve
forabdominal circumference (AC) against gestational age with
5th,50th and 95th centiles ( ) in fetuses with gastroschisis
andnormal AC chart of Chitty
et al.8 with 5th, 50th and 95th centiles
( ) (SE, 14.27 mm).
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
Umb
ilicalarterypulsatilityindex
160 180 200 220 240 260 280
Gestational age (days)
Figure 4 Functional linear discriminant analysis curve for
umbilicalartery pulsatility index (UA-PI) against gestational age
with 5th,50th and 95th centiles ( ) in fetuses with gastroschisis
andnormal UA-PI chart of Parra-Cordero et al.7 with 5th, 50th and
95th
centiles ( ) (SE, 0.157).
median P-values and their interquartile range after
Bonferroni correction for multiple testing remainedsignificant
when at least 20 points were sampled at eachgestational age.
Umbilical artery pulsatility index
For UA-PI, the interaction term was significant (P
