How are fetal biometry charts formulated

Full text

asktheexpert
how are fetal biometry graphs
formulated?
BY SUSAN CAMPBELL WESTERWAY, AMS, NORTH SHORE OBSTETRIC AND GYNAECOLOGIC ULTRASOUND,
NSW.
Have you ever wondered exactly how a fetal biometry graph is formed? In this edition we ask our expert to explain how they are created.
“Knowledge of fetal size has two main
applications in obstetric practice.
The fi rst is to compare the size of
a fetus of unknown gestational age
with normal fi gures and so obtain an
estimate of the maturity of the fetus.
The second application is to compare
the size of a fetus of known gestational
age with known normals either as
a single reading to tell whether the
fetus in question is larger or smaller
than normal or, better, as a series of
readings. A series of readings is to be
preferred since it not only checks the
accuracy of a single reading but also
gives most useful information as to
the rate of growth of the fetus.” (Bill
Garrett and David Robinson, pioneers of
Australian ultrasound, 1971.)
Compared with the physical examination
of the pregnant uterus the most accurate
method for assessing and tracking
fetal size and growth is with the use of
ultrasound imaging and measuring of
the various fetal parameters. Gestational
age is determined by measuring the fetal
parameters of interest, such as a long
bone, from the frozen ultrasound image.
The collection of measurements that are
obtained are then compared with fetal
measurement charts, which relate the
observed value of a fetal parameter to
gestational age or growth. The earliest
measurement of gestational age taken in
pregnancy should usually be accepted as
the defi nitive assessment with subsequent
examinations refl ecting only fetal growth in
Graph 1. Raw Data for Femur Length.
the intervening period. Until the mid 1980’s
there were no generally accepted criteria
to follow for chart production, for example
de Crespigny et al [1] defi ned the week
of gestation as being completed weeks,
whilst Hadlock [2] used half weeks as the
end point. This factor alone gave variations
between charts of half a week.
Ultrasonic fetal measurement charts are
formulated using a number of steps:
• Data collection using either a cross
sectional or longitudinal study with equal
number of data at each week of gestation
• The raw data obtained is plotted onto a
graph with the observed gestational age on
the x-axis and measured fetal parameter
on the y-axis
• The mean value with plus and minus two
standard deviations is calculated for each
weekly interval
• A mathematical model, or regression
analysis is applied to describe the
relationship of fetal size to gestational age.
Data Collection - Longitudinal versus
Cross Sectional Analysis
The two types of studies that can be used
for fetal measurement data collection are
cross sectional and longitudinal. A cross
sectional study assumes that there is a
common growth process subject only to
random variation. Hadlock [2] defi ned a
properly designed cross sectional analysis
of fetal parameters as being one that
measures a large number of fetuses, once
only, that are evenly distributed over the
entire range of gestational age. The main
20 issue 4, 2005 soundeffects

an ultrasonic review of scanning carotid stents
advantage of this type of study is that
adequate data can be collected in a short
period of time and this was the method
used to collect the 3,800 measurements for
the ASUM 2001 charts [3]. In a longitudinal
study a number of fetuses are assessed
for size and growth serially throughout
the gestation [4]. The fi rst ultrasound
determines or confi rms the gestational
age of the fetus and any subsequent
scans map the growth of the fetus. This
mapping allows for easier detection of any
abnormal growth pattern but the two main
disadvantages of a longitudinal study are
the commitment by the pregnant woman to
return for regular ultrasound examinations
and the time needed by the sonographer to
map the growth of the individual fetuses.
Graph Formation
The collection of measurements is called
raw data, which is entered into a software
program to enable graphs to be produced.
This raw data is the most important
information of the study as it can be
manipulated in many ways to describe the
fetal parameter of interest.
Graph 1 shows an example of a chart of
raw data being constructed for the femur
length overlaid with a regression line. The
higher the curvilinear regression the better
the line will fi t the raw data.
The size of the normal fetus increases
with gestational age and it is this
relationship that requires defi nition with
the use of a mathematical model. The
most common method used to formulate
fetal measurement graphs is to put the
measured parameter (the size) plotted
on the Y-axis and the observed value
(gestational age) on the X-axis. The mean
values, with two standard deviations,
are calculated at weekly intervals.
Regression analysis is then applied to
the raw point-to-point data to produce an
analytic description. Linear, quadratic and
cubic functions are used to fi nd the best
relationship between the measured fetal
parameter and gestational age according
to least squares criteria (graphs 2 & 3).
Due to the tendency for linear regression
to average raw data, Jeanty [5] suggested
that a curvilinear (or polynomial) quadratic
or cubic regression should be used.
Given that this only indicates how well
the function reproduces the measured
data points, it is not necessarily the
most appropriate function to use when
trying to predict the gestational age
(GA) from a single measured value.
What is required is the function that
best represents the expected biological
growth pattern. The effectiveness of the
functions are evaluated by measurement
of the coeffi cient of determination R2.
The nearer this coeffi cient is to one, the
better the correlation. Predicted parameter
values / GA are calculated using the most
appropriate models. When the relationship
between the two variables is known it is
then possible to predict gestational age
from a given parameter and vice versa.
Graph 2 is the mean of the raw data.
Simple averaging has been applied for
each week of gestation by taking the sum
of the total data points at each completed
week and dividing by the number of
data for that week of gestation. Graph 3
shows the linear regression, which misses
many of the data points, whilst graph 4
demonstrates how the application of a
polynomial regression, in this case cubic,
closely follows the original raw data.
Graph 2. Mean of Raw Data.
Graph 3. Mean Data with Linear Regression Overlay.
Graph 4. Mean Data with Cubic Regression Overlay.
how are fetal biometry graphs formulated?
soundeffects issue 4, 2005 21

Standard Deviations and Confi dence
Intervals
The accuracy of the fetal parameter
calculation is described using standard
deviations (graphs 5 & 6) and standard
error. Confi dence intervals are calculated
for each curve, using standard error, and
set at the 90th or 95th percentile.
95% confi dence interval
= mean +/- (standard error x critical t)
critical t is read from a look up chart.
standard error (se)
= standard deviation divided by the
square root of the number of samples
in the set.
The 95% confi dence interval (CI) provides
an estimate of the true population value of
the mean, that is, there is a 95% chance
that the true population mean lies within
these limits. The confi dence interval is a
statement about the mean population, not
the spread of scores and as such is used
to identify for example a macrosomic fetus
(> 95th percentile).
The most important thing to remember
when using fetal measurement charts is to
use a chart formulated from the population
that is to be examined and to use the same
imaging plane and measuring technique
used by the original author of the graph
being used.
References
1. de Crespigny L, Speirs A. A New Look
at Biparietal Diameter. Aust NZ J Obstet
Gynaecol. 1989; 29:26.
2. Hadlock F. Ultrasound determination
of menstrual age. In Callen P (editor)
Ultrasonography in Obstetrics and
Gynaecology. Philadelphia: W B Saunders;
1991.
3. Westerway SLC. Ultrasonic fetal
measurements - new Australian standards
for the new millenium. Aust NZ J Obstet
Gynaecol. 2000: 40(3):297.
4. Deter R, Harrist R. Growth standards for
anatomic measurements and growth rates
derived from longitudinal studies of normal
fetal growth. J Clin Ultrasound. 1992; 20:
381-388.
5. Jeanty P. Fetal Biometry. In Fleischer A,
Romero R, Manning F, Jeanty P, (editors).
Principles and Practice of Ultrasonography
in Obstetrics and Gynaecology. 6th ed.
New York: McGraw Hill; 2001.
Graph 5. Mean Data with +/- 2 Standard Deviations.
Graph 6. Final Graph with Regression and Standard Deviations.
OH&S POSTERS
A limited number of laminated posters
“Prevent Injury - Stretching for Sonographers” & “Prevent Injury - Posture is Important”
are available for purchase from the National Office
For further details contact the Member Services Officer on: 03 9585 2996
or via email on: members@A-S-A.com.au
how are fetal biometry graphs formulated?
22 issue 4, 2005 soundeffects