256 Am J Clin Pathol 2006;126:256-265

256 DOI: 10.1309/FK9D5WBA1UEPT5BB

American Society for Clinical Pathology
Anatomic Pathology / FETAL AND NEONATAL AUTOPSY STANDARDS
Quantitative Standards for Fetal and Neonatal Autopsy
John G. Archie,
1
Julianne S. Collins, PhD,
1
Robert Roger Lebel, MD
2
Key words: Fetal pathology; Fetal autopsy; Growth curves; Reference ranges; Regression analysis
DOI: 10.1309/FK9D5WBA1UEPT5BB
A b s t r a c t
Growth curves are essential for determining
whether growth parameters lie within normal ranges. In
the case of fetal and neonatal autopsy, relevant data are
scattered across many publications, and few sources
examine a large enough sample to be considered
definitive. To ameliorate these inadequacies,
regressions were created incorporating data from
multiple sources both to increase accuracy and to
condense available data into a single standard. When
measurements were not well studied, the best available
published standards are given. These regressions
provide a valuable tool for clinicians who need to
understand the significance of measurements obtained
during autopsy.
Pathologists obtain and report numerous anatomic details
as part of fetal and neonatal autopsies to identify specific diag-
noses and to allow meaningful assessment of recurrence risk
for adverse pregnancy outcomes. Family and personal health
history of the parents, obstetric events, potentially teratogenic
exposures, anthropometric measurements, external examina-
tion, dissection, radiography, histologic examination, and lab-
oratory studies all have a role in the search for answers.
Nevertheless, even after an exhaustive effort, a clinician may
not determine a specific cause in about one third of cases.
1
Lack of a specific diagnosis leaves family members with an
inadequate explanation for their misfortune and may con-
tribute to dysfunctional grieving.
2
Professionals strive to discover specific syndromic diag-
noses because families are better served by having direct
insight into future risks. Consequently, the interpretation of
the masses of data obtained at autopsy is of paramount impor-
tance, and without normative tables, mere clinical impressions
may overinterpret or overlook signs of hypoplasia, hypertro-
phy, and other potentially important findings.
Normative data available for such interpretations are nei-
ther extensive nor easily available. Consequently, most practi-
tioners depend on a single source or simply report the weights
and measures without offering interpretation. A review of the
literature revealed a short list of major publications offering
data for such interpretations. These findings typically are sum-
mary data in which the means and, often, SDs for a particular
measurement are given by each week or fortnight of gestation.
Furthermore, sample size for each week or fortnight some-
times is omitted. The observed means occasionally do not
increase as smoothly as one would hope for growth curves,
and matters are far worse for SDs.
Am J Clin Pathol 2006;126:256-265 257
257 DOI: 10.1309/FK9D5WBA1UEPT5BB 257
American Society for Clinical Pathology
Anatomic Pathology / ORIGINAL ARTICLE
To illustrate the discrepancies between standards, consid-
er a 26-week fetus whose brain weighs 130 g. Assuming a
normal distribution, one can use existing standards to estimate
percentiles for the organ weight, but depending on the stan-
dard used, the estimate varies: 56th,
3
69th,
4
98th,
5
95th,
6
81st,
7
and 88th percentile.
7
Thus, a pathologist might conclude that
the brain weight was abnormally high or well within the nor-
mal range. This problem can be reduced by making 2 reason-
able assumptions. First, the data given by various authors are
only an approximation of the correct values and tend to vary
around the correct values. Second, the correct values follow a
trend (eg, the average 12-week fetus should weigh somewhere
between the average 11-week fetus and average 13-week
fetus). Regression analysis makes both of these assumptions,
allowing the incorporation of available data into a single
curve, thereby providing an opportunity to interpret autopsy
findings against many sources and reducing inaccuracies from
any single source. Thus, standards constructed by regressions
provide an invaluable tool for pathologists and offer a signifi-
cant improvement over existing published data.
Materials and Methods
Selection of Studies
A literature review revealed publications listing weights
and linear measurements by week of gestation Table 1. In
each study, paired organs (lungs, adrenals, and kidneys) were
weighed in aggregate, and all measurements were derived from
nonmacerated tissue. From these data, a selection of clinically
useful measurements was chosen and divided into 2 groups.
The first group includes measurements in which many
summary data have been published or the only available data
are summary data (weight of body and internal organs, crown-
heel length, crown-rump length, and foot length). For these
measurements, all published summary data obtained during
autopsy were selected for incorporation into regressions.
The second group includes measurements in which the
best or only data come from published studies using polynomi-
al regressions (head circumference, biparietal diameter, outer
canthal distance, inner canthal distance, philtrum length, chest
circumference, internipple distance, abdominal circumference,
small intestine length, and large intestine length). Because
these regressions could not be improved on by the method pre-
sented herein, they are included for convenience and to show
their concordance with other published data, where possible.
Two measurements did not fall neatly into either of these
groups. First, no normative autopsy data were found for hand
length at greater ages; consequently, normal liveborn data
from Madhulika et al
15
were combined with data from
Chambers et al
9
for incorporation into a regression. Second,
the data given by Chambers et al
9
for abdominal circumfer-
ence are a linear regression, but because the data agree with
other data from Bhat et al,
8
the Chambers et al
9
regression was
extended to cover a larger range.
Regression Analysis
Because the mechanisms of fetal growth are likely to include
many factors and not follow a simple equation, a nonparametric
method seems best suited to model these data. For data in the
first group, locally weighted polynomial regressions were con-
structed by using the loess function of the statistical program-
ming language R.
19
For each measurement, 2 scattergrams were
Table 1
Country of Origin, Data Type, and Population Studied for Each Data Set
*
Data Set Country Data Type Population Studied
Bhat et al
8
Zambia Summary Normal term infants
Chambers et al
9
Australia Linear Autopsy
Cussen et al
3
Australia Summary Autopsy
FitzSimmons et al
10
United States Linear Autopsy
Gruenwald and Minh (autopsy)
4
United States Summary Autopsy
Gruenwald and Minh (survivors)
11
United States Summary Consecutive single births of infants discharged alive
Guihard-Costa et al
12
France Polynomial Autopsy
Hansen et al
5
United States Summary Autopsy
Kulkarni et al
13
and Kulkarni and Rajendran
14
India Summary Consecutive liveborn infants
Larroche
6
France Summary Autopsy
Madhulika et al
15
India Summary Liveborn infants with no congenital anomalies
Merlob et al
16
Israel Summary Liveborn infants with no congenital anomalies
Munsick
17
United States Summary Electively aborted fetuses
Potter (liveborn)
18
United States Summary Consecutive single-born white infants
Potter (young)
18
United States Summary Not stated
Wigglesworth and Singer (stillborn)
7
United States Summary Stillborn autopsy
Wigglesworth and Singer (liveborn)
7
United States Summary Liveborn autopsy
*
For the data type column, summary indicates that means and SDs were calculated for measurements grouped by gestational age; linear indicates linear regression; and
polynomial indicates polynomial regression.
258 Am J Clin Pathol 2006;126:256-265
258 DOI: 10.1309/FK9D5WBA1UEPT5BB
American Society for Clinical Pathology
Archie et al / FETAL AND NEONATAL AUTOPSY STANDARDS
createdone for age vs measurement and one for age vs SD.
The loess function then was used to create models to fit each
scattergram. In both cases, the span parameter, which controls
the degree of smoothing, was increased gradually until all
noise seemed to disappear by visual inspection Figure 1.
For the age vs measurement curves, measurements with
smaller SDs probably were closer to the true mean value and,
thus, were weighted more. So, the points in each curve were
weighted by the inverse of the SD. (The inverse of the SE per-
haps would have been a better choice for weighting, but the
SE was not available in many cases.) The age vs SD curves
were weighted in the same manner. This approach forces the
curve to fit more tightly near smaller SDs, which is essential
for the regression to be useful at earlier ages. Furthermore, the
loess model occasionally did not fit the data well at earlier
ages. For this reason, different span parameters were applied
for fetuses younger than 19 weeks in a few cases.
Use of Existing Data
For measurements from the second group, the authors
who seemed to publish the best (or the only) data used polyno-
mial regressions. To use these models, the reference tables for
each author were fit to polynomial equations with the Gauss-
Newton algorithm implemented in the nls function of R.
19
Results
Figure 2, Figure 3, Figure 4, Figure 5, Figure 6,
Figure 7, Figure 8, Figure 9, Figure 10, Figure 11,
Figure 12, Figure 13, Figure 14, Figure 15, Figure 16,
Figure 17, Figure 18, Figure 19, Figure 20, and Figure 21
show the results for each measurement plotted against the data
from each author; the SD regressions are displayed as error bars.
No figures are shown for philtrum length,
9
abdominal circumfer-
ence,
9
large intestine length,
10
and small intestine length
10
because these data are from only 1 source. Table 2 and Table
3 show the predicted means and SDs at various ages.
The brain, heart, and liver (Figures 3, 6, and 7) show rel-
atively small ranges of normal weight, even late in gestation,
whereas the thymus, lungs, spleen, adrenals, pancreas, and
kidneys (Figures 4, 5, and 8-11) are more variable.
Furthermore, the thymus arrests in absolute growth during the
final weeks of gestation, reflecting the well-known natural
history of the gland.
15 20 25 30 35 40 45
3,000
2,000
1,000
0
W
e
i
g
h
t

(
g
)
Age (wk)
Span = 0.1
15 20 25 30 35 40 45
800
600
400
200
0
W
e
i
g
h
t


S
D

(
g
)Span = 0.2
Age (wk)
15 20 25 30 35 40 45
3,000
2,000
1,000
0
W
e
i
g
h
t

(
g
)
Span = 0.2
Age (wk)
15 20 25 30 35 40 45
800
600
400
200
0
W
e
i
g
h
t


S
D

(
g
)Span = 0.4
Age (wk)
15 20 25 30 35 40 45
3,000
2,000
1,000
0
W
e
i
g
h
t

(
g
)
Span = 0.3
Age (wk)
15 20 25 30 35 40 45
800
600
400
200
0
W
e
i
g
h
t


S
D

(
g
)Span = 0.6
Age (wk)
A D
B E
C F
Figure 1 Constructing loess regressions for body weight. Each circle represents a published age vs measurement observation,
and the lines represent the loess regressions on those data. For means (A, B, and C) and SDs (D, E, and F), the span
parameter was increased to reduce local noise without oversmoothing.
Am J Clin Pathol 2006;126:256-265 259
259 DOI: 10.1309/FK9D5WBA1UEPT5BB 259
American Society for Clinical Pathology
Anatomic Pathology / ORIGINAL ARTICLE
4,500
4,000
3,500
3,000
2,500
2,000
1,500
1,000
500
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald (autopsy)
Gruenwald (survivors)
Hansen
Potter (liveborn)
Potter (young)
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
500
450
400
350
300
250
200
150
100
50
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald
Hansen
Larroche
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
16
14
12
10
8
6
4
2
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald
Hansen
Larroche
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
80
70
60
50
40
30
20
10
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald
Hansen
Larroche
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
Figure 2 Body weight data and regression (mean 1 SD). The
curves were created with data from Cussen et al,
3
Gruenwald
and Minh,
4,11
Hansen et al,
5
Potter,
18
and Wigglesworth and
Singer.
7
The body weight regression was derived by loess
modeling; the mean and SD spans were 0.3 and 0.6,
respectively. The Potter young
18
data were not included in the
regression because there evidently was no attempt to select
normal fetuses. Similarly, the Gruenwald and Minh
survivors
11
and Potter liveborn
18
data were excluded
because the populations studied were not autopsy populations.
The liveborn curves are shown to illustrate the difference
between autopsy populations and populations of typical children.
Figure 3 Brain weight data and regression (mean 1 SD).
The curves were created with data from Cussen et al,
3
Gruenwald and Minh,
4
Hansen et al,
5
Larroche,
6
and
Wigglesworth and Singer.
7
The brain weight regression was
derived by loess modeling; the mean and SD spans were 0.3
and 0.6, respectively.
Figure 4 Thymus weight data and regression (mean 1
SD). The curves were created with data from Cussen et al,
3
Gruenwald and Minh,
4
Hansen et al,
5
Larroche,
6
and
Wigglesworth and Singer.
7
The thymus weight regression
was derived by loess modeling; the mean and SD spans were
0.4 and 0.6, respectively. For fetuses younger than 19 weeks,
the mean and SD spans were 0.2 and 0.3, respectively.
Although Hansen et al
5
did not provide an SD for the thymus
weight at 12 weeks, an SD of 0.01 was assumed.
Figure 5 Lung weight (both lungs) data and regression
(mean 1 SD). The curves were created with data from
Cussen et al,
3
Gruenwald and Minh,
4
Hansen et al,
5
Larroche,
6
and Wigglesworth and Singer.
7
The lung weight
regression was derived by loess modeling; the mean and SD
spans were 0.4 and 0.6, respectively.
260 Am J Clin Pathol 2006;126:256-265
260 DOI: 10.1309/FK9D5WBA1UEPT5BB
American Society for Clinical Pathology
Archie et al / FETAL AND NEONATAL AUTOPSY STANDARDS
35
30
25
20
15
10
5
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald
Hansen
Larroche
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
250
200
150
100
50
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald
Hansen
Larroche
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
16
14
12
10
8
6
4
2
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald
Hansen
Larroche
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
14
12
10
8
6
4
2
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald
Hansen
Larroche
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
Figure 6 Heart weight data and regression (mean 1 SD).
The curves were created with data from Cussen et al,
3
Gruenwald and Minh,
4
Hansen et al,
5
Larroche,
6
and
Wigglesworth and Singer.
7
The heart weight regression was
derived by loess modeling; the mean and SD spans were 0.4
and 0.6, respectively. For fetuses younger than 19 weeks, the
mean span was 0.2.
Figure 7 Liver weight data and regression (mean 1 SD).
The curves were created with data from Cussen et al,
3
Gruenwald and Minh,
4
Hansen et al,
5
Larroche,
6
and
Wigglesworth and Singer.
7
The liver weight regression was
derived by loess modeling; the mean and SD spans were 0.3
and 0.5, respectively.
Figure 8 Spleen weight data and regression (mean 1 SD).
The curves were created with data from Cussen et al,
3
Gruenwald and Minh,
4
Hansen et al,
5
Larroche,
6
and
Wigglesworth and Singer.
7
The spleen weight regression
was derived by loess modeling; the mean and SD spans
were 0.3 and 0.4, respectively. For fetuses younger than 19
weeks, the mean span was 0.2. Although Hansen et al
5
did
not provide an SD for the thymus weight at 12 weeks, an SD
of 0.01 was assumed.
Figure 9 Adrenal weight (both adrenals) data and
regression (mean 1 SD). The curves were created with
data from Cussen et al,
3
Gruenwald and Minh,
4
Hansen et
al,
5
Larroche,
6
and Wigglesworth and Singer.
7
The adrenal
weight regression was derived by loess modeling; the mean
and SD spans were 0.4 and 0.7, respectively. For fetuses
younger than 19 weeks, the mean and SD spans were 0.2
and 0.3, respectively.
Am J Clin Pathol 2006;126:256-265 261
261 DOI: 10.1309/FK9D5WBA1UEPT5BB 261
American Society for Clinical Pathology
Anatomic Pathology / ORIGINAL ARTICLE
6
5
4
3
2
1
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
40
35
30
25
20
15
10
5
0
W
e
i
g
h
t

(
g
)
10 15 20 25 30 35 40 45 50
Cussen
Gruenwald
Hansen
Larroche
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
600
500
400
300
200
100
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Bhat
Chambers
Cussen
Hansen
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
400
350
300
250
200
150
100
50
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Chambers
Cussen
Hansen
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
Figure 10 Pancreas weight data and regression (mean 1
SD). The curves were created with data from Wigglesworth
and Singer.
7
The pancreas weight regression was derived by
loess modeling; the mean and SD spans were 0.4 and 0.8,
respectively. Because Wigglesworth and Singer
7
did not
provide SDs for age 41 weeks in liveborn infants and stillborn
infants of 21 weeks gestation, SDs of 2.0 and 0.4,
respectively (the average of the values for the surrounding
weeks) were assumed.
Figure 11 Kidney weight (both kidneys) data and regression
(mean 1 SD). The curves were created with data from
Cussen et al,
3
Gruenwald and Minh,
4
Hansen et al,
5
Larroche,
6
and Wigglesworth and Singer.
7
The kidney weight
regression was derived by loess modeling; the mean and SD
spans were 0.3 and 0.6, respectively.
Figure 12 Crown-heel length (CHL) data and regression
(mean 1 SD). The curves were created with data from Bhat
et al,
8
Chambers et al,
9
Cussen et al,
3
Hansen et al,
5
and
Wigglesworth and Singer.
7
The CHL regression was derived
by loess modeling; the mean and SD spans were 0.4 and 0.6,
respectively. Data from Chambers et al
9
were not used in the
regression because those points are from a linear regression
model and not actual data. Furthermore, data from Bhat et al
8
were omitted from the regression because they studied
normal liveborn infants and their data do not agree well with
the other curves.
Figure 13 Crown-rump length data and regression (mean
1 SD). The curves were created with data from Chambers et
al,
9
Cussen et al,
3
Hansen et al,
5
and Wigglesworth and
Singer.
7
The crown-rump length regression was derived by
loess modeling; the mean and SD spans were 0.4 and 0.7,
respectively. Data from Chambers et al
9
were not used in the
regression because those points are from a linear regression
model and not actual data.
262 Am J Clin Pathol 2006;126:256-265
262 DOI: 10.1309/FK9D5WBA1UEPT5BB
American Society for Clinical Pathology
Archie et al / FETAL AND NEONATAL AUTOPSY STANDARDS
400
350
300
250
200
150
100
50
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Bhat
Chambers
Cussen
Guihard-Costa
Regression
Age (wk)
120
100
80
60
40
20
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Chambers
Guihard-Costa
Regression
Age (wk)
80
70
60
50
40
30
20
10
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Chambers
Guihard-Costa
Kulkarni
Regression
Age (wk)
25
20
15
10
5
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Chambers
Guihard-Costa
Kulkarni
Regression
Age (wk)
Figure 16 Outer canthal distance (OCD) data and regression
(mean 1 SD). The curves were constructed with data from
Chambers et al,
9
Guihard-Costa et al,
12
and Kulkarni et al.
13
The data from Guihard-Costa et al
12
seemed to be a close
match to the other curves, and they studied a rather large
population. So, the OCD regression is simply their model.
Figure 14 Head circumference data and regression (mean
1 SD). The curves were created with data from Bhat et al,
8
Chambers et al,
9
Cussen et al,
3
and Guihard-Costa et al.
12
The
data from Guihard-Costa et al
12
seemed to be a close match
to the other curves, and they studied a large population. So,
the head circumference regression is simply their model.
Note that Bhat et al
8
have data only for term delivery, and
their curve is relatively flat.
Figure 15 Biparietal diameter (BPD) data and regression
(mean 1 SD). The curves were created with data from
Chambers et al
9
and Guihard-Costa et al.
12
As with the head
circumference, the BPD regression is the model of Guihard-
Costa et al.
12
Figure 17 Inner canthal distance (ICD) data and regression
(mean 1 SD). The curves were constructed with data from
Chambers et al,
9
Guihard-Costa et al,
12
and Kulkarni et al.
13
As
with the outer canthal distance, the ICD regression is the
model of Guihard-Costa et al.
12
Am J Clin Pathol 2006;126:256-265 263
263 DOI: 10.1309/FK9D5WBA1UEPT5BB 263
American Society for Clinical Pathology
Anatomic Pathology / ORIGINAL ARTICLE
350
300
250
200
150
100
50
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Bhat
Chambers
Regression
Age (wk)
90
80
70
60
50
40
30
20
10
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Chambers
Guihard-Costa
Kulkarni
Regression
Age (wk)
80
70
60
50
40
30
20
10
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Chambers
Madhulika
Regression
Age (wk)
90
80
70
60
50
40
30
20
10
0
L
e
n
g
t
h

(
m
m
)
10 15 20 25 30 35 40 45 50
Chambers
Hansen
Madhulika
Merlob
Munsick
Wigglesworth (stillborn)
Wigglesworth (liveborn)
Regression
Age (wk)
Figure 18 Chest circumference data and regression. The
data from Chambers et al
9
are from a simple linear
regression. Because, when extended, the line fits nicely with
data from Bhat et al,
8
the model of Chambers et al
9
was
simply extended to include a larger range.
Figure 19 Internipple distance (IND) data and regression (mean
1 SD). The curves were constructed with data from Chambers
et al,
9
Guihard-Costa et al,
12
and Kulkarni and Rajendran.
14
As
with the outer canthal distance, the IND regression is simply
the polynomial model of Guihard-Costa et al.
12
Figure 20 Hand length (HL) data and regression (mean 1
SD). The HL regression was derived by loess modeling; the
mean and SD spans were 0.6 and 0.9, respectively. For the
purposes of weighting the mean length, data from Chambers
et al
9
were assumed to have an SD of 1; for computing SDs,
only data from Madhulika et al
15
were considered.
Figure 21 Foot length (FL) data and regression (mean 1
SD). The curves were constructed with data from Chambers
et al,
9
Hansen et al,
5
Madhulika et al,
15
Merlob et al,
16
Munsick,
17
and Wigglesworth and Singer.
7
The FL regression
was derived by loess modeling; the mean and SD spans were
0.4 and 0.8, respectively. Because data from Chambers et al
9
were a linear regression, their data were excluded from the
model. Furthermore, the data from Madhulika et al,
15
Merlob
et al,
16
and Munsick
17
were excluded because they did not
study autopsy populations.
In contrast with the aforementioned observations regarding
internal organs, most external anthropometric measurements
have narrow ranges of normal variation. The most thoroughly
studied are the crown-heel length, crown-rump length, and foot
length (Figures 12, 13, and 21). The work of Guihard-Costa et
al
12
is much more extensive than the few other authors who pro-
vide data on head circumference, biparietal diameter, outer can-
thal distance, inner canthal distance, and internipple distance
(Figures 14-17 and 19), so the Guihard-Costa data
12
have been
taken as normative. A paucity of data on philtrum length, chest
circumference (Figure 18), abdominal circumference, hand
length (Figure 20), small intestine length, and large intestine
length yields limited interpretive power.
Discussion
Figure 2 shows total body weight data derived from 5
major sources. Although the authors generally agree, one of the
264 Am J Clin Pathol 2006;126:256-265
264 DOI: 10.1309/FK9D5WBA1UEPT5BB
American Society for Clinical Pathology
Archie et al / FETAL AND NEONATAL AUTOPSY STANDARDS
most widely trusted authors, Potter,
18
has data that seem
skewed. Her liveborn data, however, are based on consecutive
observations, and no attempt was made to select for normal
individuals. While not stated explicitly, her data for earlier ages
may have been obtained similarly. In the Potter young curve,
body weights generally are smaller than expected compared
with other curves, possibly owing to an overrepresentation of
fetuses with intrauterine growth restriction. In the Potter live-
born curve, a few points seem skewed between 30 and 35
weeks, presumably owing to fewer children born at those ages,
resulting in smaller samples.
Despite the aforementioned shortcomings, Potters curves
18
and the survivor curve for Gruenwald and Minh
11
provide
some insight into the character of autopsy populations, which, by
nature, are abnormal. Potters
18
liveborn infants and the survivors
studied by Gruenwald and Minh
11
tend to weigh more than the
autopsy populations, suggesting that autopsied children tend to
be smaller and that normative data derived from autopsies are
biased toward smaller values on average. Nevertheless, authors
providing observed data to be used as standards carefully select
normal fetuses, and although the definition of normal varies
from author to author, the care taken minimizes bias.
Even so, the abnormal nature of autopsy data still affects the
mean and SD estimates. Gruenwald and Minh,
4
noting the hetero-
geneity of autopsy populations, suggested that normal ranges can
be defined as plus or minus 1 SD (16th to 84th percentile, assum-
ing a normal distribution); another reasonable choice might be
plus or minus 1.28 SDs (10th to 90th percentile, assuming a nor-
mal distribution). Possibly less significant sources of bias include
sex, race, socioeconomic status, maternal factors, and autopsy
technique, but because not all authors give extensive descriptions
of their samples, no complete discussion of these biases and their
effects on the presented standards can be provided.
For these reasons, pathologists may choose to collect their
own set of measurements for incorporation into a regression,
giving them the ability to compare a fetus with others in the same
population, examined with the same procedure, and meeting a
consistent definition of normal.
8
Nevertheless, where locally
constructed standards are unavailable, the regressions presented
herein provide an external standard incorporating the best avail-
able data, and, even if these standards have some bias, they are
far superior to none at all.
From the
1
Office of Epidemiology and
2
Center for Anatomic
Studies, Greenwood Genetic Center, Greenwood, SC.
Table 2
Expected Weights and SDs at Various Postmenstrual Gestational Ages
*
Age
(wk) Body Brain Thymus Lungs Heart Liver Spleen Adrenals Pancreas Kidneys
12 20.9 6.6 3.20 1.44 0.01 0.01 0.50 0.28 0.15 0.02 1.01 0.38 0.01 0.01 0.10 0.03 0.16 0.04
13 31.2 10.1 5.19 1.95 0.03 0.01 1.08 0.45 0.20 0.06 1.38 0.57 0.01 0.01 0.15 0.05 0.22 0.07
14 49.1 14.5 8.14 2.58 0.05 0.02 1.79 0.67 0.31 0.11 2.18 0.84 0.03 0.02 0.23 0.08 0.36 0.13
15 74.7 19.8 12.0 3.3 0.09 0.04 2.64 0.92 0.50 0.17 3.41 1.18 0.05 0.03 0.33 0.12 0.59 0.19
16 108 26 16.9 4.2 0.14 0.06 3.61 1.21 0.76 0.24 5.06 1.60 0.09 0.05 0.47 0.16 0.90 0.28
17 149 33 22.8 5.2 0.20 0.08 4.70 1.55 1.10 0.31 7.14 2.10 0.15 0.07 0.64 0.22 1.30 0.39
18 197 42 29.7 6.3 0.28 0.12 5.92 1.92 1.50 0.40 9.65 2.66 0.21 0.10 0.84 0.30 1.79 0.51
19 255 51 37.2 7.6 0.41 0.17 7.30 2.34 1.88 0.49 12.8 3.3 0.30 0.14 1.03 0.34 2.36 0.65
20 319 61 45.7 8.9 0.54 0.23 8.84 2.80 2.41 0.59 16.5 4.0 0.41 0.18 1.29 0.41 0.50 0.14 3.00 0.81
21 389 72 54.6 10.4 0.72 0.29 10.4 3.3 2.89 0.71 19.9 4.8 0.54 0.22 1.51 0.49 0.54 0.21 3.63 0.99
22 452 84 63.7 12.0 0.92 0.37 12.0 3.8 3.38 0.82 22.7 5.7 0.66 0.28 1.73 0.57 0.60 0.26 4.23 1.18
23 510 97 72.3 13.8 1.15 0.46 13.5 4.4 3.81 0.96 24.3 6.5 0.75 0.32 1.88 0.66 0.68 0.31 4.77 1.39
24 579 115 82.8 15.6 1.38 0.58 15.0 5.0 4.23 1.12 26.4 7.1 0.91 0.36 2.00 0.74 0.77 0.34 5.65 1.63
25 660 134 93.4 17.4 1.63 0.71 16.8 5.6 4.80 1.31 29.4 7.8 1.11 0.44 2.16 0.82 0.85 0.36 6.55 1.91
26 744 163 105 19 1.96 0.86 18.7 6.2 5.50 1.57 33.2 8.8 1.38 0.55 2.36 0.90 0.92 0.38 7.46 2.21
27 839 199 118 21 2.37 1.02 20.6 6.8 6.28 1.84 37.8 9.9 1.78 0.71 2.58 0.99 1.01 0.38 8.53 2.53
28 946 239 135 24 2.85 1.22 22.7 7.3 7.13 2.11 42.6 11.5 2.26 0.96 2.83 1.10 1.08 0.37 9.75 2.85
29 1,064 286 154 26 3.44 1.49 25.1 7.9 7.95 2.44 46.9 13.3 2.73 1.19 3.09 1.21 1.14 0.37 11.1 3.2
30 1,211 330 173 30 4.02 1.85 27.4 8.4 8.84 2.71 51.3 14.8 3.20 1.36 3.36 1.34 1.27 0.39 12.5 3.7
31 1,351 373 191 33 4.52 2.17 29.2 8.8 9.83 2.86 55.9 15.8 3.74 1.58 3.71 1.42 1.46 0.42 13.8 4.0
32 1,492 406 206 35 4.91 2.43 31.2 9.0 10.8 3.0 61.2 17.0 4.37 1.87 4.07 1.50 1.77 0.47 15.0 4.4
33 1,650 433 222 36 5.40 2.63 34.1 9.4 11.9 3.2 66.3 18.8 5.06 2.18 4.42 1.56 1.95 0.55 16.5 4.9
34 1,832 457 242 37 6.03 2.84 37.5 10.1 13.1 3.5 72.8 20.9 5.76 2.51 4.77 1.63 2.11 0.63 18.0 5.3
35 2,040 487 265 39 6.87 3.06 41.7 11.0 14.5 3.7 81.8 22.3 6.47 2.79 5.19 1.76 2.36 0.69 19.6 5.7
36 2,246 511 292 42 7.85 3.22 45.1 12.2 16.0 4.0 92.8 22.9 7.21 3.07 5.74 1.92 2.61 0.77 21.3 6.0
37 2,424 535 319 44 8.95 3.41 47.0 13.2 17.6 4.3 104 23 8.11 3.30 6.46 2.10 2.84 0.85 22.5 6.4
38 2,603 559 340 46 9.61 3.60 48.4 14.0 18.6 4.5 116 26 9.15 3.53 7.01 2.31 3.04 0.94 23.9 6.8
39 2,787 582 355 49 9.98 3.78 49.4 14.8 19.4 4.8 124 29 9.83 3.73 7.44 2.55 3.33 1.04 24.9 7.1
40 2,942 603 368 51 10.2 3.9 50.8 15.5 20.3 5.0 130 32 10.2 3.9 7.75 2.82 3.65 1.15 25.7 7.5
41 3,098 623 382 53 10.2 4.1 52.3 16.1 21.3 5.2 136 36 10.5 4.0 7.99 3.11 4.01 1.26 26.4 7.8
42 3,267 641 395 55 10.1 4.3 54.0 16.5 22.4 5.3 141 40 10.8 4.0 8.14 3.44 4.40 1.39 27.0 8.2
43 3,444 657 408 57 9.83 4.46 55.9 16.8 23.6 5.4 145 45 10.9 4.0 8.21 3.79 27.6 8.5
44 3,633 671 421 59 9.44 4.64 57.8 16.9 24.8 5.5 149 50 11.1 4.0 8.22 4.17 28.3 8.8
*
Lungs, adrenals, and kidneys were weighed in pairs. Data are given as mean SD. Weights are given in grams.
Am J Clin Pathol 2006;126:256-265 265
265 DOI: 10.1309/FK9D5WBA1UEPT5BB 265
American Society for Clinical Pathology
Anatomic Pathology / ORIGINAL ARTICLE
Address reprint requests to Dr Lebel: Greenwood Genetic
Center, 101 Gregor Mendel Circle, Greenwood, SC 29646.
The data used in this article and other supplementary
materials are available online (http://autopsy.jarchie.com/).
References
1. Incerpi MH, Miller DA, Samadi R, et al. Stillbirth evaluation:
what tests are needed? Am J Obstet Gynecol. 1998;178:1121-1125.
2. Lebel RR, Elejalde BR, de Elejalde MM. Pregnancy loss and
unsanctioned grief. In: VR Pine, ed. Unrecognized and
Unsanctioned Grief: the Nature and Counseling of Unacknowledged
Loss. Springfield, IL: CC Thomas; 1990:41-46.
3. Cussen L, Scurry J, Mitropoulos G, et al. Mean organ weights of
an Australian population of fetuses and infants. J Paediatr Child
Health. 1990;26:101-103.
4. Gruenwald P, Minh HN. Evaluation of body and organ weights
in perinatal pathology, I: normal standards derived from
autopsies. Am J Clin Pathol. 1960;34:247-253.
5. Hansen K, Sung CJ, Huang C, et al. Reference values for second
trimester fetal and neonatal organ weights and measurements.
Pediatr Dev Pathol. 2003;6:160-167.
6. Larroche J-C. Developmental Pathology of the Neonate. London,
England: Excerpta Medica; 1977.
7. Wigglesworth JS, Singer DB. Textbook of Fetal and Perinatal
Pathology. Amsterdam, the Netherlands: Blackwell Publishers;
1991.
8. Bhat GJ, Mukelabai K, Shastri GN, et al. Anthropometric
parameters of Zambian infants at birth. J Trop Pediatr.
1989;35:100104.
9. Chambers HM, Knowles S, Staples A, et al. Anthropometric
measurements in the second trimester fetus. Early Hum Dev.
1993;33:45-59.
10. FitzSimmons J, Chinn A, Shepard TH. Normal length of the
human fetal gastrointestinal tract. Pediatr Pathol. 1988;8:633-641.
11. Gruenwald P, Minh HN. Evaluation of body and organ weights
in perinatal pathology, II: weight of body and placenta of
surviving and of autopsied infants. Am J Obstet Gynecol.
1961;82:312-319.
12. Guihard-Costa A-M, Menez F, Delezoide AL. Standards for
dysmorphological diagnosis in human fetuses. Pediatr Dev Pathol.
2003;6:427-434.
13. Kulkarni ML, Rajendran NK, Sangam DK. Inner canthal, outer
canthal and inter pupillary distance in newborns. Indian Pediatr.
1992;29:759-763.
14. Kulkarni ML, Rajendran NK. Internipple distance in the
newborns. Indian Pediatr. 1992;29:619-620.
15. Madhulika, Kabra SK, Barar V, et al. Upper and lower limb
standards in newborn. Indian Pediatr. 1989;26:667-670.
16. Merlob P, Sivan Y, Reisner SH. Lower limb standards in
newborns. Am J Dis Child. 1984;138:140-142.
17. Munsick RA. Similarities of Negro and Caucasian fetal
extremity lengths in the interval from 9 to 20 weeks of
pregnancy. Am J Obstet Gynecol. 1987;156:183-185.
18. Potter EL. Pathology of the Fetus and Infant. 3rd ed. Chicago, IL:
Year Book Medical Publishers; 1975.
19. The R Foundation for Statistical Computing. The R Project for
Statistical Computing. 2004. Available at http://www.r-
project.org/. Accessed July 6, 2005.
Table 3
Expected Linear Measurements and SDs at Various Postmenstrual Gestational Ages
*
Age
(wk) CHL CRL HC BPD OCD ICD PL CC IND AC HL FL SIL LIL
12 93.0 9.7 76.1 6.7 67.7 11.6 19.5 3.5 14.0 4.0 6.04 1.55 14.1 3.4 9.74 1.11 194 20
13 114 11 86.7 7.8 82.1 11.7 23.2 3.6 16.7 4.0 6.93 1.56 2.80 76.7 16.3 3.5 59.8 11.3 12.3 1.1 282 39
14 134 12 97.7 8.8 96.2 11.9 26.8 3.6 19.3 4.0 7.80 1.57 3.09 86.1 18.5 3.7 68.9 13.7 15.1 1.1 370 58
15 154 14 109 10 110 12 30.4 3.7 21.9 4.0 8.64 1.58 3.39 95.4 20.7 3.9 78.0 16.1 17.9 1.1 458 77
16 174 15 121 11 123 12 33.9 3.8 24.4 4.0 9.45 1.59 3.68 105 22.9 4.1 87.2 18.6 20.9 1.2 547 96
17 193 16 133 11 136 12 37.3 3.8 26.8 4.0 10.2 1.6 3.98 114 25.0 4.3 96.3 21.0 24.0 1.2 635 115
18 212 17 145 12 149 13 40.7 3.9 29.2 4.0 11.0 1.6 4.27 124 27.2 4.4 105 23.4 27.2 1.4 723 134
19 230 18 158 13 161 13 43.9 4.0 31.5 4.1 11.7 1.6 4.57 133 29.4 4.6 115 25.8 30.5 1.5 811 152
20 247 19 171 14 173 13 47.1 4.1 33.8 4.1 12.5 1.6 4.86 142 31.6 4.8 124 28.2 33.9 1.7 900 171
21 264 19 184 14 185 13 50.2 4.1 35.9 4.1 13.1 1.7 5.16 152 33.8 5.0 133 30.6 37.2 1.9 988 190
22 278 20 195 15 196 13 53.3 4.2 38.1 4.1 13.8 1.7 5.45 161 36.0 5.1 142 32.9 40.0 2.1 1,076 209
23 291 20 204 15 207 14 56.2 4.3 40.1 4.1 14.4 1.7 5.75 170 38.2 5.3 151 35.3 41.7 2.4 1,164 228
24 303 21 213 16 218 14 59.1 4.3 42.1 4.1 15.0 1.7 6.04 180 40.4 5.5 160 37.8 43.8 2.6 1,253 247
25 316 22 223 17 228 14 61.9 4.4 44.0 4.1 15.6 1.7 6.33 189 42.6 5.7 169 40.4 46.0 3.0 1,341 266
26 328 23 232 18 238 14 64.7 4.5 45.9 4.1 16.2 1.7 6.63 199 44.8 5.8 179 43.0 48.0 3.5
27 340 26 242 19 248 14 67.3 4.5 47.7 4.2 16.7 1.7 208 47.0 6.0 45.7 50.0 3.9
28 351 30 250 21 257 14 69.9 4.6 49.4 4.2 17.2 1.7 217 49.2 6.2 48.4 52.0 4.3
29 362 33 259 24 266 15 72.4 4.7 51.1 4.2 17.7 1.7 227 51.4 6.4 51.0 54.1 4.9
30 374 35 267 27 275 15 74.8 4.8 52.7 4.2 18.1 1.8 236 53.5 6.6 53.4 56.2 5.4
31 386 37 276 30 283 15 77.2 4.8 54.2 4.2 18.6 1.8 245 55.7 6.7 55.6 58.2 6.0
32 397 38 284 32 291 15 79.4 4.9 55.7 4.2 19.0 1.8 255 57.9 6.9 57.6 2.1 60.4 6.3
33 408 40 292 33 298 15 81.6 5.0 57.1 4.2 19.3 1.8 264 60.1 7.1 59.2 2.9 62.5 6.4
34 419 41 301 33 306 16 83.7 5.0 58.5 4.2 19.7 1.8 274 62.3 7.3 60.5 4.1 64.7 6.6
35 432 43 310 33 312 16 85.8 5.1 59.8 4.2 20.0 1.8 283 64.5 7.4 61.9 4.8 66.9 6.7
36 444 44 318 33 319 16 87.8 5.2 61.0 4.3 20.3 1.8 292 66.7 7.6 63.2 5.1 69.2 6.7
37 457 44 327 32 325 16 89.6 5.2 62.2 4.3 20.6 1.8 302 68.9 7.8 64.4 5.4 71.3 6.7
38 470 44 336 32 331 16 91.5 5.3 63.3 4.3 20.8 1.9 311 71.1 8.0 65.4 5.5 73.4 6.7
39 482 44 344 30 336 17 93.2 5.4 64.3 4.3 21.0 1.9 321 73.3 8.1 66.3 5.5 75.6 6.7
40 493 42 352 29 342 17 94.9 5.5 65.3 4.3 21.2 1.9 330 75.5 8.3 67.0 5.3 77.8 6.6
41 505 41 360 27 346 17 96.4 5.5 66.2 4.3 21.4 1.9 77.7 8.5 67.6 4.9 80.1 6.4
42 516 38 367 25 351 17 97.9 5.6 67.0 4.3 21.5 1.9 79.9 8.7 68.0 4.2 82.5 6.3
AC, abdominal circumference; BPD, biparietal diameter; CC, chest circumference; CHL, crown-heel length; CRL, crown-rump length; FL, foot length; HC, head circumference; HL,
hand length; ICD, inner canthal distance; IND, internipple distance; LIL, large intestine length; OCD, outer canthal distance; PL, philtrum length; SIL, small intestine length.
*
Data are given as mean or mean SD. Linear measurements are given in millimeters.