American Journal of Epidemiology
© The Author 2013. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of
Public Health. All rights reserved. For permissions, please e-mail: [email protected].
Vol. 178, No. 8
DOI: 10.1093/aje/kwt176
Advance Access publication:
August 21, 2013
Practice of Epidemiology
Customized Versus Population-Based Birth Weight Charts for the Detection of
Neonatal Growth and Perinatal Morbidity in a Cross-Sectional Study of Term
Neonates
Angela E. Carberry*, Camille H. Raynes-Greenow, Robin M. Turner, and Heather E. Jeffery
* Correspondence to Dr. Angela Carberry, Sydney School of Public Health, University of Sydney, Edward Ford Building (A27), Sydney 2006,
New South Wales, Australia (e-mail: [email protected]).
Initially submitted November 7, 2012; accepted for publication April 17, 2013.
Customized birth weight charts that incorporate maternal characteristics are now being adopted into clinical
practice. However, there is controversy surrounding the value of these charts in the prediction of growth and perinatal outcomes. The objective of this study was to assess the use of customized charts in predicting growth,
defined by body fat percentage, and perinatal morbidity. A total of 581 term (≥37 weeks’ gestation) neonates born
in Sydney, Australia, in 2010 were included. Body fat percentage measurements were taken by using air displacement plethysmography. Objective composite measurements of perinatal morbidity were used to identify neonates
who had poor outcomes; these data were extracted from medical records. The value of customized charts was
assessed by calculating positive predictive values, negative predictive values, and odds ratios with 95% confidence intervals. Customized versus population-based charts did not improve the prediction of either low body fat
percentage (59% vs. 66% positive predictive value and 87% vs. 89% negative predictive value, respectively) or
high body fat percentage (48% vs. 53% positive predictive value and 90% vs. 89% negative predictive value,
respectively). Customized charts were not better than population-based charts at predicting perinatal morbidity
(for customized charts, odds ratio = 1.02, 95% confidence interval: 1.01, 1.04; for population-based charts, odds
ratio = 1.03, 95% confidence interval: 1.01, 1.05) per percentile decrease in birth weight. Customized birth weight
charts do not provide significant improvements over population-based charts in predicting neonatal growth and
morbidity.
birth weight; fetal development; infant; infant nutrition disorders; malnutrition; newborn; nutrition assessment;
perinatal care; predictive value of tests
Abbreviations: ADP, air displacement plethysmography; LGA, large for gestational age; ROC, receiver operating characteristic;
SGA, small for gestational age.
Editor’s note: An invited commentary on this article
appears on page 1309.
or as having weights that are less than the 10th birth weight
percentile, and many are in fact constitutionally small but normal. Alternatively, birth weight that is higher than the 10th
percentile does not necessarily suggest “normal” growth for
a neonate (2–4). Traditionally, growth has been assessed by
using birth weight percentiles according to population-based
birth weight charts that correct for gestational age and sex.
Recently, it has been proposed that customized birth weight
charts that account for maternal characteristics are better at
determining neonatal growth and the risk of perinatal morbidity
One of the biggest global public health challenges is the accurate assessment of neonatal growth both in utero and postnatally (1). Of particular importance is the determination of
whether a neonate has been pathologically growth restricted
in utero as opposed to being a small but healthy neonate. These
neonates are often classified as small for gestational age (SGA),
1301
Am J Epidemiol. 2013;178(8):1301–1308
1302 Carberry et al.
(5). Customized birth weight charts additionally adjust for
maternal height, weight, ethnicity, and parity, all of which are
factors known to influence growth (3); however, there is ongoing
debate as to whether this is appropriate.
The current literature is inconsistent regarding the true effectiveness of customized charts in predicting growth and risk
(1). Several studies have shown that neonates identified as
SGA by customized birth weight charts have higher associated
risks of perinatal morbidity and mortality, including stillbirth
(5, 6), compared with those identified as SGA by populationbased charts. These findings have prompted recommendations
for customized birth weight charts to be adopted for use in clinical practice worldwide (3, 5, 6). On the other hand, further
studies have shown that the use of customized birth weight charts
results in overestimation of perinatal mortality and morbidity
risks because of the inclusion of preterm neonates in the analyses (7–9). A deficiency in these studies is the lack of a “gold
standard” determination of neonatal growth.
Few studies exist that assess the efficacy of customized birth
weight charts against an accurate measure of neonatal growth
restriction. It is well known that birth weight alone is not sufficient as a measure of neonatal growth. Key indicators of the
adequacy of neonatal growth, including body fat and lean muscle
tissue, are not evident by using measures of birth weight (10).
Thus, measurement of newborn body fat percentage can provide an accurate assessment of substrate depletion and an estimate of morbidity risk, such as failure of alternative substrates
for brain metabolism, hypothermia, and feeding difficulties (11).
Currently, there are no studies available that investigate neonatal body fat percentage and associated perinatal morbidity
in relationship to customized versus population-based birth
weight chart percentiles in a population of term neonates. Therefore, the aims of this study were to a) predict body fat percentage as an indicator of growth, and specifically undernutrition and
overnutrition, by using customized versus population-based birth
weight charts, and b) determine which birth weight charts better
predict perinatal morbidity.
MATERIALS AND METHODS
Sampling and design
By using a cross-sectional design, we recruited subjects from
term singleton neonates born at the Royal Prince Alfred Hospital in Sydney, Australia, between September and October
2010. Royal Prince Alfred Hospital is a major public teaching
hospital with a predominantly inner-urban, multicultural population and more than 5,000 obstetric deliveries per year.
Eligible subjects were healthy neonates born at term (37–42
weeks’ gestation), who were recruited within 48 hours after
birth. Exclusions included neonates with major congenital
abnormalities, those who were part of multiple births, preterm
neonates, or those with admissions to the neonatal intensive
care unit for longer than 48 hours, because illness or invasive
monitoring prevented measurement collection. The sample
size was based on the main study goal of estimating the number
of neonates with low body fat percentage (11); a minimum
sample size of 384 was required to provide a 95% confidence
interval of plus or minus 3%, assuming a 10% prevalence of
neonates with low body fat percentage, producing 95% confidence interval bounds of 7% and 13% (11). A sample size
of 384 ensured the study was sufficiently powered at 90% to
detect differences between an area under the curve of 0.80 and
0.85 (12).
Data collection
Neonatal data included birth weight (in grams), length, sex,
gestational age (in weeks) and body fat percentage. SGA was
defined as less than the 10th percentile of birth weight, and
large for gestational age (LGA) was defined as greater than the
90th percentile of birth weight (13). Body fat percentage measurements were performed within the first 48 hours after birth
by using the PEA POD air displacement plethysmography
(ADP) system (COSMED USA, Inc., Concord, California).
Before testing, the infant’s clothes were removed and a head
cap was applied or the hair smoothed to reduce the amount of
air behaving isothermically. The infant’s mass was measured
by using the integrated scale. Mass and volume calibrations
took into account the presence of 2 hospital identification
bracelets and an umbilical cord clip on each neonate at birth.
The body fat percentage was computed on the basis of a 2compartment model—fat and fat-free compartments (14).
Neonatal length measurements required for ADP calculations were obtained to the nearest 0.1 cm from heel to crown by
2 trained personnel by using an Easy-Glide bearing infantometer (Perspective Enterprises, Inc., Portage, Michigan). The
intraclass correlation coefficient was 0.968, showing good
interrater reliability (15). Birth weight was measured to within
0.1 g by using the digital scales on the PEA POD system. The
method for measuring length and anthropometry was taught
and assessed by a neonatologist (H.E.J.), and competency was
confirmed before data collection. Gestational age (in weeks)
was calculated from the first day of the mother’s last menstrual period by using Naegele’s rule. Undernutrition (low body
fat percentage) and overnutrition (high body fat percentage)
were defined as 1 standard deviation above or below the mean
(for males, ≤4.2% or ≥12.7%; for females, ≤5.8% or ≥14.3%).
The threshold of 1 standard deviation below the mean provided
the best balance between maximum sensitivity without unnecessarily low specificity to detect morbidity compared with 1.5 standard deviations and 2 standard deviations below the mean (11).
The population-based birth weight percentile charts were
derived from population data obtained from the New South Wales
Midwives Data Collection, which includes more than 400,000
births (13). The customized birth weight percentiles were based
on the Australian version of the bulk percentiles by using the
principles of the Gestational Related Optimal Weight method
(www.gestation.net). Mothers completed a short intervieweradministered questionnaire to collect detailed demographic
data, including prepregnancy weight, height, ethnicity, and parity,
along with details regarding the dating ultrasonography conducted in the first trimester of pregnancy and the total number
of ultrasonography examinations during pregnancy. Information about height, weight, and ultrasonography details was
based on maternal recall and was checked against the medical records. We excluded cases with insufficient information
to calculate a customized birth weight percentile. Perinatal morbidity was defined as having all 3 of the following outcomes:
Am J Epidemiol. 2013;178(8):1301–1308
Customized Charts for Detecting Growth and Morbidity
hypothermia, longer length of hospital stay, and poor feeding,
as described in a previous publication (11).
1303
criteria (Table 1). The total number of neonates with low or high
body fat percentages and perinatal outcomes classified as SGA,
appropriate for gestational age, or LGA is shown in Table 2.
Statistics
Population-based versus customized birth weight charts in
predicting body fat percentage. Univariate logistic regres-
sion was used to assess whether population-based birth weight
charts or customized birth weight charts better predicted low
and high body fat percentages in the neonates. The linearity
of the associations between percentiles and the log odds of
the outcomes were assessed by using plots. The sensitivity
and specificity at varying thresholds of weight percentiles were
used to calculate the receiver operating characteristic (ROC)
curve. The ability of population-based and customized birth
weight charts to predict neonates with low and high body fat
percentages was measured by using the area under the ROC
curve. We investigated how well the commonly used thresholds for both charts, of less than the 10th birth weight percentile
(for SGA infants) and greater than the 90th birth weight percentile (for LGA infants), predicted body fat percentage by
using positive predictive values, negative predictive values,
sensitivity, and specificity. The ROC curves for low and high
body fat percentages were tested to see if they were significantly
different (16).
Population-based versus customized birth weight charts
in predicting perinatal morbidity. To determine which chart
better predicted morbidity, we conducted a univariate logistic regression assessing the association of customized and
population-based birth weight charts (continuous variables)
with perinatal morbidity (based on the composite morbidity
measure) and estimated the area under the ROC curve. For
both charts, we investigated how well categories of less than
the 10th percentile (for SGA infants) and greater than the 90th
percentile (for LGA infants) identified neonates with perinatal morbidity by calculating the sensitivity and specificity.
The ROC curves for morbidity were tested to see if they were
significantly different (16).
P values of less than 0.05 were considered statistically significant. All analyses were conducted in SPSS, version 19, software(SPSS,Inc.,Chicago,Illinois),SAS,version9.2,software
(SAS Institute, Inc., Cary, North Carolina), and Stata, version
11, software (StataCorp LP, College Station, Texas).
The study was approved by the human research ethics committees of Royal Prince Alfred Hospital and the University
of Sydney. Informed written consent was obtained from parents, and participation was voluntary.
Population-based versus customized birth weight charts
in predicting body fat percentage
The accuracy of customized and population-based percentiles in identifying infants with low body fat percentage is shown
in Figure 2. There was no significant difference between the
ROC curves for customized and population-based charts (P =
0.08). The positive and negative predictive values were similar
for customized versus population-based charts, with 59% versus 66% positive predictive value and 87% versus 89% negative predictive value, respectively. For SGA infants, the sensitivity
and specificity analyses showed minimal if any differences
between customized and population-based charts with 23%
versus 33% sensitivity and 97% versus 97% specificity, respectively. Similarly, the ROC curve demonstrating high body fat
percentage as an indication of overnutrition in customized versus population-based charts is shown in Figure 3. There was
a significant difference between the ROC curves for both charts,
with population-based charts having a better area under the
curve than customized charts (for population-based charts,
area under the curve = 0.83, 95% confidence interval: 0.79,
0.88; for customized charts, area under the curve = 0.80, 95%
confidence interval: 0.75, 0.85) (P = 0.002). The positive and
negative predictive values were similar for customized versus
population-based charts with 48% versus 53% positive predictive value and 90% versus 89% negative predictive value,
respectively. For LGA neonates, the sensitivity and specificity analyses showed nominal differences between customized
and population-based charts with 38% versus 37% sensitivity and 93% versus 94% specificity, respectively. The estimates
of the positive predictive value, negative predictive value, sensitivity, and specificity for both charts are shown in Table 3.
A comparison of customized versus population-based birth
weight charts found that, for the neonates classified as having
low body fat percentage, 19 were detected by both charts, 10
additional neonates were detected by population-based charts,
1 was detected by customized charts only, and 57 were not
detected by either chart. For neonates classified as having high
body fat percentage, 25 were detected by both charts, 5 additional neonates were detected by the population-based charts,
6 were detected by customized charts only, and 46 were missed
by both charts.
RESULTS
Population-based versus customized birth weight charts
in predicting perinatal morbidity
There were 815 term neonates born during the study period.
Of the 782 eligible neonates, 581 were enrolled in the study,
and 550 had sufficient data for the analysis (Figure 1). The
maternal and neonatal characteristics of our cohort are shown
in Table 1. Ninety-four percent of women had early ultrasonography dating scans (at <20 weeks’ gestational age), allowing
confirmation of the accuracy of gestational age. Seventy-four
percent of women had 3 or more ultrasonography scans during their pregnancies (Table 1). Of the neonates, 3.4% were
categorized with perinatal morbidity on the basis of our defined
Customized birth weight charts were associated with perinatal morbidity with an odds ratio of 1.02 (95% confidence
interval: 1.01, 1.04) per percentile decrease in birth weight.
Similarly, population-based birth weight charts were associated with perinatal morbidity with an odds ratio of 1.03 (95%
confidence interval: 1.01, 1.05), indicating that, for both charts,
for every 1-percentile decrease in weight, the odds of morbidity
increased 1.02–1.03 times; thus, the smaller the neonates, the
more likely they were to have morbidity. Model discrimination
measured by the area under the ROC curve for customized versus
Am J Epidemiol. 2013;178(8):1301–1308
1304 Carberry et al.
Figure 1. Recruitment flow chart, Sydney, Australia, September–October 2010.
population-based charts was 0.67 (95% confidence interval:
0.54, 0.80) versus 0.71 (95% confidence interval: 0.58, 0.83),
indicating moderate discrimination for both charts. There was
no significant difference between these ROC curves (P = 0.221).
The precision of customized and population-based percentiles
in identifying infants with poor perinatal morbidity is shown
in Figure 4. At the <10th-percentile threshold, the sensitivity
and specificity analyses between customized and populationbased charts were 16% versus 32% for sensitivity and 94%
versus 93% for specificity, respectively. By using the >90thpercentile threshold in the customized versus population-based
charts, sensitivity was 5.2% versus 5.3% and specificity was
88% versus 89%, respectively.
Comparison of customized versus population-based birth
weight charts when using the <10th-percentile threshold found
that, in neonates with perinatal morbidity, 3 were correctly
classified by both charts, 3 additional neonates were detected
by the population-based charts, no additional neonates were
detected by customized charts, and 13 were missed by both
charts. In addition, when using the >90th-percentile threshold, both charts detected 1 neonate with no differences found
between the 2 charts, and 18 were missed by both charts.
DISCUSSION
Customized birth weight charts have been postulated as the
best method to predict neonates at risk of morbidity, particularly morbidity related to undernutrition. However, studies have
been based on the assumption that birth weight measurement
is the best measure of growth. Body fat percentage as measured by ADP is highly accurate and is considered the gold
standard (17). We used body fat percentage and found that
customized percentiles did not improve the prediction of
undernutrition, overnutrition, or perinatal morbidity compared
with population-based charts. Our findings do not support previous findings favoring customized birth weight charts (5, 18).
Am J Epidemiol. 2013;178(8):1301–1308
Customized Charts for Detecting Growth and Morbidity
Table 1. Characteristics of Neonates Included in the Analysis,
Sydney, Australia, September–October 2010
Characteristic
Mean (SD)
%
Table 2. Number of Neonates With Low or High Body Fat
Percentages and Poor Perinatal Outcomes Classified by SGA, AGA,
and LGA by Customized and Population-Based Birth Weight Charts,
Sydney, Australia, September–October 2010
Maternal characteristics
Height, cm
No. With
No. With
No. With
No. With
Low Body
Normal
High Body
Poor
Fat
Body Fat
Fat
Outcome
Percentage Percentage Percentage
164.7 (6.8)
Prepregnancy weight, kg
62.0 (12.2)
Parity
0
55
1
34
2
7
≥3
4
a
Ethnicity
European
60
African
0.5
Middle Eastern
3
Chinese
25
Samoan/Tongan
1.5
Indian/Pakistani
5
Other
5
b
1305
Early ultrasonography dating scans
94
≥3 Ultrasonography scans during
pregnancy
74
Customized
birth weight
charts
SGAa
20
14
0
3
b
AGA
65
337
51
15
LGAc
2
32
31
1
Populationbased birth
weight
charts
SGA
29
15
0
6
AGA
57
342
52
12
LGA
1
26
30
1
Abbreviations: AGA, appropriate for gestational age; LGA, large for
gestational age; SGA, small for gestational age.
a
Less than the 10th percentile for birth weight.
b
Between the 10th and 90th percentiles for birth weight.
c
Greater than the 90th percentile for birth weight.
Fetal characteristics
Birth weight, g
Gestational age, weeks
3,428.6 (465.9)
39.3 (1.1)
Male sex
Body fat percentage at birth
51.9
9.2 (4.3)
Perinatal morbidity
3.4
Abbreviation: SD, standard deviation.
a
Ethnicity groups were based on those reported in the Australian
version of the principles of the Gestational Related Optimal Weight
method (www.gestation.net).
b
Before 20 weeks’ gestation.
To our knowledge, this is the first study to investigate customized versus population-based birth weight charts by using
ADP to measure growth by body fat percentage in term neonates. One small study by Law et al. (10) assessed the use of
customized birth weight charts in a high-risk preterm cohort.
The authors reported that customized birth weight charts
are better than population-based charts at identifying SGA
infants with low body fat percentage in a population of preterm
infants with low birth weight (10). The authors measured body
fat percentage by using ADP but, because of the prematurity
of the neonates, the measurement was not undertaken until,
on average, 10 weeks after birth, at which point it may have
been significantly different from the body fat percentage at
birth (10). Other studies have used birth weight or fetal weight
estimated by ultrasonography to determine growth. A major
difficulty with these studies is that neonates who have failed
to reach their growth potential, such as pathologically small
or small-but-healthy neonates, are not detected by using birth
weight alone (1). In our study, we were able to overcome this
weakness by using an objective diagnostic tool (ADP) to
Am J Epidemiol. 2013;178(8):1301–1308
determine body fat percentage, providing accurate anthropometric measurements without observer error. ADP has been
validated as a highly accurate tool to measure body fat percentage when compared with gold standard techniques, including
the 4-compartment model and physical and biological phantoms (17, 19–23).
Figure 2. The receiver operating characteristic curve showing how
many neonates with low body fat percentage were correctly identified
by customized versus population-based growth charts, Sydney, Australia, September–October 2010. One hundred minus the specificity
is displayed on the x-axis.
1306 Carberry et al.
Figure 3. The receiver operating characteristic curve showing how
many neonates with high body fat percentage were correctly identified by customized versus population-based growth charts, Sydney,
Australia, September–October 2010. One hundred minus the specificity is displayed on the x-axis.
Figure 4. The receiver operating characteristic curve showing how
many neonates with perinatal morbidity were correctly identified by
customized versus population-based growth charts, Sydney, Australia, September–October 2010. One hundred minus the specificity is
displayed on the x-axis.
Similarly, we were able to assess whether customized or
population-based charts were better at predicting perinatal morbidity based on a composite morbidity measure including hypothermia, extended length of hospital stay, and poor feeding. We
found no differences between the 2 charts in risk prediction by
using this conservative morbidity measure. The effectiveness of
customized birth weight charts in predicting perinatal morbidity has mixed support in the literature, and currently there are no
randomized controlled trials to assess the true benefits and harms
associated with customized versus population-based birth weight
charts (2, 18). In particular, Cha et al. (8) examined 9,052 normal singleton deliveries in South Korea and investigated
the differences between perinatal morbidity in customized versus population-based charts. The perinatal morbidity definition
included 5-minute Apgar scores below 7, admission to the neonatal intensive care unit, and composite morbidity, which included
a long list of serious morbidity outcomes including intraventricular hemorrhage, necrotizing enterocolitis, and sepsis and
was not significantly different between the 2 charts (8). Conversely, Ego et al. (24) undertook a multicenter study with
56,606 births in France and compared perinatal outcomes
between customized and population-based birth weight charts.
The perinatal outcomes included cesarean delivery before the
onset of labor, 5-minute Apgar score of less than 7, admission
Table 3. The Positive Predictive Values, Negative Predictive Values, Sensitivity, and Specificity of Customized and Population-Based Birth
Weight Charts, Sydney, Australia September–October 2010
Sensitivity, %
95% CI
Specificity, %
95% CI
Positive
Predictive
Value, %
95%
CI
Negative
Predictive
Value, %
95% CI
SGAa neonates with low
body fat percentage
23
15, 33
97
95, 98
59
42, 75
87
84, 90
LGAb neonates with high
body fat percentage
38
27, 49
93
90, 95
48
35, 59
90
87, 92
SGA neonates with low body
fat percentage
33
24, 44
97
95, 98
66
52, 80
89
86, 91
LGA neonates with high body
fat percentage
37
26, 48
94
92, 96
53
40, 66
89
87, 92
Neonatal Weight Category
by Chart Type
Customized birth weight charts
Population-based birth weight
charts
Abbreviations: CI, confidence interval; LGA, large for gestational age; SGA, small for gestational age.
a
Less than the 10th percentile for birth weight.
b
Greater than the 90th percentile for birth weight.
Am J Epidemiol. 2013;178(8):1301–1308
Customized Charts for Detecting Growth and Morbidity
to a neonatal intensive care unit, length of hospital stay, stillbirth, and neonatal death (24). The authors found better prediction of stillbirth and perinatal death in customized versus
population-based charts (24). However, their study included
births as early as 22 weeks’ gestation, which may have overestimated the perinatal risk. It has been shown in the literature that the inclusion of a preterm population skews perinatal
risk, particularly among infants classified as SGA; thus, inclusion of this group may overstate the findings (7, 25).
Our study has limitations. ADP technology is relatively new
and is not yet available in all clinical settings. Some parents
initially found the use of the machine to be intimidating, thus
reducing our total cohort size and limiting the generalizability of our findings. The generalizability of our findings is further
limited by our restriction to term births. Originally, we were most
interested in using body fat percentage to identify undernutrition in term babies who are at increased risk compared with
preterm babies of not being recognized as undernourished; thus,
we excluded preterm births. Maternal height, weight, and ultrasonography details were self-reported, which may increase
the risk of bias, although for most mothers these were crosschecked with their antenatal data. The use of a composite measure of perinatal morbidity including 3 or more outcomes (11)
may mean that some individual perinatal outcomes were not
included as part of the measure. In our population, the number
of neonates with perinatal morbidity was low (3.4%). A priori,
we excluded neonates who were admitted to the neonatal
intensive care unit for longer than 48 hours because of inability to access the neonates for data collection. We therefore may
have excluded some neonates who are most at risk, although,
of the 33 excluded neonates, only 12% were found to be SGA.
In many clinical settings in different countries, customized
birth weight charts are being introduced into clinical practice
to assess neonatal growth and the risk of perinatal morbidity
without appropriate evidence to support implementation (2, 26).
For this reason, our results are both important and novel for
clinicians globally. Our findings highlight the ineffectiveness
of customized charts in predicting aberrant growth and perinatal morbidity in a healthy term population by using body fat
percentage as an improved measure of growth. In particular,
population-based charts had a better area under the curve than
customized charts, and population-based charts detected more
neonates with low body fat percentage and subsequent morbidity. We speculate that this difference is due to customized
charts controlling for maternal weight, height, and ethnicity,
and this could be overadjusting for any intergenerational growth
stunting. A mother who is short (because of her own stunting
in fetal or early neonatal life) may have a low-birth-weight baby
that also has a low body fat percentage. The population chart
would give a low birth weight percentile, but the customized
chart would give the baby a higher birth weight percentile
because of the mother’s own low weight and height. We suggest
that this may explain why customized charts have diminished
usefulness in predicting low body fat percentage and morbidity
compared with population-based charts. Most studies that support
the use of customized birth weight charts have focused on or
included a high-risk preterm population (5). Our study is the
first in a term population to use a validated objective measure
(body fat percentage) to assess growth and a conservative perinatal morbidity score with a relatively large sample that can
Am J Epidemiol. 2013;178(8):1301–1308
1307
be generalized to other populations. Our findings are consistent with Hutcheon et al. (27), who found that customized birth
weight charts’ weight-for-gestational-age data do little to improve
accurate prediction of perinatal mortality (and in a simulated
model, customized charts did not improve the detection of growth
restriction) (9). In addition, it has been found that customized
birth weight percentiles appear to be an unnecessary addition
to clinical care with no meaningful improvements in the detection of neonates who are truly growth restricted (pathologically
small) as opposed to those who are normally grown (constitutionally small) (1). Future work, including large randomized
control trials, is needed to assess the improvement of both
antenatal and postnatal detection of risk and growth in term
neonatal populations.
ACKNOWLEDGMENTS
Author affiliations: School of Public Health, University of
Sydney, Sydney, Australia (Angela E. Carberry, Heather E.
Jeffery, Camille H. Raynes-Greenow, Robin M. Turner);
Department of Newborn Care, Royal Prince Alfred Hospital,
Sydney Australia (Angela E. Carberry, Heather E. Jeffery).
Tenix Sydney (Sydney, Australia) and an anonymous donor
from the University of Sydney funded the purchase of the PEA
POD air displacement plethysmography equipment. R.M.T.
was supported by the National Health and Medical Research
Council ( program grant 633003 to the Screening and Test
Evaluation Program). C.H.R.G. is supported by a National
Health and Medical Research Council Early Career Fellowship (511481). A.E.C. is supported by an Australian government postgraduate PhD award.
We thank the Royal Prince Alfred Hospital medical and
nursing staff and Lucia Wang, Cheryl Au, Elizabeth Hayles,
and Erin Donnelley for assisting with data collection.
Conflict of interest: none declared.
REFERENCES
1. Hutcheon JA, Walker M, Platt RW. Assessing the value of
customized birth weight percentiles. Am J Epidemiol.
2011;173(4):459–467.
2. Carberry AE, Gordon A, Bond DM, et al. Customised versus
population-based growth charts as a screening tool for
detecting small for gestational age infants in low-risk pregnant
women. Cochrane Database Syst Rev. 2011;(12):CD008549.
3. Resnik R. To customise or not to customise: That is the
question. Paediatr Perinat Epidemiol. 2011;25(1):17–19.
4. Zhang J, Merialdi M, Platt LD, et al. Defining normal and
abnormal fetal growth: promises and challenges. Am J Obstet
Gynecol. 2010;202(6):522–528.
5. Gardosi J, Figueras F, Clausson B, et al. The customised
growth potential: an international research tool to study the
epidemiology of fetal growth. Paediatr Perinat Epidemiol.
2011;25(1):2–10.
6. Resnik R. One size does not fit all. Am J Obstet Gynecol.
2007;197(3):221–222.
1308 Carberry et al.
7. Zhang X, Platt RW, Cnattingius S, et al. The use of customised
versus population-based birthweight standards in predicting
perinatal mortality. BJOG. 2007;114(4):474–477.
8. Cha HH, Lee SH, Park JS, et al. Comparison of perinatal
outcomes in small-for-gestational-age infants classified by
population-based versus customised birth weight standards.
Aust N Z J Obstet Gynaecol. 2012;52(4):348–355.
9. Hutcheon JA, Zhang X, Platt RW, et al. The case against
customised birthweight standards. Paediatr Perinat Epidemiol.
2011;25(1):11–16.
10. Law TL, Katikaneni LD, Taylor SN, et al. Customized versus
population-based growth curves: prediction of low body fat
percent at term-corrected gestational age following preterm
birth. J Matern Fetal Neonatal Med. 2012;25(7):1142–1147.
11. Carberry AE, Raynes-Greenow CH, Turner RM, et al. Is body
fat percentage a better measure of undernutrition in newborns
than birth weight percentiles? Pediatr Res. In press.
12. Pepe MS. The Statistical Evaluation of Medical Tests for
Classification and Prediction. New York, NY: Oxford
University Press; 2003.
13. Beeby PJ, Bhutap T, Taylor LK. New South Wales populationbased birthweight percentile charts. J Paediatr Child Health.
1996;32(6):512–518.
14. Carberry AE, Colditz PB, Lingwood BE. Body composition
from birth to 4.5 months in infants born to non-obese women.
Pediatr Res. 2010;68(1):84–88.
15. Wood AJ, Raynes-Greenow CH, Carberry AE, et al. Neonatal
length inaccuracies in clinical practice and related percentile
discrepancies detected by a simple length-board. J Paediatr
Child Health. 2013;49(3):199–203.
16. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the
areas under two or more correlated receiver operating
characteristic curves: a nonparametric approach. Biometrics.
1988;44(3):837–845.
17. Deierlein AL, Thornton J, Hull H, et al. An anthropometric
model to estimate neonatal fat mass using air displacement
plethysmography. Nutr Metab. 2012;9:21.
18. Gelbaya TA, Nardo LG. Customised fetal growth chart:
a systematic review. J Obstet Gynaecol. 2005;25(5):
445–450.
19. Ellis KJ, Yao M, Shypailo RJ, et al. Body-composition
assessment in infancy: air-displacement plethysmography
compared with a reference 4-compartment model. Am J Clin
Nutr. 2007;85(1):90–95.
20. Ma G, Yao M, Liu Y, et al. Validation of a new pediatric
air-displacement plethysmograph for assessing body
composition in infants. Am J Clin Nutr. 2004;79(4):
653–660.
21. Sainz RD, Urlando A. Evaluation of a new pediatric
air-displacement plethysmograph for body-composition
assessment by means of chemical analysis of bovine
tissue phantoms. Am J Clin Nutr. 2003;77(2):
364–370.
22. Urlando A, Dempster P, Aitkens S. A new air displacement
plethysmograph for the measurement of body composition in
infants. Pediatr Res. 2003;53(3):486–492.
23. Yao M, Nommsen-Rivers L, Dewey K, et al. Preliminary
evaluation of a new pediatric air displacement plethysmograph
for body composition assessment in infants. Acta Diabetol.
2003;40(suppl 1):55S–58S.
24. Ego A, Subtil D, Grange G, et al. Customized versus
population-based birth weight standards for identifying growth
restricted infants: a French multicenter study. Am J Obstet
Gynecol. 2006;194(4):1042–1049.
25. Hemming K, Hutton JL, Bonellie S. A comparison of
customized and population-based birth-weight standards: the
influence of gestational age. Eur J Obstet Gynecol Reprod
Biol. 2009;146(1):41–45.
26. Dobbins TA, Sullivan EA, Roberts CL, et al. Australian
national birthweight percentiles by sex and gestational age,
1998–2007. Med J Aust. 2012;197(5):291–294.
27. Hutcheon JA, Zhang X, Cnattingius S, et al. Customised
birthweight percentiles: Does adjusting for maternal
characteristics matter? BJOG. 2008;115(11):1397–1404.
Am J Epidemiol. 2013;178(8):1301–1308