See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/51173891

A global reference for fetal-weight and

birthweight percentiles
ARTICLE in THE LANCET MAY 2011
Impact Factor: 45.22 DOI: 10.1016/S0140-6736(11)60364-4 Source: PubMed

CITATIONS

READS

110

402

7 AUTHORS, INCLUDING:
Rafael Mikolajczyk

Joo Paulo Souza

Helmholtz Centre for Infection Research

University of So Paulo

231 PUBLICATIONS 2,571 CITATIONS

206 PUBLICATIONS 2,938 CITATIONS

SEE PROFILE

SEE PROFILE

Rintaro Mori

Ahmet Metin Glmezoglu

National Center for Child Health and Devel

World Health Organization WHO

178 PUBLICATIONS 5,978 CITATIONS

322 PUBLICATIONS 9,817 CITATIONS

SEE PROFILE

SEE PROFILE

Available from: Rintaro Mori

Retrieved on: 16 March 2016

Articles

A global reference for fetal-weight and

birthweight percentiles
Rafael T Mikolajczyk, Jun Zhang, Ana Pilar Betran, Joo Paulo Souza, Rintaro Mori, A Metin Glmezoglu, Mario Merialdi

Summary
Background Denition of small for gestational age in various populations worldwide remains a challenge. References
based on birthweight are decient for preterm births, those derived from ultrasound estimates might not be applicable
to all populations, and the individualised reference can be too complex to use in developing countries. Our aim was
to create a generic reference for fetal weight and birthweight that overcame these deciencies and could be readily
adapted to local populations.
Methods We used the fetal-weight reference developed by Hadlock and colleagues and the notion of proportionality
proposed by Gardosi and colleagues and made the weight reference easily adjustable according to the mean birthweight
at 40 weeks of gestation for any local population. For application and validation, we used data from 24 countries in Africa,
Latin America, and Asia that participated in the 200408 WHO Global Survey on Maternal and Perinatal Health
(237 025 births). We compared our reference with that of Hadlock and colleagues (non-customised) and with that of
Gardosi and colleagues (individualised). For every reference, the odds ratio (OR) of adverse perinatal outcomes (stillbirths,
neonatal deaths, referral to higher-level or special care unit, or Apgar score lower than 7 at 5 min) for infants who were
small for gestational age versus those who were not was estimated with multilevel logistic regression.
Findings OR of adverse outcomes for infants small for gestational age versus those not small for gestational age
was 159 (95% CI 153166) for the non-customised fetal-weight reference compared with 287 (273301) for our
country-specic reference, and 284 (271299) for the fully individualised reference.
Interpretation Our generic reference for fetal-weight and birthweight percentiles can be easily adapted to local
populations. It has a better ability to predict adverse perinatal outcomes than has the non-customised fetal-weight
reference, and is simpler to use than the individualised reference without loss of predictive ability.
Funding None.

Introduction
Infants who are small for gestational age, generally
dened as having birthweight below the tenth percentile
at a particular gestational week, have a higher risk of
various adverse outcomes in perinatal period1,2 and in the
long term.36 The fth or third percentiles are also used
sometimes to assess eects of more pronounced growth
restriction.7 However, denition of the status of small for
gestational age in various populations worldwide has
been a longstanding challenge.8 Although a birthweight
reference based on a mixed ethnic population of
California was proposed for international comparisons,9
whether this reference is suitable for countries with
dierent ethnic composition remains unclear.
Furthermore, birthweight references based on neonates
born at various gestational weeks have been used for
nearly half a century.10 Researchers have come to realise
that birthweight references are awed at early gestational
weeks because infants born before term are more likely
to be growth restricted. When the weight at the tenth
percentile is based on preterm neonates only, it is
substantially lower than that based on all unborn fetuses
and neonates at a particular gestational week.11 The
discrepancy could be large in early gestational weeks,
which could substantially underdiagnose small for
www.thelancet.com Vol 377 May 28, 2011

gestational age in preterm births or fetuses. From this

point of view, ultrasound-based estimated references of
fetal weight are a better choice.
Yet, ultrasound-based references were mostly developed
in women from the European Continental Ancestry
Group.12 They might not be applicable to ethnic groups
living in most developing countries. Gardosi and
colleagues1320 proposed an individualised approach that
took into account ethnic origin, maternal height and
weight, parity, and sex of the infant. However, the
corresponding references are available only for selected
ethnic groups and are based on populations (including
ethnic minorities and large groups of immigrants)
residing in developed countrieseg, the UK, USA, New
Zealand, Australia, and Spain.2124 Although individualisation was shown to improve the classication of children
small for gestational age,13 applications of the
individualised references in low-technology settings and
busy clinical practice are less practicable. Furthermore,
ethnic origin itself might partly portray maternal weight
and height. For instance, women from the European
Continental Ancestry Group are generally taller and
heavier than Asian women. Thus, in an international
comparison, whether variables other than ethnic origin
can have a substantial eect remains unknown.

Lancet 2011; 377: 185561

See Comment page 1812
Department of Clinical
Epidemiology, Bremen
Institute for Prevention
Research and Social Medicine,
Bremen, Germany
(R T Mikolajczyk MD);
Department of Public Health
Medicine, School of Public
Health, Bielefeld University,
Bielefeld, Germany
(R T Mikolajczyk); Epidemiology
Branch, Eunice Kennedy Shriver
National Institute of Child
Health and Human
Development, National
Institutes of Health, Bethesda,
MD, USA (Prof J Zhang MD);
MOE and Shanghai Key
Laboratory of Childrens
Environmental Health, Xinhua
Hospital, Shanghai Jiaotong
University School of Medicine,
Shanghai, China (J Zhang);
Department of Reproductive
Health and Research, World
Health Organization, Geneva,
Switzerland (A P Betran MD,
J P Souza MD,
A M Glmezoglu MD,
M Merialdi MD); and
Department of Global Health
Policy, Graduate School of
Medicine, University of Tokyo,
Tokyo, Japan (R Mori MD)
Correspondence to:
Dr Jun Zhang, MOE and Shanghai
Key Laboratory of Childrens
Environmental Health,
Xinhua Hospital,
Shanghai 200092, China
junjimzhang@gmail.com

1855

Articles

We aimed to create an ultrasound-based generic

global reference for fetal weight and birthweight that
would be applicable to low-income and middle-income
countries, as well as high-income countries. Such
reference is important for international studies that
address causes of small for gestational age and compare
outcomes of this disorder across dierent populations.

Methods
Creation of the generic global reference
To create a generic global reference, we adapted the
following widely accepted, ultrasound-based reference
for estimated fetal weight proposed by Hadlock and
colleagues25
Fetal weight (g) = exp(0578 + 0332 GA 000354 GA)
GA refers to gestational age in exact weeks
(eg, 39 weeks + 5 days = 397 weeks). Hadlock and

Term births
Median
(3742 weeks, %) maternal
height*
(cm, IQR)

Median
maternal
weight*
(kg, IQR)

colleagues25 used ultrasound measurements between 10 and 41 weeks of gestation of 392 pregnant
women of the European Continental Ancestry Group
living in the USA to create this optimum growth
equation. They also showed that the statistical variation
of fetal weight in a given gestational week was a
constant fraction of the mean. On the basis of this
information, they provided fetal-weight percentiles for
each gestational week.
Gardosi and colleagues13 created the individualised
reference by using the Hadlock formula and transforming
it into a so-called proportionality function (individual
weight expressed as a percentage of the expected weight
based on Hadlocks formula) and adjusting it for ethnic
group, parity, sex of the infant, and maternal height and
weight. Once the mean birthweight and standard
deviation (SD) of birthweight at 40 weeks is identied,
the mean and SD of fetal weight and birthweight for all
gestational weeks are xed; and so are the weight

Coecient of
Mean (SD)
Birthweight
reported only in birthweight at variation (%)
full 100s g* (%) 40 weeks of
gestation* (g)

Adverse
perinatal
events (%)

Deaths (%)

SGA
according to
Hadlock24 (%)

Africa
Algeria

14 775

95%

162 (159165)

72 (6580)

77%

3511 (467)

133%

40%

20%

14%

Angola

3358

59%

160 (155165)

62 (5669)

69%

3202 (463)

145%

96%

24%

19%

Congo

8354

84%

156 (152161)

59 (5565)

49%

3091 (440)

142%

57%

28%

35%

Kenya

2658

93%

158 (154163)

65 (6073)

83%

3176 (448)

141%

52%

23%

32%

Niger

7985

96%

160 (158165)

65 (5873)

51%

3103 (429)

138%

46%

31%

42%

Nigeria

7567

88%

160 (156165)

73 (6483)

97%

3298 (498)

151%

101%

39%

22%

Uganda

10 457

91%

159 (155164)

64 (5872)

95%

3336 (456)

137%

41%

24%

16%

Latin America
Argentina

7074

91%

159 (155163)

72 (6580)

26%

3494 (428)

122%

28%

09%

15%

Brazil

4847

92%

1585 (154163)

68 (6175)

10%

3331 (439)

132%

60%

15%

25%

Cuba

12 489

95%

159 (155163)

68 (6276)

34%

3374 (446)

132%

22%

11%

22%

Ecuador

11 397

93%

154 (150159)

65 (5972)

53%

3222 (422)

131%

54%

13%

31%

Mexico

19 394

91%

156 (152160)

69 (6276)

47%

3288 (432)

131%

25%

10%

24%

Nicaragua

5576

93%

155 (151158)

66 (6073)

55%

3214 (414)

129%

22%

13%

24%

Paraguay

3060

91%

158 (155161)

71 (6579)

37%

3506 (441)

126%

41%

15%

14%

13 692

92%

154 (150158)

66 (6173)

27%

3456 (422)

122%

53%

14%

15%
47%

Peru
Asia
Cambodia

5362

93%

155 (151158)

58 (5464)

91%

3126 (402)

129%

66%

19%

China

14 286

95%

159 (156162)

66 (6172)

64%

3410 (411)

121%

12%

03%

21%

India

23 960

78%

152 (150155)

55 (5059)

78%

2790 (396)

142%

88%

41%

60%

Japan

3204

95%

158 (154162)

62 (5767)

2%

3160 (357)

113%

44%

01%

44%

Nepal

8268

88%

150 (148153)

58 (5263)

87%

3016 (448)

149%

66%

21%

50%

Philippines

10 533

92%

155 (152158)

60 (5468)

48%

3052 (408)

134%

37%

16%

51%

Sri Lanka

14 708

93%

154 (150158)

60 (5467)

30%

3079 (399)

130%

18%

05%

47%

Thailand

9334

90%

156 (152160)

66 (6073)

14%

3237 (412)

127%

30%

04%

30%

Vietnam

13 167

95%

155 (152159)

60 (5565)

93%

3255 (385)

118%

11%

02%

36%

*Sample restricted to women giving birth at 40 weeks of gestation. Gestational age is measured in completed weeks. Coecient of variation is SD divided by mean birthweight times 100%. Adverse perinatal
events include any of: fresh stillbirth, dead within or after 24 h of birth, alive on 7th day postpartum but referred to a higher level or special care unit, or Apgar score lower than 7 at 5 min. Deaths include: fresh
stillbirth and dead within or after 24 h of birth.

Table 1: Characteristics of the study population in the 200408 WHO Global Survey on Maternal and Perinatal Health

1856

www.thelancet.com Vol 377 May 28, 2011

Articles

Birthweight (g)

4000

3000

2000
0

37
38 39 40
Gestational weeks
Local reference mean birthweight

41

37
38 39 40 41
Gestational weeks
Local reference 10th and 90th percentile

37
38 39 40 41
0
37
38 39 40
Gestational weeks
Gestational weeks
Empirical mean birthweight
Empirical 10th and 90th percentile

41

Figure 1: Comparison between recorded birthweight and birthweight estimated from the generic reference in the four countries with lowest fraction of
rounded birthweight
Argentina (A), Brazil (B), Japan (C), and Thailand (D) were the four countries with the lowest fraction of rounded birthweight in table 1.

percentiles under the assumption of normal distribution

of these weights.
With the approach proposed by Gardosi and
colleagues,13 we assumed that Hadlocks growth equation
could be used to derive percentiles of fetal weight in a
given gestational week for a dierent population by
anchorage of the formula to a mean birthweight at
400 weeks. Note that if gestational age is recorded in
completed weeks, the mean birthweight at 40 completed
weeks is equivalent to that at 405 weeks of gestation in
Hadlocks formula. A detailed technical approach is
described in webappendix 1 p 1.
Briey, we rst assumed that a sample of births in
40 weeks of gestation from a country existed (the
average of exact gestational age was 405 weeks). The
mean birthweight was estimated. This mean birthweight
(MW.GA=40) was then divided by the constant of 3705 g,
which is the mean birthweight at 405 weeks of
gestation in Hadlocks fetal growth equation. The
obtained ratio was assumed to be constant across
gestationie, if the mean birthweight at 40 weeks in a
particular population was 09 (or 90%) of those of the
European Continental Ancestry Group, then it was
also 09 at 34 weeks. Next, we multiplied fetal-weight
estimates based on Hadlocks reference for each
gestational week by this ratio and obtained mean fetalweight estimates across gestation for the specic
country. Finally, we used the estimate of SD from the
sample of births at 40 weeks of gestation to obtain
weight percentiles across gestation. As with Hadlocks
method, we assumed that SD expressed in percentage
of the mean weight was constant across gestation. For
easy use of the customisation for country, we created a
calculation sheet for Microsoft Oce Excel software
that could be readily used by any country or local
institution (webappendix 2). A similar approach was
used to further customise the reference by inclusion of
additional individual maternal and fetal characteristics,
up to the fully individualised reference according to
Gardosi and colleagues13 (webappendix 1 p 2).
www.thelancet.com Vol 377 May 28, 2011

Application and comparison with other references

We used data from the 200408 WHO Global Survey on
Maternal and Perinatal Health (WHOGS), for which a
detailed description has been published elsewhere.26
Briey, this was a multinational, facility-based survey
that obtained data for all delivering women in randomly
selected facilities from randomly selected countries. Data
were abstracted from medical records and transcribed
into a form by trained data collectors. These forms were
completed after delivery and before hospital discharge of
the mother. Data related to outcomes were only obtained
until discharge from the hospital. Denitions for data
extraction were described in a manual of operation and
were available at all sites. With respect to recording
birthweight, local customs of rounding were followed
ie, some hospitals rounded the birthweight to kilograms
with a decimal point (eg, 29 kg) whereas others used
grams (2900 g). Ethics approval was given by the
institutional or national review committees. Data
collection took place in 200405 in Africa and Latin
America and in 200708 in Asia.
The Global Survey26 included 290 610 births from
Africa, Latin America, and Asia. For the aim of this
analysis, the sample was restricted in consecutive steps:
singleton pregnancies (n=283 521), available information
on birthweight (n=282 606), sex of the baby (n=282 231),
maternal weight and height (n=237 633). Maternal weight
was available only from the last antenatal visit in Africa
and Latin America and before delivery in Asia.
Furthermore, the sample was restricted to data with
complete information for the following birth outcomes:
status at birth (alive, fresh stillbirth, or macerated
stillbirth), status of the newborn baby at discharge or at
the seventh day after delivery (alive and well or with
obstetric trauma, alive but referred to a higher-level or a
special care unit, dead within or after 24 h of birth) and
Apgar score less than 7 at 5 min (n=237 025, 82% of the
original sample). Events with fresh stillbirth (n=2534),
dead within (n=946) or after 24 h (n=522) of birth, on the
seventh day after delivery alive but referred to a

For more on WHOGS see

http://www.who.int/
reproductivehealth/topics/
best_practices/globalsurvey

See Online for webappendix

1857

Articles

Percentage of SGA

Among SGA (%)

Among non-SGA (%)

Adjusted OR (95% CI) AUC

Adverse perinatal events

(1)

353%

64%

36%

159 (153166)

(2)

106%

109%

39%

287 (273301)

0679
0699

(3)

107%

109%

39%

289 (275303)

0698

(4)

105%

110%

39%

288 (274302)

0698

(5)

105%

109%

39%

285 (272299)

0698

(6)

106%

109%

39%

284 (271299)

0698

Perinatal mortality
(1)

353%

25%

13%

177 (165189)

0711

(2)

106%

49%

14%

363 (339390)

0737

(3)

107%

48%

14%

365 (340391)

0737

(4)

105%

49%

14%

361 (337388)

0737

(5)

105%

49%

14%

359 (335385)

0737

(6)

106%

49%

14%

365 (340392)

0738

SGA=small for gestational age. OR=odds ratio. AUC=area under the curve. *References used for SGA classication were
dened as 10th percentile of the following denitions: in (1) original fetal growth formula developed by Hadlock and
colleagues.25 Because the global survey24 reported gestational age in completed weeks, 05 weeks were added to each
gestational age. In (2)(6) the birthweight at 40 weeks of gestation is predicted by an increasing number of variables:
(2) country, (3) country+sex of the infant, (4) country+sex of the infant+maternal height+maternal weight,
(5) country+sex of the infant+maternal height+maternal weight+maternal weight squared, (6) country+sex of the
infant+maternal height+maternal weight+maternal weight-squared+parity (reference 6 is the individualised reference
by Gardosi et al12). ORs for infants with SGA compared with those without SGA were obtained by means of random
eects logistic regression with country as random eect. Estimates for the random eects are not presented. AUC was
estimated from a model with country included as a xed eect, which was nearly the same as the random eects model
used for estimation of ORs. Adverse perinatal events include any of the following conditions: fresh stillbirth, dead
within or after 24 h of birth, alive on 7th day postpartum but referred to a higher-level or special-care unit, or Apgar
score lower than 7 at 5 min. Deaths include: fresh stillbirths and deaths within or after 24 h of birth.

Table 2: Comparison of references (16) for classication* of infants SGA with respect to the ability to
predict adverse perinatal outcomes

Gestational age* (weeks)

97th

95th

24

820

786

25

957

918

26

1110

1064

25th

10th

90th

75th

Mean

5th

3rd

1st

768

741

695

644

593

547

520

502

897

865

812

752

692

639

607

586

547

1040

1003

941

872

803

741

703

679

634

468

27

1278

1225

1198

1155

1083

1004

924

853

810

782

730

28

1461

1401

1369

1320

1238

1147

1057

975

926

894

834

29

1658

1590

1554

1498

1405

1302

1199

1106

1051

1015

947

30

1869

1792

1751

1689

1584

1468

1352

1247

1184

1144

1067

31

2091

2005

1960

1890

1773

1643

1513

1395

1325

1280

1194

32

2324

2228

2178

2100

1970

1825

1681

1551

1473

1422

1327

33

2564

2459

2403

2317

2173

2014

1854

1711

1625

1569

1464

34

2809

2694

2632

2538

2381

2206

2032

1874

1780

1719

1604

35

3056

2930

2864

2761

2590

2400

2210

2039

1937

1870

1745

36

3301

3165

3093

2983

2798

2593

2387

2203

2092

2020

1885

37

3540

3395

3318

3199

3001

2781

2561

2362

2244

2167

2021

38

3770

3615

3533

3407

3196

2961

2727

2516

2390

2308

2153

39

3987

3823

3736

3603

3380

3132

2884

2660

2527

2440

2276

40

4186

4014

3923

3783

3549

3288

3028

2794

2653

2562

2390

41

4365

4185

4090

3944

3700

3428

3157

2913

2766

2671

2492

Figure 2: Selected percentiles of fetal weight and birthweight for a population with the mean birthweight at
40 weeks of gestation of 3288 g in Mexico
Data obtained from the calculation sheet for Microsoft Oce Excel software (webappendix B). *Gestational age
in completed weeks. Standard deviation of 132% of the corresponding mean weight was used for calculation
of percentiles.

1858

Role of the funding source

There was no external funding source for the study. The
corresponding author had full access to all the data in the
study and had nal responsibility for the decision to
submit for publication.

Results

Weight percentiles for the local population

99th

higher-level or special care unit (n=2319), and none of the

above but Apgar score less than 7 at 5 min (n=4031) were
dened as adverse perinatal outcomes (total n=10 352);
and those with none of the above conditions as normal
outcomes (n=225 153). Macerated stillbirths were
excluded (n=1520) because their weight at stillbirth might
not be informative of their growth, and their gestational
age at death is unknown.
To assess the performance of dierent references of
small for gestational age, we compared the odds ratios
(OR) of adverse perinatal events between children with
and without the disorder using logistic regression with
inclusion of country as random eect. In a second
analysis, adverse events were restricted to deaths only
(fresh stillbirths, deaths at less than 24 h or more than
24 h before discharge). The following references were
compared: the original fetal growth equation proposed
by Hadlock and colleagues25 (with no adjustment), our
generic reference customised to every country (adjustment only for country) and references including an
increasing number of adjustment factors (country, sex
of infant, maternal height and weight, and parity) up to
the fully individualised reference according to Gardosi
and colleagues.13

Data from the Global Survey26 showed that birthweight

varied between countries (table 1). Mean birthweight at
40 completed weeks of gestation varied between 2790 g in
India and 3511 g in Algeria, well below the mean
birthweight for the women from the European
Continental Ancestry Group in the original study sample
used to develop the ultrasound reference by Hadlock and
colleagues.25 (3705 g). The SD of birthweight expressed in
terms of percentage of mean birthweight (coecient of
variation) across countries was 132% and varied slightly
between countries.
There was also a substantial variation with respect to
maternal weight and height as determinants of fetal
weight. In women who gave birth at 40 weeks of
gestation, median maternal height varied between
150 cm and 162 cm across the countries. Median
maternal weight ranged from 55 to 73 kg. Incidence of
adverse perinatal events was between 11% in Vietnam
and 101% in Nigeria. When fresh stillbirths or neonatal
deaths only were considered, the variation was between
01% in Japan and 41% in India. Figure 1 shows the
local references based on our global reference method in
relation to empirical data in the four countries that had
the lowest fraction of rounded birthweight (used here as
an indicator of data accuracy). In three of those countries
www.thelancet.com Vol 377 May 28, 2011

Articles

(Argentina, Brazil, and Japan), means and percentiles

for the empirical distribution and those obtained from
the local reference matched well. By contrast, in Thailand
the empirical mean birthweight at 37 weeks of gestation
was noticeably higher than the predicted birthweight.
On further examination of the gestational age distribution
at birth in Thailand, we noted that the distribution was
shifted towards early term pregnancies compared with
most other countries, suggesting that errors in the
estimation of gestational age had probably produced the
picture of less steep increase in birthweight from 37 to
39 weeks of gestation.
Application of Hadlocks reference to the population of
the Global Survey study26 resulted in a much higher
proportion of infants classied as small for gestational age
(35%) than the nominal 10% (table 2). However, after we
applied our global reference, which adjusted for country,
the proportion of infants small for gestational age
decreased to 11% and the incidence of adverse perinatal
events in those infants increased from 6% to 11%. The OR
of adverse outcomes in infants small for gestational age
compared with those not small for gestational age
increased from 16 to 29. Adding further components of
the individually customised model did not change the
results noticeably (table 2). The results were similar when
adverse outcomes were restricted to perinatal deaths only.
Figure 2 shows the local weight reference values for
Mexico as an example with mean birthweight of 3288 g at
40 completed weeks of gestational age (webappendix B).
When the mean birthweight at 40 completed weeks for
the local population is entered, the programme will
display the corresponding reference values. The mean
birthweight has to be estimated from the local population,
preferably excluding women with known risk factors for
small for gestational age such as maternal malnutrition,
smoking, or diseases known to be associated with either
high (eg, poorly controlled diabetes) or low (eg, preeclampsia) birthweight. Similarly, the SD of the
birthweight distribution at 40 weeks in the particular
population can be entered, if desired. Otherwise, the
average SD of 132% is used.

Discussion
We showed that although the more complex, individualised reference improved prediction of adverse
events, a simpler generic reference adjusted for country
(or geographical area) or ethnic origin could achieve the
same eect (panel). Admittedly, the comparison with a
fully customised reference was rather hampered by the
fact that only maternal weight from late pregnancy was
available. The use of weight measured before pregnancy
or in early pregnancy as requested by Gardosi and
colleagues13 might somewhat improve the performance
of the customised reference. Nevertheless, the absence of
measurements before pregnancy is a clinical reality and
reliance on self-report instead could again oset the gain
of including maternal weight in the reference.
www.thelancet.com Vol 377 May 28, 2011

Despite the fact that children small for gestational age

have an increased risk of adverse outcomes irrespective
of the reference used for classication, most adverse
outcomes occur in infants without this disorder.
Consequently, the predictive value of status of small for
gestational age for adverse perinatal outcomes is very
low. Our analysis showed that country as a proxy for the
local ethnic mix was much more important than other
variables. Although further customisation beyond
country or ethnic origin could be theoretically appealing,
additional gains were few, which was in agreement with
studies at both short-term (perinatal outcomes)27,28 and
long-term (intelligence quotient at 5 years of age).29
Although an improved prediction of perinatal mortality
by individual customisation was reported in specic
strata dened by maternal characteristics,17 this result is
apparently not generalisable to the population level. A
possible explanation for the absence of residual eects
of this further customisation could be that the
reclassication happens only for some infants who are
just above or below the threshold of small for gestational
age, and therefore their risk of perinatal mortality is
only slightly increased. Nevertheless, in multiethnic
populations, further specication of ethnic-specic
standards might be useful because these standards
aect whole subpopulations. However, their necessity

Panel: Research in context

Systematic review
Denition of infants small for gestational age in various
populations worldwide remains a challenge.7 References
based on birthweight are decient for preterm births;11
available fetal-weight references based on ultrasound
estimates24 might not be applicable in all populations; and
the complexity of the individualised1223 reference hinders its
wide use in developing countries. To identify relevant
references, we searched Medline using the terms
denition and classication with small for gestational
age. Further publications were derived from the reference
lists of the identied articles. Our aim was to create a
generic reference for fetal weight and birthweight that
overcame the above deciencies and could be readily
adapted to local populations.
Interpretation
We compared the performance of our global reference with
that of the individualised reference using data from
24 countries in Africa, Asia, and Latin America that
participated in the 200408 WHO Global Maternal and
Perinatal Health Survey.26 We showed that in the case of large
ethnic dierences, adjustment for ethnicity improves the
classication of small for gestational age substantially,
whereas addition of more parameters for individualisation
provides little further improvement against the
ethnicity-adjusted reference.

1859

Articles

depends on how large the dierences in mean

birthweight are between ethnic groups.
The most important advantage of our method is its
simplicity and exibility, which are important for
application in low-resource settings. Our global reference
was validated in a large multinational study that included
various ethnic groups across Africa, Latin America, and
Asia, and could be easily adjusted to any mixture of ethnic
groups. But our study also has several limitations.
First, like the method proposed by Gardosi and
colleagues,13 we assumed a similar fetal growth pattern
for all ethnic groups, but this assumption might not
always be true.7 Longitudinal ultrasound studies are
needed to examine racial variation in fetal growth and
how the variation might aect classication of this
disorder. Second, with a population-specic cuto for
small for gestational age (10th percentile) in various
populations, one implicitly assumes that fetal growth
restriction is similarly common across populations. This
assumption might not be true in populations heavily
aected by undernutrition, for example. However, our
generic reference can easily correct this deciency by
restriction of the analysis to pregnant women who have
adequate nutrition. For instance, in a hypothetical
malnourished population, the mean birthweight at
40 weeks of gestation is 2800 g, compared with 3000 g in
women who have adequate nutrition. One can use 3000 g
to construct a reference of fetal weight or birthweight for
this population, which corresponds to an optimum
weight standard.7 This optimum standard does not
automatically classify 10% of children in every subsample
as small for gestational age, but rather the fraction can
vary according to the local risk prole.
Third, data quality could be questionable in some
countries included in the Global Survey.26 For example, a
high fraction of preterm deliveries in some countries
suggests errors in pregnancy dating. These errors might
have resulted in a slightly larger SD of the birthweight
distribution at 40 weeks of gestation. For instance, a
study by Gardosi and colleagues13 with more accurate
dating showed an SD of 11% of the mean birthweight,
whereas our SD was 132%. The SD in Hadlock and
colleagues study25 was 127%. In countries where early
ultrasound dating is common practice, a smaller SD
provides more accurate classication of small for
gestational age. However, in countries where gestational
age is largely based on the last menstrual period, a small
SD will classify more babies as small for gestational age.
In such cases, a slightly larger SD might be acceptable.
Therefore, our programme provided in the webappendix
B allows investigators to enter their own SD if needed.
Rounding of birthweight to a full 100 g in some countries
poses another challenge in the data analysis. However, if
the rounding was mathematically correct, it is likely to
introduce only a non-dierential misclassication.
Additionally, despite all eorts to standardise data,
dierent local practice variation might aect Apgar score.
1860

Similarly, the referral of neonates to higher-level care can

depend on the local facilities. Fourth, the issue of longterm eects of status of small for gestational age could
not be assessed because no follow-up data existed. Finally,
our country-specic estimates might not be representative
for the countries, but portray the population included in
this study. A denition of small for gestational age should
take ethnic origin into account. Adaptation of a generic
reference for fetal weight and birthweight to meet local
needs is simple and can be accomplished with the
approach described in this study.
Contributors
RTM and JZ participated in the study design. RTM participated in the
data analysis and drafted the manuscript. JZ co-wrote the report. APB,
JPS, RM, AMG, and MM provided critical comments and valuable
suggestions, and contributed to writing of the report.
Conicts of interest
We declare that we have no conicts of interest.
Acknowledgments
The 200408 WHO Global Survey on Maternal and Perinatal Health
(WHOGS) was a research project implemented by the WHO in a global
network of health facilities between 2004 and 2008. The authors of this
secondary analysis are grateful to all those who contributed to the project
design and implementation, including researchers, study coordinators,
data collectors, data clerks, and other partners such as the sta from the
Ministries of Health and WHO oces. The WHOGS project was
nancially supported by the UNDP/UNFPA/WHO/World Bank Special
Programme of Research, Development, and Research Training in
Human Reproduction (HRP); WHO; United States Agency for
International Development (USAID); Ministry of Health, Labour and
Welfare of Japan; Ministry of Public Health of the Peoples Republic of
China; and the Indian Council of Medical Research, India. JZ was
supported by the Intramural Programme of Eunice Shriver Kennedy
National Institute of Child Health and Human Development, National
Institutes of Health, USA. The named authors alone are responsible for
the views expressed in this manuscript, which does not necessarily
represent the decisions or the stated policy of the WHO.
References
1
Ott WJ. Small for gestational age fetus and neonatal outcome:
reevaluation of the relationship. Am J Perinatol 1995; 12: 396400.
2
Doctor BA, ORiordan MA, Kirchner HL, Shah D, Hack M.
Perinatal correlates and neonatal outcomes of small for gestational
age infants born at term gestation. Am J Obstet Gynecol 2001;
185: 65259.
3
Smedler AC, Faxelius G, Bremme K, Lagerstrom M. Psychological
development in children born with very low birth weight after
severe intrauterine growth retardation: a 10-year follow-up study.
Acta Paediatr 1992; 81: 197203.
4
Sung IK, Vohr B, Oh W. Growth and neurodevelopmental outcome
of very low birth weight infants with intrauterine growth
retardation: comparison with control subjects matched by birth
weight and gestational age. J Pediatr 1993; 123: 61824.
5
Verkauskiene R, Figueras F, Deghmoun S, Chevenne D, Gardosi J,
Levy-Marchal M. Birth weight and long-term metabolic outcomes:
does the denition of smallness matter? Horm Res 2008; 70: 30915.
6
von Ehrenstein OS, Mikolajczyk RT, Zhang J. Timing and
trajectories of fetal growth related to cognitive development
in childhood. Am J Epidemiol 2009; 170: 138895.
7
Zhang J, Merialdi M, Platt LD, Kramer MS. Dening normal
and abnormal fetal growth: promises and challenges.
Am J Obstet Gynecol 2010; 202: 52228.
8
Chauhan SP, Gupta LM, Hendrix NW, Berghella V. Intrauterine
growth restriction: comparison of American College of
Obstetricians and Gynecologists practice bulletin with other
national guidelines. Am J Obstet Gynecol 2009; 200: 409 e16.
9
Williams RL, Creasy RK, Cunningham GC, Hawes WE, Norris FD,
Tashiro M. Fetal growth and perinatal viability in California.
Obstet Gynecol 1982; 59: 62432.

www.thelancet.com Vol 377 May 28, 2011

Articles

10

11

12

13
14
15
16

17

18
19
20

Alexander GR, Himes JH, Kaufman RB, Mor J, Kogan M. A United

States national reference for fetal growth. Obstet Gynecol 1996;
87: 16368.
Zhang X, Platt RW, Cnattingius S, Joseph KS, Kramer MS. The use
of customised versus population-based birthweight standards in
predicting perinatal mortality. BJOG 2007; 114: 47477.
MEDLINE Data Changes2004. http://www.nlm.nih.gov/pubs/
techbull/nd03/nd03_med_data_changes.html (accessed April 26,
2011).
Gardosi J, Mongelli M, Wilcox M, Chang A. An adjustable fetal
weight standard. Ultrasound Obstet Gynecol 1995; 6: 16874.
Gardosi J. Ethnic dierences in fetal growth.
Ultrasound Obstet Gynecol 1995; 6: 7374.
Gardosi J. Fetal growth: towards an international standard.
Ultrasound Obstet Gynecol 2005; 26: 11214.
Gardosi J, Clausson B, Francis A. The use of customised versus
population-based birthweight standards in predicting perinatal
mortality. BJOG 2007; 114: 130102.
Gardosi J, Clausson B, Francis A. The value of customised centiles
in assessing perinatal mortality risk associated with parity and
maternal size. BJOG 2009; 116: 135663.
Figueras F, Gardosi J. Should we customize fetal growth standards?
Fetal Diagn Ther 2009; 25: 297303.
Gardosi J. New denition of small for gestational age based on fetal
growth potential. Horm Res 2006; 65 (suppl 3): 1518.
Clausson B, Gardosi J, Francis A, Cnattingius S. Perinatal outcome
in SGA births dened by customised versus population-based
birthweight standards. BJOG 2001; 108: 83034.

www.thelancet.com Vol 377 May 28, 2011

21

22

23
24

25
26

27

28

29

McCowan L, Stewart AW, Francis A, Gardosi J. A customised

birthweight centile calculator developed for a New Zealand
population. Aust N Z J Obstet Gynaecol 2004; 44: 42831.
Mongelli M, Figueras F, Francis A, Gardosi J. A customized
birthweight centile calculator developed for an Australian
population. Aust N Z J Obstet Gynaecol 2007; 47: 12831.
Gardosi J, Francis A. A customized standard to assess fetal growth
in a US population. Am J Obstet Gynecol 2009; 201: 25 e17.
Figueras F, Meler E, Iraola A, et al. Customized birthweight
standards for a Spanish population. Eur J Obstet Gynecol Reprod Biol
2008; 136: 2024.
Hadlock FP, Harrist RB, Martinez-Poyer J. In utero analysis of fetal
growth: a sonographic weight standard. Radiology 1991; 181: 12933.
Shah A, Faundes A, Machoki M, et al. Methodological
considerations in implementing the WHO Global Survey for
Monitoring Maternal and Perinatal Health. Bull World Health Organ
2008; 86: 12631.
Zhang J, Mikolajczyk R, Grewal J, Neta G, Klebano M. Prenatal
application of the individualized fetal growth reference.
Am J Epidemiol 2011; 173: 53943.
Hutcheon JA, Walker M, Platt RW. Assessing the value
of customized birth weight percentiles. Am J Epidemiol
2011; 173: 45967.
Neta G, Grewal J, Mikolajczyk R, Klebano M, Zhang J. Does
the individualized birthweight reference outperform a simple
ultrasound reference in predicting long-term child
neurodevelopment? Ultrasound Obstet Gynecol 2010; published
online Dec 10. DOI:10.1002/uog.8902.

1861