Ministry of Education and Shanghai Key Laboratory of Children's Environmental Health, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
Division of Neonatology, Department of Paediatrics, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
Correspondence: Dr J Zhang, MOE and Shanghai Key Laboratory of Children's Environmental Health, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, 1665 Kongjiang Road, 200092 Shanghai, China. Email email@example.com
To determine whether the recently published A global reference for fetal-weight and birthweight percentiles (Global Reference) improves small- (SGA), appropriate- (AGA), and large-for-gestational-age (LGA) definitions in predicting infant mortality.
Population-based cohort study.
The US Linked Livebirth and Infant Death records between 1995 and 2004.
Singleton births with birthweight >500 g born at 24–41 weeks of gestation.
We compared infant mortality rates of SGA, AGA, and LGA infants classified by three different references: the Global Reference; a commonly used birthweight reference; and Hadlock's ultrasound reference.
Main outcome measures
Infant mortality rates.
Among 33 997 719 eligible liveborn singleton births, 25% of preterm and 9% of term infants were classified differently for SGA, AGA, and LGA by the Global Reference and the birthweight reference. The Global Reference indicated higher mortality rates in preterm SGA and preterm LGA infants than the birthweight reference. The mortality rate was considerably higher in infants classified as preterm SGA by the Global Reference but not by the birthweight reference, compared with the corresponding infants classified by the birthweight reference but not by the Global Reference (105.7 versus 12.9 per 1000, RR 8.17, 95% CI 7.38–9.06). Yet, the differences in mortality rates were much smaller in term infants than in preterm infants. Black infants had a particularly higher mortality rate than other races in AGA and LGA preterm and term infants.
In respect to the commonly used birthweight reference, the Global Reference increases the identification of infant deaths by improved classification of abnormal newborn size at birth, and these advantages were more obvious in preterm than in term infants.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Abnormal fetal growth is easy to understand but hard to define. Most literature continues to use small for gestational age (SGA) or large for gestational age (LGA) as a surrogate for abnormal fetal growth. SGA is generally defined as a fetal weight or birthweight below the tenth percentile of a certain reference at a given gestational week, whereas LGA is a weight above the 90th percentile. It is well established that fetal growth abnormalities are often associated with increased risks of various adverse outcomes in the perinatal period or even in the long term.[2-4]
Over the last half century birthweight references for gestational age have been commonly used in most clinical and epidemiological research.[5-8] These references were developed with very large, mostly population-based databases, and they provided birthweight percentiles by each gestational week. However, infants born preterm are more likely to be growth restricted, and their birthweight does not represent all fetuses in utero at a given gestational week for preterm. Therefore, birthweight references would substantially under-diagnose SGA infants in preterm births.[9-11] Yet, ultrasound-based fetal-weight references were mostly developed in white women from Europe or North America,[12-15] which might not be applicable to the ethnic groups living in most developing countries.
Recently, Mikolajczyk et al. created a Global Reference that adjusted for country or ethnic origin and could be readily adapted to local populations. Using data from the 2004–2008 World Health Organization (WHO) Global Survey on Maternal and Perinatal Health, they have proved that the Global Reference has a better ability to predict adverse perinatal outcomes (stillbirths, neonatal deaths, referral to higher-level or special care unit, or low Apgar score) than does the non-customised fetal-weight reference (by Hadlock et al.), and is more simple to use than the individualised reference, without losing any predictive ability.[16-18] The most important advantages of this method are its simplicity and flexibility, which are important for low-resource settings, and could be easily adjusted to any mixture of ethnic groups.
In the present study, we aimed to further examine the validity and superiority of the Global Reference to predict infant mortality. Using a large data set, we tested the hypothesis that this reference has better ability to distinguish between normal and abnormal newborn size at birth determined by infant mortality, compared with a commonly used birthweight reference.
The US Linked Livebirth and Infant Death files from 1995 to 2004 were used. These data were from a national live birth registry linked to infant deaths and compiled by the National Center for Health Statistics (NCHS), US Centers for Disease Control and Prevention. A detailed description of this data set is provided elsewhere. The available information in this data set included demographic characteristics of mothers, obstetric history, birth outcomes, and infant death.
In the current study, maternal age was categorised into three groups: 19 years or less; 20–34 years; and 35 years or older. Maternal race/ethnicity was classified as non-Hispanic white, non-Hispanic black, Hispanic, and Asian. Maternal educational attainment was categorised into less than high school, complete high school, college, and graduate school. Marital status of the mother (married/unmarried), and infant birthweight (in grams), and sex (male/female) were also included. Maternal behaviours and obstetric complications included: cigarette use during pregnancy (yes/no); pregestational or gestational diabetes (yes/no); and chronic or pregnancy-associated hypertension, eclampsia (yes/no). The starting month of prenatal care was categorised into first trimester (months 1–3), second trimester (months 4–6), third trimester (months 7–9), and no prenatal care. The outcome of interest in this study was infant death, which was further divided into three components: early neonatal death (0–6 days), late neonatal death (7–27 days), and postneonatal death (28–364 days).
For most women (86.8%), two different types of gestational age estimates were recorded on the birth certificate: the clinical estimate (CE) and the gestational age based on the self-reported last menstrual period (LMP). The deficiencies of LMP-based gestational age are well established.[20-23] Recently, Qin et al. used a simple method in which the CE of gestational age is substituted for LMP-based gestational age when the difference between the two estimates is greater than 2 weeks (LMP/CE method). They demonstrated that the LMP/CE method, when compared with other techniques, almost eliminated the second mode in the gestational age distribution. Thus, this method appears to be effective in correcting large errors in gestational age estimates. It has the further benefit that records are reclassified, rather than excluded altogether. Given these strengths, we adopted the LMP/CE method for the purpose of our analysis. A total of 10.7% of women had a replacement in the current study.
To define SGA, AGA, and LGA, we applied three references to our study population. The first was a commonly used birthweight reference proposed by Alexander et al., which was based on all singleton live births to US women, irrespective of ethnicity, and in proportion to their separate contribution to the total US population. The second reference was a Global Reference, which was recently created according to the Hadlock's fetal-weight reference and the Gardosi's notion of proportionality. With the approach proposed by Gardosi and colleagues, we assumed that Hadlock's growth equation could be used to derive percentiles of fetal weight in a given gestational week for a different population by anchoring the curve to a mean birthweight at 40 weeks of gestation. Details of the method were described elsewhere. In the current analysis, we first separated the study population by race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and Asian). Second, we calculated the mean birthweight and standard deviation at 40 completed weeks of gestation for each race/ethnicity (white, 3564.8 ± 446.7 g; Hispanic, 3484.7 ± 441.8 g; black, 3366.4 ± 445.2 g; Asian, 3398.3 ± 425.9 g). Using the Global Reference weight percentiles calculator, we produced race-specific birthweight reference percentiles. In addition, we added the Hadlock's reference, a non-customised fetal-weight reference, which was developed from a population of 392 predominantly middle-class white pregnant women with well-known LMPs, who were evaluated with ultrasound between gestational weeks 10 and 41. The tenth and 90th percentiles were used as cut-offs for SGA, AGA, and LGA, and we compared the infant mortality rates in these groups classified by the three references. In order to compare the predictability of infant mortality by different references, we first have to classify each infant into an appropriate group (SGA, AGA, and LGA). To do so, establishing a local standard based on the Global Reference method, therefore, is the first necessary step for classification. Although the procedures are the same, the purpose of this study is not to establish a new standard. We have limited ourselves to conducting an external validation of the Global Reference. Given the huge number of subjects in our analysis, all differences were statistically significant (P < 0.001).
We restricted our analysis to singleton live births at gestational ages of 24–41 weeks and with birthweights of 500 g or more. Our upper limit of gestational age was set at 41 weeks because the Global Reference for fetal-weight and birthweight provides weight percentiles only up to 41 weeks of gestation. There were 39 956 864 live births in the linked 1995–2004 birth and infant death data set (Figure 1). We excluded 1 227 780 multiple gestations, 1 962 048 newborns delivered before 24 or after 41 weeks of gestation, and 8754 births of less than 500 g. Subjects with missing information on important variables were further excluded, except for smoking during pregnancy, for which a new category of missing value was created, leaving 33 997 719 subjects (85%) for the final analysis.
Table 1 presents maternal and infant characteristics in the study population. It also shows that the Global Reference classified substantially more SGA and LGA newborns than the birthweight reference did, particularly for LGA and among the infant deaths. However, the Global Reference classified substantially fewer SGA but more LGA newborns than the Hadlock's reference did, especially for SGA and among the infant deaths. Figure 2 compared reference curves at the tenth and 90th percentiles by the Alexander reference, Hadlock reference, and the Global Reference for white women. As expected, the latter two curves are very similar. The curves of the Alexander reference have a quite different shape and wider variation than the other curves. The tenth percentiles of the Global Reference are generally higher than those of the Alexander reference throughout 24–41 weeks of gestation. The 90th percentiles of the Global Reference are lower before 39 weeks of gestation, but are higher after 39 weeks of gestation than those of the Alexander reference.
Table 1. Maternal and infant characteristics of the study population, Birth and Infant Death Cohort, 1995–2004
Live birth (n = 33 997 719)
Early neonatal (n = 52 870)
Late neonatal (n = 27 423)
Postneonatal (n = 78 899)
Early indicates death within 0–6 days of birth, late indicates death within 7–27 days of birth, and postneonatal indicates death within 28–364 days of birth.
<12 years (less than high school)
12 years (high school)
13–16 years (college)
≥17 years (graduate school)
Mother's marital status
Month that prenatal care started
No prenatal care
Cigarette use during pregnancy
Unknown or not stated
<37 weeks (preterm)
37–41 weeks (term)
Table 2 shows that the Global Reference and the birthweight reference classified 75% of preterm infants (n = 2 264 823) and 91% of term infants (n = 28 170 024) into the same category of SGA, AGA, or LGA, i.e. a discrepancy of classification occurred in 25% preterm and 9% term infants. Preterm infants who were classified as SGA by the birthweight reference but as AGA by the Global Reference (in green) had a mortality rate of 12.9 per 1000. In contrast, preterm infants who were classified as AGA by the birthweight reference but as SGA by the Global Reference (in red) had a mortality rate of 105.7 per 1000 (RR 8.17, 95% CI 7.38–9.06). The contrast is striking across all periods (early, late, and post-neonatal) of preterm infant death.
Table 2. Classification overlap between the birthweight and global references and infant mortality, US Linked Livebirth and Infant Death Cohort, 1995–2004
Global Reference: mortality/1000 live births (95% CI)
CI, confidence interval.
Early indicates death within 0–6 days of birth, late indicates death within 7–27 days of birth, postneonatal indicates death within 28–364 days of birth, and infant indicates death within 0–364 days of birth.
1 737 832
2 345 303
22 817 860
1 686 161
3 006 861
The differences of mortality rates between the two references were much smaller in term infants than in preterm infants, however. The corresponding mortality rates were 5.0 per 1000 (in green) versus 3.2 per 1000 (in red) among term births (RR 0.63, 95% CI 0.59–0.68), and 1.4 per 1000 (in yellow) versus 2.1 per 1000 (in blue) among term births (RR 1.56, 95% CI 1.42–1.72). The contrast is moderate across all periods (early, late, and post-neonatal) of term infant death.
We constructed a similar table to compare the Global Reference with the Hadlock's reference. Because the reference values listed by Hadlock et al. are consistently higher than those listed by the Global Reference, no infants were classified as AGA by Hadlock's reference but as SGA by the Global Reference, or as LGA by Hadlock's reference but as AGA by the Global Reference. Thus, no appropriate comparisons could be made (Table S1).
Figure 3 illustrates that the Global Reference indicated higher mortality rates in preterm SGA (61.6 versus 45.1 per 1000, RR 1.37, 95% CI 1.34–1.39) and preterm LGA (13.8 versus 12.4 per 1000, RR 1.12, 95% CI 1.07–1.17) infants than the birthweight reference, and lower mortality rates in preterm AGA (20.2 versus 22.7 per 1000, RR 0.89, 95% CI 0.88–0.90). However, the differences of mortality rates were much smaller in term SGA (7.8 versus 8.3 per 1000, RR 0.94, 95% CI 0.92–0.96), term AGA (2.3 versus 2.3 per 1000, RR 1.01, 95% CI 0.99–1.02), and term LGA (1.9 versus 1.7 per 1000, RR 1.11, 95% CI 1.08–1.15) infants between the two references. With regards to each race/ethnicity group, these trends were the same as described above.
We further examined infant mortality in SGA, AGA, and LGA classified by the Global Reference and by Hadlock's reference. As shown in Figure 3, the Global Reference indicated much higher infant mortality in SGA (for preterm SGA, 61.6 versus 55.4 per 1000, RR 1.11, 95% CI 1.09–1.13; for term SGA, 7.8 versus 6.4 per 1000, RR 1.22, 95% CI 1.20–1.24) than Hadlock's reference, whereas the differences in mortality rates were much smaller for LGA (for preterm LGA, 13.8 versus 14.4 per 1000, RR 0.96, 95% CI 0.93–0.99; for term LGA, 1.9 versus 1.9 per 1000, RR 1.00, 95% CI 0.96–1.04) and for AGA (for preterm AGA, 20.2 versus 17.8 per 1000, RR 1.13, 95% CI 1.11–1.15; for term AGA, 2.3 versus 2.2 per 1000, RR 1.06, 95% CI 1.05–1.08) between the two references. This figure further shows that black infants had almost twice as high a mortality rate as other races in both preterm and term AGA and LGA infants. But the difference in mortality was much smaller among SGA preterm and term births.
This study, using data from the 1995–2004 US Livebirth and Infant Death cohort, compares two methods of classifying normal and abnormal newborn size at birth (SGA, AGA, and LGA): a commonly used birthweight reference and a recently published Global Reference customised to race/ethnicity.[7, 16] Our results indicate that the race-specific Global Reference not only identifies more SGA and LGA infants, but also improves the classification of abnormal newborn size at birth determined by substantially higher infant mortality, and these advantages were more obvious in preterm than in term births.
Strengths and interpretation
This phenomenon is probably attributable to two features of the Global Reference: using fetal weight as a reference and customising the data by race/ethnicity. First, evidence has shown that infants born preterm are more likely to be growth restricted. The tenth percentile of the birthweight reference in preterm is substantially lower than the tenth percentile of the ultrasound-based fetal-weight reference: thus, the birthweight reference significantly under-diagnoses SGA infants in preterm births. The Global Reference, which was based on intrauterine fetal weight, overcomes the above deficiency. It had a better ability to include more preterm births in the SGA group, and resulted in a higher infant mortality rate.
Second, birthweight differs significantly across race/ethnicity. For example, Asian infants are approximately 170 g lighter than white infants in the USA, but these two groups have similar infant mortality rates. On the other hand, black infants have similar birthweights to Asian infants, whereas the mortality rate of black infants is much higher than that of Asian infants. Thus, it is important to adjust for race/ethnicity in classifying fetal growth. The race-specific reference produced by the Global Reference contributed to the improvement in the classification of normal and abnormal fetal growth.
Although the increased risk of infant mortality with SGA is well documented,[26-28] the impact of LGA on mortality is less frequently reported. LGA fetuses are often, but not always, at an increased risk of a number of perinatal complications. These include birthing difficulties as well as metabolic problems resulting in neonatal or infant death.[29-31] Our results indicate that, compared with the birthweight reference, the Global Reference reclassified substantially more LGA infants, and the infant mortality rate was significantly higher among the infants redefined as LGA. The difference in infant mortality suggested that the Global Reference was a better tool to identify pathologically excessive growth.
Weaknesses and interpretation
The improved prediction of infant mortality by the Global Reference compared with the birthweight reference is mainly derived from preterm births. First, and most importantly, the infant mortality rate of term births is much lower than that of preterm births (2.7 versus 24.9 per 1000). Birthweight is no longer a good predictor for term infant deaths. Second, the differences in birthweight and infant mortality across races/ethnicities in the USA may not be as large as those in other parts of the world. It may have reduced the overwhelming superiority of the Global Reference observed in our previous study from different countries. Thus, the Global Reference may be particularly valuable in populations with high variation in birth size and mortality rate.
Recent studies compared the Hadlock's reference (without adjustment for maternal and fetal characteristics) versus a fully customised classification of birthweight centiles and found that the latter offered little benefit in terms of predicting stillbirth or early neonatal death.[32-37] However, our results (Figure 3) indicate that there are large differences in infant mortality rates among races/ethnicities, even within the same fetal size category, which supports the necessity to adjust for race/ethnicity. Unfortunately, the current data do not allow us to look at some other components of customisation, such as maternal height and weight. Finally, it is important to point out that most previous validation studies used severe adverse outcomes such as neonatal and infant mortality. Consequently, conclusions were limited to this context. If the ultimate goal of refining the definition of abnormal fetal growth is to predict obstetric and more long-term effects, such as miscarriage, stillbirth, chronic hypertension, and metabolic syndrome, then the future validation study would preferably use such outcomes or their surrogates as the ‘gold standard’.
In summary, our study demonstrated that, compared with the commonly used birthweight reference, the Global Reference has an improved ability to identify abnormal fetal growth associated with an increased risk of infant death, and these advantages are mainly derived from preterm births. With a large and diverse population with regards to race/ethnicity, future research focusing on more subtle perinatal and long-term adverse outcomes may be needed to further examine whether the Global Reference still holds the advantage over the birthweight-based reference.
Disclosure of interests
None to declare.
Contribution to authorship
J.Z. conceived the study and designed it with G.D. G.D. participated in the data analysis with Y.P., and drafted the article. J.Z. and Y.T. substantially revised the article. J.Z. and G.D. participated in data interpretation. Y.Z. and J.S.Z. provided critical comments and valuable suggestions, and contributed to the writing of the report. All authors agreed to its submission, and had full access to the original data.
Details of ethics approval
This observational cohort study did not require ethics committee approval.
The US Linked Livebirth and Infant Death research project was conducted by the National Center for Health Statistics (NCHS), US Centers for Disease Control and Prevention. The authors of this secondary analysis are grateful to all those who contributed to the project design and implementation, including the researchers, study coordinators, data collectors, data clerks, and other partners.