An estimated 20 million infants are born globally with low birthweight (<2500 g) every year.¹ Low birthweight is an important population indicator for tracking neonatal health and includes babies born preterm (<37 completed weeks of gestation) and infants with intrauterine growth restriction. These components of low birthweight have different causes and risks of mortality, morbidity, impaired growth, and non-communicable diseases. Hence, for us to guide interventions to address both prevention and care, we must delineate low birthweight according to preterm birth, intrauterine growth restriction, and their overlap.

National estimates of preterm birth for 184 countries have been published for the year 2010,² showing a total of 14·9 million preterm births. In the Global Burden of Disease Study,³ 77 million (3·1%) disability-adjusted life-years were attributed to preterm birth, similar to the burden of HIV or malaria. In 1998, de Onis and colleagues⁴ reported estimates of intrauterine growth restriction, using babies born full term and with low birthweight as a proxy measure. They estimated that 13·7 million babies were born at term and with low birthweight every year, but they did not provide national estimates.⁴ Furthermore, no estimates are available for the co-occurrence of intrauterine growth restriction and preterm birth, or the relation between intrauterine growth restriction and the widely used metric of low birthweight.

The classification of small for gestational age was defined by a 1995 WHO expert committee as infants below the 10th centile of a birthweight-for-gestational-age, gender-specific reference population.^5,6 A major challenge is selection of an appropriate global reference. Small for gestational age is a commonly accepted proxy measure of intrauterine growth restriction.⁵ However, small for gestational age includes babies who are constitutionally small and in the lower tail of the growth curve, in addition to infants who were growth-restricted in utero because of maternal and environmental factors, such as chronic undernutrition, multiple pregnancy, placental insufficiency, pregnancy complications (eg, pre-eclampsia), infections, and other toxic exposures.⁷ In settings with high rates of small-for-gestational-age births, growth restriction accounts for a high proportion of these,⁴ justifying its use as a proxy for intrauterine growth restriction.

Our aim is to estimate the national prevalence and numbers of neonates born small for gestational age at full term (≥37 weeks; term-SGA), and the co-occurrence of small for gestational age with preterm birth (preterm-SGA), in 138 countries of low and middle income. We focus on this group of countries in view of their high burden of disease and the urgent need for data to direct, monitor, and assess public health planning in these regions.

The appendix shows data inputs for the estimation process (p 12), study characteristics of the 22 CHERG cohorts included in our study^13,17–36 (pp 2–4), and survey characteristics for the WHO datasets⁹ (p 6). From the literature review, we identified six studies reporting prevalence of babies born small for gestational age and the proportion of low-birthweight infants (appendix p 5); however, none of these studies could be used because term-SGA or preterm-SGA rates were not reported. Table 1 shows the final model for term-SGA. Logit(term-SGA prevalence) increased with rising rates of low birthweight and neonatal mortality (figure 1). Regional random effects were applied to account for regional variations of the relations. With low birthweight and neonatal mortality included in the model, socioeconomic covariates were not significant and, thus, not retained in the final model. Data source covariates were retained to control for selection bias and data quality. Regression diagnostic plots are shown in the appendix (p 13); the overall model fit was good (adjusted R² 0·8237).

Meta-analyses for the prevalence of babies born small for gestational age are presented in the appendix for moderate-to-late preterm infants (32 weeks to <37 weeks of gestation; pp 14–16) and for early preterm infants (<32 weeks of gestation; pp 17–18). Overall, in moderate-to-late preterm infants, 22·0% were small for gestational age (Asia 24·4%; Africa 17·0%; Latin America and the Caribbean 22·7%; appendix p 14). Prevalence of babies born small for gestational age in the moderate-to-late preterm group was similar in community-based and facility-based studies in Asia (appendix p 15), and when restricted to high-quality studies (p 16). For early preterm infants (born at <32 weeks of gestation), the potential for selection bias was greater, in view of early mortality before weighing in community-based cohorts and incomplete data capture. In Asia, prevalence of preterm-SGA was highest in facility-based studies (9·0% in nine facility studies vs 2·1% in six community studies; appendix p 17). With high-quality datasets, the overall prevalence of babies born small for gestational age before 32 weeks of gestation was 11·0% (appendix p 18). In sensitivity analyses, we noted no effect of imputation of missing birthweight data on the prevalence of preterm-SGA in four Asian and two African datasets (appendix p 9).

Figure 2 shows the estimated national prevalences of small-for-gestational-age births in low-income and middle-income countries for the year 2010 (a complete list of national estimates is available in the appendix pp 19–23). Table 2 presents the numbers and prevalence of term-SGA, preterm-SGA, and all small-for-gestational-age births, by UN-MDG region. Prevalence of term-SGA ranged from 5·3% in east Asia to as high as 41·5% in south Asia, and preterm-SGA ranged from 1·2% in north Africa to 3·0% in southeast Asia. For all small-for-gestational-age births, the highest prevalence was recorded in south Asia (44·5%), followed by sub-Saharan Africa (25·5%) and southeast Asia (24·3%). The greatest numbers of term-SGA infants were born in south Asia (16·2 million) and sub-Saharan Africa (7·5 million). Although their absolute numbers are lower, preterm-SGA infants carry a higher risk of mortality in the newborn and infant periods than do term-SGA infants;¹⁰ numbers of preterm-SGA babies totalled 1·2 million in south Asia and 0·6 million in sub-Saharan Africa. The vast majority of small-for-gestational-age infants (87%, 28·2 million) were born in south Asia, southeast Asia, and sub-Saharan Africa.

Figure 3 presents prevalence data for term-SGA, preterm-SGA, and preterm appropriate-for-gestational-age births, compared with the prevalence of babies born with low birthweight. Table 2 also shows the estimated numbers of term-SGA infants who weighed less than 2500 g at birth (term-SGA and with low birthweight), by UN-MDG region for 2010. In all regions, the majority (>50%) of term-SGA infants weighed 2500 g or heavier, with high proportions of babies not low birthweight but small for gestational age in Africa (74%) and Latin America and the Caribbean (71%). The highest regional proportion of low-birthweight babies was recorded in south Asia (26%), and the prevalence of term-SGA infants was also very high in this region (42%). Term-SGA accounted for 65% of low-birthweight babies in south Asia and preterm birth accounted for 35%. In sub-Saharan Africa, although preterm birth rates were similar to those in south Asia, the rate of low-birthweight babies was lower (14%) and preterm birth made a relatively larger contribution to the low-birthweight metric (57% preterm birth vs 43% term-SGA). Similarly in Latin America and the Caribbean, preterm birth comprised a larger proportion of the low-birthweight metric (60% preterm birth vs 40% term-SGA). In east Asia, the proportion of low-birthweight infants was very low (2·6%) and consisted mainly of preterm-SGA infants. In regions with lower rates of low-birthweight babies, such as north Africa or east Asia, preterm birth seems the more influential contributor to the low-birthweight metric.

Overall, in countries of low and middle income in 2010, an estimated 43·3 million infants (36% of livebirths) were born either preterm or small for gestational age, or both (figure 4). Of 18 million low-birthweight infants, 59% were term-SGA whereas 41% were preterm (16% preterm-SGA, 25% preterm and appropriate size for gestational age).

Table 3 shows the ten countries with the largest numbers of small-for-gestational-age infants born in 2010. An estimated 12·8 million babies were born small for gestational age in India alone (95% CI 11·5–14·3 million), with a prevalence of 47%. Pakistan, Nigeria, Bangladesh, China, and Indonesia had more than 1 million small-for-gestational-age babies.

Our data provide national and regional estimates for the prevalence and number of babies born small for gestational age and the co-occurrence of small for gestational age with preterm birth. 43·3 million infants (36% of livebirths) in countries of low or middle income were born either too small (small for gestational age) or too soon (preterm), or both, in 2010. The estimated burden of babies born small for gestational age is very high; 32·4 million neonates (27% of livebirths) are affected, of whom 29·7 million infants were born at full term (≥37 weeks) and 10·6 million were born at term and with low birthweight (<2500 g). Almost 3 million infants (2%) were born preterm and small for gestational age.

The highest rates and numbers of babies born small for gestational age were in south Asia, where more than half of babies small for gestational age are born and nearly one in two infants are born too small. The prevalence of babies born small for gestational age reached almost 50% in Pakistan and India, predicted largely on national rates for low birthweight, which were very high. The cutoff for small for gestational age at the 10th centile of the reference population was recommended by a WHO expert committee;⁵ however, a lower cutoff could be considered at the 3rd centile, which would indicate especially severe cases of small for gestational age, particularly in high-burden settings. With a 3rd centile³⁸ cutoff, the prevalence of severe small-for-gestational-age births was 23% in south Asia, affecting 3·9 million infants (analysis not shown). The lowest rates of babies born small for gestational age were noted in east Asia, largely affected by data for China, where the reported low-birthweight rate was 2·4% (WHO Regional Offices, 2008).

A major challenge in estimating the global burden of babies born small for gestational age is selection of a common reference population. The Williams³⁹ reference of Californian livebirths from 1970–76 (n=2 288 806) was recommended in 1995 by WHO⁶ in view of the multiracial population, representation at lower gestational ages, and association with survival. We chose the 1991 US birth reference population,⁸ which was published after the original WHO recommendation, because it is more recent than the Williams reference, has a large sample size (n=3 134 879) that better represents low gestational ages, covers a national and diverse multiethnic population, has well characterised methods to smooth centile curves and exclude outlying values, and is the most frequently cited reference in scientific literature. Choosing a common reference for burden estimates is important, since the estimated prevalence of babies born small for gestational age varies substantially depending on the reference population chosen. For example, within a south Indian cohort, the estimated prevalence of babies born small for gestational age ranged from 10·5%⁴⁰ to 72·5%⁴¹ using the 10th centile cutoff of different reference populations, with the Alexander reference providing a prevalence of 56%⁸ (Joanne Katz, Johns Hopkins Bloomberg School of Public Health; personal communication). Another consideration is use of a birthweight-for-gestation curve versus an ultrasound-based fetal-weight curve. For preterm infants, a birthweight-for-gestation reference might underestimate the true prevalence of intrauterine growth restriction because preterm infants could be small at birth because of pathological effects, which led to the preterm birth, compared with babies who remain in utero.⁴² However, ultrasound-based fetal-weight estimation methods also have limitations. A standard proposed by WHO⁴³ shifts the Hadlock distribution of fetal weights for every gestational age by a particular country's mean birthweight at 40 weeks, thus setting by default any population-based small-for-gestational-age prevalence close to 10%. This strategy only identifies the most growth-restricted infants in that particular population, rather than establishing the burden of suboptimum growth. Most limitations of available fetal growth references are being addressed in the WHO Intergrowth study, which is currently taking place in eight geographically diverse settings and aims to develop international growth standards to describe optimum fetal growth and newborn nutritional status (completion in 2014).⁴⁴

Our analyses show important regional differences in babies born small for gestational age and the composition of low birthweight. In south Asia, rates of low birthweight are high and many (65%) low-birthweight births are attributable to term-SGA infants. However, in sub-Saharan Africa and Latin America and the Caribbean, just over 50% of low-birthweight babies are preterm. Furthermore, low birthweight might not fully capture the increased risk of babies born too soon or too small. The median birthweight of an infant born at 33 weeks of gestation is around 2500 g for the Alexander distribution;⁸ thus, many late preterm infants could weigh 2500 g or heavier. Two-thirds of term-SGA infants weigh 2500 g or more, although these babies are at lower risk of morbidity and mortality than their low-birthweight counterparts, particularly from non-communicable diseases in adulthood.

Estimates of intrauterine growth restriction were reported by de Onis and colleagues in 1998.⁴ These researchers estimated that 13·7 million infants (11% of births) in low-income and middle-income countries were born at term and with a low birthweight, an indicator that was a proxy for intrauterine growth restriction. By comparison, we estimated that a total of 10·6 million babies were born at term with low birthweight in countries of low and middle income in 2010. However, our estimation of small for gestational age also included two groups missing in the term and low-birthweight indicator: preterm-SGA infants who are at substantially higher risk of adverse outcomes;¹⁰ and babies born small for gestational age but weighing 2500 g or more. The estimates made by de Onis and colleagues were based on 1996 rates of low birthweight from WHO and on older data from 1960–96, which used inputs from 60 datasets in low-income and middle-income countries at a time when less attention was paid to metrics for gestational age. Recent findings show temporal changes in the distribution of small-for-gestational-age and preterm births in low-birthweight babies.⁴⁵

Estimates of preterm, low-birthweight, and small-for-gestational-age rates are imperfect because of gaps and biases in data. The methods used to ascertain gestational age varied between studies and might affect estimation of gestational length. We included studies meeting a priori data-quality criteria for gestational age, and nine studies included ultrasound measures of gestational age. In several studies, last menstrual period was recorded prospectively during monthly pregnancy surveillance and, thus, this information was subject to less recall bias. We included information from both facility-based and community-based or population-based studies, and we attempted to assess biases. National data were available from Chile only.²¹ The WHO Global Survey was a facility-based survey, which could be biased depending on the nature of the facility, the number of facilities in an area, and the proportion of deliveries that took place in the home. We included a covariate to control for facility bias. In community-based studies, neonatal weight is measured after birth and, therefore, a high proportion of birthweight data can be missing for early neonatal deaths. We excluded datasets that had a substantial amount of missing birthweight data (>25%), and we did sensitivity analyses with imputation of missing birthweight data.¹⁰ The prevalence of term-SGA and preterm-SGA did not change substantially. However, data for birthweight might have been missing more frequently among preterm-SGA babies, because these infants are at a higher risk of mortality and they might have died before weighing. Thus, our projections could underestimate the prevalence of preterm-SGA. Furthermore, in view of the use of birthweight rather than an ultrasound growth reference, the prevalence of preterm-SGA could be underestimated, because growth restriction has a relatively higher frequency in babies who are born preterm versus those who remain in utero for the full gestation period. Data for maternal HIV status were limited; HIV infection can be a risk factor for babies born small for gestational age, although risk is not so clearly defined for preterm birth.⁴⁶ Finally, most of our datasets did not include data on stillbirths, which are more likely to be associated with fetal growth restriction and preterm birth, and our estimates do not capture this burden.

The dearth and quality of data on both birthweight and gestational age in countries of low and middle income have been key barriers to quantification of the burden of small-for-gestational-age babies or intrauterine growth restriction (panel). More than half of infants in low-income and middle-income countries are never weighed at birth, particularly those born outside of facilities,¹ and facility-based data are subject to selection biases. Inclusion of birthweight in household surveys (eg, demographic and health survey, multiple indicator cluster survey) since the 1990s has improved data availability, and methods to adjust data quality have been developed.¹ Serial fetal ultrasonography is the gold standard for diagnosis of intrauterine growth restriction in high-resource settings, but small for gestational age at birth is the most commonly used indicator in countries of low and middle income. Data for gestational age are also troublesome. In low-income and middle-income countries, ultrasound is generally not available and last menstrual period is used to date most pregnancies, which can be affected by poor maternal recall, lactational amenorrhoea, variation in length of menstrual cycle, or injectable contraception. Last menstrual period has an estimated error of SD 3 weeks, clinical assessment SD 2 weeks, and ultrasound done before 20 weeks of gestation has an error of SD 7 days. Under-registration of very preterm births attributable to early death is also a problem.²

Our findings have important programmatic and research implications for newborn health and survival, particularly because 43% of under-5 deaths happen during the neonatal period. Evidence for the primary prevention of preterm birth and fetal growth restriction is limited. In an analysis modelling high coverage of five evidence-based interventions in countries with a high development index, little reduction was seen in preterm birth rates.⁴⁷ However, the findings underline the importance of further research in settings of low-income and middle-income countries about birth spacing and treatment of maternal infections. Nutritional supplementation programmes (balanced protein–energy supplementation) for women during pregnancy can reduce the risk of small-for-gestational-age births by a third,⁴⁸ although evidence of effectiveness at scale is scarce. Multiple micronutrient supplementation reduces the risk of babies born small for gestational age by 17%;⁴⁹ however, the effect of supplementation varies across populations, with differing baseline rates of malnutrition and access to obstetric care.

By contrast, interventions that improve the care and survival of preterm and small-for-gestational-age infants have major potential for immediate effects and should be prioritised—eg, early feeding support (initiation of breastfeeding, alternative oral feeding methods), kangaroo mother care,⁵⁰ early detection and treatment of neonatal infections,⁵¹ and neonatal resuscitation.⁵² These common interventions improve the management of small babies—whether due to preterm birth or intrauterine growth restriction—and have been proven to have great effect, or are even judged standard care, in high-income settings. Yet, these simple low-cost interventions do not reach those small babies in the settings of greatest need.

Moving beyond birthweight metrics alone and delineating preterm birth and intrauterine growth restriction are important for advancing a healthy start in life. In 2010 in countries of low and middle income, 32·4 million neonates, or one in every four babies, were classified as small for gestational age, closely linked to 13·7 million babies born too soon. Half of infants born small for gestational age were in south Asia, where one of two babies was born too small. To improve the epidemiology and adequately monitor the effect of interventions, systems are needed urgently to better capture and track pregnancy outcomes and to increase the quantity and quality of both birthweight and gestational age data. Effective low-technology interventions are available now to deliver care to these most vulnerable babies born too small or too soon.

This online publication has been corrected. The corrected version first appeared at thelancet.com/lancetgh on July 3, 2013