Intrauterine growth restriction, prematurity, and low birth weight: risk phenotypes of neonatal death, Rio de Janeiro State, Brazil

Abstract: Intrauterine growth restriction and prematurity determine low birth weight. The combination of the three conditions results in different neonatal phenotypes that interfere with child survival. Neonatal prevalence, survival and mortality were estimated according to neonatal phenotypes in the cohort of live births in 2021 in the state of Rio de Janeiro, Brazil. In this study, live births of multiple pregnancies, with congenital anomalies and inconsistencies in the information of weight and gestational age were excluded. The Intergrowth curve was used to classify weight adequacy. Mortality (< 24 hours, 1-6 and 7-27 days) and survival (Kaplan-Meier) were estimated. In total, 6.8%, 5.5%, and 9.5% of the 174,399 live births were low birth weight, small for gestational age (SGA), and premature, respectively. Considering low birth weight live births, 39.7% were SGA and 70% were premature. The neonatal phenotypes were heterogeneous according to maternal, delivery, pregnancy, and newborn characteristics. The mortality rate per 1,000 live births was high for low birth weight premature newborns, both SGA (78.1) and AGA (adequate for gestational age: 61.1), at all specific ages. Reductions in the survival rate were observed when comparing non-low birth weight and AGA term live births. The estimated prevalence values were lower than those of other studies, partly due to the exclusion criteria adopted. The neonatal phenotypes identified children who were more vulnerable and at higher risk of death. Prematurity contributed more to mortality than SGA, and its prevention is necessary to reduce neonatal mortality in the state of Rio de Janeiro.

Ashorn et al. 4 proposed the analysis of mortality risks of more vulnerable newborns, considering birth weight, prematurity and weight adequacy for gestational age in different phenotypes, according to their combination.
Birth weight remains one of the most important markers of maternal and child health, and 2,500g remains the cutoff point to define low birth weight 10 .The World Health Organization (WHO) 11 proposed a 30% reduction births with low weight by 2025, however, the 2015 evaluation showed slow progress towards this goal 10 .About 20 million births with low weight were estimated, resulting in a 14.6% prevalence, with regional differences 10 .It is thus a complex event, and efforts to identify its proximal determinants -prematurity and fetal growth restriction -are recommended 11 .
In 2014, the estimated prevalence of prematurity -births under 37 weeks of gestational age -worldwide was 10.6%, ranging from 8.4% in Europe to 13.4% in North Africa, and 9.8% in Latin America 12 .In 2019, a slight drop to 10.2% was observed worldwide 5 .In Brazil, the estimate from 2011 to 2018 was 9.4% 9 .On the other hand, the emergence of COVID-19 led to an increase in preterm births and severe morbidity and mortality of the mother and child binomial due to the association with maternal infection, in a cohort of pregnant women in early 2020 13 , however, such results are not uniform 14 .
Regarding fetal growth restriction, the prevalence of infants small for gestational age (SGA) is often used as proxy, considering the 10th percentile of growth curves as the cutoff.The estimate for low-and middle-income countries, in 2012, based on the Intergrowth curve, was 19.3%, with 34.2% as the highest value, found in South Asia 8 .Latin America had a prevalence of 8.6% and Brazil of 9% 8 .The Birth in Brazil study, using data from a similar period and population percentiles developed with the data itself, reported 11.1% prevalence 15 .However, another national evaluation, from 2011 to 2018, using percentiles of the Intergrowth curve that restricts the population to the range of 24 to 42 weeks, reported a 9.2% prevalence 9 .
Social and health care inequities such as unfavorable socioeconomic conditions, absence of a partner, extremes of age, non-white skin/color, low maternal schooling, smoking, newborn with previous low weight, hypertensive syndrome during pregnancy, chronic morbidities (lupus and kidney disease), low gestational weight gain, and inadequate prenatal care were associated with fetal growth restriction in the country 15,16 .In Birth in Brazil, the highest attributable fractions were nulliparity, hypertensive syndrome during pregnancy, low gestational weight gain, and smoking 15 .Many of the factors for SGA birth are also related to prematurity, especially extremes of maternal age, low schooling, indigenous ethnicity, black skin color, and absence of a partner 9 .Furthermore, cesarean section is implicated in the increase of preterm birth prevalence, even after adjusting for maternal characteristics such as age, schooling, marital status and parity 17 .
This study estimated the prevalence of low birth weight, fetal growth restriction, prematurity, as well as survival rates and specific neonatal mortality by age according to these phenotypes in the 2021 live birth cohort in the state of Rio de Janeiro, Brazil.

Methodology
This is a retrospective cohort study of live births in 2021 in the state of Rio de Janeiro followed from birth to 27 full days of life.The outcome was specific neonatal death (< 24 hours, 1-6 days, 7-27 days) without nominal and residential identification, were deterministically related using number of the Declaration of Live Birth as key variable.The losses were about 12% and not selective regarding age at death and maternal and neonatal characteristics 19 .
Eligible newborns had birth weight ≥ 500g, gestational age ≥ 22 weeks, and were children of single pregnancy.Records with values of inconsistent weight for gestational age, that is, values outside the consistent ranges of the lowest 3rd percentile value and the highest 97th percentile value 19 , and with reports of congenital anomalies (10th revision of the International Classification of Diseases -ICD-10 -code described in field 34 of the Declaration of Live Births), were excluded, due to their relationship with birth weight, growth restriction, and prematurity 9 .
To classify the newborn according to fetal growth, the adequacy of birth weight for gestational age was used, considered a proxy for this estimate.Although the use of adequacy of weight is criticized for not presenting a perfect correspondence with fetal growth 20,21 , it is a relatively accessible and pragmatic measure.The newborn was defined as having adequate weight for gestational age (AGA: 10th-90th percentiles), small for gestational age (SGA: < 10th percentile), and large for gestational age (LGA: > 90th percentile), using gender-specific curves from the INTERGROWTH-21st 22,23 .This methodology does not consider gestational age of 22 and 23 weeks or greater than 42 weeks.Thus, records with these values were excluded from the analysis.
For the following analyses, in which growth restriction and prematurity were the variables of interest, LGA newborns were excluded.The resulting population was classified as: (1) birth weight -not low birth weight (≥ 2,500g) and low birth weight (< 2,500g); (2) gestational age -term (≥ 37 weeks) and prematurity (< 37 weeks); (3) adequacy of weight for gestational age -AGA and SGA.The resulting combinations were eight: (a) not low birth weight (term AGA, preterm AGA, term SGA, and preterm SGA), (b) low birth weight (term AGA, preterm AGA, term SGA, and preterm SGA).The reference category was not low birth weight term SGA, due to the lower risk of death.
After classification, live births were described according to maternal characteristics.Sociodemographic variables included age group (10-19, 20-34 and ≥ 35 years old); skin color (white, black and brown), and schooling (0-3; 4-7 and ≥ 8 years of study).Parity was evaluated according to the number of previous deliveries (zero: primiparous and ≥ 1: multiparous).The adequacy of access to prenatal care 19 was analyzed in a dichotomous way (no prenatal care and beginning of prenatal care ≥ 4th month or beginning of prenatal care ≤ 3rd month); as well as the mode of delivery (vaginal and cesarean section).Characteristics of the newborn were also evaluated: cephalic presentation at the time of delivery (yes and no), gender (female and male), and fifth minute Apgar score (< 7 and ≥ 7).

Statistical analyses
Absolute and percentage distributions of the variables, adequacy of weight and gestational age, prematurity, and low weight were described for neonatal survivors and by age at death.Mean and the respective 95% confidence intervals (95%CI) were calculated for the variables weight and gestational age according to phenotype.Pearson's chi-square test, Fisher's exact test and ANOVA (statistical significance of 5%) were used.The rate of specific neonatal mortality by age i (NMR -quotient between number of neonatal deaths at age i by the number of live births in 2021) per 1,000 live births, the relative risks and the 95%CI according to the combinations of low weight, prematurity, and adequacy of weight for gestational age were estimated.
The Kaplan-Meier method was used to analyze the survival curves 24 of newborns according to low weight, adequacy of weight and gestational age, and prematurity, as well as the resulting combinations.The survival time (day) was calculated by the difference between the date of death or censoring due to the end of the neonatal follow-up period (27 days) and the date of birth.Neonatal deaths in less than 24 hours were considered as contributing (0.5 day) to the survival estimates.To test the difference between the survival curves, the log-rank statistical test (5% statistical significance) was used.The computer program used was Stata SE, version 12 (https://www.stata.com).
This study was approved by the Ethics Research Committee of the Fluminense Federal University (n.29721320.0.0000.5243,opinion 4.091.556,from June 16 th , 2020).

Results
The 2021 cohort consisted of 189,663 live births, of which 178,733 were eligible.The prevalence rates of SGA and LGA newborns were, respectively, 5.5 and 14.5% (Figure 1).
The prevalence rates of preterm births were: 9.5% for the total, 13.2% for SGA live births, 8.8% for AGA live births, and 13.7% for LGA live births.Regarding low weight, the prevalence rates were: 6.8% for the total, 39.7% for SGA, 5.6% AGA, and 1.1% LGA.Among live births with low birth weight, 70% were premature.
Live births classified in the categories of birth weight, prematurity, SGA, and AGA were heterogeneous according to all maternal sociodemographic characteristics, pregnancy, delivery, and newborn (p < 0.0001) (Table 1).Among the newborns weighing ≥ 2,500g (not low birth weight) and no SGA live births was classified as preterm.Thus, this category was excluded from the analyses.

Figure 1
Diagram of the live birth cohort of 2021, state of Rio de Janeiro, Brazil.
AGA: adequate for gestational age; LGA: large for gestational age; SGA: small for gestational age.
Source: Brazilian Information Systems on Live Births from the Rio de Janeiro State Health Department (databases provided in June 2022 in physical digital format -CD-ROM).
* The phenotypic groups were heterogeneous according to all characteristics analyzed (Pearson's chi-square test p < 0.0001).
Cad. Saúde Pública 2023; 39(6):e00231022 The highest proportions of mothers ≥ 35 years old were among preterm live births, regardless of the adequacy of weight for gestational age.On the other hand, the highest proportion of adolescents were among not low birth weight and SGA full term live births.The lowest schooling level was observed among mothers of SGA and term live births, both low birth weight and not low birth weight.Black skin color presented a higher percentage between SGA with low birth weight and preterm live births.Inadequate prenatal care predominated among low birth weight live births, especially SGA and preterm.Cesarean section and non-cephalic presentation were mostly observed in preterm live births, especially with low birth weight and SGA.Asphyxia occurred mostly in preterm live births, with low weight, regardless of adequacy of weight and age.Boys predominated among preterm and girls among low birth weight live births (Table 1).
Mean weight and gestational age showed significant differences between the seven phenotypes analyzed.Particularly, for the two phenotypes of term and low weight live births, the mean gestational age of SGA live births was about one week higher than AGA live births (p < 0.00001) (Table 2).Low birth weight live births, term and AGA were 37 gestational weeks.Among low birth weight live births, term and SGA, gestational age ranged between 37 and 40 weeks, with the highest frequency, at the 39th week (47.4%), when it reaches 78.8% of the entire distribution (Supplementary Material: https:// cadernos.ensp.fiocruz.br/static//arquivo/supl-een231022_3491.pdf).
The analysis of neonatal mortality showed higher rates for preterm newborns with low birth weight, both SGA and AGA, at all specific ages of neonatal death.In the not low birth weight group, prematurity increased mortality rates more than SGA.The highest risk of neonatal death was among SGA preterm newborns with low birth weight (78.1 per 1,000 live births), followed by AGA preterm newborns with low birth weight (61.1 per 1,000 live births).Regarding the specific age at death, SGA preterm newborns with low birth weight had the highest mortality from 1 to 6 days (47.4 per 1,000 live births), and AGA preterm newborns with low birth weight from 7 to 27 days (20.2 per 1,000 live births).All combinations showed high mortality rates when compared to the not low birth weight AGA category.The highest relative risk (RR = 79) was observed for mortality from 1 to 6 days of preterm SGA low birth weight libe births (Table 3).Note: low birth weight (< 2,500g); not low birth weight (≥ 2,500g); preterm (< 37 weeks), term (≥ 37 weeks).
Neonatal survival rate was higher than 92% in the live birth cohort.The survival proportions of low birth weight and not low birth weight newborns were, respectively, 94.74% (95%CI: 94.32; 95.13) and 99.85% (95%CI: 99.82; 99.86); preterm and term, 95.41% (95%CI: 95.05; 95.75) and 99.85% (95%CI: 99.83; 99.87); and SGA and AGA, 98.66% (95%CI: 98.41; 98.87) and 99.55% (95%CI: 99.52; 99.59)The smallest difference in estimated survival was between AGA and SGA live births (0.84%) and about 5% between term and preterm live births, and not low birth weight and low weight.When the three conditions are combined, significant reductions in neonatal survival are observed, especially for preterm newborns with low birth weight, both for AGA and SGA, the latter being the one with the lowest survival, when compared to not low birth weight and AGA term newborns (p < 0.001).Figure 2 shows the survival of the phenotypes.Preterm SGA live births with low birth weight only presented a higher survival rate than the preterm AGA live births with low birth weight in the first two days of life, although the values were close.Then, survival values are distant, always higher for preterm AGA live births with low birth weight (Figure 2).

Discussion
In the cohort of live births from the state of Rio de Janeiro, in 2021, excluding congenital anomalies, the prevalence of prematurity was 9.8%, 6.8% for low birth weight, and 5.5% for SGA.The risk of neonatal death was more influenced by prematurity than by fetal growth restriction, regardless of the phenotypes analyzed.The highest risk of neonatal death occurred in the combination of the three conditions -low birth weight, SGA and preterm live births.
The prevalence of SGA newborns (without exclusion criteria and using the Intergrowth curve) was equal to the one of this study in the cohort of pregnant women from 2009 to 2012 in the capital of Rio de Janeiro 25 .However, the proportional distribution of phenotypes differs from other Brazilian studies, which estimated higher values for low birth weight and SGA, and similar values for prematurity 9,15 .Souza et al. 15 used data from Birth in Brazil -a survey based on hospital records, between 2011 and 2012 -and estimated the adequacy of weight for gestational age by a curve generated with the data, without using exclusion criteria for live births.Paixão et al. 9 evaluated SINASC and SIM data from 2011 to 2018, using the Intergrowth curve (≥ 24 and < 43 weeks) and excluding live births with weights considered implausible (< 350g or > 6,500g).Applying the criteria of Paixão et al. 9

Figure 2
Neonatal cumulative survival curve according to phenotypes based on weight adequacy for gestational age, as well as prematurity and low birth weight.
Live birth cohort of 2021, state of Rio de Janeiro, Brazil.
AGA: adequate for gestational age; SGA: small for gestational age.
the 2021 cohort data of the present study, the prevalence rates of prematurity, low weight, and SGA would increase to 10.4%, 8.4%, and 6.7%, respectively (data not presented in the tables), but would still be slightly different.Thus, factors, probably populational, may interfere in the distribution of phenotypes.In the cohort of Brazilian live births of single pregnancy, without congenital anomaly, and extremely poor (2012/2015) 16 , the prevalence of SGA (Intergrowth) was 7.8%, the highest.The predominance of prematurity over low weight in the profile of newborns confirms what a study observed in Pelotas (Rio Grande do Sul State) and Argentina in the early 2000s 26 .The constitution of live births with low birth weight was also similar, totaling 70% of preterm live births.
Among the factors that may affect the comparability of phenotype studies are population and methodological factors.The population distribution of gestational ages, birth weight, and fetal growth restriction varies between countries and regions within a same country 8,10,11,12,26,27 .In the study of the Child Health Epidemiology Reference Group SGA -Preterm Birth Working Group, the prevalence of the low birth weight, preterm and SGA phenotype was 0.7% in Latin America, 1.6% in Africa, and 2.3% in Asia 27 .Regarding the methodological choices, the factors include different diagnostic approaches to gestational age 28 , different curves to assess the adequacy of weight for gestational age 22,29,30 , the type of basis of the study -population or hospital -the inclusion and exclusion criteria of newborns, especially at risk, such as twins and those with congenital malformations, who usually have higher frequencies of low weight, prematurity, and fetal growth restriction 31,32,33 .Notably, in Brazil, most deliveries occur in hospitals, which would be less relevant for national studies 15 .
Although the prevalence rates in the state of Rio de Janeiro were different from those of the study by Paixão et al. 9 -both using Intergrowth percentiles, although with different newborn weight restrictions, respectively, below 500g and 350g -the behavior of each phenotype regarding mortality was very similar.In the group of low birth weight live births term, AGA showed higher mortality rate than SGA.This can be explained by the difference in gestational ages: although all were full-term infants, the AGA group was composed only of preterm live births (37 weeks) and were about one week less than the SGA group (Supplementary Material: https://cadernos.ensp.fiocruz.br/static//arquivo/supl-een231022_3491.pdf).Preterm newborns have a higher risk of morbidity and mortality than full-term newborns with a gestational age greater than 39 weeks (not preterm) 34 .
Cad. Saúde Pública 2023; 39(6):e00231022 In African countries, prematurity, alone or combined with fetal growth restriction, presented a higher risk of infant mortality, especially in the neonatal component 6,7 .Wilcox et al. 21tested the validity of low weight, prematurity, and SGA separately to predict neonatal mortality, from the estimation of the area under the ROC curve, concluding that only prematurity presented a good performance to identify at-risk newborns.In the cohort of Danish live births of single pregnancy, weight ≥ 500g, and ≥ 22 weeks of gestation (1981 to 2015), the hypothesis of a mediating effect of the variables weight and gestational age on the association between low maternal education and neonatal mortality was evaluated.Even in the mediating position, preterm birth seemed to have a greater influence than fetal growth restriction 35 .
The concentration of deaths of newborns with higher risk phenotypes in the early component of neonatal mortality of this study corroborates the national pattern of the retrospective cohort of live births followed for longer, up to five incomplete years (2011 to 2018) 9 .In accordance with our results, the greatest impact on neonatal survival rates occurred in the presence of low weight, fetal growth restriction, and prematurity, followed by the phenotypic combination of low weight, AGA, and preterm newborns 9 .In Ethiopia, by only including live births of low birth weight, Debere et al. 36 showed that the largest number of deaths was observed in the preterm and SGA phenotype.Moreover, contrary to our findings, survival rate was lower among the full-term SGA live births with low birth weight than among preterm SGA live births with low birth weight.However, in the Ethiopian study, difficulties in estimating gestational age are described, with 12% of losses due to lack of this information.These two studies also used the Intergrowth curve and the Kaplan-Meier method to estimate survival 9,36 .
Furthermore, the conditions of greater sociodemographic vulnerability were present in the three outcomes.Adolescence and low schooling were more related to fetal growth restriction in full-term infants, corroborating studies in other low-and middle-income countries 37,38 .Older maternal age emerged among preterm live births, according to a study in the city of Rio de Janeiro 19 .Black people and inadequate prenatal care were frequent in the extreme combination -preterm SGA -which was an expected result since both are corroborated factors for these conditions.However, these results were adjusted and the factors cannot be assumed as determinants.
This study presents some limitations due to losses, since information was either lacking or inconsistent regarding weight for gestational age.A possible lack of measurement accuracy, particularly of gestational age 39 and the possible underestimation of SGA with the use of the Intergrowth curve, when compared to other growth curves 40,41 , may have led to the imprecision of some phenotypes analyzed.Notably, other studies identify that, despite the restrictions of this curve, the results of accuracy to predict neonatal mortality, compared to those of other curves, are similar 42,43 .
A strong point of the study was highlighting the representativeness of the results considering the reconstitution of the cohort of live births from the data available in the health information systems that have good national coverage and, mostly, in the state of Rio de Janeiro 39 .The Intergrowth fetal growth curve is universal and provides greater specificity in the classification of newborns according to the adequacy of birth weight by gestational age, resulting in a lower frequency of false SGA live births 40 .
Therefore, the estimated prevalence rates of the conditions of risk of low weight neonatal death, SGA, and prematurity showed lower values compared to other studies, partly due to methodological nuances.Furthermore, population differences may have contributed to different phenotypic combinations, as reported in other studies 26,27 .
Phenotypes resulting from the combination of these conditions identified more vulnerable children in the present study, highlighting the contribution of prematurity.Fetal growth restriction and prematurity share causes and consequences and are interlinked in the determination of low weight and, consequently, neonatal morbidity and mortality 27,44,45 .There are differences in the profile of newborns at risk according to the distribution of these conditions, which may require different interventions 27 .In the population studied, the lowest survival rate was in the presence of all three conditions, but prematurity, regardless of the presence of fetal growth restriction, showed greater magnitude, reinforcing its validity as a predictor of neonatal death 21 , and the need for its prevention for a greater reduction in neonatal mortality in the state of Rio de Janeiro.

Table 1
Maternal and live birth characteristics classified according to the conditions: adequacy of weight and gestational age, prematurity and low birth weight.Live birth cohort of 2021, state of Rio de Janeiro, Brazil.
AGA: adequate for gestational age; SGA: small for gestational age.Source: Brazilian Information Systems on Live Births from the Rio de Janeiro State Health Department (databases provided in June 2022 in physical digital format -CD-ROM).

Table 2
Mean gestational age (weeks) and birth weight (grams) by components according to prematurity, low weight, and adequacy of weight and gestational age.Live birth cohort of 2021, state of Rio de Janeiro, Brazil.

Table 3
Mean gestational age (weeks) and birth weight (grams) by components according to prematurity, low weight, and adequacy of weight and gestational age.Live birth cohort of 2021, state of Rio de Janeiro, Brazil.
95%CI: 95% confidence interval; AGA: adequate for gestational age; SGA: small for gestational age.Source: Brazilian Information Systems on Live Births and Brazilian Mortality Information System from the Rio de Janeiro State Health Department (databases provided in June 2022 in physical digital format -CD-ROM).