Our objectives were to examine the associations of neonatal and infant mortality with preterm birth and intrauterine growth restriction (IUGR), and to estimate the partial population attributable risk per cent (pPAR%) of neonatal and infant mortality due to preterm birth and IUGR.
Participants were HIV-negative pregnant women and their infants enrolled in Dar es Salaam, Tanzania. Gestational age calculated from date of last menstrual period was used to define preterm, and small for gestational age (SGA) was used as proxy for IUGR. Survival of infants was ascertained at monthly follow-up visits. Cox proportional hazard models were used to estimate the associations of preterm and SGA with neonatal and infant mortality.
Study included 7225 singletons, of whom 15% were preterm and 21% were SGA; majority of preterm or SGA babies had birthweight ≥2500 g. Compared to term and appropriately sized babies (AGA), relative risks (RR) of neonatal mortality among preterm-AGA was 2.6 [95% CI 1.8, 3.9], RR among term-SGA was 2.3 [95% CI 1.6, 3.3], and the highest risk was among the preterm-SGA babies (RR 15.1 [95% CI 8.2, 27.7]). Risk associated with preterm was elevated throughout the infancy, and risk associated with SGA was elevated during the neonatal period only. The pPAR% of neonatal mortality for preterm was 22% [95% CI 17%, 26%] and for SGA it was 26% [95% CI 16%, 36%].
Preterm and SGA birth substantially increased the risk of mortality. Interventions for prevention and management of these conditions are likely to reduce of infant mortality in Tanzania.
Low birthweight (LBW) is a major predictor of neonatal and infant death. LBW is caused by shortened duration of pregnancy (preterm birth) or intra-uterine growth restriction (IUGR), or both. Small for gestational age (SGA) is frequently used as a proxy for IUGR, which is an end result of the ongoing process of growth restriction and is defined by a certain cut-off values of birthweight for gestational age.[3, 4] Since gestational age is not known for most births in developing countries, the relative contribution of preterm and IUGR to deaths attributable to LBW has not been well characterised. Furthermore, a limited number of studies have examined the mortality risk among preterm and growth restricted infants who are born with normal birthweight.
Distinguishing between prematurity and IUGR is important for understanding and preventing LBW-associated mortality risk in a country. Although there are risk factors in common, the relative importance of these risk factors for IUGR and preterm birth are different.[5, 6] While maternal undernutrition are related to both IUGR and preterm birth, the hypertensive disorders of pregnancy and infections including malaria, HIV, and UTI during pregnancy are the immediate risk factors leading to preterm birth in developing countries.[2, 7] Therefore, combining preterm and IUGR together as LBW may impede development of preventive interventions. Some studies suggest that LBW in developing countries is mostly due to IUGR rather than preterm birth,[8, 9] while more recent studies report that the rates of preterm birth are actually quite high in these countries. Studies from developed countries show that preterm and SGA infants are at higher risk of mortality in first year and the highest risk is observed among infants who are both preterm and SGA.[11, 12] However, the mortality risk associated with SGA and preterm birth, and the population attributable risk due to these conditions, are likely to be different in low-resource settings than high-resource settings, due to differences in rates of SGA and preterm births, and also the limited health care facilities to support small or preterm babies. In the present study, we examined the association of preterm and SGA births with infant mortality in a cohort of infants born to HIV-negative mothers in Dar es Salaam, Tanzania. Understanding preterm and SGA-associated infant mortality risk is important for the prioritisation of interventions aimed at achieving the MDG 4, especially in the context of Tanzania, which is among the top 10 countries with highest child mortality.
Study design and population
This study is nested within a randomised trial which examined the benefits of maternal multiple micronutrient supplements on pregnancy outcomes. From August 2001 to July 2004, 8428 pregnant women were enrolled in the trial from four antenatal clinics in Dar es Salaam, Tanzania. Women were eligible for the trial if they had a negative test for HIV infection, an estimated gestational age at enrolment between 12 and 27 weeks, and if they planned to stay in the city until delivery and for 1 year thereafter.
Information on socio-demographic characteristics and obstetric history was collected at baseline. A study nurse measured height and weight of the pregnant women and a study physician performed a complete clinical examination. Women's hemoglobin concentration was measured at baseline. Women were then followed up monthly until the 32nd week of pregnancy, then every 2 weeks till 36th week, and then weekly until delivery. Women who did not come to the clinic within 3 days of their expected delivery date were visited at home to assess the outcome of pregnancy.
Birthweight of the babies was measured to 10-g precision by trained research midwives at delivery. Gestational age of the babies was calculated based on the mother's date of last menstrual period (LMP) as reported at baseline. The vital status of the child was recorded at discharge for hospital deliveries. During the postpartum period, mothers came to the study clinics with their babies monthly until 18 months following delivery. The survival status of the child was ascertained at these visits. Women who missed their appointment were visited at home. If they left Dar es-Salaam, the vital status of the child was ascertained from neighbours and relatives.
Using a recent US population-based standard developed by Oken et al. newborns were grouped based upon their weight for gestational age into the following categories: small for gestational age (SGA), appropriate for gestational age (AGA), or large for gestational age (LGA). SGA was defined as a birthweight <10th percentile for gestational age; AGA was defined as birthweight ≥10th percentile for gestational age; LGA was defined as birthweight >90th percentile. In order to assess mortality risk in extremely growth-restricted infants, we further classified the severe SGA babies as birthweight <3rd percentile, and moderate SGA as birthweight between 3rd and <10th percentile. We defined preterm as birth before 37 weeks of gestation. Late preterm was defined as births at 34 through 36 completed weeks of gestation, and early preterm was defined as birth before 34 completed weeks.
The Kaplan–Meier method was used to construct cumulative incidence curves of mortality during the first year. Crude and adjusted relative risks of mortality were estimated for the three exposure categories defined by combinations of preterm and weight for gestational age: (1) preterm-AGA, (2) term-SGA, and (3) preterm-SGA. Each of these categories was compared to term-AGA as the reference category. We used Cox proportional hazards model with time at death as the outcome to estimate the hazard ratios and 95% confidence intervals of neonatal, post-neonatal, and infant mortality. We also estimated the risk of mortality for early and late preterm births and severe and moderate SGA separately. To control confounding, multivariable models were adjusted for maternal age, BMI, height, hemoglobin level, marital status, education, multivitamin intake, and household possessions. To control as finely for confounding as possible all known or suspected risk factors for infant and neonatal mortality measured in the study were included in the multivariate models. Risk factor data were collected at the time of enrolment. We used missing indicator terms in the multivariate models for model covariates with missing data.
We estimated the point and confidence intervals for the partial population attributable risk of mortality due to preterm birth and SGA using the method described by Spiegelman et al. This method provides estimates of the percent of deaths associated with the exposure category from the multivariable model, assuming that the distribution of the other risk factors in the population remains unchanged.
We performed sensitivity analysis after including the twins (total 251, 32 deaths). Further analysis was conducted after excluding the LGA babies (total 304, 14 deaths). However, because the results did not change after the exclusion of the LGA babies, we retained them in the main analysis. All statistical analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC, USA).
During the study period, 8137 babies were born live. Survival status was known for 7752 babies. We excluded babies for whom birthweight was not recorded (182, 2.3%), gender was not recorded (20, 0.26%), and babies with improbable gestational age and birthweight combinations (72, 0.93%). We excluded the twins as they carry higher risk of prematurity and mortality compared to the singletons. After exclusion of twins (251, 3.2%), 7225 singleton remained for primary analysis. There were 184 deaths during the neonatal period (neonatal mortality rate = 25.5/1000 live births) and 291 deaths during first year (infant mortality rate = 40.3/1000 live births). Most (169, 92%) of the neonatal deaths occurred within first 7 days of life.
Table 1 shows the maternal characteristics at baseline. About 45% of mothers were 25 years of age or older, over 60% were currently married, and more than half of them had had at least one prior pregnancy. About 67% of mothers had 5–7 years of education and about half of the families owned more than three common household possessions.
Table 1. Basic characteristics of the mothers (n = 7725)
Of the 7225 singletons, 1089 (15%) were preterm and 1581 (22%) were SGA. A large proportion of babies with normal birthweight were either preterm (867, 12%) or SGA (1272, 17.6%). Only 474 (6.6%) were LBW; among them 252 (3.5%) were SGA, 165 (2.3%) were preterm, and 57 (0.8%) were both preterm and SGA (Figure 1). None of the normal birthweight babies were both preterm and SGA; and the LBW babies who were both preterm and SGA, the majority were late preterm (data not shown). The standard definition of LBW (<2500 gm) identified a group infants at very high risk of mortality, but this classification left out the 12% of infants who were preterm-AGA ≥2500 g and the 18% of infants who were term-SGA ≥2500 g. These two large groups form the majority of infants who were at a high risk of mortality compared with term-AGA infants (Figure 2).
The risk ratios for mortality throughout infancy are given in Table 2. Compared to term-AGA infants, preterm-AGA infants had two to three times higher risks of neonatal, post-neonatal, and infant mortality. Approximately twofold higher mortality risk was observed among the term-SGA infants compared with term-AGA infants. Among the term-SGA infant,s the risk of mortality was not significantly elevated during the post-neonatal period. An extremely high RR of mortality was observed among the preterm-SGA group both during the neonatal (RR 15.1 [95% CI 8.2, 27.7]) and infancy periods (RR 10.0 [95% CI 5.8, 17.4]. After inclusion of the twins in the analysis, the RR of neonatal mortality among preterm-AGA infants rose to 3.4 [95% CI 2.4, 4.8], while the estimate for the term-SGA group did not change substantially (results not shown).
Table 2. Mortality outcomes by gestational age-size at gestational age category and birthweight in the early neonatal, neonatal, post-neonatal, and infancy periods
aHazard ratios and 95% CI were obtained from Cox proportional hazards models.
bAdjusted for maternal age (<20, 20–24, 25–29, and ≥30 years), BMI (<18.5, 18.6–24.99, 25–29.99, and >30 kg/m2), height (<145, 145–149, 150–154, and ≥155 cm), hemoglobin level (<8.5, 8.5–11, and ≥11 g/dl), marital status (married, cohabiting, poly married, and unmarried), parity (0, 1, 2, and ≥3), maternal smoking (ever and never smoker), maternal education (0–4, 5–7, 8–11, and ≥12 years), household possession (0, 1–3, and ≥3), and multivitamin intake (iron and folate only, multivitamin with iron and folate).
AGA, appropriate for gestational age; SGA, small for gestational age.
2.5 [1.7, 3.8]
2.6 [1.8, 3.9]
2.5 [1.6, 3.9]
2.6 [1.9, 3.5]
2.3 [1.6, 3.3]
2.3 [1.6, 3.3]
0.9 [0.6, 1.6]
1.7 [1.3, 2.3]
12.3 [6.2, 24.3]
15.1 [8.2, 27.7]
2.6 [0.6, 11.9]
10.0 [5.8, 17.4]
6.9 [4.9, 9.6]
7.5 [5.5, 10.3]
3.2 [1.9, 5.3]
5.8 [4.5, 7.6]
The risk ratios for mortality among early preterm, late preterm, moderate SGA, and severe SGA babies are shown in Table 3. Compared to term infants, early preterm infants were seven times more likely and late preterm infants were two times more likely to die in the neonatal period. Similar elevations of risk were observed for post-neonatal and infant mortality. The risk of mortality among the SGA infants was elevated during neonatal period only (Table 3). Severe SGA was associated with more than fourfold higher risk of neonatal mortality and moderate SGA was associated with a doubling of neonatal mortality compared to AGA infants.
Table 3. Mortality outcomes by gestational age and size at gestational age in the early neonatal, neonatal, and post-neonatal and infancy period
aHazard ratios and 95% CI were obtained from Cox proportional hazards models.
bAdjusted for maternal age (<20, 20–24, 25–29, and ≥30 years), BMI [ < 18.5, 18.6–24.99, 25–29.99, and >30 kg/m2), height (<145, 145–149, 150–154, and ≥155 cm), hemoglobin level (<8.5, 8.5–11, and ≥11 g/dl), marital status (married, cohabiting, poly married, and unmarried), parity (0, 1, 2, and ≥3), maternal smoking (ever and never smoker), maternal education (0–4, 5–7, 8–11, and ≥12 years), household possession (0, 1–3, and ≥3), and multivitamin intake (iron and folate only, multivitamin with iron and folate).
AGA, appropriate for gestational age; SGA, small for gestational age.
Term (≥37 weeks)
Late preterm (34 to <37 weeks)
1.8 [1.2, 2.9]
2.0 [1.3, 3.2]
1.5 [0.9, 2.7]
1.9 [1.3, 2.7]
Early preterm (<34 weeks)
6.9 [4.4, 10.9]
7.6 [4.9, 11.6]
5.8 [3.4, 10.1]
6.9 [4.9, 9.7]
Size at gestational age
AGA (birthweight >10%)
Moderate SGA (birthweight 3–<10%)
1.8 [1.1, 2.8]
1.9 [1.2, 2.9]
1.1 [0.6, 1.9]
1.5 [1.1, 2.2]
Severe SGA (birthweight <3%)
4.2 [2.8, 6.2]
4.4 [2.9, 6.5]
0.9 [0.5, 1.9]
2.9 [2.1, 4.0]
Table 4 shows the partial population attributable risk percent (pPAR%) of neonatal and infant mortality due to preterm and SGA. Twenty-one per cent of neonatal mortality in this cohort was attributable to preterm birth. The estimated pPAR% of neonatal mortality in relation to early preterm birth was 15% [95% CI 13%, 17%], and for late preterm birth, the pPAR% was 7% [95% CI 4%, 9%]. The pPAR% for neonatal mortality in relation to SGA was quite high (26% [95% CI 16%, 36%]). Forty-six per cent [95% CI 33%, 57%] of all neonatal deaths in this cohort could be prevented by jointly eliminating preterm and SGA births.
Table 4. Population attributable risk percent of preterm birth, small for gestational age for mortality during early neonatal, neonatal, and infant mortality
aAll models included early preterm, late preterm, severe SGA, moderate SGA, and interaction terms between preterm and SGA variables. The variable for which pPAR% was estimated was treated as modifiable variable in the respective models; the remaining variables were treated as fixed variables.
bAdjusted for maternal age (<20, 20–24, 25–29, and ≥30 years), BMI (<18.5, 18.6–24.99, 25–29.99, and >30 kg/m2), height (<145, 145–149, 150–154, and ≥155 cm), hemoglobin level (<8.5, 8.5–11, and ≥11 g/dL), marital status (married, cohabiting, poly married, and unmarried), parity (0, 1, 2, and ≥3), maternal smoking (ever and never smoker), maternal education (0–4, 5–7, 8–11, and ≥12 years), household possession (0, 1–3, and ≥3), and multivitamin intake (iron and folate only, multivitamin with iron and folate).
pPAR%, partial population attributable risk percent; SGA, small for gestational age.
14.1 [12.6, 15.5]
14.9 [13.1, 16.8]
15.7 [15.3, 16.1]
5.7 [2.6, 8.8]
6.6 [4.0, 9.2]
6.0 [4.1, 7.9]
19.8 [14.2, 24.5]
21.5 [17.2, 25.8]
21.7 [19.0, 24.4]
18.1 [11.9, 24.3]
18.9 [13.0, 24.7]
12.2 [3.8, 20.4]
6.8 [2.0, 11.5]
7.5 [2.8, 12.1]
5.1 [2.2, 7.9]
24.9 [14.4, 34.8]
26.3 [16.3, 35.9]
17.3 [4.7, 29.3]
Preterm and SGA
43.1 [29.5, 54.9]
45.9 [33.3, 56.9]
38.2 [24.1, 50.7]
Our study evaluated the risk of neonatal and infant mortality among the groups of infants defined by preterm birth and weight for gestational age. Although all the preterm-SGA babies were LBW, a large proportion of normal birthweight babies were preterm-AGA or term-SGA. Compared with term AGA infants, mortality risk was elevated in all the three groups, with the highest magnitude of risk among the babies who were both preterm and SGA. Similar to most developing country settings, a large proportion of LBW in our study population was SGA.[8, 9] However, our study population also had a higher incidence of preterm birth (15%) compared to estimated average rate of preterm birth in sub Saharan Africa (12.3 %).
LBW is often referred to as a marker of growth restriction in developing country context. The magnitude of relative risk of death among LBW infants observed in our study is similar to the estimates reported in the past studies from Tanzania and other developing countries.[21, 22] Our estimate of the association of SGA (birthweight <10th percentile) with the risk of mortality is consistent with the results of the meta-analysis by Marchant et al. conducted in four East African countries. Our finding that the risk of mortality among term-SGA infants was more than twofold greater than term-AGA infants is also consistent with results from several studies conducted in developed countries.[12, 24, 25] Similar to those studies, we found that the risk of mortality was very high among SGA infants who were also preterm, demonstrating the need for extra care for these newborns. Similar to our findings, Chen et al. found that the elevated risk among SGA infants was limited to the neonatal period.
There have been few previous efforts to examine the association between preterm birth and neonatal mortality in the developing countries. In a study of 795 mother–infant pairs in rural Malawi, the odds of neonatal mortality among preterm babies was 11 times greater than that of term babies. In the meta-analysis by Marchant et al., the odd ratios of neonatal mortality were 6.3 and 58.7 among early and late preterm babies, respectively. Although the overall conclusions are similar to our findings, the results may not be directly comparable as different methods of gestational age estimation were used in these studies. In the Malawi study, gestational age was estimated from a single measurement of fundal height. Because fundal height is affected by both gestational age and fetal growth, this method provides a highly inaccurate estimate. Only 4% of the study population was classified as preterm in the study conducted by Marchant et al. which used ultrasound-based assessment of gestational age in one site and neonatal assessment in three other sites. While ultrasound assessment is considered the most accurate among all methods of gestational age estimation, neonatal assessment is less precise than the LMP-based estimates.
Our results show that the increased mortality risk associated with preterm birth was highest during the neonatal period, and the risk was attenuated but remained significantly high during post-neonatal period. Compared to term infants, the risk of neonatal mortality among early preterm and late preterm babies was five and two times higher, respectively (Table 3). Contrary to the notion that the risk of mortality among preterm infants would be higher in settings where health systems lack facilities to support preterm babies, the magnitude of the association estimated from our adjusted model was much lower than reported estimates from developed countries. An analysis of British Columbia Perinatal Database showed that late preterm babies had 5.5-fold higher risk of neonatal mortality compared to term births. Another study reported a risk ratio of 5.2 for the US and 7.9 for Canada among late preterm infants, with the risk of mortality even higher for early preterm infants (RR 6.6 and 15.6 for the US and Canada, respectively). The very low mortality rate among the term infants in developed country settings is the likely explanation for the difference in the risk ratios observed in the two settings. The difference between the rate of stillbirth in our study population (3.5%) compared to stillbirth rate in the population in the US or Canada (1.13%) is a possible explanation for the comparatively smaller proportion of early preterm birth in our study population. Our estimates may also be an underestimate of the true risk, as we excluded the newborns with unknown weight (born at home) and gestational age from our analysis. Some studies suggest that mothers with low socioeconomic status have the poorest LMP recall and are more likely to present for a late antenatal check-up and deliver at home, and babies born to these women are at higher risk of adverse outcomes.[31, 32]
Our estimates of the pPAR% for neonatal and infant mortality showed that preterm birth and SGA together are responsible for approximately 45% of the neonatal death. SGA was associated with a larger fraction of neonatal death, while preterm birth was associated with a larger fraction of infant death. Our findings indicate that investing in interventions for preventing IUGR and preterm births as well as for managing babies born SGA and preterm is likely to contribute to a major reduction of infant mortality in Tanzania. Scaling up interventions with proven efficiency, e.g. antenatal corticosteroid for preterm labor, Kangaroo mother care, early and exclusive breastfeeding, and treatment of neonatal infections may be feasible in the context of Tanzania and other developing countries.[33-36]
Our study has both strengths and limitations. Gestational age estimation based on the date of LMP depends on women's recall ability and, therefore, was subject to measurement error. This might lead to differential misclassification in preterm birth leading to an underestimate of the true risk of mortality due to preterm birth. Another potential limitation is that we used US standards for classification of SGA and AGA infants. There is a weight for gestational age standard based upon liveborn Tanzanian children available. However, this standard was developed in 1979 and has not been updated. US population-based standard may provide a reasonable reference population as it is a better reflection of growth potential not affected by nutritional deprivation, compared to local norms.
We have examined the associations of prematurity and IUGR on neonatal and infant mortality in a large sample of individual level data in sub Saharan Africa. Gestational age specific-mortality data for such large sample size is not very common in developing countries. Data used to determine SGA births was measured immediately after birth by the trained midwives. We had detailed information on many potential confounders, unlike most studies from developing countries, and some studies from developed countries which used registry data and therefore were not able to adjust for some potential confounders. However, since our estimates did not change much after adjusting for the confounders, there may have been limited confounding in the reported estimates from registry studies.
In conclusion, preterm and SGA substantially increased the risk of mortality, and a large proportion of neonatal deaths were attributable to these conditions. Incorporating gestational age and SGA information in routine newborn assessment should be considered in resource-limited settings. Further research on risk factors for preterm and SGA birth in this region is needed to guide resource allocation for scaling up interventions aiming at reduction of infant mortality and thereby achieving the MDG 4, which calls for reduction of under-five mortality rate by two-thirds by 2015.
We thank the mothers and children, and the field team including nurses, midwives, supervisors, and laboratory personnel who made this study possible. We also thank Ellen Hertzmark of Harvard School of Public Health and Anne CC Lee of Brigham and Women's Hospital, Boston for their inputs in the analysis of this study.
Funding Source: National Institute of Child Health and Human Development (NICHD R01 37701).
Authors do not have any financial relationships relevant to this article to disclose.
Conflict of interest
Authors do not have any conflict of interest to disclose.