Background
Recent global estimates suggest that more than 1 in 10 or an estimated 15 million babies born in 2010 were preterm, of which more than 1 million died as a result of preterm birth and related complications [
1]. Although neonatal mortality rates have fallen globally between 1990 and 2009 [
2], the absolute numbers and rates of preterm birth have increased during this period [
3]. Preterm birth complications account for 35% of the estimated 3.1 million global neonatal deaths [
4], and are the second leading cause of death in children under 5 years of age. The vast majority (85%) of global preterm births occur in Asia and Africa [
5] where health systems are weak and access to and utilization of health services are limited, contributing to the higher risks of death and disabilities in preterm babies [
6,
7]. Approximately one-third of preterm survivors suffer from severe long-term neurological disabilities, such as cerebral palsy or mental retardation [
8]. Furthermore, preterm infants carry increased risk of a range of neurodevelopmental impairments and disabilities, including behavioral problems, school learning difficulties, chronic lung disease, retinopathy of prematurity, hearing impairment, and lower growth attainment [
9]. Preterm birth affects not only infants but also their families who may have to spend substantial time and financial resources to ensure care for their preterm infants; thus, preterm birth has increasing cost implications for families and health services [
10].
Identification of at-risk women and their risk factors for preterm birth is important for targeting of services and initiation of risk-specific interventions and/or preventive measures. Study of risk factors might also provide important insights leading to new discoveries for prevention and management of preterm births. We describe the burden and associated risks factors of preterm birth in a cohort of rural Bangladeshi women.
Methods
Study design
We analyzed prospectively collected data from a large community-based cluster-randomized trial conducted in Sylhet district of Bangladesh. Data for our study were primarily collected to evaluate the impact of two regimens of umbilical cord cleansing - single-day vs. 7-day - with 4.0% chlorhexidine solution on all-cause neonatal mortality and incidence of cord infections [
11,
12].
Study setting and population
The study was implemented during June 2007- September 2009 in 22 unions (the smallest administrative unit with a health center) in 3 rural sub-districts (called upazila) of Sylhet district (Beanibazar, Zakiganj and Kanaighat) in north-eastern Bangladesh with an estimated total population of 546,000 people. The study area was divided into 133 working units (clusters) each served by a female community health worker (CHW). CHWs implemented the interventions and collected data from respondent women and their babies.
Study implementation
Details of the study designs, interventions, delivery strategies and the map of the area have been published elsewhere [
11,
12]. Briefly, CHWs followed a complete map of all households and thus prepared a complete list of all married women of reproductive age (MWRA) through house-to-house visitation and recorded their names, addresses and pregnancy status. The list was updated and new pregnancies were identified every two months by conducting home visits. All newly identified pregnant women were invited to participate in the study and explained the study procedures. Those agreeing to participate gave informed oral consent and provided data on age, parity, date of last menstrual period, literacy, a brief pregnancy history, and socio-economic information about the household.
All enrolled women were provided with a package of maternal and newborn health interventions, delivered by CHWs through two antenatal home visits. The first session was conducted at the time of enrolment at 12–16 weeks of pregnancy and the second occurred at approximately 32 weeks of pregnancy. The intervention package included a clean delivery kit (CDK), messages on birth and newborn care preparedness (BNCP), and advice on essential newborn care (immediate breastfeeding, thermal care and clean cord care) and on neonatal danger sign recognition and care-seeking [
11,
13]. At each visit, information was collected by CHWs on status of birth and neonatal care preparedness, antenatal care (ANC), complications during pregnancy and care seeking for those complications. BNCP included practice of the following 6 steps: 1) selection of birth attendant, 2) selection of newborn care personnel, 3) arrangement for three pieces of cloth for drying/wrapping of the newborn, 4) arrangement for transport for any emergency need, 5) savings for management of complications, and 6) having a CDK for use during delivery.
Inclusion and exclusion criteria
All reported live births within the study area for which data was available on the first day of the last menstrual period (LMP) were included in this study. Reported stillbirths and abortion (spontaneous, induced or therapeutic) were excluded. Pregnancies terminated before 28 weeks of gestation were defined as miscarriage/abortion. A stillbirth was defined as an infant born without any signs of life (no spontaneous crying, breathing, and/or movement) at 28 weeks of gestation or later.
Assessment of exposure variables
Socio-demographic and economic information (women’s age at delivery, educational attainment of women and their husbands, basic housing construction, household belongings, religion) and previous pregnancy history were collected by CHWs using a structured instrument in face-to-face interviews during the enrolment visit. Relevant data on antenatal care seeking, compliance with BNCP, consumption of supplied iron tablets, TT immunization dosage and antenatal complications (history of fever, severe abdominal pain, swelling of hand, leg or face, vaginal bleeding, convulsion, severe headache, blurring of vision) were also collected from all women during BNCP visits or the first postpartum visit. Compliance with BNCP was categorized as “fully compliant” (woman reported practice all 6 of the above-mentioned steps), “partially compliant” (1–5 steps), or “non-compliant” (0 steps). CHWs measured the mid upper arm circumference (MUAC) of the enrolled mothers during enrollment visits. Data on antenatal complications (except fever) were self-reported by respondent women. CHWs measured axillary temperature from women who reported having fever during the interview.
Assessment of outcome variable
The primary outcome was preterm birth as defined by the World Health Organization as: “Any birth before 37 completed weeks of gestation or fewer than 259 days since the first day of the women’s LMP” [
14]. Gestational age at birth was computed from the difference between the date of pregnancy outcome and the date of the first day of the LMP recorded at enrolment. Date of first day of the LMP was determined through maternal report to a CHW during a two-monthly pregnancy surveillance visit at the household, when the CHW asked the pregnant women to recall LMP with the assistance of calendars and memory aids. Women for whom no date of LMP was estimated after several attempts using various approaches were excluded from analysis. Date and type of pregnancy outcome was recorded by a CHW on her first visit after women delivered, usually within 24 hours or as soon as possible after birth.
Data quality assurance
CHWs received 6 weeks of classroom-based and hands-on supervised training. These training sessions followed a structured curriculum including skills development for behavior change communication, provision of BNCP and essential newborn care, clinical assessment of neonates, and identification and management of sick newborns using an algorithm adapted from Integrated Management of Childhood Illness. Quality of data collected by CHWs was ensured through direct supervision by respective Field Supervisors. Supervisory visits and standardization exercise sessions were organized routinely to ensure quality of data collected. Every reported neonatal death was confirmed by a repeat visit to the household by a supervisory staff. A sample (5%) of newborns who survived the neonatal period was revisited for quality assurance of vital status reporting by CHWs.
CHWs submitted data forms to their supervisor, who checked the forms for completeness and consistency. Data entry system was custom-designed with built in range and consistency checks. Field verifications were conducted to resolve identified inconsistencies and incompleteness if required.
Statistical analyses
Births at ≥37 weeks were classified as term births. Preterm births were sub-categorized as: 1) Very preterm (28 – 31 weeks of gestation), 2) Moderate preterm (32–34 weeks of gestation) and 3) Late Preterm (35–36 weeks of gestation). We estimated the incidence of preterm birth by dividing all live preterm births, whether singleton, twin or higher order multiples, by all live births in the population; 95% confidence intervals (CI) were calculated for the estimated proportion of preterm birth incidence.
Based on published reports and considering biological plausibility, we categorized,
a priori, the potential risk factors for preterm birth into three groups: 1) Proximal factors: antenatal care seeking and antenatal complications during the index pregnancy; 2) Intermediate factors: previous pregnancy history; 3) Distal factors: socio-demographic characteristics. We constructed a wealth index score [
15] for each household by employing principal component analysis of basic housing construction materials (i.e., materials used to construct the walls, roof, and floor of houses), sources of water, sanitation facilities and household belongings.
Covariates showing moderate strength of association (
P <0.10) in bivariate analysis were included in multivariate models, which were constructed in a three-step sequence starting with proximal factors [Model1] and progressively adding previous pregnancy/birth information [Model2] and thereafter distal/socio-demographic factors) [Model3]. The association (risk ratio) between potential risk factors and preterm birth was modeled using binomial regression with a log link function; generalized estimating equations with exchangeable correlation structure were used to adjust standard errors to account for clustering [
16,
17]. In case of convergence failure, poisson models with robust standard error estimation were used [
18,
19]. Imputation of missing data was done using the “hotdeck” method by cluster [
20]. Data were analyzed by using STATA (version 11) statistical package [
21].
Ethical approval
We received ethical approval for the study from the Institutional Review Board of Johns Hopkins Bloomberg School of Public Health and from the Ethical Review Committee of International Centre for Diarrheal Disease Research, Bangladesh (icddr,b). The study was registered at ClinicalTrials.gov (NCT00434408).
Discussion
We have presented the incidence of and risk factors for preterm birth using prospectively collected data from a large cohort of 32,126 pregnant women in a rural population of Bangladesh who delivered a live-born infant. Among live-born babies, more than one-fifth was preterm (22.3%) and the majority of preterm births (55.1%) were late preterm. If we could support these pregnancies to continue an additional 1–2 weeks, that could lead to a substantial decrease in the preterm birth toll and burden of disease due to preterm. A small of number of behavioral (e,g. smoking cessation), clinical (e,g. progesterone supplementation) and health system interventions (e,g. reducing non-medically indicated labor induction or caesarean delivery) have been shown to reduce the preterm birth rate [
3,
22]. Bangladesh was ranked 7
th on the top-10 country list for high preterm birth rates in 2010 [
1]. Although recent global estimates reported that sub-Saharan Africa and South Asia account for the majority (60%) of the globally estimated 14.9 million annual preterm births [
1], available data on preterm birth rates from South Asian countries are scarce. Our estimate of 22.3% is consistent with data from similar regional community-based research sites in southern Nepal (NNIPS, Sarlahi, Nepal - 19%, [
23,
24]) and northwestern Bangladesh (JiVITA, Gaibandha, Bangladesh – 23%) [
25].
Maternal factors increasing the risk of preterm birth in our population included socioeconomically poorer status, poor nutritional status (lower MUAC), antenatal iron consumption, history of a previous child death, multiple pregnancies and having any antenatal complication. Factors which were protective for preterm birth included practicing all 6 BNCP steps, first child, education at primary level (grade 5) or above, TT immunization, and female sex of the baby. Study results from Ahmedabad, India [
26] reported previous child death as a risk factor for preterm birth, ranging from 1.5 times (1 child death; 95% CI: 0.9, 2.2) to 3.1 times (2 child deaths; 95% CI: 1.5, 6.4), compared to those who had no previous child death. Unlike our study, they found that first pregnancy was a risk factor as well (RR: 1.3; 95% CI: 0.9 – 1.9). Multiple gestations—accounting for only 2–3% of infants—carry a substantial risk of preterm delivery, resulting in 15–20% of all preterm births; nearly 60% of twins are born preterm [
27]. A Zimbabwe study [
28] reported similar increased risk of preterm birth attributable to multiple gestations (RR: 3.45; 95% CI: 3.1, 3.8) as ours.
Maternal nutritional status before and during pregnancy may contribute to the risk for preterm birth [
29]. In the Preterm Prediction Study, a low pre-pregnancy body mass index (BMI) was strongly associated with increased risk of preterm birth, with the RRs being greater than 2.5 [
30]. In contrast, a recent meta-analysis found that pre-pregnancy BMI had little or no relationship with the risk of preterm birth overall [
31]. In our study, we found women having lower MUAC were at higher risk of preterm birth. Preterm birth can be caused by maternal thinness associated with decreased blood volume and reduced uterine blood flow [
32]. Further research is required to understand the relationship between BMI and preterm birth risk.
Our finding on consumption of IFA during pregnancy as a risk factor for preterm birth differs from conventional knowledge and previous study reports. Traditionally, gestational anemia has been prevented with the provision of daily iron supplementation throughout pregnancy [
33]. Results from a recent systematic review [
34] which included 48 randomized trials and 44 cohort studies, revealed significant effect of prenatal iron consumption on reducing risk of low birth weight (RR: 0.81; 95% CI: 0.71, 0.93) but non-significant effect on preterm births (RR: 0.84; 95% CI: 0.68, 1.03).
However, many studies also fail to show beneficial effects of antenatal iron supplementation on pregnancy outcomes. Cochrane review and meta-analyses [
35‐
37] concluded that iron supplementation during pregnancy was neither beneficial nor harmful in terms of preterm birth [
38]. There is evidence to suggest that increasing iron intake is not always beneficial [
39]; iron availability influences the severity and chronicity of maternal infections and thus might lead to negative pregnancy outcome, including preterm birth [
40]. Because iron allays the fall in hemoglobin during pregnancy, iron-induced macrocytosis could increase blood viscosity to a degree that would impair utero-placental blood flow, decrease placental perfusion and increase risk of placental infarction [
41,
42]. This patho-physiological pathway may partially explain the results found in our study.
Education is the dimension of socioeconomic status that most strongly and consistently predicts health status [
43,
44]. A low level of education limits a person's access to employment and other social resources, which in turn limits his/her capacity to integrate within society and thereby increases the risk of subsequent poverty [
43‐
45]. Similar to ours, studies in India and Brazil [
26,
46] also reported maternal education below primary level as a risk factor for preterm birth (RR: 1.4; 95% CI: 1.1, 1.8). We found that male sex was associated with preterm birth relative to girls, which is consistent with previous studies in which 55% of all preterm births are boys [
47,
48].
Studies from several developing countries have found that “no ANC visit” is a significant risk factor for preterm birth, ranging from 1.3 times to 7 times higher than for women having any ANC visit [
26,
28,
46,
49,
50]. Visits to ANC centers and/or receiving ANC may raise awareness of the need for skilled delivery care [
51] or give women and their families familiarity with the health services available at health centers or the skills of the service providers, thus enabling them to navigate and receive necessary care when a crises arises [
52].
The main strengths of this study are that we analyzed prospectively collected, population-based data from a large sample (n = 32,126) of live births. In addition, since we collected data through visiting study women at home, the common concerns about selection bias in hospital-based studies from developing countries were avoided.
A limitation of our study was reliance on LMP to determine gestational age. Some of the common criticisms of this method are possible inaccurate recall of LMP, including heaping on certain dates, and reliance of the calculations on a “normal” menstrual cycle of 28 days with ovulation on day 14 [
53‐
55]. However, mounting evidence shows that LMP dates can be utilized to estimate gestational age accurately and reliably [
56] with a precision of 86%-90% [
57]. In a multi-country trial in developing countries that enrolled 799 women from China, Cuba, and India, 99.9% of women were able to provide an exact date for their LMP, and 92.4% of LMP dates were within 1 week of physician clinical assessments [
58]. In a study of 355 preterm neonates born in Dhaka Shishu Hospital, LMP estimates of gestational age were on average only 1 day lower than first or second trimester ultrasound determined gestational age (+/- 11 days) [
57]. Compared to ultrasound, use of LMP may over or underestimate preterm delivery depending on characteristics of the sample, timing of ultrasound, and LMP recall period [
53,
57,
59,
60]. Such misclassification of gestational age estimation using the LMP method, along with differential misclassification across the risk factors of interest, could lead to over or underestimation of the population level burden of preterm birth.
Since preterm birth is the leading cause of neonatal deaths globally [
61], and the second leading cause of deaths in children under five years of age, progress towards achieving MDG4 for child survival requires achieving higher coverage of evidence-based interventions to prevent preterm birth and/or to improve survival for preterm newborns [
62]. Global experts announced a “Goal for reduction of preterm birth rate by 2025” on World Prematurity Day (Nov 11) in 2012 [
3]. For countries, like Bangladesh, with a neonatal mortality rate above 5 per 1000 live births in 2010, “the goal is to reduce their preterm birth-attributable mortality by 50% between 2010 and 2025” [
3]. Thus, it is important to ensure effective planning and design of community-based programs focusing on preterm births, specifically in low resource settings. Such a focus will require a clearer understanding of associated risk factors, especially those which can be intervened upon.
Acknowledgements
The study was conducted by the ProjAHNMo study group in Bangladesh. ProjAHNMo is a partnership of the ICDDR,B; the Bangladesh government’s Ministry of Health and Family Welfare; Bangladeshi nongovernmental organizations, including Shimantik, Save the Children-USA, Dhaka Shishu Hospital and the Institute of Child and Mother Health; and the Johns Hopkins Bloomberg School of Public Health. We thank the members of the ProjAHNMo study team and colleagues at the Bangladesh Ministry of Health and Family Welfare at the sub-district, district and central levels for their valuable help and advice. We thank the many individuals in Sylhet district who gave their time generously as well as ProjAHNMo field and data management staff who worked tirelessly.
Funding for the ProjAHNMo was provided by the United States Agency for International Development, Office of Health, Infectious Diseases, and Nutrition, Global Health Bureau and the Dhaka Mission through the Global Research Activity Cooperative Agreement (GHS-A-00-03-00019-00), and the Saving Newborn Lives initiative of Save the Children Federation – USA through a grant from the Bill and Melinda Gates Foundation.
Rashed Shah was supported through Fogarty Training Grant (Grant # D43TW7587) during the period of data analyses and manuscript writing.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
RS, AHB and LCM were primarily responsible for conceptualizing and designing this study. LCM, AHB, SEA and GLD were responsible for protocol development and study design of the main study (Chlorhexidine trial). AHB and SEA were the principal investigators of the ProjAHNMo Chlorhexidine study. RS, IM, SMR, GLD, LCM were co-investigators of the main study. RS conducted data analyses and drafted the manuscript. RRT, JAA, DM and NB contributed in literature review, interpretation of results and manuscript editing. DM and NB assisted in data management and analyses. All authors provided critical intellectual input in editing and revising the manuscript and approved the manuscript for submission.