Risks of stillbirth and neonatal death with advancing gestation at term: A systematic review and meta-analysis of cohort studies of 15 million pregnancies

Background Despite advances in healthcare, stillbirth rates remain relatively unchanged. We conducted a systematic review to quantify the risks of stillbirth and neonatal death at term (from 37 weeks gestation) according to gestational age. Methods and findings We searched the major electronic databases Medline, Embase, and Google Scholar (January 1990–October 2018) without language restrictions. We included cohort studies on term pregnancies that provided estimates of stillbirths or neonatal deaths by gestation week. We estimated the additional weekly risk of stillbirth in term pregnancies that continued versus delivered at various gestational ages. We compared week-specific neonatal mortality rates by gestational age at delivery. We used mixed-effects logistic regression models with random intercepts, and computed risk ratios (RRs), odds ratios (ORs), and 95% confidence intervals (CIs). Thirteen studies (15 million pregnancies, 17,830 stillbirths) were included. All studies were from high-income countries. Four studies provided the risks of stillbirth in mothers of White and Black race, 2 in mothers of White and Asian race, 5 in mothers of White race only, and 2 in mothers of Black race only. The prospective risk of stillbirth increased with gestational age from 0.11 per 1,000 pregnancies at 37 weeks (95% CI 0.07 to 0.15) to 3.18 per 1,000 at 42 weeks (95% CI 1.84 to 4.35). Neonatal mortality increased when pregnancies continued beyond 41 weeks; the risk increased significantly for deliveries at 42 versus 41 weeks gestation (RR 1.87, 95% CI 1.07 to 2.86, p = 0.012). One additional stillbirth occurred for every 1,449 (95% CI 1,237 to 1,747) pregnancies that advanced from 40 to 41 weeks. Limitations include variations in the definition of low-risk pregnancy, the wide time span of the studies, the use of registry-based data, and potential confounders affecting the outcome. Conclusions Our findings suggest there is a significant additional risk of stillbirth, with no corresponding reduction in neonatal mortality, when term pregnancies continue to 41 weeks compared to delivery at 40 weeks. Systematic review registration PROSPERO CRD42015013785

pregnancies continued beyond 41 weeks; the risk increased significantly for deliveries at 42 versus 41 weeks gestation (RR 1.87, 95% CI 1.07 to 2.86, p = 0.012). One additional stillbirth occurred for every 1,449 (95% CI 1,237 to 1,747) pregnancies that advanced from 40 to 41 weeks. Limitations include variations in the definition of low-risk pregnancy, the wide time span of the studies, the use of registry-based data, and potential confounders affecting the outcome.

Conclusions
Our findings suggest there is a significant additional risk of stillbirth, with no corresponding reduction in neonatal mortality, when term pregnancies continue to 41 weeks compared to delivery at 40 weeks.

PROSPERO CRD42015013785
Author summary Why was this study done?
• A third of stillborn babies in the UK are born at term (>37 weeks) and were previously considered to be healthy.
• Prolongation of pregnancy at term is a known risk factor for stillbirth. Currently women are routinely offered induction of labour after 41 weeks gestation to avoid stillbirth. But 1 in 3 women have a stillborn baby prior to this gestational age.
• Mothers need robust estimates of potential risks of stillbirth and newborn death at term to make decisions on timing of delivery.

What did the researchers do and find?
• We collated data from all relevant studies found in a systematic review and determined the additional risks of stillbirth and newborn death in mothers at term gestation, by comparing pregnancies that continued versus delivered at various gestational ages,.
• We found a steady increase in the risk of stillbirth with advancing gestation at term. In mothers who continued their pregnancy to 41 weeks, there was a 64% increase in the risk of stillbirth compared to those who delivered at 40 weeks, with 1 additional mother having a stillborn baby for every 1,449 women.
• The risks of newborn death remained constant between 38 and 41 weeks, and only increased beyond 41 weeks.

Introduction
Despite advances in antenatal and intrapartum care, stillbirth continues to be a major burden [1]. More than 3,000 babies are stillborn every year in the UK-with a third of them considered to be apparently healthy term infants (37 weeks gestation or beyond) [2]. Stillbirth at term in an otherwise low-risk pregnancy [3,4] devastates parents with its unexpectedness. The UK's recent Maternity Safety Strategy initiative aims to halve the stillbirth rate by 2025 [5]. Such an effort requires an understanding of the magnitude of the problem through collation of large datasets, particularly for key factors like gestational age, for which existing information is imprecise [6]. Prolonged pregnancy is a known risk factor for stillbirth. To avoid this adverse outcome, women are routinely offered induction of labour after 41 weeks gestation [6][7][8]. This recommendation is based on evidence of increased stillbirth risk beyond 41 weeks [9]. However, 1 in 3 stillbirths occur prior to 41 weeks gestation [2][3][4]. The stillbirth risks before 41 weeks are not routinely discussed with women who have no clinical indication for delivery. This is in part because of how 'term pregnancy' is defined as normal in standard texts [10], and in part because of concerns about adverse neonatal outcomes that may occur from delivery before 41 weeks [11].
Individual studies on the risk of stillbirth in what is considered as normal term gestation vary in the magnitude and consistency of findings by gestational week [12][13][14]. Corresponding neonatal mortality estimates are imprecise [13,15,16]. We undertook a systematic review to evaluate the additional weekly risks of stillbirth in term pregnancies that continue versus deliver at various gestational ages. We also assessed the week-specific risks of neonatal death by gestational age at birth.

Methods
We undertook the review using a prospectively registered protocol (PROSPERO CRD42015013785) and report our findings in line with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) recommendations [17]. Ethics approval was not needed.

Literature search and study identification
We searched Medline, Embase, and Google Scholar from January 1990 to March 2017 for studies reporting rates of stillbirth and/or neonatal death at various gestational ages in apparently low-risk term pregnancies, and updated the search again to October 2018. We used the following search terms for the population: 'term pregnancy', 'prolonged pregnancy', 'post term', and 'postdates'; these were combined with terms relevant to the outcomes such as 'stillbirth', 'intrauterine death', 'fetal death', 'perinatal death', and 'perinatal mortality'. We undertook a separate search for studies reporting only neonatal death using the terms 'neonatal mortality', 'newborn death', and 'neonatal death' and combined these with 'term pregnancy', 'singleton', and 'low-risk pregnancy' (S1 Appendix). There were no language restrictions. We manually searched the reference lists of relevant studies for more relevant data, and contacted the authors and researchers in the field for additional studies or relevant information where required.

Study selection
We selected the studies in a 2-stage process. First 2 independent reviewers (JM and HR) screened the titles and abstracts to identify eligible studies, and then they retrieved the relevant full texts for detailed assessment. Any disagreements on the eligibility of the studies were resolved with a third reviewer (ST). We included cohort studies (including those nested within randomised trials) on pregnant women at term gestation without a prespecified indication for early delivery, if they provided weekly estimates of stillbirths. We excluded abstracts, letters, case reports, case series, and animal studies, and studies that only included women with preexisting medical conditions, congenital fetal malformations, complications such as preeclampsia, gestational diabetes, or small-for-gestational-age fetuses, or women who needed planned delivery before 37 weeks for maternal or fetal reasons. We defined term pregnancies as pregnancies with a gestational age of 37 completed weeks or beyond [18]. Stillbirth was defined as the death of a baby before birth, which included both antenatal and intrapartum deaths [19]. Any newborn death before 28 days of age was classed as a neonatal death [20]. We defined a low-risk pregnancy as that in which a healthy woman with apparently uncomplicated pregnancy enters labour with a low risk of developing intrapartum complications [21].

Quality assessment and data extraction
Two independent reviewers (JM and HR) assessed the quality of the individual studies, both for internal (risk of bias) and external (the representativeness of the population) validity [22]. For internal validity, we studied the individual features of the study such as the design, method of sampling, ascertainment of the outcome, appropriate determination of gestational age, and adequacy of follow-up [23]. We considered studies with a prospective design, random or consecutive sampling, use of first-trimester ultrasound to determine gestational age [24], and follow-up rates of over 80% to have a low risk of bias. For external validity, we considered a population to be clearly defined as representative of low-risk pregnancy if it met the following criteria: a clear definition of low-risk pregnancy, exclusion of pregnancies with congenital fetal malformations, and exclusion of multiple pregnancies. Any discrepancies were resolved after discussion with a third reviewer (ST). Data were extracted in duplicate by 2 reviewers (JM and HR). We extracted the number of ongoing pregnancies, number of deliveries, and number of events (stillbirths or neonatal deaths) per week.

Analysis
In the first step, for each study we calculated the gestation-week-specific prospective risk of stillbirth from the number of stillbirths that occurred in that week divided by the number of pregnancies that were considered to be at risk. The 'at risk' pregnancies were determined from the number of women who were still pregnant at the beginning of the week minus half the number who delivered that week [23]. We obtained pooled week-specific risks by using a multilevel (studies and women) mixed-effects logistic regression model without covariates and with random intercepts [25]. The overall week-specific rates of neonatal death were calculated using the same model: The number of neonatal deaths that occurred in a particular week were divided by the number of deliveries in that period.
In the next step, we compared the change in overall week-specific risk of the event (either stillbirth or neonatal death) between 2 consecutive weeks by calculating the risk ratio (RR). We calculated the RR by dividing week-specific risks that were obtained after fitting the corresponding logistic models. Non-parametric 95% confidence intervals (CIs) were obtained after fitting the logistic models on each of the 1,000 bootstrap samples (not stratified by study). After calculating the distribution of RRs for all gestational ages, we chose the 2.5th and 97.5th percentiles to represent the non-parametric limits of the 95% CI. For each gestational week, we also estimated the number of pregnancies at risk, i.e., the number of pregnancies that if continued to the next week will experience 1 additional stillbirth, compared to delivery at that gestational week [26,27].
We planned subgroup and sensitivity analyses a priori to determine whether the risks of stillbirth and neonatal death at term varied according to maternal characteristics such as race (White, Black, Asian, Other), body mass index (normal, overweight, obese), and age; study characteristics such as quality (risk of bias), country income status (low, middle, high), and time period; or restriction of assessments to those studies that excluded fetuses with congenital malformations and studies that used a strict definition (criterion) of low-risk pregnancy. For subgroup analysis, we compared the week-specific risks of stillbirth in women of Black versus White race by including race as a covariate in the logistic model, with White race as reference. We reported the estimates as odds ratios (ORs) with 95% CIs for various gestational ages.
We estimated the heterogeneity using the tau-squared statistic, with a value of 0 indicating no between-study variance. Publication bias and small study effect were assessed with funnel plots representing weekly event rate (logit scale) versus its standard error. Begg's and Egger's tests were used to determine funnel asymmetry [28,29]. All analyses were carried out in Stata version 13.1.

Risk of stillbirth at term
The risk of stillbirth at term in the studies varied from 1.1 [34] to 3.2 [12] per 1,000 pregnancies. The overall gestation-week-specific prospective risk of stillbirth steadily increased with gestational age, from 0.11 per 1,000 pregnancies at 37 weeks (95% CI 0.07 to 0.15) to 3.18 per 1,000 at 42 weeks gestation (95% CI 1.84 to 4.35) (Fig 3).  The stillbirth risk increased by 64% (RR 1.64, 95% CI 1.51 to 1.77, p < 0.001) when pregnancies are continued to 41 weeks-as currently recommended-compared to delivery at 40 weeks. One additional stillbirth occurred for every 1,449 women (95% CI 1,237 to 1,747) who continued the pregnancy from 40 to 41 weeks (Table 2). S2 Appendix provides individual study estimates on week-specific risks of stillbirth for 40 weeks and 41 weeks.
Our sensitivity analyses restricted to studies with a strict definition of low-risk pregnancy (Table 3), pregnancies without congenital fetal malformations (Table 3), last participant recruitment after 1990 (S3 Appendix), and a low risk of bias (S4 Appendix) showed a consistent increase in risk of stillbirth at each gestational week after 37 weeks. Subgroup analyses by race showed that compared to White women, Black women at term were 1.5 to 2 times more likely to have a stillbirth at all gestational ages (S5 Appendix) [12,32,34]. The week-specific prospective risks of stillbirth are provided separately for Black and White women in S6 Appendix. There were no statistically significant differences in the odds of stillbirth at any gestational age between Asian and White women, except for a lower risk at 42 weeks in mothers of Asian race (RR 0.49, 95% CI 0.29 to 0.83, p = 0.008) (S7 Appendix) [34].
There were insufficient data to undertake other planned subgroup analyses on maternal body mass index, age, and country income status. We did not observe evidence of small study effect for stillbirth (Begg's test Kendall's score p > 0.05; Egger's test p > 0.05) (S10 Appendix).

Risk of neonatal mortality at term
The risk of neonatal death was unchanged for births between 38 and 41 weeks of gestation; the risk increased beyond 41 weeks (RR 1.87, 95% CI 1.07 to 2.86, p = 0.012). Table 2 provides the estimates of week-specific risk of neonatal death for births at various gestational ages at term. Sensitivity analysis performed by only including studies on singleton pregnancies uncomplicated by congenital fetal malformations (S8 Appendix), and only high-quality studies (S9 Appendix), showed a similar pattern, with increased risks observed for births beyond 42 weeks compared to the previous week.

Discussion
We found that the prospective risk of stillbirth increased with gestational age in pregnancies at term; neonatal mortality risk remained unchanged until 41 weeks, but increased beyond this gestation. Pregnancies that continued to 41 weeks-currently still considered normal term gestation-had a small but significant increase in the risk of stillbirth compared to those delivered at 40 weeks, with no differences in neonatal mortality. To our knowledge, ours is the largest review to date on risks of stillbirth and neonatal death at various gestational ages in term pregnancies. The review was based on a prospective protocol with predefined inclusion criteria. We registered the review protocol with PROSPERO prior to completion of the detailed search and data extraction. We contacted the individual authors for relevant data when it was required for the analysis and where it was possible. When we included unpublished data, the relevant researchers were not involved in the data extraction, quality assessment, or analysis to minimise bias. The large sample size achieved with these efforts allowed us to generate results with high precision. By reporting both relative and absolute increases in the risks, our findings provide the appropriate context for interpretation. We assessed the qualities of the included studies and the validity of the evidence. Our sensitivity analyses demonstrated that our findings were not sensitive to the assumptions made. Unlike previous studies in this area, our robust analytical approach [38,39] avoided the inappropriate use of Kaplan-Meier method [40,41]. Table 3. Risks of stillbirth in pregnancies that continue to the next week versus deliver in studies with a strict definition of low-risk pregnancy and those without congenital fetal malformations. Low-risk pregnancy defined as singleton pregnancies, absence of congenital fetal malformations, and absence of any medical conditions in the mother. � Between 2 consecutive weeks.

Gestational age (weeks) Number of studies Number of stillbirths Number of pregnancies Risk ratio
The inclusion criteria varied between studies. But all studies included women whose pregnancy continued to term and beyond, an indication of their low-risk status in that early delivery was not required [42,43]. Some of these apparently 'low risk' pregnancies may also have had undetected fetal growth restriction. But continuation of such pregnancies to term is in line with current practice, where there is no routine ultrasound monitoring of fetal growth [43].
Some of the included cohorts extended before 1990, and the risk of outcomes might have changed over time. We consider the effect of study time span on our stillbirth estimates to be minimal for the following reasons. First, current definitions of 'term' and 'post-term' pregnancies have remained unchanged over the decades, with very little change in the criteria that label pregnancies as high or low risk [44,45]. Second, evidence behind the current recommendations on the timing of delivery in term pregnancies with no obvious reasons for early delivery stem from both recent and past data [6,46] Third, in apparently low-risk pregnancies at term, the standard of antenatal care, including regular blood pressure checks and auscultation of fetal heart, has not changed over time [44]. Lastly, we observed very little between-study variance in the reported risks of stillbirth for various gestational ages, irrespective of the year of data collection or inclusion criteria.
Although the database registries are prone to biases, we expect the outcome of death to be well recorded [47]. It is possible that an intrauterine death recorded as being in a particular gestational week may have occurred in the previous week. But in women who undergo weekly monitoring of fetal heart rate at term gestation, the time interval from fetal demise to birth is considered about 2 days on average [48].
Regarding maternal characteristics, we only analysed the risk of stillbirth at various gestational ages by race. Due to the paucity of the published information, we were unable to explore in detail if there were variations in risks by socioeconomic status, maternal age, and parity [49]. Our approach is similar to that of previous studies that studied the 'real life' risk of stillbirth with advancing gestation at term irrespective of the presence or absence of risk factors [30]. The higher risk of stillbirth at all gestational ages in Black women compared to White women could be attributed to upstream determinants such as low educational and socioeconomic status, reduced access to antenatal care, and increased rates of fetal growth restriction [50][51][52].
We did not observe significant changes in neonatal mortality for births between 38 and 41 weeks gestation, a finding that was consistent with previous studies [53,54]. The developmental outcomes for children born at term but before 40 weeks gestation are not known to be significantly different from those for children born at 40 weeks [55]. In a recently published randomised trial (ARRIVE) on induction versus expectant management in low-risk nulliparous women, the reduction in the composite of perinatal death or serious neonatal complication by 20% with induction fell just short of statistical significance (95% CI 0.64 to 1.00) [56]; our findings also support the evidence of benefit for delivery versus expectant management. Trials and systematic reviews have assessed the effects of induction of labour at prespecified gestational cutoffs such as 39 weeks [57,58]. This approach limits the information on risks of stillbirth at various gestational time points, and the effects of intervention.
Any discussion with women considering prolonging their pregnancy beyond 41 weeks gestation should be include information on the absolute risk increase, and the effects of induction of labour on mode of delivery and perinatal outcomes [58]. There is a need to assess the acceptability of early delivery at term to parents and healthcare providers to avoid the small risk of stillbirth. Better stratification of apparently low-risk pregnancies for complications using individualised prediction models could reduce the number of women who need to be delivered to avoid 1 additional stillbirth. Decision analytic modelling with economic evaluation is required to assess the cost-effectiveness of offering delivery at various gestational ages at term. While our comprehensive systematic review provided robust quantitative estimates of the risks of stillbirth and neonatal death at various gestational ages in term pregnancies, the findings were limited by the heterogeneity in the definition of low-risk pregnancies (which might have included women with undiagnosed fetal growth restriction), loss of data due to exclusion of studies that did not provide stillbirth estimates in weekly intervals, and the inability to adjust for confounding variables.
In conclusion, there is a significant increase in the risk of stillbirth, without a corresponding reduction in the risk of neonatal death, in mothers at term when pregnancies continue to the current recommended gestation of 41 weeks compared to delivering in the previous week.