Planned delivery or expectant management in preeclampsia: an individual participant data meta-analysis

We analyzed all the available data for each prespeciﬁed outcome on an intention-to-treat basis. For primary individual patient data analyses, we used a 1-stage ﬁxed effects model. RESULTS: We included 1790 participants from 6 trials in our analysis. Planned delivery from 34 weeks gestation onward signiﬁcantly reduced the risk of maternal morbidity (2.6% vs 4.4%; adjusted risk ratio, 0.59; 95% conﬁdence interval, 0.36 e 0.98) compared with expectant management. The primary composite perinatal outcome was increased by planned delivery (20.9% vs 17.1%; adjusted risk ratio, 1.22; 95% conﬁdence interval, 1.01 e 1.47), driven by short-term neonatal respiratory morbidity. However, infants in the expectant management group were more likely to be born small for gestational age (7.8% vs 10.6%; risk ratio, 0.74; 95% conﬁdence interval, 0.55 e 0.99). CONCLUSION: Planned early deliveryinwomenwithlate pretermpreeclampsia provides clear maternal beneﬁts andmay reducetheriskofthe infant being born small for gestational age, with a possible increase in short-term neonatal respiratory morbidity. The potential beneﬁts and risks of prolonging a pregnancy complicated by preeclampsia should be discussed with women as part of a shared decision-making process.


Introduction
Pregnancy hypertension is responsible for at least 27,800 maternal deaths 1 worldwide every year and 500,000 infant deaths, 2 including approximately 200,000 stillbirths. 3 Although the prevalence of preeclampsia varies throughout the world, it complicates between 2% and 3% of pregnancies in a high-income setting. 4 Estimates for low-and middleincome countries are higher, with up to 12% of pregnancies affected in these settings. 2 Delivery is the only definitive management for this progressive and unpredictable condition, and it is routinely recommended for all women with preeclampsia from 37 weeks gestation onward. 5 At gestations up to 34 weeks, if there are no immediate indications for delivery, expectant management is preferable because of the neonatal risks associated with early preterm birth. 5 It is less clear whether a policy of expectant management in the late preterm period (34e37 weeks) should be pursued, although if severe features of preeclampsia develop or the woman reaches 37 weeks, delivery is indicated. However, there is uncertainty as to whether a policy of routine immediate delivery at this gestational window (34e37 weeks) could reduce maternal complications without serious neonatal consequences. Several studies have compared these 2 strategies in women with hypertensive disorders of pregnancy (including preeclampsia) from 34 weeks. 6e12 However, it has not been possible to draw firm conclusions from individual studies alone. Recent metaanalyses 13,14 and individual participant data (IPD) meta-analyses 15 of women with hypertensive disorders of pregnancy have shown that planned early delivery from 34 weeks gestation reduces maternal complications, but the neonatal impact remains unclear. These reviews generally grouped all hypertensive disorders of pregnancy together, combining women with chronic hypertension, gestational hypertension, and preeclampsia. However, the underlying pathophysiology of preeclampsia is distinct, with maternal endothelial dysfunction leading to multiorgan complications and potentially severe maternal and fetal outcomes. The optimal timing of delivery in preeclampsia may therefore differ compared with other hypertensive disorders of pregnancy, and the balance of risks and benefits for the infant should also be considered within the context of this rapidly progressive and unpredictable disease. A limited subgroup analysis conducted as part of the previous IPD meta-analysis 15 in women with all types of pregnancy hypertension identified women with preeclampsia as a population in whom planned delivery may confer significant benefit. The authors therefore highlighted a need to evaluate the impact of this intervention specifically in women with preeclampsia. Since this meta-analysis was published, a new trial has been reported, 6 enrolling more women with preeclampsia than all previously included trials combined. This enabled us to conduct an IPD metaanalysis evaluating the timing of delivery on a wider set of maternal and perinatal outcomes in this high-risk group of women with preeclampsia. A meta-analysis evaluating early delivery or expectant management for late preterm preeclampsia was recently published. 16 However, this study was limited by its inclusion of just 3 randomized controlled trials, only 2 of which were used to evaluate the coprimary outcome of neonatal intensive care unit admission. Our IPD meta-analysis is strengthened by its ability to harmonize data to overcome inconsistencies in outcome definitions between trials and to evaluate key outcomes such as neonatal morbidity, in more detail.

Objective
The objective of this study was to undertake an IPD meta-analysis focusing on women with preeclampsia alone. In women with preeclampsia from 34 weeks gestation onward, this study aimed to evaluate the effect of planned early delivery on maternal mortality or morbidity and perinatal mortality or morbidity compared with expectant management using IPD from randomized controlled trials. The use of IPD enabled us to target our review to women with late preterm preeclampsia and to perform subgroup analyses and adjustments that would not be possible with the use of aggregate data, for example, using blood pressure values to reflect the severity of disease. This is clinically relevant, because the presence of additional risk factors in women with preeclampsia may alter management options.

Methods
Search strategy and study selection We followed a protocol and statistical analysis plan published in the PROS-PERO registry in accordance with PRISMA-IPD guidance. 17 We included studies that were randomized controlled trials comparing planned early delivery with expectant management in women presenting with preeclampsia from 34 weeks gestation onward. Cluster randomized trials or studies with a quasirandomized design were excluded. To identify the eligible studies, we electronically searched the Cochrane Central Register of Controlled Trials (CEN-TRAL), PubMed, MEDLINE, and ClinicalTrials.gov using the search terms "pre-eclampsia" OR "preeclampsia" AND "delivery" OR "birth" with the limits "human" and "randomized controlled trial." The final search date was December 18, 2021. We did not restrict our search by language. We excluded trials published before the year 2000. This was because of changes in clinical practice, care of women with preeclampsia, and neonatal care over time such that the findings from earlier trials may be difficult to interpret. To ensure that the search was comprehensive, we also hand-searched the reference lists of the retrieved studies and any relevant reviews identified. Two independent review authors (A.B.G. and J.F.) assessed all the studies identified by the search strategy against the study-level inclusion criteria. Any disagreement was resolved through discussion or with a third review author (not required), if necessary.

Eligibility criteria
We included women with singleton or multifetal pregnancies presenting with preeclampsia or superimposed preeclampsia from 34 weeks gestation onward. The definition of preeclampsia or superimposed preeclampsia was that used by the study at the time. All the definitions used would now be encompassed by the current International Society for the Study of Hypertension in Pregnancy (ISSHP) 2018 diagnostic criteria. 18

Data extraction
We sought participant-level data from the authors of all eligible trials. The available data were extracted from trial databases (provided via a data-sharing agreement) according to prespecified variables by 2 of the review authors (A.B.G. and P.S.). The data were recoded into a common format, and the definitions of key characteristics, diagnoses (eg, preeclampsia), and outcomes were harmonized. A final dataset was then produced and rechecked for accuracy and completeness.

Assessment of risk of bias
Two review authors (A.B.G. and J.F.) independently assessed the included trials for risk of bias using the Cochrane risk-of-bias tool. 19 Outcomes The primary maternal outcome was a composite of maternal mortality and severe maternal morbidity (adapted from a previously published composite derived by Delphi consensus). 20 The presence of severe maternal morbidity was defined as 1 or more of the following individual components: maternal death, eclampsia, stroke, pulmonary edema, HELLP (hemolysis, elevated liver enzymes, low platelets) syndrome, acute renal insufficiency, and placental abruption. The primary perinatal outcome was a composite of perinatal mortality or morbidity. This was defined as any 1 of perinatal death, neonatal death, or neonatal morbidity. The selection of components was guided by recent recommendations for core outcome sets in preeclampsia. 21 Neonatal morbidity was defined as 1 or more of respiratory disease (any one of respiratory distress syndrome, need for respiratory support, neonatal unit admission for respiratory disease or bronchopulmonary dysplasia), central nervous system complications (any 1 of intraventricular hemorrhage, intracerebral hemorrhage, periventricular leukomalacia, hypoxic ischemic encephalopathy, cerebral infarction, or convulsions), culture-proven sepsis, necrotizing enterocolitis, hypoglycemia requiring intravenous glucose or neonatal unit admission, or jaundice requiring neonatal unit admission. If data were missing (ie, not collected for a particular component) for either of the composite outcomes, we treated it as absent. The secondary maternal outcomes included severe postpartum hemorrhage, progression to severe hypertension, thromboembolic disease, hepatic dysfunction, onset of delivery, and admission to maternal intensive care unit. The secondary perinatal outcomes were gestational age at delivery, mode of delivery, birthweight, birthweight centile, baby sex, small for gestational age (<3rd centile or <10th centile), admission to neonatal unit, admission to neonatal intensive care unit, 5-minute Apgar score <7, and arterial pH <7.05.

Data synthesis
We analyzed all available data for baseline maternal characteristics at enrolment, related process outcomes (such as time from randomization to delivery) and the data for each prespecified outcome on an intention-to-treat basis. In each study, all the outcomes of interest were either reported completely with <5% missingness or not reported at all. Under these circumstances, multiple imputation is not feasible or recommended, and we therefore analyzed all the outcomes without imputation. For primary IPD meta-analyses, we used a 1stage fixed-effect model. Standard errors, confidence intervals (CIs), and P values were adjusted for clustering within studies. In addition, we used robust standard errors to correct for clustering of twin pregnancies by the mother for the perinatal outcomes. 22 We set out to calculate the odds ratios using multilevel models as originally outlined in the statistical analysis plan. However, this multilevel model structure did not converge, as there were not sufficient datapoints at each of the levels. We therefore performed a multivariate Systematic Reviews ajog.org analysis, calculating risk ratios for binary outcomes and mean differences for continuous outcomes using a simpler fixed-effects model. We also calculated unadjusted risk differences. A fixedeffects, 1-stage analysis such as this is appropriate where there are small studies with rare event numbers. We gave a separate intercept for each trial but assumed the same treatment effect (ie, we used fixed effects for each trial).
The numbers needed to treat or harm with 95% CIs were calculated for outcomes where a significant difference between the management groups was found. The analysis was adjusted for study, gestational age at randomization (34 þ0 e34 þ6 weeks, 35 þ0 e35 þ6 weeks, 36 þ0 e36 þ6 weeks, 37 þ0 e37 þ6 weeks, 38 þ0 e38 þ6 weeks, 39 þ0 e39 þ6 weeks, 40 þ0 weeks and above), severity of systolic hypertension at study entry (<150 vs ! 150 mm Hg), parity (primiparous vs multiparous), and number of fetuses (singleton vs all other). The severity of systolic hypertension at study entry was chosen, because it is an objective marker of disease severity consistently available across studies, and there is a known doseeresponse relationship between increasing blood pressure and adverse pregnancy outcomes. 23e25 We calculated and used the average value (or proportion for categorical variables) across all studies, where these prespecified adjustment variables were missing. We did not use multiple imputation methods, as they are not recommended in this scenario. Subgroup analysis was conducted if there were at least 10 events in each subgroup; this was also done using a 1stage, fixed-effects model. The prespecified subgroups were study, gestational age at randomization, parity, singleton vs multifetal pregnancy, previous cesarean delivery, prerandomization diabetes of any type, superimposed preeclampsia, and suspected fetal growth restriction at enrolment. Because many of the subgroups concerned the same adjustment variables used for our main analysis (including some additional subgroups of clinical relevance), our subgroup analysis was unadjusted to better delineate the effect of these variables. Heterogeneity was assessed using I 2 (the proportion of the total variance of the outcome that is between studies rather than between subjects within studies) as part of the subgroup analysis. We have also presented values for tau. 2 No additional analyses were undertaken.

Study selection
We identified 1617 references after duplicates were removed ( Figure 1). A total of 1567 references were excluded after title screening, and 43 were excluded after abstract and full-text screening. Seven trials (3791 participants) were considered eligible for inclusion at study-level. One trial (100 participants) was subsequently excluded, as the trial authors did not respond to our request for participant-level data despite several attempts. 26 The only published data available from this trial were a conference abstract, and therefore we were not able to include any aggregate data for this trial. Six trials 6e11 with participant-level data were available. Following data extraction and review by 2 authors, 1901 participants were deemed ineligible for inclusion in this IPD meta-analysis principally because of women being enrolled with conditions other than preeclampsia or before 34 weeks gestation, with the reasons given for exclusion in Table 1. The remaining 1790 participants from 6 trials were therefore included in our analysis.

Study characteristics
A summary of characteristics of included studies, including details of the interventions, can be found in Table 1 and  Supplementary Tables S1 and S2. Two trials (GRIT and DIGITAT) enrolled ajog.org a Preeclampsia defined as a diastolic blood pressure of 90 mm Hg or higher measured on 2 occasions at least 6 hours apart, combined with proteinuria.
Beardmore-Gray. Timing of delivery in late preterm preeclampsia. Am J Obstet Gynecol 2022.

Systematic Reviews
ajog.org women with suspected fetal growth restriction on ultrasound, including those with pregnancies complicated by preeclampsia, over a wide gestational age range. The HYPITAT and HYPITAT II trials enrolled women with any hypertensive disorder of pregnancy from 36 þ0 and 34 þ0 weeks gestation onward, respectively. The PHOENIX trial and Deliver or Deliberate trial focused specifically on women with preeclampsia (without severe features) between 34 þ0 and 36 þ6 weeks gestation. None of the trials enrolled women with severe features of preeclampsia or any other indications for immediate delivery. This was stated in each of their inclusion criteria (Table 1), with severe features defined in accordance with the relevant guidelines at the time (primarily American College of Obstetricians and Gynecologists or ISSHP criteria). These are consistent with current definitions. 27 For the purposes of this IPD metaanalysis, we selected only those participants who met our eligibility criteria as described in the section above.

Risk of bias of included studies
The results of our risk of bias assessment using the Cochrane Risk of Bias 2 tool can be found in Figures 2 and 3 ajog.org was judged to be at a high risk of bias. This was primarily because of limited reporting regarding the randomization process and an imbalance in the final analysis population suggesting postrandomization exclusions. Supplementary Tables S3 and S4 describe the missing data for each maternal and perinatal variable by study.
Missing data were usually because of the outcome not being collected, with very few cases of missing data because of incomplete reporting or exclusion.

Synthesis of results
The baseline maternal characteristics at enrolment were similar across the planned delivery and expectant management groups (Table 2). Importantly, the proportion of women with suspected fetal growth restriction and severe hypertension at enrolment ( Table 2) was balanced between the 2 management groups as expected with randomization. None of the trials enrolled women with severe features of preeclampsia. However, we acknowledge that some participants may have transiently had high blood pressure readings before enrolment. This alone would not be an indication for delivery. 18 The difference in median time between the 2 groups from randomization to delivery was 4.0 (95% CI, 3.0e4.0) days. One-stage meta-analysis found that planned delivery from 34 weeks gestation onward significantly reduced the risk of major maternal morbidity (2.6% vs 4.4%; adjusted risk ratio [aRR], 0.59; 95% CI, 0.36e0.98; P¼.041) compared with expectant management (Table 3). This direction of effect was also consistent across the secondary maternal outcomes (  5) were consistent with the main results. Higher degrees of heterogeneity were seen when analyzed by study and by twin or singleton pregnancy. Subgroup analysis was only undertaken if there were 10 or more events in each subgroup, which meant that the overall effect by study was different to that reported for the overall IPD metaanalysis because of the exclusion of certain trials from the subgroup analysis. A summary of findings and the numbers need to treat and harm are presented in supplementary tables S9 and S10.

Comment
Principal findings In this IPD meta-analysis, we show that planned early delivery from 34 weeks gestation onward in women with preeclampsia significantly reduces adverse maternal outcomes and the number of infants born small for gestational age. This was balanced against an increase in the composite perinatal outcome driven by short-term neonatal respiratory morbidity; there was no significant Systematic Reviews ajog.org impact of gestational age on this primary outcome. These results indicate clinically important maternal benefits, and in particular, a reduction in severe hypertension and HELLP syndrome among women allocated to planned delivery. Importantly, the intervention did not increase the risk of cesarean delivery. Information on medical comorbidities was not consistently available across all studies. However, other than singleton or twin pregnancy subgroup analysis for the primary perinatal outcome, there was no significant test of interaction for any pre-enrolment characteristics such that we could not predefine a particular group of pregnant women in whom the impact of the intervention might be different. Most of the participants included in this analysis were classified as White European, which should be taken into account when considering the generalizability of these findings to other populations. The differences in the incidence of respiratory disease between the management groups was mainly seen among infants born to women from 2 trials, namely HYPITAT II 7 and Deliver or Deliberate, 8 conducted earlier in the time period considered for this metaanalysis. In HYPITAT II, only 8.6% of women randomized to planned delivery received antenatal corticosteroids. Steroid use was not reported in the Deliver or Deliberate trial, though planned delivery took place within 12 hours of randomization, leaving little time for optimal steroid administration. In comparison, 65% of the women in the PHOENIX trial 6 allocated to planned delivery received antenatal corticosteroids; this likely influences the much lower incidence of adverse respiratory outcomes among infants in this trial, with no difference between the 2 management groups. Although we acknowledge that our analysis was not specifically powered to address this question, it is likely that the difference in administration of steroids observed between different time epochs and trial settings explains our perinatal findings. This suggests that appropriately timed antenatal corticosteroid administration mitigates the short-term risk of respiratory complications for infants of women with preeclampsia, as previously demonstrated by a large systematic review. 28 Antenatal corticosteroids have also been shown to reduce infant intraventricular hemorrhage, 28 which is a rare  ajog.org Systematic Reviews outcome in infants at this late preterm gestation, providing further potential benefit in ameliorating the risk of central nervous system complications at this gestational age. Although some authors have raised concerns over the association between maternal antenatal corticosteroid treatment and childhood behavioral disorders in term-born children (on the basis of a population-based study 29 ), the most recent Cochrane systematic review of randomized controlled trials reported that antenatal corticosteroids probably lead to a reduction in developmental delay in childhood (RR, 0.51; 95% CI, 0.27e0.97). 28 The rates of other serious neonatal complications such as sepsis and necrotizing enterocolitis were low, as expected in this population. The relatively high rates of neonatal admission across both groups highlights the additional care that this high-risk population of infants may require irrespective of the timing of delivery. In addition, infants born to mothers in the expectant management group were significantly more likely to be born small for gestational age. As low birthweight is a risk factor for long-term neurodevelopmental delay 30,31 and has been shown to be a more important predictor of long-term infant outcomes than gestational age at delivery, 32 avoidance of ongoing growth restriction may influence management choices. Use of ultrasound to accurately evaluate gestational age and presence of growth restriction should therefore be an integral part of assessment of a woman with preeclampsia. Although the average difference between the 2 groups was 4 days, the third quartile was 10 days. It remains difficult to identify the women (and infants) who are most likely to require delivery within the following 7 days using clinical risk factor or biomarker prognostication, 33 but for a progressive and unpredictable condition such as preeclampsia, this degree of pregnancy prolongation could be associated with a biologically plausible and clinical relevant difference in fetal growth restriction and neonatal outcomes. An increased awareness that expectant management increases the risk of a small for gestational age infant, most likely by perpetuating growth restriction within an adverse intrauterine environment, may lower the threshold for considering planned delivery from 34 weeks onward. These findings raise interesting questions regarding the influence of expectant management on fetal growth restriction and the impact that this may have on the infant, which should be addressed by future research.

Comparison with existing literature
In the United States, current guidelines recommend planned early delivery in women with late preterm preeclampsia with severe features 34 but advise expectant management in women without severe features up to 37 weeks gestation. The guidelines acknowledge that this latter recommendation is based on limited and inconsistent evidence. 27-Current United Kingdom 35 and international 18 guidelines provide similar recommendations but again note the uncertainty in clinical practice around thresholds for intervention and the limited evidence base. Many reviews, including a recent Cochrane review, have therefore called for evidence focusing on optimal timing of delivery in different types of pregnancy hypertensive disease. Our findings confirm clear maternal benefits associated with planned early delivery in women with preeclampsia from 34 weeks gestation onward and provide a greater understanding of the perinatal benefits and risks, including factors (such as antenatal steroid use) that mitigate these. Our analysis extends the current evidence base and quantifies the benefiterisk balance specific to women with preeclampsia in the late preterm period. The important lack of increased risk in operative delivery is in keeping with other recent clinical studies comparing induction of labor with expectant management 36e38 ; women and clinicians may perceive similar rates of vaginal delivery in both groups as important to their decision-making. The perinatal results are consistent with interpretation by a systematic review evaluating planned early delivery for suspected fetal compromise that Systematic Reviews ajog.org highlighted an increased short-term risk of respiratory complications and neonatal unit admission. 39 However, the varying use of antenatal corticosteroids across the different trials included in our analysis should be considered when interpreting these results. Planned subgroup analysis showed that there was no difference in the primary perinatal outcome in the most recent trial, 6 where most of the women allocated to planned delivery received antenatal corticosteroids. Given that the universal administration of antenatal corticosteroids is not routinely recommended for women considered at risk of late preterm birth, 40 demonstrating benefit in certain clinical scenarios such as planned delivery for preeclampsia may guide clinical practice. Furthermore, we have demonstrated an increased risk of small for gestational age births associated with expectant management, a finding that is consistent with similar studies and is known to be associated with longer term impaired neurodevelopmental outcomes. 30,31 In addition, on the basis of the largest and most recent trial in this population, 6 clinicians and women should be aware that there is an average prolongation of pregnancy of around 3 days only with expectant management, with 74% progressing to severe preeclampsia (compared with 64% with planned delivery) and 55% requiring expedited delivery before 37 weeks gestation. The high proportion of women who were delivered early is in keeping with an expectant management strategy and highlights the rapidly progressive nature of preeclampsia often resulting in a constellation of maternal and fetal complications. Data from this IPD meta-analysis (which included the trial discussed above) supported this finding with a difference in median time from randomization to delivery of only 4 days between the 2 management groups. This study therefore strengthens the current evidence supporting a policy of considering planned early delivery for maternal benefit in late preterm preeclampsia. Planned delivery has been shown to be cost-saving in the UK National Health Service setting compared with expectant management (£1478 per woman) when the total maternal and infant costs were considered, but the decision-making should reflect clinical and health economic factors together.

Strengths and limitations
Following guidance on the use of IPD meta-analysis, 41 we did not adopt an overly restrictive approach when selecting trials for inclusion, and this study is therefore strengthened by the inclusion of several large, wellconducted randomized clinical trials, most of which were assessed as being at a low risk of bias. For most outcomes, heterogeneity between studies was low, though some important differences have been highlighted above. Furthermore, the use of a 1-stage IPD meta-analysis approach allows the relative influence of multiple trial and participant characteristics on any intervention effect to be considered simultaneously. 41 We had full access to the trial data and were able to include all the eligible participants for most of the studies. We were able to include complete data for most of our outcomes of interest but were limited by differences in outcome reporting between trials such that data were not available for every variable. This low missingness for most of the variables and broad consistency between trials means that we have confidence in our results. The limitations include changes in clinical practice during the time period of the trials included such that external factors (such as uptake of antenatal corticosteroid use) may impact the main outcomes directly. Certain perinatal outcomes such as bronchopulmonary dysplasia, cerebral infarction, and intracerebral hemorrhage were not collected across a large proportion of included studies likely because of the rarity of these outcomes and the availability of more objective ajog.org measures. Ideally, all trials should include longer term follow-up of the women and infants, but retention within a study can be challenging and expensive to undertake. We were not able to report the indications for delivery, as this information was not consistently available across the included trials. However, given the randomized nature of the data, we would not expect significant differences between the 2 management groups at baseline. The PHOENIX trial reported indications for delivery for both the management groups. In the planned delivery group, 99% of women had allocation to planned delivery arm as their recorded indication for delivery, consistent with trial procedures. Women in the expectant management group were delivered more frequently for both maternal and fetal indications, with over 50% requiring expedited delivery, compared with the planned delivery group.

Clinical implications
Delivery is already known to improve maternal outcomes in preeclampsia. However, this review quantifies the effect, specific to gestation, on outcomes and addresses the balance between maternal and fetal effects.
We also addressed the question specifically in women who have preeclampsia without severe features. By synthesizing and presenting the available data on this topic, we aim to provide as much information as possible on the balance of risks and benefits associated with each management strategy so that women and their caregivers can make fully informed decisions. For clinicians who already have a low threshold for planned delivery in women with late preterm preeclampsia, this meta-analysis provides new evidence that could support this approach. Other clinicians may consider that although maternal benefit of planned delivery is clear, there is a trade-off with short-term perinatal morbidity. However, this may be ameliorated by judicial use of antenatal corticosteroids.

Conclusions
This meta-analysis of IPD from 6 randomized controlled trials synthesizes the available evidence pertaining to timing of delivery in late preterm preeclampsia. We have clearly demonstrated that planned delivery in women with preeclampsia from 34 weeks onward provides maternal benefit with no increased risk of operative delivery compared with expectant management. Planned delivery reduces the likelihood of infants being born small for gestational age but increases shortterm respiratory morbidity. The administration of antenatal corticosteroids was observed to reduce this risk such that perinatal morbidity was no different between the groups in the most recent trial; the potential benefits of antenatal corticosteroids should be discussed with women undergoing late preterm delivery. Further research is needed to identify the optimal methods of determining the women and infants who are at the greatest risk of adverse outcomes, enabling the stratification of surveillance and targeted intervention. A similar need for accurate prognostic strategies has been identified for planning delivery in pregnancies with suspected fetal compromise 39 and preterm prelabor rupture of membranes 42 , as the challenges are common across these scenarios. Longer-term infant outcome data (including infants born with and without growth restriction) from large randomized controlled trials are also needed, as outcomes cannot be extrapolated from population-level databases comparing delivery at preterm gestations with term gestations in healthy pregnancies. There is also a need to establish the most clinically meaningful neonatal outcomes to measure when conducting preeclampsia trials, particularly those focused on timing of delivery. The impact of the intervention is likely to be very different in low-resource settings, where most of the maternal and perinatal disease Systematic Reviews ajog.org burden associated with preeclampsia lies. 43 Because antenatal stillbirth is much more common in these settings, 44,45 it is possible that early delivery in women with preeclampsia in low-and middle-income countries may reduce not just adverse maternal outcomes but fetal and perinatal deaths associated with severe maternal disease. However, this must also be balanced against the resource constraints in these environments. A multicenter randomized controlled trial evaluating this is currently underway 46 and may shed further light on this clinical dilemma in a different context. Our findings provide further information to guide women and clinicians in a highincome setting, who must consider the balance of benefits and risks associated with planned delivery for women and their infants with late preterm preeclampsia. In line with recent recommendations, 47 we recommend that clinicians discuss the trade-off with earlier delivery (better for maternal outcomes but with increased admissions to the neonatal unit) with women, fully supporting them to understand their options and consider both management strategies. Maternal: composite of maternal morbidity of fullPIERS 20 outcomes, with the addition of recorded systolic BP of at least 160mmHg post randomization, up to primary maternal hospital discharge Perinatal: composite of neonatal deaths within 7 days of delivery and perinatal deaths or neonatal unit admissions before infant primary hospital discharge Individual components of the composite primary outcome, use of antihypertensive drugs, progression to severe pre-eclampsia (systolic BP of at least 160mmHg, platelet count <100, abnormal liver function enzymes -ALT or AST >70), time and mode of onset, confirmed thromboembolic disease, confirmed sepsis, primary and additional indications for delivery; and placental abruption. Stillbirth, NND within 7 days of delivery, NND before hospital discharge, admissions to NNU, number of nights in each category of care, total number of nights in hospital, BW, BW centile, BW less than 10th or 3rd centile, GA at delivery, Apgar score at 5 min after birth, umbilical arterial and venous pH at birth, need for supplementary oxygen before discharge, number of days required, need for respiratory support, other indications and main diagnoses resulting in NNU admission and health resource use outcomes ALT, alanine aminotransferase; AST, aspartate transaminase; BW, birthweight; GA, gestational age; GMH, Germinal matrix hemorrhage; HELLP syndrome, Hemolysis, elevated liver enzymes, low platelet count syndrome; IVH, intraventricular hemorrhage; NICU, neonatal intensive care unit; NND, neonatal death; NNU, neonatal unit, PPH, post-partum hemorrhage; PVL, Periventricular leukomalacia; VM, ventriculomegaly.
Beardmore-Gray. Timing of delivery in late preterm preeclampsia. Am J Obstet Gynecol 2022.   ajog.org Systematic Reviews