Sonographic surveillance and perinatal outcomes among pregnancies with periviable fetal growth restriction

Introduction

Identification of fetuses with pathologically decreased in utero growth has remained a challenge in the field of prenatal diagnosis. Classifications have been proposed to aid in the differentiation between non-pathologically small fetuses (i.e., constitutionally-small fetuses) and fetuses with pathologically decreased growth (i.e., fetal growth restriction [FGR]) [1], 2].

Population-based studies have revealed that perinatal mortality significantly increases with EFW<50 % for gestational age (GA), or below +1 standard deviation (SD) from the mean, with the highest risk for perinatal mortality demonstrated <3 % for GA, or approximately −2 to −2.5 SD below the mean [3], [4], [5], [6]. Constitutionally-small fetuses and FGR have also been associated with a number of adverse perinatal outcomes. Adverse obstetric outcomes include hypertensive disorders of pregnancy (HDP), placental abruption, cesarean delivery (CD), and preterm delivery (PTD) [7], [8], [9], [10], [11], [12], [13]. Neonatal adverse outcomes include intrauterine fetal demise (IUFD), hypoglycemia, respiratory distress syndrome (RDS), hyperbilirubinemia, sepsis, hypocalcemia, hypothermia, intraventricular hemorrhage (IVH), necrotizing enterocolitis (NEC), neonatal seizures and death [14], [15], [16], [17], [18]. Adverse outcomes such as HDP, PTD, and neonatal adverse outcomes secondary to chronic fetal hypoxia are more likely in FGR diagnosed <32 weeks [19], [20], [21], [22], [23], [24], [25], [26], [27].

Previous studies have investigated the utility of fetal Doppler indices and biomarkers to monitor fetal status and predict obstetric and neonatal prognosis [21], 25], [28], [29], [30]. These studies have shown that an elevated umbilical artery (UA) Doppler pulsatility index (PI)≥95 % for GA is associated with worse prognosis. Prognosis progressively worsens when absent end diastolic flow (AEDF) and/or reversed end diastolic flow (REDF) of the UA is identified [21], 25], [28], [29], [30]. These findings are useful to guide the management of pregnancies beyond the limits of viability or, more specifically, beyond the limits at which active management of the pregnancy (i.e., delivery and neonatal support) would be considered. The clinical utility of these findings can be questioned near the limits of viability when delivery for fetal benefit may be futile if the GA and/or birthweight preclude neonatal resuscitative care. At the limits of periviability, often GA between 22 and 25 weeks and a birthweight (prenatally, EFW)<400 g, there is a dearth of evidence to guide the timing of initiation or frequency of UA Doppler surveillance among pregnancies with constitutionally-small fetuses or FGR [2], [30], [31], [32]. Figueras et al. (2014) proposed a management strategy for early-onset FGR; however, recommendations <26 weeks were omitted [28].

There is no standardized practice for timing and frequency of UA Doppler surveillance among pregnancies with constitutionally-small fetuses and FGR diagnosed at or below periviability at the study institutions. Management of pregnancies complicated by perviable FGR (pFGR) 1) an EFW<500 g and <10 % for GA and 2) a GA between 22 0/7 and 27 6/7 weeks has been individualized based on patient-specific characteristics. Management of pFGR has typically involved either initiation of weekly UA Doppler surveillance at pFGR diagnosis (irrespective of GA and EFW at diagnosis) or initiation of serial fetal growth assessments (every 3–4 weeks) until the fetus reaches an EFW≥500 g – a cutoff above which neonatal resuscitative efforts were more successfully implemented and neonatal prognosis improved after accounting for EFW variance from birthweight. This study aimed to compare antenatal surveillance strategies by comparing perinatal death and adverse outcomes between pregnancies with pFGR that underwent weekly UA Doppler surveillance between 22 0/7–27 6/7 weeks gestation vs. pregnancies that underwent serial fetal growth assessments during the same time frame.

Materials and methods

A retrospective cohort study was performed between January 2017 and June 2023 at two tertiary care centers in Cleveland, Ohio – The MetroHealth System and University Hospitals Cleveland Medical Center. IRB approvals were obtained from The MetroHealth IRB (IRB#00000382) and University Hospitals IRB (IRB#20230844). All complete (CPT code 76805), detailed fetal (CPT code 76811) and follow-up (CPT code 76816) obstetric ultrasounds were evaluated. Pregnant individuals with singleton gestations complicated by periviable fetal growth restriction (pFGR) were included. pFGR was defined as 1) EFW<500 g and <10 % for GA and 2) a GA between 22 0/7 and 27 6/7 weeks by best obstetric estimate. An EFW<10 % for GA was selected to represent growth restriction [1] as this was institutional practice. Classification of growth restriction by parameters of the Delphi consensus [2] was not standard institutional practice. Pregnancies were excluded if 1) delivery occurred outside of the included institutions, 2) gestational dating was confirmed at ≥16 weeks gestation, and/or 3) the fetus had a known structural or genetic anomaly. Pregnancies with abnormal cell-free DNA prenatal screening were also excluded. Among individuals with>1 pregnancy who met eligibility criteria within the study period, the earliest pregnancy was selected for inclusion. During the study period, neonatal resuscitation was offered at both institutions beginning at 22 0/7 weeks.

The primary outcome was a perinatal death which was comprised of IUFD or neonatal death prior to hospital discharge (NND). Secondary outcomes included IUFD, NND, an adverse neonatal outcome composite and its individual components, and an adverse obstetric outcome composite and its individual components. A pregnancy met criteria for the adverse neonatal adverse outcome composite if any one of the following were diagnosed in the neonate prior to hospital discharge: hypoglycemia, hypothermia, IVH, NEC, RDS, hyperbilirubinemia, hypocalcemia, seizures, and sepsis. A pregnancy met criteria for the adverse obstetric outcome composite if any one of the following were diagnosed prior to hospital discharge: HDP, hemolysis, elevated liver enzymes, low platelets (HELLP) syndrome, placental abruption, PTD, classical CD, and postpartum hemorrhage (PPH).

Baseline characteristics, prenatal and delivery outcomes, ultrasound findings, neonatal characteristics, the primary outcome, and secondary outcomes were compared between pregnancies that initiated weekly UA Doppler surveillance between 22 0/7 and 27 6/7 weeks (surveillance group) and pregnancies that underwent serial fetal growth surveillance without weekly UA Doppler surveillance between 22 0/7 and 27 6/7 weeks (serial growth group, SG group). Allocation to the surveillance group vs. the SG group was at the clinical discretion of the performing provider in conjunction with patient counseling and shared decision making. UA Doppler surveillance ≥28 weeks in the SG group was considered standard of care for the purposes of this study and did not alter the study group allocation.

Counts (%) and medians (interquartile ranges) were used, where applicable, for population description. Categorical characteristics and primary and secondary outcomes were compared between the surveillance and SG groups using Chi-squared tests. ANOVA and Kruskal-Wallis test were also used for comparison of parametric and non-parametric continuous variables, respectively. A p value of<0.05 was deemed significant. Post-hoc exploratory analyses were conducted. Uni- and multivariate, binomial logistic regression were performed to assess the association between weekly UA surveillance on the primary outcome and secondary outcome composites when adjusted for maternal history of pre-existing hypertension (CHTN) and a history of IUFD. Using Stata18 statistical software [33], time to delivery was analyzed via survival analysis between the surveillance and SG groups using competing risk regression with livebirth defined as the terminal event, stillbirth the competing event, and no instances of censoring.

Results

Between January 2017 and August 2023, 447 pregnancies met criteria for pFGR. Two hundred twenty-three pregnancies met exclusion criteria which resulted in 224 pregnancies (50.1 %) for final analyses. Eighty-one (36.2 %) pregnancies were in the surveillance group (Figure 1). In the surveillance group, weekly surveillance began at a median GA of 23 weeks (22, 24.5) and a median EFW of 428 g (349, 482.5).

Figure 1:

Study population selection.

There was no difference in maternal demographics between the surveillance and SG groups (Table 1). CHTN (18.5 % vs. 9.1 %, p=0.04) and a prior history of IUFD (8.6 % vs. 2.1 %, p=0.02) occurred more frequently in the surveillance group than the SG group (Table 1). There was no difference in the GA at pFGR diagnosis between the groups (Table 1). More ultrasounds were performed in the surveillance group (11 ultrasounds [6], 14]) when compared to the SG group (6 ultrasounds [3], 9]; p<0.05; Table 1). Aspirin had been prescribed more often in the surveillance group than the SG group (34.6 % vs. 18.9 %, p<0.05, respectively). First trimester pregnancy-associated plasma protein-A (PAPP-A) was performed in 42 % (94/224) of the study population with 49.4 % (40/81) performance in the surveillance group and 37.8 % (54/143) performance in SG group (p=0.09). Multiples of the median (MoM) for first trimester PAPP-A were lower in the surveillance group (0.6 MoM [0.4, 1]) than the SG group (0.8 MoM [0.4, 1.3], p=0.02; Table 1). Second trimester alpha-fetoprotein (AFP) was performed in 14.7 % (33/224) of the study population with less frequent performance in the surveillance group than the SG group (7.4 % vs. 18.9 %, p=0.03). MoM for AFP were higher in the surveillance group (1.6 MoM [1.2, 3.3]) than the SG group (0.8 MoM [0.6, 1.1], p<0.05).

Delivery occurred at an earlier GA in the surveillance group (34 weeks [27, 37.5]) than the SG group (37.2 weeks [37, 39], p<0.05; Table 1) which also correlated to a higher rate of PTD (56.8 % vs. 22.4 %, p<0.05, respectively). PTDs in the surveillance group were more likely to receive a course of antenatal corticosteroids when compared to those in the SG group (76.1 % vs. 50 %, p=0.02). Among all PTDs, the proportion of deliveries for fetal indications did not differ between the surveillance and SG groups when stratified by GA at delivery (Table 2). CD occurred more often in the surveillance group (56.8 %) than the SG group (32.2 %, p<0.05). Of the pregnancies culminating in CD, a larger proportion in the surveillance group received a classical hysterotomy (34.8 % vs. 15.2 %, p=0.03; Table 1).

There was no difference in the distribution of neonatal sex within each group (Supplementary Table 1). Neonates born to pregnancies within the surveillance group had lower birthweights (1,935 g (665, 2,582.5) vs. 2,560 g ([2,130, 2,915], p<0.05), were more likely to be admitted to the neonatal intensive care unit (58.1 % vs. 21.6 %, p<0.05), and were less likely to be discharged from the hospital without a co-morbid diagnosis (23.9 % vs. 53.9 %, p<0.05). Among the neonates admitted to the neonatal intensive care unit, there was no difference in the length of stay between the groups (Supplementary Table 1).

There was no difference in the primary outcome – perinatal death – or in IUFD or NND between the surveillance and SG groups (Table 3). The neonatal adverse outcome composite occurred more frequently in the surveillance group (60.8 %) than the SG group (28.4 %, p<0.05; Table 3). Hypoglycemia (24.3 % vs. 11.2 %, p=0.01), RDS (41.9 % vs. 9.7 %, p<0.05), hyperbilirubinemia (36.5 % vs. 17.2 %, p<0.05), and hypocalcemia (5.4 % vs. 0.7 %, p=0.04) were diagnosed more often among neonates of the surveillance group than the SG group (Table 3). The obstetric adverse outcome composite occurred more frequently in the surveillance group (65.4 %) than the SG group (42.7 %, p<0.05; Table 3). Specifically, HDP (45.7 % vs. 25.2 %, p<0.05), PTD (56.8 % vs. 22.4 %, p<0.05) and classical CD (19.8 % vs. 4.9 %, p=0.03) occurred more frequently in the surveillance group than the SG group (Table 3).

Based on univariate analyses and clinical relevance, multivariate logistic regression analysis demonstrated no association between surveillance frequency and perinatal death (Table 4). When adjusted for weekly Doppler surveillance, CHTN, a prior history of IUFD, prescription of low-dose aspirin in index pregnancy, and pre-existing diabetes, this multivariate logistic regression analysis did demonstrate an association between a history of pregnancy-induced hypertension and perinatal death (adjusted odds ratio [aOR] 3.34, 95 % CI 1.00, 10.09, p=0.04). Both weekly Doppler surveillance (aOR 1.95, 95 % CI 1.06, 3.61, p=0.03) and CHTN (aOR 3.51, 95 % CI 1.38, 9.72, p=0.01) were associated with obstetric adverse composite outcome after adjusting for weekly Doppler surveillance, a prior history of IUFD, CHTN, a history of pregnancy-induced hypertension, prescription of low-dose aspirin in index pregnancy, and pre-existing diabetes (Table 4). Neonatal adverse outcome composite was not associated with weekly Doppler surveillance; however, regression analysis demonstrated an association between receipt of antenatal steroids (aOR 7.20, 95 % CI 1.75, 34.5, p=0.01), GA at delivery (aOR 0.72, 95 % CI 0.55, 0.92, p=0.01), and mode of delivery (aOR 2.39, 95 % CI 1.07, 5.33, p=0.03).

Table 5 compares the fetal characteristics and UA Doppler findings between groups at pFGR diagnosis. More severe FGR (EFW<3 %) with abnormal Doppler indices was seen in the surveillance group (22.2 % vs. 10.5 %, p=0.02). Any abnormal UA Doppler finding was seen more frequently in the surveillance group (46.9 %) than the SG group (24.5 %, p<0.05; Tables 5 and 6). Further, severe UA Doppler findings (i.e., persistent REDF) were seen more often in the surveillance group than the SG group (42.1 % vs. 11.4 %, p<0.05; Tables 5 and 6). For all pregnancies with any abnormal UA Doppler index, the median time from worst UA Doppler finding to delivery was 4 weeks ([1, 10.1], Table 6). When stratified by UA Doppler abnormality, there were no differences in the interval from diagnosis of elevated PI, AEDF, or REDF to delivery. The interval (weeks) between diagnosis of intermittent absent end diastolic flow (iAEDF) and delivery was shorter in the surveillance group (1.3 weeks [0.9, 3.4]) than the SG group (12 weeks, p=0.02); however, only one pregnancy met criteria for iAEDF in the SG group. Competing risk regression did not identify differences in time to delivery of a livebirth between groups (p=0.093, Figure 2).

Figure 2:

Time to delivery analysis: Time to delivery was analyzed via survival analysis between the surveillance and serial growth (SG) groups using competing risk regression with livebirth defined as the terminal event, stillbirth the competing event, and no instances of censoring. Figure 2 demonstrates that there were no differences in time (days) from periviable fetal growth restriction (pFGR) diagnosis to delivery of a liveborn between the surveillance and SG groups.

Discussion

This study demonstrated that there was no difference in perinatal death between fetuses with pFGR that underwent weekly UA Doppler surveillance vs. serial fetal growth evaluation from 22 0/7 to 27 6/7 weeks gestation. Adverse obstetric and neonatal outcomes occurred more frequently among pregnancies that underwent weekly UA Doppler surveillance. After adjusting for relevant clinical factors, the strength of the association between UA Doppler surveillance and neonatal outcomes was mitigated; however, an association between UA Doppler surveillance and adverse obstetric outcomes remained. These findings may suggest that baseline patient characteristics, rather than antenatal surveillance frequency, are the predominant determinant of the adverse outcomes that were observed. However, the independent association between weekly Doppler surveillance and adverse obstetric outcomes suggests that antenatal surveillance frequency may have an influence on perinatal management (e.g., delivery timing, route of delivery). Despite no difference in GA at pFGR diagnosis between the groups, fetuses in the surveillance group were more likely to demonstrate features of severe growth restriction at diagnosis – EFW<3 % for GA, UA PI>95 % for GA, EFW<3 % for GA and UA PI>95 % for GA – and were more likely to demonstrate abnormal UA Doppler findings, specifically REDF, by delivery. There was no difference in time from pFGR diagnosis to delivery of a livebirth between the groups.

Previous studies on periviable or early-onset FGR demonstrated similar prevalence of CHTN (12.5 % vs. 2–13.8 % when clearly stated) and history of prior IUFD (4.5 % vs. 4 %) [19], 20], 25]. Substance use has not been routinely reported in previous studies on early-onset FGR, though Meler et al. (2023) report on the prevalence of substance use with smoking being the most commonly-reported substance at approximately 16 % and Mendoza et al. (2021) report a smoking rate of 18.2 % [19], 20]. This appears similar to rates reported within this study – 27.7 % with any substance use and 16.5 % with tobacco use. For the primary outcome, this study demonstrated similar rates of IUFD (7.1 %) and NND (4.3 %) as reported in previous studies (2–12 % and 3–18 %, respectively) [19], [20], [21, [24], [25], [26]. When compared to previously reported adverse perinatal outcomes, components of both the secondary obstetric and neonatal outcome composites were similar (Supplementary Table 2). The GA at delivery reported in previous studies ranges from 27.7 to 37 weeks [19], [20], [21], [22, 25], 29]. The GA at delivery in this study lands at the higher end of previously reported rates. The cited studies include populations with more severe features upon FGR diagnosis (e.g., AC<3 %, abnormal UA Dopplers at diagnosis, etc.) when compared to this study which likely explains the earlier deliveries [19], [20], [21], [22, 25], 29].

Pre-existing maternal co-morbidities – CHTN and history of IUFD – were more prevalent among the surveillance group. This may have introduced selection bias in our study population, demonstrating a tendency to assign high-risk pregnancies into the surveillance group while lower-risk pregnancies were assigned to the SG group. This potential selection bias was further suggested by the following high-risk characteristics seen more frequently among the surveillance group: lower PAPP-A, higher AFP, more fetuses demonstrating an EFW<3 % and abnormal UA Dopplers at diagnosis. This bias may reflect a predisposition of the surveillance group to develop obstetric complications such as HDP, leading to increased rates of medically-indicated PTD, and subsequent neonatal morbidity, when compared to the SG group. This is supported by the persistent association between weekly surveillance and adverse obstetric and neonatal adverse composite outcomes after adjusting for CHTN and prior history of IUFD (Supplementary Table 1).

Alternatively, increased adverse obstetric and neonatal outcomes in the surveillance group with no difference in perinatal death between the groups may suggest that exposure to weekly UA Doppler surveillance allowed for timely delivery which may have been protective against IUFD. Timely intervention (delivery and antenatal corticosteroids) may also have been protective against severe fetal deterioration, supportive of improved immediate neonatal transition, and beneficial to neonatal survival rates. Nonetheless, these interventions are likely at the cost of obstetric (PTD, classical CD) and neonatal morbidity. Should this theory hold true, the authors assumed PTDs for fetal indications would occur more frequently among the surveillance group than the SG group. Investigation, however, revealed no difference in the rates of PTDs for fetal indications at ≤28, ≤32, ≤34, and ≤36 weeks (Table 2, Figure 2). Additionally, no differences were identified in time from pFGR diagnosis to delivery of liveborn. Therefore, the authors conclude that the increased baseline risk among the surveillance group more likely contributed to worse obstetric, fetal, and neonatal morbidities, though not at the cost of increased fetal or neonatal mortality.

The strengths of this study include the evaluation of pregnancies from multiple tertiary care centers and the considerable cohort size. Larger cohort studies have been published by both Li et al. (2014) and Gairabekova et al. (2021); however, both populations presented in these studies were selective of high-risk pregnancies only (i.e., history of HDP or IUFD) [22], 25] or included pregnancies with FGR diagnosis up to 32 weeks [25]. Most importantly, this study aims to address the gap in periviability management recommendations through investigation of the overall utility of weekly UA Doppler surveillance at the limits of periviability. Previous studies have, instead, described the pattern of various UA Doppler trends following periviable, or early onset, FGR or provided surveillance recommendations ≥26 weeks [20], 22], 25], 28], 29]. Limitations of this study include the retrospective design and, therefore, lack of case randomization. The included institutions did not universally interrogate middle cerebral artery (MCA) or ductus venosus (DV) Doppler profiles or computerized cardiotocography which may limit comparison to previous studies of early FGR [34]. Additionally, the study sample size precludes powered analysis of severe perinatal outcomes among this relatively rare clinical cohort, especially sub-analyses involving of individual maternal and fetal characteristics. Post-hoc calculation indicates that approximately 800–6,000 pregnancies would be required to detect between 20 % and 50 % decrease in the primary outcome of perinatal death from the population rate of 10.5 % in the SG group with 80 % power and 0.05 type I error. A multi-center, prospective, randomized controlled trial should be considered to investigate optimal Doppler initiation and frequency among pregnancies complicated by pFGR. However, based on available data, the authors propose that in pregnancies with increased baseline risk, weekly Doppler surveillance may be considered after a diagnosis of pFGR to help with prognostication. Nonetheless, decisions aimed at intervention (such as delivery) should be deferred until the fetus reaches an EFW that aligns with the institution’s resuscitation thresholds.

In conclusion, weekly surveillance was associated with a higher rate of perinatal complications despite no differences in perinatal death. As an observational, cohort study, these findings suggest that pre-existing maternal and obstetric risk factors may have driven surveillance stratification of higher-risk patients into the weekly surveillance group. Prospective studies are necessary to better understand the clinical benefits of weekly ultrasound surveillance among fetuses with pFGR.