Eligibility criteria, information sources, and search strategy

Given the rarity of vasa previa, studies included in this review were cohort studies, case series, and case reports. Because of the limited literature on neonatal outcomes from pregnancies affected by vasa previa, conference abstracts were not strictly excluded from the eligibility criteria; however, they were excluded if they were too brief to draw relevant conclusions from or if they did not contain enough data on the outcomes of interest. Studies were only included if they were available in English and if they were human studies. Studies only proceeded to the quantitative synthesis if they had more than 5 cases and if a detailed breakdown of complications by gestational age was available.

A systematic search of the Ovid MEDLINE, PubMed, CINAHL, Embase, Scopus, and Web of Science databases from inception to January 2022 was performed to identify studies that analyzed prenatally diagnosed cases of vasa previa and neonatal outcomes. The reference lists of relevant studies were also searched to identify any additional studies that may not have been captured in the initial searches.

A combination of medical subject heading terms and key terms and variants of “prenatally diagnosed,” “vasa previa,” “delivery,” “cesarean,” and “outcomes” were used to search the aforementioned databases. Variations of search terms were combined with the Boolean operator “OR,” and the different elements of the Population, Intervention, Comparator, and Outcome framework were combined with the Boolean operator “AND.” The complete search strategy is included in Appendix A .

Study selection

The study selection process was performed using Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). After duplicates were removed, 2 independent reviewers (G.N. and S.J.M.) screened the title and abstract of each study for eligibility; studies that were considered relevant by both reviewers, and all studies over which there was disagreement, proceeded to the next stage of screening. Full texts of these studies were then assessed for eligibility by the same 2 independent reviewers with strict adherence to the predefined inclusion and exclusion criteria. Disagreements were discussed, and any persistent conflicts were resolved through discussion with a third reviewer (D.L.R.).

Studies used in the qualitative synthesis were excluded from the quantitative synthesis if they had a sample size <5, because the presence of outcomes in studies with smaller sample sizes were likely to cause overly significant shifts in the proportions. When data breakdown by gestational age was not reported, authors were contacted and additional data were requested. Per the eligibility criteria, if sufficient data could not be obtained, the study was excluded from the quantitative analysis.

Data extraction

Two authors (G.N. and S.J.M.) independently extracted the author names, year of publication, study design, sample sizes, and outcomes directly from the manuscripts. In addition, for the quantitative synthesis, the counts of relevant outcomes at each gestational age were also extracted, either directly from the manuscript or from a raw data set requested from authors. Gestational ages were grouped as <32 weeks, 32 weeks (ie, 32+0 to 32+6 weeks), 33 weeks, 34 weeks, 35 weeks, 36 weeks, 37 weeks, and ≥38 weeks. Once the extraction was complete, the spreadsheets were then directly cross-checked between the 2 authors to identify any discrepancies; these were resolved through consensus among the authors.

Assessment of risk of bias

The quality assessment was completed using the National Institutes of Health Study Quality Assessment Tools, as recommended by Ma et al, for nonrandomized studies. The tool for observational cohort and cross-sectional studies was used for cohort studies, and the tool for case series was used for both case series and case reports. These checklists were completed by 2 independent reviewers (G.N. and S.J.M.) and compared. Discrepancies were again resolved through discussion. Study quality was globally rated as poor, fair, or good.

Outcomes of interest and data synthesis

Data extracted from each study for the quantitative analysis included the total number of neonates delivered at each gestational age and the number of neonates with the outcomes of interest These included perinatal mortality, respiratory distress syndrome (RDS) requiring intubation, Apgar scores <7 at 5 minutes, neonatal intensive care unit (NICU) admission, low birthweight (<2500 g), and neonatal blood transfusions. We then analyzed the aggregate rates of each outcome of interest overall and by gestational age. Pooled proportions of different adverse pregnancy outcomes, overall and stratified by gestational age, were calculated with random-effects models using inverse-variance weights and the arcsine transformation to achieve stabilization of the variances. Publication bias was investigated by inspection of funnel plots and the Egger’s test for outcomes reported by 10 studies or more. Finally, we performed a sensitivity analysis of the aggregate rates of adverse pregnancy outcomes and included only studies considered to be of good quality to investigate the impact of studies at high risk of bias on the results. Analyses were conducted with the package metafor in the statistical software package R (R Core Team, Vienna, Austria).

Study selection

A total of 3435 studies were identified through the search, 1264 of which were duplicates. After screening 2171 titles and abstracts, 140 studies were deemed potentially relevant and were included in the full-text screen, but 9 reports could not be retrieved. A further 94 were then excluded because of disagreement with the predefined eligibility criteria or irrelevance. This left 37 studies to be included in the systematic review. The study selection process is presented in Figure 1 , and the characteristics of included studies are included in Supplemental Table 1 .

PRISMA flow diagram

PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses

Mitchell. Timing of birth in cases of antenatally diagnosed vasa previa. Am J Obstet Gynecol 2022.

Three studies were excluded from the quantitative synthesis because they had a study sample size of <5. ^{, ,} Of the remaining 34 studies, 4 presented a data breakdown by gestational age. ^{, , ,} The remaining 30 studies only reported the total incidence of complications across all deliveries and did not specify the gestational age at which these occurred. ^{, , , , , , ,} Authors of these articles were contacted to request the breakdown of complications by gestational age. A total of 19 authors either could not be contacted, did not respond, or did not have the data that we requested available. ^{, , , , , , , , , ,} The corresponding author of 1 study reported that their data was included in another, larger study and so this was treated as a duplicate and was removed from the quantitative synthesis. A total of 10 sets of raw, stratified data were obtained from corresponding authors. ^{, , , , , , , , ,} Hence, 14 studies were included in the quantitative synthesis. ^{, , , , , , , , , ,}

Risk of bias

In total, there were 3 case reports, 21 case series, and 13 cohort studies. The scoring and overall rating of the studies are shown in Supplemental Tables 2 and 3 .

In terms of case series and case reports, the criterion that was consistently poor was the use of statistical analysis; however, reviewers believed that this was not necessary in most cases given the small sample sizes and the descriptive nature of the series. Fifteen of the 21 case series were rated as good, and 6 were rated as fair. All 3 case reports were rated as fair.

There were some criteria that consistently were not applicable to cohort studies. These included variation in the amount or level of an exposure (because one can only have vasa previa or not), whether outcome assessors were blinded (because most studies were retrospective), and whether there was a sample size justification (this was considered unnecessary because most studies included all consecutive cases within a reasonable period). Overall, 6 cohort studies were rated as good, and 7 were rated as fair.

There was difficulty in assessing the overall quality of conference abstracts because of the lack of essential information for this. For this reason, it was difficult for the global assessments of conference abstracts to exceed fair.

Qualitative synthesis

The most commonly reported outcomes included perinatal mortality, ^{, , , , , ,} RDS requiring intubation, ^{, , , , , , , , ,} 5-minute Apgar scores <7, ^{, , , , , , , , , , , , , , ,} requirement for blood transfusion by the neonate, ^{, , , , , , , , , , ,} NICU admission, ^{, , , , , , , , , , , , , ,} and low birthweight. ^{, , , , , , , , , , , , , ,}

The number of cases of vasa previa in the included studies ranged from 2 to 586 with a mean of 63.7 and a median of 23. Three of the 37 studies were prospective, 2 of which were case series and 1 was a cohort study. ^{, ,} Broadly, cohort studies that compared the outcomes of prenatally diagnosed vasa previa with vasa previa that was undiagnosed prenatally reported a reduced risk of neonatal mortality among diagnosed cases. ^{, , ,} Studies comparing women without vasa previa with women with vasa previa demonstrated an increased risk for neonatal mortality, RDS requiring intubation, and a need for neonatal blood transfusions. Perinatal mortality ranged from 0% to 16.6% in the studies analyzing prenatally diagnosed vasa previa. NICU admission was a common event for neonates delivered from pregnancies affected by vasa previa with the rates ranging from 53.5% to 100%.

Other complications of prematurity were also included in a small number of studies. Six studies reported on intraventricular hemorrhage (IVH) (0%–16.7%), ^{, , ,} 2 studies reported on cases of bronchopulmonary dysplasia (BPD) (0.02%–8.7%), ^, and 5 studies reported cases of necrotizing enterocolitis (NEC) (0%–8.7%). ^{, , , ,} However, there was not enough data available to perform a quantitative analysis on these neonatal outcomes.

Of the 37 included studies, 14 provided recommendations on the timing of delivery, which ranged from 33 to 37 weeks’ gestation. The most common recommendations were 34 to 35 weeks ^{, ,} and 35 weeks’ gestation. ^{, ,} Only 1 study suggested delivery earlier than this at 33 to 34 weeks’ gestation. Broader recommendations included 34 to 36 weeks, 34 to 37 weeks, and 35 to 37 weeks’ gestation. ^, Some studies also proposed delivery at 35 to 36 weeks ^, and 36 weeks’ gestation.

Of note, although all studies largely defined vasa previa as fetal vessels running unprotected over or close to the internal cervical os, 4 studies defined the distance of the fetal vessels to the internal cervical os explicitly. ^{, , , , , ,} Six studies defined vasa previa as fetal vessels within 2 cm of the cervical os, and the last defined close proximity as being within a distance of 4 cm.

Quantitative synthesis

A total of 490 neonates across the 14 studies were included in the analysis. This included 44 (9.0%) neonates born before 32 weeks’ gestation, 20 (4.1%) born at 32 weeks, 34 (6.9%) born at 33 weeks, 122 (24.9%) born at 34 weeks, 145 (29.6%) born at 35 weeks, 80 (16.3%) born at 36 weeks, 33 (6.7%) born at 37 weeks, and 12 (2.4%) bor at 38 weeks’ gestation or later. Figure 2 demonstrates the unweighted pooled rates of perinatal complications, whereas Supplemental Table 4 summarizes the rates (numerically) of different perinatal complications. Supplemental Figure 1 presents the forest plots of the inverse-variance weighted proportions of the morbidity outcomes overall, and Supplemental Figure 2 presents the proportions by gestational age at birth. Meta-analyses were not possible for the outcome of perinatal mortality given the low number of events.

Pooled rates of adverse pregnancy outcomes

Numbers and percentages provided in Supplemental Table 4 .

NICU , neonatal intensive care unit

Mitchell. Timing of birth in cases of antenatally diagnosed vasa previa. Am J Obstet Gynecol 2022.

There was no evidence of publication bias for the outcomes of respiratory distress requiring intubation, Apgar score <7 at 5 minutes, low birthweight, and blood transfusion (Egger’s test P values .196, .833, .132, and .817, respectively) ( Supplemental Figures 3 , A, B, D, and E, respectively). There was some evidence of small study effects for the outcome of NICU admission with a tendency of smaller studies reporting lower rates of NICU admission (Egger test P value.033) ( Supplemental Figure 3 , C). A sensitivity analysis restricted to good quality studies showed similar trends as those of the main analysis ( Supplemental Table 5 and Supplemental Figure 4 ).

Principal findings

In this systematic review, we analyzed 37 studies. The qualitative synthesis identified relevant neonatal outcomes relating to vasa previa, including perinatal mortality, RDS, 5-minute Apgar scores <7, NICU admissions, low birthweight, and requirement for neonatal blood transfusion; other complications of prematurity (including IVH, BPD, and NEC) were less commonly reported neonatal outcomes. Moreover, the qualitative synthesis reiterated the importance of a prenatal diagnosis of vasa previa.

A total of 14 studies were included in the quantitative synthesis to investigate the gestational age with the lowest rate of complications. This analysis found a downward trend in the rate of complications until 36 weeks’ gestation; this gestational age saw no perinatal deaths or 5-minute Apgar scores <7, low rates of RDS (5.3%) and blood transfusion (2.9%), and relatively low rates of NICU admission (29.2%) and low birthweight (30.9%). At 37 weeks’ gestation, the rates of most complications slightly increased again (RDS, 5-minute Apgar score <7, blood transfusions, and NICU admission), but generally remained low. The lowest absolute risk of complications seemed to be at 38 weeks’ gestation or later when the rate of all complications was 0.0%, except for neonatal blood transfusion (8.3%). However, it must be considered that there was only a small number of neonates born after 38 weeks (n=12), which was likely because of guidelines suggesting delivery before this time. The small sample size at 38 weeks’ gestation and later significantly limits the reliability of the conclusions about the rates of adverse outcomes at these gestational ages; hence, it cannot be confidently stated that delivery at this time is safe. Given the decrease in complications until 36 weeks and the small increase in complications at 37 weeks, this may suggest that waiting until late preterm may be the best time to deliver in otherwise uncomplicated pregnancies with prenatally diagnosed vasa previa.

In total, there were 2 neonatal deaths (0.4% of all neonates in this study), both of which were caused by complications of prematurity. Because we only included cases of vasa previa with a prenatal diagnosis, the low rate of perinatal mortality is in line with previous studies demonstrating much lower death rates in cases diagnosed prenatally than in those without a prenatal diagnosis.

When examining the safest time to deliver in cases of prenatally diagnosed vasa previa, it is also important to consider the individual patient. For example, some women may be at higher risk for preterm premature rupture of membranes (PPROM) than others; a relevant predictor of PPROM is short cervical length, which can be assessed using a transvaginal ultrasound. Hence, this relatively simple assessment could provide valuable patient-specific information about the risk of fetal vessel rupture at earlier gestational ages and subsequently guide the timing of delivery. Recent literature has also suggested elective hospitalization for women in the weeks before a planned delivery, which would allow closer monitoring for signs of preterm labor and timely access to an emergency cesarean delivery if indicated. However, data to endorse this as standard practice are lacking, and the cost-effectiveness of this strategy has not been adequately explored.

Comparison with existing literature

There have been some recent systematic reviews on vasa previa that have explored risk factors and neonatal outcomes in cases of vasa previa. ^{, ,} This review differs from these in that we examined a broader range of outcomes and included both a qualitative and quantitative synthesis.

We found 14 studies that provided recommendations on the timing of delivery in cases of vasa previa. Suggestions ranged from as early as 33 to 34 weeks’ gestation up until 37 weeks of gestation. Half of the 14 studies recommended delivery at 35 weeks’ gestation or earlier. ^{, , , , , , ,} One study recommended 36 weeks’ gestation. The remaining studies proposed a window of time that extended to at least 36 weeks’ gestation, but also included gestations at 35 weeks or earlier. ^{, , ,} Our data suggest that it is generally safe and beneficial to prolong otherwise uncomplicated pregnancies with a diagnosis of vasa previa to at least 36 weeks’ gestation, which is slightly later than most recommendations made by previous studies.

Another important resource to compare these findings with are current guidelines. The American College of Obstetricians and Gynecologists and the Society for Maternal-Fetal Medicine both recommend birth at 34 to 37 weeks’ gestation, and the Royal College of Obstetricians and Gynaecologists (United Kingdom) and the Royal Australian and New Zealand College of Obstetricians and Gynaecologists recommend delivery at 34 to 36 weeks’ gestation. ^{, ,} Given the lack of large studies to base these recommendations on, most guidelines rely on expert consensus of a gestational age period at which the rates of adverse neonatal outcomes are usually low in high-resource settings. Our finding of relatively low complication rates at 36 weeks’ gestation supports these guidelines and could also add greater specificity.

Strengths and limitations

Existing literature on this topic commonly analyzed the cohort as a whole and reported mean event rates. This study analyzed the rate of complications at each discrete gestational age and the trends in these adverse outcomes and stratified neonatal outcomes in cases of vasa previa by gestational age to find the age with the lowest rate of complications.

There were also weaknesses in this study. Because the incidence of vasa previa is low, there is only limited literature available. Although this study combined results of many previous studies, the stratified data from some of the largest studies could not be obtained. This left us with a relatively small sample (490 cases) from which to obtain reliable estimates and this potentially introduced some bias into our results. Future studies should hence focus on prospective data collection through multicenter collaborations to increase the confidence and have clear and uniform criteria for a vasa previa diagnosis.

When interpreting these results, it is important to consider that not all outcomes are equal in severity and that individual patients would likely have differing opinions on which risks (prematurity or vasa previa) they are more willing to take. However, the rates of all complications seem to be well balanced at 36 to 36+6 weeks, thereby minimizing the need for subjective data interpretation. Moreover, long-term outcomes were not reported and therefore could not be accounted for in this study. For example, neonates born with RDS are at risk of developing bronchopulmonary dysplasia. Prolonged fetal anemia, which could be secondary to preterm delivery or fetal hemorrhage, can lead to ischemic brain lesions and, consequently, neurodevelopmental delay or other forms of disability. These could form further important considerations that women and clinicians may have to take into account during the decision-making process regarding the timing of birth.

Because the data we received from authors were aggregate patient data, we were unable to obtain the indications for delivery, including whether the delivery was planned or emergent. This is a limitation to our study, because emergent or medical indications for earlier delivery may bias the data toward more complications. A subgroup analysis examining emergent deliveries and the specific indications before planned deliveries would have been valuable but was not possible for this review. Formal meta-analysis techniques were used to consider clustering and intra- and between-study variability in the calculation of weights and in the assessment of statistical heterogeneity. Further adjustments for clustering are only possible in individual participant data meta-analyses, which were outside the scope of this study. It was also unclear which particular management protocols were being followed in most studies. We acknowledge that the comparison of management protocols could also have been of great interest, but we lacked the data to do so.

This study only analyzed pregnancies that had been diagnosed prenatally. One of the most significant prognostic markers for poor neonatal outcomes in pregnancies affected by vasa previa is the lack of a prenatal diagnosis; these pregnancies are likely to be prolonged until the labor occurs spontaneously or until birth is indicated for other reasons, which places the fetus at high risk for exsanguination. Unfortunately, the benefits of screening the general population remain unclear and routine antenatal screening is currently not performed in many settings. However, with advances in ultrasound technology and guidelines for more thorough evaluation of high-risk pregnancies, many cases can be identified. ^,

Conclusions

In otherwise uncomplicated cases of prenatally diagnosed vasa previa, perinatal mortality is low and it seems safe and beneficial to prolong pregnancy until 36 weeks’ gestation. Given the possible small increase in complications at term, 36 to 36+6 weeks’ gestation may represent the gestational age that best balances the risks of complications from vasa previa with those of prematurity.