Dr L Saraswat, Department of Obstetrics and Gynaecology, Aberdeen Maternity Hospital, Aberdeen AB25 2ZL, UK. Email firstname.lastname@example.org
Please cite this paper as: Saraswat L, Bhattacharya S, Maheshwari A, Bhattacharya S. Maternal and perinatal outcome in women with threatened miscarriage in the first trimester: a systematic review. BJOG 2010;117:245–257.
Background Threatened miscarriage is a common complication in the first trimester of pregnancy and is often associated with anxiety regarding pregnancy outcome.
Objective We undertook a systematic review to explore the effects of threatened miscarriage in the first trimester on maternal and perinatal outcomes.
Search strategy An electronic literature search using MEDLINE and EMBASE, and bibliographies of retrieved primary articles. No language restrictions were applied.
Selection criteria All studies analysing outcomes of first-trimester bleeding where viability was confirmed on ultrasound or the pregnancy continued beyond viability.
Data collection and analysis Two review authors independently selected studies and extracted data on study characteristics, quality and accuracy. Meta-analysis was performed using Review Manager software
Main outcome measures The outcome was broadly categorised into maternal and perinatal outcomes. The chief maternal outcomes included pre-eclampsia/eclampsia or pregnancy-induced hypertension, antepartum haemorrhage, preterm prelabour rupture of membranes (PPROM) and mode of delivery. The perinatal outcomes evaluated were preterm delivery, low birthweight, intrauterine growth restriction, perinatal mortality, indicators of perinatal morbidity (Apgar scores and neonatal unit admission) and presence of congenital anomalies.
Main results Fourteen studies met the inclusion criteria. Women with threatened miscarriage had a significantly higher incidence of antepartum haemorrhage due to placenta praevia [odds ratio (OR) 1.62, 95% CI 1.19, 2.22] or antepartum haemorrhage of unknown origin (OR 2.47, 95% CI 1.52, 4.02) when compared with those without first-trimester bleeding. They were more likely to experience PPROM (OR 1.78, 95% CI 1.28, 2.48), preterm delivery (OR 2.05, 95% CI 1.76, 2.4) and to have babies with intrauterine growth restriction (OR 1.54, 95% CI 1.18, 2.00). First-trimester bleeding was associated with significantly higher rates of perinatal mortality (OR 2.15, 95% CI 1.41, 3.27) and low-birthweight babies (OR 1.83, 95% CI 1.48, 2.28).
Authors’ conclusions Threatened miscarriage in the first trimester is associated with increased incidence of adverse maternal and perinatal outcome.
First-trimester bleeding is a common complication which affects 16–25% of all pregnancies.1 Threatened miscarriage is diagnosed on the basis of documented fetal cardiac activity on ultrasound with a history of vaginal bleeding in the presence of a closed cervix. Bleeding during pregnancy can cause maternal anxiety and emerging evidence suggests that it may be associated with poor fetal and maternal outcomes.2–9
It is hypothesised that first-trimester bleeding may indicate an underlying placental dysfunction, which may manifest later in pregnancy causing adverse outcomes such as increased risk of pre-eclampsia, preterm delivery, preterm prelabour rupture of membranes (PPROM), placental abruption and intrauterine growth restriction (IUGR).8
Knowledge about the outcome of ongoing pregnancies following first-trimester bleeding is relevant to both women and their obstetricians in order to plan antenatal care and consider clinical interventions in pregnancy. Several primary studies have sought to identify adverse fetal outcomes, but very few have commented on maternal complications. A previous meta-analysis summarised the evidence on adverse perinatal outcome following vaginal bleeding in the first and second trimesters of pregnancy,10 but language restrictions in its search strategy and the lack of quality assessment of the studies included potentially limit the strength of its inferences. Moreover, several new primary studies have been published since that warrant inclusion in an updated systematic review. To date, we know of no systematic review that has looked into both maternal and perinatal outcomes following first-trimester bleeding. We therefore undertook a comprehensive systematic review on maternal and perinatal outcomes following first-trimester bleeding.
A study protocol stating the research question to be addressed, the population and conditions of interest, the exposure and outcomes considered, the criteria used for identifying and selecting or excluding studies, and the methods used for extracting and analysing data preceded this systematic review. We followed the guidelines of the Meta-analysis of Observational Studies in Epidemiology (MOOSE) group.11
Prospective and retrospective observational studies evaluating the association between threatened miscarriage and maternal and perinatal outcomes were identified using the computerised databases MEDLINE (US National Library of Medicine, Bethesda, MD, USA) and EMBASE (Elsevier, Amsterdam, the Netherlands). The searches were conducted for published literature from January 1976 to April 2009, without language restrictions. The search strategy was written in Ovid, then modified and run in each database. Adjacency operators and truncation were used. Our search term combination for electronic databases was MeSH headings (Medical Subject Headings, US National Library of Medicine), text words, and word variants for threatened miscarriage and for maternal and perinatal outcomes.
The citation lists were independently reviewed by two authors (LS and SB). Titles and abstracts were screened and articles were retrieved if they passed the relevance filter or if there was uncertainty as to whether or not they were relevant. References from identified studies were also screened for relevant citations. Retrieved articles were then reviewed for inclusion/exclusion criteria. Those articles that met the criteria were then kept for critical appraisal and data collection using a standard data-capture form. A review of literature suggested that the use of ultrasound scan in early pregnancy diagnosis only became popular in the late 1970s and early 1980s. Therefore, following a general consensus among the authors, we decided to include only the published literature from 1976 to April 2009. As more than one database was searched, there was some degree of duplication in the citations identified. Therefore references were managed using ‘RefWorks’software (RefWorks-COS, Proquest, Ann Arbor, MI, USA) and duplicates were removed.
Inclusion and exclusion criteria
There were variations in the definitions of ‘threatened miscarriage’ and ‘first trimester’ among published studies. We therefore adopted an inclusive approach and selected all studies that took into account pregnant women with first-trimester bleeding where viability was confirmed on ultrasound or the pregnancy continued beyond viability. Only case–control or cohort studies were included in the review. Case series and studies without controls were excluded.
We categorised outcomes broadly into maternal and perinatal outcomes. The maternal outcomes included, pre-eclampsia/eclampsia or pregnancy-induced hypertension (PIH), antepartum haemorrhage (APH; placenta praevia, abruption, other APH), PPROM, mode of delivery (instrumental and caesarean deliveries), postpartum haemorrhage (PPH) and retained placenta. The perinatal outcomes evaluated were preterm delivery (delivery before 37 completed weeks), low birthweight (birthweight ≤2500 g), IUGR, perinatal mortality, indicators of perinatal morbidity (Apgar scores and neonatal unit admission) and presence of congenital malformations.
Data synthesis and statistical analysis
We assessed the methodological quality of each study using the Newcastle–Ottawa scale.12 All studies that met our inclusion criteria were independently evaluated by two reviewers (LS and SB). We designed a data abstraction form, and the two reviewers abstracted the data separately. Discrepancies regarding the inclusion of studies or abstracted data were resolved by discussion. Where necessary, we contacted researchers to obtain additional information about study methods or outcome measures. We entered and analysed the data using RevMan 4.2 (Cochrane Collaboration, Oxford, UK). For each outcome, data were only pooled if there were at least two studies available for a particular outcome measure. As adjustments for confounding variables varied among different studies, we used the raw numbers from each study to calculate the crude odds ratios (OR) and 95% CI for each outcome before pooling the data. A random effect model was used (because of statistical heterogeneity in the outcome data) to calculate combined OR, 95% CI. Tests of heterogeneity were performed before pooling the data.
Thirty-one publications evaluating the effect of threatened miscarriage on maternal or perinatal outcome were identified. A hand search of references of the above papers identified another seven potentially useful references.
These 38 references were reviewed for inclusion and exclusion criteria. Figure 1 summarises the process of literature identification and selection.
Of these references, 16 met the inclusion criteria, although there was variation in the definition of threatened miscarriage in terms of gestational age. However, two of these articles were excluded because of insufficient data.13,14 In both cases, the authors were contacted but no response was obtained. As a consequence, a total of 145–9,15–23 studies were included in the meta-analysis. Thirteen of the 14 studies included in the meta-analysis employed a cohort design. The report by Hossain et al.17 was a case–control study. Of the cohort studies, six had a retrospective design and the other seven used prospective cohorts. The study characteristics are described in Table 1.
Cases: women with vaginal bleeding before 12 weeks where pregnancy continued to a minimum of 24 weeks Controls: Women delivering after 24 weeks within the same time period, but without first-trimester bleeding
Women with vaginal bleeding before 12 weeks where pregnancy continued to a minimum of 24 weeks
Women with complete, incomplete, or missed miscarriage. Women opting for termination Multiple pregnancies Pregnancies with fetal malformation Hydatidiform moles Second-trimester miscarriages
1. Pre-eclampsia 2. Eclampsia 3. Placental abruption 4. Placenta praevia 5.Other APH 6. PPROM 7. Induced labour 8. Instrumental delivery 9. Elective caesarean 10. Emergency caesarean 11. Postpartum haemorrhage 12. Manual removal of placenta 13. Preterm delivery 14. Malpresentation 15. Stillbirth 16. Neonatal death 17. Birthweight <2500 g 18. Apgar at 5 min <7 19. Admission to neonatal unit
Cases: women with first-trimester bleeding who delivered after 20 weeks Controls: women with no vaginal bleeding
Vaginal bleeding or spotting limited to first trimester
Pregnancies complicated with diabetes, placenta praevia, placental abruption, or idiopathic bleeding beginning in trimesters other than the first
1. Low birthweight 2. Preterm birth 3. Term low birthweight 4. Stillbirth 5. Neonatal death
Women with first-trimester bleeding had an elevated risk of adverse maternal and perinatal outcome.
Pregnancy-induced hypertension, pre-eclampsia and eclampsia
Six of the 14 studies reported on PIH, pre-eclampsia and eclampsia. The incidence of PIH, pre-eclampsia or eclampsia was not significantly altered by bleeding in first trimester; OR (95% CI) of 0.99 (0.84, 1.17). These results could be attributed to the influence of two large studies by Weiss et al.8 and Wijesiriwardana et al.9 There was no significant heterogeneity (P = 0.19) in results across the different studies (Figure 2A).
Women with first-trimester bleeding were prone to subsequent APH in pregnancy. The meta-analysis included all those studies that analysed outcomes following first-trimester bleeding where the pregnancy continued beyond viability. In this context we have defined APH as bleeding beyond viability (≥24 weeks). These women were more likely to have placenta praevia (OR 1.62, 95% CI 1.19, 2.22) as well as placental abruption (OR 1.46, 95% CI 1.00, 2.14). Antepartum haemorrhage of unknown origin was twice as likely in those with threatened miscarriage (OR 2.47, 95% CI 1.52, 4.02) as in women without first-trimester bleeding. Except for placental abruption (P = 0.03) there was no significant statistical heterogeneity (placenta praevia P = 0.46, APH of unknown origin P = 0.07) when results were pooled across the different studies (Figure 2B).
Preterm prelabour rupture of membranes
The incidence of PPROM was significantly higher in women whose pregnancy was complicated by first-trimester bleeding (OR 1.78, 95% CI 1.28, 2.48). The test of heterogeneity for PPROM was significant (P = 0.01) (Figure 2C).
Mode of delivery
First-trimester bleeding did not appear to influence the mode of delivery. The risk of instrumental delivery (OR 1.01, 95% CI 0.96, 1.07) or caesarean section (OR 0.92, 95% CI 0.73, 1.16) was not significantly altered. There was evidence of significant statistical heterogeneity (P = 0.00001) in results relating to the risk of caesarean section (Figure 2D).
Other maternal outcomes
The only study in the meta-analysis that evaluated PPH and retained placenta as an outcome is by Wijesiriwardana et al.9 They reported increased incidence of PPH (OR 1.13, 95% CI 1.04, 1.23) and manual removal of retained placenta (OR 1.45, 95% CI 1.26, 1.68). However, in the original study, after adjusting for potential confounders the increase in PPH was no longer significant.
First-trimester bleeding was a predictor of poor perinatal outcome.
The reported risk of preterm delivery in women with threatened miscarriage varied between 1.5 and 4.5 across the different studies. The overall adjusted risk of preterm delivery was 2.05 (95% CI 1.76, 2.4) in women who experienced first-trimester bleeding. There was evidence of significant statistical heterogeneity in reported results (P < 0.0001) (Figure 3A).
Intrauterine growth restriction
A significant association existed between first-trimester bleeding and IUGR. The risk of having a baby with IUGR was 1.54 (95% CI 1.18, 2.0) times in women with first-trimester threatened miscarriage. Once again there was evidence of significant statistical heterogeneity (P = 0.0002) in the reported results (Figure 3B).
The overall risk of having a low-birthweight baby was higher in women who bled in the first trimester (OR 1.83, 95% CI 1.48, 2.28) than in women who did not. The risk varied from 1.1 to 3.7 across the different studies. The test for heterogeneity was highly significant (P < 0.0001) (Figure 3C).
Perinatal deaths were observed to be nearly twice as frequent in women who experienced threatened miscarriage when pooled across different studies (OR 2.15, 95% CI 1.41, 3.27). The results displayed evidence of significant statistical heterogeneity (P = 0.001) (Figure 3D).
The women with history of early pregnancy bleeding were more likely to deliver babies with Apgar score <7 at 5 minutes after birth (OR 1.2, 95% CI 1.03, 1.4) and babies that were admitted to the neonatal unit (OR1.13, 95% CI 1.03, 1.23) (Figure 3E).
Four studies reported the incidence of congenital anomalies in babies born to mothers with early pregnancy bleeding. The odds of having a baby with a congenital anomaly was 1.26 (95% CI 0.89, 1.79) (Figure 3F).
In this comprehensive review we evaluated 12 different maternal and perinatal outcomes and found a consistent association between first-trimester bleeding and adverse fetal and maternal outcomes.
To our knowledge this is the first report to systematically review and pool data on both maternal and perinatal outcomes associated with first-trimester bleeding. It was rigorously carried out without language restrictions and met the criteria laid down in the MOOSE statement.11 We paid careful attention to quality assessment of studies and collected information important for evaluation of the validity of the observed associations, potential for bias, and causality.
Our literature search identified one previous meta-analysis by Ananth and Savitz,10 which evaluated the effect of vaginal bleeding up to 28 weeks and focused on perinatal outcomes only. This systematic review included 28 studies published between 1950 and 1992 and found that vaginal bleeding was associated with increased risk of low birthweight, preterm birth, stillbirth, perinatal death and congenital malformations in infants. However, with changes in practice and advances in medical technology, the limit of viability is now 20 weeks (World Health Organization) or 24 weeks (UK) and therefore the 28-week cutoff used by Ananth and Savitz10is no longer compatible with the current practice as there would be overlap between exposure and outcome with this approach. Moreover, the objective of our meta-analysis was to evaluate the association of bleeding primarily in the first trimester with both maternal as well as perinatal outcome. Only those studies that have used first-trimester bleeding as inclusion criteria for the women or have performed an independent analysis for first-trimester or any subsequent bleeding were included in the review. Among the included studies, five studies evaluated outcome following first-trimester and second-trimester bleeding and performed separate analysis for each trimester. For the purpose of the review, we have only used data relating to first-trimester bleeding. As a consequence, this meta-analysis aims to provide information regarding pregnancy outcomes for women who had threatened miscarriage in the first trimester.
Women with threatened miscarriage have a higher likelihood of miscarrying. Some of the studies included in the review reported miscarriage rates whereas others only included women where pregnancy continued beyond viability. Five out of the 14 studies reported miscarriage rates. Davari-Tanha et al.16 quoted a figure as high as 42.7% spontaneous pregnancy loss in first trimester whereas other studies reported miscarriage incidence of 7.8% by 14 weeks,18 9.3% in first trimester,19 5.5% by 20 weeks22 and Weiss et al.8 reported a rate of 1% for light bleeding and 2% for heavy bleeding by 24 weeks. However, none of these studies excluded these women from the denominator when reporting results.
One of the challenges of performing this systematic review was the fact that the definition of threatened miscarriage is rarely stated in explicit terms. Some studies have defined first trimester up to 12 weeks,5,9 some up to 14 weeks8,19 whereas others have just mentioned first trimester without defining gestational age in terms of weeks.6,16,22,23 Moreover, it is possible that the risk of adverse outcome may be different in women who experience ‘light’ versus ‘heavy’ bleeding. While some studies have attempted to distinguish between light and heavy bleeding,6–8 others have failed to do so. However, in this context it is important to realise that subjective assessment of blood loss is often erroneous in such situations and objective assessment is often impractical.
Limitations and potential bias
Meta-analyses are limited by biases introduced through individual studies as well as through the processes of systematic review and quantitative summary. Although we rigorously carried out an extensive literature search, we were unable to search the grey literature and unpublished data. Hence publication bias may have an impact on our results. Moreover, in meta-analyses of observational data secondary researchers are unable to adjust for potential confounders; however, the present topic does not lend itself to experimental studies including randomised trials. Despite this, the consistent associations seen in large primary studies included in our review would suggest that the direction of association would remain even if we had missed smaller studies with equivocal or negative associations.
Other concerns relate to the use of non-standard definitions with questionable validity or reliability to discriminate between the exposure and the outcome. Often studies were lacking in one or another quality feature. For these reasons, associations that are not strong and consistent should be viewed as no more than hypothesis generating. Also, confidence in some of the findings may be constrained as certain outcomes have been evaluated only by one or two studies. The study by Wijesiriwardana et al.9 was the only one that looked at outcomes like PPH, and retained placenta.
Diverse factors are associated with poor pregnancy outcome such as maternal age, social class, ethnicity and previous obstetric history so it is difficult to compare directly the results of individual outcomes across all studies because of varying degrees of control for potential confounders.
For certain outcomes like placental abruption, PPROM, preterm delivery, IUGR and low birthweight the assumption of homogeneity was violated when the overall risks were adjusted for different studies and designs. The way around this would be to perform a meta-regression, but the number of studies looking at each individual outcome was too small.
Inter-related risk factors
It is important to disentangle the relative importance of key outcomes that may be inter-related. For instance, women with PPROM are more likely to have preterm babies who may in turn be of low birthweight. We could not perform multivariate analysis in our meta-analyses to explore such interactions between factors. Pooling of raw data from relevant studies in meta-analysis from individual women might help to clarify the causality of some observed associations. Moreover, for certain outcomes like preterm labour/delivery, most studies have given the overall risk of preterm labour/delivery and not made an attempt to distinguish between spontaneous labour or iatrogenic preterm delivery and therefore the association should be interpreted with caution.
Meaning of findings
Reasons for the association between first-trimester bleeding and adverse pregnancy outcomes are poorly understood. Bleeding in the first trimester may be associated with a chronic inflammatory reaction in the decidua. It is known that in about two-thirds of early pregnancy failures, there is evidence of defective placentation, characterised by thinner and fragmented trophoblast shell and reduced cytotrophoblast invasion of the spiral arterioles. Later pregnancy complications such as pre-eclampsia, preterm labour and PPROM have been shown to be associated with impaired placentation and failure of physiological invasion of the spiral arterioles. Problems with placental development may therefore explain why women with threatened miscarriage are more likely to have placenta praevia, placental abruption and APH of unknown origin.
Our data highlight the fact that first-trimester bleeding increases the risk of prematurity, growth restriction and perinatal deaths. While some of the incidences of prematurity can be linked to maternal complications such as APH, growth restriction suggests a degree of placental compromise.
Overall, our results suggest that pregnancies with first-trimester bleeding are at a higher risk for poor fetal and maternal outcome compared with women without bleeding. However owing to the risk of adverse outcomes being relatively modest (OR ≤2) and the lack of availability of any specific interventions to prevent these adverse events, it would be premature to suggest a policy of increased fetal and maternal surveillance.
Our review consists of six retrospective cohort studies and seven studies with a prospective cohort design as well as one case–control study. Prospective cohort studies are a more reliable way of establishing a causal association because retrospective designs are subject to recall bias. However, bearing in mind the significant cost (both financial and manpower) implications of implementing a programme of increased surveillance and the limitations of the studies included in the meta-analysis, perhaps what is needed are more prospective studies on women with and without vaginal bleeding in early pregnancy that are large enough to allow subgroup analyses based on gestation, severity and duration of bleeding to be performed with a degree of confidence. Another possibility is to aggregate raw data from existing studies to perform individual women data meta-analysis, which will permit adjustment for confounders and meaningful subgroup analyses.
In conclusion, the current meta-analysis reports that women with first-trimester threatened miscarriage are at increased risk of adverse maternal and perinatal outcome, although in the majority of women the risks are low (OR ≤2). As a consequence, in the interim, it would be rational to use the findings of our review to reassure women with first-trimester bleeding and at the same time alert clinicians for the signs of the possible complications.
Disclosure of interests
None of the authors report any conflict of interest or financial interest.
Contribution to authorship
L.S. prepared the protocol, collected data, assessed eligibility and methodological quality of studies and wrote the review. S.B. conceived the idea, conducted searches, assessed eligibility and quality of studies, and provided comments on the manuscript. A.M. performed the statistical analysis and S.B. conceived the idea, provided comments on the manuscript and supervised the review.