Rheumatic heart disease in pregnancy and neonatal outcomes: A systematic review and meta-analysis

Purpose Associations between rheumatic heart disease (RHD) in pregnancy and fetal outcomes are relatively unknown. This study aimed to review rates and predictors of major adverse fetal outcomes of RHD in pregnancy. Methods Medline (Ovid), Pubmed, EMcare, Scopus, CINAHL, Informit, and WHOICTRP databases were searched for studies that reported rates of adverse perinatal events in women with RHD during pregnancy. Outcomes included preterm birth, intra-uterine growth restriction (IUGR), low-birth weight (LBW), perinatal death and percutaneous balloon mitral valvuloplasty intervention. Meta-analysis of fetal events by the New-York Heart Association (NYHA) heart failure classification, and the Mitral-valve Area (MVA) severity score was performed with unadjusted random effects models and heterogeneity of risk ratios (RR) was assessed with the I2 statistic. Quality of evidence was evaluated using the GRADE approach. The study was registered in PROSPERO (CRD42020161529). Findings The search identified 5949 non-duplicate records of which 136 full-text articles were assessed for eligibility and 22 studies included, 11 studies were eligible for meta-analyses. In 3928 pregnancies, high rates of preterm birth (9.35%-42.97%), LBW (12.98%-39.70%), IUGR (6.76%-22.40%) and perinatal death (0.00%-9.41%) were reported. NYHA III/IV pre-pregnancy was associated with higher rates of preterm birth (5 studies, RR 2.86, 95%CI 1.54–5.33), and perinatal death (6 studies, RR 3.23, 1.92–5.44). Moderate /severe mitral stenosis (MS) was associated with higher rates of preterm birth (3 studies, RR 2.05, 95%CI 1.02–4.11) and IUGR (3 studies, RR 2.46, 95%CI 1.02–5.95). Interpretation RHD during pregnancy is associated with adverse fetal outcomes. Maternal NYHA III/IV and moderate/severe MS in particular may predict poor prognosis.


Introduction
The global prevalence of rheumatic heart disease (RHD) is 1%, and is twice as common in women than men, particularly in women of childbearing age [1,2]. This figure is likely underestimated in developing countries [2]. RHD accounts for approximately 30% of cardiac disease in pregnancy in developed countries, and 90% of cardiac disease in non-industrialized regions [3,4].
Normal hemodynamic changes of pregnancy impose an additional 30-50% cardiac load. This is well tolerated by a normal heart but can result in morbidity and mortality in women with pre-existing RHD [5][6][7]. Mitral stenosis (MS) is especially sensitive to cardiac insufficiency in pregnancy [8,9]. The placental-fetal heart circulation is likely affected [10], and hemodynamic insufficiency poses a risk to the developing fetus. Complications such as intrauterine growth restriction (IUGR) and prematurity may have lasting developmental effects into childhood and beyond [11].
The New York Heart Association (NYHA) functional classification of heart failure is used worldwide, with four categories (I-IV) based on limitations during physical activity; Class Ino limit, to Class IV-symptoms at rest [12]. In addition, MS severity can be graded using echocardiography based on mitral-valve area (MVA) into mild (>1.5cm 2 ), moderate (1.0-1.5cm 2 ) and severe (<1.0cm 2 ) [13]. Increasing severity of these indicators (NYHA, MVA) is associated with increased frequency of maternal cardiac complications [9]. In contrast, the association with adverse fetal and neonatal outcomes is often unreported.
The purpose of this study was to review rates of adverse fetal and neonatal outcomes for women with RHD in pregnancy and investigate the association between increasing severity of RHD using the NYHA and MVA scales with fetal outcomes. Additionally, the effects of percutaneous balloon mitral valvuloplasty (PBMV) on fetal events is reported.

Methods and analysis
This systematic review and meta-analysis is reported in accordance with the PRISMA guidelines [14], and registered with PROSPERO (CRD42020161529) [15].

Search strategy
An electronic search of Medline (Ovid), Pubmed, EMcare, Scopus, CINAHL, Informit, and WHO ICTRP was performed on 15 July 2020, limited to studies published in English language between 01 January 1990-15 July, 2020.
The complete search strategy (S1 Fig) used combined controlled vocabulary with free-text words related to population, intervention/exposure, and outcome (PICO). Studies were eligible for inclusion if they were conducted at a tertiary centre and reported associations between RHD in pregnancy and one or more pre-specified fetal outcomes. Studies with non-specific pregnancy-related cardiac disease, concordant congenital heart disease, isolated pulmonary or aortic valve involvement were excluded. Randomized controlled trials, intervention studies, cohort studies, case-control studies were eligible for inclusion. Case reports, case series, reviews, and duplicates were excluded.
Titles and abstracts were screened by the primary author (JL) on selection criteria. A second reviewer (BW) screened a sample until agreement reached >0.8 using Cronbach alpha [16]. For all selected articles, the full text were retrieved and evaluated by primary author and independent second reviewer (BW) for eligibility. In case of disagreements, a third reviewer was consulted (CH), and a decision agreed by consensus. Additional studies were identified from a manual search of references of included studies.

Type of outcome measures
Studies reporting one or more of the following outcomes were included: preterm birth (live delivery before 37 weeks gestation), low birth weight (LBW) (<2500 grams), small for gestational age (SGA) or intra-uterine growth restriction (IUGR) (estimated weight <10% percentile for gestational age), miscarriage (non-viable products of conception <20 weeks gestation) or perinatal death (including stillbirths (fetal demise after 20 weeks) and neonatal deaths (within the first 28 days of life)).

Data extraction and risk of bias assessment
Data were extracted into custom data collection forms by two independent reviewers (JL, BW). Authors were contacted for further information if required. Information extracted included: authors, setting, location, study design, study period and population characteristics (maternal age, gravida, parity, gestational age, rheumatic valvar lesions, mitral valve area severity, baseline NYHA classifications and mode of delivery).
Two authors (JL and BW) independently scored the risk of bias with a modified Quality in Prognostic Studies (QUIPS) tool [17] (S2 Fig). A risk of bias assessment was based on criteria for study participation, study attribution, prognostic factor measurement, outcome measurement, study confounding and statistical analysis. Each category was classified as low, medium, high or unknown risk of bias. Discrepancies were resolved by consensus or by a third reviewer (LH).

Statistical analysis
Rates of neonatal outcomes were recorded and compared based on maternal baseline NYHA status and MVA severity at the time of first antenatal visit. The effect of minimal invasive intervention (PBMV) during pregnancy on neonatal outcomes was narratively synthesised. An apriori decision was made to perform meta-analysis if sufficient data was available. Weight of the studies in the meta-analysis was calculated based on the Mantel-Haenszel test using Revman v5.4 [18]. The random effects model was chosen to account for inter-and intra-study variability [19,20]. Between-study heterogeneity was assessed using the I 2 test [21]. Risk ratios (RR) were reported with 95% confidence intervals (CI). Subgroup analysis was conducted based on country and non-PBMV/PBMV cohorts and sensitivity analysis was conducted with exclusion of outliers. Small study bias (including publication bias) was examined using funnel plots and Egger's test [22] if 10 or more studies were available, with statistical significance set at 10%.

Results
The initial database search identified 9719 papers. A further 4 papers were obtained from other sources. After removal of duplicates (n = 3774) and ineligible studies from title and abstract screening (n = 5813) and full text review (n = 114), 22 studies were included in the review (all cohort study designs) (Fig 1), and 11 in the meta-analysis (Table 1).

PLOS ONE
Rheumatic heart disease in pregnancy and neonatal outcomes: A systematic review and meta-analysis  [36], and higher perinatal deaths in those with NYHA III/IV who did not undergo PBMV vs those who did; (GA 37.15+/-1.06 vs 43.8+/-3.61, p = 0.002), and (0% perinatal death vs 19.4%, p = 0.08) respectively [36]. Lower preterm births in women who underwent PBMV during pregnancy was also seen in another study [29] but did not reach statistical significance. No meta-analysis was conducted for this outcome.
RHD was first diagnosed during pregnancy in 66.5% of patients in one study [32]. High rates were also seen in Australia (14.2%) [42], South Africa (42%) [28] and 24.9% in the ROPAC study [9]. Limited antenatal care in multiple studies [25,34,39,42] was associated with poor fetal outcomes and late optimisation of anticoagulants during pregnancy in select women.  One study [9] conducted adjusted analysis and found severe MS was independently associated with adverse fetal outcomes (OR 3.62, 95%CI 1.45-9.05), when adjusted for atrial fibrillation, severe mitral regurgitation, and anticoagulation during pregnancy. Pre-pregnancy NYHA>1 did not show univariate significance with adverse fetal outcomes (OR 1.10, 95% CI 0.59-2.02, p = 0.10), but was an independent predictor of maternal cardiac events in women with MS [9].
Funnel plot and Eggers test was not conducted as less than 10 studies were included per meta-analysis. The GRADE system rated the overall certainty of evidence as low for MVA and NYHA as markers for preterm, and very low for NYHA as markers of SGA/IUGR, LBW and perinatal death (Table 3). There was also low certainty for MVA and SGA/IUGR.

Discussion
Evidence from the 22 included studies suggest RHD in pregnancy is associated with high rates of adverse fetal outcomes (preterm birth, LBW, SGA, IUGR, miscarriage and perinatal death). Additional outcomes found high rates of NICU admissions [26,32,40] low rates of antenatal care and late diagnosis of RHD in many women.
On meta-analysis, both severe MS and NYHA III/IV were significantly associated with preterm birth. Additionally, NYHA III/IV were also associated with higher rates of perinatal death. A Nepalese study [34] appeared to be an outlier in reporting consistently higher rates of fetal adverse outcomes in their patient cohort. Post-hoc sensitivity analyses excluding this study lowered statistical heterogeneity and reduced the RR. This could indicate that this site in Nepal may have wider health care inequities (low health resources, health access, co-morbidities) compared to other developing countries (India, Egypt etc.).
The association between rheumatic MS during pregnancy and adverse fetal outcomes is biologically plausible. Early pregnancy is associated with a 30-40% increase in cardiac preload [43], decreased systematic vascular resistance and systolic blood pressure. These changes are poorly tolerated in MS and restricted left ventricular inflow with increasing atrial pulmonary pressures often precipitates cardiac decompensation and pulmonary edema [10]. Adverse fetal outcomes are likely due to uteroplacental insufficiency secondary to left heart obstruction [10]. Poor oxygen and nutrient transfer may lead to stunted fetal growth.
Mitral stenosis carries a high risk of chronic fetal hypoxia and early onset (<32 weeks) IUGR in pregnancy. These fetuses are more likely born preterm, and are high risk of rapid deterioration, fetal demise in-utero and stillbirth [44]. There are also recognised links of IUGR with cardio-vascular remodelling, sub-optimal renal and neurological development, and altered glucose metabolism; collectively known as the fetal origin hypothesis [45]. Such outcomes are currently unexplored in neonates born to mothers with RHD.
The prognostic value of NYHA classification for neonatal outcomes is likely a reflection of the severity of the pressure gradient across the mitral valve and underlying pulmonary edema. As such, it is well established in predicting maternal cardiac events, but less so for adverse fetal events. This review found NYHA class III/IV had significant associations with prematurity and perinatal death, but not LBW or IUGR/SGA. Conversely, mitral valve area (MVA) determined by echocardiogram could more directly indicate cardiac output and uteroplacental perfusion as moderate/severe MS was significantly associated with both SGA/IUGR and prematurity.
RHD remains the predominant form of maternal heart disease in pregnancy in developing nations [3,4]. In this study, developing countries [46] (India, Nepal, Egypt, South Africa) exhibited relatively higher rates of adverse neonatal outcomes compared to developed countries (Australia, New Zealand). Shortage of health services and delayed access to tertiary centres may be more evident in these developing nations, with further limited capacity of hospitals in surgical intervention [29] and neonatal intensive care [34].
Poorer education among women with RHD was reported in one study in Chandigarh, India [36]; 14% illiterate and 10.6% only receiving a primary school education. Downstream health behaviours associated with low education status, such as younger maternal age and multiparity are also predictors for adverse perinatal events [47].
Within-country variations in birth outcomes were observed in western developed nations. One study found higher rates of preterm and perinatal death in Aboriginal Australians or Torres Strait Islanders, and Maori or Pasifika mothers compared to non-Indigenous counterparts [42]. Indigenous mothers with RHD were significantly younger, [39,42] more likely to present >20 weeks to antenatal clinic, be socioeconomically disadvantaged, and smoke during pregnancy compared to non-Indigenous mothers [42]. While the disparity in fetal outcomes is likely a combination of these bio-psychosocial factors, there is evidence of an independent association of RHD in pregnancy. For example, in Australia, one study [42] reported an overall preterm birth rate of 21%; much higher than the overall rate of Australia (9%) and of babies born to Indigenous mothers (14%) [48] As such, closing the gap between health inequities among disadvantaged populations is a priority in the improvement of global neonatal health, and eradication of RHD among women of child-bearing age.
Antenatal care remains a critical component of neonatal outcomes in RHD patients [49]. Sub-optimal antenatal visits were common among studies with relatively higher adverse fetal events. In one study [25] over 90% of women presented for first time in labour and reported a high IUD rate (6.25%). In Durban (South Africa) [38] 62% had first cardiac evaluation in 3 rd trimester and had 42.97% prematurity deliveries. Greater emphasis on pregnancy planning, particularly after an index pregnancy would be beneficial.
Pregnancy planning and early initial antenatal consultation is also important for women with a surgical valve replacement and on lifelong anticoagulants such as warfarin. Delayed initial antenatal visits and low uptake of contraceptives [50] were reported among this sub-group of women in several studies [34,38]. This is concerning as warfarin is teratogenic and has strong associations with fetal malformation, abortion and stillbirth. Early optimisation with heparin or low-molecular weight heparin should be a priority in these patients [51]. PBMV remains the treatment of choice for isolated non-calcified MS and is safe to perform during pregnancy, with few adverse maternal or fetal events [52][53][54]. This systematic review suggests PBMV in women with severe symptoms (NYHA III/IV) is associated with reduced rates of preterm births; however further comparative studies are required. No long-term effects on child development have been reported to date [52,55,56].
Mitral valve surgery involving cardiac bypass was not assessed in this review as such interventions are avoided where possible during pregnancy due to the significantly high associated fetal mortality (ranging from 5-33%) [54,57].

Limitations
Neonatal outcomes are influenced by a complex interplay of known and unknown factors. RHD is associated with socio-economically disadvantage, and many important potentially confounding variables such as smoking, poor antenatal care, chronic disease [58] and extent of RHD-related antenatal services in hospitals of different countries, which were not measured in the included studies. Insufficient reporting on outcomes of women on warfarin during pregnancy also limited analysis of a clinically important sub-group.
Only studies based at tertiary hospitals were included, and it is likely that lower-resourced and rural areas experience even poorer pregnancy outcomes that are underreported.
There is some methodological limitation in this research. First, most studies in this review had moderate risk of bias in multiple domains. In particular, outcome definitions were not clearly explained, and influence of confounders was also uncertain. Second, there is a possibility of publication bias among studies with small samples, particularly in outcomes around perinatal death. Overestimation of the true effect size may have resulted from smaller studies with

PLOS ONE
non-significant findings not being published. Third, there was clinical and statistical heterogeneity between studies. The overall certainty of the evidence generated from meta-analysis was low or very low (Table 3), although the GRADE system allows a maximum of low-quality evidence for meta-analysis of cohort studies [59].
These findings are important from a national, international, public health policy perspective, highlighting increased perinatal morbidity and mortality in infants born to women with RHD. As our results indicate that moderate or severe MS, symptomatic NYHA, have worse outcomes, we recommend early specialist involvement in these cases. While no definitive management for IUGR exists besides delivery, neonatal USS Doppler could be useful for early identification. PBMV in pregnancy is an effective, low risk procedure for symptom relief in MS during pregnancy but requires further research. Finally, our findings add support to large scale echocardiographic screening of RHD in pregnancy in high-risk populations.
Large, well-designed prospective studies of pregnancy in women with RHD are required. Associations with NYHA and MVA severity on neonatal outcomes need to be calculated based on adjusted rates. One study [9] in this systematic review demonstrated robust methodology that could be modelled in future studies.  � Meta-analysis of observational studies has an initial confidence estimate of 'Low confidence' [43]. No outcome was rated up in confidence as analysis did not yield a sufficiently large effect size.

Supporting information
A-The evidence was kept as low after sensitivity analyses (95%CI 1.56-3.64, I 2 = 12%) removing an outlier study.
B-The evidence was downgraded to very low due to the significant difference in clinical implications of the lower and upper limits of the overall CI (0.84-2.80). C-The evidence was downgraded as the CI is wide and crosses no significant effect CI (0.98-3.10).
D-The evidence was downgraded to very low as perinatal death did not specify between women with PBMV vs non-PBMV, or women on anticoagulant. Both factors have shown to significantly influence maternal mortality (and fetal death) [44,45]. https://doi.org/10.1371/journal.pone.0253581.t003