Study Site

Moi Teaching and Referral Hospital (MTRH) is one of only two national referral hospitals in Kenya. The standalone maternity hospital at MTRH performs over 12,000 deliveries annually. Cardiovascular care and research are organized through the infrastructure of a National Heart, Lung and Blood Institute (NHLBI)-funded Center of Excellence [1]. Ethical approval for the study was obtained from both the Institutional Review and Ethics Committee (IREC) at the Moi University School of Medicine and the University of Toronto.

Study Design

This was a retrospective case-control study of pregnant women hospitalized at MTRH between January 2011 and April 2016. Cases included all women with a known cardiac diagnosis prior to or during the pregnancy period who were admitted to the hospital for any reason during pregnancy, or up to 6 weeks postpartum. Cases were identified using ICD-9 and ICD-10 codes from inpatient admissions, echocardiogram orders and pharmacy records. Cases were matched to pregnant controls without any known history of cardiac disease based on age (+/– 2 years) and parity (+/– 2). Controls were excluded if insufficient data were able to be located from the medical record.

Medical charts were reviewed by two data collectors (JA, RL). Information pertaining to the incident pregnancy and cardiac disease were obtained from provider notes, standardized labor and delivery forms, and transthoracic echocardiogram (TTE) and electrocardiogram (ECG) reports. All cardiac diagnoses listed during the clinical encounter were recorded. Objective data from the first medical encounter during pregnancy were collected if present, including: blood pressure, heart rate, oxygen saturation, New York Heart Association (NYHA) class of heart failure, and TTE and ECG findings. Left ventricular ejection fraction (LVEF), right ventricular systolic pressure (RVSP), and type and severity of valvular lesions were obtained from hospital TTE reports. Left ventricular ejection fraction (LVEF) was recorded as reported using the following categories: normal (>55%), mild dysfunction (45–55%), moderate dysfunction (30–44%), or severe dysfunction (<30%). All ECGs were interpreted by a single cardiologist (FB). Cases with discrepancies between diagnoses were reviewed by two cardiologists (FB, GSB). Those without objective cardiac disease were excluded from the study.

We generated a standardized definition of RHD based on adapted criteria from the World Heart Federation and data available from TTE reports [9]. We defined RHD as: any mitral stenosis (MS) or aortic stenosis (AS); presence of both mitral regurgitation (MR) and aortic regurgitation (AR); or isolated MR or AR, if moderate or severe. All cases under 20 years old or with equivocal lesions were individually reviewed by two cardiologists (FB, GSB) to determine inclusion using World Heart Federation criteria for borderline RHD [9].

Our primary outcome was maternal death during pregnancy or up to 6 weeks postpartum. Secondary outcomes included: cardiac adverse events (cardiac arrest, new arrhythmia, heart failure, myocardial infarction, stroke, endocarditis, or admission to the cardiac intensive care unit [CCU]), obstetric adverse events (Cesarean-section, vacuum-assisted delivery, induction of labor, postpartum hemorrhage, pre-eclampsia/eclampsia, or venous thrombosis event), or neonatal adverse events (intrauterine fetal demise (IUFD, fetal death ≥28 weeks gestational age). Other secondary outcomes included: neonatal death (<28 days after live birth), preterm delivery (<38 weeks gestational age), low birth weight (<2500 grams), admission to the intensive newborn unit, or APGAR score <7 at 1, 5 or 10 minutes).

Statistical Analysis

A minimum of 145 participants were required in each arm to achieve 80% power in detecting a difference in the primary outcome, using the maternal mortality rate at MTRH of 0.3% and an estimated rate of 6% among cases based on literature [2, 3, 16, 23]. Controls were matched to cases (2:1) in order to detect differences in secondary maternal and neonatal outcomes. Missing data points for objective measures pertaining to the incident pregnancy were dropped, whereas missing historical medical diagnoses were assumed to be “no.” Univariate analyses were performed using the Kruskal-Wallis test and Fisher’s exact test for continuous and categorical variables respectively.

We applied both the CARPREG I score and mWHO classification to estimate the risk of adverse events in pregnancy [8, 13, 20]. A CARPREG I score of 0–4 was generated for each case, with one point each for: prior cardiac event (history of heart failure, stroke, or arrhythmia before pregnancy) NYHA class III or IV or oxygen saturation <80% at presentation; left ventricular inflow and outflow obstruction (mitral valve area (MVA) < 1.5 cm² or severe aortic stenosis); or reduced left ventricular systolic function (LVEF < 45% in our sample) [8]. We adapted the mWHO classification of maternal cardiovascular risk (class I–IV), as outlined in the European Society of Cardiology guidelines, as follows: [6, 23] mWHO class I included repaired simple, congenital lesions; class II included all unrepaired congenital cases and any arrhythmia on ECG; class III included mild LVEF impairment (LVEF < 45%) and any valvular heart disease not otherwise considered class I or IV. Class IV included pulmonary hypertension of any cause (RVSP by echocardiogram > 35 mmHg), severe systemic ventricular dysfunction (LVEF < 30% or NYHA class III or IV), previous peripartum cardiomyopathy with any residual LV impairment, severe mitral stenosis, or severe, symptomatic aortic stenosis. Cases were categorized into the higher class if multiple features were present. Logistic regression was performed to calculate the observed risk of cardiac and neonatal outcomes based on number of predictor variables for both CARPREG I and mWHO scores, independently. Analyses were carried out using STATA (Version 15, College Station, TX: Stata Corporation).

Demographics

A total of 97 cardiac cases in pregnancy were identified and matched to 242 controls. Median age of cardiac cases was 26 years compared to 28 for controls (p = 0.007, Table 1). There was no difference in gravidity or parity between groups. Nearly two-thirds (63.9%) of cardiac cases had a previous delivery, while 30% had three or more. Almost all participants had at least one routine antenatal visit. There was no significant difference in rates of comorbid conditions between groups, including hypertension, diabetes, HIV, pulmonary disease or renal disease (9.3% vs. 4.6%, p = 0.128).

Clinical profile of cardiac disease in pregnancy

Acquired causes of cardiac disease were present in 90 of 97 cases (92.3%), while electrical and congenital causes were less common (15.5% and 7.2% respectively, Table 2). Most cases (75.3%) had RHD, of which half had mitral stenosis, and one-third of those were severe. Pulmonary hypertension, defined as RVSP ≥ 35 mmHg on TTE, was present in 61% of cases, with a median RVSP of 64mmHg (IQR 39–120). Only 10% of cases had an LVEF < 45%. Of the 57 cases with NYHA class recorded at the initial encounter, 35 (61.4%) were class III or IV.

Only half of cases (53.8%) were diagnosed with cardiac disease prior to the incident pregnancy (Table 2). Of the 43 cases diagnosed during pregnancy, the majority (n = 37, 86.0%) were diagnosed late in the 3^rd trimester, or after delivery.

Outcomes

The maternal mortality rate among cardiac cases was 9.3%, with no deaths in the control group (p = 0.001, Table 3). Half of all maternal deaths had severe mitral stenosis (n = 4) and nearly all had pulmonary hypertension (n = 7). Despite the severity of disease among these cases, four of the nine were undiagnosed prior to the incident pregnancy. Six of the nine maternal deaths happened in the postpartum period, with four occurring more than 14 days after delivery (Table 4).

Nearly all cardiac cases (79.8%) had at least one cardiac, obstetric, or neonatal adverse event related to pregnancy, as compared to only 25% of controls (p < 0.001, Table 3). Congestive heart failure was the most common cardiac event, occurring in 42.3% of cases, followed by new arrythmia in 7%, and 2% had a stroke. Nearly one-quarter (21.7%) of all cases were admitted to the cardiac ICU at least once during the pregnancy period. Significantly higher rates of neonatal events were seen among cardiac cases as compared to controls, including: intrauterine fetal demise (11.2% vs 3.0%, p = 0.009), low birth weight (30.4% vs 7.8%, p < 0.001), and preterm delivery (41.6% vs 10.8%, p < 0.001). Cases had higher rates of induction of labor (37.0% vs. 10.3%, p < 0.001) and vacuum-assisted deliveries (19.8% vs. 0%, p < 0.001) compared to controls, but there was no difference in C-section rates between groups.

Prediction Scores

Observed rates of adverse events were higher than predicted by the CARPREG I and mWHO risk scores. Cardiac events occurred in 29%, 48%, and 81% of cases with CARPREG I scores of 0, 1, and ≥2 respectively, as compared to expected rates of 5%, 27%, and 75% (Figure 1) [13]. Among those with a CARPREG I score of 0, 59% had some adverse event (cardiac, obstetric or neonatal Figure 2). A CARPREG I score ≥2 carried a 100% rate of any adverse event.

Similarly, high rates of adverse events occurred among cases with low mWHO class. Of those with mWHO class I, 27% experienced an adverse cardiac event (Figure 1). Observed rates of cardiac events in these groups were similar to predicted from existing data in emerging countries for mWHO class I (27% vs. 26%) and class IV (67% vs. 57%) (20). However, most cases were classified as either mWHO class I (23%) or mWHO class IV (68%), with very few cases classified as class II or III (1% and 7%, Table 5).

Outcomes

Maternal mortality among cardiac cases was nearly 10-times higher than all-cause mortality among pregnant women in Kenya. With up to one-third of maternal deaths in western Kenya attributable to non-obstetric causes, our findings indicate that cardiac disease may be a significant, under-recognized threat to persistent maternal mortality in this region [25, 26, 27, 28]. Notably, half of maternal deaths occurred in the late postpartum period, greater than 14 days after delivery, suggesting that the risk of significant adverse events extends beyond the traditional timeframe of routine hospitalization following obstetric delivery. Similar results were found in a prospective cohort of South African women with cardiac disease, where the highest rates of maternal death occurred between 44–150 days postpartum [16]. Thus, close postpartum monitoring and follow up may be as critical as early antenatal care in this population.

Risk Prediction

Understanding and quantifying risk of cardiac disease in pregnancy is critical to pregnancy planning and early management strategies, especially in low-resource settings where surgical or other invasive interventions for cardiac disease are limited, but existing tools are inadequate. We found that the CARPREG I and mWHO risk prediction tools underestimated adverse events in our population, with high rates of cardiac events despite low CARPREG I and mWHO risk scores. Previous attempts to validate the CARPREG I score in other LMICs have demonstrated both over- and underestimation of risk of adverse events, likely due to the unique spectrum of disease specific to LMICs as compared to the North American cohort from which the CARPREG I score was derived [12, 14]. Over half of our cases had pulmonary hypertension, a relative contraindication to pregnancy with significant maternal and neonatal risk, which is not captured in the CARPREG I score, and thus could account for underestimate of risk [8, 14, 17]. The mWHO risk score, which includes more disease-specific cardiac lesions, is still regionally limited. Sub-analysis from the ROPAC registry reveals that the mWHO risk score was still less effective in predicting cardiac events among women in LMICs compared to high income countries, and was particularly poor for women with acquired heart disease in LMICs [20].

Both scores fail to capture unique barriers present in low-resource settings that may be contributing to poor outcomes, including late presentation to routine, antenatal care, lack of adequate diagnostic technologies, and limited access to coordinated, subspecialized care [12, 20]. While the severity of the underlying cardiac lesions remains the major driver of poor outcomes in pregnancy, late presentation to care and delayed diagnosis likely exacerbate the risk of maternal and neonatal complications in our population. Nearly all of our cases attended an antenatal clinic, and most had at least one prior pregnancy; yet, over one-third of cases were first diagnosed with cardiac disease after the second trimester, suggesting that identification of cardiac disease within routine antenatal care is low.

The CARPREG II score was recently derived from and validated in a Canadian population and found that late pregnancy assessment and lack of cardiac intervention prior to pregnancy were independent risk factors for cardiac event in pregnancy [12]. This newer score now includes these as weighted, predictor variables, suggesting that poor access and late presentation to antenatal care, on top of disease type and severity, are significant risk factors to adverse cardiac events in pregnancy in Kenya. Therefore, clinical strategies for earlier detection of cardiac disease during routine antenatal care, strengthened postpartum follow-up, and improved risk prediction scores are needed in the sub-Saharan population.

Limitations

Overall, we found fewer cardiac cases than we anticipated based on the known burden of cardiac disease in the region. This may be due to the retrospective study design, which resulted in a smaller sample than expected in which to detect a statistical difference in the primary outcome between groups. Despite this, our data highlights the high absolute maternal death rate among women with cardiac disease. Given that we included only women hospitalized at a tertiary care facility, selection bias may contribute to overestimation of adverse events, and may affect the application of risk prediction models. Inconsistent record keeping from paper medical charts resulted in the potential for missing data, which we attempted to correct for by imputing historical medical data. However, it could have led to underestimation of outcomes among the control group. Similarly, this likely contributed to small subgroup sizes in risk prediction modeling (ex: few cases of mWHO class II and III) and limited use of the newer CARPREG II score in this population. Lastly, there were statistical differences between cardiac and non-cardiac cases with regards to both age and parity due to our phased approach to matching. However, these small statistical differences are unlikely to have clinically meaningful influences on our results. While these limitations suggest more robust, prospective studies need to be conducted to better guide our clinical management of these high-risk patients, this study provides critical information as one of the first to attempt to characterize the impact of cardiac disease in pregnancy in East Africa.