Late Gestation Predictors of a Postnatal Biventricular Circulation after Fetal Aortic Valvuloplasty

Abstract

Objectives:

Fetal aortic valvuloplasty (FAV) for severe aortic stenosis (AS) in mid-gestation has shown promise in averting progression to hypoplastic left heart syndrome (HLHS). After a technically successful FAV, predicting prenatally which fetuses will achieve a biventricular circulation after birth remains a challenge. Identifying predictors of postnatal biventricular circulation on late gestation fetal echocardiography will improve our understanding of the disease and parental counseling.

Methods:

Liveborn patients who underwent FAV and had a late gestation fetal echocardiogram available were included (2000-2017, n=96). Multivariable logistic regression and classification and regression tree analysis were utilized to identify independent anatomic and physiologic predictors of biventricular circulation on late gestation echocardiogram.

Results:

Among 96 fetuses, 50 (52.1%) had BiV circulation at the time of neonatal discharge. On univariate analysis, anatomic parameters associated with postnatal biventricular circulation included a larger z-score for: aortic valve diameter, ascending aorta diameter, left ventricular (LV) long axis, and mitral valve diameter (p≤0.001 for all). Physiologic parameters included higher LV ejection fraction (p=0.001), longer mitral valve inflow time (p=0.006), anterograde flow around the aortic arch (p<0.001), and bidirectional or right-to-left flow across the foramen ovale (compared to left-to-right flow, p<0.001). In multivariable analysis, independent predictors of biventricular circulation included LV long axis z-score (OR 3.2, 95% CI 1.8-5.7, p<0.001), LV ejection fraction (OR 1.3 per 5% increase, 95% CI 1.0-1.8, p=0.023), anterograde aortic arch flow (OR 5.0, 95% CI 1.2-20.4, p=0.024), and bidirectional or right-to-left foramen ovale flow direction (OR 4.6, 95% CI 1.4-15.8, p=0.015).

Conclusion:

Late gestation independent predictors of a postnatal biventricular circulation after FAV include LV long axis z-score, LV ejection fraction, anterograde flow around the aortic arch, and bidirectional or right-to-left flow across the foramen ovale. Identifying these predictors adds to our understanding of LV growth and hemodynamics after FAV and may improve parental counseling.

Introduction

Natural history studies have demonstrated that mid-gestation fetuses with moderate to severe aortic stenosis (AS) and a normal-sized or dilated left ventricle (LV) often progress to HLHS by birth. This progression occurs particularly in those with LV systolic dysfunction, retrograde aortic arch flow, and left-to-right flow at the foramen ovale (FO).^1-6 Since the 1990s, fetal aortic valvuloplasty (FAV) has been performed in an attempt to halt the evolution of AS to HLHS in utero.^7,8 Technically successful FAV results in improved left heart dimensions and LV function, and a subset of fetuses undergoing FAV achieve biventricular (BiV) circulation.^9-12 Rates of technical success have improved since the initial era of FAV, and selection criteria have undergone iterative refinement.^4,13 Overall, ~45% of fetuses who underwent FAV at our institution achieved BiV circulation postnatally, with a higher proportion of patients achieving BiV circulation (59%) in the more recent era (2009-2017).¹³

A prior study from our institution described changes in left heart hemodynamics when comparing fetal echocardiograms pre-intervention and 2 months post-intervention.¹⁰ Technically successful FAV resulted in an increased likelihood of biphasic mitral valve inflow, bidirectional flow across the FO, anterograde flow in the transverse aortic arch, and improvement in LV ejection fraction. Due to a small sample size (26 fetuses), there was insufficient power to detect late gestation echocardiographic findings that were associated with BiV circulation postnatally. It is not yet known which echocardiographic parameters in late gestation are associated with BiV circulation after birth. We aim to determine whether there are anatomic and physiologic parameters in late gestation that are associated with neonatal BiV circulation.

Methods

Patients

We reviewed the records of all fetuses with AS in which FAV was attempted at our institution from March 2000 to June 2017. Patient selection criteria have evolved and undergone revision over time as our experience has grown. Refinement in the selection criteria has focused on identifying patients at high risk of HLHS, but with a salvageable LV that might support systemic circulation after FAV.^9,11

The technical aspects of FAV, procedural complications, and factors impacting the technical success of the procedure, have been previously described.^11,14-16 A technically successful FAV was defined as positioning and dilation of the balloon across the aortic valve with subsequent improvement in anterograde flow.

FAV was attempted in 138 consecutive patients with evolving HLHS from March 2000 to June 2017, and was technically successful in 115 (Figure 1). Subjects were excluded if the neonatal circulation was indeterminate (due to fetal or neonatal demise) or if a late gestation echocardiogram was not available for review. Twelve patients were excluded from analysis after fetal demise (n=10) or termination of pregnancy (n=2). Additionally, five neonatal deaths were excluded, three due to prematurity and one transitioned to comfort care after birth. Fourteen patients were excluded from the study due to lack of available late gestation echocardiograms, among whom 6 underwent technically successful FAV, and 2 of these patients had BiV circulation. Among the 107 patients who had complete echocardiographic data available for review, 96 patients had a technically successful FAV and constituted the cohort for analysis in this study.

Figure 1:

Study population. Patients were excluded from analysis if they did not survive to neonatal discharge, if there was no late gestation fetal echocardiogram available for review, or if FAV was technically unsuccessful. FAV, fetal aortic valvuloplasty.

Patients were categorized as having single ventricle or BiV circulation based on their status at the time of neonatal hospital discharge. BiV circulation was defined as one in which the entire systemic output was ejected from the LV. Single ventricle (SV) circulation was defined as requiring a palliative procedure (such as Stage I or hybrid operation) to utilize the right ventricle as the systemic ventricle. Neonatal management was determined by the providers at the birth institution, without a standardized postnatal algorithm. Of note, six patients who were initially discharged with SV circulation subsequently underwent conversion to BiV circulation. These patients were classified as having SV circulation in our analysis, based on their circulation outcome at neonatal discharge. One patient with BiV circulation at neonatal discharge was subsequently converted to SV circulation. Two fetuses underwent a second FAV prior to the late gestation echo included in the analysis.

Echocardiography

Fetal echocardiograms were performed at our institution using commercially-available ultrasound equipment within one day prior to FAV, as well as 24 to 48 hours after FAV. Subsequent fetal echocardiograms were obtained approximately monthly by the referring fetal cardiologist. Late gestation follow-up echocardiograms were collected, and the last fetal echocardiogram obtained before birth was analyzed for this study. All echocardiograms were reviewed by a single investigator blinded to the neonatal circulation of the subject. The Institutional Review Boards at Boston Children’s Hospital and Brigham and Women’s Hospital approved this retrospective study with a waiver of informed consent.

The anatomic parameters selected for study included diameter measurements of the aortic valve, ascending aorta, and mitral valve, LV long and short axis dimensions, LV end diastolic volume, and LV sphericity. Institutional normative data were used to calculate z-scores indexed to gestational age. Physiologic parameters assessed included LV ejection fraction (LVEF) using the 5/6^th area-length method, mitral valve inflow time and inflow pattern, estimated LV pressure (estimated using aortic stenosis maximum instantaneous gradient or mitral regurgitation jet velocity, as available), qualitative degree of LV systolic dysfunction (< moderate or ≥ moderate), direction of flow across the foramen ovale (left-to-right, bidirectional, or right-to-left), and direction of systolic flow in the transverse aortic arch (anterograde or retrograde).

Statistical Methods

Analysis was conducted on the cohort of liveborn patients who underwent technically successful FAV, had a late gestation echocardiogram available for review, and survived to neonatal discharge. Data were reported as median (interquartile range) for continuous variables or frequency (%) for categorical variables. Where median comparisons are presented, a Wilcoxon rank sum test was performed, and for comparisons of categorical variables, a Fisher exact test was performed.

Univariate logistic regression was utilized to identify associations with biventricular circulation. To identify factors that are independently associated with biventricular circulation, we performed multivariable logistic regression modeling. The predictors with a p-value < 0.2 in Tables 2 and 3 were included as candidates with stepwise selection. Classification and regression tree (CART) analysis was used to identify potentially important interactions that may describe high-association subgroups. Each parameter listed in Tables 2 and 3 was examined and recursive partitioning was utilized through a series of dichotomous splits to maximize the sensitivity and specificity of classification, resulting in a decision tree. The classification tree is automatically developed to forecast BiV/SV circulation by considering every possible cut point on every independent predictor at every node.

Table 2.

Late Gestation Anatomic Parameters by Circulation Outcome

Variable	Overall (n=96)	Biventricular (n=50)	Single Ventricle (n=46)	p-value
AoV diameter z-score	−2.95 (−3.67,−2.19)	−2.82 (−3.09, −1.94)	−3.42 (−4.20, −2.70)	<0.001
AscAo diameter z-score	0.01 (−1.27, 1.30)	0.71 (−0.63, 1.91)	−1.02 (−1.65, 0.02)	<0.001
MV diameter z-score	−2.60 (−3.36, −1.25)	−1.84 (−2.93, −0.86)	−3.14 (−3.81, −2.13)	<0.001
LV long axis z-score	−1.38 (−2.45, −0.47)	−0.80 (−1.48, 0.44)	−2.42 (−3.13, −1.16)	<0.001
LVEDD z-score	−0.65 (−2.18, 0.93)	−0.05 (−1.84, 1.26)	−1.23 (−2.35, 0.14)	0.027
LVEDV z-score	−1.55 (−3.30, −0.05)	−0.70 (−2.05, 0.84)	−2.92 (−3.54, −1.13)	<0.001
LVESV (ml)	2.04 (1.23, 3.85)	2.32 (1.46, 4.66)	1.69 (1.05, 3.30)	0.056
LV sphericity	0.63 (0.53, 0.74)	0.60 (0.47, 0.73)	0.67 (0.59, 0.74)	0.069

Data presented as median (IQR). AoV, aortic valve; AscAo, ascending aorta; MV, mitral valve; LV, left ventricular; LVEDD, left ventricular end diastolic dimension; LVEDV, left ventricular end diastolic volume; LVESV, left ventricular end systolic volume.

Table 3.

Late Gestation Physiologic Parameters by Circulation Outcome

Variable	Overall (n=96)	Biventricular (n=50)	Single Ventricle (n=46)	p-value
LV ejection fraction, %	41.1 (30.0, 54.4)	48.6 (36.8, 57.6)	36.8 (26.1, 44.8)	<0.001
LV dysfunction (≥ moderate)	63 (65.6%)	24 (48%)	39 (86.7%)	0.001
MV inflow time (msec)	179 (153, 200)	184 (158, 200)	169 (147, 202)	0.293
MV inflow time z-score	−1.01 (−1.76, −0.35)	−0.87 (−1.61, −0.37)	−1.34 (−1.95, −0.26)	0.393
MV inflow pattern				0.037
Biphasic	68 (81.9%)	40 (88.9%)	28 (73.7%)
Partial fusion	5 (6.0%)	0 (0%)	5 (13.2%)
Monophasic	10 (12.0%)	5 (11.1%)	5 (13.2%)
Aortic stenosis peak gradient (mmHg)	29 (10, 42)	30.0 (20, 45)	22.5 (4, 40)	0.069
LV pressure estimate (mmHg)	68.47 (46.14, 80.0)	70.78 (54.71, 86.71)	59.81 (39.64, 74.50)	0.022
Mitral regurgitation (≥ moderate)	15 (15.8%)	10 (20.0%)	5 (11.4%)	0.28
Aortic regurgitation (≥mild)	21 (22.8%)	13 (26.5%)	8 (18.6%)	0.46
Anterograde aortic arch flow	33 (34.7%)	26 (52.0%)	7 (15.6%)	<0.001
FO flow (bidirectional or right-to-left)	32 (38.6%)	25 (59.5%)	7 (17.1%)	<0.001

Data presented as n (%) or median (IQR). LV, left ventricular; MV, mitral valve; AoV, aortic valve; FO, foramen ovale.

A p-value of less than 0.05 was considered statistically significant. Analyses were performed using SAS version 9.4 (SAS Institute, INC., Cary, NC) and R 3.5.1.

Results

The demographic features of the 96 patients in the study cohort are listed in Table 1. Of the 96 fetuses included in the study, 50 (52.1%) had BiV circulation at the time of neonatal discharge. Late gestation fetal echocardiograms were performed at a median of 36 weeks (interquartile range [IQR] 34.7, 36.9 weeks). The anatomic parameters assessed on the late gestation fetal echocardiogram are shown in Table 2, both for the overall cohort and by postnatal circulation outcome. In univariate analysis, larger mitral valve, aortic valve, and ascending aorta diameter z-scores were associated with BiV circulation. Additionally, larger LV long axis z-score (p<0.001), end diastolic dimension z-score (p=0.027), and end diastolic volume z-score (p<0.001) were associated with BiV circulation.

Table 1.

Demographic Characteristics of Study Cohort and Association with Postnatal Circulation

Variable	Overall (n=96)	Biventricular (n=50)	Univentricular (n=46)	p-value
Male fetal sex	76 (79.2%)	39 (78%)	37 (80.4%)	0.806
Gestational age at FAV, weeks	24.1 (22.8, 26.9)	25.6 (23.0, 27.6)	23.5 (22.0, 25.1)	0.002
Gestational age at LGE, weeks	36.0 (34.7, 36.9)	36.2 (35.0, 37.1)	35.5 (33.9, 36.8)	0.167
Intervention Era				0.005
2000-2005	22 (22.9%)	5 (10.0%)	17 (37.0%)
2006-2011	33 (34.4%)	18 (36.0%)	15 (32.6%)
2012-2017	41 (42.7%)	27 (54.0%)	14 (30.4%)

Data presented as n (%) or median (IQR). FAV, fetal aortic valvuloplasty; LGE, late gestation echo.

Several physiologic parameters were also associated with BiV circulation after FAV (Table 3). Better LV function (higher ejection fraction or a less than moderate degree of qualitative systolic dysfunction), biphasic mitral valve inflow pattern, anterograde flow within the transverse aortic arch, and bidirectional or right-to-left flow through the FO (compared to left-to-right flow) were associated with BiV circulation. Mitral valve inflow time z-score was not significantly different between groups. In multivariable analysis with both anatomic and physiologic parameters as candidate predictors, larger LV long axis dimension z-score, higher LVEF, bidirectional or right-to-left FO flow, and anterograde aortic arch flow were independently associated with BiV outcome (c-statistic 0.91, R²=0.61; Table 4). LV long axis dimension z-score was the most significant variable, with an adjusted odds ratio for biventricular circulation of 3.2 per unit increase in z-score (95% CI 1.8-5.7, p<0.001).

Table 4.

Multivariable Model for Biventricular Circulation in Liveborn Neonates Following Technically Successful Fetal Aortic Valvuloplasty (n=88, 46 with biventricular circulation)

Variable	OR	95% CI	p-value
LV long axis dimension z-score	3.2	1.8-5.7	<0.001
LVEF, % (per 5% increase)	1.3	1.0-1.8	0.023
Bidirectional (or right-to-left) PFO flow	4.6	1.4-15.8	0.015
Anterograde aortic arch flow	5.0	1.2-20.4	0.024

c-statistic = 0.91, R² = 0.61. LV, left ventricular; LVEF, left ventricular ejection fraction; PFO, patent foramen ovale; OR, odds ratio; CI, confidence interval.

Classification and regression tree (CART) analysis for BiV circulation is shown in Figure 2. LV long axis dimension threshold of z > −2.3 was the strongest binary discriminator of biventricular vs. single ventricle circulation, with a sensitivity of 94.0%, specificity of 60.9%, and accuracy of 78.1%. Within the subgroup with normal or near-normal LV long axis dimension (z > −2.3), the other factors with the highest discriminating power for BiV outcome were aortic valve diameter (cm) and mitral valve inflow pattern. Patients with the highest likelihood of BiV circulation were those with a normal or near-normal LV long axis dimension, aortic valve diameter ≥ 0.45 cm, and biphasic mitral valve inflow pattern (88%, 36/41). The CART models using all four branches shown in Figure 2 had a sensitivity of 72%, specificity of 89%, positive predictive value of 88%, negative predictive value of 75%, and accuracy of 80%.

Figure 2.

Classification and regression tree analysis for prediction of biventricular circulation outcome; sensitivity 72%, specificity 89%, accuracy 80%. LV, left ventricular; MV, mitral valve.

Discussion

In this study, we identify late gestation anatomic and physiologic echocardiographic predictors of BiV circulation after FAV. Most notably, LV long axis dimension z-score, LV ejection fraction, anterograde transverse aortic arch flow, and bidirectional or right-to-left flow across the FO were independent predictors of biventricular circulation. Our study adds to the understanding of the evolution of left ventricular growth and improved hemodynamics after FAV and delineates features in late gestation that signify a favorable prognosis towards biventricular circulation.

Prior studies have described the pre-FAV echocardiographic predictors of BiV circulation, contributing to the refinement of the selection criteria over time.^4,9,13 Patient selection continues to focus on identifying fetuses that (1) are at high risk of progression to HLHS and (2) have features demonstrating potential for LV recovery after FAV. Iterative improvement in selection criteria has increased the proportion of patients undergoing FAV who achieve BiV circulation from 23% (2000-2005) to 66% (2012-2017). The aim of FAV is to relieve aortic valve stenosis, thereby decreasing the pressure load on the LV, improving diastolic function, and increasing right-to-left flow across the foramen ovale to increase preload into the LV. Prior investigation of physiologic changes within 24-48 hours after FAV demonstrated that features indicative of increased forward flow through the left heart, such as anterograde systolic flow through the transverse aortic arch and improved bidirectional flow across the FO, were associated with neonatal BiV circulation.¹⁷ Our results demonstrate that similar features in late gestation were strongly associated with BiV circulation. Additionally, we found an association of both larger LV long axis dimension z-score and higher LV ejection fraction with BiV outcome. These findings support the physiologic hypothesis that successful relief of outflow tract obstruction allows increased flow through the left heart, promoting longitudinal growth of an LV that will be capable of supporting systemic cardiac output after birth. Our results corroborate the findings of Kovacevic et al., who demonstrated that fetuses who had sustained improvement in hemodynamic parameters such as anterograde arch flow, biphasic mitral valve inflow, and reversal of left-to-right foramen flow through gestation had a higher likelihood of BiV circulation after birth.¹⁸

By delineating the late gestation anatomic and physiologic characteristics associated with a postnatal BiV outcome, we improve parental counseling with regards to prognosis and may contribute to decision-making regarding the initial postnatal management strategy. Prior to undergoing FAV, each family must be extensively counseled regarding the risks and potential benefits of this intervention. Several prior studies have contributed to the current body of literature informing patient selection and supporting an individualized risk-benefit assessment for each fetus.^4,9,12,13 By applying our CART decision tree and results from the multivariable analysis to a fetus in late gestation, we will be able to stratify patients into those with high or low likelihood of achieving BiV circulation. This insight will allow parental counseling to be tailored to the individual patient and to guide parental expectations.

There are several limitations of this study. Our primary outcome variable, circulation at the time of neonatal discharge, is a short-term measure. A number of patients who underwent FAV at our institution delivered at their local centers, and individual variations in management strategy of borderline cases may have impacted their circulation outcome. Some patients who underwent single ventricle palliation may have been able to support BiV circulation. Alternatively, other patients who underwent multiple catheterizations and surgeries to maintain a BiV outcome may have had more optimal circulation with stage I palliation. We acknowledge that decision-making for neonates with a borderline left heart is challenging and should not be clouded by the antecedent fetal cardiac intervention. Decisions in the neonatal period were made at each institution using their best judgment.

Conclusion

In fetuses that have undergone technically successful FAV, larger LV long axis, higher LV ejection fraction, anterograde aortic arch flow, and right-to-left or bidirectional FO flow on late gestation echocardiography were independently associated with BiV circulation after birth. Further investigation into the association between fetal parameters and long-term outcomes after FAV will improve parental counseling and may inform decision-making regarding postnatal management strategies.

What’s already known about this topic?

• Midgestation fetuses with aortic stenosis often progress to hypoplastic left heart syndrome by birth, requiring single ventricle palliation.

• Fetal aortic valvuloplasty has the potential to improve fetal left heart growth and increase the likelihood of achieving postnatal biventricular circulation.

What does this study add?

• Our study identifies late gestation predictors of postnatal biventricular circulation after fetal aortic valvuloplasty, including left ventricular long axis dimension z-score, left ventricular ejection fraction, anterograde aortic arch flow, and bidirectional or right-to-left flow across the foramen ovale.

• Determining which patients are likely to achieve biventricular circulation after birth will improve parental counseling and may inform delivery planning.

Data availability statement:

The data that support the findings of this study are available on request from the corresponding author.