Single-Ventricle Outcomes After Neonatal Palliation of Severe Ebstein Anomaly With Modified Starnes ProcedureClinical Perspective
Background: We have previously shown that neonates in profound cardiogenic shock caused by a severe Ebstein anomaly can be successfully salvaged with fenestrated right ventricular exclusion and systemic to pulmonary shunt (modified Starnes procedure). The long-term outcome of single-ventricle management in these patients is not known.
Methods: We retrospectively reviewed the records of patients who underwent neonatal Starnes procedure between 1989 and 2015. Patient demographics, clinical variables, and outcome data were collected.
Results: Twenty-seven patients (13, 48% boys) underwent the Starnes procedure at 7 (5–9) days of life. All were intubated and on prostaglandin, 24 (89%) were inotrope dependent, and 22 (81%) had no antegrade flow from the right ventricle. Three patients underwent nonfenestrated right ventricular exclusion, 2 (67%) of whom died. Of the remaining 24, 3 (13%) died during the same hospitalization. The 22 neonatal survivors have been followed for 7 (6–8) years: 1 patient is awaiting a Glenn procedure; 1 died after undergoing a Glenn procedure; and the remaining 20 patients have successfully undergone Fontan completion. Their indexed pulmonary vascular resistance was 1.8 (1.2–2.3) W/m2, and mean pulmonary pressure was 12 (9–18) mm Hg. At last follow-up, 1 patient had died, and the remaining patients had normal left ventricular function, and all but 1 have New York Heart Association class I symptoms. Two patients have required pacemaker implantation, whereas the rest are in sinus rhythm. Survival for the entire cohort at 1, 5, and 10 years is 81±4%, 81±5%, and 76±3%, respectively, whereas for those with fenestrated right ventricular exclusion, survival at 1, 5, and 10 years is 87±2%, 87±2%, and 81±4%, respectively.
Conclusions: Long-term single-ventricle outcomes among neonatal survivors of the modified Starnes procedure are excellent. There is reliable remodeling of the excluded right ventricle and good function of the left ventricle.
Ebstein anomaly encompasses a spectrum of congenital heart defects characterized by apical displacement of the septal leaflet of the tricuspid valve, resulting in a variable degree of tricuspid regurgitation and right ventricular (RV) dysfunction. In its most severe form, neonates present in refractory cardiogenic shock associated with high mortality rates. These patients tend to have highly malformed tricuspid valve leaflets, with a significant atrialized portion of the RV, resulting in severe tricuspid regurgitation and heart failure. There is varying degree of obstruction to antegrade flow out the RV and, frequently, profound cyanosis with ductal-dependent pulmonary circulation. We have previously shown that neonatal palliation with RV exclusion and modified systemic to pulmonary shunting (Starnes procedure) can effectively salvage these neonates.1,2 After a Starnes procedure, the excluded RV remodels over time,3 resulting in improved left heart function. Such a strategy, however, commits the patients to a single-ventricle pathway and Fontan circulation. We aimed to analyze the long-term outcomes of single-ventricle palliation in neonates undergoing a modified Starnes procedure for severe Ebstein anomaly.
We retrospectively reviewed the charts of 27 consecutive patients who underwent a Starnes procedure between 1989 and 2015 under an institutional review board–approved protocol. One patient had undergone neonatal Starnes procedure (by Vaughn Starnes) at another institution and underwent subsequent stages of palliation and follow-up at Children’s Hospital, Los Angeles. All other patients underwent all interventions and were also followed at Children’s Hospital, Los Angeles. Data on demographics, surgical interventions, hospital course, and follow-up were collected.
The neonatal Starnes procedure includes sternotomy with aortic and bicaval cannulation. Under moderate hypothermic bypass and cardioplegic arrest, RV exclusion is accomplished through the right atrium by patching the tricuspid valve at the anatomic annulus by using glutaraldehyde-fixed autologous pericardium. The coronary sinus is retained on the atrial side of the patch. In all but the first 3 patients, an ≈4-mm fenestration was made in the patch by using a coronary punch (modified Starnes procedure). Free atrial communication is ensured, and a right atrial reduction plasty is routinely performed. If significant pulmonary insufficiency exists, the main pulmonary artery (PA) is ligated. Whereas early in the series, RV plication was frequently performed, we have abandoned this practice in the past 17 patients. The ductus is ligated, and pulmonary blood flow is provided by a modified Blalock-Taussig shunt (3.5–4.0 mm). The size of the shunt is chosen largely based on patient weight; we tend to use a 3.5-mm shunt in patients weighing up to ≈3200 g. Subsequent stages of single-ventricle palliation followed current standard practice patterns. At ≈6 months of age, bidirectional superior cavopulmonary anastomosis (Glenn procedure) is performed under cardiopulmonary bypass and the shunt is ligated. At ≈3 years of age, total cavopulmonary connection (Fontan procedure) is established with an extracardiac 18-mm Goretex conduit. Fenestration of the conduit is undertaken on an as needed basis.
Standard echocardiographic analyses were undertaken as clinically indicated and at the discretion of the treating physicians. The Great Ormond Street (GOS) ratio was calculated as described by Celermajer et al.4 This ratio is obtained by dividing the combined area of the right atrium and atrialized portion of the RV by the sum of the areas of the trabeculated RV, left atrium, and left ventricle (LV). A GOS ratio of >1 was associated with 40% mortality, and no patient with a GOS ratio >1.5 survived in their study.4 The degree of septal protrusion into the LV cavity was assessed by using the LV septal impingement ratio. This ratio is obtained by dividing the upper septal curvature vertical dimension from the RV wall by the lower septal curvature vertical dimension from the LV wall as measured in the short-axis, left parasternal echocardiographic view.3,5 For the purposes of this study, the last echocardiographic assessment performed preoperatively before the Starnes, Glenn, and Fontan procedures was used to compare the GOS and septal impingement ratios. To determine the rapidity of RV regression after the Starnes procedure, serial echocardiograms obtained during the index hospitalization were analyzed. Initial and follow-up echocardiograms were reanalyzed for the purposes of this study by 2 authors (N.C., G.K.). Using standard anterior- posterior radiographic studies, the span of the cardiac silhouette was divided by the length of the thoracic cage at the same level. This number provided the cardiothoracic ratio. All patients underwent catheterization for hemodynamic and anatomic assessments before the Glenn and Fontan procedures.
Data were tabulated and analyzed using SAS 9.4. Continuous data are presented as median with 25th and 75th percentiles. Global comparison of cardiac ratios was undertaken for each patient at different time points using repeated-measures analysis of variance and post hoc Tukey test for multiple comparisons. Pairwise comparison of values at each subsequent time point to corresponding pre-Starnes values was undertaken using the Wilcoxon signed-rank test. Kaplan-Meier curves were used to assess survival. Significance was defined as P<0.05.
The demographic, clinical, and echocardiographic characteristics of the 27 patients are shown in Table 1. All patients had ductal-dependent pulmonary circulation, and the majority were ventilator and inotrope dependent. On echocardiography, only 5 patients had demonstrable antegrade pulmonary blood flow, 3 of whom also had significant pulmonary regurgitation. Four patients (15%) had RV pressure (estimated by tricuspid regurgitation jet) >40 mm Hg; 2 of these patients had an open ductus and severe pulmonary regurgitation. The median GOS, septal impingement, and cardiothoracic ratios are shown in Table 2. Twenty-two patients (81%) had a GOS ratio >1.
Two neonates underwent an initial systemic to pulmonary shunt, and, in 1 neonate, tricuspid valve repair was attempted. Because of persistent heart failure and lack of clinical improvement, these patients subsequently underwent RV exclusion. The other 24 neonates underwent primary RV exclusion (Table 3). In the first 3 patients, the pericardial patch was not fenestrated. One of these patients underwent a transannular RV outflow tract patch and another an RV-to-PA homograft placement in an attempt to decompress the persistently dilated RV into the pulmonary circulation. Both patients died. In the remaining 24 patients, the pericardial patch was fenestrated to decompress the RV into the low-pressure right atrium. In 6 of the first 10 patients, RV plasty/plication was performed to reduce the size of the dilated and thin-walled RV. This step was excluded once we started noticing significant RV remodeling with fenestration of the RV exclusion patch.3 In 1 patient with severe arrhythmias preoperatively, a concomitant maze procedure was performed. Because of significant pulmonary regurgitation, the main PA was ligated in 3 patients.
Hospital Course After Starnes Procedure
Three of the 24 patients (13%) who underwent fenestrated RV exclusion died 20, 42, and 71 days, respectively, after surgery during the same hospitalization. The cause of death was multiorgan dysfunction in 2 patients and sepsis in the third patient. There were 4 (15%) morbidity events noted: 1 shunt dysfunction requiring shunt revision, 1 complete heart block in the child who underwent the maze procedure requiring pacemaker insertion, 1 phrenic nerve paresis requiring hemidiaphragm plication, and 1 chylothorax managed with dietary restriction alone. In all but 1 patient, the chest was left open after the Starnes procedure and closed a median of 3 (3–4) days later. Median duration to first extubation was 6 (4–9) days after chest closure. Two patients required reintubation, and 1 patient required prolonged ventilatory support and subsequent tracheostomy before discharge. Median postoperative duration of stay in the intensive care unit was 14 (13–17.8) days, and postoperative hospital stay was 35 (27.3–44) days.
All patients underwent follow-up echocardiography during their hospitalization on the basis of clinical need. Of the 3 patients who had nonfenestrated RV exclusion, only 1 demonstrated RV regression. The other 2 had persistent RV dilation requiring additional procedures, and both eventually died. Of the 24 patients who had fenestrated RV exclusion, all patients had at least 1 examination within the first 2 weeks postoperatively. By postoperative day 15, all but 1 patient demonstrated significant reduction in their RV size. Median GOS score was 0.58 (0.39–0.73), with an average reduction of 0.35 points from preoperative values (P<0.001, Table 3). The only patient who did not demonstrate significant reduction in RV size had a preoperative GOS score of 0.82 and a postoperative score of 0.72. By the time of discharge (32 days after surgery), this patient also demonstrated significant RV decompression with a GOS score of 0.59. All patients had obligatory regurgitation through the fenestration with a median pressure gradient of 14.4 mm Hg (10.2–18.3). No patient had significant pulmonary regurgitation.
Subsequent Stages of Single-Ventricle Palliation
Of the 22 patients who were discharged after the Starnes procedure, 1 is awaiting a Glenn procedure, and 21 underwent a Glenn procedure at a median 7 months of age (Table 2). By the time of their Glenn procedure, all patients demonstrated significant remodeling of their hearts (Table 3, Figures 1 and 2). Three patients required PA plasty at the time of the Glenn procedures. An LV lead was added to the patient who had a pacemaker. One patient required reexploration for bleeding, and 1 patient had chylothorax that was managed with dietary changes alone. There was no mortality.
After the Glenn procedure, 1 patient died at 1 year of age of a noncardiac cause. The remaining 20 patients underwent a Fontan completion procedure at a median age of 2.8 years (Table 2). The favorable cardiac remodeling persisted to the time of Fontan completion (Table 3, Figures 1 and 2). Median indexed pulmonary vascular resistance was 1.8 (1.2–2.3) Wood units/m2 and mean PA pressure was 12 (9–18) mm Hg. One patient required fenestration at the time of the Fontan procedure, and 1 patient with residual antegrade flow out the RV underwent main PA ligation at the Fontan procedure. There was no operative morbidity or mortality.
The 20 patients who have reached Fontan circulation have been followed for a median of 7 years (Table 4). One patient (5%) died at ≈6 years of age of Fontan- related complications. Overall survival for the entire patient cohort at 1, 5, and 10 years is 81±4%, 81±5%, and 76±3%, respectively, whereas survival for those with fenestrated RV exclusion is 87±2%, 87±2%, and 81±4%, respectively. Among patients who survived to discharge after the Starnes procedure, 1-, 5-, and 10-year survival was 95±3%, 95±3%, and 89±4%, respectively (Figure 3). Two patients have required additional interventions: 1 Fontan fenestration and 1 pacemaker implantation for sinus node dysfunction and bradycardia. At last follow-up, all patients had normal LV function. One patient has exercise intolerance and desaturation, whereas the remaining patients have New York Heart Association grade I symptoms. The 18 patients without a pacemaker are in sinus rhythm, and no patient has required long-term antiarrhythmic therapy.
Our study is the first to report long-term single-ventricle outcomes after neonatal palliation with a modified Starnes procedure in patients with severe Ebstein anomaly. We demonstrate excellent physiology, favorable hemodynamics, and a salutary symptom profile in patients who successfully reach Fontan circulation.
Patients with an Ebstein anomaly have a widely variable clinical presentation. Patients who present in childhood or later with symptoms from tricuspid insufficiency can be managed with tricuspid valve repair, with or without a concomitant Glenn procedure (1.5 ventricle repair).6–8 However, neonates with Ebstein anomaly continue to present management challenges.2,9,10 Outcomes of neonatal tricuspid valve repair are variable in the literature, depending on institutional preference and potential differences in Ebstein morphology between the patient cohorts. Boston et al11 reported tricuspid valve repair in 26 neonates and infants with excellent long-term durability. They advocate a strong preference toward primary tricuspid valvuloplasty and biventricular repair, with the exception of patients with pulmonary atresia in whom there was 45% survival after biventricular repair. In contrast, Shinkawa et al12 report disappointing results after neonatal tricuspid valve repair with 75% mortality. They have largely abandoned this approach in favor of RV exclusion in neonates with cyanosis and heart failure. Our practice has been to tailor therapy to each individual patient based on the clinical and echocardiographic parameters. When the RV is of reasonable size (low GOS score) with significant antegrade pulmonary blood flow, we trial patients off prostaglandin infusion. If saturations and cardiac output are well maintained, they are followed with the plan for anatomic repair later in life. In contrast, the cohort of neonates in this study could not be managed medically. All these patients had severe tricuspid regurgitation, and 89% were in inotrope-dependent heart failure. Only 19% had any antegrade pulmonary blood flow, and all required ductal patency to maintain oxygen saturation. We maintain that, in this difficult subset of patients, the most appropriate approach is RV exclusion and management along a single-ventricle pathway. Our data clearly demonstrate that RV exclusion can be safely accomplished with a very favorable morbidity and mortality profile. As we have gained experience with this procedure, hospital mortality for the Starnes procedure has significantly improved with only 1 death in the past 22 patients over 19 years. This compares very favorably with single right-ventricle lesions that require complex neonatal palliation such as the Norwood procedure.
One of the most important features of neonatal palliation is successful RV exclusion and decompression. We believe that the RVs in these patients are incapable of functioning adequately to maintain cardiac output. Our initial approach was to place a nonfenestrated patch to exclude the RV. In 2 of the 3 patients who underwent this procedure, there was no significant decompression of the RV. Attempts to open the RV outflow tract by either a transannular patch or a RV-to-PA homograft were unsuccessful in decompressing such a dysfunctional RV into the high-pressure neonatal pulmonary vascular bed. In the subsequent 24 patients, fenestrating the patch led to effective decompression of the RV into the low- pressure right atrium, resulting in significantly improved outcomes. RV decompression is so effective that it has obviated the need for RV plasty/plication, which we frequently used in the early part of this series. Whereas patch fenestration seems to strongly correlate with better outcomes, given the small numbers of patients with events in our series, we were unable to identify any clinical predictors of mortality after RV exclusion.
The excluded and decompressed RV undergoes rapid regression and remodeling. Within 2 weeks after the Starnes procedure, we document reduction in GOS scores and improvement in septal impingement. There is rapid reestablishment of normal LV function in the postoperative period, which emphasizes the importance of ventricular septal impingement leading to severe heart failure in these neonates. The decompressed RV also becomes depressurized as demonstrated by low regurgitation velocities. We interpret this rapid reduction in RV pressures to indicate that these ventricles were likely intrinsically incapable of maintaining adequate cardiac output to the lungs, thereby reiterating our preference toward single-ventricle palliation. Whether there exists a subset of patients with less severe RV hypoplasia/dysfunction whose ventricles could subsequently be integrated back into circulation cannot be ascertained on the basis of our study.
The excluded RV allows adequate filling and normal physiology of the LV in the long term. There is excellent function of the LV and very favorable pulmonary hemodynamics, resulting in successful palliation of the vast majority of neonatal survivors to Fontan circulation. Such a favorable clinical profile continues past Fontan completion as well. We have lost only 1 patient to Fontan-related complications. All the others have favorable symptom profiles and quality of life. Such a long-term outcome compares very favorably to other single LV results, such as in patients with tricuspid atresia. In addition, we have not seen any long-term issues with supraventricular arrhythmias, which is frequently associated with Ebstein anomaly.
As a single-center retrospective study, our work suffers from some inherent limitations. We have studied only patients who underwent neonatal palliation with the Starnes procedure. Therefore, our results cannot be universally applied to all patients with Ebstein anomaly. Although this is the largest series of neonates with severe Ebstein anomaly, it is still composed of a relatively small number of patients. Despite these limitations, our study clearly establishes that the modified Starnes procedure is an effective approach to palliate neonates with severe Ebstein anomaly and cardiogenic shock. RV exclusion combined with decompression into the low-pressure right atrium leads to effective RV remodeling, normalization of LV morphology, and improved LV function that is durable in the long term. Long-term outcomes for single-ventricle palliation are excellent.
Sources of Funding
Sources of Funding, see page 1264
Circulation is available at http://circ.ahajournals.org.
- Received December 29, 2015.
- Accepted July 26, 2016.
- © 2016 American Heart Association, Inc.
- Celermajer DS,
- Cullen S,
- Sullivan ID,
- Spiegelhalter DJ,
- Wyse RK,
- Deanfield JE
- Satomi G,
- Nakamura K,
- Takao A,
- Imai Y
- da Silva JP,
- Baumgratz JF,
- da Fonseca L,
- Franchi SM,
- Lopes LM,
- Tavares GM,
- Soares AM,
- Moreira LF,
- Barbero-Marcial M
- Knott-Craig CJ,
- Goldberg SP,
- Ballweg JA,
- Boston US
What Is New?
Neonates with severe Ebstein anomaly who present in refractory cardiogenic shock can be effectively palliated with a Starnes procedure. Ours is the first study to comprehensively analyze the long-term single-ventricle outcomes in this patient population.
Our results show that the modified Starnes procedure results in very favorable short-term outcomes, similar to those seen in other high-risk neonates requiring complex palliative procedures.
The excluded and decompressed right ventricle undergoes rapid regression and stable remodeling.
Long-term single-ventricle outcomes among neonatal survivors of a modified Starnes procedure are excellent.
What Are the Clinical Implications?
In neonates with severe Ebstein anomaly who present in extremis, the right ventricles are incapable of functioning adequately to maintain cardiac output and right ventricular exclusion, and decompression is crucial to successful outcome. Right ventricular decompression leads to rapid reestablishment of normal left ventricular function, emphasizing the importance of ventricular septal impingement leading to severe heart failure in these neonates.
In the long term, this leads to excellent function of the left ventricle and favorable pulmonary hemodynamics, resulting in successful palliation to Fontan circulation. The favorable clinical profile continues past Fontan completion as well.