The optimal timing of Stage-2-Palliation for Hypoplastic Left Heart Syndrome: An analysis of the Pediatric Heart Network Single Ventricle Reconstruction Trial public dataset

Abstract

Background

In infants requiring three-stage single ventricle palliation for hypoplastic left heart syndrome, attrition after the Norwood procedure remains significant. The effect of the timing of stage-2-palliation (S2P), a physician-modifiable factor, on long term survival is not well understood. We hypothesized that an optimal interval between the Norwood and S2P that both minimizes pre-S2P attrition and maximizes post-S2P survival exists and is associated with individual patient characteristics.

Methods

The NIH/NHLBI Pediatric Heart Network Single Ventricle Reconstruction Trial public dataset was used. Transplant-free survival (TFS) was modeled from (1) Norwood to S2P and (2) S2P to three years, using parametric hazard analysis. Factors associated with death or heart transplantation were determined for each interval. To account for staged procedures, risk-adjusted, three-year, post-Norwood TFS (the probability of TFS at three years given survival to S2P) was calculated using parametric conditional survival analysis. TFS from the Norwood to S2P was first predicted. TFS after S2P to three years was then predicted and adjusted for attrition before S2P by multiplying by the estimate of TFS to S2P. The optimal timing of S2P was determined by generating nomograms of risk-adjusted, three-year, post-Norwood, TFS versus the interval from the Norwood to S2P.

Results

Of 547 included patients, 399 survived to S2P (73%). Of the survivors to S2P, 349 (87%) survived to three-year follow-up. The median interval from the Norwood to S2P was 5.1 (IQR 4.1–6.0) months. The risk-adjusted, three-year, TFS was 68±7%. A Norwood-S2P interval of three to six months was associated with greatest three-year TFS overall and in patients with few risk factors. In patients with multiple risk factors, TFS was severely compromised, regardless of the timing of S2P and most severely when S2P was performed early. No difference in the optimal timing of S2P existed when stratified by shunt type.

Conclusions

In infants with few risk factors, progressing to S2P at three to six months after the Norwood procedure was associated with maximal TFS. Early S2P did not rescue patients with greater risk factor burdens. Instead, referral for heart transplantation may offer their best chance at long term survival.

ClinicalTrials.gov number: NCT00115934.

In infants with hypoplastic left heart syndrome (HLHS) requiring three-stage single ventricle palliation, mortality between the Norwood procedure and stage-2-palliation (S2P) remains high, ranging from 2–20%. In the Single Ventricle Reconstruction (SVR) Trial, 16% died prior to Norwood hospitalization discharge and another 12% died after discharge and before S2P.^1–3 The National Pediatric Cardiology Quality Improvement Collaborative recently reported similar 10% post-discharge mortality after the Norwood hospitalization before S2P.^4–6

It remains unknown whether shortening an infant’s exposure to Norwood-associated cardiovascular physiology would decrease pre-S2P attrition. The parallel circulation present after initial palliation requires the single ventricle to perform excess volume work and may expose the pulmonary vasculature to elevated pressures. Progression to S2P eliminates the risk of systemic to pulmonary artery shunt failure, greatly reduces the excess volume work performed by the ventricle, normalizes coronary arterial diastolic blood pressure in patients with a modified Blalock-Taussig shunt (MBTS), and reduces pulmonary overcirculation. However, very early conversion to S2P is not without risk, including potentially reduced pulmonary arterial growth, pulmonary vascular resistance higher than desirable for S2P, and increased resource utilization in younger patients.⁷

The current evidence regarding the timing of S2P is limited to single-institution studies that only report outcomes after S2P, without accounting for pre-S2P mortality. Two decades ago, small case series focused on how early S2P can be performed with acceptable outcomes.^8–11 Recent studies have yielded contradictory conclusions, regardless of how timing was analyzed.^12–16 Secondary analyses of the SVR and Infant Single Ventricle Trials showed no association of age at S2P with survival, hypoxemia, or ejection fraction.^{17, 18} The timing of S2P remains at the discretion of institutional or individual surgeon/cardiologist preferences.

The primary objectives of this study were to investigate the clinical factors associated with the timing of S2P and to determine the optimal timing of S2P that both minimizes pre-S2P attrition and maximizes post-S2P survival.

Methods

Study Design

The NIH/NHLBI Pediatric Heart Network SVR Trial dataset was used in preparation of this work. Data collected at enrollment through three-year follow-up were downloaded from http://www.pediatricheartnetwork.org/ForResearchers/PHNPublicUseDatasets/SingleVentricleReconstructionTrial.aspx on August 6, 2015. The SVR Trial was a prospective, multicenter, randomized, controlled trial comparing one-year transplant-free survival in patients undergoing the Norwood procedure, randomized to receive a modified Blalock-Taussig shunt (MBTS) or a right-ventricular-to-pulmonary artery (RVPA) conduit, with ongoing long-term follow-up. The details of recruitment, inclusion and exclusion, patient management, and data collection have been previously reported.^{19, 20} The study protocol was approved by each participating site’s institutional review board. Prior to randomization, each patient’s parent or guardian provided written informed consent. The pre-Norwood medical history, Norwood procedure and post-operative course, S2P operation and post-operative course, post-natal echocardiograms (pre-Norwood, post-Norwood, and pre-S2P), pre-S2P angiography and cardiac catheterization, and three-year outcomes were analyzed. Shunt type was defined as the shunt in place at the end of the Norwood hospitalization. No data regarding Fontan completion were available in the public data set.

Study Population

Of these 549 patients who were included in the original analysis of the trial’s primary endpoint, 547 were included in this study. Two patients were excluded: one who underwent S2P but had no operative data available and one who had an excessively long Norwood-S2P interval (22.8 months).

Clinical and Operative Data

The baseline characteristics, operative details, and outcomes of the Norwood procedure, and S2P in the SVR Trial participants have been reported previously.^{5, 18, 20, 21} The majority of patients underwent a right bidirectional superior cavopulmonary connection or a right hemi-Fontan operation. The elective status of the S2P operation was coded as elective or non-elective, at the discretion of each site investigator.¹⁸ Specific indications for non-elective S2P were coded as progressive hypoxemia, failure to thrive, shunt occlusion, neoaortic arch obstruction, moderate or greater systemic atrioventricular valvular insufficiency, ventricular dysfunction, or other, again at the discretion of the site investigators.

Statistical Analysis

Survival analysis and multivariable modeling

Time-related analyses of the primary outcome, a composite of death or heart transplantation, were performed over two intervals, after the Norwood and after S2P (Supplemental Figure 1). Transplant-free survival was modeled for both intervals using parametric risk hazard analysis. Briefly, parametric hazard analysis models the hazard function by decomposing it into multiple phases of risk. Three phases of risk may exist, though are not required for model validity (an early peaking, rapidly decreasing hazard, a constant phase, and late, rising hazard). The SAS procedure PROC HAZARD uses the maximum likelihood method to resolve the phases of the distribution of times until an event.²² Additional details can be found at http://www.lerner.ccf.org/qhs/software/hazard.

Multivariable risk hazard analysis was performed to identify factors associated with death or heart transplantation after each operation. Bootstrap aggregation, or sampling with replacement, (n=500 resamples) was used to guide variable selection.²³ The final model was built using forward stepwise selection, with p < 0.07 for entry and p < 0.05 for retention. Variables with greater than 50% missing values were excluded. For variables with less than 50% missing data, multiple informative imputation using PROC MI was performed. To graphically depict the effects of risk factors on survival, representative curves were generated in which values for the risk factors were specified in the multivariable models. Specifically, the multivariable parametric equations were solved for specified values of the associated factors. “Summary” or “average” curves were created by generating predicted survival curves for each individual study patient, using their actual values for the factors in the multivariable equations, and then summarized into a single curve.²⁴

Determinants of timing of S2P

Linear regression, in which the outcome was the length of the Norwood-S2P interval, was used to determine the patient characteristics significantly associated with the timing of S2P. Bootstrap aggregation was used to aid in variable selection (n=500 resamples) and covariables with greater than 50% reliability were included for final selection (stepwise selection, p=0.07 for entry, p=0.05 for retention).

Analysis of the timing of S2P

Locally Weighted Scatterplot Smoothing (LOESS) logistic regression was first used to model the relationship between the timing of S2P and transplant-free survival after S2P in patients who survived to S2P. Parametric conditional survival analysis was used to analyze survival through the Norwood and S2P. Conditional survival is defined as the probability of surviving to time t given survival to time s [CS(t|s)], or the probability of surviving to three years after the Norwood procedure given survival to S2P. ^{25, 26} Transplant-free survival after the Norwood procedure to the day of S2P was predicted using the parametric model for transplant-free survival after the Norwood, as described above. The model for transplant-free survival after S2P was next used to predict transplant-free survival to three years, which was then adjusted for attrition before S2P by multiplying by the estimated transplant-free survival to the day of S2P obtained from the first model. The optimal timing was determined by creating nomograms in which the predicted three-year, post-Norwood, transplant-free survival was plotted as a function of the Norwood-S2P interval. Detailed information regarding analytic techniques can be found in the Supplementary Materials and Supplementary Figures 2–7. All statistical analyses were performed using SAS version 9.2 (SAS Institutes, Cary, NC).

Results

Patient outcomes and risk factors before S2P

Of the 547 patients who underwent a Norwood procedure, 139 (25%) died and 9 (2%) underwent cardiac transplantation. Factors associated with death or transplant after Norwood included MBTS versus RVPA conduit, requiring extracorporeal membrane oxygenation (ECMO) at the end of the Norwood, the use of steroids during the Norwood, smaller pre-Norwood tricuspid valve diameter z-score, the presence of an aberrant right subclavian artery, the use of aprotinin during the Norwood, site volume less than or equal to 30 Norwood procedures per year, longer cardiopulmonary bypass time, longer deep hypothermic circulatory arrest time, and higher pre-Norwood lactate value (Table 1).

Table 1.

Risk factors for outcomes after the Norwood procedure

	Parameter Estimate*	P-value	Reliability†
Factors associated with death/heart transplantation
Required ECMO at Norwood	1.50	< 0.0001	99%
MBTS vs. RVPA conduit	0.67	0.0001	97%
Use of intraoperative steroids during the Norwood procedure	1.69	0.0003	85%
Smaller tricuspid valve anteroposterior diameter z-score on baseline echocardiogram	0.17	0.008	80%
Presence of an aberrant right subclavian artery	0.94	0.001	72%
Use of intraoperative aprotinin during the Norwood procedure	0.47	0.03	67%
Site volume 16–20 Norwood procedures per year	0.62	0.004	61%
Longer cardiopulmonary bypass time (min)	0.01	0.001	58%
Higher maximal pre-Norwood lactate level (mmol/L)	0.06	0.006	58%
Longer deep hypothermic circulatory arrest time (min)	0.01	0.009	51%
Factors associated with reaching stage-2-palliation
Smaller percentage below the federal poverty level (%)	0.01	0.003	76%
Coarctectomy performed during the Norwood procedure	0.26	0.02	57%
Shorter deep hypothermic circulatory arrest time (min)	0.01	0.04	69%
Larger tricuspid valve anteroposterior diameter z-score on baseline echocardiogram	0.08	0.007	54%

ECMO=extracorporeal membrane oxygenation, MBTS=modified Blalock-Taussig shunt, RVPA=right ventricle to pulmonary artery

^*Within the single phase of risk present in this model, the parameter estimates derived from parametric analysis can be interpreted analogously to hazard ratios from Cox proportional hazards methods.

^†Reliability indicates the fraction of bootstrapped samples in which the covariable was present

Of the remainder, 399 (73%) underwent S2P. Factors associated with reaching S2P included a smaller percentage below the federal poverty line, shorter deep hypothermic circulatory arrest time, performing a coarctectomy during the Norwood procedure, and smaller pre-Norwood tricuspid valve diameter z-score (Table 1).

Patient outcomes and risk factors after S2P

Of the 399 who underwent S2P, 349 (87%) survived to three-year follow-up, 39 (10%) died, and 11 (3%) underwent cardiac transplantation. The median interval between the day of the Norwood and the day of S2P was 5.1 (IQR 4.1–6.0) months. Among the 399 patients who underwent S2P, 66% were elective operations. Parametric risk hazard analysis resolved two phases of risk after S2P, an early and a constant phase. Factors associated with death or transplant during the early phase included smaller neo-aortic annular area z-score on the post-Norwood echocardiogram, the presence of pulmonary artery stenosis before S2P, greater number of post-Norwood complications, failure to thrive as an indication for S2P, severe systemic atrioventricular valve regurgitation, and requiring ECMO after the Norwood. Specifically, a shorter interval between the Norwood procedure and S2P was associated with an increased risk of death or heart transplantation. The only significant associated factor during the constant phase was a lower RV ejection fraction on the pre-S2P echocardiogram (Table 2).

Table 2.

Risk factors for death or heart transplant after Stage-2-Palliation

	Parameter Estimate*	P-value	Reliability†
Early phase:
Smaller neo-aortic annular area z-score on post-Norwood echo	0.33	0.0007	35%
Presence of pulmonary artery stenosis	3.00	< 0.0001	48%
Greater number of post-Norwood complications	0.20	< 0.0001	48%
Failure to thrive, indication for S2P	1.76	0.005	45%
Shorter Norwood-S2P interval (days)‡	1.89	0.001	33%
Severe systemic atrioventricular valve regurgitation on pre-S2P echo	1.68	0.001	32%
Required ECMO post-Norwood procedure	1.48	0.002	31%
Constant phase:
Lower right ventricular ejection fraction on pre-S2P echo	0.14	0.0009	45%

^*Within each phase of risk in this model, the parameter estimates derived from parametric analysis can be interpreted analogously to hazard ratios from Cox proportional hazards methods.

^†Reliability indicates the fraction of bootstrapped samples in which the covariable was present

^‡Natural log transformation

S2P=Stage-2-Palliation, ECMO=Extracorporeal Membrane Oxygenation

Risk-adjusted, transplant-free survival at three years

Overall, the unadjusted, transplant-free survival at three years post-Norwood was 63±2%. Using the actual values for the factors associated with death or heart transplantation, the risk-adjusted, three-year, post-Norwood, transplant-free survival through both the Norwood procedure and S2P for all 547 infants was 68±7% (Figure 1).

Figure 1.

Risk-adjusted, transplant-free survival at three years after the Norwood operation in the Single Ventricle Reconstruction Trial. Using the parametric models for transplant-free survival after the Norwood and after S2P, curves representing transplant-free survival were calculated for all 547 patients in the trial, using their actual risk factor values. All curves were then averaged into a single curve, which is displayed here. The risk-adjusted, transplant-free survival at three years was 68±7%. The dashed lines represent the 70% confidence limits.

Determinants of the timing of S2P

The interval from the Norwood to S2P was modeled and the results are shown in Table 3. Covariables associated with a shorter interval included shunt stenosis at the time of S2P, ventricular dysfunction as an indication for S2P, the use of intraoperative alpha blockade, center volume of 20 or fewer Norwood procedures per year, male sex, and a higher weight-for-age z-score at the pre-S2P catheterization. Covariables associated with a longer interval included smaller left pulmonary artery diameter on the pre-S2P catheterization, a greater right ventricular end diastolic volume on the pre-S2P echocardiogram, a greater total number of operations during the interstage period, the use of aprotinin during the Norwood, the use of steroids during the Norwood, and electively-performed S2P.

Table 3.

Factors associated with the length of the Norwood-S2P interval

	Parameter Estimate (Days)	P-Value	Reliability
Presence of shunt stenosis pre-S2P	−33.7	0.0008	69%
Ventricular dysfunction, indication for S2P	−27.9	0.0005	64%
Use of intraoperative alpha blockade during the Norwood procedure	−18.8	< 0.0001	94%
Site volume ≤20 Norwood procedures per year	−17.7	0.002	63%
Male sex	−13.5	0.003	62%
Greater weight-for-age z-score at the pre-S2P cardiac catheterization	−8.4	< 0.0001	86%
Greater proximal LPA diameter on the pre-S2P cardiac catheterization (mm)	3.1	0.03	78%
Greater RV end diastolic volume on the pre-S2P echocardiogram (cm2)	4.6	< 0.0001	89%
Greater total number of interstage operations	9.9	0.0009	74%
Use of intraoperative aprotinin during the Norwood procedure	13.7	0.007	80%
Use of intraoperative steroids during the Norwood procedure	17.9	0.01	53%

S2P=Stage-2-Palliation, LPA=Left Main Pulmonary Artery, RV=Right Ventricular

The optimal timing of S2P in the SVR Trial

Exploratory analysis using LOESS regression demonstrated that the probability of death or heart transplantation was initially just greater than 20% (Supplemental Figure 8). It then decreased as the Norwood-S2P interval lengthened and rose again to approximately 20% as the interval approached 13 months.

The optimal timing of S2P was then determined by plotting calculated three-year, risk-adjusted, post-Norwood transplant-free survival versus the Norwood-S2P interval. Using the actual patient values for all 547 patients in both multivariable models, an averaged nomogram was created and is shown in Figure 2. Calculated transplant-free survival at three years was stable at 68±7% during a Norwood-S2P interval of three to six months. Calculated transplant-free survival decreased rapidly when S2P was performed before three months after the Norwood and then again, though more gradually, after six months.

Figure 2.

The optimal timing of Stage-2-Palliation (S2P) in the Single Ventricle Reconstruction Trial. In this nomogram, the probability of transplant-free survival at three years after the Norwood procedure is plotted versus the interval between the Norwood and S2P operations in months (solid line). Calculated transplant-free survival is severely compromised until after three months, at which it stabilizes at 68±7%, and begins to decrease again after six months. This curve was generated by producing nomograms for all 547 patients in the trial and averaging them into a single curve. The dashed lines represent the 70% confidence limits.

The effect of performing S2P outside of the optimal timing

The effect of performing S2P outside of the optimal window on transplant-free survival can be demonstrated by plotting survival over time, stratified by day of S2P. In Supplemental Figure 9A, transplant-free survival was calculated at three years after the Norwood procedure and was stratified by the day of S2P for a “hypothetical average” patient (all associated factor values in the multivariable models were set at their mean values). Performing S2P before 1.5 (red curve) or two (pink curve) months after the Norwood resulted in three-transplant-free survival estimates of 63±11% and 69±6%, respectively. Calculated three-year, transplant-free survival estimates when S2P was performed at three (orange curve), four (yellow curve), and five (green curve) months after the Norwood were 73±3%, 73±3%, and 73±3%, respectively (Supplemental Figure 9B). Steady decrements in calculated transplant-free survival was seen when S2P was performed after Norwood-S2P intervals greater than six months, with three-year transplant-free survival estimates of 73±3%, 72±3%, 71±4%, 69±4%, and 65±4% when S2P was performed at six (blue curve), seven (purple curve), eight (brown curve), nine (grey curve), or 12 (black curve) months, respectively (Supplemental Figure 9C). The nomogram representing these estimates as a function of the Norwood-S2P interval is shown separately in Supplemental Figure 4B.

The effect of specific risk factors on the timing of S2P

The nomograms depicting calculated risk-adjusted transplant-free survival can be stratified by a selected associated factor from the multivariable models. In this manner, the effect of shunt type at the end of the Norwood hospitalization was investigated. Figure 3A demonstrates that the optimal timing of S2P does not differ by shunt type, as the confidence limits of each curve overlap for any Norwood-S2P interval. Therefore, the same Norwood-S2P interval of three to six months is associated with maximal calculated transplant-free survival in patients regardless of shunt type. Similarly, the interval associated with maximal calculated three-year, transplant-free survival in patients with failure to thrive was investigated. In Figure 3B, undergoing S2P prior to six months after the Norwood procedure resulted in calculated three-year, transplant-free survival less than 50%.

Figure 3.

A. The optimal timing of Stage-2-Palliation (S2P) stratified by shunt type. No Norwood-S2P is associated with differential calculated, risk-adjusted, transplant-free survival whether the patients had a modified Blalock-Taussig shunt (MBTS) or a right-ventricle-to-pulmonary-artery (RVPA) conduit. This curve was generated by generating nomograms for all 547 patients in the trial, then averaging them into a single curve, and stratifying by shunt type at the end of the Norwood hospitalization.

B. The optimal timing of S2P, stratified by the presence of Failure to Thrive (FTT) as an indication for progression to S2P. Performing S2P earlier than six months after the Norwood in patients with FTT is associated with calculated risk-adjusted, transplant-free survival of less than 50%. This curve was generated by producing nomograms for all 547 patients in the trial, then averaging them into a single curve, and stratifying by FTT. The dashed lines represent the 70% confidence limits.

The optimal timing of S2P based on individual patient risk profiles

In Figure 4, curves solved using the actual values for the associated factors for individual patients with few and multiple risk factors are shown, along with the averaged curve for all patients in the study cohort. When comparing a patient with few risk factors (blue curve) to the cohort average (black curve), there was no difference in the Norwood-S2P interval that was associated with maximal calculated three-year, transplant-free survival. However, in a patient with multiple risk factors (orange curve), performing S2P before a Norwood-S2P interval of six months was associated with calculated three-year, transplant-free survival of less than 50% (Figure 4).

Figure 4.

The optimal timing of Stage-2-Palliation (S2P), stratified by risk profile. In low-risk patients (blue line), the optimal timing of S2P that maximizes calculated risk-adjusted, transplant-free survival at three years after the Norwood does not differ from the cohort average (black line, generated using all 547 patients, averaged into a single curve). In high risk patients (orange line), earlier S2P is associated with much lower calculated risk-adjusted, transplant-free survival, 50% or lower.

Risk factor values, patient with few risk factors: Did not require extracorporeal membrane oxygenation (ECMO) at the end of the Norwood procedure, modified Blalock-Taussig shunt, received steroids, pre-Norwood tricuspid valve z-score=4.3, no aberrant right subclavian artery, received aprotinin, site volume < 30 Norwood procedures/year, cardiopulmonary bypass time=111 minutes, highest pre-Norwood lactate=4.0 mg/dL, deep hypothermic circulatory arrest time=0 minutes, post-Norwood neo-aortic annular area z-score=10.85, no pulmonary arterial stenosis, three post-Norwood complications, no diagnosis of failure to thrive, diagnosed with severe atrioventricular valve regurgitation on pre-S2P echocardiogram, did not require ECMO before Norwood hospitalization discharge, and right ventricular ejection fraction=43.9% on the pre-S2P echocardiogram.

Risk factor values, patient with multiple risk factors: Did not require extracorporeal membrane oxygenation (ECMO) at the end of the Norwood procedure, Right-Ventricle-to-Pulmonary-Artery conduit, received steroids, pre-Norwood tricuspid valve z-score=0.4, no aberrant right subclavian artery, received aprotinin, site volume < 30 Norwood procedures/year, cardiopulmonary bypass time=140 minutes, highest pre-Norwood lactate=1.9 mg/dL, deep hypothermic circulatory arrest time=60 minutes, post-Norwood neo-aortic annular area z-score=1.63, no pulmonary arterial stenosis, one post-Norwood complication, diagnosed with failure to thrive, no diagnosis of severe atrioventricular valve regurgitation on pre-S2P echocardiogram, did not require ECMO before Norwood hospitalization discharge, and right ventricular ejection fraction=56.2% on pre-S2P echocardiogram.

Discussion

We investigated the optimal timing of S2P associated with minimal post-Norwood attrition and maximal post-S2P survival. A shorter Norwood-S2P interval was associated with an increased risk for death or heart transplantation after S2P. In low- or average-risk patients, an interval of three to six months after the Norwood was associated with maximal calculated transplant-free survival at three years post-Norwood. In high risk patients, three-year transplant-free survival was substantially compromised, regardless of the timing of S2P or shunt type. Patients with high risk characteristics prior to the Norwood who required early progression to S2P demonstrated especially dismal outcomes.

Accounting for transplant-free survival through staged procedures

We utilized parametric conditional survival analysis to account for staged procedures. The non-parametric or semi-parametric forms, based on Kaplan-Meier or Cox methods, have been increasingly applied in oncology over the last decade.^27–29 They have been used to predict survival beyond a given earlier time point, e.g. survival to five years, given survival to one year after cancer resection. Survival after heart transplantation for congenital heart disease has been examined using conditional survival analysis.^{30, 31}

In contrast, single ventricle palliation requires multiple operations. Previous studies have analyzed survival after S2P only. Non-parametric conditional survival analysis was inadequate for this study, as it can only graphically display survival following the conditional event (i.e. S2P), and thereby fails to provide important clinical information regarding the risk of significant attrition prior to S2P. Using a novel application of this method, we were able to more completely describe transplant-free survival from the Norwood procedure, through S2P, and up to three-year follow-up.

The optimal timing of S2P in patients without risk factors

Since 2000, several single-center series have examined the timing of S2P, although many were limited by sample size and/or lack of risk adjustment. Jaquiss et al. examined early and late outcomes in patients who underwent S2P either before or after four months old, and found no difference in survival or Fontan completion. However, resource utilization was higher in the early S2P group.^{12, 13} Petrucci et al. also reported equivalent survival between patients who underwent S2P before and after three months of age.¹⁵ Two studies determined that age less than three months at S2P was associated with an increased risk for death, heart transplantation, or inability to proceed with Fontan completion.^{32, 33} Many other recent studies of the outcomes after S2P have not found the timing of S2P to be associated with outcomes.^{14, 16, 34, 35} Notably, risk factors for death or heart transplantation after S2P have not yet been reported in SVR Trial study participants. Schwartz, et al. instead reported risk factors for prolonged hospital length of stay after S2P. While many were similar to those described in this analysis, neither age at S2P nor the Norwood-S2P interval were identified.¹⁸

Our results indicate that S2P performed between three and six months after the Norwood is associated with maximal calculated three-year survival. While the majority of patients who underwent elective S2P did actually undergo the operation within the three to six-month window, 30 patients underwent S2P prior to three months after the Norwood and 52 underwent S2P after six months (Figure 5A). Four of the 30 patients underwent elective S2P at less than three months after the Norwood. It is unclear what prompted early progression to S2P among these patients. These may also simply represent coding errors for their indications for S2P. Of perhaps greater interest are the 14 patients who underwent non-elective progression to S2P after a Norwood-S2P interval longer than six months. Even though a single death occurred in these patients, this death may have been avoided, without harming these patients’ chance of three-year survival, by performing S2P electively prior to six months (Figure 5B). Alternatively, the true clinical reasons for the delay of S2P may not have been adequately captured in the SVR Trial data set. In this case, although exhaustively performed in this analysis, incomplete risk adjustment may have occurred. Furthermore, the possibility exists that the factors identified may differentially affect the risk of death or heart transplantation in patients who undergo S2P earlier and later than the three to six-month window.

Figure 5.

A. The elective status of the Stage-2-Palliation (S2P) by month after the Norwood procedure in which S2P was performed. B. The outcomes of patients by month after the Norwood procedure in which S2P was performed.

Transplant-free survival over time when S2P is performed outside of the optimal timing

Parametric conditional survival analysis also enabled us to examine the effect of performing S2P outside of the optimal timing, via the graphical depiction of calculated survival through several hypothetical times for S2P. These plots clearly demonstrate the much higher hazard for death and heart transplantation before S2P than after S2P (Supplemental Figures 9A–C). However, the hazard for death or transplant after S2P changes with the time at which S2P is performed. The non-linear association between the timing of S2P and three-year transplant-free survival is displayed. Performing S2P 1.5 months after the Norwood yields calculated survival estimates much closer to performing it at 12 months post-Norwood versus at three months. This family of curves demonstrates the clinical paradox – how to balance the risk of early S2P with the continuing attrition in the post-Norwood period.

The optimal timing of S2P does not differ for average-risk patients by shunt type

The goal of the SVR Trial was to compare transplant-free survival at one year in patients who underwent the Norwood procedure randomized receive either a MBTS or RVPA conduit. In this analysis, we compared the timing of S2P associated with maximal calculated transplant-free survival. Although a MBTS was associated with death or heart transplant prior to S2P, the optimal interval did not differ between shunt types.

A single study has described a difference in the timing of S2P based on shunt type. Rüffer and colleagues noted greater survival in patients with an RVPA conduit until four months old and that, after four months, the risk of death increased markedly at their institution. Therefore, they advocated catheterization at two months post-Norwood and progression to S2P by four months.³⁶ This was a single-institution study that only included 58 patients, was analyzed non-parametrically and without risk adjustment. While our analysis is in accordance with their recommendation of performing S2P at four months in patients with RVPA conduits, our data does not support their conclusion that patients require distinct timing of S2P based on shunt type.

S2P does not rescue patients with risk factors

Three-year transplant-free survival was substantially decreased in high-risk patients undergoing S2P. Calculated transplant-free survival was often less than 50% when S2P was performed before six months post-Norwood. This survival cost was especially striking when high-risk patients (in whom survival after the Norwood was already compromised) required S2P at a Norwood-S2P interval less than three months, indicating the early S2P does not rescue ill patients in the post-Norwood period.

It is tempting to conclude that delaying S2P in patients with significant risk factor burden would maximize calculated three-year transplant-free survival. Although the curve depicting this type of patient in the nomogram in Figure 4 flattened past six months, this is likely due to length-time bias. Our data certainly does not suggest that non-intervention in a deteriorating, high-risk infant would eventuate in higher survival. By six months, most of the highest-risk patients will have either died or undergone heart transplantation, thereby increasing the calculated survival estimates. We hypothesize that the patients undergoing early non-elective S2P are qualitatively different than those who undergo non-elective S2P late. Those undergoing S2P prior to three months may have pathology that was not amenable to surgical intervention (e.g. ventricular dysfunction) versus those who undergoing S2P after six months (e.g. progressive cyanosis). We did not have data at a sufficiently granular level (oxygenation saturations, weights) to test this hypothesis.

Instead of recommending the avoidance of intervention in the ill post-Norwood patient who does not have pathology amenable to surgical intervention, prompt referral for heart transplantation may offer a greater chance at long term survival. These patients may be better served by cardiac replacement, given the failure of S2P as a rescue that we have demonstrated. Two recent studies suggest that acceptable outcomes are achievable in infants undergoing single ventricle palliation who require salvage heart transplantation.^{37, 38} Further formal analysis of this treatment pathway versus high-risk progression to S2P is warranted.

Determinants of the timing of S2P

This study addresses another knowledge gap in this field, the determinants of the timing of S2P. Patients with impaired ventricular function underwent S2P earlier, but not as soon as those with more acute anatomic issues such as shunt stenosis. Furthermore, a more complicated course after the Norwood also was associated with later progression to S2P, as a greater number of interstage operations was associated with a longer Norwood-S2P interval.

The intra-operative use of steroids during the Norwood was associated with longest increase in the Norwood-S2P interval. The use of intraoperative steroids during the Norwood procedure is discussed in greater detail in the Supplementary Results. A significant association between intraoperative steroid use and death or heart transplantation was seen in multivariable analysis. Our results harmonize with a recent study of intraoperative steroid use during the Norwood in the SVR Trial, in which intraoperative steroids were associated with greater hospital mortality in a multivariable logistic regression.³⁹ In a meta-analysis including 232 patients, intraoperative steroid use was not associated with significant reductions in morbidity or mortality in children undergoing cardiac surgery.⁴⁰ However, other published studies from the last decade present contradictory results.^{41, 42} The exact mechanism by which intraoperative steroid use may lengthen the time between the Norwood and S2P remains unclear.

The association of intra-operative aprotinin with later S2P was both statistically significant (p=0.007) and reliable (80%). No secondary analysis of aprotinin use in the SVR Trial has been performed. More detailed discussion can be found in the Supplementary Results. The drug was removed from the market in 2007 after several analyses suggested it was associated with greater rates of renal dysfunction in adults undergoing cardiac surgery on cardiopulmonary bypass.^43–45 Evidence for its potential link with renal dysfunction in congenital heart surgery is mixed. Four single-center observational studies from the last decade examined the effect of aprotinin in children undergoing cardiac surgery and their findings were evenly split. No difference in mortality was found in patients who did or did not receive the drug. ^46–49 Our multivariable analysis did identify an association between aprotinin use and death or heart transplantation after the Norwood. If the incidence of renal dysfunction with aprotinin was greater in the SVR Trial, then a longer ICU or hospital length of stay may have delayed S2P.

Finally, a site volume of 16–20 Norwood procedures per year was associated with a shorter Norwood-S2P interval. The SVR Trial public data set includes center volume of Norwood procedures per year as a variable with four levels, ≤15 per year (17%), 16–20 per year (20%), 21–30 per year (32%), or ≥30 per year (31%). The median Norwood-S2P intervals did differ by site volume, 5.3 versus 4.2 versus 5.1 versus 4.8 months, respectively (p< 0.0001). While the Norwood-S2P interval was shortest at sites with a yearly volume of 16–20 Norwood procedures per year, transplant-free survival was lowest at the sites performing 15 or fewer Norwood procedures per year (59% versus 75% versus 79% versus 73%, respectively, p=0.006) Site and surgeon volume-outcomes relationships with the Norwood procedure have been well described ⁵⁰. The reasoning behind earlier S2P at these institutions is not immediately clear.

Limitations

This study is subject to several important limitations. Although it utilizes the SVR Trial’s publicly available data set, it is a secondary analysis and uses the data set to answer a question for which it was not originally designed. In reality, the timing of S2P is determined currently by multiple factors, including institutional protocols, surgeon preferences, and patient factors, such as oxygen saturation, weight, and pulmonary vascular resistance (PVR) over time. In this data set, some potentially important variables were not present (e.g. weight at S2P), had only one measurement recorded (e.g. PVR on cardiac catheterization), or were represented using proxies, such as a diagnosis of failure to thrive. We were therefore unable to perform analyses using repeated measures of, e.g. weight, oxygenation saturations, or PVR over time. The SVR Trial data also lacked information regarding interstage home monitoring, which may also affect the timing of S2P based on specific clinical parameters. Finally, while the advantages of large, multicenter RCTs are obvious, they can also create an “artificial” environment to test a hypothesis. For example, nearly 90% of the patients included in the SVR Trial were diagnosed with HLHS. This strict enrollment criteria, while creating a homogenous patient population, also avoids the diagnostic and procedural complexity seen in the “real world” clinical practice in patients undergoing the Norwood procedure.

From a methodologic perspective, the analysis is likely complicated by informative censoring, as the children who were rapidly transitioned to S2P may have also been likely to die soon after the Norwood. This risk of death likely affected the time to S2P and could bias the conditional survival analysis. Although multistate modeling may represent an attractive alternative approach, current available methods (such as Cox Markov models) are not inadequate for the complexity of this analysis, as the rates of transition between the Norwood to S2P and death/transplantation are both competing and time-varying.

Clinical Implications

Several implications exist for the management of patients with few risk factors undergoing single ventricle palliation. Defining the timing of S2P associated with greatest three-year transplant-free survival allows for strategic planning after the initial Norwood procedure. At the time of discharge after the Norwood hospitalization, the adoption or modification of a formalized approach, in which S2P is explicitly planned and scheduled within three to six months of the Norwood, may lead to improved outcomes. Reducing variability in care across centers could lead to gains in survival through the staged management of single ventricle palliation. However, flexibility and perspective is needed when managing individual infants given that the interstage period is rife with risk and may necessitate interval procedures, hospitalizations, and more that may affect the timing of S2P. Further refinement that enables individual, patient-level predictions for optimal timing cannot be provided by this analysis, but merits additional investigation using a data source that can provide highly granular patient data, such as from the Congenital Heart Surgeons’ Society Data Center.

In addition to allowing for operative planning, our findings, particularly with regard to infants with multiple risk factors who are likely to fail single ventricle palliation can be used to counsel parents and ought to be incorporated into preoperative and prenatal consultations. The current practice of rapidly transitioning these patients out of Norwood circulation via early S2P resulted in especially poor transplant-free survival. This raises serious questions regarding this strategy. In these infants, early referral for heart transplantation ought to be considered, given the severely limited supply of organs for neonatal and infant transplantation. Further analyses are needed to weigh the potential benefits of early transplantation in these infants in light of the known scarcity of available organ donors in this age group.

Conclusions

In conclusion, in this analysis of the SVR Trial public dataset, the optimal timing for S2P is significantly influenced by patient characteristics and clinical status. For patients with few risk factors, scheduling S2P within three to six months of the Norwood at the time of discharge after the Norwood hospitalization may improve outcomes and reduce center-level variation in care. Survival in patients with greater risk factor burdens, especially those who require early S2P, will be severely adversely impacted. Progressing to S2P does not fundamentally change the risk profile of these patients. While earlier referral for heart transplantation may offer the greatest chance of long term survival for high-risk patients, the supply of available organs will limit this approach. The detailed study of outcomes in heart transplantation in poor candidates for staged single ventricle palliation and developing methods to increase the number of available organs represent the next important steps in maximizing long term survival in this precarious group of patients. Finally, the optimal timing did not differ by shunt type, regardless of risk profile.

Clinical Perspectives.

What is new?

• Mortality in infants with hypoplastic left heart syndrome undergoing single ventricle palliation remains high, especially prior to stage-2-palliation (S2P).

• The effect of the timing of the S2P – a physician-modifiable factor – is not well understood.

• In low- or average-risk infants, S2P between three and six months post-Norwood was associated with maximal three-year transplant-free survival.

• Optimal timing of S2P did not differ between patients who had Blalock-Taussig shunts or right-ventricle-to-pulmonary-artery conduits.

• Identifiable characteristics exist that severely compromise three-year transplant-free survival, irrespective of timing of S2P or shunt type. (91/100 words)

What are the clinical implications?

• Adoption of formal clinical protocols to ensure that operative planning for S2P in infants with few risk factors occurs prior to discharge after the Norwood procedure may lead to improved outcomes.

• In Infants at high risk of failing three-stage single ventricle palliation, the practice of proceeding rapidly to S2P after the Norwood incurs a major survival cost.

• Although organ donor supply is severely limited, early cardiac transplantation may provide these high-risk infants with their greatest chance at long-term survival. (82/100 words)