Introduction
Children with complex congenital heart anomalies resulting in single ventricle CHD with aortic arch hypoplasia face the highest morbidity and mortality of any patients with CHDs. Reference Boneva, Botto, Moore, Yang, Correa and Erickson1 These patients typically undergo a three-staged surgical palliation culminating in total cavo-pulmonary anastomoses and a Fontan circulation. The highest risk periods for these patients include the perioperative period surrounding the Stage 1 palliation, which occurs in the first few days of life, as well as the “interstage” period between the Stage 1 palliation and the Stage 2 palliation, which occurs at 4–6 months of age. Reference Tabbutt, Ghanayem and Ravishankar2–Reference Cross, Harahsheh and McCarter5 The Pediatric Heart Network and the National Pediatric Cardiology Quality Improvement Collaborative have spearheaded national research and quality improvement initiatives to optimise outcomes for single ventricle CHD and aortic arch hypoplasia, resulting in improved interstage survival, reduced morbidity with improved growth, and reduced unplanned hospital readmissions for major clinical events. Reference Anderson, Beekman and Kugler6–Reference Oster, Kellerman, McCracken, Ohye and Mahle10
The “NEONATE” score was developed by National Pediatric Cardiology Quality Improvement Collaborative investigators to risk stratify patients at the time of initial hospital discharge following the Stage 1 palliation, to reduce interstage mortality or heart transplant from the most recently reported incidence of 8–18%. Reference Ahmed, Anderson and Bates11 The score was derived from modelling performed on a subset of patients (Phase 1 learning cohort) in the Phase 1 National Pediatric Cardiology Quality Improvement Collaborative database and was validated internally in the remaining patients (Phase 1 validation cohort). The NEONATE risk stratification tool allocated points at the time of Stage 1 palliation discharge based on anatomic, surgical, and discharge characteristics (Norwood type—Norwood/Blalock–Taussig shunt: 3 points, Hybrid: 12 points; extracorporeal membrane oxygenation post-op: 9 points, Opiates at discharge: 6 points, No Digoxin at discharge: 9 points, Arch Obstruction on discharge echo: 9 points, Tricuspid regurgitation ≥ moderate on discharge echo: 12 points; Extra oxygen use plus ≥moderate tricuspid regurgitation: 28 points). A NEONATE score of ≤17 points as a cut-off defined patients as “low risk” predicted freedom from interstage mortality or heart transplant with good specificity.
Although the NEONATE score was validated in an independent subset of the Phase I National Pediatric Cardiology Quality Improvement Collaborative database, the score has not been evaluated in more modern cohorts (i.e., after 2016). We sought to evaluate the performance of the NEONATE score in the National Pediatric Cardiology Quality Improvement Collaborative Phase II database, which began enrolling infants with single ventricle CHD and aortic arch hypoplasia planned for Stage 1 palliation in 2016.
Methods
Data source
This study was an analysis of data collected prospectively on patients enrolled in the National Pediatric Cardiology Quality Improvement Collaborative Phase II registry between 2016 and 2020, as part of a national learning collaborative of over 60 surgical sites to reduce mortality and improve quality of life for infants with single ventricle CHD and aortic arch hypoplasia. The methods of subject recruitment and data acquisition and maintenance were identical to those employed in the Phase I registry and have been described previously. Reference Anderson, Beekman and Kugler6 The criteria for eligibility were broadened in the Phase 2 registry and are detailed below. After verification of eligibility, informed consent was obtained according to institutional guidelines at each National Pediatric Cardiology Quality Improvement Collaborative centre. Data were collected at fixed intervals and entered by each site into a Research Electronic Data Capture database, which was maintained by a central data management site (Cincinnati Children’s Hospital). After verification of eligibility by a participating centre, consent was obtained if required, and multiple data elements were collected from the time of the patient’s birth through discharge from their Stage 2 palliation operation or transplant. Data quality control was conducted using a combination of Research Electronic Data Capture system programmed edit checks and Statistical Analysis System reports. The programmed edit checks flag for discrepant data, out of range values, and incorrect data types in real time.
This study was performed according to a protocol approved by the Institutional Review Board of each participating institution, including the Committee for Clinical Investigation at Boston Children’s Hospital. The authors had full access to a de-identified version of data and take full responsibility for its integrity. All authors have read and agree to the manuscript as written.
Study population
The Phase 1 National Pediatric Cardiology Quality Improvement Collaborative registry only included patients with single ventricle CHD and aortic arch hypoplasia and aortic arch obstruction needing intervention with the intention to undergo Stage 1 palliation with a modified Blalock–Taussig shunt, right ventricle to pulmonary artery conduit, or hybrid approach; and survived to hospital discharge in the interstage period without requiring listing for a heart transplant. As referenced above, the population included in the Phase 2 registry was broadened to include all patients with single ventricle CHD with aortic arch obstruction, including those who underwent alternative initial surgical palliations, and those who either did not survive to Stage 1 palliation discharge or remained inpatient until the Stage 2 palliation. To reduce the impact of confounding variables, the analysis cohort of the Phase 2 validation was limited to those patients meeting the anatomic and surgical criteria defined for eligibility in the Phase 1 registry, who survived to hospital discharge in the interstage period. As in the NEONATE development study, patients were also excluded from the analysis if lost to follow-up, transferred to a non-National Pediatric Cardiology Quality Improvement Collaborative facility, or were converted to a biventricular repair strategy.
Study design and statistical analysis
The primary outcome was a composite endpoint of death or transplant after initial hospital discharge post-Stage 1 palliation.
Student’s t-test was used to compare normally distributed data of patients undergoing Stage 2 palliation versus death or transplant and is presented as mean ± standard deviation. A Wilcoxon rank-sum test was used to evaluate the non-normally distributed variables, with data presented as median (interquartile range). Categorical unordered and ordered variables were analysed with a Fisher exact test and Mantel–Haenszel test for linear trend, respectively, to assess association with interstage mortality. Two-tailed p < 0.05 was considered statistically significant.
The statistical methods used for derivation and weighting of the risk factors identified in the learning cohort of the NEONATE derivation study have been described previously. Reference Ahmed, Anderson and Bates11 Briefly, this included a classification and regression tree analysis to identify potential interactions among variables that define particularly high- or low-risk subgroups. The identified high-risk classification and regression tree variable, along with all variables with a p value <0.2 in univariate logistic regression analysis, were included as candidate predictors for the multivariable analysis. Using the learning cohort point cut-off of ≤17 points, sensitivity, specificity, positive predictive value, negative predictive value, and accuracy (true positive and true negative results) were calculated in the present, external validation cohort.
Results
Patient characteristics
Of the 1806 patients from 60 participating surgical centres in the National Pediatric Cardiology Quality Improvement Collaborative Phase 2 registry who were eligible for inclusion, 104 patients did not yet have Stage 1 palliation information and were thus excluded from the analysis cohort. An additional 481 patients were excluded because they died (n = 124) or were referred to heart transplant (n = 18) prior to being discharged post-Stage 1 palliation, did not have follow-up information after the Stage 1 palliation (n = 137), or remained inpatient until Stage 2 palliation (n = 202). Finally, an additional 195 patients were excluded because of a non-qualifying primary diagnosis or first surgery, conversion to a biventricular pathway, loss of candidacy for Stage 2 palliation, loss to follow-up including transfer to a different centre, or study withdrawal. The remaining 1026 patients comprised the analysis cohort, whose median gestational age and birthweight were 39 (IQR 38,39) weeks and 3.2 (2.9,3.5) kg, respectively. Four hundred (39%) were female and 721 (70%) were white. The most common native cardiac anatomy was a form of hypoplastic left heart syndrome (n = 802, 75%); 4.7% (n = 48) of patients had a major anomaly of at least one additional organ system. At birth, the atrial septum was restrictive in 11% (n = 111) of patients, and the prevalence of ≥ moderate tricuspid regurgitation (n = 20, 1.9%) or ≥ moderate ventricular systolic dysfunction (n = 15, 1.5%) on the initial postnatal echocardiogram was low.
Interstage events
The overall composite outcome event rate was 6%, with 44 (4.3%) interstage deaths, and 17 (1.7%) orthotopic heart transplants. Baseline patient characteristics by primary outcome are detailed in Table 1. In univariate analysis, the subgroup of patients who experienced interstage mortality or heart transplant had a lower mean gestational age and birthweight, along with more female patients, compared to their counterparts who survived to Stage 2 palliation. There was no difference in the primary anatomical diagnosis between the groups, but more patients who experienced interstage mortality or heart transplant had anomalous pulmonary venous return of any sort and a genetic diagnosis of 22q11 deletion.
Table 1. Pre-operative patient characteristics by primary outcome
IM/HTx = interstage mortality/heart transplant; S2P = Stage 2 palliation; MA = mitral atresia; AA = aortic atresia; MS = mitral stenosis; AS = aortic stenosis; NOS = not otherwise specified; AV = atrioventricular; DORV = double outlet right ventricle; PV = pulmonary venous; TR = tricuspid regurgitation.
Bold values are the significant p values.
There was an association between type of Norwood procedure and rate of interstage mortality or heart transplant, with an incidence of 16.7%, 5.9%, and 4.7% for patients who had undergone a Hybrid procedure, modified Blalock–Taussig shunt, and right ventricle to pulmonary artery conduit, respectively (Table 2). Institutional standard was the most frequently reported indication for undergoing a Hybrid procedure. However, more patients who underwent a Hybrid were premature (defined as a gestational age <37 weeks) compared to those who did not [16/70 (22.8%) versus 68/942 (18.2%), p < 0.0001]. Undergoing a Hybrid procedure was also associated with a lower gestational age at birth [median 38 (IQR 37,39) versus 39 (IQR 38,39), p = 0.002]. The median birthweight was not different between the two [Hybrid 3.05 kg (IQR 2.75,3.54) vs not-Hybrid 3.19 kg (2.89,3.15), p = 0.0502]. However, when assessing low birthweight (<2.8 kg) as a categorical variable, more patients who underwent a Hybrid had a low birthweight [16/70 (22.8%) versus 168/922 (18.2%), p = 0.008]. These data do suggest that prematurity and low birthweight are associated with selection of the Hybrid procedure for the Stage 1 palliation.
Table 2. Stage 1 palliation procedural and post-operative characteristics by outcome
S2P = Stage 2 palliation; S1P = Stage 1 palliation; RV-PA = right ventricle to pulmonary artery; BT = Blalock–Taussig; NOS = not otherwise specified; V = ventricle; vs = versus; IQR = interquartile range; NEC = necrotising enterocolitis; G-tube = gastrostomy tube; ECMO = extracorporeal membrane oxygenation.
The most common post-operative complication was an arrhythmia requiring drug therapy, occurring in about ¼ patients; the incidence did not differ based on primary outcome. Compared to their peers who survived to Stage 2 palliation, more interstage mortality or heart transplant patients required post-operative extracorporeal membrane oxygenation (18 versus 6%, p = 0.001), re-operation for aortic arch repair (4.9 versus 0.9%, p = 0.005) or a tracheostomy (3.3% versus 0.4%, p = 0.004), or experienced a post-operative neurologic deficit (4.9 versus 0.3%, p < 0.001) or acute renal failure (9.8 versus 1.3%, p < 0.001) requiring dialysis (8.2 versus 0.9%, p < 0.001).
At initial hospital discharge following Stage 1 palliation, patients were in the second month of life, weighed 3.8–3.9 kg, and had a median oxygen saturation of 81–82% (Table 3). A smaller percentage of patients who experienced interstage mortality or heart transplant had been prescribed digoxin (61% versus 70%) at discharge, and a greater percentage had been prescribed opiates (8.2 versus 4.7%), although neither difference was statistically significant. There was a greater prevalence of moderate-to-severe tricuspid regurgitation on the pre-discharge echocardiogram in patients who experienced interstage mortality or heart transplant compared to those who survived to Stage 2 palliation (27.9 versus 15.9%, p = 0.020). This difference was compounded in the subset of patients with moderate-to-severe tricuspid regurgitation who were discharged home on supplemental oxygen, with an ∼8 times higher event rate in this group (4.9 versus 0.6%, p < 0.001). There was no difference in the prevalence of pre-discharge clinical or echocardiographic arch obstruction, or moderate-to-severe single ventricular systolic dysfunction or atrial septal defect restriction on the pre-discharge echocardiogram (Fig 1).
Figure 1. NEONATE factors that remain associated with interstage mortality or heart transplant in the Phase 2 cohort.
Table 3. Stage 1 palliation discharge characteristics by primary outcome
S2P = Stage 2 palliation; MA = mitral atresia; IQR = interquartile range; TR = tricuspid regurgitation; d/c = discharge; ASD = atrial septal defect.
NEONATE score
The median NEONATE score was significantly lower in patients who survived to Stage 2 palliation [0 (0,9)] compared to their counterparts who experienced interstage mortality or heart transplant [9, (0,15); p < 0.0001). The rate of interstage mortality or heart transplant increased with increasing NEONATE score, ranging from a low of 3.8% in patients with a score of 0 to a high of 33.3% in patients with a score >30 (Fig 2). The highest score observed in the entire cohort was 49 points, out of a maximum possible 61 points. Applying the NEONATE score cut-off of 17 points that was optimal in the learning cohort for separating patients into low and high-risk groups provided 91% specificity, a negative predictive value of 95%, and overall accuracy of 87% (Fig 3).
Figure 2. The rate of IM/HTx increases exponentially with increasing NEONATE score. IM/HTx = interstage mortality/heart transplant.
Figure 3. Predictive performance and rate of interstage mortality or transplant after hospital discharge, using a NEONATE score of 17 points as a primary stratification of risk.
Discussion
In this multi-centre registry study, we successfully validated the NEONATE score using an external, modern-day cohort. In the 10 years between the creation of the National Pediatric Cardiology Quality Improvement Collaborative Phase 1 and 2 registries, the rate of interstage mortality or referral for heart transplantation decreased ∼33% from a rate of 9% in the Phase 1 registry to 5.9% in the Phase 2 registry (p = 0.002). Overall, patients in the Phase 2 cohort who were discharged home following Stage 1 palliation had a lower median NEONATE score (0 points) compared to those in the Phase 1 learning cohort (17 points).
The significant differences seen in the rate of interstage mortality or heart transplant and median NEONATE score between the Phase 1 and Phase 2 cohorts may have been due in part to rapid, national practice changes at National Pediatric Cardiology Quality Improvement Collaborative centres following the derivation and internal validation of the NEONATE score. Reference Ahmed, Anderson and Bates11 This is supported by a decrease in the percentage of Phase 2 validation patients who were discharged home with modifiable risk factors identified in the initial publication. Phase 2 patients were more likely to be discharged home on digoxin (70% versus 30%, p < 0.00001), less likely to be discharged with opiates (4.9% versus 7.7%, p = 0.004), and less likely to have arch obstruction on the pre-discharge echocardiogram (2.2% versus 7.8%, p < 0.0001) compared to their counterparts in the Phase 1 learning cohort. A similar and related phenomenon occurred following the seminal National Pediatric Cardiology Quality Improvement Collaborative publications that suggested an association between digoxin use and improved interstage survival. Reference Brown, Mangeot and Anderson9–Reference O’Byrne, Song and Huang15
In contrast, there were no changes in other, less modifiable factors. A similar percentage of patients in both cohorts had undergone post-operative extracorporeal membrane oxygenation (4.6 versus 7.6%, p = 0.600) and were discharged with supplemental oxygen in the setting of moderate-to-severe tricuspid regurgitation (0.9 versus 1.1%, p = 0.500). The breakdown of Stage 1 palliation type was also similar between the two groups. Notably, a greater percentage of patients in the validation cohort had moderate-to-severe tricuspid regurgitation on the pre-discharge echocardiogram (16.6% versus 10.7%, p < 0.001).
Clinical implications
The NEONATE score is the first and only risk score for patients with single ventricle CHD and aortic arch hypoplasia with aortic arch hypoplasia in the interstage period, and it remains a useful tool for identifying patients at the lowest risk for interstage events, even in a modern, external cohort of patients. The present data confirm that a NEONATE score ≤17 points can be used as threshold to identify patients at lower risk for interstage events with high specificity, accuracy, and negative predictive value. The data continue to indicate a higher incidence of interstage events in patients with moderate-to-severe tricuspid regurgitation at the time of hospital discharge, particularly in those discharged on supplemental oxygen. While certainly not simple, tricuspid regurgitation in the single ventricle patient presents an important target for future endeavours to further decrease interstage mortality. This could be achieved by a multi-pronged approach that includes innovations in neonatal valve repair, earlier referral to advanced heart failure and transplant evaluation, and/or reconsideration of the safety of discharging these patients prior to the Stage 2 palliation.
In our original NEONATE score publication, we had considered that interstage mortality could be decreased further by addressing the modifiable risk factors identified in the model – particularly, discharge of patients on opiates and digoxin. While it may have been unrelated, it is interesting to note the association between the lower rate of discharge with opiates and the higher rate of discharge with digoxin, and the overall decrease in interstage mortality. While performing a prospective randomised controlled trial would be the gold standard for a definitive evaluation of the effect of digoxin in the interstage period, a number of factors challenge the feasibility of this type of investigation, including the rarity of the disease, the short duration (generally less than 6 months) of the interstage period, the relative low frequency of the composite outcome, and thus the large number of patients who would need to be enrolled to detect differences in mortality. Furthermore, it may be unethical to withhold this generally well-tolerated medication in such a vulnerable population. To examine the effect of opiates, one may alter practice to ensure that patients remain hospitalised until they have been weaned off opiates and then perform an analysis of mortality using previous studies as a historical control.
Limitations
This registry study should be viewed considering its limitations. As with all multi-centre registries, it was not possible to account for inherent intra- and inter-centre variability in operative technique and standards of post-operative care. An important aspect of this is significant variability in which patients are discharged home at all in the interstage period. Given that this study only includes patients who were discharged home post-Stage 1 palliation, this may introduce significant exclusion and/or inclusion bias, by virtue of a subset of the sickest patients potentially being excluded from the analysis and derivation of the NEONATE score. Similarly, the association between interstage mortality or heart transplant and the hybrid procedure is difficult to assess given that at most National Pediatric Cardiology Quality Improvement Collaborative centres, this strategy is reserved for premature and/or sick patients. The patients who experienced interstage mortality or heart transplant prior to hospital discharge could not be evaluated in this study because no post-Stage 1 palliation medication or echocardiographic data were available in this subset.
Some variables that may impact clinical outcomes, such as measures of socio-economic status, were not collected in either phase of the database. Finally, full reports of diagnostic and interventional catheterisations were not available, and some of the variables collected in our dataset – particularly related to echocardiographic findings – included subjective measures of severity. This would have been mitigated by using a core imaging laboratory with a single reader, which was not employed in the National Pediatric Cardiology Quality Improvement Collaborative registry. However, some bias was eliminated since all echocardiographic reports were completed prior to hospital discharge, thereby blinding readers to a patient’s ultimate outcome.
Conclusion
In this modern-day validation cohort, infants with a low NEONATE score at the time of initial hospital discharge following the Stage 1 palliation experienced a low rate of interstage mortality or heart transplant. The NEONTE score should be used as an adjunctive, risk stratification tool at the time of initial hospital discharge post-Stage 1 palliation for patients with single ventricle CHD and aortic arch hypoplasia and aortic arch hypoplasia, to predict freedom from interstage mortality or heart transplant.
