Ventricular assist device (VAD) is a standard therapy to provide long-term mechanical circulatory support in patients with end-stage heart disease as either a bridge to heart transplantation or an alternative to transplantation that could improve the functional status, quality of life, and survival of the recipients. Over the past decade, there have been significant advancements in VAD technology and growing overall experience with VAD support in the pediatric population that yielded greatly improved survival.

Previous research based on adult recipients indicated that appropriate timing and candidate selection are extremely critical for effective use of mechanical circulatory support (1- 2). In the literature, several pediatric VAD studies have reported risk factors for the post-operative survival of children (3- 4). Nevertheless, the patients included in these studies were mainly adolescents with adult-size devices implanted, and the number of patients receiving long-term VAD support was relatively small, which might limit the generalizability of their results to the pediatric population. More important, these studies included a limited analysis of patient characteristics; therefore, many potentially important pre-operative risk assessment parameters might have been missed. Because of the increased use of VAD support in pediatric patients with advanced heart failure, risk factors and clinical indications for VAD implantation in this population need to be further clarified. In the present study, we retrospectively analyzed our pediatric VAD experience to define associated risk factors after VAD implantation in pediatric patients with advanced heart failure.

Pediatric subjects 18 years of age and younger who underwent an initial implantation of a long-term VAD at Germany Heart Institute Berlin between June 1996 and December 2009 (N = 92) were identified retrospectively from a computerized database. All these children had end-stage heart failure at the time of VAD insertion and were supported with intravenous inotropic agents or low-output syndrome developed after a cardiac surgery that required subsequent implantation of a VAD to separate them from cardiopulmonary bypass. All parents of the patients provided written informed consent for device implantation. The study population was followed until heart transplantation, myocardial recovery, or death on mechanical support. Approval by the ethics committee of the present study was not required, because the retrospective data analysis did not vitiate human rights, according to the German law.

During the study period, multiple VADs were inserted in the pediatric patients with advanced heart failure. Implanted pumps included the Berlin Heart EXCOR VADs (Berlin Heart GmbH, Berlin, Germany) (n = 87, 95%), the Berlin Heart Incor VADs (Berlin Heart GmbH, Berlin, Germany) (n = 2, 2%), the Novacor VADs (World Heart Inc., Oakland, California) (n = 2, 2%), and the HeartWare VAD (HeartWare Inc., Miramar, Florida) (n = 1, 1%). For small infants and young children, the Berlin Heart EXCOR pediatric VADs were implanted, which were designed specifically to provide mechanical circulatory support in all age groups of the pediatric population. The adult-sized VADs (Incor, Novacor, and HeartWare) were only used in the fully grown adolescent patients. Detailed information concerning the surgical procedure of VAD insertion, implantation and weaning criteria, and post-operative anticoagulation therapy was described in previous reports (5- 7).

Because most of the in-hospital deaths occurred during the first 3 months after VAD implantation, we chose 90-day in-hospital mortality as the primary endpoint for the risk factor analysis (data are shown in the Online Appendix). Subsequently, subjects were categorized into 2 groups: children who survived during the first 3-month hospitalization after VAD implantation (n = 62) and those who did not (n = 30). Patients' medical records were retrospectively reviewed, and the demographic, clinical, echocardiographic, hemodynamic, and laboratory data were obtained and assessed. Details of data collection are shown in the Online Appendix. Data were analyzed with PASW Statistic 18 (SPSS Inc., Chicago, Illinois). Categorical variables were described by percentages and frequencies, and numerical variables were represented as mean ± SD and/or median range(s). Qualitative data were compared using the chi-square or Fisher exact test, as appropriate. For continuous data, between-group comparisons were performed by either the Mann-Whitney or the Student t test, depending on the normality of each variable. All statistical comparisons were 2 sided. The variables that were significant by univariate analysis were used as potential risk factors for the stepwise multivariable model, with an entry criterion of a p value ≤0.1; receiver-operating characteristic curves were applied to identify cutoff values for continuous variables in this series that best predicted post-VAD mortality. The Kaplan-Meier method was applied to estimate survival probability for children continuing on mechanical support, and log-rank analysis was performed to assess the statistical significance of survival difference. Patient survival was calculated from the day of VAD implantation until death on mechanical support, censoring the time of heart transplantation or recovery of the native heart. Competing outcomes methods were applied to analyze outcome after implantation.

From June 1996 to December 2009, 44 girls and 48 boys were implanted with a long-term VAD at Germany Heart Institute (Table 1). At the time of VAD insertion, the median age was 7 years (mean 7.4 years; range 12 days to 18 years). The median weight was 18.8 kg (mean 27.5 kg; range 3.22 to 115 kg), and the mean body surface area was 0.91 m² (median 0.80 m²; range 0.2 to 2.38 m²). Of the 92 pediatric patients, 66 (72%) had a diagnosis of cardiomyopathy, 21 (23%) congenital heart disease (CHD), and 5 (5%) acute myocarditis. The diagnoses of CHD included transposition of the great arteries (n = 10), complete atrial ventricular septal defect (n = 3), anomalous left coronary artery arising from the pulmonary artery (n = 2), single ventricle (n = 2), Ebstein's anomaly (n = 1), hypoplastic left heart syndrome (n = 1), interrupted aortic arch (n = 1), and atrial septal defect, ductus arteriosus with left ventricular tumor (n = 1).

Of the 92 patients analyzed, 42 patients were successfully bridged to heart transplantation, and 16 were explanted after myocardial recovery. The overall survival rate to transplantation or recovery of ventricular function was 63%. The Kaplan-Meier estimates of survival on mechanical support were 76 ± 4.7% and 56.7 ± 6.4% at 30 days and 6 months after VAD implantation, respectively (Figure 1). Fifty-eight children were finally discharged from the hospital. Of the 34 deaths in patients with a VAD, 32 patients died before hospital discharge (34.8% in-hospital mortality), and 93.8% of these (30 of 32) occurred within 90 days of VAD implantation (Figure 2). Among these children, the main causes of post-operative deaths included multisystem failure in 11 children (36.7%), stroke in 5 (16.7%), bleeding in 4 (13.3%), circulatory failure in 3 (10%), respiratory failure in 3 (10%), infection in 2 (6.7%), and other causes in 2 (6.7%).

The median time of mechanical circulatory support was 35 days (mean 67 days; range 1 to 591 days). Post-operative adverse events included infection in 34 patients (37%), bleeding requiring reoperation in 24 (26.1%), stroke in 22 (23.9%), pump exchange in 19 (20.7%), and renal failure in 17 (18.5%).

In the present study, 43 patients younger than 5 years of age were on VAD support. The median weight was 7.2 kg (mean 7.8 kg; range 3.2 to 18 kg), and the median body surface area was 0.38 m² (mean 0.39 m²; range 0.20 to 0.72 m²). Eighteen patients were successfully bridged to transplantation, and 12 were weaned from VAD support with cardiac recovery. The overall survival rate to transplantation or recovery of ventricular function was 69.8%, which was similar to that in older children. We did an additional analysis by only including patients weighing <10 kg; a total of 22 of the 32 children (68.8%) successfully survived during the support period.

Twenty-one patients with CHD underwent long-term VAD implantation. Of the 21 CHD patients, 15 died before VAD explantation (71%) compared with 19 of the 71 children with a noncongenital etiology (27%). The 30-day and 6-month survival rate was 86.1 ± 4.3% and 68.9 ± 6.8% for patients with noncongenital heart disease and 40.2 ± 12.2% and 16.1 ± 10.1% for children with CHD, respectively. Compared with patients with CHD, noncongenital recipients had a significantly better survival rate (p < 0.0001).

Thirty-two patients received biventricular support with a biventricular VAD (BVAD). During mechanical support, 16 of 32 patients with a BVAD (50%) and 18 of 60 (30%) of patients with a left ventricular assist device (LVAD) died. The Kaplan-Meier estimates of 30-day and 6-month survival were 82.9 ± 5.3% and 67.8 ± 7.5% for LVAD implantation, respectively, and 63.5 ± 8.9% and 37.9 ± 10.5%, respectively, for children requiring biventricular support (p = 0.02).

Nine patients were supported by extracorporeal membrane oxygenation, 4 underwent implantation of an intra-aortic balloon pump, and 4 received temporary support from centrifugal pumps as a bridge to VAD. Four of them were bridged to heart transplantation, 4 were weaned from VAD support with myocardial recovery, and 9 patients died before VAD explantation. The overall survival rate to transplantation or myocardial recovery was 47.1%.

(Table 2) shows the patient characteristics and pre-operative medication use of the study population. When dichotomizing the etiology of heart failure into congenital or noncongenital (cardiomyopathy or myocarditis), the incidence of in-hospital death for patients with CHD was significantly higher than for those without congenital lesions (p < 0.0001). Previous cardiac surgery was associated with significantly increased in-hospital mortality (p = 0.001). There was significantly increased mortality after VAD insertion in children receiving pre-operative intravenous norepinephrine infusion (p = 0.008). Additionally, death during mechanical support was less common in patients supported with pre-operative milrinone (p = 0.04).

In (Table 3), the results for comparisons of clinical signs and hemodynamic indexes between the 2 groups are presented. Among those pre-operative hemodynamic variables, elevated central venous pressure (CVP) was the single parameter that portended an increased risk of in-hospital mortality (p = 0.02). Several hemodynamic parameters could vary according to different age (such as systematic blood pressure, heart rate, and cardiac output). An additional analysis was done by separately assessing these factors in infants (younger than 1 year), younger children (1 to 5 years of age), and older children (5 years of age and older). The effects of the variables were not statistically significant in each subgroup.

(Table 4) summarizes the laboratory profile of the study cohort. There was increased mortality in children with a higher C-reactive protein (CRP) level (p = 0.04). There was no significant difference in other laboratory parameters between the survivor and nonsurvivor groups.

Variables that were univariately significant were further entered into a multivariable logistic regression model. Independent predictors associated with increased mortality in pediatric patients after VAD insertion included congenital etiology (odds ratio [OR]: 11.2; 95% confidence interval [CI]: 2.6 to 47.5; p = 0.001), a norepinephrine requirement (OR: 6.9; 95% CI: 1.4 to 31; p = 0.02), CRP level >6.3 mg/dl (OR: 4.9; 95% CI: 1.1 to 22.1; p = 0.041), and CVP >17 mm Hg (OR: 4.6; 95% CI: 1.1 to 20; p = 0.04) (Table 5). The Hosmer and Lemeshow test showed that this model provided good fit with the data (p = 0.89). The multicollinearity analysis indicated that our risk factors were not confounded by the multicollinearity between these parameters (variance inflation factor ranged from 1.02 to 1.07).

The present analysis demonstrated that children with severe right heart failure needing BVAD implantation had higher post-VAD mortality than those who did not. Decreased milrinone use (p = 0.001) and increased dose of epinephrine use (p < 0.001) were associated with requirement for BVAD support. The study period spans 13 years, during which significant improvement in the field of pediatric mechanical circulatory support has been achieved, which might also cause potential bias to our analysis. Furthermore, pump size, CHD, and patient age/size tended to be correlated in pediatric patients; thus, potential confounding between these factors could exist. We evaluated the possible confounding by these factors in logistic regression models that included each of these covariates (biventricular support, implantation year, and pump size). All of the identified risk factors remained statistically significant after adjustment for the potential confounding factors, and the effects of these confounding factors were not significant in the multivariate analysis.

Optimal patient selection for VAD therapy is the key to improved survival in patients with end-stage heart disease. Although many risk factors have been previously reported based on adult recipients, the predictors of post-VAD survival in children have not been adequately defined. Compared with previous pediatric VAD reports from either our institution or other clinical heart centers in the world, the current study included the largest number of children on long-term VAD support, and we have collected more complete patient profiles aiming at identifying potential factors that were closely correlated with outcomes of pediatric VAD surgery (3- 8). Our analysis showed that pre-operative characteristics associated with increased risk of in-hospital death were the congenital etiology, pre-operative norepinephrine requirement, CRP level >6.3 mg/dl, and CVP >17 mm Hg.

Reports from the International Society for Heart and Lung Transplantation and other multicenter studies have demonstrated significantly worse outcomes in children with CHD undergoing heart transplantation compared with other pediatric patients (9- 10). Similarly, in this study, the presence of a congenital diagnosis was associated with the highest negative impact on survival in children after VAD implantation, with a 67% in hospital mortality rate in children with CHD compared with 23% in those with a noncongenital etiology. Others have found similarly poorer outcomes in CHD patients receiving VAD support (3- 4). There might be several reasons for the reduced survival in this patient group: first, the pre-operative cardiac lesions and physiological states of CHD patients could be more complex and considerably worse than those in the noncongenital population. In addition, children with CHD are more likely to have received repeated surgical interventions to repair cardiac anomalies before VAD implantation, thus increasing the operative risk encountered on repeat sternotomy.

The pre-implantation need for vasopressor support could be a reflection of more advanced hemodynamic instability in these patients. Currently, the American College of Cardiology/American Heart Association guidelines recommend dopamine and norepinephrine as the first-line vasopressor agents in the treatment of cardiogenic shock and the use of norepinephrine, in particular, for only the most severely hypotensive patients; thus, children requiring norepinephrine administration could have a worse pre-operative circulatory status and more advanced stage of disease compared with others (11). A previous study also showed that a higher dose norepinephrine requirement was associated with significantly increased mortality in patients with refractory cardiogenic shock (12).

CRP, one of the most sensitive markers of inflammation and infection, has also been recently indicated as a potential risk factor for cardiovascular disease in pediatric and adult patients (13- 15). A subgroup analysis of the Framingham study demonstrated that older patients with serum CRP levels ≥5 mg/dl experienced significantly increased risk of chronic heart failure (14). In pediatric patients with heart failure, an elevated CRP level was reported to be correlated with more severe symptoms and worse cardiac characteristics (15). The 2 recent cohort analyses including adult VAD recipients both showed that higher CRP levels were associated with significantly increased post-VAD mortality (16- 17). Consistent with their findings, our study highlighted increased CRP as an incremental predictor for death in children after VAD implantation.

Increased CVP could be a reflection of the worsening right ventricular function in patients with cardiovascular disease and has been reported to be predictive of the development of post-implantation right heart failure. Right heart failure in adults and children is associated with high post-VAD morbidity and mortality, likely due to the impact of a dysfunctional right ventricle on hepatic and renal function and, therefore, overall operative risk. In this analysis, pediatric patients with increased CVP had a significantly higher risk of death. Corroborating our finding, a cohort study of adult VAD recipients by Rao et al. (18) also identified increased CVP as an independent predictor of post-VAD mortality.

Although currently there are no consensus guidelines for pediatric VAD implantation, it should be used selectively to ensure reasonable benefits for the associated risks of this therapy. In the present analysis, we identified several pre-operative risk factors closely related to in-hospital death that might provide a valuable reference for candidate selection and outcome prediction in pediatric VAD recipients. Although no 1 predictor is a contraindication of device insertion, children with these factors might have higher mortality exceeding expectation, which should be carefully considered by physicians before making a decision regarding implantation, and this would be also useful during the pre-operative family education encounter. Previous adult VAD studies have indicated that healthier candidates without evidence of severe functional impairment and end-organ failure could obtain better outcomes after VAD implantation; the viewpoint was also confirmed in our study. In pediatric patients with end-stage heart failure, complex CHD and increased CVP are often associated with seriously impaired right ventricular, pulmonary, hepatic, and renal functions. Norepinephrine requirement and increased CRP level represent more advanced circulatory failure and worse clinical status, which might cause subsequently irreversible organ shock sequelae due to malperfusion. In our experience, earlier VAD implantation has resulted in significantly better survival in children. Thus, one of the most important aspects of pediatric VAD recipient selection might be that physicians should choose the device insertion before further deterioration of a patient's condition. Postponing VAD surgery until progression to end-organ failure might significantly worsen the outcomes after implantation. Stabilizing patients' circulation and restoring organ function using aggressive medical treatment or temporary mechanical support might render a child with a serious condition suitable for VAD operation. Additionally, for patients with increased pulmonary vascular resistance and impaired right heart function, milrinone might need to be considered to avoid additional right VAD insertion, which was also shown to be associated with higher post-implantation mortality.

Notably, in contrast to previous pediatric VAD reports, age and age-dependent factors were not predictive of post-VAD survival in our study on univariable or multivariable analyses (3). In our cohort, approximately 70% of the children younger than 5 years of age or those weighing <10 kg could successfully survive during the mechanical support period. The mortality is much lower than that in previous studies using mostly adult-sized devices, which could lead to increased hemorrhage and thromboembolic complications in small patients due to device mismatch. Our study suggests that, with the pediatric-specific VAD, younger children could enjoy comparable benefits from VAD implantation as do older adolescents.

First, this was a single-center, retrospective cohort study and therefore subject to inherent deficiencies. Additionally, the relatively small patient sample limits study power: in our analysis, comparatively more variables were submitted to the multivariate logistic regression analyses, and, thus, some degrees of “overfitting” might exist; furthermore, unadjusted p values provided for univariable analysis should be interpreted in the context of risk for a type I error. However, we believe that Bonferroni p value adjustment in this exploratory analysis may lead to a dismissal of clinically relevant variables (19). In addition, in pediatric studies of long-term mechanical circulatory support, power limitations are inherently difficult to avoid due to the current low device use rates. Finally, the present study included children supported with LVAD and BVAD, which might cause bias in our analysis. However, the strategy in our center for BVAD implantation is to first implant an LVAD combined with pharmacological right heart support and then decide whether there is a need for an additional right VAD, according to the performance of right heart function; thus, the majority of our children underwent LVAD implantation initially. Further analysis has shown that the statistical significance of our risk factors was not altered after adjustment for this confounding factor, and the effect of this variable in the regression model was not significant. In addition, previous reports indicated that risk factors for outcomes in BVAD patients were similar to those in LVAD recipients (20- 21).

The finding of this study suggests that VAD mortality is increased in children with CHD, pre-operative requirement for norepinephrine, CRP level >6.3 mg/dl, and CVP >17 mm Hg. Appropriate patient selection might be critical to improved outcomes after VAD implantation in children with advanced heart failure.

For supplemental material, please see online version of this article.