For children admitted with symptomatic heart failure (HF), the risk of death/need for transplantation (D/Tx) is high. Data from adult studies suggest serum measurements, such as percent lymphocytes, are valuable predictors of outcomes. The aim of this study was to identify risk factors for D/Tx in hospitalized pediatric patients with symptomatic HF. Retrospective analysis of children admitted to an academic center from January 1994 to June 2008 with clinical HF was undertaken. The most common cause of HF was dilated cardiomyopathy (58 of 99, 59%). Echocardiographic and serum measurements were collected from admission. Factors independently associated with risk of D/Tx were evaluated by a stepwise multivariate Cox regression model. There were 99 children with 139 hospitalizations. Median age at admission was 3 years (range 0 to 22). Mean systemic ventricular ejection fraction was 23% ± 11. Risk of D/Tx per hospitalization was 60 of 139 (43%). In multivariate analysis, lymphocytopenia, lower ejection fraction, low serum sodium, and higher serum creatinine were independent predictors of D/Tx. These variables correctly predicted those subjects at risk of D/Tx in 82.1% of cases. Subgroup analysis found that brain natruretic peptide did not improve the model’s accuracy markedly. In conclusion, serum measurements (percent lymphocytes, sodium, and creatinine) and echocardiographic assessment routinely obtained at admission are predictive of D/Tx in children hospitalized for HF. Significant lymphocytopenia was predictive of adverse outcomes.
Pediatric heart failure (HF) is a rare but serious entity, with an annual incidence of approximately 6 in 1,000,000. Mortality or risk of transplantation (D/Tx) is as high as 50% at 10 years after diagnosis. Data from adult HF studies, such as the Seattle Heart Failure Study, have also shown that serum parameters can be as important as echocardiography in predicting outcomes. Although adult models are often extrapolated to pediatrics, there are significant differences between the natural history of HF between these populations. In light of limited pediatric data, we examined the value of routinely obtained serum and echocardiographic measurements in predicting short-term risk of D/Tx. An understanding of these variables could aid in management of this high-risk population.
Methods
The study was approved by the Children’s Healthcare of Atlanta (Atlanta, Georgia) institutional review board. We undertook a retrospective chart review of children with moderately to severely impaired ventricular function admitted to Children’s Healthcare of Atlanta from January 1994 to June 2008 with symptomatic HF. We collected variables including age, symptomatology, and echocardiographic and serum measurements obtained within the first 24 hours of admission. The primary outcome variable (D/Tx) was defined as death or transplantation during the hospitalization.
Inclusion criteria were (1) symptomatic HF (e.g., dyspnea, edema, abdominal pain, emesis, and failure to thrive) and an echocardiographically determined ejection fraction (EF) <0.40 or qualitatively “moderately” or “severely” decreased function when ventricular geometry precluded EF calculation; (2) age <18 years at presentation; and (3) admission to Children’s Healthcare of Atlanta from January 1994 to June 2008. Exclusion criteria included a history of malignancy, primary renal disorders, restrictive cardiomyopathy, previous heart Tx or listing, glycogen storage disorders, inborn errors of metabolism, lack of serum measurements, or sepsis at admission.
Consistent with the natural history of HF, several patients were admitted with HF on multiple occasions. Each admission was considered an independent event that could end in D/Tx. Studies routinely obtained in our subjects at admission included a complete blood count with manual differential, a complete metabolic panel, and an echocardiogram. Tissue Doppler measurements of myocardial function were incorporated into the routine examination in 2003 and, hence, are not included in the analysis. Serum parameters obtained at admission also varied over time. After 2001, brain natruretic peptide (BNP) was frequently obtained at admission for symptomatic patients. These encounters were included in a separate subgroup analysis.
Each covariate and categorical, continuous, and transformed univariate variables were entered into a stepwise forward multivariate Cox model using a probability value of 0.10 for inclusion or deletion. The odds ratio and confidence interval were reported for each variable. Furthermore, a receiver operator curve was created to evaluate the combined utility of echocardiographic and serum parameters in predicting D/Tx. For continuous variables, missing values were imputed based on the mean of the entire sample.
Results
In the screening of 250 children, there were 99 children with 139 hospitalizations that fulfilled inclusion criteria and were analyzed. Most children not included in analysis were excluded due to lack of admissions for HF, history of malignancy, and lack of serum/echocardiographic parameters at admission. Table 1 lists baseline clinical characteristics of the group.
| Male gender | 56 (56%) |
| Days of hospitalization | median 15 (range 0–146) |
| Cardiac diagnosis (n = 99) | |
| Dilated cardiomyopathy | 58 (59%) |
| Left-sided obstructive lesions | 8 (8%) |
| D-transposition of great arteries | 2 (2%) |
| Tetralogy variant | 6 (6%) |
| Atrioventricular canal defect | 9 (9%) |
| Ebstein’s anomaly | 1 (1%) |
| Anomalous left coronary artery arising from left pulmonary artery | 2 (2%) |
| Hypertrophic cardiomyopathy | 1 (1%) |
| Muscular dystrophy | 2 (2%) |
| Single ventricular physiology | 8 (8%) |
| Other | 10 (10%) |
| Symptom at presentation (n = 139) | |
| Shortness of breath | 77 (57%) |
| Abdominal symptoms | 24 (17%) |
| Edema | 7 (5%) |
| Fatigue | 9 (7%) |
| Decreased function | 7 (5%) |
| Cardiac arrest | 2 (1%) |
| Chest pain | 1 (1%) |
| Other | 12 (9%) |
| Number of medications (n = 139) | |
| >1 diuretic | 33 (24%) |
| 4-drug heart failure regimen (angiotensin-converting enzyme inhibitor, β blocker, diuretics, digoxin) | 22 (16%) |
| β blocker | 37 (27%) |
Median age at admission was 3 years (range 0 to 18). Consistent with the natural history of HF, 78% of patients (n = 78) had 1 hospitalization, 11% (n = 11) had 2 hospitalizations, and 10% (n = 10) had ≥3 hospitalizations. Physical examinations were performed by 2 providers at admission, an attending and fellow or midlevel provider. Physical examination findings associated with significant HF include hepatomegaly and a gallop. The documented presence of these findings is included in Table 1 . Admission telemetric and/or electrocardiographic data were available on 56% of hospitalizations (n = 79). Most patients were in sinus rhythm (n = 69). Two subjects presented with significant arrhythmias: atrial flutter (n = 1) and ventricular tachycardia (n = 1).
Most patients (85/99) had an echocardiogram performed within 24 hours of the hospitalization ( Table 2 ). Mean left ventricular end-diastolic dimension (LVEDD) was 4.9 ± 1.5 cm (z-score 5.1 ± 3.2). Mean left atrial size was 2.9 ± 1.2 cm (z-score 2.2 ± 1.9).
| Variable (within 24 hours of admission) | No D/Tx | D/Tx | p Value |
|---|---|---|---|
| Serum | |||
| Bicarbonate (mmol/L) | 23.0 ± 4.5 | 21.2 ± 4.5 | 0.02 |
| Creatinine (mg/dl) | 0.6 ± 0.3 | 0.8 ± 0.8 | 0.004 |
| Brain natruretic peptide (pg/ml) (n = 51) | 1,038 ± 978 | 1,884 ± 1501 | 0.02 |
| Lymphocyte ratio (<1 relative lymphocytopenia) | 0.8 ± 0.4 | 0.6 ± 0.3 | 0.0005 |
| Echocardiographic | |||
| Ejection fraction (%) | 25.7 ± 11.6 | 18.9 ± 8.8 | 0.0006 |
| Tricuspid regurgitation (1 = trace, 5 = severe) | 2.1 ± 1.2 | 2.6 ± 1.1 | 0.02 |
There were 60 D/Tx events, which accounted for 43% of all HF admissions. Thirty-five patients were listed for heart Tx, 7 of whom died while waiting on the list. Two patients were bridged to Tx, 1 with a ventricular assist device and the other with extracorporeal membrane oxygenation.
A comparison was performed between serum and echocardiographic parameters for hospitalizations that lead to D/Tx versus discharge home. In univariate analysis, statistically significant differences were found in admission bicarbonate, creatinine, lymphocyte count, sodium, and BNP between groups ( Table 2 ). Patients admitted for symptomatic HF and had D/Tx had lower bicarbonate, higher creatinine, and higher BNP at admission. In addition, a statistically significant difference was found in peak Doppler tricuspid regurgitation velocity and EF between hospitalizations that ended in D/Tx versus discharge home.
Because normal lymphocyte count varies with age, we created a ratio of patients’ lymphocyte count to the midpoint of their normal age-based range. A ratio <1.0 indicated relative lymphocytopenia. In comparing this ratio, a statistically significant difference was also found between degrees of lymphocytopenia in the 2 groups ( Table 2 ).
In the multivariate logistic regression model, we found that lymphocytopenia, lower EF, sodium <138 mg/dl, and creatinine were independent predictors of D/Tx ( Table 3 ). Bicarbonate had a borderline association with outcome (p = 0.05) and was included in the model because it improved classification accuracy. This model correctly classified subjects 82% of the time ( Table 4 ). The receiver operator curve is shown in Figure 1 . The received operator curve was incorporated to determine how well the selected laboratory/echocardiographic predictors were at determining which patients with symptomatic HF would have the adverse outcome of D/Tx listing during a hospitalization. The area under the curve provides an estimate of the diagnostic accuracy of our model. The area under the curve in our analysis is 0.898. There were 51 subjects who had serum BNP at admission. In constructing a risk model using this cohort, BNP essentially replaced EF. The 2 variables, EF and BNP, had a modest correlation (r = −0.23, p = 0.03). All other predictive variables from the model remained independently predictive of the adverse outcome end point. The risk model in which BNP replaced EF correctly classified subjects 86% of the time.
| Serum/Echocardiographic Measurement (<24 hours of admission) | Odds Ratio | 95% CI | p Value |
|---|---|---|---|
| Ejection fraction | 0.94 | 0.90–0.98 | 0.009 |
| Lymphocytopenia | 5.40 | 1.67–17.4 | 0.005 |
| Creatinine | 9.77 | 1.97–48.3 | 0.005 |
| Bicarbonate | 0.91 | 0.83–0.99 | 0.05 |
| Sodium <138 mmol/L | 5.16 | 2.14–12.4 | 0.001 |
| Era ⁎ | 0.62 | 0.26–1.49 | 0.285 |
⁎ Although not a serum/echocardiographic measurement, era has been included in the multivariate analysis to demonstrate there was no predictive value or impact of era on death/transplantation.
| Sensitivity | 73.3% |
| Specificity | 88.8% |
| Positive predictive value | 83.0% |
| Negative predictive value | 81.6% |
| Correctly classified | 82.1% |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
