N-terminal pro-B-type natriuretic peptide (NT-proBNP) is an important predictor of outcome in adults with heart failure. In children with heart failure secondary to dilated cardiomyopathy (DC) markers that reliably predict disease progression and outcome during follow-up are scarce. We investigated whether serial NT-proBNP measurements were predictive for outcome in children with DC. All available NT-proBNP measurements in children with DC were analyzed. Linear mixed-effect models and Cox regression were used to analyze the predictive value of NT-proBNP on the end point of cardiac death (death, heart transplantation, or mechanical circulatory support). During 7 years, 115 patients were included. At diagnosis, median NT-proBNP was high and not predictive for outcome. At any time during follow-up, a twofold higher NT-proBNP resulted in a 2.9 times higher risk in the first year (p <0.001) and a 1.8 times higher risk thereafter (p <0.001). Furthermore, at any time, the slope of log10(NT-proBNP) was significantly predictive for the risk of an end point (0 to 30 days hazard ratio [HR] 3.5, >30 days HR 2.9; >1 year HR 6.4). In patients with idiopathic DC (IDC) at 30 days after diagnosis, NT-proBNP ≥7,990 pg/ml showed a 1- and 2-year event-free survival of 79% and 71% and >1 year after diagnosis NT-proBNP ≥924 pg/ml showed a 2- and 5-year event-free survival of 50% and 40%, whereas below both thresholds event-free survival was 100%. In non-IDC, these thresholds were not predictive for outcome. In conclusion, NT-proBNP at any time during follow-up and its change over time were significantly predictive for the risk of cardiac death in children with DC. In children with IDC >1 year after diagnosis, NT-proBNP >924 pg/ml identified a subgroup with a poor outcome.
Dilated cardiomyopathy (DC) in children causes heart failure and is characterized by a poor prognosis. Previous reports have shown that 1 year after diagnosis, 18% to 31% of the patients have reached the end point of death or heart transplantation after a median time of 1 to 2 months. Event rates decrease after the first year, and 5 years after diagnosis, transplant-free survival is from 54% to 72%. Some children have stable disease and 22% to 38% of the patients recover. There is a strong need for predictors that identify children with a poor prognosis, both at diagnosis and during follow-up. N-terminal pro-B-type natriuretic peptide (NT-proBNP) is a biomarker that is predominantly released by myocytes in the ventricular walls in response to increased wall stress. In adults with chronic heart failure, single NT-proBNP measurement is an important predictor for short-term mortality. Repeated NT-proBNP measurements have been proved a superior method for risk stratification. More recently, NT-proBNP values have been used to guide pharmacologic treatment and have shown beneficial effects on outcome. In children with heart failure, NT-proBNP has been associated with the clinical severity of heart failure and with echocardiographic measures of left ventricular (LV) dysfunction. However, until now, no studies have investigated the predictive value of repeated NT-proBNP measurements on hard end points as death and heart transplantation in a relatively large pediatric DC cohort. Therefore, the aim of this study was to investigate the predictive value of serial NT-proBNP measurements from diagnosis onward and during follow-up.
Methods
All children (0 to 18 years) who fulfilled the criteria of DC in a participating tertiary referral center from July 1, 2006, to October 1, 2013, were selected for this study. DC was defined as the presence of 2 of 3 of the following criteria: (1) symptomatic heart failure, (2) impaired systolic function (fractional shortening ≤25%), and (3) LV dilation (LV end-diastolic dimension z-score >2 for body surface area). DC was idiopathic or secondary to other causes. Patients with structural heart disease or neuromuscular disease were excluded. All available NT-proBNP values were collected.
Additional information was obtained from medical charts, that is, age, time since DC diagnosis, and presence of symptomatic heart failure at diagnosis. The estimated glomerular filtration rate (eGFR) was calculated using the modified Schwarz formula (children 1 to 17 years) or the MDRD formula (17 years). eGFR <90 ml/min/1.73 m 2 was considered renal dysfunction. No eGFR formula is available for children aged <1 year; thus, we constructed the variable “renal dysfunction present or absent” based on serum creatinine. Serum creatinine >2 SDS for age was considered renal dysfunction. Serum creatinine and NT-proBNP were preferably measured on the same day but never >1 week apart.
Follow-up continued either until patients reached the primary end point, or the age of 18 years, or until the end of the study on October 1, 2013. The primary end point was cardiac death, defined as death, heart transplantation [HTx], or need for mechanical circulatory support (MCS, ventricular assist device, or extracorporeal membrane oxygenation).
We analyzed the data in 2 groups: (1) NT-proBNP values from DC diagnosis onward in children presenting with symptomatic heart failure and (2) NT-proBNP values of children who survived at least the first year after diagnosis and still met the criteria of DC. Children could be represented in both study groups. Analyses in both groups were adjusted for renal function; in the first group, renal dysfunction was dichotomous because this group comprised children <1 year of age and in the second group eGFR was used. The type of DC was analyzed in 2 groups: idiopathic DC (IDC) and DC with underlying cause identified (non-IDC). The study was approved by the medical ethical committee of the Erasmus University Medical Center, Rotterdam, The Netherlands. Informed consent was waived.
For descriptive statistical analysis, we used IBM SPSS Statistics for Windows, version 21.0 (IBM Corp., Armonk, New York). Continuous data are reported as mean ± SD if normally distributed or otherwise as median and interquartile range (IQR). Medians were compared using the Mann-Whitney U test. Based on our previous observations, we determined thresholds at 1 month and 1 year after diagnosis to identify patients with a very low risk of an end point versus those with an increased risk. Receiver-operating characteristic curves were generated of available NT-proBNP values and points with a relatively high sensitivity were chosen. Using these thresholds, end point–free survival was estimated with the Kaplan-Meier method and the log-rank test was used to determine statistical significance. At 30 days, the Kaplan-Meier analysis only took end points in the first 2 years into account. For advanced statistical analysis of longitudinal and survival data, we used the R environment for statistical computing and graphics (R, version 3.0.2 [2013-09-25]). NT-proBNP was log transformed in all analyses because it was non-normally distributed. To assess changes in NT-proBNP levels over time, while accounting for the correlation between repeated follow-up measurements in each patient, a linear mixed-effects model analysis was used. To account for the expected differences in evolution of NT-proBNP, we have included 2 separate slope parameters in the fixed- and random-effect parts of the model, 1 for the period from DC diagnosis to 30 days and 1 from 30 days to the end of follow-up (segmented regression). Subsequently, a joint model was constructed to explore how serial NT-proBNP measurements were associated with the risk of an end point. Here, renal function was included as covariate. Residual plots were used to validate the models’ assumptions. The subject-specific regression coefficients from the mixed model that correspond to the slopes of the NT-proBNP profile for the 2 periods were incorporated in a Cox regression model to measure the strength of the association with the hazard of an end point; age and type of DC were included as covariates. The significance level in all analyses was set to 5%.
Results
We included 115 patients with DC, in whom a total of 2,048 NT-proBNP values were analyzed. In 79 patients, NT-proBNP data were available from diagnosis onward and were used in the analysis for the first year after presentation. For the analysis >1 year after presentation, data of 36 patients with previous presentation and of 32 of 79 patients followed from diagnosis onward were available (see Figure 1 ).
