Author + information
- Received December 16, 2008
- Revision received May 13, 2009
- Accepted June 2, 2009
- Published online October 6, 2009.
- ↵*Reprint requests and correspondence:
Dr. Yuk M. Law, Children's Hospital and Regional Medical Center, University of Washington, 4800 Sand Point Way NE, M/S G-0035, Seattle, Washington 98105
Objectives The purpose of this study was to assess the ability of plasma B-type natriuretic peptide (BNP) to diagnose significant cardiovascular disease (CVD) in the pediatric population.
Background BNP has been shown to be reliable in detecting ventricular dysfunction and heart failure in adults. Timely and accurate identification of significant pediatric heart disease is important but challenging. A simple blood test could aid the front-line physician in this task.
Methods Subjects without a history of heart disease with findings possibly attributable to significant CVD in the acute care setting requiring a cardiology consult were enrolled. Clinicians were blinded to the BNP result, and confirmation of disease was made by cardiology consultation.
Results Subjects were divided into a neonatal (n = 42, 0 to 7 days) and older age group (n = 58, >7 days to 19 years). CVD was present in 74% of neonates and 53% of the older age group. In neonates with disease, median BNP was 526 pg/ml versus 96 pg/ml (p < 0.001) for those without disease. In older children with disease, median BNP was 122 pg/ml versus 22 pg/ml in those without disease (p < 0.001). Subjects with disease from an anatomic defect, a longer hospital stay, or who died had higher BNP. A BNP of 170 pg/ml yielded a sensitivity of 94% and specificity of 73% in the neonatal group and 87% and 70% in the older age group, respectively, using a BNP of 41 pg/ml.
Conclusions BNP is a reliable test to diagnose significant structural or functional CVD in children. Optimal cutoff values are different from adult values.
Cardiovascular disease (CVD) of childhood is important to recognize at its presentation, but this is often a challenging task. The spectrum of diseases is heterogeneous, and no single entity predominates. To complicate matters, the physiology of the cardiovascular system in children is not static and changes with growth and development. Furthermore, children are poor historians, and with the high prevalence of respiratory infections in the pediatric population, significant underlying cardiac disease can be obscured. In acute care settings, pediatric cardiologists may not always be available to assist front-line providers in the proper triage of the compromised patient. Traditional diagnostic studies such as chest radiography (CXR) and electrocardiography are also not accurate in identifying cardiovascular abnormalities (1,2). Consequently, patients with significant CVD are sometimes diagnosed with delay or are transported long distances unnecessarily.
Although a simple screening test (pulse oximetry) exists for cyanotic heart disease, the same cannot be said for hemodynamically significant cardiac lesions. Plasma B-type natriuretic peptide (BNP) is a hormone secreted by the atria and ventricles (3) and is part of an intricate homeostatic network that regulates the circulating blood volume. It is elevated under conditions of increased wall stress (4,5), such as with increased preload or afterload and decreased systolic or diastolic ventricular function. Large prospective studies performed in the acute care setting in adults have demonstrated the value of circulating BNP as a method to diagnose ventricular dysfunction and heart failure (6–10). The measurement of BNP is available as a rapid automated point-of-care antibody fluorescent test, requiring <0.5 ml of whole blood, which can be obtained from finger stick. BNP values in the healthy population are age-related, and although BNP values for healthy children have been published (11–14) and small studies have shown BNP to be elevated in various types of congenital cardiac defects (15–17), neither the accuracy nor the utility of BNP in diagnosing hemodynamically significant heart disease in children has been rigorously examined. The Better Not Pout Children! study was a field test to evaluate the accuracy of BNP as a rapid triage tool to diagnose significant CVD in the pediatric population. The hypothesis is that a single plasma BNP measurement can be used to accurately identify significant CVD in infants and children using the cardiology consultant's evaluation as the gold standard.
Subjects and diagnosis of significant CVD
This was a prospective study from a single pediatric center with blinding of the diagnosticians to the BNP test results. Subjects were between 36 weeks' gestational age (corrected) to 19 years of age and had no known cardiovascular abnormality. Subjects were eligible for enrollment upon request by a noncardiologist for an urgent pediatric cardiology consult to evaluate a patient presenting with signs or symptoms that may be attributable to a significant cardiovascular abnormality, such as dyspnea, failure to thrive, swelling, hypotension, and so on. Similarly, subjects with an isolated murmur or spontaneously resolved palpitations who were otherwise asymptomatic were not eligible. Other exclusion criteria included sepsis, renal dysfunction requiring therapy, insufficient data for the cardiologist to form a diagnosis, or cardiovascular-specific treatment, for example, diuretics, vasodilators, inotropes, vasoconstrictors, and the like, having been started more than 6 h before sample collection.
Because the spectrum of significant CVD in children is broad, the common defining feature of “significant” was whether the function or the structure of the heart had been altered by the cardiovascular abnormality as officially interpreted by the consulting cardiologist. Abnormal function included systolic dysfunction with shortening fraction <28% or a qualitative decrease and diastolic dysfunction, which was defined as indirect signs of elevated filling pressures, such as enlarged atria, or the presence of heart failure in the absence of systolic dysfunction. Altered structure included an abnormality in the chamber size (e.g., from increased pre-load) or myocardial thickness (e.g., from increased afterload) attributable to a cardiac defect (e.g., ventricular septal defect) or a cardiovascular abnormality without an anatomic defect (e.g., pulmonary hypertension or systemic hypertension). Because pulmonary arterial hypertension is a common manifestation of hemodynamically significant CVD, it was defined by echocardiography directly as a right ventricular or pulmonary systolic pressure that is over one-half systemic or indirectly by flattening of the interventricular septum or right-to-left shunting at the ventricular or aortic level during any part of the cardiac cycle. Lastly, CVD is further divided into those with an anatomic defect and those without. An anatomic defect is defined as an inherent structural abnormality in the circulation that is either congenital or acquired, such as coarctation of the aorta or rheumatic valve disease.
A cardiac finding was classified as insignificant if it was thought to be unrelated to the symptoms that prompted the initial request for consultation. For example, a subject with a bicuspid aortic valve and a 2-m/s gradient without alteration in structure or function, and the illness not consistent with the cardiovascular abnormality found, would not be placed in the group of significant CVD.
The consulting cardiologist determined the diagnostic tests required (echocardiogram was not required), made the official diagnosis, and filed a study form to define whether an alteration of function or structure was present. The study population was divided into the neonatal group (≤7 days of age) and the older age group (the rest of the subjects up to 19 years of age). Consent and/or assent were obtained from all subjects and their legal guardians. The study was approved by the local institutional review board.
Blood samples collected from indwelling vascular catheters, venipuncture, or finger/heel stick were transported immediately in potassium (K2) ethylenediaminetetra acetic acid tubes and assayed within 4 h. Samples were collected within 24 h of the consult. Subjects were enrolled from any acute care facility of the medical center (e.g., emergency department, inpatient wards, nursery, and intensive care units). Plasma BNP was measured using the Triage BNP point-of-care assay (Biosite Inc., San Diego, California). This assay has a linear dynamic range from <5 to 5,000 pg/ml, with an intraday precision standard deviation of 8.8% to 11.6% over this range. Maintenance and quality performance assurance of the instrument were strictly followed according to company instructions.
Descriptive data are presented as median with range and mean ± SD. Comparison of the groups was done by nonparametric methods using the Mann-Whitney Utest between 2 groups and the Kruskal-Wallis H test among more than 2 groups. The distribution of BNP values is illustrated by box-and-whisker plots (Tukey), where the heavy solid line indicates the median, the box represents the interquartile range (25% to 75%), and whiskers are drawn to 1.5 times the interquartile range. Bivariate analysis for length of stay was performed using the method of Spearman. Receiver-operating characteristic curves, accuracy calculations, figures, and all statistical analyses were generated using SPSS version 15.0 (SPSS Inc., Chicago, Illinois). Statistical significance was accepted for p < 0.05.
The study was conducted over a 20-month period, with 102 subjects enrolled and 100 fulfilling the final study criteria. Table 1describes the demographic characteristics of the study population. All subjects had clinical features that posed the possibility of significant CVD to the referring physician. The majority of the subjects were neonates and infants, as shown by the age and weight distribution. In addition, 28% of subjects were from the inpatient ward, 15% from the pediatric intensive care unit, 46% from the neonatal intensive care unit/nursery, and 11% from the emergency department. This distribution by location of subjects also reflected the large number of young infants in the study. An echocardiogram was performed in 97% of the subjects. All subjects with CVD had an echocardiogram. CXR was performed in 92% of the subjects, 93% in the disease group, and 90% in the nondisease group. There was no statistical difference between the disease and nondisease groups (total and age groups) in regard to having had an echocardiogram or CXR. Abnormalities on CXR included enlarged heart size (30%), increased/decreased vascularity (9%), edema (8%), or parenchymal disease (50%). Patients with an abnormal CXR had a median BNP of 263 pg/ml (range 5 to 5,000 pg/ml); patients with a normal CXR had a median BNP of 47 pg/ml (range 5 to 2,890 pg/ml, p = 0.009).
To be enrolled in the study, subjects were required to present with signs and/or symptoms that led the front-line physician to suspect significant CVD. All subjects had symptoms possibly referable to the cardiovascular system, with the most common being dyspnea (neonatal 64%, older age 60%), hypoxemia (neonatal 45%, older age 19%), feeding/gastrointestinal/growth difficulties (neonatal 5%, older age 17%), fatigue or lethargy (neonatal 10%, older age 14%), and collapse (neonatal 0%, older age 7%). There was a correlation between the symptoms recognized by the front-line physician and consulting cardiologist. The most common abnormal physical examination findings referable to the cardiovascular system included an abnormal pre-cordial impulse, S3/S4, loud single S2, edema, jugular venous distension, hepatomegaly, decreased perfusion or pulses, and breath sounds (Table 2).There was a weak correlation in physical findings as reported between the front-line physician and cardiologist. The prevalence of 1 or more abnormal physical findings, as detected by the cardiology consultant, was 76% in those with significant CVD, whereas these findings were present in 33% of those without disease. In the older age group with disease, 67% had 1 or more abnormal physical findings; 36% had 1 or more abnormal physical findings in those without disease. For the entire group of subjects, 74% had 1 or more abnormal physical findings in the disease group versus 34% in those without significant CVD.
BNP by age groups and cardiovascular diagnosis
Plasma BNP was higher overall in the neonatal group, with or without significant cardiac disease, than in the older age group. Figure 1illustrates the distribution of BNP by age groups and presence of disease. The median BNP for neonates with disease was 526 pg/ml (mean 1,017 ± 1,128 pg/ml) versus 96 pg/ml (mean 134 ± 130 pg/ml, p < 0.001) for those without disease. The median BNP for the older age group with disease was 122 pg/ml (mean 824 ± 1,330 pg/ml) versus 22 pg/ml (mean 66 ± 96 pg/ml, p < 0.001) for those without disease. When significant CVD was further divided into those with an anatomic defect or without an anatomic defect, BNP remained elevated in both age groups compared with their nondisease group (Fig. 2).In the neonatal group, median BNP was 872 pg/ml (mean 1,350 ± 1,387 pg/ml) in those with a CVD with anatomic defect and 446 pg/ml (mean 613 ± 499 pg/ml) in those with CVD without an anatomic defect. In the older age group, median BNP was 613 pg/ml (mean 887 ± 865 pg/ml) in those with CVD from an anatomic defect and 106 pg/ml (mean 793 ± 1,521 pg/ml) in those with CVD without an anatomic defect. There is statistical significance in the comparison of the difference in BNP among the 3 groups (Fig. 2). The difference is also statistically significant in the independent comparison between each of the disease groups with the nondisease group (p < 0.05 for each of the age group comparisons). The specific cardiovascular diagnoses are shown in Tables 3 and 4.⇓The group without disease encompassed a wide range of diagnoses, the most common of which included chest infection, lung disease, and metabolic and neuromuscular diseases.
Length of hospital stay and survival
Subjects with CVD had a longer length of stay (median 7 days) compared with those without disease (median 5 days, p = 0.04). The correlation of BNP to the length of stay in the entire group of subjects with significant CVD showed an r coefficient of 0.27 with p = 0.041. Within the neonatal group, the correlation coefficient was 0.56 (p = 0.001). Within the older age group, the correlation coefficient was 0.16 (p = 0.418). In the length-of-stay analysis, the 4 subjects who died were excluded. Furthermore, the association of survival with BNP was assessed by comparing BNP in the subjects with CVD who survived (n = 58) with that of those who died (n = 4) during the same hospital stay. Those who survived had a median BNP of 314 pg/ml (mean 868 ± 1,141 pg/ml) versus 776 pg/ml (mean 1,682 ± 2,242 pg/ml) for those who died (p = 0.344). These 2 findings suggest that a higher BNP is associated with greater morbidity and severity of disease, particularly in the neonatal group.
Accuracy of the BNP assay for clinician-judged CVD
Receiver-operating characteristic curves were generated for each age group. Characteristics of these curves are depicted by the area under the curve, confidence interval, and p value compared with a diagonal reference line (Table 5).Cutoff values that favored sensitivity while maintaining an adequate combination of sensitivity and specificity are shown. Optimal BNP cutoff values to diagnose significant CVD were higher for the neonatal group than for the older age group. In the neonatal group, a BNP cutoff of 170 pg/ml produced a sensitivity of 94%, specificity of 73%, positive predictive value (PPV) of 91%, and negative predictive value (NPV) of 80%. For the older age group, the optimal BNP cutoff of 41 pg/ml delivered a sensitivity of 87%, a specificity of 70%, PPV of 77%, and an NPV of 83%. When all the subjects are combined, an intermediate cutoff of 99 pg/ml provided the best sensitivity and specificity. As shown in Table 5, the accuracy of the BNP assay is improved when separate age groups are considered, compared with lumping the entire study population into 1 group.
If the maximal NPV (the chance of those with a negative test to not have disease) was used to determine cutoff values, different cutoffs and their respective accuracy values were generated for the neonatal and total group but remained the same (41 pg/ml) for the older age group (Table 6).For the neonatal group, using a higher NPV (88%) resulted in a higher sensitivity (97%) but a lower specificity (64%) and PPV (88%). In the combined group, using a maximal NPV of 81%, the sensitivity increased to 92%, with the specificity decreasing to 55% and PPV to 77%.
In a logistic regression model, age, sex, weight, CXR, and length of stay were not found to influence BNP. A linear regression model showed all of these variables to characterize BNP. Because neonatal status was pre-defined at the outset of the study to influence BNP, the data were stratified by neonatal or older age status. In the neonatal-stratified model, the length of stay was predictive of BNP (p < 0.05). In the older age stratum, age, weight, and CXR were predictive of the BNP level (p < 0.05).
The Better Not Pout Children! study is the largest prospective study to assess the accuracy of plasma BNP to diagnose CVD in the pediatric population. It was designed as a field test to evaluate BNP as a potential diagnostic tool for front-line physicians to diagnose significant CVD when there is a suspicion of such in the acute care setting. These are clinical conditions that should be recognized and differentiated accurately from non-CVD processes without delay. The principal findings of this study are that a neonate up to 7 days of age with possible cardiac-related findings and a BNP of ≥170 pg/ml has a 91% chance of having significant CVD; in a child older than this age, a BNP of ≥41 pg/ml predicts a 77% chance that significant CVD is present. Favoring a high sensitivity is important in identifying heart disease in the acute care setting in children, especially infants, because a delayed diagnosis can affect treatment response and outcome.
Using cutoff values based on the maximal NPV to improve the ability to rule out significant CVD in the presence of a negative test also delivered good sensitivities for both groups. However, the specificity and PPV do decline, and the maximal NPVs generated are not sufficiently high enough that they would be recommended over the ones generated from favoring a more balanced accuracy profile that still favors sensitivity. One possible explanation for why the maximal NPVs did not approach >95% is the high prevalence of disease in the study population. Enrolled subjects presented with signs and symptoms that made the front-line physician suspicious of significant CVD. It is also possible that the acuity and severity of the disease process in the non-CVD group could have raised the basal level of BNP production.
In summary, at an acceptable level of sensitivity and NPV, the specificity and PPV were not compromised. These results demonstrated that plasma BNP, when used with a clinical suspicion of significant CVD by the front-line physician, can be an accurate biomarker in the pediatric population.
Prospective studies that attempt to measure the accuracy of BNP in children are scarce, are limited by small sample size, and use different enrollment and diagnostic criteria (18–20). In the Better Not Pout Children! study, enrollment of subjects without known pre-existing CVD was critical because it eliminates selection bias. We enrolled subjects from acute care settings because these are the places where the initial presentation of significant CVD is most likely to occur. Furthermore, blinding caregivers to the BNP results and using the cardiologist's consult as the gold standard for diagnosis resembles how the BNP assay would be applied in clinical practice and also parallels the methods used in previous landmark adult studies (6–9). Separation of the subjects by age groups based on previously known physiologic data allows for a more precise comparison.
One of the most important aspects of the current study is the selection of the nondisease group. A recently published pediatric prospective study by Maher et al. (20) also attempted to measure the accuracy of the assay. In that study, the nondisease population was sampled from the emergency department, whereas the disease population required an intensive care admission, allowing for the possibility of an ascertainment bias in that the control subjects were less ill than the disease population. Perhaps not surprisingly, a remarkably high sensitivity and specificity of >97% were obtained. In the Better Not Pout Children! study, the nondisease and disease subjects required a similar clinical presentation and were sampled concurrently from the same source, thereby ensuring that both populations possessed a similar disease acuity and severity. Although the current study finds a lower sensitivity and specificity compared with that of Maher et al. (20), it probably provides a more valid measurement of the accuracy of the BNP assay in the pediatric population.
It should also be emphasized that in the Better Not Pout Children! study, BNP was not used as a screening test of all acute care patients. The enrolled subjects presented with signs and symptoms suggestive of significant CVD as discerned by the front-line physician using all available data deemed necessary without a cardiology consult. Therefore, the population studied is enriched with disease subjects, which can increase the sensitivity and PPV while possibly disfavoring the specificity and NPV. On the basis of this study design and entry criteria, we view the application of the BNP test as a way to strengthen and not replace clinical acumen.
BNP results by age groups
It should be underscored that different cutoff values for different age groups is important in the pediatric population. The decline in the accuracy of the assay is particularly large in the neonatal population if these subjects were to be grouped with older children. This was not unexpected because the combined group's cutoff (99 pg/ml) is significantly lower than the optimal cutoff (170 pg/ml) generated for the neonates. In other words, in using a combined age group cutoff, even though neonates with disease will still be identified, the false-positive rate and therefore specificity will suffer significantly. Similarly, the sensitivity will also be lessened in the older age group if a higher cutoff value is used.
The higher cutoff value in neonates is not surprising, as studies have shown higher natriuretic peptide levels in healthy neonates compared with those of other age groups. Natriuretic peptide levels decrease after birth, reaching a steady level by 4 to 7 days of age (12–14). The natriuretic peptides may play an important role in the homeostasis of the fetal-neonatal-mature transitional circulation. Total body water is increased in the newborn, and the acute and dramatic increases in the systemic pre-load and afterload to the pulmonary arterial circulation in conjunction with a stiff neonatal myocardium (21,22) could lead to a physiologic release of natriuretic peptides. Because the newborn may be less responsive to natriuretic peptides (23), a higher secreted level is required to promote the well-recognized natriuresis that commences in the first week of life. During this time, the precipitous decrease in pulmonary vascular resistance may also be promoted by natriuretic peptides (24–26). Altogether, the elevated BNP values followed by their decrease after the first week of life may be an adaptive response to these cardiophysiologic requirements. This age variation was 1 reason we chose to separate the age groups at 7 days of age. Although the levels continue to decline after the first week of life in healthy individuals, the change is small after this time until adulthood, when the values increase gradually again. However, from the multivariate analysis, it does appear that there remains an age factor in the older age group. This is not altogether surprising, because the physiologic decline is a continuum (albeit very small), and in CVD, the decline may be delayed. This could affect the accuracy of the BNP assay at the lower age range of the non-neonates. The design of future studies to improve upon the current study's assessment of BNP to diagnose significant CVD should take this into consideration.
BNP and type of CVD
When BNP was examined by disease category, the disease patients either with or without an anatomic defect were shown to have significantly higher values than those without disease. This suggests that BNP can be used to diagnose significant CVD regardless of etiology. Neither our study nor any others to date have been adequately powered to use BNP to differentiate between anatomic versus functional heart disease when contributing factors are properly controlled. Moreover, the current study did not attempt to stratify the severity of the cardiovascular condition to the magnitude of BNP elevation. These are important areas of investigation in the future.
To further improve the utility of BNP as a diagnostic biomarker for the pediatric population, future studies should assess the accuracy of BNP when it is added to the physical examination and CXR as interpreted by the front-line physician. For the cardiologist who is consulting with a distant primary caregiver, the BNP level may also aid in determining whether an immediate echocardiogram or referral is necessary. Given our findings that suggest a higher BNP level is associated with length of stay and possibly survival, it is important to test whether BNP has a role in prognostication (27). Lastly, future studies can culminate in clinical trials to assess whether adding BNP to other routine monitoring tests can positively impact treatment response or outcome.
Although large compared with that of existing studies, the sample size in this study remains a limitation. Pediatric CVDs and their presentations are heterogeneous, resulting in a wide range of ventricular dysfunction and heart failure severity, the latter being particularly difficult to grade appropriately. The necessity of lumping subjects of different ages into only 2 groups can also affect the accuracy. Furthermore, a clear-cut definition of significant CVD and the means to accurately diagnose it 100% correctly (the gold standard) does not exist in pediatric cardiology. Hence, the consulting cardiologist's official diagnosis was used in this study, even though this would not guarantee standardization and complete objectivity. The clinician must cautiously interpret a simple cutoff in a quantitative test that is not 100% accurate, such as we propose in this study. Not only was the patient population heterogeneous in terms of the demographics (age, sex, and weight), but noncardiovascular processes could also potentially affect the individual patient's BNP result. Therefore, the BNP result must be interpreted under the clinical context for which it was ordered. Nevertheless, the pediatric cutoffs generated by this study do provide a much-needed point of reference for the clinician to determine whether a cardiology consult is necessary, especially when these cutoffs are different from those of the adult population.
The Better Not Pout Children! study validates the utility of a single plasma BNP measurement to diagnose a variety of significant CVD in the pediatric acute care setting. In this heterogeneous population, it is important to identify cutoff values applicable to neonates and older children. The cutoffs offering optimal accuracy are 170 pg/ml for neonates to 7 days of age and 41 pg/ml for those who are older than 7 days to the age of 19 years. The cutoffs are different between these 2 age groups. Equally important, the study results are also different from published and company-recommended cutoffs of 80 to 100 pg/ml for adults (6–9). Because congenital heart disease can present early in the neonatal period (e.g., ductal-dependent lesions), and children are less forthcoming with their complaints than adults, the timely recognition of significant CVD is important, and the measurement of BNP can be useful in the diagnostic process.
The investigators thank Dr. Tracy Bumsted of Oregon Health & Science University for her expert review of the paper from the perspective of a pediatric hospitalist.
This study was supported by a grant from the Friends of Doernbecher Foundation, Oregon Health & Science University, Portland, Oregon. The Triage BNP assay and reagents were an unrestricted gift for the purpose of research and education from Biosite Inc., San Diego, California. It was a 1-time transaction, and Biosite was never involved in the design nor has any knowledge of the results of the study.
- Abbreviations and Acronyms
- B-type natriuretic peptide
- cardiovascular disease
- chest X-ray
- negative predictive value
- positive predictive value
- Received December 16, 2008.
- Revision received May 13, 2009.
- Accepted June 2, 2009.
- American College of Cardiology Foundation
- Hama N.,
- Itoh H.,
- Shirakami G.,
- et al.
- McCullough P.A.,
- Nowak R.M.,
- McCord J.,
- et al.,
- BNP Multinational Study Investigators
- Morrison L.K.,
- Harrison A.,
- Krishnaswamy P.,
- Kazanegra R.,
- Clopton P.,
- Maisel A.
- Dao Q.,
- Krishnaswamy P.,
- Kazanegra R.,
- et al.
- Koch A.,
- Singer H.
- Mir T.S.,
- Laux R.,
- Hellwege H.H.,
- et al.
- Mir T.S.,
- Marohn S.,
- Laer S.,
- Eiselt M.,
- Grollmus O.,
- Weil J.
- Cohen S.,
- Springer C.,
- Avital A.,
- et al.
- Maher K.O.,
- Reed H.,
- Cuadrado A.,
- et al.
- Rudolph A.M.
- Silberbach M.,
- Stejskal E.,
- Foker J.,
- et al.
- Zhao L.,
- Mason N.A.,
- Strange J.W.,
- Walker H.,
- Wilkins M.R.
- Price J.F.,
- Thomas A.K.,
- Grenier M.,
- et al.