Growth Differentiation Factor 15: A Novel Growth Biomarker for Children With Congenital Heart Disease

Abstract

Background:

Failure to thrive (FTT), defined as weight or height less than the lowest 2.5 percentile for age, is prevalent in up to 66% of children with congenital heart disease (CHD). Risk stratification methods to identify those who would benefit from early intervention are currently lacking. We aimed to identify a novel growth biomarker to aid clinical decision-making in children with CHD.

Methods:

This is a cross-sectional study of patients 2 months to 10 years of age with any CHD undergoing cardiac surgery. Preoperative weight-for-age Z scores (WAZ) and height-for-age Z scores (HAZ) were calculated and assessed for association with preoperative plasma biomarkers: growth differentiation factor 15 (GDF-15), fibroblast growth factor 21, leptin, prealbumin, and C-reactive protein (CRP).

Results:

Of the 238 patients included, approximately 70% of patients had WAZ/HAZ < 0 and 34% had FTT. There was a moderate correlation between GDF-15 and WAZ/HAZ. When stratified by age, the correlation of GDF-15 to WAZ and HAZ was strongest in children under 2 years of age and persisted in the setting of inflammation (CRP > 0.5 mg/dL). Diagnoses commonly associated with congestive heart failure had high proportions of FTT and median GDF-15 levels. Prealbumin was not correlated with WAZ or HAZ.

Conclusions:

GDF-15 represents an important growth biomarker in children with CHD, especially those under 2 years of age who have diagnoses commonly associated with CHF. Our data do not support prealbumin as a long-term growth biomarker.

Introduction

Congenital heart disease (CHD) is the most common congenital malformation with an incidence of 14 per 1000 live births in the United States.¹ Congenital heart defects can vary widely with regard to their impact on a child’s health and well-being, and approximately 25% of children with CHD require a cardiac procedure within the first year of life.² Children with CHD are at increased risk for failure to thrive (FTT), defined as weight or height less than the lowest 2.5 percentile for age, with up to 66% of children with CHD developing FTT.³ FTT is associated with higher perioperative morbidity and mortality, suboptimal nutritional recovery even after corrective cardiac surgery, and longer hospitalizations.^4,5 In a study of 74 children with CHD who died after surgery, the odds of death was 13.5 times higher in children with FTT than those without.⁶ In a study of 100 single ventricle patients, a lower weight-for-age Z score was associated with increased hospital length of stay after the bidirectional Glenn procedure.⁵ There are a number of factors thought to contribute to FTT in children with CHD including the type of defect, degree of cyanosis, presence of shunts, increased energy expenditure, decreased energy intake, disturbances in gastrointestinal function, and presence of genetic disorders.^7,8

Although certain defects are more highly associated with FTT, there is heterogeneity for FTT between children with the same defects and physiology.⁵ Therefore, it would be clinically useful to be able to identify those children with CHD who are at increased risk for FTT to initiate early nutritional intervention and potentially prevent FTT and its adverse outcomes. In diseases associated with left to right shunts and congestive heart failure, failure to thrive despite maximal medical management is an indication for earlier correction of underlying disease. Plasma biomarkers, which are simple to obtain and measure repeatedly, serve as a potential predictive tool for FTT.⁹ A plasma biomarker related to the underlying physiology created by CHD and associated with FTT may provide a better biological measure of the child’s current state of growth potential than other methods like daily weights, which are prone to inaccuracies and overestimation in patients with congestive heart failure, and total calorie consumption, which though seemingly adequate may not always produce growth. Traditionally, prealbumin has been considered the gold standard biomarker of inpatient nutritional status^10,11; however, a long-term growth biomarker that can be used to predict FTT has not been identified.

In a small cohort study, Wang and colleagues found children with CHD and FTT have higher levels of growth differentiation factor (GDF-15) than children with CHD alone.¹² Fibroblast growth factor 21 (FGF-21) expression has been found to be elevated in states of starvation.¹³ Previous studies have also suggested that elevated plasma leptin levels may play a role in the development of FTT in patients with CHD.¹⁴ In the present study, we sought to combine and compare these molecules in a large CHD surgical study to identify a reliable plasma biomarker associated with FTT in children with CHD.

Patients and Methods

Study Population

This cross-sectional study included patients between 2 months and 10 years of age with any CHD diagnosis who underwent cardiac surgery from September 2016 to July 2019 at Monroe Carell Jr. Children’s Hospital at Vanderbilt University, Nashville, TN. Patients with genetic syndromes were excluded from this study due to the confounding effect of syndromes on FTT. Characteristics recorded included age, sex, diagnosis (see Supplement S1 for all primary diagnoses with ICD-10 codes) type of surgery, height, and weight. This study was approved by the Vanderbilt University and Johns Hopkins Medicine Institutional Review Boards. Written informed consent was provided by parents of all participating patients.

Sample Collection

The samples of 238 children were randomly chosen from biorepository samples collected preoperatively from patients with CHD and stored in the Core Laboratory for Cardiovascular Translational and Clinical Research at Vanderbilt University. The preoperative blood samples are collected in EDTA-coated tubes and these tubes are centrifuged at 1000 × g for approximately 15 min to collect separated plasma and catalogued in the Core Laboratory for Cardiovascular Translational and Clinical Research at Vanderbilt University. All samples are stored at −80 °C.

Enzyme-Linked Immunosorbent Assay

Plasma samples were thawed, and multiplex analysis of GDF-15, FGF-21, and leptin was performed on a fluorescently labeled and microsphere bead-based Luminex 200 system using MILLIPLEX Human Aging Magnetic Bead Panel 1 multiplex assay kit (Millipore Sigma). Prealbumin and C-reactive protein (CRP) were also measured as gold standard markers of nutritional status and inflammatory status, respectively, via ELISAs (Eagle Biosciences Human Prealbumin ELISA Assay Kit, Catalog # HPA31-K01, Millipore Human C-Reactive Protein ELISA Kit, Catalog # RAB0096). Elevated CRP was defined as greater than 0.5 mg/dL. Samples for all markers were run according to the kit protocols in duplicates, and samples with higher-than-expected values were noted from each batch and repeated at the end of the assay. The interassay coefficients of variability were as follows: GDF-15 (<10%), FGF-21 (<15%), leptin (<10%), CRP (<12%), and prealbumin (<10%). Values above or below the detectable threshold were considered missing. Data are presented as the median and interquartile range.

Statistical Analysis

The primary outcome was the association between the candidate biomarkers and weight-for-age Z score (WAZ) and height-for-age Z score (HAZ) just prior to surgery. Failure to thrive was defined as WAZ or HAZ < −2, representing the lowest 2.5% of values within a normal distribution. For patients up to 2 years of age, a preoperative WAZ and HAZ was calculated using the World Health Organization (WHO) growth charts for 0 to 2 years. For patients older than 2 years of age, WAZ and HAZ were calculated using the Center for Disease Control growth charts for 2 to 20 years. This decision was based on the Center for Disease Control and Prevention (CDC), National Institutes of Health (NIH), and American Academy of Pediatrics (AAP) recommendations.¹⁵

Categorical variables are presented as number and percentage. Continuous variables were assessed for normality using histograms and log-transformed as appropriate. Associations between the biomarkers and WAZ/HAZ were assessed using Pearson’s correlation coefficient, in the overall group and stratified by age group (0-12 months, 12-24 months, >24 months) and also by inflammatory status (CRP ≤ 0.5 mg/dL, >0.5 mg/dL). A Wilcoxon rank-sum test was used to compare biomarker levels between FTT groups. Biomarker robustness for individual plasma markers was analyzed by receiver operating characteristic (ROC) analysis using the area under the ROC curve (AUC) to assess and integrate the sensitivity and specificity of the biomarkers for FTT (WAZ or HAZ <−2), which informs classification performance. Prior to ROC analysis, biomarkers were considered to be independent of FTT, and samples were randomly selected and considered to be representative of the population of children with CHD, thus meeting basic assumptions for ROC analysis. Multivariable logistic regression models analyzing the presence of FTT with adjustment for age and sex were performed separately for each biomarker of interest, and goodness of fit was assessed for each model. A P value of <.05 was considered statistically significant for all analyses. All analyses were performed using SAS 9.4 (SAS Institute).

Results

Patient Characteristics

Anthropometric data and plasma samples were obtained for 238 patients. Baseline characteristics of the cohort and median levels of the candidate biomarkers are included in Table 1. There were no missing data for the patient characteristics, but the following percentage of biomarkers were missing from the sample: GDF-15 < 1%, FGF-21 8%, leptin 17%, prealbumin <1%, and CRP 20%. Fifty-three percent of patients were female, and 59% were less than 2 years of age. Approximately 70% of patients had WAZ or HAZ < 0, 34% had FTT (WAZ or HAZ < −2). Sixty-nine percent of children had a prealbumin <11 mg/dL, and 22% had elevated CRP.

Table 1.

Baseline Characteristics.

Characteristics	Patients (n = 238)
Females	127 (53.4%)
Age (months)
0-6	67 (28.2%)
6-12	46 (19.3%)
12-24	27 (11.3%)
>24	98 (41.2%)
WAZ
>0	68 (28.6%)
0 to −2	108 (45.4%)
<−2	62 (26.1%)
HAZ
>0	74 (31.1%)
0 to −2	105 (44.1%)
<−2	59 (24.8%)
Single ventricle physiology	35 (14.7%)
Failure to thrive	81 (34%)
Elevated CRP	52 (22%)
Biomarker	Median [IQR]
GDF-15 (pg/mL)	907 [617.8–1620.9]

Abbreviations: CRP C-reactive protein; GDF-15, growth differentiation factor 15; HAZ, height-for-age Z score; IQR, interquartile range; WAZ, weight-for-age Z score.

GDF-15 Association With WAZ and HAZ

Table 1 displays the baseline characteristics of the sample. As shown in Figure 1, there was a moderate negative correlation between GDF-15 and WAZ (r = −0.369, P<.0001) and HAZ (r = −0.359, P < .0001), which constitutes the major finding of this study. Prealbumin was not correlated with WAZ or HAZ. FGF-21, leptin, and CRP had no or weak correlations with WAZ and HAZ; results for these biomarkers can be found in Supplement S2, Supplemental Figure 1 and Supplemental Figure 2. The normal reference range of GDF-15 in healthy children has been reported to be 155 to 584 mg/dL.¹⁶ In categorical analysis of FTT with a WAZ or HAZ <−2, GDF-15 was significantly elevated in children with FTT (median 1587 mg/dL [933-2548] vs. 776 mg/dL [512-1201] in children without FTT, P < .001). Figure 2 provides the ROC for GDF-15, which could significantly discriminate FTT with a AUC of 0.75 (0.68-0.82). When stratified by age, the correlation of GDF-15 to WAZ and HAZ was strongest in children 0 to 12 months (r = −0.326, P = .0004; r = −0.350, P = .0002) and 12 to 24 months (r = −0.404, P = .037; r = −0.407, P = .035), respectively.

Figure 1.

Correlation plots with line of best fit for log-transformed GDF-15 and WAZ/HAZ. GDF-15, growth differentiation factor 15; HAZ, height-for-age Z score; WAZ, weight-for-age Z score.

Figure 2.

ROC curve for GDF-15 and FTT. FTT, failure to thrive; GDF-15, growth differentiation factor 15; ROC, receiver operating characteristic.

Inflammation and FTT

Overall inflammatory state likely plays an important role in modifying growth. CRP > 0.5 mg/dL was used as a measure of whole-body inflammation. When stratified by inflammatory status, the GDF-15/WAZ correlation was −0.380 in the normal CRP group and −0.403 in the elevated CRP group, and the GDF-15/HAZ correlation was −0.346 in the normal CRP group and −0.423 in the elevated CRP group (all P < .0001).

Disease Subgroups

FTT for the overall cohort was 34%. As shown in Table 2, the proportion of children with FTT was highest for children with Tetralogy of Fallot (TOF) with pulmonary atresia (69%), atrioventricular septal defect (63%), ventricular septal defect (VSD), and/or patent ductus arteriosus (PDA) (43%), all of which can produce physiologic states of diastolic runoff and/or left-to-right shunting leading to symptoms of congestive heart failure.

Table 2.

FTT and GDF-15 Levels by Type of CHD.

Group	Age (months, median)	% FTT	GDF-15 (pg/mL, median [IQR])
All patients/Categories	14.1	34.5	907 [613.20-1628.42]
Left-sided obstruction (n = 16)	30.4	6.3	528.96 [361.69-737.62]
ASD (39)	43.5	23.1	804.02 [579.84-1096.33]
AVSD (26)	6.0	65.4	2244.00 [1504.60-5774.00]
TOF/DORV (TOF type) (27)	5.5	37.0	838.68 [713.84-1484.04]
VSD/PDA (51)	12.6	43.1	1175.36 [532.18-2137.00]
Single ventricle CHD (31)	8.8	29.0	921.54 [775.88-1463.72]
TOF/PA (13)	10.8	69.2	1076.90 [789.74-1737.15]
Other (35)	18.3	14.3	848.26 [520.59-1306.66]

Abbreviations: ASD, atrial septal defect; AVSD, atrioventricular septal defect; CHD, congenital heart disease; DORV, double-outlet right ventricle; FTT, failure to thrive; GDF-15, growth differentiation factor 15; PA, pulmonary atresia; PDA, patent ductus arteriosus; TOF, tetralogy of Fallot; VSD, ventricular septal defect.

Adjusted Analysis

Multivariable logistic regression models predicting the presence of FTT, including age, sex, and quartiles of biomarker levels are presented in Table 3. Higher GDF-15 quartiles remain significantly associated with the presence of FTT even after adjustment for age and sex.

Table 3.

Age- and Sex-Adjusted Associations Between Failure to Thrive and Biomarkers.

Biomarker quartile	Odds ratio	95% confidence interval	P value
GDF-15 quartile 4 (1620-30,550)	10.407	3.806-28.452	<.0001
GDF-15 quartile 3 (907-1620)	3.438	1.279-9.246	0.0144
GDF-15 quartile 2 (618-906)	1.820	0.659-5.029	0.2480
GDF-15 quartile 1 (110-617)	Reference

Discussion

FTT is common in children with CHD, has important implications for timing of surgery, and contributes to adverse long-term outcomes. In this study of multiple potential growth markers, only increasing levels of GDF-15 were correlated with lower weight-for-age and height-for-age z-scores in our population of CHD patients, even in the setting of inflammation, which has not been previously studied. Chronic inflammation is common in children with CHD and has been associated with many CHD-related comorbidities; however, its role as a contributor or mediator of FTT has not been previously explored.¹⁷ The results of our study suggest GDF-15 has good classification performance for FTT based on its AUC, which holds promise for GDF-15 as a potential growth biomarker for children with CHD.

Identifying a plasma growth biomarker is important. Currently, we rely on patients objectively failing to thrive based on WAZ/HAZ before we diagnose them with FTT and subsequently intervene. It would be helpful to identify a plasma biomarker within the metabolic pathway of FTT that is simple to measure and obtain, which would allow us to identify patients who are at the beginning of failing to thrive but who have not yet crossed into the FTT diagnosis based on WAZ/HAZ. By identifying these patients prior to the clinical manifestations of FTT, we can initiate nutritional interventions sooner or operate earlier and potentially stave off FTT altogether.

GDF-15 is a stress response cytokine belonging to the transforming growth factor β (TGF-β) superfamily and has been shown to be elevated in states of cellular stress, including cardiac failure, renal failure, liver disease, hypoxia, mitochondrial dysfunction, and cancers.¹⁸ We hypothesize that the association between GDF-15 and FTT in CHD patients demonstrated in our study could be mediated by 2 possible mechanisms. First, previous animal studies have demonstrated that even a moderate increase in GDF-15 can lead to the suppression of appetite,¹⁹ independently of other satiety or hunger hormones, including leptin, glucagon-like peptide 1 (GLP-1), or the melanocortin 4 receptor.¹⁸ This effect is thought to be mediated through the binding of GDF-15 to glial cell-derived neurotrophic factor (GDNF) family receptor alpha-like (GFRAL) in the area postrema and nucleus tractus solitarius of the hindbrain–brainstem.²⁰ Second, an important study by Wang and colleagues demonstrated that cardiac-derived GDF-15 acts at the liver to inhibit growth hormone (GH), decreasing the release of insulin-like growth factor 1 (IGF1).¹²

Previous studies of CHD patients have associated GDF-15 levels with severity of heart failure^21-24 as well as increased clinical events and mortality in Fontan patients.²⁵ Additionally, Wang and colleagues found that GDF-15 was elevated in children with heart disease compared to healthy controls and even more elevated in children with CHD who are in heart failure and have FTT.¹² Our findings corroborate these results as diagnoses known to be associated with heart failure including TOF with pulmonary atresia, AVSD, and VSD had high proportions of FTT and high GDF-15 levels in our cohort.

The currently accepted gold standard nutritional marker is prealbumin¹⁰; however, while it is a marker of short-term nutritional status with a half-life of 2 days, there is no correlation between prealbumin and WAZ or HAZ in our cohort, and therefore, it does not appear to be a marker of sustained growth. Additionally, prealbumin is a negative acute-phase reactant and cannot be used reliably in the setting of inflammation,¹⁰ a common phenomenon in children with CHD, which is supported by our data. The ability to predict FTT, or conversely growth, using a plasma biomarker would fill a critical need for children with CHD as it could assist clinicians in determining the optimal timing for surgical intervention and allow for intervention with an individualized nutritional plan or additional medical therapy. In addition, repeating a growth biomarker measurement after the implementation of an intervention might allow for rapid assessment of the intervention’s future growth effects, which would not be clinically evident for some time.

Limitations

Given the cross-sectional study design, longitudinal changes in both GDF-15 serum concentrations and anthropometric measurements were not assessed or correlated. Future prospective studies that follow patients longitudinally should be conducted to further confirm these results. Additionally, there is significant heterogeneity in the anatomy and physiology of different CHD diagnoses, which may limit the generalizability of these findings to specific subsets of CHD patients. We recognize that certain CHD anatomy and physiology such as AVSD, single ventricle physiology, and VSD/PDA are at higher risk for FTT, and in fact, our results show that these groups have higher median GDF-15 levels than the overall cohort median GDF-15 level. This finding supports further work in these CHD subsets. Our multivariable analyses were able to account for only age and sex, while future modeling should consider diagnoses, type of surgery, and additional factors.

Future Research

Further describing the pathways in which GDF-15 plays a role is an important next step as there may be other important cytokines/metabolites involved in the FTT pathway. The results of this study support the need for longitudinal measurement of GDF-15 levels to determine its predictive ability for FTT, which is the ultimate goal of this work. Future work should be directed at obtaining longitudinal data and gathering more detailed patient information. Additionally, noninvasive detection of GDF-15 from saliva and urine will be investigated to eliminate phlebotomy required to obtain plasma samples and improve the ease of testing.

Conclusion

GDF-15 holds promise as a novel growth biomarker for children with CHD based on its association with FTT. Further work is needed to determine its clinical applicability.

Abbreviations and Acronyms

AAP

American Academy of Pediatrics

AUC

area under the receiver operating characteristic curve

CDC

Center for Disease Control and Prevention

CHD

congenital heart disease

CRP

C-reactive protein

FGF-21

fibroblast growth factor 21

FTT

failure to thrive

GDF-15

growth differentiation factor 15

GDNF

glial cell-derived neurotrophic factor

GFRAL

GDNF family receptor alpha-like

GH

growth hormone

GLP-1

glucagon-like peptide 1

IGF-1

insulin-like growth factor 1

HAZ

height-for-age Z score

NIH

National Institutes of Health

WAZ

weight-for-age Z score