You are here

Article Text

Article menu

Download PDFPDF

Congenital heart disease
Original article
Exercise capacity in single-ventricle patients after Fontan correlates with haemodynamic energy loss in TCPC
  1. Reza H Khiabani1,
  2. Kevin K Whitehead2,
  3. David Han1,
  4. Maria Restrepo1,
  5. Elaine Tang1,
  6. James Bethel3,
  7. Stephen M Paridon2,
  8. Mark A Fogel2,
  9. Ajit P Yoganathan1
  1. 1Wallace H. Coulter Department of Biomedical Engineering, Georgia Institute of Technology & Emory University, Atlanta, Georgia, USA
  2. 2Division of Cardiology and Department of Radiology, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, USA
  3. 3Westat, Rockville, Maryland, USA
  1. Correspondence to Ajit P Yoganathan, Wallace H. Coulter Department of Biomedical Engineering, Georgia Institute of Technology & Emory University, 387 Technology Circle, Suite 232, Atlanta, GA 30313, USA; ajit.yoganathan{at}bme.gatech.edu

Abstract

Objective Elevated energy loss in the total cavopulmonary connection (TCPC) is hypothesised to have a detrimental effect on clinical outcomes in single-ventricle physiology, which may be magnified with exercise. This study investigates the relationship between TCPC haemodynamic energy dissipation and exercise performance in single-ventricle patients.

Methods Thirty consecutive Fontan patients with TCPC and standard metabolic exercise testing were included. Specific anatomies and flow rates at rest and exercise were obtained from cardiac MR (CMR) and phase-encoded velocity mapping. Exercise CMR images were acquired immediately following supine lower limb exercise using a CMR-compatible cycle ergometer. Computational fluid dynamics simulations were performed to determine power loss of the TCPC anatomies using in vivo anatomies and measured flows.

Results A significant negative linear correlation was observed between indexed power loss at exercise and (a) minute oxygen consumption (r=−0.60, p<0.0005) and (b) work (r=−0.62, p<0.0005) at anaerobic threshold. As cardiac output increased during exercise, indexed power loss increased in an exponential fashion (y=0.9671x3.0263, p<0.0001).

Conclusions This is the first study to demonstrate the relationship between power loss and exercise performance with the TCPC being one of the few modifiable factors to allow for improved quality of life. These results suggest that aerobic exercise tolerance in Fontan patients may, in part, be a consequence of TCPC power loss.

  • CONGENITAL HEART DISEASE

http://dx.doi.org/10.1136/heartjnl-2014-306337

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    The palliation for single-ventricle patients generally involves three stages of surgery culminating in the total cavopulmonary connection (TCPC) with, most commonly, either an extracardiac (EC) or intra-atrial (IA) connection. While there have been advancements in clinical management, these patients may develop significant complications in the long term including decreased exercise tolerance, pulmonary arteriovenous malformations, heart failure or protein-losing enteropathy.1–3

    The complex physiology makes it difficult to predict postoperative haemodynamic performance during rest or exercise conditions.4–6 The haemodynamics is significantly different from the biventricular circulation; the single ventricle is forced to perform the work of two ventricles. Several studies have modelled and characterised TCPC haemodynamic parameters such as energy dissipation, wall shear stress or flow resistance.7–11 It has been suggested that patient outcome may be related to TCPC haemodynamic parameters such as power loss (PL)8; however, it is unclear to what extent the haemodynamics of the surgically constructed TCPC impacts the ability of the patient to exercise. Understanding these dynamics might have critical implications for the management of these patients; it could identify specific markers that control outcomes, can potentially elucidate mechanisms of Fontan failure and may be used for the development of optimal surgical and/or clinical management strategies12 that may allow for these children to exercise better and improve their ability to keep up with their peers, ultimately improving their quality of life.

    The goal of this study was to use patient-specific data to investigate the association between haemodynamic energy dissipation within the TCPC and metabolic exercise test performance.

    Methods

    A total of 30 single-ventricle patients who underwent TCPC were included. Consecutive patients were taken from Georgia Tech—Children's Hospital of Philadelphia (CHOP) Fontan database that had completed standard graded metabolic exercise testing as well as resting and exercise cardiac MR (CMR) at CHOP using a 1.5 T Avanto Whole Body system (Siemens Medical Solutions, Malvern, Pennsylvania, USA). Patient data are summarised in table 1. Informed consent was obtained from all patients and all the study protocols complied with the Institutional Review Boards of CHOP and the Georgia Institute of Technology.

    View this table:
    Table 1

    Patient data during rest and exercise at VAT

    Clinical examination and data acquisition

    The inclusion criteria were (1) TCPC with no other sources of pulmonary blood flow and (2) the ability to undergo the metabolic exercise stress test using a stationary cycle ergometer. Exclusion criteria included a pacemaker or implanted metal that precluded adequate imaging and segmentation of the anatomy. Patients enrolled in this study first completed a routine maximal metabolic exercise test using a ramp cycle protocol. Minute oxygen consumption (VO2) was obtained on a breath-by-breath basis. Maximal VO2 and heart rate (HR), as well as VO2, HR and work rate at the ventilatory anaerobic threshold (VAT) were obtained using a metabolic cart (SensorBiomedics V29, Yorba Linda, California, USA). VAT was measured by the V-slope method. The weight and height of all patients were measured to calculate their body surface area (BSA).

    After the metabolic test, the patients completed the resting and exercise CMR study. The imaging protocol, which used parallel imaging, began with anatomic CMR and phase-contrast MR (PCMR) acquisitions under resting conditions. The acquisition protocol consisted of several components: (1) a contiguous axial stack of static steady-state-free precession images used for the 3D reconstruction of the TCPC and (2) through-plane retrospectively gated PCMR with respiratory compensation across the superior (SVC) and inferior vena cava (IVC), descending aorta (DAO), across the aortic (or neoaortic) valve, and right and left pulmonary arteries (RPA and LPA). The CMR data were gated to the cardiac cycle using breath-holding.

    After the resting CMR acquisition, the patients were slid partially out from the MRI bore to perform lower limb exercise using an MRI-compatible supine bicycle ergometer (Lode BV, Groningen, the Netherlands). This ergometer allows revolutions per minute (RPM)-independent workload ranging from 10 to 250 W while the patient maintains their position by bracing themselves with hand grips. The goal was to bring the patients from resting condition up to a steady work rate near their VAT as measured in their metabolic exercise test, which was a sustainable work rate for the completion of the CMR data acquisition. HR was monitored continuously. Initially, the workload was set to 20 W. It was then increased progressively at a rate of 20 W/min to obtain an HR corresponding to that of the HR at VAT on their prior metabolic exercise test. Exercise was then suspended, their feet quickly removed from the ergometer pedals and then returned to isocenter for imaging (generally within 5–10 s). Using this method, PCMR measurements of the AO, SVC and DAO were acquired with repeated exercise performed in between for the patient to return to the target HR. Since patients could not breath-hold after exercise uniformly, velocity maps were averaged (generally three excitations). Due to difficulty in maintaining IVC position at exercise, DAO was substituted for IVC flow at exercise. In the remaining text, the CMR exercise condition at VAT is referred as exercise condition.

    Anatomic and flow segmentation

    Patient-specific 3D anatomies and flow conditions were obtained using CMR data. The resting MRI slices were first interpolated using the adaptive control grid interpolation technique, and then segmented using a bouncing ball algorithm. The 3D geometries were then obtained by surface fitting performed in Geomagic Studio (Geomagic, North Carolina, USA).13 Time-varying velocity fields at rest and exercise, from through-plane PCMR slices, were integrated over vessel cross-sectional areas (eg, at venae cavae and pulmonary arteries) to calculate the associated flow rates throughout a cardiac cycle.14

    Computational fluid dynamics

    An in-house transient flow solver, developed based on a sharp interface immersed boundary method, was used in the numerical simulations.15 Rigid walls were assumed for the TCPC and blood was modelled as a single-phase Newtonian fluid. The inflow boundary conditions were taken directly from PCMR measurements, and the mean flow rates were imposed as a uniform velocity profile at the inlets (IVC and SVC). Outflow boundary conditions (at the pulmonary arteries (PAs)) were prescribed using the in vivo LPA/RPA flow splits obtained from PCMR to ensure mass conservation. Since the movement of patients during exercise limited the accuracy of direct flow measurements in PAs, in modelling exercise haemodynamics LPA/RPA flow split was taken from the resting data as suggested in previous studies.16 ,17

    Calculation of ‘indexed’ power loss

    The blood flow entering the TCPC from IVC and SVC (inlets) loses part of its haemodynamic energy across TCPC due to friction and flow disturbances. As a result, the blood flow exiting TCPC through LPA and RPA (outlets) has lower total energy compared with the entering blood flow. Here, TCPC PL was defined as the rate at which energy is dissipated in the fluid across the TCPC. Therefore, PL was calculated as the difference between the total entering haemodynamic energy (first term in equation 1) and the total exiting haemodynamic energy (second term in equation 1):Embedded Image 1

    where ρ, p, v and Q respectively stand for blood density, hydrostatic pressure, mean velocity and flow rate across each vessel cross section.

    To account for differences in flow and BSA between different patients, all the comparisons and correlations were conducted using indexed power loss (iPL):Embedded Image 2

    where Qs=QIVC+QSVC was the total systemic return. Because PL scales as a cubic function of flow, the resulting iPL can be viewed as a flow-independent resistance index, making it ideal for comparing different patients whose raw PL would be influenced by cardiac output. Dasi et al11 previously demonstrated the flow independence of this parameter.

    Statistical analysis

    All parameters were presented as mean values with SD unless otherwise stated. Regression was used to investigate the correlation between haemodynamic parameters (iPL, Qs) with metabolic test results (Work, VO2). Student t test (or Mann–Whitney test) was used to compare iPL, haemodynamics and exercise parameters among groups. Normality was tested by Shapiro–Wilk test. Fisher's z transformation was used to assess the significance of the difference between two correlation coefficients found in two independent groups (male/female, IA/EC or patients younger/older than 20 years of age). In all analyses, p value<0.05 was considered significant (two-tailed).

    Results

    The metabolic and flow measurements at resting and exercise (at VAT) conditions are summarised in table 1. The average work rate at VAT was 0.9±0.3 W/kg, which was associated with 60% increase in HR and 240% increase in VO2. As expected, there was a significant linear correlation (r=0.82, p<0.0001) between VO2 at exercise and work at exercise.

    Correlation between exercise VO2 and iPL

    As shown in figure 1A, there was a significant negative linear correlation between VO2 at VAT and iPL at exercise (r=−0.60, p<0.0005). The correlation between exercise VO2 at VAT and iPL was not statistically significantly different between male and female groups (p=0.56), IA and EC groups (p=0.7) or the two age groups (p=0.31). As presented in table 2, male patients had significantly (p<0.05) higher VO2 and lower iPL at exercise compared with the female patients. In addition, older (>20 years old) patients had significantly (p<0.05) lower exercise VO2 compared with younger (<20 years old) patients. No significant difference was observed in exercise VO2 between IA and EC groups. There was no significant correlation between iPL and resting VO2.

    View this table:
    Table 2

    Patient data and haemodynamic parameters in male/female, intra-atrial/extra-cardiac and age groups

    Figure 1
    Figure 1

    Correlation between indexed power loss (iPL) at ventilatory anaerobic threshold (VAT) and (A) exercise minute oxygen consumption (VO2) at VAT and (B) exercise work at VAT.

    Correlation between normalised exercise work and iPL

    A significant negative linear correlation was observed (figure 1B) between work (per kg) at VAT and iPL at exercise (r=−0.62, p<0.0005). No statistically significant difference in the correlation was observed between male and female groups (p=0.45), IA and EC groups (p=0.44) or the two age groups (p=0.1). As presented in table 2, older (>20 years old) patients had significantly (p<0.05) lower exercise work (per kilogram) compared with younger (<20 years old) patients. However, no significant difference was observed in exercise work between IA/EC and male/female groups.

    Correlation between Qs and PL

    At exercise, Qs increased significantly (85% on average) compared with the resting condition. As shown in figure 2, there was a cubic relationship between the ratio of raw PL at exercise to rest with the ratio of Qs at exercise to rest. In addition, as presented in figure 3, there was no significant difference (p=0.87, paired t test) between iPL at VAT (iPL_exe) and iPL at rest (iPL_rest). There was a very high intraclass correlation coefficient between resting and exercise iPL (0.99) compared with an intraclass correlation coefficient of 0.2 between resting and exercise raw PL.

    Figure 2
    Figure 2

    Correlation between ratio of power loss (PL) at ventilatory anaerobic threshold (VAT) to PL at rest and ratio of systemic return (Qs) at VAT to Qs at rest.

    Figure 3
    Figure 3

    Comparison between indexed power loss at ventilatory anaerobic threshold (iPL_exe) and indexed power loss at rest (iPL_rest).

    There was no significant correlation between cardiac index and VO2 at rest and exercise (p>0.05). Similarly, there was no significant correlation between EF and VO2 at rest and exercise (p>0.05).

    Discussion

    Fontan patients usually suffer from several long-term complications, some of which have been hypothesised to be related to the TCPC haemodynamics. Flow rates, pressure drops and haemodynamic deficiencies can be magnified during exercise, impeding the child's ability to play normally with peers. Therefore, studying exercise conditions may make such relationships more evident and be a key to improving the quality of life for these youngsters.

    This is the first study to demonstrate the relationship between PL and exercise capacity. The TCPC is one of the few modifiable factors in these children to allow them to exercise better and keep up with their peers. As shown in figure 1, there was a strong negative correlation between iPL, and both VO2 and work at VAT. These relationships suggest that iPL within the Fontan circuit may be linked to aerobic exercise performance and this work leads to the speculation that surgical modification of the TCPC may allow teenagers to exercise better. Data from previous studies suggest that limited capacity to increase ventricular preload with exercise is in large part a result of limited pulmonary blood flow.3 Any increase in resistance in the cavopulmonary pathway has the potential to significantly impair flow and hence ventricular preload. Therefore, the finding that VAT is best preserved in those subjects with the lowest iPL should not be surprising.

    Those subjects with a greater PL will be less able to maintain pulmonary blood flow at tolerable systemic venous pressures during exercise and thus limit their ventricular preload. This will result in decreased ability to increase Qs and the VO2 at which VAT occurs will be decreased. This unique preload dependency during exercise was also shown in the recent large Pediatric Heart Network cross-sectional study of exercise performance in Fontan physiology. In that study, they assessed the contribution to the variance in aerobic capacity and physical working capacity of the three determinants of oxygen delivery during exercise; stroke volume (measured by oxygen pulse), chronotropy (as measured by chronotropic index) and oxygen content (as measured by oxygen saturation). Chronotropy and oxygen content accounted for <5% of the variance in aerobic performance at both VAT and maximal exercise despite markedly low peak HRs. The investigators concluded that differences in chronotropic response or oxygen saturation were minimally responsible for the range of aerobic performance observed in their population. Rather, limited ability to maintain stroke volume was responsible for the vast majority of the variance in aerobic capacity in this population and was felt to reflect the effects of the transpulmonary pressure gradient limiting ventricular preload.3

    TCPC PL has been used in previous studies as a measure of haemodynamic efficiency in TCPC, assuming that higher haemodynamic efficiency may lead to better outcomes.8–11 Using patient-specific data, the present results (figure 2) suggested a strong correlation between PL and increased exercise systemic return: (PL_exer/PL_rest)≈(Qs_exer/Qs_rest)3. This confirms the utility of using the derived iPL, which indexes raw PL to the cube of QS, thus establishing a dimensionless parameter that can be viewed as a flow-independent resistance index. The insignificant difference between resting and exercise iPL and high intraclass correlation between resting and exercise iPL (figure 3) confirms the flow independence of iPL. This flow-independent resistance index is critical in establishing a relationship between Fontan haemodynamics and exercise performance since patients with higher cardiac output will have higher raw PL simply due to the above relationship between Qs and raw PL. It is, in fact, likely the reason previous studies have failed to establish this relationship.

    The current results suggested that the TCPC and its associated haemodynamics may play a role in the aerobic capacity of patients with Fontan physiology and suggest that iPL may be a clinically relevant haemodynamic parameter to help predict exercise tolerance in TCPC patients. The importance of these findings is that the TCPC is a modifiable factor in the outcome of single-ventricle patients. Establishing clinically relevant haemodynamic parameters (such as iPL, hepatic flow distribution between the lungs and others) may help predict which patients may benefit from a modification of their TCPC connection. It may also affect how surgeons approach new and novel TCPC designs. In addition to predetermined templates, past experience and intuition, the surgeon may systematically assess the most haemodynamically efficient option for each patient and potentially consider this extra information prior to the surgery. Modification of surgeries using computational fluid dynamics (CFD) and surgical planning18 may be the first step in designing ‘individualised’ medicine to optimise a youngster's ability to exercise and play with others, trying to lead as normal a life as possible.

    Limitations

    Every effort was made to take the in vivo data with the highest possible accuracy; however, the accuracy of the flow data is limited to CMR resolution especially during exercise. Therefore, the CFD predictions are valid within these limits. For example, the effects of respiration and wall motion were not considered here. In addition, since the patients included in this study were consecutively taken from the available data in our database, the number of cases in different groups was not necessarily equal. For example, there were less number of EC compared with IA TCPCs within the present cohort. A follow-up study with a larger cohort can be useful in identifying the effect of secondary factors such as anatomy or gender on the current findings.

    The metabolic exercise test was performed in the sitting position and the CMR exercise in the supine position, and there are known differences between these physiologies. There was a strong correlation between the VO2 at VAT observed in the metabolic exercise test and exercise work. Wattage in supine exercise required to achieve the goal HR was generally 10–20 W higher than the upright ergometer. This would suggest that the relationship of increased VO2 to changes in the haemodynamics observed in supine ergometry should be at least similar to those that occur in more conventional upright ergometry.

    Conclusion

    This study establishes for the first time the clinical importance of TCPC PL as an indication of exercise performance in Fontan patients. We demonstrate a clear relationship between higher indexed PL and decreased aerobic exercise performance. This has broad implications for the ability of children and adolescents to live a normal life and keep up with their peers at play as the TCPC is one of the few modifiable factors in single-ventricle patients. Further research is required to determine whether it impacts other morbidities related to Fontan physiology. The results may improve the predictive capability of physicians in the management of TCPC patients and affect the treatment planning for these patients.

    Key messages

    • What is already known on this subject?

    • Children with single ventricles have a decreased quality of life and limited exercise capacity for which there are little, if any, options. This is especially important in young people where keeping up with their peers and playing with others are important factors for socialisation.

    • What might this study add?

    • Our study is the first investigation to demonstrate a link between energy loss in the Fontan baffle—a surgical creation—and exercise performance. Furthermore, our study will lay the groundwork for different surgical approaches to the single ventricle, such as changing Fontan geometry, to improve exercise ability in this group of patients, which is so important in youngsters; this paper has the potential to change clinical practice.

    • How might this impact on clinical practice?

    • There are many complications associated with the single ventricle and the Fontan operation. We present important data demonstrating an association between the energy characteristics of the Fontan pathway and exercise performance, suggesting that at least one of these problems can potentially be improved. This avenue of investigation will open up a new approach to a patient population that take up a substantial portion of time of any healthcare worker involved in the care of congenital heart disease. As adults with congenital heart disease now outnumber children, this study has broad applicability to both adult and paediatric cardiology as well as cardiothoracic surgery.

    References

    1. ↵
      1. Mair DD,
      2. Puga FJ,
      3. Danielson GK
      . The Fontan procedure for tricuspid atresia: early and late results of a 25-year experience with 216 patients. J Am Coll Cardiol 2001;37:933–9.
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Brassard P,
      2. Bedard E,
      3. Jobin J,
      4. et al
      . Exercise capacity and impact of exercise training in patients after a Fontan procedure: a review. Can J Cardiol 2006;22:489–95.
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      1. Paridon M,
      2. Mitchell P,
      3. Colan S,
      4. et al
      . A cross-sectional study of exercise performance during the first 2 decades of life after the Fontan operation. JACC 2008;52:99–107.
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Hjortdal V,
      2. Christensen T,
      3. Larsen S,
      4. et al
      . Caval blood flow during supine exercise in normal and Fontan patients. Ann Thorac Surg 2008;85:599–603.
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Hsia T,
      2. Cosentino D,
      3. Corsini C,
      4. et al
      . Use of mathematical modeling to compare and predict hemodynamic effects between hybrid and surgical Norwood palliations for hypoplastic left heart syndrome. Circulation 2011;124:S204–10.
      OpenUrlAbstract/FREE Full Text
    6. ↵
      1. Marsden AL,
      2. Vignon-Clementel IE,
      3. Chan FP,
      4. et al
      . Effects of exercise and respiration on hemodynamic efficiency in CFD simulations of the total cavopulmonary connection. Ann Biomed Eng 2007;35:250–63.
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      1. DeGroff CG
      . Modeling the Fontan circulation: where we are and where we need to go. Pediatric Cardiol 2008;29:3–12.
      OpenUrlCrossRef
    8. ↵
      1. Whitehead KK,
      2. Pekkan K,
      3. Kitajima HD,
      4. et al
      . Nonlinear power loss during exercise in single-ventricle patients after the Fontan: insights from computational fluid dynamics. Circulation 2007;116:I165–71.
      OpenUrlPubMedWeb of Science
    9. ↵
      1. Dur O,
      2. DeGroff CG,
      3. Keller BB,
      4. et al
      . Optimization of inflow waveform phase-difference for minimized total cavopulmonary power loss. J Biomech Eng 2010;132:031012.
      OpenUrlCrossRefPubMed
    10. ↵
      1. Khiabani RH,
      2. Restrepo M,
      3. Tang E,
      4. et al
      . Effect of flow pulsatility on modeling the hemodynamics in the total cavopulmonary connection. J Biomech 2012;45:2376–81.
      OpenUrlCrossRefPubMed
    11. ↵
      1. Dasi LP,
      2. Pekkan K,
      3. Katajima HD,
      4. et al
      . Functional analysis of Fontan energy dissipation. J Biomech 2008;41:2246–52.
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Pekkan K,
      2. Whited B,
      3. Kanter K,
      4. et al
      . Patient-specific surgical planning and hemodynamic computational fluid dynamics optimization through free-form haptic anatomy editing tool (SURGEM). Med Biol Eng Comput 2008;46:1139–52.
      OpenUrlCrossRefPubMedWeb of Science
    13. ↵
      1. Frakes DH,
      2. Dasi LP,
      3. Pekkan K,
      4. et al
      . A new method for registration-based medical image interpolation. IEEE Trans Med Imaging 2008;27:370–7.
      OpenUrlCrossRefPubMed
    14. ↵
      1. Sundareswaran K,
      2. Frakes D,
      3. Fogel M,
      4. et al
      . Optimum fuzzy filters for phase-contrast magnetic resonance imaging segmentation. J Magn Reson Imaging 2009;29:155–65.
      OpenUrlCrossRefPubMed
    15. ↵
      1. De Zelicourt D,
      2. Ge L,
      3. Wang C,
      4. et al
      . Flow simulations in arbitrarily complex cardiovascular anatomies—an unstructured Cartesian grid approach. Comput Fluids 2009;38:1749–62.
      OpenUrlCrossRefWeb of Science
    16. ↵
      1. Pedersen EM,
      2. Stenbog EV,
      3. Frund T,
      4. et al
      . Flow during exercise in the total cavopulmonary connection measured by magnetic resonance velocity mapping. Heart 2002;87:554–8.
      OpenUrlAbstract/FREE Full Text
    17. ↵
      1. Cheng CP,
      2. Herfkens RJ,
      3. Lightner AL,
      4. et al
      . Blood flow conditions in the proximal pulmonary arteries and vena cavae: healthy children during upright cycling exercise. Am J Physiol Heart Circ Physiol 2004;287:921–6.
      OpenUrlCrossRef
    18. ↵
      1. Fogel MA,
      2. Khiabani RH,
      3. Yoganathan AP
      . Imaging for pre intervention planning: pre and post-Fontan procedures. Circ Cardiovasc Imaging 2013;6:1092–101.
      OpenUrlFREE Full Text

    Footnotes

    • RHK and KKW contributed equally.

    • Contributors All authors were fully involved in the study design, data analysis and interpretation of results, preparation of the manuscript and final approval of the manuscript. All authors have approved the submitted manuscript.

    • Funding This study was supported by the National Heart, Lung, and Blood Institute Grants HL67622, HL098252 and HL089647.

    • Competing interests None.

    • Patient consent Obtained.

    • Ethics approval Institutional Review Boards of Children's Hospital of Philadelphia and Georgia Tech.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    Linked Articles