Quantification of lymphatic burden in patients with Fontan circulation by T2 MR lymphangiography and associations with adverse Fontan status

Abstract

Aims

To quantify thoracic lymphatic burden in paediatric Fontan patients using MRI and correlate with clinical status.

Methods and results

Paediatric Fontan patients (<18-years-old) with clinical cardiac MRI that had routine lymphatic 3D T2 fast spin echo (FSE) imaging performed from May 2017 to October 2019 were included. ‘Lymphatic burden’ was quantified by thresholding-based segmentation of the 3D T2 FSE maximum intensity projection image and indexed to body surface area, performed by two independent readers blinded to patient status. There were 48 patients (27 males) with median age at MRI of 12.9 (9.4–14.7) years, time from Fontan surgery to MRI of 9.1 (5.9–10.4) years, and follow-up time post-Fontan surgery of 9.4 (6.6–11.0) years. Intraclass correlation coefficient between two observers for lymphatic burden was 0.96 (0.94–0.98). Greater lymphatic burden correlated with post-Fontan operation hospital length of stay and duration of chest tube drainage (r_s = 0.416, P = 0.004 and r_s = 0.439, P = 0.002). Median lymphatic burden was greater in patients with chylous effusions immediately post-Fontan (178 (118–393) vs. 113 (46–190) mL/m², P = 0.028), and in patients with composite adverse Fontan status (n = 13) defined by heart failure (n = 3), transplant assessment (n = 2), recurrent effusions (n = 6), Fontan thrombus (n = 2), and/or PLE (n = 6) post-Fontan (435 (137–822) vs. 114 (51–178) mL/m², P = 0.003). Lymphatic burden > 600 mL/m² was associated with late adverse Fontan status with sensitivity of 57% and specificity of 95%.

Conclusion

Quantification of MR lymphatic burden is a reliable tool to assess the lymphatics post-Fontan and is associated with clinical status.

Graphical Abstract

Open in new tabDownload slide

Introduction

Advances in patient selection, operative technique, and post-operative care has significantly improved survival in patients with Fontan circulation with 20-year survival in the current surgical era near 85%.^1–3 However, management of associated late morbidity is an ongoing challenge.^4,5 Late complications can have varied presentations including the phenotype of the failing Fontan circulation, which is frequently characterized by some combination of arrhythmias, recurrent effusions and ascites, chronic low cardiac output state, thromboembolic complications, peripheral oedema, and lymphatic dysfunction, or much less commonly ventricular pump failure and its end-organ consequences. The involvement of lymphatic dysfunction in the deterioration of the Fontan circulation is increasingly becoming evident, however the exact mechanisms are less clear. While elevated central systemic venous pressure is ubiquitous in Fontan patients, frank intractable derangement of the lymphatic circulation is selective and often not directly related to the magnitude of central venous congestion. Hence, it is difficult to prognosticate which patients will develop these complications, but important to do so given that outcomes can be poor in Fontan patients with lymphatic insufficiency.^6–8

Non-invasive T₂-weighted magnetic resonance lymph angiography (T2MRL) is widely available and can be used to routinely evaluate the lymphatic circulation in single ventricle disease.^9,10 The distribution of lymphatic abnormality on pre-Fontan T2MRL correlates with immediate post-operative Fontan outcomes.^10,11 In our experience, post-Fontan children and adolescents also show a wide range of lymphatic abnormalities on routine T2MRL imaging, with failing Fontan patients exhibiting particularly abnormal lymphatics. Beyond the pattern of abnormal lymphatic distribution, the aggregate volume of abnormal lymphatics may allow for more nuanced risk stratification and follow-up. However, quantification of the burden of lymphatic disease in individual patients has not been previously attempted.

We hypothesize that lymphatic burden in Fontan patients can be reliably quantified on T2MRL and may be a useful tool for patient surveillance and prognostication. The objectives of our study were thus to (i) test the feasibility and inter-observer reproducibility of a thresholding-based quantitative T2MRL imaging measure of lymphatic burden in children with Fontan circulation and (ii) investigate associations between T2MRL lymphatic burden and clinical and haemodynamic parameters.

Methods

Study population

This is a retrospective study approved by the local research ethics board with waiver of informed consent. Consecutive patients < 18-years-old status post-Fontan surgery who had clinical cardiac magnetic resonance imaging (MRI) performed at our institution between May 2017 and October 2019 were included. May 2017 was when T2MRL was adopted into our routine Fontan protocol. Patients without T2MRL were excluded as some studies early in the adoption period were missing the sequence. Indications for MRI included either (i) routine assessment prior to fenestration closure, (ii) routine Fontan surveillance per institutional protocol (typically performed twice in adolescence, between the ages of 12–14 and 16–18), (iii) investigation for failing Fontan physiology, or (iv) for other specific clinical queries, such as exercise intolerance or vascular stenoses.

Cardiac MRI technique

All exams were acquired on a 1.5-T system (Avanto Fit; Siemens Healthineers, Erlangen, Germany) using a standard institutional Fontan protocol with patients fasting for at least 4 h and typically not requiring general anaesthesia for the purposes of lymphatic imaging. T2MRL was performed with a commercially available, heavily T2-weighted, respiratory-navigated, 3D fast spin-echo sequence (sampling perfection with application optimized contrasts using different flip angle evolution; SPACE) acquired in coronal plane including the neck and thorax. Typical sequence parameters were TR 3500–5000 ms, TE 700 ms, flip angle 140°, isotropic voxel size 1.0–1.4 mm, NEX 2, integrated parallel acquisition technique acceleration factor of 2. A single maximum intensity projection (MIP) image of the entire volume was automatically generated by the scanner. All examinations also included ventricular and blood flow assessment in line with established guidelines.¹² In brief, ventricular volumetry was performed with balanced steady-state free precession short-axis cine images and blood flow was assessed with through-plane velocity-encoded phase-contrast images in vessels of interest.

Lymphatic burden quantification

‘Lymphatic burden’ was quantified by thresholding-based segmentation of the scanner-generated T2MRL coronal MIP using the myocardial T2W module in QMass (Medis Suite v3.1, Medis Medical Imaging Systems, Leiden, Netherlands), as depicted in Figure 1. Specifically, the thoracic region was defined as the part of the image including the entire thorax, axilla, supraclavicular regions and lower neck. The inferior limit was to just above the dome of the highest hemi-diaphragm, to exclude any intra-abdominal signal. An uninvolved region within the defined thoracic region was contoured to set the inner spatial boundary for thresholding analysis. Automated thresholding of signal intensity > 5 standard deviations beyond the minimal region of interest (ROI) was done to include hyperintense areas with dilated lymphatics. Contours were manually corrected by the reader via correlation with source images to include all visible thoracic lymphatics and exclude hyperintense non-lymphatic structures. This provides a surrogate measurement of lymphatic volume in mL, which was indexed to BSA to provide a quantitative measure of ‘lymphatic burden’. Lymphatic burden measurements were done by two independent readers (A.V.H. and J.A., 1 year of cardiac MRI experience) blinded to patient status.

Clinical and haemodynamic data

Clinical and haemodynamic data were collected via chart review. Pre-Fontan surgery cardiac catheterization pressures were recorded. Immediate post-Fontan morbidity was evaluated including post-operative length of stay, post-operative days with chest tube, and presence of chest tube chylous drainage, which was based on presence of chylomicrons or a lymphocyte fraction ≥80%.⁴ The most recent echocardiographic exam at the time of last follow-up was reviewed for severity of atrioventricular valve regurgitation. Cardiac MRI ventricular volumetry and phase-contrast blood flow values at time of T2MRL were recorded.

Adverse Fontan status

Clinical charts were reviewed for post-Fontan PLE, plastic bronchitis (PB), readmissions for management of recurrent pleural or pericardial effusions, thrombus in the Fontan circulation, heart failure post-discharge after Fontan surgery, or heart transplant assessment at any time since Fontan surgery. Diagnosis of PLE was based on an elevated a-1antitrypsin level in the stool together with the presence of serum hypoalbuminemia and symptoms of oedema without another identified cause.⁷ Diagnosis of PB was based on expectoration of casts and histologic examination (when possible), and bronchoscopy when clinically indicated. Heart failure was defined as initiation of heart failure therapy, specifically diuretics, ace-inhibitors, beta blockers, or inotropic agents, in the inpatient or outpatient setting, for either echocardiographic evidence of ventricular dysfunction or for cardiologist’s assessment of clinical heart failure. Referral and acceptance by the heart transplant team for a complete transplant assessment was counted as heart transplant assessment. Adverse status was divided into early, late, and combined composite groups. Early adverse status was defined as any of the above in the immediate 6 months following Fontan surgery. Late adverse status was defined as the any of the above occurring more than 6 months following Fontan surgery. Composite adverse Fontan status was defined as presence of any one or more of these features at any time point post-Fontan surgery.

Statistical analysis

The mean quantified lymphatic burden for each individual patient between the two observers were used as the lymphatic burden for statistical analysis. Variables were expressed as median with interquartile ranges in parentheses. Lymphatic burden was analysed in relation to clinical and hemodynamic variables using Wilcoxon rank-sum, Kruskal–Wallis or Fisher’s exact tests as applicable for categorical variables, and Spearman rank correlation coefficients for continuous variables. Inter-observer reproducibility was assessed using intraclass correlation coefficients (ICCs) and Bland–Altman plots. All analyses were conducted assuming a significance level of 5% and implemented using R v3.5.3.

Results

Fifty-four unique paediatric patients with Fontan circulation underwent a clinical cardiac MRI during the study period (Table 1). Six were excluded as T2MRL was not performed, leaving a final cohort of 48 patients. Twenty-seven of the 48 (44%) were male. Median age at MRI was 12.9 (9.4–14.7) years. Median age at Fontan surgery was 3.4 (2.9–3.8) years, median duration from Fontan operation to MRI was 9.1 (5.9–10.4) years, and median follow-up post-Fontan surgery was 9.4 (6.6–11.0) years. Hypoplastic left heart syndrome (HLHS) was the most common diagnosis constituting 15/48 (31%) of the cohort. Twenty-eight of 48 (58%) had a dominant right ventricle. Five patients had an open fenestration at time of MRI. Thirty-six MRIs were performed per institutional protocol for routine surveillance, six for failing Fontan physiology, three for routine pre-fenestration closure, and three for other specific clinical queries (two to investigate desaturation/exercise intolerance, one to follow-up a stenotic left pulmonary artery stent).

T2MRL lymphatic burden quantification

Median T2MRL scan time was 5:55 (5:18–6:45) minutes. Median lymphatic burden for the entire cohort was 136 (64–255) mL/m². Quantification of lymphatic burden showed good interrater reliability with an ICC of 0.96 (95% CI: 0.94–0.98). Bland–Altman plots are depicted in Figure 2 and show no significant systemic bias with a mean difference of 22 mL/m² and 95% limits of ± 170 mL/m² but with greater variation along the severe spectrum of lymphatic burden (which appears to drive the relatively wide limits). The individual measurements for the four data points beyond the limits of agreement were above the 87th percentile for the mean lymphatic burden between observers. There was no significant difference in lymphatic burden based on any categorical demographic features. This data are summarized in Table 1.

Lymphatic burden and pre-Fontan catheterization

Lymphatic burden at post-Fontan MRI did not correlate with pre-Fontan mean pulmonary artery pressures (r_s = 0.083, P = 0.58), transpulmonary gradient (r_s = 0.108, P = 0.53), or systemic ventricle end-diastolic pressure (r_s = −0.061, P = 0.69), (Table 2).

Lymphatic burden and immediate post-Fontan morbidity

Lymphatic burden correlated with post-operative length of stay (r_s = 0.416, P < 0.01) and post-operative days with chest tube (r_s = 0.439, P < 0.01). Lymphatic burden was also greater in those with post-operative chylous chest tube drainage compared to those without chylous drainage, with median lymphatic burden of 178 (118–393) vs. 113 (46–190) mL/m², P = 0.03. This data are depicted in Figure 3.

Lymphatic burden and cardiac MRI and echocardiographic haemodynamics

Lymphatic burden did not correlate with end-diastolic volume, ejection fraction, ascending aorta flow, superior vena cava flow, Fontan flow, pulmonary artery flow (Qpa), pulmonary vein flow (Qpv), ratio of pulmonary venous to systemic venous flow (Qpv:Qs), or aortopulmonary collateral flow as measured by cardiac MRI. Lymphatic burden was not different between subjects with moderate or greater atrioventricular volume regurgitation compared with those with less than moderate atrioventricular volume regurgitation by echocardiography at most recent follow-up, with median lymphatic burden of 213 (247–388) vs. 122 (58–220) mL/m², P = 0.26. The correlations of lymphatic burden with continuous clinical and hemodynamic variables are summarized in Table 2.

Lymphatic burden and adverse Fontan status

Nine of 48 (19%) patients had early adverse Fontan status, specifically 6/48 (12%) with recurrent effusions, 1/48 (2%) with thrombus, and 2/48 (4%) with heart failure. Seven of 48 (15%) patients had late adverse Fontan status, specifically 6/48 (12%) with PLE, 1/48 (2%) with thrombus, 1/48 (2%) with heart failure, and 2/48 (4%) with heart transplant assessment. Three of these patients had multiple late adverse events, one patient with PLE and heart failure, one with PLE and a heart transplant assessment, and one with thrombus and a heart transplant assessment. These three patients were also the only patients that had both an early and late adverse event, with a median lymphatic burden of 822 (629–980) mL/m². No patients had PB. This data with lymphatic burden per sub-group is summarized in Table 3.

Overall, patients with early adverse status had higher lymphatic burden than those who did not [435 (137–810) vs. 118 (53–207) mL/m², P = 0.03]. Similarly, those with a late adverse status had higher lymphatic burden than those who did not (822 (273–1,137) vs. 119 (53–190) mL/m², P = 0.004). Finally, when we considered those who had composite adverse Fontan status (n = 13); those with an adverse status had a higher lymphatic burden than those who did not [435 (137–822) vs. 114 (51–178) mL/m², P < 0.001], depicted in Figure 4.

Lymphatic burden > 600 mL/m² was associated with early adverse Fontan status with a sensitivity of 44% and specificity of 95%, late adverse Fontan status with a sensitivity of 57% and specificity of 95% and composite adverse Fontan status with a sensitivity of 46% and specificity of 100%.

Discussion

In this study, thresholding-based segmentation of T2MRL MIP images was used to generate quantitative measurements of lymphatic burden in children and adolescents with a Fontan circulation. The key findings of our work are (i) quantification of T2MRL lymphatic burden is feasible in paediatric Fontan patients and shows reasonable inter-observer reproducibility, (ii) patients with evidence of early lymphatic leak with chylous effusions and prolonged chest tube drainage immediately post-Fontan show high T2MRL lymphatic burden in later follow-up, and (iii) patients with late adverse Fontan status have higher concurrent T2MRL lymphatic burden.

The potential advantages of a quantitative approach over previous pattern-based approaches¹⁰ are that quantification may allow for better discrimination between patients and permit better follow-up of lymphatic disease within individuals. This method also accounts for possible unusual distributions of thoracic lymphatic disease that do not strictly conform to common patterns of lymphatic derangement. The challenges of this technique include accurately contouring the lymphatic burden on the MIP image, which requires close correlation with source images. The inter-observer reproducibility was reasonable in this study, with outliers in agreement only at the extreme end of severe lymphatic burden, where measurements from either observer were nonetheless greater than the 87th percentile, and so obviously abnormal. It should be emphasized that as these measurements are made on the MIP image, these values do not represent the true lymphatic volume but rather a surrogate volume. Quantifying the true lymphatic volume is unlikely to be feasible for practical use given the greater processing demand of contouring 1 vs. 80 to 100 images per patient and inherent difficulty in accurately contouring tiny individual serpiginous lymphatic vessels on thin slices. These limitations are overcome by the MIP technique and moreover, the resultant surrogate marker of lymphatic volume generated appears to be useful for discrimination between Fontan patients of differing clinical status.

Using this technique, we found that greater lymphatic burden post-Fontan correlates with early adverse Fontan status, particularly presence of chylothorax and longer duration of chest tube drainage. Our data complements the work of others who have shown that abnormal patterns of lymphatic distribution seen by pre-Fontan MRI are associated with early post-Fontan morbidity and complications, such as recurrent effusions and ascites, Fontan takedown, and even death.^10,11,13 It is theorized that certain single ventricle patients have an increased underlying susceptibility to lymphatic insufficiency, which may be due to congenital variation in lymphatic anatomy, variable patterns of lymphatic collateralization in response to elevated central venous pressure, genetic/syndromic predisposition, or previous insults to the lymphatic system such as in chronic in-utero pulmonary venous obstruction.^{10,11,14–18} These factors persist and are exacerbated after the Fontan operation, which imposes a host of associated circulatory stressors, resulting in worse lymphatic burden in follow-up. While the serial progression of lymphatic insufficiency after the Fontan operation is still unknown, of note, time from Fontan had no correlation with lymphatic burden in our cohort. Our work thus suggests that the lymphatic derangement occurs early after Fontan completion and persists.

Expectedly, lymphatic burden was also greater in those with worse late adverse status post-Fontan, with median lymphatic burden nearly seven times greater in patients with vs. without late adverse status, and lymphatic burden > 600 mL/m² representing a highly specific marker associated with late adverse Fontan status. This is not surprising, as in our experience, patients with systemic Fontan failure often have severely abnormal lymphatics on MRI compared with the clinically well Fontan patient. These lymphatic changes can be directly seen and quantified by T2MRL, and in some patients may manifest as specific clinical phenotypes associated with Fontan failure, such as chylous effusions, PB, and PLE. Indeed, our group previously studied 320 Fontan patients and found that patients who developed chylothorax in the acute post-operative period were at higher risk of late death and adverse events such as PLE and PB.¹⁹ Schumacher et al.²⁰ reported a similar trend toward a higher incidence of chylothorax among patients who subsequently developed PLE or PB. Tellingly, while we had no patients with PB, those with chylothorax and PLE had among the greatest lymphatic burden in our study. Of note, the sensitivity of this technique for a late adverse event using a 600 mL/m2 threshold was only 57%, which may reflect the multifactorial nature of Fontan failure. Nonetheless, our work shows that lymphatic imaging can provide insight into the mechanism of Fontan failure and has potential as an adjunctive non-invasive imaging biomarker of Fontan health.

Limitations

The study has the inherent limitations of being a single center retrospective study Because there was a variable amount of time between the Fontan operation and the catheterization, echocardiographic, MRI, and clinical outcome variables, it is difficult to differentiate the effects of time; this will be the focus of future research. In addition, because our imaging field of view did not consistently include the entire abdomen we were unable to include the abdominal lymphatic burden in our study. Due to the sample size and large number of investigated parameters, multivariable analysis was not able to be effectively performed. Further, only clinically indicated cardiac MRIs performed during Fontan follow-up were included with variable indications, raising potential for referral selection bias. Patients who already died or had a heart transplant would not be included.

Conclusion

In this study, we have shown that quantification of T2MRL lymphatic burden in the Fontan patient is feasible, reproducible, and results in larger values in patients with adverse Fontan status. Lymphatic burden > 600 mL/m² is associated with late adverse Fontan status with 95% specificity. Future research is needed to understand the evolution of lymphatic derangement in the Fontan circulation and determine the predictive utility of lymphatic quantification for the management of Fontan patients.

Acknowledgements

Kyle Runeckles and Steve Fan for invaluable statistical support.

Funding

None.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

References

Author notes