Individualized Computer-Based Surgical Planning Addressing Pulmonary Arteriovenous Malformations in Single-Ventricle Patients with Interrupted Inferior Vena Cava and Azygous Continuation

Abstract

BACKGROUND

Single-ventricle patients with interrupted inferior vena cava (IVC) can develop pulmonary arterio-venous malformations due to abnormal hepatic flow distribution (HFD). However, preoperatively determining the hepatic baffle design that optimizes HFD is far from trivial. In this study, we combine virtual surgery and numerical simulations to identify potential surgical strategies for patients with interrupted IVC.

METHODS

Five patients with interrupted IVC and severe PAVMs were enrolled. Their in vivo anatomies were reconstructed from MRI (n=4) and CT (n=1), and alternate virtual-surgery options (intra/extra-cardiac, Y-grafts, hepato-to-azygous and azygous-to-hepatic shunts) were generated for each. HFD was assessed for all options using a fully validated computational flow solver.

RESULTS

For patients with a single superior vena cava (SVC, n=3), intra/extra-cardiac connections proved dangerous, as even a small left or right offset led to a highly preferential HFD to the associated lung. Best results were obtained with either a Y-graft spanning the Kawashima to split the flow, or hepato-to-azygous shunts to promote mixing. For patients with bilateral SVCs (n=2), results depended on the balance between the left and right superior inflows. When those were equal, connecting the hepatic baffle between the SVCs performed well, but other options should be pursued otherwise.

CONCLUSION

This study demonstrates how virtual-surgery environments can benefit the clinical community, especially for rare and complex cases such as single-ventricle patients with interrupted IVC. Furthermore, the sensitivity of the optimal baffle design to the superior inflows underscores the need to characterize both pre-operative anatomy and flows to identify the best suited option.

4. Introduction

Pulmonary arteriovenous malformations (PAVMs) are an uncommon but lethal complication following the superior cavopulmonary palliation of single-ventricle heart diseases, which is characterized by the appearance of intrapulmonary arterial-to-venous shunts that bypass the pulmonary gas exchange units and result in a net decrease in oxygen saturation. These shunts lead to a drop in pulmonary vascular resistance, thereby directing more flow to the diseased lung, and can create a positive feedback loop of increasing hypoxemia. One single-ventricle subgroup that is especially at risk for PAVMs is children having an interrupted inferior vena cava (IVC) with azygous (AZ) continuation, for whom PAVM occurrence can be as high as 18–21% after the Kawashima procedure(1). Although the underlying mechanism leading to PAVMs is unknown, studies have shown that liver-derived factors present in the hepatic venous blood prevent the formation of PAVMs(1–4). Following the Kawashima procedure with the completion of the total cavopulmonary connection (TCPC) was thus suggested as an efficient mean to restore hepatic flow to the lungs and resolve PAVMs(5).

Unfortunately, the TCPC does not completely eliminate the risk of PAVM development. Poor design of the IVC/hepatic venous baffle may lead to an unbalanced hepatic flow distribution (HFD) to the left and right lungs, and in turn to unilateral PAVMs. As discussed by Sundareswaran and de Zélicourt et al.(6), the low flow rate coming through the hepatic baffle in patients with interrupted IVC (which only accounts for ~20–25% of the systemic venous return, as opposed to the normal 50–60% carried by the IVC) increases the complexity of the flow interactions at the center of the TCPC, which in turn increases the difficulty of identifying the best surgical approach for a specific patient based on anatomical considerations alone. Hence, interrupted IVC patients remain at greater risk than other single-ventricle cases for the development of PAVM, even with a completed TCPC. Once the extent of PAVMs is such that oxygen saturation is critically low, the only palliative option is to re-operate and re-orient the hepatic baffle to achieve a better HFD(1).

Our group recently presented a novel surgical-planning framework(6) that enables surgeons to virtually perform multiple surgical scenarios and determine the one that will yield the best hemodynamic performance before even entering the operating room. Such a surgical-planning platform offers a unique solution for cases such as these where hemodynamics (and in particular HFD) could make or break the surgery, but where the small patient population and large number of anatomical variations pose a severe obstacle to the establishment of surgical guidelines from clinical studies alone.

In this study, we review our experience in the planning and optimization of the Fontan surgery in order to characterize the options available to the surgeons to prevent/revert PAVMs in single-ventricle patients with interrupted IVC. Although based on a small sample size, the systematic review of the TCPC variations investigated for each patient sheds light on the impact of caval offset and other design parameters on TCPC hemodynamics, and allows for a better understanding of the anatomy- or flow-based characterizations that can be made for the definition of “global” surgical strategies.

5. Patient Data

Five single-ventricle patients with interrupted IVC and severe PAVMs were enrolled at the Children’s Hospital of Philadelphia or Children’s Healthcare of Atlanta for surgical planning at the Georgia Institute of Technology. The study was approved by the institutional review boards of all institutions. Patients were separated into two groups based on the configuration of the superior venous returns. Group A, or single SVC group, included the three patients for whom the superior vena cava (SVC), AZ and inominate vein (IV) merged together at or before the pulmonary arteries (PA). Conversely, Group B included the two patients for whom the IV did not adjoin the SVC but connected directly to the PAs in the form of a persistent left SVC (LSVC). PAVMs were diagnosed by angiographic appearance and pulse oximetry (72±4% at the time of referral). Table 1 summarizes the clinical, anatomical and flow information relevant to this study.. Patients A1, A2 and B2 had a completed TCPC and were diagnosed with unilateral PAVMs. Patient A3 was still at the Kawashima stage. Patient B1 had previously undergone a TCPC that had been taken down due to a clot in the hepatic baffle. Both A3 and B1 were diagnosed with bilateral PAVMs.

Table 1.

Summary of the clinical and flow characteristics of the five patients enrolled in the study. EC TCPC: completed TCPC with an extra-cardiac baffle. IA TCPC: completed TCPC with an intra-atrial baffle

		Group A: Single SVC			Group B: Persistent LSVC
		A1	A2	A3	B1	B2
	Gender	Female	Male	Female	Female	Male
	Age (years)	4	6	3	4	11
	BSA (m2)	0.61	0.75	0.635	0.64	1.22
	TCPC Stage	EC TCPC	IA TCPC	Kawashima	Kawashima (TCPC taken down)	EC TCPC
Systemic Venous Return	QS	3.43 L/min	3.43 L/min	3.86 L/min	2.93 L/min	3.95 L/min
	QHepV (%QS)	21%	21%	32%	21%	10%
	QAZ (%QS)	24%	24%	17%	15%	30%
	QSVC (%QS)	33%	33%	51%	n/a	n/a
	QIV (%QS)	22%	22%		n/a	n/a
	QRSVC (%QS)	n/a	n/a	n/a	38%	40%
	QLSVC (%QS)	n/a	n/a	n/a	26%	20%
PAVM Location		Left	Right	Bilateral	Bilateral	Right
Oxygen Saturation	Pre-op	72%	67%	75%	76%	70%
	Post-op (follow-up time)	94% (5 months)	80% (<1month)	97% (15.5 months)	92% (25 months)	87% (4.5 months)

6. Methods

The pillars of the framework used in this study include: clinical imaging and volume rendering of the in vivo anatomy and flow; virtual surgery; and performance quantification using numerical modeling.

Clinical Imaging

Most patients underwent a cardiac magnetic resonance imaging (MRI) evaluation, including an anatomical dataset of the entire thorax to reconstruct the cardiovascular geometry(7); and through-plane velocity maps acquired across all inflows (AZ, IV, SVC, hepatic veins (HepV)) and outflows (left and right PAs, LPA and RPA, respectively) for flow boundary conditions. Patient A2 featured coils and pacer wires that prevented exposure to magnetic radiations; therefore anatomical information for that patient was instead obtained using X-ray computed tomography (CT). The reconstructed anatomies of all five patients are illustrated in Figure 1. To compensate for the lack of flow data from CT for patient A2, inflow conditions were imposed using the inflow conditions of patient A1 who had a fairly similar body surface area and age. Flow information extrapolated from other patients is shown in light gray in Table 1 to be easily differentiated from the patient-specific flow measurements.

Figure 1.

Patient anatomies seen from the posterior side. The heart is shown in red for the two patients without completed TCPC to illustrate the constraints imposed by the surrounding vascular structures. Orientation axis: S-superior; I-inferior; L-left; R-right.

Virtual Surgery

Multiple corrective anatomies were generated for each patient using a novel virtual-surgery interface(8). These surgical options (including intra/extra-cardiac grafts, bifurcated Y-shaped grafts, HepV-to-AZ or AZ-to-HepV shunts) were carefully selected based on the currently available surgical techniques and the “available space” around the connection. The geometrical constraints imposed by the surrounding organs were accounted for by including the heart and great vessels into the interface, as illustrated in Figure 1 for patients A3 and B1.

Performance Quantification

Numerical simulations were conducted on all envisioned options, using an in-house unstructured immersed-boundary method(9) to model the complex anatomical flows with elevated spatial resolution. Inflow boundary conditions were prescribed using the in vivo MRI flow rates averaged over the cardiac cycle. Outflows on the other hand, tightly depend on the surgical design retained. If an option successfully increases HFD to the diseased lung, PAVMs will regress, increasing the lung resistance on that side and subsequently decreasing the flow to that lung. Accordingly, each surgical option was tested over a wide range of LPA/RPA flow ratios to best predict HFD (and ultimately PAVM) evolution. HFD was assessed as a post-processing by uniformly seeding weightless point particles across the HepV cross-section and quantifying the flux of hepatic particles exiting through the LPA and RPA, respectively.

7. Results and Discussion

7.1. Patients with single SVC (Group A)

Figure 2 shows selected options generated for the single SVC patients, along with the achieved HFD. The superimposed black arrows show the main flow direction of the superior venous returns, while the dashed lines denote the axis of the PAs and hepatic baffle. The HFD is qualitatively represented by the red and blue streamtraces and quantified in percent under each option. Options yielding an HFD that closely follows the global flow distribution are highlighted in green, as potential candidates for the surgery, at least in terms of HFD performance.

Figure 2.

HFD for selected TCPC options for the three single SVC patients. Percentages indicate HFD to the left lung. Superimposed black arrows show the main flow direction of the superior inflows. Dashed lines denote vessel axes. Orientation axis: S-superior; I-inferior; L-left; R-right.

The first two columns compare the effect of different anastomosis - or offset - locations between the hepatic baffle and the Kawashima connection. As is demonstrated in column (a), any offset to the left of the Kawashima connection results in an almost uni-sided HFD to the left lung, with more than 90% of the flow going to the LPA for all three patients. A symmetric behavior was observed when shifting the hepatic baffle to the right of the Kawashima connection, resulting in an almost uni-sided HFD to the RPA. Such offset configuration corresponds to what had originally been implemented for Patient A2 (Option A2(a)) and had led to severe PAVMs in the contra-lateral lung, which is in line with previous literature relating PAVMs to a low supply in hepatic nutrients (1–4).

The second column of Figure 2 displays the results obtained when seeking to avoid any offset and aiming for the geometrical center of the connection in an effort to improve mixing between the superior inflows and the hepatic blood flow. This approach is only successful for Patient A3, who had the highest hepatic flow rate and for whom the superior and hepatic flows effectively collide in a head-on fashion. For Patients A1 and A2 on the other hand, the azygous flow slightly skews the superior inflows towards the LPA, such that the hepatic flow enters the connection to the right of the superior inflows, resulting in a highly biased HFD to the RPA. Aiming for the geometrical center of the connection thus does not warrant an adequate HFD; rather, it is the position of the baffle relative to the center of the superior blood streams that appears to be key in achieving the desired HFD. However, as is illustrated with Patients A1 and A2, this “flow center” is difficult to precisely locate based on anatomical considerations alone, and may evolve with time depending on the configuration of the superior vessels and their relative contribution to the systemic venous return.

In order to circumvent the difficulties posed by the definition of an optimal offset relative to the “flow center”, three other options were attempted: 1) dividing the hepatic baffle into two branches, reaching to the left and right of the Kawashima connection (Y-graft); 2) merging the azygous flow into the hepatic baffle, to increase the momentum of the flow coming through the baffle and increase mixing at the center of the connection (AZ-to-HepV); or conversely, 3) routing the hepatic flow to the azygous vein to force the mixing of all venous returns in the Kawashima connection (HepV-to-AZ). Success of the Y-graft for Patient A2 stems from the fact that the left and right branches successfully spanned the Kawshima connection, and that the graft bifurcation was designed with a relatively small angle allowing for a proper splitting of the hepatic flow into the two branches. For Patient A1, the overriding aorta did not allow the left branch of the graft to reach beyond the Kawashima connection, resulting in a poor HFD performance, similar to the pre-operative “no offset” configuration. Similarly, the overriding aorta prevented the implementation of a bifurcated graft for Patient A3. It is noteworthy that these anatomical constraints were circumvented using an intra-atrial approach for the left branch in Patient A2.

AZ-to-HepV shunts (column (d)) decrease the sensitivity to offset, increasing mixing at the center of the connection, as is illustrated by the hepatic streamtraces that penetrate deeper into the connection area than without the shunt. However, this procedure was not among the best performers for any of the patients. Part of the reason is that the AZ-to-HepV shunts were designed in combination with the existing pre-operative baffles in order to minimize the surgical complexity. Better performances might have been obtained by seeking to also modify the baffle offset or by doing surgical-planning prior to the first surgery to optimize the original baffle with combined AZ and HepV flows.

Connecting the HepV to the azygous (column (e)) was established as a suitable option in terms of HFD for all three patients. Because that option produced thorough mixing of all inflows, HFD was found to closely follow the global cardiac output (CO) distribution. In patient A3 in particular, the Hep-to-AZ graft lead to a 36% HFD to the left lung when 50% CO went to that lung. On average, HFD to the left lung scaled as (0.9±0.2)*CO to the LPA. However, that option also increased energy expenditure due to the increased viscous dissipation in the AZ. This issue might become prevalent in young patients with small vasculature where small AZ dimensions might lead to elevated power losses.

Based on the above, the approaches that were recommended for surgery for the three single SVC patients were a Y-graft for Patient A2, a no-offset intra-atrial graft for Patient A3, since those combined both HFD and energy efficiency, and an HepV-to-AZ shunt for Patient A1, setting the priority on optimizing HFD even at a higher energy expenditure.

7.2. Patients with bilateral SVCs (Group B)

Figure 3 shows the HFD performance associated with selected surgical planning options envisioned for Patient B1. The top row (Options a-c) demonstrates the impact of offset variations. As for single SVC patients, it may be noted that small changes in the baffle anastomosis location relative to the SVC led to large variations in HFD. For single SVC patients, an option that appeared to work systematically was to circumvent the problem altogether using a HepV-to-AZ shunt. However, as is illustrated by Option (d), this approach fails for bilateral SVC patients, resulting in a configuration similar to a strong offset towards the LPA (or the RPA if the azygous was connected to the RSVC instead of the LSVC) with a uni-sided HFD, which may ultimately favor PAVMs in the contra-lateral lung.

Figure 3.

HFD for selected TCPC options for Patient B1, with under a global cardiac output distribution of 50/50 RPA/LPA. HFD to the right and left lung are indicated below the figures (in blue and red, respectively). Orientation axis: S-superior; I-inferior; L-left; R-right.

On the other hand, while the most challenging question for the single SVC cases was where the center of the superior inflows was, the separation of the superior inflows in bilateral SVC patients (with the RSVC on the right and the AZ and LSVC on the left) might provide an easier answer to that question, with the mid-PA acting as the ideal “no offset” location. For Patient B1 for example, the RSVC flow rate (1.1 L/min) almost exactly matched the LSVC and AZ flows combined (1.2 L/min). As a result, when the baffle was connected to the mid-PA segment as in Option (c), the RSVC flows to the RPA, the AZ+LSVC flows go to the LPA, and it was the hepatic flow that adjusted its distribution between the LPA and RPA to achieve the desired global flow distribution, resulting in an HFD of 57/43 RPA/LPA for a global flow distribution of 50/50. The two other positive approaches in terms of HFD were to divide the hepatic flow into two branches, using either a Y-graft (Option (f)), or an H-connection combining an extra-cardiac graft offset towards the RPA and an HepV-to-AZ shunt to reach the LPA (Option (g)), both of which had very much the same effect as the mid-PA connection.

Patient B2 represents the most challenging anatomical and flow configurations encountered in the course of this study. As can be visualized from Figure 1, the RSVC attached to the PAs very close to the LSVC anastomosis site, such that the mid-PA segment was not well defined. Furthermore, the LSVC only accounted for 20% of the systemic venous return, while the AZ and RSVC combined accounted for 70%, yielding highly unbalanced left and right contributions. Finally, the hepatic blood flow represented only 10% of the cardiac output.

Given the extremely low hepatic flow rate, HFD for Patient B2 was extremely sensitive to the TCPC design and to the competition with the superior inflows. Figure 4 illustrates the flow patterns and HFD that were observed in four selected options for Patient B2. The detrimental consequence of the extremely low hepatic flow rate is apparent in Options (a) and (b), where the RSVC does not encounter any resistance from the hepatic blood flow, penetrates deep into the baffle, and forces the hepatic flow towards the LPA, thereby preventing any hepatic flow to the RPA. After investigating a number of alternate options, the two best approaches were to either perform an H-connection using the pre-operative extra-cardiac graft and a HepV-to-AZ shunt (Option (c)), or to use a Y-graft in combination with an AZ-to-HepV shunt (Option (d)).

Figure 4.

Flow patterns and HFD for selected options for Patient B2, under a global cardiac output distribution of 50/50 RPA/LPA. Orientation axis: S-superior; I-inferior; L-left; R-right.

Based on the results obtained for Patient B1, the mid-PA connection might appear as an attractive “by default” option, all the more since parametric variations revealed that the HFD did not depend on the type of graft retained (extra-cardiac or intra-atrial), nor on the exact anastomosis location as long as the graft connected somewhere between the RSVC and LSVC. However, it should be emphasized that the success of this approach for Patient B1 relied on the fact that the three following conditions were met: 1) the mid-PA segment was well defined, preventing any interaction between the left and right superior venous return, and allowing for an easy connection of the hepatic baffle; 2) the hepatic flow rate was not too low, accounting for about 20% of the total systemic venous return; 3) and, most importantly, the right superior venous return (here the RSVC) was almost exactly equal to the left superior venous inflows (here the AZ+LSVC). This last condition should be considered with care, especially when optimizing the TCPC for younger patients. Indeed, having equal left and right superior venous return at the time of TCPC surgery in younger patients does not warrant that this balance will hold as the patient grows. Azygous, hepatic, RSVC and LSVC flow rates will most likely evolve with age, which might alter the balance of the left and right superior inflows and could lead to a severely imbalanced state, such as in Patient B2.

Alternate options of interest are the Y- or H-graft configurations that seek to divide the hepatic flow between a left and right branch. A particularity of patients with bilateral SVCs is that, as pointed out for Patient B1, HepV-to-AZ shunts have the same effect as an offset towards the PA closest to the azygous. An H-connection for these patients thus circumvents the difficulties encountered with Y-grafts, allowing the hepatic blood flow to reach the PA farthest away from the hepatic veins without the spatial constraints imposed by the aortic arch. Finally, in patients where the hepatic flow is very low, or where a vast majority of a flow comes in from the side where the AZ is connected (as was the case for patient B2), diverting the AZ flow from the Kawashima into the hepatic baffle and then optimizing the baffle design might represent a suitable approach.

8. Conclusion

In summary, we presented the surgical planning studies conducted by our group for single-ventricle patients with interrupted IVC, reviewing the pre-operative performance, a range of re-operation scenarios, and the recommended surgical strategies. Although post-operative MRI data are lacking, the recommended approaches led to increased oxygen saturations (going from 72±4% pre-operatively to 90±7% after surgery), testifying to improved HFD and successful surgical implementation. Contrasting the recommended re-operative scenarios to the pre-operative anatomies that were determined by the surgeon based on anatomical considerations alone, this study demonstrates how clinical MR imaging and new virtual-surgery environments can benefit the clinical community, especially for rare and complex cases such as single-ventricle patients with interrupted IVC.

Next, this first systematic surgical-planning investigation provides a set of guidelines to be followed when treating patients with interrupted IVC to optimize HFD and ultimately minimize the chances of PAVMs. Due to the low flow rate coming through the TCPC baffle, HFD was very sensitive to the offset between the hepatic baffle and the “flow center” of the superior venous returns, and even a small offset towards the LPA or RPA led to a highly preferential HFD to the associated lung. For patients with interrupted IVC and a single SVC, “classic” intra-atrial or extra-cardiac approaches thus proved dangerous as it is not necessarily evident to identify the anastomosis location minimizing the offset between the hepatic baffle and the “flow center” of the superior venous returns; all the more that this location highly depends on the complex interaction of the AZ, IV and SVC flows. This fact alone underscores the value of the present techniques with regard to pre-operative surgery planning. In absence of a patient-specific offset optimization, Y-grafts or HepV-to-AZ shunts appeared as two attractive alternatives. HepV-to-AZ shunts ensured a thorough mixing of all venous returns, although at higher energy expenses. When successful, a Y-graft allowed for both energy and HFD efficiency, but their implementation was in practice hindered by the aorta. This technique is still being explored and, with further refinement, may become a viable and successful option for a larger patient pool.

For patients with bilateral SVC, results depended on the balance between the left and right superior inflows. When those were close to equal, optimal results were obtained by centrally connecting the hepatic baffle between the SVCs, along the mid-PA segment. Alternate options included H-connections or routing the azygous flow into the hepatic baffle and subsequently optimizing the baffle design.

Finally, all of the cases reported herein demonstrate a strong sensitivity of the optimal baffle design upon the distribution of the systemic venous return between the different vessels and the resultant flow interaction. Determining the best suited option for a given patient thus requires a detailed analysis of not only anatomy but also the flow distributions, keeping in mind that the measured flow rates may evolve as the patient ages. Future work should thus involve a detailed error analysis, to determine the sensitivity of a potential optimum to variations in the systemic flow rates.