Mechanical circulatory support is a valuable option for patients suffering from end-stage heart failure, both as a bridge to transplant and as destination therapy. Now, small implantable continuous flow pumps for left ventricular support with a low complication rate and good long-term durability are available.1–3
However, up to 30% of patients suffer from biventricular failure requiring additional RV support.4 For these patients, only two long-term options are available: the total artificial heart (TAH), requiring that the native heart be excised, and the biventricular ventricular assist device (VAD), using extracorporeal or implantable displacement pumps. Both alternatives enable only limited quality of life for the patients, and the long-term results are dismal compared with those of patients with univentricular support.4
It was our goal to evaluate a method that would allow the use of the commercially available implantable centrifugal left ventricular assist device (LVAD) HeartWare HVAD (HeartWare Inc., Framingham, MA) as a right ventricular assist device (RVAD).
In making this attempt, three issues had to be solved:
1. The fluid-mechanical design of the HVAD-LVAD is calculated with the performance requirements of the systemic circulation, pumping 3–10 L of blood per minute with a rotational speed range of 1800–3800 rpm against a pressure head of 50–200 mm Hg. With lower afterload, as in the pulmonary circulation, even with the lowest adjustable speed the flow would be too high, theoretically resulting in severe pulmonary edema (Figure 1).
2. The inflow cannula of the HVAD is 35 mm in length. Less the height of the implantation ring of 5 mm and an estimated median wall thickness of the right ventricle of 4 mm, there is an effective length of 26 mm, which would extend into the cavity of the right ventricle. Although the RV dimensions are usually increased in RV failure, this might be too much, especially when the patient is hypovolemic with subsequently reduced RV dimensions.
3. Because there is only little experience with connecting continuous flow VADs to the right ventricle, the optimal anatomical site for this connection had to be determined.
A mock circulation was used to examine the flow characteristics of the HeartWare HVAD (Figure 2). This mock circulation model allowed the simultaneous measurement of the delivered pump flow and the pressure gradient across the VAD with different levels of pump speed. The results were plotted as a performance diagram (flow-pressure curves).
According to the Hagen-Poiseuille equation, the flow resistance is mainly dependent on the diameter of the outflow tube, with an impact of the 4th power.
A family of flow versus pressure curves with the rotational speed (rotations/min) as a parameter has been evaluated with variations of the outflow diameter (10–4 mm).
Furthermore, we looked for a commercially available material, which would allow safe and satisfactory reduction of the effective length of the inflow cannula.
Finally, to estimate the optimal implant position of the VAD, we investigated echocardiographic analyses of the hearts of 15 candidates for biventricular VAD support. The RV diameters were measured in images obtained in parasternal long-axis and short-axis views (RV end-diastolic outflow tract diameter) and apical four-chamber view (maximum transverse diameter at the base, near the tricuspid valve ring and maximum longitudinal diameter, measured from the tricuspid valve ring to the RV apex).
Optimization of the Flow-Rotational Speed Relationship
In the case of a normal pulmonary resistance of 100–200 dyn · s · cm−5, the pump would deliver between 6.5 and 8 L/min even with the lowest adjustable pump speed of 1800 rpm.
To allow for a “physiological” flow range of 3–6 L/min within a pump speed setting of between 1800 and 3500 rpm, as usually set when the HVAD is used as an LVAD, the effective resistance has to be artificially increased. Best hydrodynamical performances of impeller and bearings are achieved with a pump speed of >2200 rpm, as recommended by the manufacturer. Therefore, we decided to reduce the outflow graft diameter stepwise to such a degree that the afterload of the VAD reaches the usual levels of the systemic circulation. The length of the section where the outflow graft was narrowed was 35 mm.
The resulting additional resistance for different outflow graft diameters is depicted in Figure 3. As this resistance is flow dependent, for every flow level, a curve can be drawn that illustrates the relationship between the graft diameter and the resulting additional resistance.
According to this data, we decided to reduce the outflow graft to an inner graft diameter of ∼5 mm. This was achieved by side clamping and narrowing the graft with a simple suture (6 × 0 Prolene), as shown in Figure 4, A and B. To ensure a definite and reproducible reduction of the diameter to close to 5 mm, we used a 5 mm Hegar bar for calibration (Figure 4B). The artificial stenosis was made 3 cm distal of the beginning of the outflow graft.
As after this procedure, the performance diagram of the HeartWare HVAD in RV position overlaps that for its use as an LVAD (Figure 5), the pump could now work with optimal speed settings of between 2200 and 3500 rpm as an RVAD.
Reduction of the Effective Inflow Cannula Length
To reduce the effective length of the inflow cannula, we decided to add two 5 mm silicon suture rings covered with Dacron Velour (DHZB in-house product, made by Berlin Heart GmbH, Berlin, Germany) to the original implantation ring (Figure 6, A and B). The two rings were fixed to the original “apical” fixation ring using Bioglue (CryoLife, Atlanta, GA). These additional rings prevent the cannula from deep penetration into the RV cavity.
Evaluation of the Optimal Implant Position
In contrast to the left ventricle, the right ventricle has no well-defined apex where a VAD could be connected. Many patients have, especially in the apical RV region, many trabeculae. As a consequence, two different pump positions have been under consideration: connection to the anterior free wall or to the diaphragmatic wall of the right ventricle.
Two-dimensional echocardiographic analyses of the hearts of 15 candidates for biventricular VAD support were investigated and diameters of the right ventricle in the long axis and different short axes were measured. The mean right ventricular end-diastolic diameter (RVEDD) was 36 ± 6.8 mm.
As a result of these measurements, we decided to connect the pump to the anterior free wall of the right ventricle. With this implantation site, there is the greatest distance between the tip of the inflow cannula and the opposite interventricular septum.
As the HeartWare HVAD is intended for use as an LVAD, it is constructed for pumping against a systemic vascular resistance of 600–3500 dyn · s · cm−5. The normal pulmonary resistance is found between 100 and 200 dyn · s · cm−5. With our idea of performing a “banding” procedure at the outflow graft of the VAD, we added a second resistance which is in serial connection to the patient's own pulmonary resistance. Depending on the patient's pulmonary resistance, it could be necessary to add up to 1100 dyn · s · cm−5 with the “banding” of the graft.
The most effective way to achieve an additional resistance is to reduce the diameter of the outflow graft, because according to the Hagen-Poiseuille equation, its internal radius influences the resistance to the 4th power. However, in many patients with end-stage biventricular heart failure, the native pulmonary resistance is—depending on the cardiac output— increased up to systemic levels. Banding performed in a clinical setting should consider this. For patients with pulmonary hypertension, a reduction of the outflow graft diameter to 6 or even only to 7 mm can be appropriate. Otherwise, an HVAD pump in right ventricular position would have to run with much higher pump speed than normally as an LVAD.
In clinical practice, the patient's pulmonary resistance is a dynamic value. During the weeks or months of left ventricular assistance, an initially increased pulmonary resistance can become lower or can even completely drop to normal values. How far this process can go, and which final level of pulmonary resistance will be reached, is unpredictable. However, the range of possible recovery of the pulmonary circulatory bed, the reserve of the pump speed spectrum after our “banding” procedure should be sufficient to guarantee an appropriate flow within usual settings. From our measurements we can calculate that, with an outflow graft diameter of 6 mm during a pulmonary vascular resistance (PVR) reduction from 800 to 160 dyn · s · cm−5, a constant pump flow of 6 L/min could be assured by reduction of the pump speed from 3200 to 2400 rpm, which is entirely within the usual clinical range of speed settings.
As a consequence, it is not probable that in patients with pulmonary hypertension, initial banding to a diameter of 6 mm can become insufficient during the further postoperative cause. However, a theoretically optimal solution for this issue could be a dynamic restriction device, which can be adjusted transcutaneously.
Generally, it is a limitation of this study that measurements from a mock circulation model cannot be applied directly to clinical situations.
The reduction of the effective length of the inflow cannula by simply increasing the thickness of the implantation ring by the use of additional distance rings leads to an increased height of the extraventricular part of the whole device. In this case, the pump is not advanced as deeply into the right ventricle, and therefore, more of the device stays in front of the ventricle. This requires enough space between the chest wall and the anterior wall of the right ventricle. Especially in slim patients with a short distance between the sternum and the vertebral body, this can be critical. On the other hand, the majority of right ventricles are enlarged in the case of RV failure, and after mechanical unloading by a VAD, their diameter should reduce, resulting in some extra space for the pump.
In our candidates for biventricular support, the RVEDD was significantly increased with a mean of 36 mm. This, however, would not be enough to implant the original pump without reduction of the effective length of the inflow cannula. A longitudinal orientation of the inflow cannula within the right ventricle by implantation from the inferior aspect of the right ventricle, however, would probably facilitate suction of the thin free RV wall toward the interventricular septum by a Bernoulli mechanism.
Right ventricular dimensions vary during the postoperative course. As the left ventricle becomes smaller in most patients after LVAD implantation, this can also be expected for the right ventricle when supported by an RVAD. This could potentially lead to a situation where even the reduced effective length of the inflow cannula could be too long for the RV diameter. Having this in mind, further clinical studies should elaborate whether an inferior approach to the RV cavity might be better.
As the reduction of RV dimensions can also be accompanied by improved RV function, less blood could flow through the right pump causing stasis with subsequent pump thrombosis. Further clinical studies should clarify whether an RVAD has then to be removed or could be simply stopped and be left in place.
The HeartWare LVAD pump in its original design seems to be also usable as a RVAD after a few, but important, modifications of the implantation procedure. A reduction of the outflow graft diameter to 5 or 6 mm together with the addition of two 5 mm rings to the original HeartWare sewing ring to reduce the effective length of the inflow cannula allows stable VAD function. The anterior free wall of the right ventricle can be used as the implantation site.
After first clinical reports,5,6 systematic evaluation of this therapeutic strategy in various clinical subsets is mandatory.
We thank the Friede Springer Foundation for supporting our experimental laboratory unit.
The authors further thank Mrs. Anne Gale for editorial assistance.