Caroticocavernous fistula or cavernous sinus dural arteriovenous fistula (CS-DAVF) has presented various treatment challenges over many years. This paper outlines these challenges in a review of the literature, and attempts to address them by analyzing the anatomical and hemodynamic characteristics of 32 consecutive patients with CS-DAVF treated between 2007 and 2016, in doing so proposing novel strategies for safe access and treatment of CS-DAVF.
- Arteriovenous Malformation
- Liquid Embolic Material
Statistics from Altmetric.com
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.
Indirect caroticocavernous fistula, more accurately described as central skull base or cavernous sinus dural arteriovenous fistula (CS-DAVF), presents a number of therapeutic challenges. These include: (1) the anatomy of the CS, which consists of a network of interconnected venous pockets, one or many of which may receive shunt arteries; and (2) access to the CS, especially when the inferior petrosal sinus (IPS), the most direct and easily accessible channel draining the CS, is absent or occluded and/or the superior orbital vein (the most common ‘plan B’ access route) is small.1–3
Traditionally, the CS has been accessed via the IPS and occluded with coils to obliterate the compartment(s) of the sinus receiving shunts. This approach has met with some success, but the literature suggests a significant rate of treatment failure (failed access, incomplete closure of the shunt, worsening of the shunt symptoms/signs), and also a significant complication rate (venous perforation and subarachnoid hemorrhage, new cranial nerve palsy or worsening of pre-existing cranial nerve palsy) with ‘standard’ strategies.3–7 The author has had similar experiences in this regard, which stimulated investigation of alternative access routes to the diseased CS and alternative choices of embolic agents for treatment of CS-DAVF.
This paper outlines a single center's experience with management of CS-DAVF between 2007 and 2016, addressing alternative approaches to accessing the CS, safety and efficacy of dimethylsulfoxide (DMSO)-soluble liquid embolic materials in closure of the CS, and immediate and 6-month angiographic and clinical outcomes including immediate and delayed complications. A proposed modification of the existing generally used classification system for DAVF is advanced in an accompanying paper.
Between 2007 and 2016, 32 consecutive patients with CS-DAVF of mean age 62 years (median age 63 years) were evaluated and treated. The patient cohort characteristics are summarized in table 1. There were 17 women and 15 men. All patients had six-vessel diagnostic cerebral angiography on admission, followed by elective treatment between 3 hours and 14 days later.
Of the 32 patients with CS-DAVF, 29 had arteriovenous shunts confined exclusively to the dura of the CS. In 29 cases, access to and closure of the affected CS was performed with either Onyx-34 (Medtronic, Irvine, California, USA) or PHIL-35 (Microvention, Tuston, California, USA) alone. Three lesions involved the dura of both the CS and the sphenoparietal sinus and/or Meckel's cave (patients 15, 19, and 22). In two of these patients (patients 15 and 19), transarterial closure of the shunts was performed with Onyx-18 without attempting access to the CS and, in one patient (patient 22), a combination of direct closure of the CS with Onyx-34 and transarterial embolization of the non-CS shunts with Onyx-18 was performed. Another patient (patient 24) had a fistula confined to an isolated pocket at the posterior aspect of the left CS, supplied by a clival branch of the ascending pharyngeal artery, with venous drainage to the leptomeningeal veins of the anterior cranial fossa and no orbital or IPS outlets. This lesion was treated with a single transarterial injection of Onyx-18.
Control angiograms were performed at the conclusion of each procedure and at 6 months post embolization. Procedure and screening times were recorded. Preoperative and postoperative independent assessments of preoperative visual fields, cranial nerve function, and intraocular pressures were performed.
In the patient cohort, one of four access/treatment strategies was adopted: (1) direct percutaneous inferolateral transorbital puncture of the CS (19 patients); (2) transvenous access to the CS (8 patients); (3) transarterial embolization of the feeding external carotid branch(es) (3 patients); (4) combined percutaneous transorbital CS puncture and transarterial embolization (2 patients).
Twenty-one patients received transorbital puncture of the CS as part of their treatment. In 19 patients this was their sole access for CS embolization. In two patients (patients 18 and 22), combined transorbital and transarterial access was employed to close shunts remote from the dura of the CS. In all 20 of these patients, the IPS on the affected side was occluded or absent. Apart from two patients (patients 2 and 5) described later, no attempts were made to access the CS through a closed/absent IPS.
Of the eight patients receiving therapy via transvenous approaches, seven were accessed from the internal jugular vein via the IPS and one (patient 2) was accessed via percutaneous ultrasound-guided puncture of the superior ophthalmic vein (SOV).
Of the three patients receiving transarterial embolization only (patients 15, 19 and 24), two (patients 15 and 19) had fistulas involving both the CS and the dura of the sphenoparietal sinus and/or Meckel's cave. The other (patient 24) had a fistula involving only a small posterior pocket of the CS, as described above.
Of the two patients receiving combined transarterial and percutaneous transorbital CS puncture and embolization, one (patient 18) had previous failure of therapy at another institution with coils in the affected sinus and IPS rendering both transvenous and percutaneous CS access difficult; in this case, percutaneous access to the CS was therefore obtained with a balloon in the ipsilateral internal carotid artery (ICA) for both visualization and protection of the vessel. In the other patient (patient 22), transorbital CS access was sufficient to occlude the cavernous portion of the DAVF, but transarterial embolization was required to occlude the shunts to the dura of the sphenoparietal sinus and Meckel's cave.
Percutaneous transorbital CS puncture
Under general anesthesia, six-vessel cerebral angiography is performed to define the arterial supply and venous drainage of the CS-DAVF. Once IPS occlusion is confirmed, a decision is made to proceed to transorbital CS puncture. The arterial catheter is placed in the common carotid artery on the side of the lesion for angiograms and roadmapping, and placed on flush. The periorbita is cleaned and covered with a fenestrated drape. The frontal imaging plate is angled into an ipsilateral oblique caudal projection to view along the orbital axis, parallel to the orbital floor, aiming for optimal visualization of the superior and inferior orbital fissures and their junction point (figure 1). A small skin incision is made lateral to the infraorbital notch (figure 2). Under ﬂuoroscopic guidance, a 9 cm or 12 cm 20 gauge needle is inserted through the skin incision along the orbital ﬂoor towards the inferomedial aspect of the superior orbital fissure (SOF) (figure 3). The needle tip is kept close to the orbital ﬂoor with the bevel inferiorly, the point superiorly, especially during the initial advancement, to facilitate passage of the needle across the irregular bone of the orbital floor (figure 4). As the needle approaches the SOF, a roadmap of the ICA is acquired, and the needle advanced gradually (with the bevel now facing superiorly, the tip inferiorly) until the tip is placed inferior to the anterior genu of the ICA on the lateral roadmap, projected over the inferior ophthalmic vein (IOV) or the confluence of the SOV and IOV on both planes. The stylet of the needle is removed and arterialized blood noted, indicating successful access to the arterialized vein. The wire from a Cook 5 F Micropuncture Access Set (Cook Medical, Bloomington, Indiana, USA) is advanced through the needle into the CS under lateral fluoroscopy, its tip placed in the most posterior venous pocket possible. The needle is then removed over the wire and the 5 F Micropuncture sheath system advanced over the wire with a gentle twisting motion until its tip is positioned in the CS. The inner dilator of the set is removed, leaving the soft 5 F (1.7 mm) outer dilator in situ. A hemostatic valve is attached to the external end of the outer dilator, which is then secured to the skin. Confirmation of sheath position is obtained with a very gentle injection of contrast through the sheath (figure 5). A DMSO-compatible microcatheter is inserted through the outer dilator into the CS. A venogram is performed through the microcatheter to conﬁrm precise placement of the catheter in relation to the shunting compartment(s) of the sinus. If necessary, the microcatheter can be navigated to the most appropriate position for embolization. The dead space of the microcatheter is ﬁlled with DMSO and Onyx-34 or PHIL-35 is then injected gradually into the CS. The status of the CS-DAVF is monitored with intermittent angiography. Following closure of the fistula, control angiography is performed before closure of the femoral access site. The microcatheter is removed from the orbital sheath under gentle traction before the sheath is also removed. Manual hemostasis at the infraorbital puncture site is then achieved and the wound cleaned and dressed.
The overall results in our cohort are summarized in table 1.
Of 32 patients with CS-DAVF, 29 were treated in one procedure. All 32 lesions were completely occluded on immediate post-embolization angiography, with no clinical or angiographic recurrence of the target lesion at or beyond 6 months.
Three patients required a second procedure (patients 2, 5, and 18). One (patient 5) was our first experience of percutaneous transorbital CS puncture. At the initial procedure, unsuccessful attempts were made to access the CS via an occluded IPS. Eventually, a microcatheter was advanced to a point in the clival plexus adjacent to the shunts and Onyx-34 embolization undertaken, failing to penetrate the CS or to reduce the degree of shunting. The patient developed a permanent ipsilateral abducens nerve palsy following this procedure, presumably due to Onyx injection adjacent to and around Dorello's canal. This patient's second procedure, our first experience of percutaneous transorbital CS puncture, was uneventful and resulted in complete and durable occlusion of the CS and the fistula with no new clinical sequelae.
The second patient (patient 2) requiring a second procedure had also had an unsuccessful attempt at rewiring an occluded IPS. She had a second transvenous procedure employing ultrasound-guided puncture of the SOV with uneventful access to the CS and closure with Onyx-34.
The third patient who required a second procedure had previously undergone coil embolization of the CS via the IPS at another center (patient 18). This treatment had failed to occlude the shunt and had caused deterioration in symptoms and signs as the coils, while failing to occlude the arteriovenous shunts, had also blocked the IPS, resulting in retrograde drainage to the orbit through a small infraorbital vein. At the initial procedure in our institution, the coils in the CS and the small size of the infraorbital vein rendered transorbital puncture challenging and visualization of the inferior surface of the carotid siphon very difficult. After several attempts, the initial procedure was abandoned for fear of needle penetration of the ICA. For the second procedure 2 days later, a Hyperglide 4×20 mm balloon (Medtronic) was placed in the carotid siphon to allow visualization of the vessel and protection of its lumen during embolization. The second procedure was uneventful and resulted in complete and durable clinical and angiographic occlusion of the fistula.
Procedure duration, measured as the time between the first and last fluoroscopic or digital subtraction angiography acquisitions in a particular procedure, and total screening time data were collected. In patients who had two procedures, screening and procedure times refer to the final curative procedure (not the times for unsuccessful access via the occluded IPS) in two cases and, in one case, cumulative time over two procedures for the patient who required a second attempt at percutaneous CS puncture with transarterial balloon protection. In our 32 patients the average procedure time was 55 min, with an average screening time of 30 min. In 19 patients receiving transorbital CS puncture only, average procedure and screening times were 40 and 21 min respectively; for transvenous approaches, 65 and 37 min respectively. For combined procedures, screening and procedure times were correspondingly longer, but the small numbers in this group render any detailed analysis of these times meaningless.
Complications in our series were independently documented and are summarized in table 2.
In 28 patients who received injections of Onyx-34 or PHIL-35 into the affected CS as their sole treatment (20 after transorbital CS puncture and 8 after transvenous access), there was no immediate postoperative cranial neuropathy. One patient (patient 25), who presented with progressive and near-complete bilateral IVth and VIth nerve palsies and a partial left IIIrd nerve palsy, progressed to a complete left IVth nerve palsy and new left ptosis at 48 hours post procedure, associated with headaches. CT and MRI were unremarkable. This was felt by the consulting neurologist and ophthalmologist to be related to ongoing venous thrombosis rather than the effects of DMSO, which should have been completely excreted by that time.8 There was no change in visual acuity and intraocular pressures had returned to normal. At 8 weeks post procedure, bilateral IVth and VIth nerve palsies had nearly resolved, ptosis had completely resolved, and there was only slight residual (pre-existing) left IIIrd nerve palsy.
As described above, one patient (patient 5) who received transvenous Onyx-34 experienced a permanent VIth nerve palsy, presumably resulting from a misplaced Onyx injection into the clival plexus at the time of the first procedure. There was no new or progressive cranial neuropathy following a second procedure, which involved transorbital occlusion of the CS with Onyx-34.
In the three patients who had transarterial embolization as their only treatment modality (patients 15, 18, and 24), all fistulas were fed by accessory meningeal branches of the external carotid artery (ECA) and/or clival branches of the ascending pharyngeal artery. In these cases it is our routine practice to inform patients of the risk to sensation in the V1 and V2 distributions, which we quote as being 50–75% with a 10–20% chance of recovery, based on our institutional experience. We do not proceed to treatment in these patients unless they accept this risk as part of their consent process. In the three patients in our series with such lesions, all three developed paresthesia post-procedurally, two patients in the V2 distribution and one in the V1 distribution. Another patient described above, who had both transorbital embolization of the CS and transarterial embolization with Onyx-18 for a large component of the DAVF along the sphenoparietal sinus and Meckel's cave, developed malar paresthesia in the V2 distribution. Of these four patients, the patient with V1 paresthesia (patient 24) had full recovery of sensation by 6 months. One patient (patient 22) has permanent V2 distribution paresthesia with no improvement at 12 months and the other two (patients 15 and 19) have experienced partial recovery (subjective return of sensation in 30–50% of the affected malar region).
In the 29 patients in this series who received Onyx-34 or PHIL-35 into their CS, there were no instances of embolic material penetration of the ICA or its dural branches. There were no instances of mechanical neuropraxia or DMSO-related neurotoxicity affecting the IIIrd, IVth, or VIth cranial nerves.
Current treatment of CS-DAVF is problematic. When evaluating a newly diagnosed patient with CS-DAVF, two main aspects must be considered: (1) whether the lesion requires treatment; and (2) what treatment approach offers the best chance of safe and effective cure. Current classification systems do not assist the operator in either regard. This issue is addressed in an accompanying paper. With regard to the need for treatment, the issue is simply reduced to whether the lesion presents a significant risk to the patient of visual impairment, oculomotor dysfunction, or intracranial hemorrhage.8 ,9 The risks to vision and oculomotor function may be assessed by evaluation of the angiographic presence of significant anterior drainage of the CS-DAVF to one or both orbits combined with clinical evaluation of intraocular pressure, visual acuity, and oculomotor function. The risk of hemorrhage may be determined by the angiographic presence or absence of retrograde leptomeningeal venous drainage. In practice, only lesions with isolated posterior drainage into the IPS with no objective abnormalities in intraocular pressure, visual acuity, or oculomotor function may be deemed ‘benign’. Nonetheless, like all dural fistulas, CS-DAVF has the potential to be a dynamic entity whose clinical and pathophysiological behavior may change over time, so even lesions assessed as ‘benign’ at one clinical assessment may become dangerous at the next. While published studies would indicate that such transformation is rare (1.4% of Borden type 1 lesions developing potentially dangerous cortical venous drainage in 409 lesion-years of follow-up in one study10), it is important to note that, in the context of CS-DAVF, resolution of eye signs may indicate opening of alternate cortical venous drainage pathways rather than spontaneous resolution of the lesion, and follow-up with either angiography or non-invasive vascular imaging is essential in any lesions being managed conservatively.
With regard to the preferred treatment approach, there are a number of important angiographic features that must be considered. These include: (1) the presence or absence of arteriovenous shunts to one or both CSs; (2) the patency or otherwise of the IPS on both sides; and (3) the presence or absence of venous drainage crossing the midline from the shunt side.
While arteriovenous shunting is unilateral in most cases of CS-DAVF, a minority have bilateral shunts from one or a combination of ECA or ICA branches, with a small subset receiving shunts primarily to the circular sinus. In cases of unilateral shunts, closure of the ipsilateral CS is sufficient to cure the lesion, irrespective of the venous drainage pattern. However, if shunting is bilateral, closure of both CSs is usually required for cure, especially if there are bilateral venous outflows. In the rare cases of isolated shunting to the circular sinus, this must be the target of embolization.
Establishing the patency or absence of the IPS on both sides is crucial. If the IPS is patent on the side of the arteriovenous shunt (in 40–50% of cases in our experience), this usually provides a straightforward path for access and closure of the affected CS. Even if the ipsilateral IPS is occluded, patency of the contralateral IPS may permit retrograde access through the contralateral IPS and circular sinus to the affected CS. Where both IPSs are occluded, options include attempted rewiring of the occluded sinus(es),11 access through the facial vein or SOV (either via ultrasound-guided access or direct cut-down), or direct percutaneous transorbital puncture of the CS.12 From a pathophysiological standpoint, the presence of IPS occlusion may be considered analogous to ‘isolation’ of a venous channel in the transverse, sigmoid, or superior sagittal sinus in DAVF in these locations, where occlusion of the sinus potentially draining the fistula induces retrograde sinus or leptomeningeal venous flow. In CS-DAVF, IPS occlusion may induce retrograde flow into the orbit or leptomeningeal veins.
When arteriovenous shunting is unilateral but there is venous drainage to the contralateral CS via the circular sinus, this may (in the experience of the author) be associated with significant distension of the circular sinus, rendering access to the diseased CS via the contralateral IPS and circular sinus more straightforward compared with the same approach through a non-dilated circular sinus.
In considering treatment approaches to CS-DAVF, the last morphological aspect to note is the presence of unilateral or bilateral SOV drainage and the degree of distension of the vein(s). The presence of a significantly dilated SOV on one or both sides again presents a potential access route to the diseased CS(s) via the facial vein, ultrasound-guided venous puncture, or direct cut-down onto the vein. The importance of this feature of a given CS-DAVF increases in the presence of occluded IPSs.12
Given the above, it is the author's view that any classification of CS-DAVF should reflect the most important of these features. In other words, a classification which indicates the presence of unilateral or bilateral arteriovenous shunting; ipsilateral, contralateral or bilateral venous drainage; the patency or otherwise of the relevant IPS(s); and the presence or absence of retrograde leptomeningeal venous drainage would be a significant improvement in describing CS-DAVF. This would also allow more accurate communication regarding treatment decisions.
Apart from the problematic nature of description and classification of CS-DAVF, in our institution we have encountered significant limitations in the ‘conventional’ therapeutic options available for patients with this disease. At most centers, in the subset of patients with occluded IPS(s), it has been routine practice to attempt retrograde rewiring of the occluded IPS to gain access to the CS.11 It is our experience, borne out in the literature, that this is frequently unsuccessful, usually time-consuming and difficult, and carries risks of venous perforation and other complications.3–7 In the largest series which includes this method as the primary approach to CS-DAVF with occluded IPS, reported rates of failed access are up to 17%.3 If access to the affected CS is successful, the conventional preference for embolic material has been detachable coils. Deployment of coils is usually commenced at the anterior aspect of the CS, near the exit of the SOV, in order to avoid worsening of intraocular pressure in the event that treatment fails to occlude the lesion—this despite the fact that most CS-DAVF receive their shunting arteries towards the posterior compartments of the affected CS. In other words, embolization is commenced remote from the actual pathology for fear of treatment failure, which is reported in between 17% and 75% of CS-DAVF cases.3 ,7 In addition to significant risks of failed access and failure to close the shunts, there is also a significant risk of development of new—or exacerbation of existing—IIIrd, IVth and/or VIth cranial neuropathy when using coils. This has been reported with a frequency of 5.5–39.4%, or around 10.2% across the major published series using this technique.3–6 Various authors have ascribed this incidence of cranial neuropathy after coil deposition in the CS as being due to one or a combination of ‘overpacking’ of the sinus with resultant mass-effect, intra-/periprocedural thrombosis and inflammation around the nerves and/or hemodynamic effects in the partially occluded CS.5 These disappointing data (summarized in table 3) and personal experience led the author to explore alternative access routes and embolic agents to treat CS-DAVF.
Transorbital percutaneous CS puncture has been described in case reports in the literature as a safe and simple means of access to the CS.1 ,13 ,14 Although the numbers in these reports are low, the author adopted this technique for the first time in a patient where rewiring of an occluded IPS had failed and had also resulted in a significant complication (permanent VIth nerve palsy). The technique is surprisingly simple, usually of short duration and relatively low risk. It has become our first choice access route to the CS in patients with occluded IPS. In 21 patients who had transorbital CS access in this series there were no access-related complications. Specifically, there were no cases of retrobulbar hematoma, intraorbital cranial nerve damage, optic nerve trauma, mechanical or chemical ophthalmoplegia, ICA perforation, or subarachnoid hemorrhage. There was one instance of failed transorbital access requiring a second attempt on another date, which was completed without incident.
With regard to choice of embolic agent in the CS, the literature again portrays a disappointing picture of treatment with the usual agent of choice—namely, detachable coils—with significant rates of failure to cure and/or new or worsened cranial neuropathy. The high rate of failure to cure CS shunts with coils3–7 is largely due to the internal architecture of the CS, which consists of a network of connected ‘chambers’ or ‘pockets’, unlike the other dural venous sinuses. This results in difficulties in accessing all chambers receiving arterial shunts when using coils, a major disadvantage when compared with liquid agents which penetrate whole interconnected vascular networks. The author's personal experience of cases of this type resulted in several ‘rescue’ procedures with Onyx-34 (prior to this series) in patients in whom coil occlusion had failed and symptoms had not resolved. This further led to adoption at our institution of Onyx-34 initially and, more recently, PHIL-35 as the first-line agents for intracavernous embolization of CS-DAVF. In early experience, the author shared the concerns of colleagues regarding the possibility of DMSO-related neurotoxicity and retrograde embolization into the ICA via its cavernous branches. However, after precautionary deployment of semicompliant balloons in the ICA in the first few patients in which liquid agents were employed, it became clear that this was an unnecessary measure in most, if not all, cases. Currently, at our institution we only deploy a balloon in the ICA if there are large ICA branches supplying the fistula or if the ICA is obscured by embolic material from previous embolization attempts. The author holds the view that injection of liquid agents freely into the CS is substantially different from injection of the same agents into small muscular arteries from a wedged or ‘plugged’ position. In the former instance, where there is a large venous ‘dead space’, the chance of ‘forcing’ the agent retrogradely into small arterial channels is considered to be far lower than in the latter. This is based on broad experience in the neurointerventional community that, unless a catheter is in a wedged position, Onyx or PHIL will have a much stronger tendency to reflux around the delivery catheter than to migrate antegradely (or retrogradely) into small vessels. In addition, the goal of liquid embolization in the CS for CS-DAVF is different from that in transarterial liquid embolization for shunting lesions. In the former, only a relatively low-volume and low-pressure injection is required with the goal of obliterating only the venous compartments receiving the arterial shunts. Unlike coil embolization of the CS, liquid embolization is commenced at the posterior aspect of the CS—the actual site of the arteriovenous shunts—resulting in more rapid and targeted embolization with a commensurate lower volume of embolysate. By contrast, in transarterial embolization of DAVF, the goal is to obliterate all feeding arteries and to penetrate all points of abnormal arteriovenous connection to achieve a cure. This necessitates use of a lower viscosity agent (eg, Onyx-18, PHIL-25), a higher volume of embolysate, delivery from a plugged/wedged position, and higher injection pressures in many or most cases. In our series of 29 patients receiving direct intracavernous embolization with DMSO-based liquid agents, there were no instances of retrograde ICA embolization and no immediate postoperative cranial neuropathy that could be ascribed to toxic effects of DMSO or to mechanical trauma from the liquid agent. The only intraprocedural incident we encountered was a short episode of marked bradycardia in a patient who had a trigeminocardiac response during priming of the microcatheter with DMSO.15 This is easily controlled by prophylactic administration of IV atropine or aminoglycopyrolate prior to DMSO injection. The sole case of worsening cranial neuropathy (patient 6) occurred 48 hours postoperatively, well after the expected timeframe for DMSO excretion,16 and was independently concluded by a neurologist and an ophthalmologist to be related to delayed venous thrombosis. This patient's pre-existing and procedure-related ophthalmoplegia had almost completely resolved at 8 weeks postoperatively.
‘Conventional’ treatment of CS-DAVF has reproducibly been shown by multiple authors to be problematic with regard to access to lesions with occluded IPSs. As a result, we have adopted transorbital percutaneous CS puncture as our primary access route in patients with occluded IPSs. In our series, this has shown promise as a safe, efficient, and straightforward alternative to conventional approaches. Similarly, ‘traditional’ use of coils to occlude the CS in patients with CS-DAVF has a significant incidence of failure and clinical complications. We have therefore adopted liquid DMSO-based embolic agents (Onyx-34 and PHIL-35) as our first choice agents for closure of the CS. In our experience, this also shows a very promising safety and efficacy profile compared with coils, although we would recommend a larger research collaboration to enable better study of these agents for this purpose. Concerns frequently raised about neurotoxicity of DMSO and risks of retrograde penetration of liquid agents into the pial circulation are not borne out in either personal experience or in peer-reviewed literature.
Contributors The author is the sole contributor to the compilation of the submission. Dr Andrew Cheung assisted in the care of three patients in the series.
Competing interests The author has paid consulting agreements with Medtronic, Irvine, California, USA and Microvention, Tustin, California, USA. The author is a stockholder in Medtronic, Irvine, California, USA and Stryker, California, USA
Ethics approval Ethics approval was obtained from the South Eastern Sydney Local Health District Human Research Ethics Committee.
Provenance and peer review Not commissioned; externally peer reviewed.