Stereotactic radiosurgery using CT cisternography and non-isocentric planning for the treatment of trigeminal neuralgia

Abstract

Objective: Frame-based radiosurgical rhizotomy has been shown in clinical studies to be effective for managing trigeminal neuralgia (TN). To date, however, only a small pilot study has been published for the frameless, image-guided CyberKnife system. We present our preliminary experience with 29 trigeminal neuralgia patients treated with the frameless CyberKnife using X-ray image-guided targeting, a novel CT method for target definition, and non-isocentric planning.

Materials and Methods: All 29 patients failed previous medical therapy and 14 had undergone prior surgical procedures. CT iohexal cisternography was used to identify the 6- to 8-mm segment of nerve to be lesioned. The marginal dose ranged from 60 to 70 Gy (median 66.4 Gy) as defined at an average 79th percentile. The corresponding Dmax varied from 71.4 to 86.4 Gy (median 77.91 Gy).

Results: After a median 10-month follow-up, 26 of 29 (90%) patients rated their pain control as excellent and 3 (10%) reported no improvement. Median time to improvement was 6 days. No or only minor progression in numbness was reported by 22 of 29 (76%) patients, 4 of 29 (14%) patients reported worsening, and 3 of 29 (10%) reported the onset of severe ipsilateral facial numbness. Two patients whose target volume inadvertently included the semi-lunar ganglion developed painful dysethesias in the distribution of their numbness.

Conclusion: Although the optimal dose and length of nerve to be lesioned are still being refined, this preliminary experience suggests that image-guided robotic radiosurgery can effectively lesion the trigeminal nerve. Further follow-up is needed to determine whether our method has advantages over the more commonly used procedure for radiosurgical trigeminal rhizotomy.

• CyberKnife
• CT cisternography
• radiosurgical rhizotomy
• stereotactic radiosurgery
• trigeminal nerve
• trigeminal neuralgia

Introduction

Trigeminal neuralgia (TN) afflicts 5 in 100,000 people. Patients typically present with sudden lancinating facial pain localized to one or more divisions of the trigeminal nerve. It is widely accepted that this condition stems from a vascular compression of the trigeminal root entry zone (most typically by the superior cerebellar artery) or the segment of the nerve that is immediately adjacent to the pontine portion of the brainstem. It is postulated that the pulsatile vascular trauma leads to a demyelination of the nerve and subsequent ephaptic transmission of pain. TN has also been associated with multiple sclerosis [1].

Antiepileptic drugs like carbamazepine, neurontin and phenytoin are the first line of therapy for TN. However, medical treatment often fails to provide long-lasting relief or produces severe side effects that preclude chronic use. Such patients are appropriate candidates for surgical intervention [2], [3].

Operative therapy for TN includes microvascular decompression (MVD) and a variety of “destructive” procedures. MVD has a high likelihood of conferring immediate pain relief and a low rate of symptom recurrence. In a surgical series of 1,185 patients, recurrent pain was reported in only 30% of cases after 10 years of follow-up [4]. Because MVD requires a major craniotomy, it is generally reserved for younger and healthier TN patients. In contrast, the frequently encountered elderly or medically infirm individual with TN is most commonly treated with a semi-selective destructive lesion.

Percutaneous destructive procedures include balloon compression and glycerol or radiofrequency rhizotomy. While each operation offers a high initial rate of success, the recurrence rate at 10 years is also very high, reportedly between 50 and 80% [5], [6]; this widely acknowledged outcome dampens enthusiasm for this approach. In addition, major facial numbness is almost the rule after a destructive procedure. Given the shortcomings of these interventions, neurosurgeons remain keenly interested in more benign and effective treatment options.

Radiosurgical rhizotomy for TN was first introduced by Leksell in 1951 [7], [8]. In fact, Leksell's invention of radiosurgery was driven in large part by the goal of lesioning the trigeminal nerve in patients with TN. However, it was not until the mid-1990s, after the emergence of MRI and modern radiosurgical technology, that radiosurgical treatment for TN was popularized. In 1996, Kondziolka et al. demonstrated a significant sustained reduction of TN pain in a multi-center trial with the GammaKnife, using a dose of 70 Gy and adjusting the isodose line to limit the brainstem to no more than 30% [9]. Goss et al. subsequently demonstrated the efficacy of the LINAC against TN pain using a dose of 90 Gy [10]. Further studies confirmed the efficacy of radiosurgery and served to refine radiosurgical techniques [11-16]. Increasing numbers of TN patients are now being treated with radiosurgical rhizotomy, based particularly on reports of relatively low rates of post-treatment facial numbness. Despite a seemingly growing enthusiasm for radiosurgical rhizotomy, the reported overall excellent response rate at one year is no better than 70%, and the median time to pain relief is 1-3 months [10], [11].

Although image-guided radiosurgery (the CyberKnife) was first developed in 1994 [17], [18], the precision of targeting with the earliest prototypes was not well defined. However, phantom studies have demonstrated that its accuracy is comparable, and sometimes even superior, to that of frame-based stereotactic radiosurgery [17]. A growing confidence in system precision stemming from clinical experience encouraged us to explore the application of image-guided robotic radiosurgery to trigeminal rhizotomy. In this paper we report our preliminary experience with this technique.

Patients and methods

Twenty-nine patients with TN were treated with the CyberKnife at Stanford University between May 2002 and February 2004. All had tried and failed previous medical therapy. Thirteen patients had undergone prior surgical procedures, including glycerol injection (3 patients), balloon compressions (1), radiofrequency ablations (2), MVD (6), or a nerve stimulator placement (1). In one case, both a glycerol rhizotomy and MVD had been performed prior to presentation (Table I). Patients who were deemed to be poor candidates for MVD by virtue of age or co-morbid conditions, or who declined a craniotomy, were offered a radiosurgical rhizotomy. Patients with atypical facial pain or multiple sclerosis were excluded from this analysis.

Table I.  Patient characteristics, pre-operative pain distribution, and prior treatments.

Male:Female		16:13
Age at time of treatment (years)	Median	67
	Min	42
	Max	92
Duration of symptoms (years)	Median	9
	Min	0.5
	Max	45
Side (L:R)		12:17
Pain Distribution	V1	0
	V2	9
	V3	7
	V1 V2	3
	V2 V3	8
	V1 V2 V3	1
	Tongue	1
Previous therapy	RF	2
	Glycerol	3
	RF + Gly	1
	Balloon	1
	MVD	6
	Nerve Stim	1
Pre-op numbness		7

Patient preparation

Subsequent to a thorough neurologic examination and obtaining patient consent, a custom Aquaplast mask was fabricated for each patient. In the radiology suite, a standard lumbar puncture was performed and 5 cc of iohexal was injected into the subarachnoid space. After keeping the patient in the Trendelenburg position for 10-20 min, a thin-section CT scan (1.25-mm contiguous slices) was made through the entire head and exported to the CyberKnife Treatment Planning workstation.

Treatment planning

Images from the CT cisternography were examined on the planning workstation and the trigeminal nerve was identified along its entire course through the prepontine cistern (Figure 1). Frequent reference to reformatted sagittal and coronal images proved useful for this task. Using the CyberKnife software drawing tools, a 5- to 8-mm segment of the fifth cranial nerve was delineated beginning 2-3 mm from the brainstem, avoiding the root entry zone. In the second half of this series, the trigeminal ganglion was also defined as a critical structure in which the dose was minimized (no more than 30 Gy). Using inverse treatment planning, a non-isocentric plan was designed with a 650-mm SAD (source-to-axis distance) node template and a “7.5-mm” collimator (defined at 80 cm), which in effect provided for a nominal 6.0-mm collimator. Dmax ranged from 71.4 to 86.4 Gy (median 77.91) while the marginal dose varied from 60-70 Gy (median 66.4) as defined at an average 79th percentile (Table II). Treatment plans were approved independently by the treating neurosurgeon and radiation oncologist.

Figure 1. Example of a treatment plan using the CyberKnife. [Color version available online]

Table II.  Description and results of TN patients treated with the CyberKnife.

Patient	Sex	Age	Side	Distribution	Duration of pain	Failed med	Previous treatment	Pre numbness	Fraction	Dmin	Dmax	Volume
1	M	79	R	V3	3	Yes	RF Rhiz	Yes	1	70	86.4	0.049
2	F	85	L	V2	30	Yes	None	No	1	66.4	81.94	0.101
3	M	83	R	V2	0.5	Yes	None	No	1	66	81.5	0.117
4	M	74	R	V1 V2	3	Yes	None	No	1	68	86.28	0.158
5	F	67	L	V2 V3	6	Yes	None	No	1	66	80.49	0.062
6	M	78	R	V2 V3	8	Yes	None	No	1	70	82.35	0.113
7	F	73	L	V2	10	Yes	RF rhiz	Yes: left V2	1	69	79.31	0.053
8	F	66	R	V2 V3	7	Yes	None	No	1	68	80	0.144
9	F	42	L	V2	10	Yes	None	No	1	60	73.17	0.033
10	F	46	R	V3	7	Yes	None	No	1	60	74.97	0.047
11	F	92	L	V2 V3	22	Yes	MVD & glycerol	No	1	64	73.56	0.051
12	F	54	L	V2	45	Yes	None	No	1	60	72.29	0.085
13	M	78	L	V1 V2	18	Yes	Glycerol × 2	Yes: left V1 V2	1	60	71.4	0.091
14	F	55	R	V2 V3	6	Yes	MVD	No	1	63	74.12	0.019
15	M	56	L	V2	10	Yes	MVD	No	1	66	82.51	0.056
16	M	49	R	V3 V2 V1	3	Yes	MVD	No	1	65.5	78	0.035
17	M	76	R	V2	4	Yes	None	No	1	65	78.31	0.04
18	M	78	R	V3	10	Yes	Glycerol	No	1	65	77.38	0.08
19	M	49	L	V3	10	Yes	RF & balloon	Yes: left V2 V3	1	66	82.5	0.03
20	F	78	R	V2 V3	15	Yes	None	No	1	70	79.55	0.044
21	F	72	R	V2 V3	0.5	Yes	None	No	1	68	77.27	0.029
22	F	48	R	V2	11	Yes	MVD	No	1	68	74.73	0.06
23	F	54	L	Tongue	25	Yes	None	No	1	68	77.27	0.03
24	M	58	R	V2	14	Yes	MVD & nerve stim	Yes: rt V1 V2 V3	1	62	72.94	0.037
25	M	49	R	V3	4,5	Yes	None	No	1	68	77.27	0.029
26	M	43	R	V3	5	Yes	None	No	1	68	77.27	0.08
27	M	82	R	V1 V2	7	Yes	None	No	1	67	77.91	0.036
28	M	58	L	V2 V3	10	Yes	MVD	Yes: left V2 V3	1	68	77.27	0.036
29	M	85	L	V3	8	Yes	Glycerol × 2	Yes: mild V3	1	70	85.37	0.056
	Mean	65.7586			11				Mean	66	78.4
	Min	42			0.5				Min	60	71.4
	Max	92			45				Max	70	86.4
	Med	67			9				Median	66.4	77.91
Patient	Follow-up (mo)	Initial relief	Time to relief	Post numbness	Time to recurrence (mo)	Time to numbness (mo)	A. dolorosa	Other	Medication status
1	24	Excellent	1 day	None	n/a	n/a	No		Off meds
2	22	Excellent	2 days	Mild L V1 V2	2	2	Yes		Resumed meds
3	20	Excellent	2 days	Mild R V2	n/a	6	No		Off meds
4	18	Excellent	1 day	Mild R V1 V2	na	4	No		Off meds
5	17	Excellent	4 days – 6 mos	Mild L V2 V3	n/a	6	No		Off meds
6	7	Excellent	4 days	Severe R V1-3	n/a	4	No
7	6	Excellent	3–5 wks	Mild L V2	6	6	No		Reduced meds
8	6	Excellent	4 wks – 2 mos	Mod V2 V3	n/a	3	No
9	15	None	n/a	None	n/a	n/a	No		Off meds
10	15	Excellent	2–3 mos	Mod R V3	n/a	3	No	Trismus	Off meds
11	14	Excellent	2 wks – 1 mo	Left V2 V3	n/a	2	No		Off meds
12	12	Excellent	n/a	Severe L V1-3	n/a	5	Yes	L masticator & corneals	Resumed meds
13	14	Excellent	1 month	Mild L V1 V2	n/a	4	No		Off meds
14	12	Excellent	8 wks – 3 mos	Mild R V2 V3	n/a	9	No		Off meds
15	10	Moderate	8 wks	Mild L V2	n/a	8	No	Diplopia, CR	Same dose
16	12	Excellent	7 days	Severe R V1-3	n/a	6	No		Off meds
17	10	Excellent	6 days	Mild R V2	n/a	8	No		Off meds
18	10	Excellent	7 days	Mild R V3	n/a	3	No		Off meds
19	9	Excellent	3 mos	Left V2 V3	8	1	No		Resumed at same dose
20	7	Excellent	2 days	Mod V1-3	n/a	5	No	Drooling, decreased CR	Reduced meds
21	5	Excellent	1 day	Mod V2 V3	n/a	3	No		Off meds
22	5	Excellent	2 days – 1 mo	None	n/a	n/a	No		Off meds
23	4	Excellent	1 month	None	n/a	n/a	No		Off meds
24	5	Excellent	2 wks – 1 mo	No new	n/a	n/a	No		Same
25	4	No change	No change	Mild R V3	n/a	1	No	Spinal leak	Resumed meds
26	3	Excellent for 2 mo	4–7 days	None	1	n/a	No		Was completely off for 2 mos
27	3	Excellent	2–10 days	None	n/a	n/a	No		Off meds
28	2	Excellent	3–10 days	None	n/a	n/a	No		Reduced meds
29	6	Excellent	1–2 mos	None	n/a	n/a	No		Off meds
	10.2		Median = 6.5 days	None 8
	2			Mild 12		Median = 4
	24			Mod 4		Mean = 4.6
	10			Severe 3

Radiosurgical treatment

Outpatient CyberKnife rhizotomy was performed either on the same day or within a few days of treatment planning, depending on the severity of symptoms. During treatment, patients were positioned supine on the CyberKnife operating table and immobilized in their Aquaplast mask. During radiosurgery, a precisely calibrated pair of X-ray cameras, with matching KV X-ray sources, repeatedly imaged the patient's head and computationally referenced these near-real-time images with digitally reconstructed radiographs (DRRs) synthesized from the treatment planning CT cisternography scan. This information was continuously fed back to the robot, which compensated for the typically small movements observed in cooperative patients. No technical difficulties were encountered with the X-ray image-guided targeting system as a consequence of using iohexal-enhanced CT scans to generate DRRs. This Target Localization System (TLS) of the CyberKnife accurately determined the spatial position of the patient throughout the procedure [19]. Treatment typically lasted approximately one hour. Patients were discharged home with 8 mg of oral dexamethasone to avoid rare cases of acute edema and nausea post-radiation.

Follow-up

All patients were retrospectively evaluated in clinic or by telephone for the extent of pain control, time to improvement, associated numbness, and anesthesia dolorosa. Each patient was asked to rate the degree of pain control as excellent (>90% relief and completely off pain medications), moderate (greater than 50% and less than 90% reduction of pain with a reduction of pain medications), mild (less than 50% relief and no change in medications) or no change. The latency to improvement was also recorded. The corneal reflex was assessed and recorded during follow-up clinic visits, as well as any other side effects.

Results

Demographics

There were 16 male and 13 female patients in the study group. Median age was 66 years (range: 42-92). Median follow-up time was 10 months (minimum 2 months; maximum 24 months). Duration of symptoms prior to CyberKnife rhizotomy ranged from 0.5 to 45 years, with an average of 9 years. The divisional distribution and laterality of pain is depicted in Table I. Seven patients had preoperative numbness; each of them had undergone a prior surgical procedure (Table I).

Pain relief

Pain control was rated as excellent by 26 of 29 (90%) patients and 3 (10%) patients reported no improvement. All but 5 patients experienced relief in less than 28 days (median 6 days). Twenty-two of the 26 patients have thus far experienced a sustained pain response; 4 (14%) patients experienced a recurrence of pain at a median of 4 months following treatment (Table II). One patient (case 15 in Table II) was treated twice with the CyberKnife. This patient complained of intractable pain, and 2 months after the first treatment was treated for a second time, which yielded minor relief.

Side effects

No or only mild progression in facial numbness was reported by 22 of 29 (76%) patients. A moderate or severe increase in trigeminal distribution anesthesia was described by 4 (14%) and 3 (10%) patients, respectively. A decreased corneal reflex was noted in 4 patients, one of whom also required lubricating ophthalmic drops. There were two instances of painful dysethesias of the face (anesthesia dolorosa). Two additional patients described ipsilateral masticator weakness, and one reported a possibly related transient diplopia (Table II).

Discussion

By virtue of the adjacent brainstem, trigeminal rhizotomy demands maximal accuracy. A growing body of published evidence now demonstrates the high spatial fidelity of CyberKnife radiosurgery and provides sufficient justification for using image guidance to perform this procedure [17], [20]. Here we present the largest series of trigeminal neuralgia patients treated with a unique radiosurgical technique that combines CT cisternography, non-isocentric treatment planning, and – of greatest relevance – image-guided CyberKnife radiosurgery [21].

Standard isocentric radiosurgery delivers a spherical dose volume, which in the case of rhizotomy is intended to lesion an elongated structure (the trigeminal nerve). This approach, which stems from the core isocentric design of nearly all radiosurgical devices, seems artificial in light of the significant dimensional variability of the targeted nerve. In our view, non-isocentric treatment planning provides a more rational radiosurgical lesion in these cases.

Prior clinical studies investigating radiosurgical rhizotomy for TN used MRI alone or a CT-MRI fusion for target delineation. The choice of CT iohexal cisternography in this series was based in part on convenience, since CyberKnife image guidance requires CT to generate the DRRs used for targeting. However, this decision was also predicated on the potential for superior distortion-free spatial accuracy. Some investigators have argued that spatial distortion is not a clinically significant concern with MR-guided stereotaxis [22]. Although this may be true under most circumstances, our own more detailed analysis has demonstrated that MRI suffers from numerous inherent spatial distortions. These potential inaccuracies can only be eliminated through methodical calibration, a series of specialized MR acquisitions, and customized image processing [20,23-25]. We have found that fine-cut CT with intrathecal iohexal reliably identified the cisternal portion of the fifth cranial nerve that extended from the root entry zone to the ganglia. Only in one case did we encounter initial difficulty identifying the cisternal portion of the nerve. We felt that the nerve was likely atrophic secondary to the patient's previous radiofrequency rhizotomy. However, by extrapolating from both the visible root entry zone and the ganglia of the trigeminal nerve, it was possible to delineate a likely target segment of nerve. Although the need for a lumbar puncture adds an invasive step to an otherwise completely non-invasive procedure, the potential gain in accuracy would seem a reasonable trade-off. It is worth noting that one patient in the present series developed a symptomatic CSF leak and needed a blood patch.

Optimal trigeminal lesion location

The optimal portion of nerve for trigeminal rhizotomy remains a point of controversy [10], [14], [15]. The most typical trigeminal target is the dorsal root entry zone adjacent to the pons [26]. This is based on data from animals that suggest that the dorsal root entry zone, which marks the transition zone between the Schwann cells and oligodendrocytes, may be particularly sensitive to radiation [9], [10], [26]. However, when a GammaKnife lesion was centered on a point 2-4 mm anterior to the dorsal root entry zone, Kondziolka et al. reported that only 37.7% of patients sustained excellent or good pain relief at 5 years [11]. In contrast, when Massager et al. targeted a range of anterior cisternal targets (5-14 mm anterior to the brainstem) with a single 4-mm isocenter [15], 100% of patients in the 5-8 mm category reported excellent pain control and minimal numbness. Although Massager et al. implied that the Gasserian ganglion received a significant amount of radiation with this procedure, they could find no correlation between the amount of radiation it received and either pain control or complications [15].

Lesion length

The optimal length of nerve to be treated was first addressed in a prospective randomized double-blind study by Flickinger et al. A mean nerve length of 5.4 ± 0.4 mm was treated in a single-isocenter group and a mean length of 8.7 mm ± 1.1 mm in a two-isocenter group (within the 50% isodose line). Pain control did not differ across groups, but complication rates were higher in the two-isocenter group [13]. The two-isocenter technique likely also exposed a larger volume of neighboring structures (such as the brainstem and Gasserian ganglion) to high-dose radiation. Furthermore, this method produces additional dose inhomogeneity within the targeted nerve segment. Our current planning technique enables greater dose homogeneity and may do a better job of limiting the irradiation of adjacent critical structures. The initial decision to target a 6- to 12-mm length (median 7.5 mm) of trigeminal nerve that began 3 mm anterior to the pons and did not include the Gasserian ganglion was partly based on published experience with conventional isocentric radiosurgical rhizotomy.

Dose

The optimal radiosurgical dose for trigeminal rhizotomy has yet to be established. Animal studies in which trigeminal nerves were treated with a Dmax of 80 Gy and 100 Gy showed that axonal necrosis occurred at doses of 100 Gy [11]. Brisman published his experience with GammaKnife trigeminal rhizotomy using both a 75- and 76.8-Gy dose. Only 68.3% of the 293 patients in this series reported 75% or greater pain relief at a median follow-up of 6 months [12]. Meanwhile, Goss et al. reported that 76% of patients experienced greater than 50% pain relief when treated by a LINAC-based technology and a maximal dose of 90 Gy; they also described a higher (32%) rate of facial numbness [10]. This observation was consistent with that of Pollock et al. that 90 Gy delivered with the GammaKnife produced permanent trigeminal nerve dysfunction in 54% of patients [10], [16].

When developing a treatment plan with the CyberKnife, it is necessary to request both a maximal and minimal value that the optimization software seeks to satisfy. In the current study, the maximal dose ranged from 71.4 to 86.4 Gy (median 77.91 Gy) and the Dmin varied from 66.4 to 70 Gy (median 66.4 Gy). Strictly in terms of maximal dose, these values are comparable to the 70-90 Gy established in previous studies of radiosurgical rhizotomy for trigeminal neuralgia (and below the 100-Gy dose known to cause axonal necrosis in animals). To date we have found no association between pain response, time to improvement, or side effects and the maximal dose or radiation.

Rate of response

After a median follow-up of 10 months, 90% of patients (26 of 29) reported good to excellent initial pain relief. This overall response rate compares favorably with previously published reports of radiosurgical rhizotomy. However, analogous to these other studies, only 76% of patients (22 of 29) in this series have had a durable response; in those patients that relapse, the average time to pain recurrence was 4 months. To date, we have been unable to discern any connection between treatment dose and the likelihood of permanent pain relief. Longer follow-up is critical to establishing the ultimate durability of any therapeutic benefit.

Latency of response

Perhaps the most remarkable observation from this study has been the relatively short interval between radiosurgery and pain control. The median time to a significant reduction of pain was only 6 days and most patients reported that improvement was virtually immediate. In contrast, prior studies using the frame-based GammaKnife or LINAC have reported a median time to improvement of 2 months, and in some instances the benefit required as long as 6 months to be detected [10-12]. It is not clear if this advantage stems from dose/volume (length of nerve) considerations or from a greater measure of targeting accuracy with the CyberKnife. It is also possible that nonisocentric treatment, which generally delivers a more homogeneous treatment within a targeted structure, may simply provide a larger integral dose to the nerve than has been administered in previous clinical studies.

Treatment complications

Ten percent of the patients in this series had developed severe facial numbness at last follow-up, and an additional 14% complained of moderate numbness; the median time to symptom onset was 4 months. We could find no correlation between Dmax or length of nerve treated and the rate of complications (see Table II). After retrospectively examining the treatment plans of patients with anesthesia dolorosa and bothersome facial numbness, it was noted that in each case the Gasserian ganglion was inadvertently included in the high-dose region (defined target volume) (Figure 2). As a result of this analysis, we now assiduously avoid the trigeminal ganglion when defining the section of nerve to be lesioned. The Gasserian ganglion is easily identified on the treatment plan generated from the CT cisternography. The length of follow-up in this series is inadequate to determine whether facial numbness, or any other side effect, improves with time.

Figure 2. Example of a treatment plan where the Gasserian ganglion was inadvertently included in the target volume. [Color version available online]

The incidence of side effects, especially facial numbness, was considerably higher than that reported by Kondziolka et al. and modestly greater than that in other comparable LINAC and Gamma Knife series [11]. Although we could not find any correlation between dose, the length of treated nerve and the rate of complications (see Table II), this may be a by-product of the relatively small number of patients in this series. We continue to believe that such a relationship is fundamental to all treatment with radiosurgery and almost certainly exists for trigeminal rhizotomy. As a consequence, we are using the current experience, and the associated rate of complications, as an upper bound for selecting dose and target volume going forward. With these values in mind, we have gradually reduced the dose of radiation and length of nerve being lesioned, so far without any noticeable decrease in the rate of efficacy. It remains to be seen if these modifications will minimize complications without lessening the advantages of CyberKnife rhizotomy.

Targeting accuracy and image-guided radiosurgery

The relatively small size of the trigeminal nerve and the large radiation dose needed to induce a partial lesion necessitate that radiosurgery rhizotomy be conducted with maximal spatial accuracy. With such requirements, the optimal assessment of system specification would measure targeting errors in vivo, i.e., in real patients. Although ideal, this goal remains elusive with all current radiosurgical devices, and the best measure of precision continues to be phantom studies. When analyzed using phantoms, the performance of CyberKnife image guidance is competitive with, and may in some circumstances be superior to, that of frame-based stereotactic targeting. In vivo, clinical outcome continues to be a useful, albeit non-quantitative, measure of system precision. In this regard, the current results also serve to subjectively validate the quantitative findings of previous phantom studies with the CyberKnife.

The results in our series are notable for the overall initial response rate (90%) and the short latency to pain relief (median 6 days). Both measures compare favorably with what has been reported for more conventional radiosurgical rhizotomy. The most likely reason is that the trigeminal nerves in the current series were treated with a larger integral dose, stemming from the fact that a longer nerve segment was treated more homogeneously. If this is the case, then treatment response time is likely to increase and the likelihood of pain relief will decrease as we seek to treat a shorter length of nerve. It is also reasonable to speculate that the short latency and excellent response obtained in this series stems from better treatment accuracy. Our use of either or both CT-based targeting and frameless image guidance may have enhanced overall accuracy, which in turn may have been manifested in greater treatment efficacy. Whether this is true is likely to become clear as we further investigate the outcome with shorter-segment trigeminal radiosurgical rhizotomy.

Conclusion

Stereotactic radiosurgery using image guidance, non-isocentric planning and CT cisternography offers an alternative method for treating trigeminal neuralgia, with the possible benefit of a higher response rate and quicker onset of pain relief. The spatial fidelity of image-guided treatment appears to be more than sufficient for a radiosurgical procedure that demands maximal targeting accuracy. The optimal radiosurgical dose and length of nerve to be lesioned with this new technique has yet to be established.