Predictors of trigeminal nerve dysfunction following stereotactic radiosurgery for trigeminal neuralgia

Abstract

Background/Aims

To evaluate clinical and dosimetric predictors of trigeminal nerve dysfunction (TND) following stereotactic radiosurgery (SRS) for Trigeminal Neuralgia (TN).

Methods

We retrospectively reviewed our cohort of 446 patients with TN who underwent SRS between 1999-2008. Median follow-up was 25.1 and 17.4 months (mo) in those with and without TND respectively. Dosimetric and anatomic measurements and clinical features including Burchiel subtype, pain quality, prior procedures, comorbidities, and medications were evaluated for their influence on the TND using univariate and multivariate logistic regression modeling.

Results

TND was observed in 44.6% of patients and was similar across facial pain types. Those with TND had prolonged time to pain relapse [(TND, 68.48 mo) vs. (No TND, 29.37 mo)]. Multivariate analysis identified sharp pain at diagnosis (OR 0.594; 95%CI 0.38-0.91), and dorsal root entry zone (DREZ) maximum dose (OR 1.022; 95%CI 1.00-1.04) as predictors of TND.

Conclusions

The presence of sharp pain and increasing DREZ maximum dose were independently associated with TND. Patients with atypical facial pain were at lower risk of TND with increasing dose relative to Type 1 and Type 2 TN.

1. Introduction

Trigeminal neuralgia (TN), also known as tic douloureux, is defined as an intense episodic or constant facial pain of spontaneous onset. It is typically caused by vascular compression of the trigeminal nerve root (95%), multiple sclerosis plaques, lacunar infarctions, or mass lesions.[1-3] Surgical and interventional treatments for trigeminal neuralgia either aim to damage the nerve [through radiofrequency lesioning, glycerol rhizolysis, balloon compression, or stereotactic radiosurgery (SRS)], or to eliminate the presumed cause of TN by separating the nerve from a compressing vessel.[4] While microvascular decompression (MVD) has been the standard treatment for TN because of its quick onset, extended durability, and >90% likelihood of obtaining pain relief, the use of SRS has gained popularity for those considered poor surgical candidates or with refractory pain following prior procedural/surgical intervention. [2, 5] [6]

Trigeminal nerve dysfunction (TND) can involve symptoms such as facial numbness, swelling, paresthesia, and hypoesthesia. Although SRS offers a favorable rate of TND relative to other neuroablative procedures [facial numbness: 60% (glycerol rhizotomy), 98% (curved electrode radiofrequency rhizotomy), motor dysfunction: 7% (curved electrode radiofrequency rhizotomy) to 66% (balloon decompression)], mild to bothersome TND is still experienced in 15-49% (mild) and 1.4-10% (bothersome) of patients and has been associated with increasing nerve length treated and total dose. [7, 8] [9-15]

While it has been accepted that higher radiation doses are associated with a higher risk of TND, our study’s objective was to determine specific factors and target sites that may predict the development of TND for SRS-treated TN and promote longer relapse-free periods.

2. Materials and Methods

2.1. Patient Population

From September 1999 to December 2008, 777 patients underwent SRS for TN at Wake Forest Baptist Health. Patients were excluded from the study if they had non-episodic atypical facial pain (somatoform etiology), multiple sclerosis-related TN, post-herpetic neuralgia, trigeminal neuropathic pain, if pre-treatment MRI was unavailable or if they lacked a post-radiosurgery follow-up visit. Four hundred forty-six patients met the above inclusion criteria and were stratified by occurrence of any TND following SRS. This study was approved by the Wake Forest University Institutional Review Board.

2.2. Pain Classifications

TN was classified as being Type I if facial pain was episodic >50% of the time and Type II if facial pain was constant >50% of the time.[1] Atypical facial pain was defined as facial pain possibly derived from a somatic symptom disorder. Those with episodic atypical facial pain were included in the study. Patient characteristics including initial pain scores are specified in Lucas et al.[17] Pain relief was graded from I-V using the Barrow Neurological Institute (BNI) scale.[16] Patients with a BNI score of I-III at three months post-radiosurgery were considered treatment responsive. Date of treatment failure was the date at which patients’ TN was rated at a BNI score of IV or V. The time to pain failure was defined as the time from GKRS to the time of BNI IV/V pain failure. TND symptoms following SRS were categorized as any facial numbness, swelling, paresthesia, hypoesthesia, and “unusual.” Unusual symptoms were defined as anesthesia dolorosa, corneal anesthesia, deafferentation pain, jaw muscle weakness or trismus.

2.3. Radiosurgery Technique

Stereotactic radiosurgery was performed using the Gamma Knife™ Model B (1999-2004) or 4C (2005-8) unit (Elekta, Norcross, GA). All patients underwent placement of a Leksell Model G stereotactic head frame and high-resolution stereotactic magnetic resonance imaging (MRI) on the day of treatment. Patients lacking MRI on the day of treatment and treated with stereotactic computed tomography (CT) imaging were excluded from this study because precise dosimetric assessment of the Dorsal Root Entry Zone (DREZ) and pons surface was not possible without MRI. Treatment plans were generated using the Leksell GammaPlan (LGP) treatment planning system (Elekta) using a single 4-mm collimator to target the trigeminal nerve. The radiation dose was prescribed at the 100% isodose level for all patients. Patients treated earlier in the study had the isocenter placed proximally, such that the 50% isodose line was tangential to the pons while those treated later in our series had the isocenter placed distally, commonly at the pars triangularis. No beam blocking was used. The mean prescription dose was 88.06 Gy (range 80-97 Gy). A collimator output factor of 0.87 was used for the 4mm collimator for dose calculations.

2.4. Follow-up

Patients were scheduled for follow-up approximately three months after their SRS procedure, and were subsequently seen on an as-needed basis if they had satisfactory pain relief. Treatment outcomes were determined from patient electronic medical records and telephone questionnaires that assessed pain relief and duration, quality of life, medications before and after SRS, additional TN procedures and toxicity.

2.5. Dosimetric Assessment

Dosimetric variables were evaluated using the LGP software paired with each patient’s treatment-day stereotactic magnetic resonance images. For patients undergoing multiple procedures, only the results and dosimetry of their first SRS procedure were included in the study. Dosimetric parameters were determined as previously described by Lucas et al.[17] Trigeminal nerve length was defined as the distance from the intersection of the nerve with the brainstem to the petrous dura. The maximum dose to each structure is outlined in Table 2.

Table 2.

Dosimetry by Trigeminal Nerve Dysfunction

		Trigeminal Nerve Dysfunction
Parameter	Unit	N	None M (IQR)	N	Present M (IQR)	p-value
DREZ Dose	Gy	240	48.65 (19.8-74.25)	196	64.15 (30.40-79.05)	0.002
Pons Max Dose	Gy	240	65.15 (37.65-80.30)	196	70.6 (48.6-81.3)	0.685
Petrous Dura Dose	Gy	240	14.8 (7.95-28.55)	196	16.5 (10.25-33.05)	0.281
Gamma Angle	deg, o	231	105 (90-110)	193	105 (90-110)	0.795
Nerve Length	mm	234	9.3 (7.5-11.1)	196	9.3 (7.65-11.15)	0.624

N: Number; M: Median; IQR: Interquartile Range; DRZ: Dorsal Root entry Zone; Gy: Gray; deg: degree.

2.6. Statistical Analysis

Two-tailed paired t-tests were used to compare mean doses between patient groups. Chi-square and Fisher’s exact tests were used to test for heterogeneity between treatment groups. Time to BNI IV-V was calculated using the Kaplan-Meier method. Log-rank tests were performed to determine statistical differences between curves. Univariate logistic regression analysis was used to generate odds ratios and assess for statistical significance of potential predictors. Interactions were accounted for using product terms. Covariate selection took place using backwards stepwise selection of covariates that met diagnostic criteria and p-values<0.2. A multivariate logistic regression model was performed following covariate selection to determine the relative value of dosimetric and patient factors in predicting for SRS-associated TND. All analyses were conducted using SAS v.9.3, Cary, NC.

3. Results

3.1. Patient Characteristics

Patient demographic information is summarized in Table 1. In our cohort, 199 of 446 patients (44.6%) experienced TND symptoms. The most common TND symptoms were facial anesthesia (n=110, 24.7%) and paresthesia (n=61, 13.7%). Less commonly experienced symptoms were hypoesthesia (n=14, 3.1%), swelling (n=4, 0.9%), and “unusual” symptoms (n=18, 4.0%), which included corneal anesthesia (n=12, 2.7%), deafferentation (n=12, 2.7%), and anesthesia dolorosa (n=3, 0.7%).Those who developed TND had longer median follow-up times (p=0.02), differed by location (<0.0001), and were more likely to have presented with “sharp” pain (0.0002). There were no significant differences between those who did and did not experience TND by age, gender, TN type or quality, laterality, or prior procedures (Table 1).

Table 1.

Patient Characteristics by Trigeminal Nerve Dysfunction

		Trigeminal Nerve Dysfunction
Descriptor	Subgroup	N	None % M (IQR)	N	Present % M (IQR)	p-value
Age	yrs	247	68.37 (57.62-77.07)	199	66.94 (55.90-74.84)	0.184
Gender	Male	98	56.65	75	43.35	0.669
	Female	149	54.58	124	45.42
TN	Type 1	217	56.36	168	43.64	0.294
	Type 2	30	49.18	31	50.82
Atypical Facial Pain	Present	16	50	16	50	0.525
Laterality	Left	100	55.25	81	44.75	0.963
	Right	147	55.47	118	44.53
Location	V1	12	66.67	6	33.33	<0.0001
	V2	64	52.46	58	47.54
	V3	34	54.84	28	45.16
	V1+V2	37	47.44	41	52.56
	V1+V3	3	42.86	4	57.14
	V2+V3	61	63.54	35	36.46
	V1+V2+V3	35	56.45	27	43.55
Presenting Pain	Burning	2	40	3	60	0.660
	Dull	2	50	2	50	1.0
	Sharp	122	65.95	63	34.05	0.0002
	Throbbing	1	100	0	0	1.0
Prior Procedures	#	247	0 (0-1)	199	0 (0-1)	0.968
	Glycerol Inj	20	46.51	23	53.49	0.259
	MVD	22	51.16	21	48.84	0.629
	RFA	27	50	27	50	0.466
Follow-Up	mo	247	17.43	199	25.13	0.020

N=Number, %=Percent, M: Median; IQR: Interquartile Range, yrs: years, TN: Trigeminal Neuralgia, V1,2,3: Branch of the trigeminal nerve, #: Number, Inj: Injection, MVD: Microvascular Decompression, RFA: Radiofrequency Ablation, mo: month.

3.2. SRS Dosimetry

Dosimetric measurements were available for >90% of our cohort (Table 2). Patients who experienced TND differed from those without TND with regards to maximum DREZ dose [64.2 Gy (IQR 30.4-79.5) vs. 48.7 Gy (IQR 19.8-74.3); p=0.002]. There were no significant differences TND experience with regards to the maximum dose to the pons or petrous dura, gamma angle or nerve length.

3.3. Treatment Outcomes

Those who experienced trigeminal dysfunction had lower BNI scores at their six-month follow-up and also had a longer median time to relapse than their counterparts who did not experience TND [(TND, 68.48 months; 95%CI, 59.87-Not Reached) vs. (No TND, 29.37 months; 95%CI, 17.73-41.31)]. There was no evidence that trigeminal dysfunction was associated with post-SRS medication use (p=0.22), the duration of their facial pain after SRS (p=0.31), or any additional procedures received after the first SRS (0.97).

Table 3.

Treatment Outcomes

		Trigeminal Nerve Dysfunction
Outcome	Subgroup	N	None % M (IQR)	N	Present % M (IQR)	p-value
Median Time to BNI 4/5 Pain Relapse	mo	247	29.37 (17.73-41.31)	198	68.48 (59.57-NR)	<0.0001
6 month BNI	1	98	51.61 (41.32-68.84)	94	75 (61.21-NR)	<0.0001
	2	44	40.63 (15.75-NR)	36	64.51 (50.89-NR)
	3	59	20.82 (12.96-60.72)	51	55.39 (22.53-NR)
Duration of Symptoms	wks	246	5.25 (3-12)	199	5 (2-10)	0.467
QOL improvement	No	64	25.9	41	20.7	0.236
	Yes	183	74.0	157	79.3
Medications	# Pre-SRS	246	3 (2-4)	199	3 (2-4)	0.857
	# Post-SRS	247	2 (1-2)	199	2 (1-2)	0.992
Further Procedures	No	173	70.0	157	79.3	0.247
	Yes	74	64.35	41	35.65

N=Number, %=Percent, M: Median; IQR: Interquartile Range mo: months; BNI: Barrow Neurologic Institute; wks: weeks; QOL: Quality of Life; #: Number; SRS: Stereotactic Radiosurgery; NR= Not Reached.

3.4. Intercorrelation of TND Symptoms

In order to assess the level of association between each complication, we determined the Pearson correlation coefficients of each complication symptom and compared it against all others (Table 4). Only facial anesthesia and paresthesias were significantly correlated (ρ=-0.2278, p<0.0001).

Table 4.

Inter-correlation of TND Symptoms

Treatment Sequelae	ρ, p-Value
	Paresthesia	Swelling	Hypoesthesia	Unusual	Facial Anesthesia
Paresthesia	1.0000	-0.0379, 0.4250	0.0780, 0.0998	0.0510, 0.2825	-0.2278, <0.0001
Swelling	-0.0379, 0.4250	1.0000	-0.0171, 0.7183	-0.1951, 0.6812	0.0559, 0.2386
Hypoesthesia	0.0780, 0.0998	-0.0171, 0.7183	1.0000	-0.0369, 0.4367	-0.0732, 0.1228
Unusual	0.0510, 0.2825	-0.0195, 0.6812	-0.0369, 0.4367	1.0000	-0.0116, 0.8067
Facial Anesthesia	-0.2278, <0.0001	0.0559, 0.2386	-0.0732, 0.1228	-0.0116, 0.8067	1.0000

ρ = Pearson Correlation Coefficient

3.5. Predictors of SRS Associated Trigeminal Nerve Dysfunction

Multivariate analysis revealed that the presence of sharp pain at diagnosis (OR 0.594; 95%CI 0.385-0.916), and increasing DREZ maximum dose (OR 1.022; 95%CI 1.004-1.041) were independently associated with TND (Table 5). In Figure 1, we expanded upon these results by plotting the probability of TND against maximum DREZ dose to illustrate the impact of both the dose itself, and also the manner in which other covariates such as presence of sharp pain and facial pain type affect the slope of the curve. In Figure 1A, we included a second x-axis at the top of the graph to aid physicians who use the Gamma Knife decide isocenter placement. It assumes a prescription of 85Gy to a point using a single 4mm collimator, and illustrates how far a physician should place the isocenter from the DREZ to yield a certain probability of complications. Those presenting with sharp pain were more likely to experience TND after SRS regardless of DREZ dose. Additionally, patients who presented with atypical facial pain had a similar incidence of TND irrespective of DREZ dose, while patients with either Type I or Type II TN exhibited an increasing odds of TND with increasing dose to the DREZ.

Table 5.

Predictors of SRS Associated Trigeminal Nerve Dysfunction

		Odds Ratio
Covariate	Subgroup	Point Estimate	95% CI		p-value
Age	Yrs	0.991	0.997	1.006	0.244
Pons Max Dose	Gy	0.980	0.960	1.001	0.058
DREZ Max Dose	Gy	1.022	1.004	1.041	0.016
Presenting Sharp Pain	Yes vs. No	0.594	0.385	0.916	0.018

CI: Confidence Interval; Odds ratios greater than 1 indicate an increased likelihood of developing TND compared across groups

Figure 1.

Probability of Trigeminal Nerve Dysfunction according to DREZ Dose, Shot Location and Presenting Symptoms

Dotted lines straddling each solid line represent 95% confidence intervals around the estimate.

a. Incidence of TND as a function of the distance from the center of a single 4 mm collimator shot to the DREZ, b. Incidence of TND across DREZ dose stratified by the presence of sharp pain pre-treatment symptomatology, c. Incidence of TND across DREZ dose by the presence of atypical facial pain or Burchiel type. Gy: Gray; Est.: Estimated.

4. Discussion

In our study, we found that TND following SRS for TN was associated with a presenting symptom of “sharp” pain as well as an increased maximum radiation dose to the DREZ. While it has been generally accepted that higher maximum radiation doses are associated with a higher proportion of TND, our study implicates an anatomic target and a volume independent effect of dose which accounts for differences in TND among patients treated to the same prescription dose.

The findings from the current study highlight a controversy of TN radiosurgery: the optimal isocenter location. The current dataset suggests that toxicity may be higher for proximal isocenter location, though Matsuda et al published results stating the contrary[18]. Moreover, several prior series have also reported that patients who experience numbness also have a greater pain response rate and more durable response after SRS. A similar association between trigeminal nerve dysfunction and increasing nerve dose has been described in multiple series, with a dependence on the volume of brainstem irradiated being an identified predictive factor.[7, 19] While Brisman described this effect as an associated consequence of brainstem dose, the current series found no association with the maximum dose to the pons and the incidence of TND.[19]

In addition to treatment-related factors, we found that patients who classified their pain as “sharp” experienced a greater degree of TND after SRS. The mechanistic etiology of this observed relationship is unclear, though it may be that “sharp” pain is more likely to represent true Type I trigeminal neuralgia. As Type I trigeminal neuralgia pain derives from the “vascular hypothesis”, vascular compression of the nerve may cause pre-treatment nerve injury or demyelination, predisposing to further injury. This is important information for pretreatment patient counseling, as some patients may opt for a higher likelihood of pain response in spite of an increased predisposition to facial anesthesia/paresthesia. Likewise, patients predisposed to a lower SRS pain response rate (Atypical Facial Pain or relapse Type 2, Figure 1) may benefit from more proximal collimator placement as this appears to be less of an impact on the odds of developing TND in this population.

5. Conclusion

Sharp pain at presentation and increasing DREZ maximum dose in Type 1 and 2 patients were significantly associated with increased odds of TND and treatment response following SRS for TN. Patients with Atypical Facial Pain may benefit from more proximal isocenter placement as this population appears to be less predisposed to TND with increasing DREZ dose. Patients should be counseled regarding their pretreatment symptomatic phenotype (Type 1/2 vs Atypical Facial Pain) as some patients may opt for a higher treatment response rate even in the setting of increased potential for TND.

6. Acknowledgments

Biostatistical services supported by the Comprehensive Cancer Center of Wake Forest University NCI CCSG P30CA012197 grant. The authors also wish to thank Bonny B. Morris, for her thoughtful manuscript review. Sincere thanks goes to Scott Isom, Doug Case and Kopriva Marshall for their important and influential contributions, in addition to the ASTRO CNS review committee, as this paper was selected as an oral presentation at ASTRO 2012 in Boston.